Contents

Azure Architecture Center

Architecture icons
Browse all Architectures
Browse Hybrid and Multicloud Architectures
What's new
Application Architecture Guide
Introduction
Architecture styles
Overview
Big compute
Big data
Event-driven architecture
Microservices
N-tier application
Web-queue-worker
Design principles for Azure applications
Overview
Design for self-healing
Make all things redundant
Minimize coordination
Design to scale out
Partition around limits
Design for operations
Use managed services
Use the best data store for the job
Design for evolution
Build for the needs of business
Technology choices
Choose a compute service
 Choose a data store
Understand data store models
Select a data store
Criteria for choosing a data store
Choose a load balancing service
Choose a messaging service
Asynchronous messaging options in Azure
Choose an Apache Kafka host
Best practices for cloud applications
API design
API implementation
Autoscaling
Background jobs
Caching
Content Delivery Network
Data partitioning
Data partitioning strategies (by service)
Message encoding considerations
Monitoring and diagnostics
Retry guidance for specific services
Transient fault handling
Performance tuning
Introduction
Scenario 1 - Distributed transactions
Scenario 2 - Multiple backend services
Scenario 3 - Event streaming
Performance antipatterns
Overview
Busy Database
Busy Front End
Chatty I/O
Extraneous Fetching
 Improper Instantiation
Monolithic Persistence
No Caching
Synchronous I/O
Responsible Innovation
Overview
Judgment Call
Harms Modeling
Understand Harm
Assess Types of Harm
Community Jury
Azure for AWS Professionals
Overview
Component information
Accounts
Compute
Databases
Messaging
Networking
Regions and Zones
Resources
Security and Identity
Storage
Service comparison
Azure for GCP Professionals
Overview
Services comparison
Microsoft Azure Well-Architected Framework
Overview
Cost Optimization
About
Principles
Design
Checklist
Cost model
Capture requirements
Azure regions
Azure resources
Governance
Initial estimate
Managed services
Performance and price options
Provision
Checklist
AI + Machine Learning
Big data
Compute
Data stores
Messaging
Networking
Cost for networking services
Web apps
Monitor
Checklist
Budgets and alerts
Reports
Reviews
Optimize
Checklist
Autoscale
Reserved instances
VM instances
Caching
Tradeoffs
Operational Excellence
Overview
Principles
Automation
Automation Overview
Repeatable infrastructure
Configure infrastructure
Automate operational tasks
Release engineering
Application development
Continuous integration
Release testing
Performance
Release deployment
Rollback
Observability (operational awareness)
Monitoring
Checklist
Performance Efficiency
Overview
Design
Application design
Application efficiency
Scalability
Capacity planning
Testing
Performance testing
Testing tools
Monitoring
Checklist
Reliability
Overview
 Application design
Design overview
Error handling
Failure mode analysis
Backup and recovery
Business metrics
Chaos engineering
Data management
Monitoring and disaster recovery
Recover from a region-wide service disruption
Resiliency testing
Checklist
Security
About
Principles
Design
Governance
Overview
Segmentation strategy
Management groups
Administration
Identity and access management
Checklist
Roles and responsibilities
Control plane
Authentication
Authorization
Best practices
Networking
Network security review
Containment strategies for organizations
Best practices
 Storage
Applications and services
Application security considerations
Application classification
Threat analysis
Securing PaaS deployments
Compliance requirements
Configuration and dependencies
Monitor
Health modeling
Tools
Security operations
Review and audit
Identity and network risks
Optimize
Automate
Replace insecure protocols
Elevate security capabilities
Design Patterns
Overview
Categories
Availability
Data management
Design and implementation
Management and monitoring
Messaging
Performance and scalability
Resiliency
Security
Ambassador
Anti-corruption Layer
Asynchronous Request-Reply
Backends for Frontends
Bulkhead
Cache-Aside
Choreography
Circuit Breaker
Claim Check
Command and Query Responsibility Segregation (CQRS)
Compensating Transaction
Competing Consumers
Compute Resource Consolidation
Deployment Stamps
Event Sourcing
External Configuration Store
Federated Identity
Gatekeeper
Gateway Aggregation
Gateway Offloading
Gateway Routing
Geodes
Health Endpoint Monitoring
Index Table
Leader Election
Materialized View
Pipes and Filters
Priority Queue
Publisher/Subscriber
Queue-Based Load Leveling
Retry
Saga
Scheduler Agent Supervisor
Sequential Convoy
Sharding
 Sidecar
Static Content Hosting
Strangler Fig
Throttling
Valet Key
Azure categories
AI + Machine Learning
Overview
Technology guide
Cognitive services
Machine learning
Machine learning at scale
Natural language processing
R developer's guide to Azure
Architectures
AI at the edge
AI enrichment in Cognitive Search
AI for Earth
Auditing and risk management
Autonomous systems
Disconnected AI at the edge
Baseball decision analysis with ML.NET and Blazor
Batch scoring for deep Learning
Batch scoring with Python
Batch scoring with Spark on Databricks
Batch scoring with R
Business Process Management
Real-time recommendation API
Chatbot for hotel reservations
E-commerce chatbot
Enterprise chatbot disaster recovery
Enterprise-grade conversational bot
Enterprise productivity chatbot
FAQ chatbot
Content Research
Contract Management
Customer churn prediction
Customer feedback
Defect prevention
Distributed deep learning training
Digital Asset Management
Energy supply optimization
Energy demand forecasting
Image classification
Image classification with CNN's
Information discovery with NLP
Interactive voice response bot
Digital text processing
Machine teaching
MLOps for Python models
MLOps technical paper
Upscale ML lifecycle with MLOps
MLOps maturity model
Azure ML service selection guide
Model training with AKS
Movie recommendations
Marketing optimization
Personalized offers
Personalized marketing solutions
Population health management
Hospital patient predictions
Vehicle telematics
Predictive maintenance
Predictive marketing
 Quality assurance
Real-time scoring Python models
Real-time scoring R models
Remote patient monitoring
Retail assistant with visual capabilities
Retail product recommendations
Scalable personalization
Speech services
Speech to text conversion
Training Python models
Vision classifier model
Visual assistant
Deploy AI and machine learning at the edge by using Azure Stack Edge
Analytics
Architectures
Advanced analytics
Anomaly detector process
App integration using Event Grid
Automated enterprise BI
Big data analytics with Azure Data Explorer
Content Delivery Network analytics
Data warehousing and analytics
Demand forecasting
Demand forecasting for marketing
Demand forecasting for price optimization
Demand forecasting for shipping
Discovery Hub for analytics
Hybrid big data with HDInsight
Monitoring solution with Azure Data Explorer
ETL using HDInsight
Interactive analytics with Azure Data Explorer
IoT telemetry analytics with Azure Data Explorer
 Interactive price analytics
Mass ingestion of news feeds on Azure
Oil and Gas tank level forecasting
Partitioning in Event Hubs and Kafka
Predicting length of stay in hospitals
Predictive aircraft engine monitoring
Real Time analytics on big data
Stream processing with Azure Databricks
Stream processing with Azure Stream Analytics
Tiering applications & data for analytics
Blockchain
Architectures
Blockchain workflow application
Decentralized trust between banks
Supply chain track and trace
Compute
HPC Overview
Architectures
3D video rendering
Computer-aided engineering
Digital image modeling
HPC risk analysis
HPC and big compute
Cloud-based HPC cluster
Hybrid HPC with HPC Pack
Linux virtual desktops with Citrix
Move Azure resources across regions
Run a Linux VM on Azure
Run a Windows VM on Azure
Solaris emulator on Azure VMs
Migrate IBM applications with TmaxSoft OpenFrame
Reservoir simulations
 CFD simulations
Containers
AKS Solution Journey
AKS Baseline Cluster
AKS Cluster Best Practices
AKS Workload Best Practices
AKS day-2 operations guide
Triage practices
Introduction
1- Cluster health
2- Node and pod health
3- Workload deployments
4- Admission controllers
5- Container registry connectivity
Common Issues
AKS Example Solutions
Microservices architecture on AKS
Microservices with AKS and Azure DevOps
Secure DevOps for AKS
Building a telehealth system
CI/CD pipeline for container-based workloads
Databases
Guides
Overview
Relational data
Extract, transform, and load (ETL)
Online analytical processing (OLAP)
Online transaction processing (OLTP)
Data Warehousing
Non-relational data
Non-relational data stores
Free-form text search
 Time series data
Working with CSV and JSON files
Big Data
Big Data architectures
Batch processing
Real time processing
Technology choices
Analytical data stores
Analytics and reporting
Batch processing
Data lakes
Data storage
Data store comparison
Pipeline orchestration
Real-time message ingestion
Search data stores
Stream processing
Application tenancy in SaaS Databases
Tenancy models
Monitor Azure Databricks jobs
Overview
Send Databricks application logs to Azure Monitor
Use dashboards to visualize Databricks metrics
Troubleshoot performance bottlenecks
Transfer data to and from Azure
Extend on-premises data solutions to Azure
Securing data solutions
Architectures
Apache Cassandra
Azure data platform
Big data analytics with Azure Data Explorer
Campaign optimization with HDInsight Spark
Campaign optimization with SQL Server
DataOps for modern data warehouse
Data streaming
Data cache
Digital campaign management
Digital marketing using Azure MySQL
Finance management using Azure MySQL
Finance management using Azure PostgreSQL
Gaming using Azure MySQL
Gaming using Cosmos DB
Globally distributed apps using Cosmos DB
Hybrid ETL with Azure Data Factory
Intelligent apps using Azure MySQL
Intelligent apps using Azure PostgreSQL
Interactive querying with HDInsight
Loan charge-off prediction with HDInsight Spark
Loan charge-off prediction with SQL Server
Loan credit risk modeling
Loan credit risk with SQL Server
Messaging
Modern data warehouse
N-tier app with Cassandra
Ops automation using Event Grid
Oracle migration to Azure
Personalization using Cosmos DB
Retail and e-commerce using Azure MySQL
Retail and e-commerce using Azure PostgreSQL
Retail and e-commerce using Cosmos DB
Running Oracle Databases on Azure
Serverless apps using Cosmos DB
Streaming using HDInsight
Windows N-tier applications
Developer Options
Microservices
Overview
Guides
Domain modeling for microservices
Domain analysis
Tactical DDD
Identify microservice boundaries
Design a microservices architecture
Introduction
Choose a compute option
Interservice communication
API design
API gateways
Data considerations
Design patterns for microservices
Operate microservices in production
Monitor microservices in Azure Kubernetes Service (AKS)
CI/CD for microservices
CI/CD for microservices on Kubernetes
Migrate to a microservices architecture
Migrate a monolith application to microservices
Modernize enterprise applications with Service Fabric
Migrate from Cloud Services to Service Fabric
Serverless applications
Serverless Functions overview
Serverless Functions examples
Plan for serverless architecture
Serverless Functions decision and planning
Serverless application assessment
Technical workshops and training
Proof of concept or pilot
 Develop and deploy serverless apps
Serverless Functions app development
Serverless Functions code walkthrough
CI/CD for a serverless frontend
Serverless Functions app operations
Serverless Functions app security
Architectures
Big data analytics with Azure Data Explorer
CI/CD pipeline using Azure DevOps
Event-based cloud automation
Microservices on Azure Service Fabric
Multicloud with the Serverless Framework
Serverless applications using Event Grid
Serverless event processing
Unified logging for microservices apps
DevOps
Checklist
Guides
Extending Resource Manager templates
Overview
Update a resource
Conditionally deploy a resource
Use an object as a parameter
Property transformer and collector
Architectures
CI/CD pipeline for chatbots with ARM templates
CI/CD for Azure VMs
CI/CD for Azure Web Apps
CI/CD for Containers
CI/CD using Jenkins and AKS
DevSecOps in Azure
DevSecOps in GitHub
 DevTest and DevOps for IaaS
DevTest and DevOps for PaaS
DevTest and DevOps for microservices
DevTest Image Factory
CI/CD using Jenkins and Terraform
Hybrid DevOps
Java CI/CD using Jenkins and Azure Web Apps
Jenkins on Azure
SharePoint for Dev-Test
Real time location sharing
Run containers in a hybrid environment
High Availability
Overview
Architectures
IaaS - Web application with relational database
Hybrid
Guides
FSLogix for the enterprise
Administer SQL Server anywhere with Azure Arc
Azure Stack stretched clusters for DR
Azure Stack for remote offices and branches
Architectures
Manage configurations for Azure Arc enabled servers
Azure Automation in a hybrid environment
Azure enterprise cloud file share
Using Azure file shares in a hybrid environment
Azure Functions in a hybrid environment
Back up files and applications on Azure Stack Hub
Azure Automation Update Management
Deploy AI and machine learning at the edge by using Azure Stack Edge
Design a hybrid Domain Name System solution with Azure
Hybrid file services
 Hybrid availability and performance monitoring
Hybrid Security Monitoring using Azure Security Center and Azure Sentinel
Manage hybrid Azure workloads using Windows Admin Center
Azure Arc hybrid management and deployment for Kubernetes clusters
Connect standalone servers by using Azure Network Adapter
Disaster Recovery for Azure Stack Hub virtual machines
On-premises data gateway for Azure Logic Apps
Run containers in a hybrid environment
Connect an on-premises network to Azure
Connect an on-premises network to Azure using ExpressRoute
Cross cloud scaling
Cross-platform chat
Extend an on-premises network using ExpressRoute
Extend an on-premises network using VPN
Hybrid connections
Troubleshoot a hybrid VPN connection
Connect using Windows Virtual Desktop
Windows Virtual Desktop for enterprises
Multiple Active Directory forests
Multiple forests with Azure AD DS
Identity
Guides
Identity in multitenant applications
Introduction
The Tailspin scenario
Authentication
Claims-based identity
Tenant sign-up
Application roles
Authorization
Secure a web API
Cache access tokens
 Client assertion
Federate with a customer's AD FS
Architectures
AD DS resource forests in Azure
Deploy AD DS in an Azure virtual network
Extend on-premises AD FS to Azure
Hybrid Identity
Integrate on-premises AD domains with Azure AD
Integrate on-premises AD with Azure
Azure AD identity management for AWS
Integration
Architectures
Basic enterprise integration on Azure
Enterprise business intelligence
Enterprise integration using queues and events
Publishing internal APIs to external users
Web and Mobile front-ends
Custom Business Processes
Line of Business Extension
On-premises data gateway for Azure Logic Apps
Internet of Things
Guides
Azure IoT Edge Vision
Overview
Camera selection
Hardware acceleration
Machine learning
Image storage
Alert persistence
User interface
Architectures
IoT reference architecture
 Condition Monitoring
IoT and data analytics
IoT using Cosmos DB
Predictive Maintenance for Industrial IoT
IoT Connected Platform
Contactless IoT interfaces
COVID-19 IoT Safe Solutions
Lighting and disinfection system
Light and power for emerging markets
Predictive maintenance with IoT
Process real-time vehicle data using IoT
Safe Buildings with IoT and Azure
Secure access to IoT apps with Azure AD
Voice assistants and IoT devices
IoT using Azure Data Explorer
Buy online, pickup in store
Project 15 Open Platform
Industrial IoT Analytics
Architecture
Recommended services
Data visualization
Considerations
IoT concepts
Introduction to IoT solutions
Devices, platform, and applications
Attestation, authentication, and provisioning
Field and cloud edge gateways
Application-to-device commands
Builders, developers, and operators
IoT patterns
Measure and control loop
Monitor and manage loop
 Analyze and optimize loop
Event routing
Solution scaling with application stamps
Management and Governance
Architectures
Archive on-premises data to cloud
Back up cloud applications
Back up on-premises applications
Centralize app configuration and security
Computer forensics
High availability for BCDR
Data Sovereignty & Data Gravity
Enterprise-scale disaster recovery
Updating Windows VMs in Azure
Highly available SharePoint Server 2016
SMB disaster recovery with Azure Site Recovery
SMB disaster recovery with Double-Take DR
Manage configurations for Azure Arc enabled servers
Azure Automation in a hybrid environment
Back up files and applications on Azure Stack Hub
Azure Automation Update Management
Hybrid availability and performance monitoring
Manage hybrid Azure workloads using Windows Admin Center
Azure Arc hybrid management and deployment for Kubernetes clusters
Disaster Recovery for Azure Stack Hub virtual machines
Media
Architectures
Instant broadcasting with serverless
Live streaming digital media
Video-on-demand digital media
Gridwich media processing system
Gridwich architecture
 Gridwich concepts
Clean monolith design
Saga orchestration
Project names and structure
Gridwich CI/CD
Content protection and DRM
Gridwich Media Services
Gridwich Storage Service
Gridwich logging
Gridwich message formats
Pipeline variables to Terraform flow
Gridwich procedures
Set up Azure DevOps
Run Azure admin scripts
Set up local dev environment
Create new cloud environment
Maintain and rotate keys
Test Media Services V3 encoding
Migration
Architectures
Adding modern front-ends to legacy apps
Banking system
Banking cloud transformation
Patterns and implementations
JMeter implementation reference
Lift and shift LOB apps
Lift and shift with AKS
Oracle database migration
Migration decision process
Cross-cloud connectivity
Lift and shift to Azure VMs
Refactor
 Rearchitect
Serverless computing LOB apps
Unlock Legacy Data with Azure Stack
Decompose apps with Service Fabric
SQL 2008 R2 failover cluster in Azure
Migrate mainframe data to Azure
Migrate Unisys mainframes to Azure
Refactor IBM z/OS mainframe CF on Azure
Mixed Reality
Architectures
Mixed reality design reviews
Facilities management with mixed reality
Training powered by mixed reality
Mobile
Architectures
Custom mobile workforce app
Scalable apps with Azure MySQL
Scalable apps using Azure PostgreSQL
Social app for with authentication
Task-based consumer mobile app
Networking
Guides
Add IP spaces to peered virtual networks
Azure Firewall Architecture Guide
Architectures
Virtual network peering and VPN gateways
Deploy highly available NVAs
High availability for IaaS apps
Hub-spoke network topology in Azure
Implement a secure hybrid network
Segmenting Virtual Networks
Azure Automation Update Management
 Design a hybrid Domain Name System solution with Azure
Hybrid availability and performance monitoring
Connect standalone servers by using Azure Network Adapter
SAP
Overview
Architectures
SAP HANA on Azure (Large Instances)
SAP HANA Scale-up on Linux
SAP NetWeaver on Windows on Azure
SAP S/4HANA in Linux on Azure
SAP BW/4HANA in Linux on Azure
SAP NetWeaver on SQL Server
SAP deployment using an Oracle DB
Dev/test for SAP
Security
Architectures
Azure AD in Security Operations
Cyber threat intelligence
Highly-secure IaaS apps
Homomorphic encryption with SEAL
Real-time fraud detection
Secure OBO refresh tokens
Securely managed web apps
Web app private database connectivity
Virtual network integrated microservices
Virtual Network security options
Hybrid Security Monitoring using Azure Security Center and Azure Sentinel
MCAS and Azure Sentinel security for AWS
Healthcare platform confidential computing
Storage
Architectures
Media rendering
 Medical data storage
HIPAA/HITRUST Health Data and AI
Using Azure file shares in a hybrid environment
Hybrid file services
Web
Architectures
Basic web application
Deployment in App Service Environments
Standard deployment
High availability deployment
E-commerce front end
E-commerce website running in ASE
Highly available SharePoint farm
Highly available multi-region web application
Hybrid SharePoint farm with Microsoft 365
Intelligent product search engine for e-commerce
Magento e-commerce in AKS
Migrate a web app using Azure APIM
Multi-region N-tier application
Multitenant SaaS
Multi-tier web application built for HA/DR
SAP S/4 HANA for Large Instances
Scalable e-commerce web app
Scalable Episerver marketing website
Scalable Sitecore marketing website
Scalable Umbraco CMS web app
Scalable and secure WordPress on Azure
Scalable order processing
Scalable web app
More Scalable web apps
Serverless web app
Simple branded website
 Simple digital marketing website
Web app monitoring on Azure
Web and mobile applications with MySQL, Cosmos DB, and Redis
Dynamics Business Central as a Service on Azure
Azure Functions in a hybrid environment
Cloud Adoption Framework
 Azure architecture icons
12/18/2020 • 2 minutes to read • Edit Online

Helping our customers design and architect new solutions is core to the Azure Architecture Center’s mission.
Architecture diagrams like those included in our guidance can help communicate design decisions and the
relationships between components of a given workload. On this page you will find an official collection of Azure
architecture icons including Azure product icons to help you build a custom architecture diagram for your next
solution.
Do’s
Use the icon to illustrate how products can work together
In diagrams, we recommend to include the product name somewhere close to the icon
Use the icons as they would appear within Azure
Don’ts
Don’t crop, flip or rotate icons
Don’t distort or change icon shape in any way
Don’t use Microsoft product icons to represent your product or service

Example architecture diagram

Browse all Azure architectures to view other examples.

Icon updates
As of November 2020, the folder structure of our collection of Azure architecture icons has changed. The FAQs and
Terms of Use PDF files appear in the first level when you download the SVG icons below. The files in the icons folder
are the same except there is no longer a CXP folder. If you encounter any issues, let us know.

Terms
Microsoft permits the use of these icons in architectural diagrams, training materials, or documentation. You may
copy, distribute, and display the icons only for the permitted use unless granted explicit permission by Microsoft.
Microsoft reserves all other rights.
I agree to the above terms
D O W NLO AD SVG
ICO NS
 What's new in the Azure Architecture Center
12/18/2020 • 4 minutes to read • Edit Online

New and updated articles in the Azure Architecture Center

December 2020
New Articles
Performance testing
Testing tools
Refactor IBM z/OS mainframe Coupling Facility (CF) to Azure
Design scalable Azure applications
Plan for capacity
Design Azure applications for efficiency
Design for scaling
Updated Articles
Baseline architecture for an Azure Kubernetes Service (AKS) cluster (#32842d421)

November 2020
New Articles
Use Azure Stack HCI stretched clusters for disaster recovery
Use Azure Stack HCI switchless interconnect and lightweight quorum for Remote Office/Branch Office
Release Engineering Application Development
Release Engineering Continuous integration
Release Engineering Rollback
Project 15 from Microsoft Open Platform for Conservation and Ecological Sustainability Solutions
Unisys mainframe migration
Security logs and audits
Check for identity, network, data risks
Security operations in Azure
Security health modeling in Azure
Azure enterprise cloud file share
Modernize mainframe & midrange data
IoT event routing
Create a Gridwich environment
Gridwich cloud media system
Gridwich CI/CD pipeline
Gridwich clean monolith architecture
Gridwich content protection and DRM
Logging in Gridwich
Gridwich request-response messages
Gridwich project naming and namespaces
Gridwich saga orchestration
 Gridwich Storage Service
Gridwich keys and secrets management
Gridwich Media Services setup and scaling
Gridwich pipeline-generated admin scripts
Gridwich Azure DevOps setup
Gridwich local development environment setup
Test Media Services V3 encoding
Gridwich variable flow
Updated Articles
Choosing a data storage technology (#4128cc2d9)
Process real-time vehicle data using IoT (#beeba69f6)
Security monitoring tools in Azure (#4f3a35043)
Building a CI/CD pipeline for microservices on Kubernetes (#c0135f775)

October 2020
New Articles
SQL Server 2008 R2 failover cluster in Azure
Kafka on Azure
Secure application's configuration and dependencies
Application classification for security
Application threat analysis
AKS triage - cluster health
AKS triage - container registry connectivity
AKS triage - admission controllers
AKS triage - workload deployments
AKS triage - node health
Azure Kubernetes Service (AKS) operations triage
Alerts in IoT Edge Vision
Camera selection for IoT Edge Vision
Hardware for IoT Edge Vision
Image storage in IoT Edge Vision
Azure IoT Edge Vision
Machine learning in IoT Edge Vision
User interface in IoT Edge Vision
Configure infrastructure
Repeatable Infrastructure
Automation overview of goals, best practices, and types in Azure
Automated Tasks
Partitioning in Event Hubs and Kafka
Retail - Buy online, pickup in store (BOPIS)
Magento e-commerce platform in Azure Kubernetes Service (AKS)
Web app private connectivity to Azure SQL database
Updated Articles
Security with identity and access management (IAM) in Azure (#2a1154709)
Regulatory compliance (#2a1154709)
 Computer forensics Chain of Custody in Azure (#909b776f4)
Overview of the performance efficiency pillar (#ed89cf6ab)
Azure Kubernetes Service (AKS) solution journey (#b63ab6a9f)
GCP to Azure Services Comparison (#a45091c00)
Resiliency patterns (#b5201626c)
Choosing an Azure compute service (#a64329288)
Security patterns (#13add8a06)

September 2020
New Articles
Administrative account security
Enforce governance to reduce risks
Security management groups
Regulatory compliance
Team roles and responsibilities
Segmentation strategies
Tenancy model for SaaS applications
Migrate IBM mainframe applications to Azure with TmaxSoft OpenFrame
Security monitoring tools in Azure
Applications and services
Security storage in Azure | Microsoft Docs
Azure Active Directory IDaaS in Security Operations
Capture cost requirements for an Azure
Tradeoffs for costs
Microsoft Azure Well-Architected Framework
Overview of the security pillar
Storage, data, and encryption in Azure | Microsoft Docs
Move Azure resources across regions
FSLogix for the enterprise
Stromasys Charon-SSP Solaris emulator on Azure VMs
Azure Kubernetes Service (AKS) solution journey
IoT analyze and optimize loops
IoT measure and control loops
IoT monitor and manage loops
Hybrid and Multicloud Architectures
Azure Arc hybrid management and deployment for Kubernetes clusters
Manage configurations for Azure Arc enabled servers
Azure Automation in a hybrid environment
Using Azure file shares in a hybrid environment
Azure Functions in a hybrid environment
Connect standalone servers by using Azure Network Adapter
Back up files and applications on Azure Stack Hub
Disaster Recovery for Azure Stack Hub virtual machines
Azure Automation Update Management
Deploy AI and machine learning at the edge by using Azure Stack Edge
On-premises data gateway for Azure Logic Apps
 Run containers in a hybrid environment
Design a hybrid Domain Name System solution with Azure
Hybrid file services
Hybrid availability and performance monitoring
Hybrid Security Monitoring using Azure Security Center and Azure Sentinel
Manage hybrid Azure workloads using Windows Admin Center
DevSecOps in GitHub
Network security review
Network security strategies
Security with identity and access management (IAM) in Azure
Azure Messaging cost estimates
Web application cost estimates
Updated Articles
DevTest and DevOps for IaaS solutions (#a2a167058)
DevTest and DevOps for microservice solutions (#a2a167058)
DevTest and DevOps for PaaS solutions (#a2a167058)
Baseline architecture for an Azure Kubernetes Service (AKS) cluster (#9b20a025d)
DevSecOps in Azure (#511e6ee92)

August 2020
New Articles
Multiple forests with AD DS, Azure AD, and Azure AD DS
Multiple forests with AD DS and Azure AD
Virtual network integrated serverless microservices
Big data analytics with Azure Data Explorer
Content Delivery Network analytics
Azure Data Explorer interactive analytics
IoT analytics with Azure Data Explorer
Azure Data Explorer monitoring
Custom Business Processes
Web and Mobile Front Ends
Line of Business Extension
Attestation, authentication, and provisioning
Field and cloud edge gateways
Condition Monitoring for Industrial IoT
Predictive Maintenance for Industrial IoT
Banking system cloud transformation on Azure
JMeter implementation reference for load testing pipeline solution
Patterns and implementations
IoT connected light, power, and internet for emerging markets
Compute
Scale IoT solutions with application stamps
Builders, developers, and operators
IoT application-to-device commands
IoT solution architecture
 IoT solutions conceptual overview
Use cached data
Criteria for choosing a data store
Data store decision tree
Updated Articles
Network security and containment in Azure | Microsoft Docs (#75626e3a7)
Data store cost estimates (#2b0e692f9)
Retry guidance for Azure services (#6c8a169c9)
Chaos engineering (#f742721a9)
Understand data store models (#4fbdd828a)
Cost governance for an Azure workload (#a3452805a)
Event-based cloud automation (#d2cca2011)
Serverless event processing (#d2cca2011)
 Azure Application Architecture Guide
12/18/2020 • 3 minutes to read • Edit Online

This guide presents a structured approach for designing applications on Azure that are scalable, secure, resilient,
and highly available. It is based on proven practices that we have learned from customer engagements.

Introduction
The cloud is changing how applications are designed and secured. Instead of monoliths, applications are
decomposed into smaller, decentralized services. These services communicate through APIs or by using
asynchronous messaging or eventing. Applications scale horizontally, adding new instances as demand requires.
These trends bring new challenges. Application state is distributed. Operations are done in parallel and
asynchronously. Applications must be resilient when failures occur. Malicious actors continuously target
applications. Deployments must be automated and predictable. Monitoring and telemetry are critical for gaining
insight into the system. This guide is designed to help you navigate these changes.

T RA DIT IO N A L O N - P REM ISES M O DERN C LO UD

Monolithic Decomposed
Designed for predictable scalability Designed for elastic scale
Relational database Polyglot persistence (mix of storage technologies)
Synchronized processing Asynchronous processing
Design to avoid failures (MTBF) Design for failure (MTTR)
Occasional large updates Frequent small updates
Manual management Automated self-management
Snowflake servers Immutable infrastructure

How this guide is structured

The Azure Application Architecture Guide is organized as a series of steps, from the architecture and design to
implementation. For each step, there is supporting guidance that will help you with the design of your application
architecture.
Architecture style
N-tier Micro services W eb -qu eu e-
w o rker

Even t driven B ig Co mp u te B ig Data

Technology choices
Co mp u te Data sto res Messagin g

Application architecture
Referen ce Design Design B est
arch itectu res p rin cip les p attern s p ractices

Microsoft Azure Well-Architected Framework

Co st O p eratio n al Perfo rman ce
o p timizatio n excellen ce efficien cy Reliab ility Secu rity
Architecture styles
The first decision point is the most fundamental. What kind of architecture are you building? It might be a
microservices architecture, a more traditional N-tier application, or a big data solution. We have identified several
distinct architecture styles. There are benefits and challenges to each.
Learn more: Architecture styles

Technology choices
Knowing the type of architecture you are building, now you can start to choose the main technology pieces for the
architecture. The following technology choices are critical:
Compute refers to the hosting model for the computing resources that your applications run on. For more
information, see Choose a compute service.
Data stores include databases but also storage for message queues, caches, logs, and anything else that an
application might persist to storage. For more information, see Choose a data store.
Messaging technologies enable asynchronous messages between components of the system. For more
information, see Choose a messaging service.
You will probably have to make additional technology choices along the way, but these three elements (compute,
data, and messaging) are central to most cloud applications and will determine many aspects of your design.

Design the architecture

Once you have chosen the architecture style and the major technology components, you are ready to tackle the
specific design of your application. Every application is different, but the following resources can help you along the
way:
Reference architectures
Depending on your scenario, one of our reference architectures may be a good starting point. Each reference
architecture includes recommended practices, along with considerations for scalability, availability, security,
resilience, and other aspects of the design. Most also include a deployable solution or reference implementation.
Design principles
We have identified 10 high-level design principles that will make your application more scalable, resilient, and
manageable. These design principles apply to any architecture style. Throughout the design process, keep these 10
high-level design principles in mind. For more information, see Design principles.
Design patterns
Software design patterns are repeatable patterns that are proven to solve specific problems. Our catalog of Cloud
design patterns addresses specific challenges in distributed systems. They address aspects such as availability,
resiliency, performance, and security. You can find our catalog of design patterns here.
Best practices
Our best practices articles cover various design considerations including API design, autoscaling, data partitioning,
caching, and so forth. Review these and apply the best practices that are appropriate for your application.
Security best practices
Our security best practices describe how to ensure that the confidentiality, integrity, and availability of your
application aren't compromised by malicious actors.

Quality pillars
A successful cloud application will focus on five pillars of software quality: Cost optimization, Operational
excellence, Performance efficiency, Reliability, and Security.
Leverage the Microsoft Azure Well-Architected Framework to assess your architecture across these five pillars.

Next steps
Architecture styles
 Architecture styles
12/18/2020 • 5 minutes to read • Edit Online

An architecture style is a family of architectures that share certain characteristics. For example, N-tier is a common
architecture style. More recently, microservice architectures have started to gain favor. Architecture styles don't
require the use of particular technologies, but some technologies are well-suited for certain architectures. For
example, containers are a natural fit for microservices.
We have identified a set of architecture styles that are commonly found in cloud applications. The article for each
style includes:
A description and logical diagram of the style.
Recommendations for when to choose this style.
Benefits, challenges, and best practices.
A recommended deployment using relevant Azure services.

A quick tour of the styles

This section gives a quick tour of the architecture styles that we've identified, along with some high-level
considerations for their use. Read more details in the linked topics.
N -tier
N-tier is a traditional architecture for enterprise applications. Dependencies are managed by dividing the
application into layers that perform logical functions, such as presentation, business logic, and data access. A layer
can only call into layers that sit below it. However, this horizontal layering can be a liability. It can be hard to
introduce changes in one part of the application without touching the rest of the application. That makes frequent
updates a challenge, limiting how quickly new features can be added.
N-tier is a natural fit for migrating existing applications that already use a layered architecture. For that reason, N-
tier is most often seen in infrastructure as a service (IaaS) solutions, or application that use a mix of IaaS and
managed services.
Web-Queue -Worker
For a purely PaaS solution, consider a Web-Queue-Worker architecture. In this style, the application has a web
front end that handles HTTP requests and a back-end worker that performs CPU-intensive tasks or long-running
operations. The front end communicates to the worker through an asynchronous message queue.
Web-queue-worker is suitable for relatively simple domains with some resource-intensive tasks. Like N-tier, the
architecture is easy to understand. The use of managed services simplifies deployment and operations. But with
complex domains, it can be hard to manage dependencies. The front end and the worker can easily become large,
monolithic components that are hard to maintain and update. As with N-tier, this can reduce the frequency of
updates and limit innovation.
Microservices
If your application has a more complex domain, consider moving to a Microser vices architecture. A
microservices application is composed of many small, independent services. Each service implements a single
business capability. Services are loosely coupled, communicating through API contracts.
Each service can be built by a small, focused development team. Individual services can be deployed without a lot
of coordination between teams, which encourages frequent updates. A microservice architecture is more complex
to build and manage than either N-tier or web-queue-worker. It requires a mature development and DevOps
culture. But done right, this style can lead to higher release velocity, faster innovation, and a more resilient
architecture.
Event-driven architecture
Event-Driven Architectures use a publish-subscribe (pub-sub) model, where producers publish events, and
consumers subscribe to them. The producers are independent from the consumers, and consumers are
independent from each other.
Consider an event-driven architecture for applications that ingest and process a large volume of data with very low
latency, such as IoT solutions. The style is also useful when different subsystems must perform different types of
processing on the same event data.
Big Data, Big Compute
Big Data and Big Compute are specialized architecture styles for workloads that fit certain specific profiles. Big
data divides a very large dataset into chunks, performing parallel processing across the entire set, for analysis and
reporting. Big compute, also called high-performance computing (HPC), makes parallel computations across a
large number (thousands) of cores. Domains include simulations, modeling, and 3-D rendering.

Architecture styles as constraints

An architecture style places constraints on the design, including the set of elements that can appear and the
allowed relationships between those elements. Constraints guide the "shape" of an architecture by restricting the
universe of choices. When an architecture conforms to the constraints of a particular style, certain desirable
properties emerge.
For example, the constraints in microservices include:
A service represents a single responsibility.
Every service is independent of the others.
Data is private to the service that owns it. Services do not share data.
By adhering to these constraints, what emerges is a system where services can be deployed independently, faults
are isolated, frequent updates are possible, and it's easy to introduce new technologies into the application.
Before choosing an architecture style, make sure that you understand the underlying principles and constraints of
that style. Otherwise, you can end up with a design that conforms to the style at a superficial level, but does not
achieve the full potential of that style. It's also important to be pragmatic. Sometimes it's better to relax a
constraint, rather than insist on architectural purity.
The following table summarizes how each style manages dependencies, and the types of domain that are best
suited for each.

A RC H IT EC T URE ST Y L E DEP EN DEN C Y M A N A GEM EN T DO M A IN T Y P E

N-tier Horizontal tiers divided by subnet Traditional business domain. Frequency

of updates is low.

Web-Queue-Worker Front and backend jobs, decoupled by Relatively simple domain with some
async messaging. resource intensive tasks.

Microservices Vertically (functionally) decomposed Complicated domain. Frequent updates.

services that call each other through
APIs.

Event-driven architecture. Producer/consumer. Independent view IoT and real-time systems

per sub-system.
 A RC H IT EC T URE ST Y L E DEP EN DEN C Y M A N A GEM EN T DO M A IN T Y P E

Big data Divide a huge dataset into small Batch and real-time data analysis.
chunks. Parallel processing on local Predictive analysis using ML.
datasets.

Big compute Data allocation to thousands of cores. Compute intensive domains such as
simulation.

Consider challenges and benefits

Constraints also create challenges, so it's important to understand the trade-offs when adopting any of these
styles. Do the benefits of the architecture style outweigh the challenges, for this subdomain and bounded context.
Here are some of the types of challenges to consider when selecting an architecture style:
Complexity . Is the complexity of the architecture justified for your domain? Conversely, is the style too
simplistic for your domain? In that case, you risk ending up with a "big ball of mud", because the
architecture does not help you to manage dependencies cleanly.
Asynchronous messaging and eventual consistency . Asynchronous messaging can be used to
decouple services, and increase reliability (because messages can be retried) and scalability. However, this
also creates challenges in handling eventual consistency, as well as the possibility of duplicate messages.
Inter-ser vice communication . As you decompose an application into separate services, there is a risk
that communication between services will cause unacceptable latency or create network congestion (for
example, in a microservices architecture).
Manageability . How hard is it to manage the application, monitor, deploy updates, and so on?
 Big compute architecture style
12/18/2020 • 3 minutes to read • Edit Online

The term big compute describes large-scale workloads that require a large number of cores, often numbering in
the hundreds or thousands. Scenarios include image rendering, fluid dynamics, financial risk modeling, oil
exploration, drug design, and engineering stress analysis, among others.

Here are some typical characteristics of big compute applications:

The work can be split into discrete tasks, which can be run across many cores simultaneously.
Each task is finite. It takes some input, does some processing, and produces output. The entire application runs
for a finite amount of time (minutes to days). A common pattern is to provision a large number of cores in a
burst, and then spin down to zero once the application completes.
The application does not need to stay up 24/7. However, the system must handle node failures or application
crashes.
For some applications, tasks are independent and can run in parallel. In other cases, tasks are tightly coupled,
meaning they must interact or exchange intermediate results. In that case, consider using high-speed
networking technologies such as InfiniBand and remote direct memory access (RDMA).
Depending on your workload, you might use compute-intensive VM sizes (H16r, H16mr, and A9).

When to use this architecture

Computationally intensive operations such as simulation and number crunching.
Simulations that are computationally intensive and must be split across CPUs in multiple computers (10-
1000s).
Simulations that require too much memory for one computer, and must be split across multiple computers.
Long-running computations that would take too long to complete on a single computer.
Smaller computations that must be run 100s or 1000s of times, such as Monte Carlo simulations.

Benefits
High performance with "embarrassingly parallel" processing.
Can harness hundreds or thousands of computer cores to solve large problems faster.
 Access to specialized high-performance hardware, with dedicated high-speed InfiniBand networks.
You can provision VMs as needed to do work, and then tear them down.

Challenges
Managing the VM infrastructure.
Managing the volume of number crunching
Provisioning thousands of cores in a timely manner.
For tightly coupled tasks, adding more cores can have diminishing returns. You may need to experiment to find
the optimum number of cores.

Big compute using Azure Batch

Azure Batch is a managed service for running large-scale high-performance computing (HPC) applications.
Using Azure Batch, you configure a VM pool, and upload the applications and data files. Then the Batch service
provisions the VMs, assign tasks to the VMs, runs the tasks, and monitors the progress. Batch can automatically
scale out the VMs in response to the workload. Batch also provides job scheduling.

Big compute running on Virtual Machines

You can use Microsoft HPC Pack to administer a cluster of VMs, and schedule and monitor HPC jobs. With this
approach, you must provision and manage the VMs and network infrastructure. Consider this approach if you
have existing HPC workloads and want to move some or all it to Azure. You can move the entire HPC cluster to
Azure, or you can keep your HPC cluster on-premises but use Azure for burst capacity. For more information, see
Batch and HPC solutions for large-scale computing workloads.
HPC Pack deployed to Azure
In this scenario, the HPC cluster is created entirely within Azure.
The head node provides management and job scheduling services to the cluster. For tightly coupled tasks, use an
RDMA network that provides very high bandwidth, low latency communication between VMs. For more
information, see Deploy an HPC Pack 2016 cluster in Azure.
Burst an HPC cluster to Azure
In this scenario, an organization is running HPC Pack on-premises, and uses Azure VMs for burst capacity. The
cluster head node is on-premises. ExpressRoute or VPN Gateway connects the on-premises network to the Azure
VNet.
 Big data architecture style
12/18/2020 • 10 minutes to read • Edit Online

A big data architecture is designed to handle the ingestion, processing, and analysis of data that is too large or
complex for traditional database systems.

Data Storage Batch

Processing
Data Analytical Analytics
Sources Data Store and
Reporting
Real-Time Message Stream
Ingestion Processing

Orchestration

Big data solutions typically involve one or more of the following types of workload:
Batch processing of big data sources at rest.
Real-time processing of big data in motion.
Interactive exploration of big data.
Predictive analytics and machine learning.
Most big data architectures include some or all of the following components:
Data sources : All big data solutions start with one or more data sources. Examples include:
Application data stores, such as relational databases.
Static files produced by applications, such as web server log files.
Real-time data sources, such as IoT devices.
Data storage : Data for batch processing operations is typically stored in a distributed file store that can
hold high volumes of large files in various formats. This kind of store is often called a data lake. Options for
implementing this storage include Azure Data Lake Store or blob containers in Azure Storage.
Batch processing : Because the data sets are so large, often a big data solution must process data files
using long-running batch jobs to filter, aggregate, and otherwise prepare the data for analysis. Usually these
jobs involve reading source files, processing them, and writing the output to new files. Options include
running U-SQL jobs in Azure Data Lake Analytics, using Hive, Pig, or custom Map/Reduce jobs in an
HDInsight Hadoop cluster, or using Java, Scala, or Python programs in an HDInsight Spark cluster.
Real-time message ingestion : If the solution includes real-time sources, the architecture must include a
way to capture and store real-time messages for stream processing. This might be a simple data store,
where incoming messages are dropped into a folder for processing. However, many solutions need a
message ingestion store to act as a buffer for messages, and to support scale-out processing, reliable
delivery, and other message queuing semantics. Options include Azure Event Hubs, Azure IoT Hubs, and
Kafka.
Stream processing : After capturing real-time messages, the solution must process them by filtering,
aggregating, and otherwise preparing the data for analysis. The processed stream data is then written to an
output sink. Azure Stream Analytics provides a managed stream processing service based on perpetually
 running SQL queries that operate on unbounded streams. You can also use open source Apache streaming
technologies like Storm and Spark Streaming in an HDInsight cluster.
Analytical data store : Many big data solutions prepare data for analysis and then serve the processed
data in a structured format that can be queried using analytical tools. The analytical data store used to serve
these queries can be a Kimball-style relational data warehouse, as seen in most traditional business
intelligence (BI) solutions. Alternatively, the data could be presented through a low-latency NoSQL
technology such as HBase, or an interactive Hive database that provides a metadata abstraction over data
files in the distributed data store. Azure Synapse Analytics provides a managed service for large-scale,
cloud-based data warehousing. HDInsight supports Interactive Hive, HBase, and Spark SQL, which can also
be used to serve data for analysis.
Analysis and repor ting : The goal of most big data solutions is to provide insights into the data through
analysis and reporting. To empower users to analyze the data, the architecture may include a data modeling
layer, such as a multidimensional OLAP cube or tabular data model in Azure Analysis Services. It might also
support self-service BI, using the modeling and visualization technologies in Microsoft Power BI or Microsoft
Excel. Analysis and reporting can also take the form of interactive data exploration by data scientists or data
analysts. For these scenarios, many Azure services support analytical notebooks, such as Jupyter, enabling
these users to leverage their existing skills with Python or R. For large-scale data exploration, you can use
Microsoft R Server, either standalone or with Spark.
Orchestration : Most big data solutions consist of repeated data processing operations, encapsulated in
workflows, that transform source data, move data between multiple sources and sinks, load the processed
data into an analytical data store, or push the results straight to a report or dashboard. To automate these
workflows, you can use an orchestration technology such Azure Data Factory or Apache Oozie and Sqoop.
Azure includes many services that can be used in a big data architecture. They fall roughly into two categories:
Managed services, including Azure Data Lake Store, Azure Data Lake Analytics, Azure Synapse Analytics, Azure
Stream Analytics, Azure Event Hub, Azure IoT Hub, and Azure Data Factory.
Open source technologies based on the Apache Hadoop platform, including HDFS, HBase, Hive, Pig, Spark,
Storm, Oozie, Sqoop, and Kafka. These technologies are available on Azure in the Azure HDInsight service.
These options are not mutually exclusive, and many solutions combine open source technologies with Azure
services.

When to use this architecture

Consider this architecture style when you need to:
Store and process data in volumes too large for a traditional database.
Transform unstructured data for analysis and reporting.
Capture, process, and analyze unbounded streams of data in real time, or with low latency.
Use Azure Machine Learning or Microsoft Cognitive Services.

Benefits
Technology choices . You can mix and match Azure managed services and Apache technologies in HDInsight
clusters, to capitalize on existing skills or technology investments.
Performance through parallelism . Big data solutions take advantage of parallelism, enabling high-
performance solutions that scale to large volumes of data.
Elastic scale . All of the components in the big data architecture support scale-out provisioning, so that you can
adjust your solution to small or large workloads, and pay only for the resources that you use.
Interoperability with existing solutions . The components of the big data architecture are also used for IoT
 processing and enterprise BI solutions, enabling you to create an integrated solution across data workloads.

Challenges
Complexity . Big data solutions can be extremely complex, with numerous components to handle data
ingestion from multiple data sources. It can be challenging to build, test, and troubleshoot big data processes.
Moreover, there may be a large number of configuration settings across multiple systems that must be used in
order to optimize performance.
Skillset . Many big data technologies are highly specialized, and use frameworks and languages that are not
typical of more general application architectures. On the other hand, big data technologies are evolving new
APIs that build on more established languages. For example, the U-SQL language in Azure Data Lake Analytics is
based on a combination of Transact-SQL and C#. Similarly, SQL-based APIs are available for Hive, HBase, and
Spark.
Technology maturity . Many of the technologies used in big data are evolving. While core Hadoop
technologies such as Hive and Pig have stabilized, emerging technologies such as Spark introduce extensive
changes and enhancements with each new release. Managed services such as Azure Data Lake Analytics and
Azure Data Factory are relatively young, compared with other Azure services, and will likely evolve over time.
Security . Big data solutions usually rely on storing all static data in a centralized data lake. Securing access to
this data can be challenging, especially when the data must be ingested and consumed by multiple applications
and platforms.

Best practices
Leverage parallelism . Most big data processing technologies distribute the workload across multiple
processing units. This requires that static data files are created and stored in a splittable format. Distributed
file systems such as HDFS can optimize read and write performance, and the actual processing is performed
by multiple cluster nodes in parallel, which reduces overall job times.
Par tition data . Batch processing usually happens on a recurring schedule — for example, weekly or
monthly. Partition data files, and data structures such as tables, based on temporal periods that match the
processing schedule. That simplifies data ingestion and job scheduling, and makes it easier to troubleshoot
failures. Also, partitioning tables that are used in Hive, U-SQL, or SQL queries can significantly improve
query performance.
Apply schema-on-read semantics . Using a data lake lets you to combine storage for files in multiple
formats, whether structured, semi-structured, or unstructured. Use schema-on-read semantics, which
project a schema onto the data when the data is processing, not when the data is stored. This builds
flexibility into the solution, and prevents bottlenecks during data ingestion caused by data validation and
type checking.
Process data in-place . Traditional BI solutions often use an extract, transform, and load (ETL) process to
move data into a data warehouse. With larger volumes data, and a greater variety of formats, big data
solutions generally use variations of ETL, such as transform, extract, and load (TEL). With this approach, the
data is processed within the distributed data store, transforming it to the required structure, before moving
the transformed data into an analytical data store.
Balance utilization and time costs . For batch processing jobs, it's important to consider two factors: The
per-unit cost of the compute nodes, and the per-minute cost of using those nodes to complete the job. For
example, a batch job may take eight hours with four cluster nodes. However, it might turn out that the job
uses all four nodes only during the first two hours, and after that, only two nodes are required. In that case,
running the entire job on two nodes would increase the total job time, but would not double it, so the total
cost would be less. In some business scenarios, a longer processing time may be preferable to the higher
cost of using underutilized cluster resources.
 Separate cluster resources . When deploying HDInsight clusters, you will normally achieve better
performance by provisioning separate cluster resources for each type of workload. For example, although
Spark clusters include Hive, if you need to perform extensive processing with both Hive and Spark, you
should consider deploying separate dedicated Spark and Hadoop clusters. Similarly, if you are using HBase
and Storm for low latency stream processing and Hive for batch processing, consider separate clusters for
Storm, HBase, and Hadoop.
Orchestrate data ingestion . In some cases, existing business applications may write data files for batch
processing directly into Azure storage blob containers, where they can be consumed by HDInsight or Azure
Data Lake Analytics. However, you will often need to orchestrate the ingestion of data from on-premises or
external data sources into the data lake. Use an orchestration workflow or pipeline, such as those supported
by Azure Data Factory or Oozie, to achieve this in a predictable and centrally manageable fashion.
Scrub sensitive data early . The data ingestion workflow should scrub sensitive data early in the process,
to avoid storing it in the data lake.

IoT architecture
Internet of Things (IoT) is a specialized subset of big data solutions. The following diagram shows a possible logical
architecture for IoT. The diagram emphasizes the event-streaming components of the architecture.

The cloud gateway ingests device events at the cloud boundary, using a reliable, low latency messaging system.
Devices might send events directly to the cloud gateway, or through a field gateway . A field gateway is a
specialized device or software, usually colocated with the devices, that receives events and forwards them to the
cloud gateway. The field gateway might also preprocess the raw device events, performing functions such as
filtering, aggregation, or protocol transformation.
After ingestion, events go through one or more stream processors that can route the data (for example, to
storage) or perform analytics and other processing.
The following are some common types of processing. (This list is certainly not exhaustive.)
Writing event data to cold storage, for archiving or batch analytics.
Hot path analytics, analyzing the event stream in (near) real time, to detect anomalies, recognize patterns
over rolling time windows, or trigger alerts when a specific condition occurs in the stream.
Handling special types of non-telemetry messages from devices, such as notifications and alarms.
Machine learning.
The boxes that are shaded gray show components of an IoT system that are not directly related to event streaming,
but are included here for completeness.
The device registr y is a database of the provisioned devices, including the device IDs and usually device
metadata, such as location.
The provisioning API is a common external interface for provisioning and registering new devices.
Some IoT solutions allow command and control messages to be sent to devices.

This section has presented a very high-level view of IoT, and there are many subtleties and challenges to
consider. For a more detailed reference architecture and discussion, see the Microsoft Azure IoT Reference
Architecture (PDF download).

Next steps
Learn more about big data architectures.
 Event-driven architecture style
12/18/2020 • 3 minutes to read • Edit Online

An event-driven architecture consists of event producers that generate a stream of events, and event
consumers that listen for the events.

Event Consumers

Event Producers Event Ingestion Event Consumers

Event Consumers

Events are delivered in near real time, so consumers can respond immediately to events as they occur. Producers
are decoupled from consumers — a producer doesn't know which consumers are listening. Consumers are also
decoupled from each other, and every consumer sees all of the events. This differs from a Competing Consumers
pattern, where consumers pull messages from a queue and a message is processed just once (assuming no
errors). In some systems, such as IoT, events must be ingested at very high volumes.
An event driven architecture can use a pub/sub model or an event stream model.
Pub/sub : The messaging infrastructure keeps track of subscriptions. When an event is published, it sends
the event to each subscriber. After an event is received, it cannot be replayed, and new subscribers do not
see the event.
Event streaming : Events are written to a log. Events are strictly ordered (within a partition) and durable.
Clients don't subscribe to the stream, instead a client can read from any part of the stream. The client is
responsible for advancing its position in the stream. That means a client can join at any time, and can
replay events.
On the consumer side, there are some common variations:
Simple event processing . An event immediately triggers an action in the consumer. For example, you
could use Azure Functions with a Service Bus trigger, so that a function executes whenever a message is
published to a Service Bus topic.
Complex event processing . A consumer processes a series of events, looking for patterns in the event
data, using a technology such as Azure Stream Analytics or Apache Storm. For example, you could
aggregate readings from an embedded device over a time window, and generate a notification if the
moving average crosses a certain threshold.
Event stream processing . Use a data streaming platform, such as Azure IoT Hub or Apache Kafka, as a
pipeline to ingest events and feed them to stream processors. The stream processors act to process or
transform the stream. There may be multiple stream processors for different subsystems of the application.
This approach is a good fit for IoT workloads.
The source of the events may be external to the system, such as physical devices in an IoT solution. In that case,
the system must be able to ingest the data at the volume and throughput that is required by the data source.
In the logical diagram above, each type of consumer is shown as a single box. In practice, it's common to have
multiple instances of a consumer, to avoid having the consumer become a single point of failure in system.
Multiple instances might also be necessary to handle the volume and frequency of events. Also, a single consumer
might process events on multiple threads. This can create challenges if events must be processed in order or
require exactly-once semantics. See Minimize Coordination.

When to use this architecture

Multiple subsystems must process the same events.
Real-time processing with minimum time lag.
Complex event processing, such as pattern matching or aggregation over time windows.
High volume and high velocity of data, such as IoT.

Benefits
Producers and consumers are decoupled.
No point-to-point integrations. It's easy to add new consumers to the system.
Consumers can respond to events immediately as they arrive.
Highly scalable and distributed.
Subsystems have independent views of the event stream.

Challenges
Guaranteed delivery. In some systems, especially in IoT scenarios, it's crucial to guarantee that events are
delivered.
Processing events in order or exactly once. Each consumer type typically runs in multiple instances, for
resiliency and scalability. This can create a challenge if the events must be processed in order (within a
consumer type), or if the processing logic is not idempotent.
Additional considerations
The amount of data to include in an event can be a significant consideration that affects both performance and
cost. Putting all the relevant information needed for processing in the event itself can simplify the processing
code and save additional lookups. Putting the minimal amount of information in an event, like just a couple of
identifiers, will reduce transport time and cost, but requires the processing code to look up any additional
information it needs.
 Microservices architecture style
12/18/2020 • 5 minutes to read • Edit Online

A microservices architecture consists of a collection of small, autonomous services. Each service is self-contained
and should implement a single business capability.

What are microservices?

Microservices are small, independent, and loosely coupled. A single small team of developers can write and
maintain a service.
Each service is a separate codebase, which can be managed by a small development team.
Services can be deployed independently. A team can update an existing service without rebuilding and
redeploying the entire application.
Services are responsible for persisting their own data or external state. This differs from the traditional
model, where a separate data layer handles data persistence.
Services communicate with each other by using well-defined APIs. Internal implementation details of each
service are hidden from other services.
Services don't need to share the same technology stack, libraries, or frameworks.
Besides for the services themselves, some other components appear in a typical microservices architecture:
Management/orchestration . This component is responsible for placing services on nodes, identifying failures,
rebalancing services across nodes, and so forth. Typically this component is an off-the-shelf technology such as
Kubernetes, rather than something custom built.
API Gateway . The API gateway is the entry point for clients. Instead of calling services directly, clients call the API
gateway, which forwards the call to the appropriate services on the back end.
Advantages of using an API gateway include:
It decouples clients from services. Services can be versioned or refactored without needing to update all of
the clients.
Services can use messaging protocols that are not web friendly, such as AMQP.
 The API Gateway can perform other cross-cutting functions such as authentication, logging, SSL
termination, and load balancing.

Benefits
Agility. Because microservices are deployed independently, it's easier to manage bug fixes and feature
releases. You can update a service without redeploying the entire application, and roll back an update if
something goes wrong. In many traditional applications, if a bug is found in one part of the application, it
can block the entire release process. New features may be held up waiting for a bug fix to be integrated,
tested, and published.
Small, focused teams . A microservice should be small enough that a single feature team can build, test,
and deploy it. Small team sizes promote greater agility. Large teams tend be less productive, because
communication is slower, management overhead goes up, and agility diminishes.
Small code base . In a monolithic application, there is a tendency over time for code dependencies to
become tangled Adding a new feature requires touching code in a lot of places. By not sharing code or data
stores, a microservices architecture minimizes dependencies, and that makes it easier to add new features.
Mix of technologies . Teams can pick the technology that best fits their service, using a mix of technology
stacks as appropriate.
Fault isolation . If an individual microservice becomes unavailable, it won't disrupt the entire application,
as long as any upstream microservices are designed to handle faults correctly (for example, by
implementing circuit breaking).
Scalability . Services can be scaled independently, letting you scale out subsystems that require more
resources, without scaling out the entire application. Using an orchestrator such as Kubernetes or Service
Fabric, you can pack a higher density of services onto a single host, which allows for more efficient
utilization of resources.
Data isolation . It is much easier to perform schema updates, because only a single microservice is
affected. In a monolithic application, schema updates can become very challenging, because different parts
of the application may all touch the same data, making any alterations to the schema risky.

Challenges
The benefits of microservices don't come for free. Here are some of the challenges to consider before embarking
on a microservices architecture.
Complexity . A microservices application has more moving parts than the equivalent monolithic
application. Each service is simpler, but the entire system as a whole is more complex.
Development and testing . Writing a small service that relies on other dependent services requires a
different approach than a writing a traditional monolithic or layered application. Existing tools are not
always designed to work with service dependencies. Refactoring across service boundaries can be difficult.
It is also challenging to test service dependencies, especially when the application is evolving quickly.
Lack of governance . The decentralized approach to building microservices has advantages, but it can
also lead to problems. You may end up with so many different languages and frameworks that the
application becomes hard to maintain. It may be useful to put some project-wide standards in place,
without overly restricting teams' flexibility. This especially applies to cross-cutting functionality such as
logging.
Network congestion and latency . The use of many small, granular services can result in more
interservice communication. Also, if the chain of service dependencies gets too long (service A calls B,
which calls C...), the additional latency can become a problem. You will need to design APIs carefully. Avoid
 overly chatty APIs, think about serialization formats, and look for places to use asynchronous
communication patterns.
Data integrity . With each microservice responsible for its own data persistence. As a result, data
consistency can be a challenge. Embrace eventual consistency where possible.
Management . To be successful with microservices requires a mature DevOps culture. Correlated logging
across services can be challenging. Typically, logging must correlate multiple service calls for a single user
operation.
Versioning . Updates to a service must not break services that depend on it. Multiple services could be
updated at any given time, so without careful design, you might have problems with backward or forward
compatibility.
Skillset . Microservices are highly distributed systems. Carefully evaluate whether the team has the skills
and experience to be successful.

Best practices
Model services around the business domain.
Decentralize everything. Individual teams are responsible for designing and building services. Avoid
sharing code or data schemas.
Data storage should be private to the service that owns the data. Use the best storage for each service and
data type.
Services communicate through well-designed APIs. Avoid leaking implementation details. APIs should
model the domain, not the internal implementation of the service.
Avoid coupling between services. Causes of coupling include shared database schemas and rigid
communication protocols.
Offload cross-cutting concerns, such as authentication and SSL termination, to the gateway.
Keep domain knowledge out of the gateway. The gateway should handle and route client requests without
any knowledge of the business rules or domain logic. Otherwise, the gateway becomes a dependency and
can cause coupling between services.
Services should have loose coupling and high functional cohesion. Functions that are likely to change
together should be packaged and deployed together. If they reside in separate services, those services end
up being tightly coupled, because a change in one service will require updating the other service. Overly
chatty communication between two services may be a symptom of tight coupling and low cohesion.
Isolate failures. Use resiliency strategies to prevent failures within a service from cascading. See Resiliency
patterns and Designing reliable applications.

Next steps
For detailed guidance about building a microservices architecture on Azure, see Designing, building, and
operating microservices on Azure.
 N-tier architecture style
12/18/2020 • 5 minutes to read • Edit Online

An N-tier architecture divides an application into logical layers and physical tiers .

Remote
Service
Middle
Tier 1

Client WAF Web Cache

Tier
Data
Tier
Messaging Middle
Tier 2

Layers are a way to separate responsibilities and manage dependencies. Each layer has a specific responsibility. A
higher layer can use services in a lower layer, but not the other way around.
Tiers are physically separated, running on separate machines. A tier can call to another tier directly, or use
asynchronous messaging (message queue). Although each layer might be hosted in its own tier, that's not
required. Several layers might be hosted on the same tier. Physically separating the tiers improves scalability and
resiliency, but also adds latency from the additional network communication.
A traditional three-tier application has a presentation tier, a middle tier, and a database tier. The middle tier is
optional. More complex applications can have more than three tiers. The diagram above shows an application with
two middle tiers, encapsulating different areas of functionality.
An N-tier application can have a closed layer architecture or an open layer architecture :
In a closed layer architecture, a layer can only call the next layer immediately down.
In an open layer architecture, a layer can call any of the layers below it.
A closed layer architecture limits the dependencies between layers. However, it might create unnecessary network
traffic, if one layer simply passes requests along to the next layer.

When to use this architecture

N-tier architectures are typically implemented as infrastructure-as-service (IaaS) applications, with each tier
running on a separate set of VMs. However, an N-tier application doesn't need to be pure IaaS. Often, it's
advantageous to use managed services for some parts of the architecture, particularly caching, messaging, and
data storage.
Consider an N-tier architecture for:
Simple web applications.
Migrating an on-premises application to Azure with minimal refactoring.
Unified development of on-premises and cloud applications.
N-tier architectures are very common in traditional on-premises applications, so it's a natural fit for migrating
existing workloads to Azure.

Benefits
Portability between cloud and on-premises, and between cloud platforms.
Less learning curve for most developers.
Natural evolution from the traditional application model.
Open to heterogeneous environment (Windows/Linux)

Challenges
It's easy to end up with a middle tier that just does CRUD operations on the database, adding extra latency
without doing any useful work.
Monolithic design prevents independent deployment of features.
Managing an IaaS application is more work than an application that uses only managed services.
It can be difficult to manage network security in a large system.

Best practices
Use autoscaling to handle changes in load. See Autoscaling best practices.
Use asynchronous messaging to decouple tiers.
Cache semistatic data. See Caching best practices.
Configure the database tier for high availability, using a solution such as SQL Server Always On availability
groups.
Place a web application firewall (WAF) between the front end and the Internet.
Place each tier in its own subnet, and use subnets as a security boundary.
Restrict access to the data tier, by allowing requests only from the middle tier(s).

N-tier architecture on virtual machines

This section describes a recommended N-tier architecture running on VMs.

Each tier consists of two or more VMs, placed in an availability set or virtual machine scale set. Multiple VMs
provide resiliency in case one VM fails. Load balancers are used to distribute requests across the VMs in a tier. A
tier can be scaled horizontally by adding more VMs to the pool.
Each tier is also placed inside its own subnet, meaning their internal IP addresses fall within the same address
range. That makes it easy to apply network security group rules and route tables to individual tiers.
The web and business tiers are stateless. Any VM can handle any request for that tier. The data tier should consist
of a replicated database. For Windows, we recommend SQL Server, using Always On availability groups for high
availability. For Linux, choose a database that supports replication, such as Apache Cassandra.
Network security groups restrict access to each tier. For example, the database tier only allows access from the
business tier.

NOTE
The layer labeled "Business Tier" in our reference diagram is a moniker to the business logic tier. Likewise, we also call the
presentation tier the "Web Tier." In our example, this is a web application, though multi-tier architectures can be used for
other topologies as well (like desktop apps). Name your tiers what works best for your team to communicate the intent of
that logical and/or physical tier in your application - you could even express that naming in resources you choose to
represent that tier (e.g. vmss-appName-business-layer).

For more information about running N-tier applications on Azure:

Run Windows VMs for an N-tier application
Windows N-tier application on Azure with SQL Server
Microsoft Learn module: Tour the N-tier architecture style
Azure Bastion
Additional considerations
N-tier architectures are not restricted to three tiers. For more complex applications, it is common to have
more tiers. In that case, consider using layer-7 routing to route requests to a particular tier.
Tiers are the boundary of scalability, reliability, and security. Consider having separate tiers for services with
different requirements in those areas.
Use virtual machine scale sets for autoscaling.
Look for places in the architecture where you can use a managed service without significant refactoring. In
particular, look at caching, messaging, storage, and databases.
For higher security, place a network DMZ in front of the application. The DMZ includes network virtual
appliances (NVAs) that implement security functionality such as firewalls and packet inspection. For more
information, see Network DMZ reference architecture.
For high availability, place two or more NVAs in an availability set, with an external load balancer to
distribute Internet requests across the instances. For more information, see Deploy highly available network
virtual appliances.
Do not allow direct RDP or SSH access to VMs that are running application code. Instead, operators should
log into a jumpbox, also called a bastion host. This is a VM on the network that administrators use to
connect to the other VMs. The jumpbox has a network security group that allows RDP or SSH only from
approved public IP addresses.
You can extend the Azure virtual network to your on-premises network using a site-to-site virtual private
network (VPN) or Azure ExpressRoute. For more information, see Hybrid network reference architecture.
If your organization uses Active Directory to manage identity, you may want to extend your Active Directory
environment to the Azure VNet. For more information, see Identity management reference architecture.
If you need higher availability than the Azure SLA for VMs provides, replicate the application across two
regions and use Azure Traffic Manager for failover. For more information, see Run Windows VMs in multiple
regions or Run Linux VMs in multiple regions.
 Web-Queue-Worker architecture style
12/18/2020 • 3 minutes to read • Edit Online

The core components of this architecture are a web front end that serves client requests, and a worker that
performs resource-intensive tasks, long-running workflows, or batch jobs. The web front end communicates with
the worker through a message queue .
Remote
Service
Identity
Provider

Web Front Cache

Client End
Worker Database

Queue

CDN Static
Content

Other components that are commonly incorporated into this architecture include:
One or more databases.
A cache to store values from the database for quick reads.
CDN to serve static content
Remote services, such as email or SMS service. Often these are provided by third parties.
Identity provider for authentication.
The web and worker are both stateless. Session state can be stored in a distributed cache. Any long-running work
is done asynchronously by the worker. The worker can be triggered by messages on the queue, or run on a
schedule for batch processing. The worker is an optional component. If there are no long-running operations, the
worker can be omitted.
The front end might consist of a web API. On the client side, the web API can be consumed by a single-page
application that makes AJAX calls, or by a native client application.

When to use this architecture

The Web-Queue-Worker architecture is typically implemented using managed compute services, either Azure App
Service or Azure Cloud Services.
Consider this architecture style for:
Applications with a relatively simple domain.
Applications with some long-running workflows or batch operations.
When you want to use managed services, rather than infrastructure as a service (IaaS).

Benefits
Relatively simple architecture that is easy to understand.
 Easy to deploy and manage.
Clear separation of concerns.
The front end is decoupled from the worker using asynchronous messaging.
The front end and the worker can be scaled independently.

Challenges
Without careful design, the front end and the worker can become large, monolithic components that are
difficult to maintain and update.
There may be hidden dependencies, if the front end and worker share data schemas or code modules.

Best practices
Expose a well-designed API to the client. See API design best practices.
Autoscale to handle changes in load. See Autoscaling best practices.
Cache semi-static data. See Caching best practices.
Use a CDN to host static content. See CDN best practices.
Use polyglot persistence when appropriate. See Use the best data store for the job.
Partition data to improve scalability, reduce contention, and optimize performance. See Data partitioning best
practices.

Web-Queue-Worker on Azure App Service

This section describes a recommended Web-Queue-Worker architecture that uses Azure App Service.

The front end is implemented as an Azure App Service web app, and the worker is implemented as an Azure
Functions app. The web app and the function app are both associated with an App Service plan that
provides the VM instances.
You can use either Azure Service Bus or Azure Storage queues for the message queue. (The diagram shows
an Azure Storage queue.)
Azure Cache for Redis stores session state and other data that needs low latency access.
Azure CDN is used to cache static content such as images, CSS, or HTML.
For storage, choose the storage technologies that best fit the needs of the application. You might use
multiple storage technologies (polyglot persistence). To illustrate this idea, the diagram shows Azure SQL
 Database and Azure Cosmos DB.
For more details, see App Service web application reference architecture.
Additional considerations
Not every transaction has to go through the queue and worker to storage. The web front end can perform
simple read/write operations directly. Workers are designed for resource-intensive tasks or long-running
workflows. In some cases, you might not need a worker at all.
Use the built-in autoscale feature of App Service to scale out the number of VM instances. If the load on the
application follows predictable patterns, use schedule-based autoscale. If the load is unpredictable, use
metrics-based autoscaling rules.
Consider putting the web app and the function app into separate App Service plans. That way, they can be
scaled independently.
Use separate App Service plans for production and testing. Otherwise, if you use the same plan for
production and testing, it means your tests are running on your production VMs.
Use deployment slots to manage deployments. This lets you to deploy an updated version to a staging slot,
then swap over to the new version. It also lets you swap back to the previous version, if there was a problem
with the update.
 Ten design principles for Azure applications
12/18/2020 • 2 minutes to read • Edit Online

Follow these design principles to make your application more scalable, resilient, and manageable.
Design for self healing . In a distributed system, failures happen. Design your application to be self healing when
failures occur.
Make all things redundant . Build redundancy into your application, to avoid having single points of failure.
Minimize coordination . Minimize coordination between application services to achieve scalability.
Design to scale out . Design your application so that it can scale horizontally, adding or removing new instances
as demand requires.
Par tition around limits . Use partitioning to work around database, network, and compute limits.
Design for operations . Design your application so that the operations team has the tools they need.
Use managed ser vices . When possible, use platform as a service (PaaS) rather than infrastructure as a service
(IaaS).
Use the best data store for the job . Pick the storage technology that is the best fit for your data and how it will
be used.
Design for evolution . All successful applications change over time. An evolutionary design is key for continuous
innovation.
Build for the needs of business . Every design decision must be justified by a business requirement.
 Design for self healing
12/18/2020 • 4 minutes to read • Edit Online

Design your application to be self healing when failures occur

In a distributed system, failures can happen. Hardware can fail. The network can have transient failures. Rarely, an
entire service or region may experience a disruption, but even those must be planned for.
Therefore, design an application to be self healing when failures occur. This requires a three-pronged approach:
Detect failures.
Respond to failures gracefully.
Log and monitor failures, to give operational insight.
How you respond to a particular type of failure may depend on your application's availability requirements. For
example, if you require very high availability, you might automatically fail over to a secondary region during a
regional outage. However, that will incur a higher cost than a single-region deployment.
Also, don't just consider big events like regional outages, which are generally rare. You should focus as much, if not
more, on handling local, short-lived failures, such as network connectivity failures or failed database connections.

Recommendations
Retr y failed operations . Transient failures may occur due to momentary loss of network connectivity, a dropped
database connection, or a timeout when a service is busy. Build retry logic into your application to handle transient
failures. For many Azure services, the client SDK implements automatic retries. For more information, see Transient
fault handling and the Retry pattern.
Protect failing remote ser vices (Circuit Breaker) . It's good to retry after a transient failure, but if the failure
persists, you can end up with too many callers hammering a failing service. This can lead to cascading failures, as
requests back up. Use the Circuit Breaker pattern to fail fast (without making the remote call) when an operation is
likely to fail.
Isolate critical resources (Bulkhead) . Failures in one subsystem can sometimes cascade. This can happen if a
failure causes some resources, such as threads or sockets, not to get freed in a timely manner, leading to resource
exhaustion. To avoid this, partition a system into isolated groups, so that a failure in one partition does not bring
down the entire system.
Perform load leveling . Applications may experience sudden spikes in traffic that can overwhelm services on the
backend. To avoid this, use the Queue-Based Load Leveling pattern to queue work items to run asynchronously. The
queue acts as a buffer that smooths out peaks in the load.
Fail over . If an instance can't be reached, fail over to another instance. For things that are stateless, like a web
server, put several instances behind a load balancer or traffic manager. For things that store state, like a database,
use replicas and fail over. Depending on the data store and how it replicates, this may require the application to
deal with eventual consistency.
Compensate failed transactions . In general, avoid distributed transactions, as they require coordination across
services and resources. Instead, compose an operation from smaller individual transactions. If the operation fails
midway through, use Compensating Transactions to undo any step that already completed.
Checkpoint long-running transactions . Checkpoints can provide resiliency if a long-running operation fails.
When the operation restarts (for example, it is picked up by another VM), it can be resumed from the last
checkpoint.
Degrade gracefully . Sometimes you can't work around a problem, but you can provide reduced functionality that
is still useful. Consider an application that shows a catalog of books. If the application can't retrieve the thumbnail
image for the cover, it might show a placeholder image. Entire subsystems might be noncritical for the application.
For example, in an e-commerce site, showing product recommendations is probably less critical than processing
orders.
Throttle clients . Sometimes a small number of users create excessive load, which can reduce your application's
availability for other users. In this situation, throttle the client for a certain period of time. See the Throttling pattern.
Block bad actors . Just because you throttle a client, it doesn't mean client was acting maliciously. It just means the
client exceeded their service quota. But if a client consistently exceeds their quota or otherwise behaves badly, you
might block them. Define an out-of-band process for user to request getting unblocked.
Use leader election . When you need to coordinate a task, use Leader Election to select a coordinator. That way,
the coordinator is not a single point of failure. If the coordinator fails, a new one is selected. Rather than implement
a leader election algorithm from scratch, consider an off-the-shelf solution such as Zookeeper.
Test with fault injection . All too often, the success path is well tested but not the failure path. A system could run
in production for a long time before a failure path is exercised. Use fault injection to test the resiliency of the
system to failures, either by triggering actual failures or by simulating them.
Embrace chaos engineering . Chaos engineering extends the notion of fault injection, by randomly injecting
failures or abnormal conditions into production instances.
For a structured approach to making your applications self healing, see Design reliable applications for Azure.
 Make all things redundant
12/18/2020 • 2 minutes to read • Edit Online

Build redundancy into your application, to avoid having single points of

failure
A resilient application routes around failure. Identify the critical paths in your application. Is there redundancy at
each point in the path? When a subsystem fails, will the application fail over to something else?

Recommendations
Consider business requirements . The amount of redundancy built into a system can affect both cost and
complexity. Your architecture should be informed by your business requirements, such as recovery time objective
(RTO). For example, a multi-region deployment is more expensive than a single-region deployment, and is more
complicated to manage. You will need operational procedures to handle failover and failback. The additional cost
and complexity might be justified for some business scenarios and not others.
Place VMs behind a load balancer . Don't use a single VM for mission-critical workloads. Instead, place multiple
VMs behind a load balancer. If any VM becomes unavailable, the load balancer distributes traffic to the remaining
healthy VMs. To learn how to deploy this configuration, see Multiple VMs for scalability and availability.

Load
Balancer

Replicate databases . Azure SQL Database and Cosmos DB automatically replicate the data within a region, and
you can enable geo-replication across regions. If you are using an IaaS database solution, choose one that supports
replication and failover, such as SQL Server Always On availability groups.
Enable geo-replication . Geo-replication for Azure SQL Database and Cosmos DB creates secondary readable
replicas of your data in one or more secondary regions. In the event of an outage, the database can fail over to the
secondary region for writes.
Par tition for availability . Database partitioning is often used to improve scalability, but it can also improve
availability. If one shard goes down, the other shards can still be reached. A failure in one shard will only disrupt a
subset of the total transactions.
Deploy to more than one region . For the highest availability, deploy the application to more than one region.
That way, in the rare case when a problem affects an entire region, the application can fail over to another region.
The following diagram shows a multi-region application that uses Azure Traffic Manager to handle failover.
 Region 1

Region 2
Azure Traffic
Manager

Synchronize front and backend failover . Use Azure Traffic Manager to fail over the front end. If the front end
becomes unreachable in one region, Traffic Manager will route new requests to the secondary region. Depending
on your database solution, you may need to coordinate failing over the database.
Use automatic failover but manual failback . Use Traffic Manager for automatic failover, but not for automatic
failback. Automatic failback carries a risk that you might switch to the primary region before the region is
completely healthy. Instead, verify that all application subsystems are healthy before manually failing back. Also,
depending on the database, you might need to check data consistency before failing back.
Include redundancy for Traffic Manager . Traffic Manager is a possible failure point. Review the Traffic Manager
SLA, and determine whether using Traffic Manager alone meets your business requirements for high availability. If
not, consider adding another traffic management solution as a failback. If the Azure Traffic Manager service fails,
change your CNAME records in DNS to point to the other traffic management service.
 Minimize coordination
12/18/2020 • 4 minutes to read • Edit Online

Minimize coordination between application services to achieve

scalability
Most cloud applications consist of multiple application services — web front ends, databases, business processes,
reporting and analysis, and so on. To achieve scalability and reliability, each of those services should run on
multiple instances.
What happens when two instances try to perform concurrent operations that affect some shared state? In some
cases, there must be coordination across nodes, for example to preserve ACID guarantees. In this diagram, Node2
is waiting for Node1 to release a database lock:

Update
Orders
Node 1

Update OrderItems

(blocked)
Node 2

Coordination limits the benefits of horizontal scale and creates bottlenecks. In this example, as you scale out the
application and add more instances, you'll see increased lock contention. In the worst case, the front-end instances
will spend most of their time waiting on locks.
"Exactly once" semantics are another frequent source of coordination. For example, an order must be processed
exactly once. Two workers are listening for new orders. Worker1 picks up an order for processing. The application
must ensure that Worker2 doesn't duplicate the work, but also if Worker1 crashes, the order isn't dropped.

Orders Worker 1

order #123 ? Process orders

Create order

Worker 2

You can use a pattern such as Scheduler Agent Supervisor to coordinate between the workers, but in this case a
better approach might be to partition the work. Each worker is assigned a certain range of orders (say, by billing
region). If a worker crashes, a new instance picks up where the previous instance left off, but multiple instances
aren't contending.

Recommendations
Embrace eventual consistency. When data is distributed, it takes coordination to enforce strong consistency
guarantees. For example, suppose an operation updates two databases. Instead of putting it into a single
transaction scope, it's better if the system can accommodate eventual consistency, perhaps by using the
Compensating Transaction pattern to logically roll back after a failure.
Use domain events to synchronize state. A domain event is an event that records when something happens
that has significance within the domain. Interested services can listen for the event, rather than using a global
transaction to coordinate across multiple services. If this approach is used, the system must tolerate eventual
consistency (see previous item).
Consider patterns such as CQRS and event sourcing. These two patterns can help to reduce contention
between read workloads and write workloads.
The CQRS pattern separates read operations from write operations. In some implementations, the read data
is physically separated from the write data.
In the Event Sourcing pattern, state changes are recorded as a series of events to an append-only data store.
Appending an event to the stream is an atomic operation, requiring minimal locking.
These two patterns complement each other. If the write-only store in CQRS uses event sourcing, the read-only
store can listen for the same events to create a readable snapshot of the current state, optimized for queries. Before
adopting CQRS or event sourcing, however, be aware of the challenges of this approach.
Par tition data. Avoid putting all of your data into one data schema that is shared across many application
services. A microservices architecture enforces this principle by making each service responsible for its own data
store. Within a single database, partitioning the data into shards can improve concurrency, because a service
writing to one shard does not affect a service writing to a different shard.
Design idempotent operations. When possible, design operations to be idempotent. That way, they can be
handled using at-least-once semantics. For example, you can put work items on a queue. If a worker crashes in the
middle of an operation, another worker simply picks up the work item.
Use asynchronous parallel processing. If an operation requires multiple steps that are performed
asynchronously (such as remote service calls), you might be able to call them in parallel, and then aggregate the
results. This approach assumes that each step does not depend on the results of the previous step.
Use optimistic concurrency when possible. Pessimistic concurrency control uses database locks to prevent
conflicts. This can cause poor performance and reduce availability. With optimistic concurrency control, each
transaction modifies a copy or snapshot of the data. When the transaction is committed, the database engine
validates the transaction and rejects any transactions that would affect database consistency.
Azure SQL Database and SQL Server support optimistic concurrency through snapshot isolation. Some Azure
storage services support optimistic concurrency through the use of Etags, including Azure Cosmos DB and Azure
Storage.
Consider MapReduce or other parallel, distributed algorithms. Depending on the data and type of work to
be performed, you may be able to split the work into independent tasks that can be performed by multiple nodes
working in parallel. See Big compute architecture style.
Use leader election for coordination. In cases where you need to coordinate operations, make sure the
coordinator does not become a single point of failure in the application. Using the Leader Election pattern, one
instance is the leader at any time, and acts as the coordinator. If the leader fails, a new instance is elected to be the
leader.
 Design to scale out
12/18/2020 • 2 minutes to read • Edit Online

Design your application so that it can scale horizontally

A primary advantage of the cloud is elastic scaling — the ability to use as much capacity as you need, scaling out as
load increases, and scaling in when the extra capacity is not needed. Design your application so that it can scale
horizontally, adding or removing new instances as demand requires.

Recommendations
Avoid instance stickiness . Stickiness, or session affinity, is when requests from the same client are always
routed to the same server. Stickiness limits the application's ability to scale out. For example, traffic from a high-
volume user will not be distributed across instances. Causes of stickiness include storing session state in memory,
and using machine-specific keys for encryption. Make sure that any instance can handle any request.
Identify bottlenecks . Scaling out isn't a magic fix for every performance issue. For example, if your backend
database is the bottleneck, it won't help to add more web servers. Identify and resolve the bottlenecks in the
system first, before throwing more instances at the problem. Stateful parts of the system are the most likely cause
of bottlenecks.
Decompose workloads by scalability requirements. Applications often consist of multiple workloads, with
different requirements for scaling. For example, an application might have a public-facing site and a separate
administration site. The public site may experience sudden surges in traffic, while the administration site has a
smaller, more predictable load.
Offload resource-intensive tasks. Tasks that require a lot of CPU or I/O resources should be moved to
background jobs when possible, to minimize the load on the front end that is handling user requests.
Use built-in autoscaling features . Many Azure compute services have built-in support for autoscaling. If the
application has a predictable, regular workload, scale out on a schedule. For example, scale out during business
hours. Otherwise, if the workload is not predictable, use performance metrics such as CPU or request queue length
to trigger autoscaling. For autoscaling best practices, see Autoscaling.
Consider aggressive autoscaling for critical workloads . For critical workloads, you want to keep ahead of
demand. It's better to add new instances quickly under heavy load to handle the additional traffic, and then
gradually scale back.
Design for scale in . Remember that with elastic scale, the application will have periods of scale in, when
instances get removed. The application must gracefully handle instances being removed. Here are some ways to
handle scalein:
Listen for shutdown events (when available) and shut down cleanly.
Clients/consumers of a service should support transient fault handling and retry.
For long-running tasks, consider breaking up the work, using checkpoints or the Pipes and Filters pattern.
Put work items on a queue so that another instance can pick up the work, if an instance is removed in the
middle of processing.
 Partition around limits
12/18/2020 • 2 minutes to read • Edit Online

Use partitioning to work around database, network, and compute limits

In the cloud, all services have limits in their ability to scale up. Azure service limits are documented in Azure
subscription and service limits, quotas, and constraints. Limits include number of cores, database size, query
throughput, and network throughput. If your system grows sufficiently large, you may hit one or more of these
limits. Use partitioning to work around these limits.
There are many ways to partition a system, such as:
Partition a database to avoid limits on database size, data I/O, or number of concurrent sessions.
Partition a queue or message bus to avoid limits on the number of requests or the number of concurrent
connections.
Partition an App Service web app to avoid limits on the number of instances per App Service plan.
A database can be partitioned horizontally, vertically, or functionally.
In horizontal partitioning, also called sharding, each partition holds data for a subset of the total data set.
The partitions share the same data schema. For example, customers whose names start with A–M go into
one partition, N–Z into another partition.
In vertical partitioning, each partition holds a subset of the fields for the items in the data store. For
example, put frequently accessed fields in one partition, and less frequently accessed fields in another.
In functional partitioning, data is partitioned according to how it is used by each bounded context in the
system. For example, store invoice data in one partition and product inventory data in another. The schemas
are independent.
For more detailed guidance, see Data partitioning.

Recommendations
Par tition different par ts of the application . Databases are one obvious candidate for partitioning, but also
consider storage, cache, queues, and compute instances.
Design the par tition key to avoid hotspots . If you partition a database, but one shard still gets the majority of
the requests, then you haven't solved your problem. Ideally, load gets distributed evenly across all the partitions.
For example, hash by customer ID and not the first letter of the customer name, because some letters are more
frequent. The same principle applies when partitioning a message queue. Pick a partition key that leads to an even
distribution of messages across the set of queues. For more information, see Sharding.
Par tition around Azure subscription and ser vice limits . Individual components and services have limits, but
there are also limits for subscriptions and resource groups. For very large applications, you might need to partition
around those limits.
Par tition at different levels . Consider a database server deployed on a VM. The VM has a VHD that is backed by
Azure Storage. The storage account belongs to an Azure subscription. Notice that each step in the hierarchy has
limits. The database server may have a connection pool limit. VMs have CPU and network limits. Storage has IOPS
limits. The subscription has limits on the number of VM cores. Generally, it's easier to partition lower in the
hierarchy. Only large applications should need to partition at the subscription level.
 Design for operations
12/18/2020 • 2 minutes to read • Edit Online

Design an application so that the operations team has the tools they
need
The cloud has dramatically changed the role of the operations team. They are no longer responsible for managing
the hardware and infrastructure that hosts the application. That said, operations is still a critical part of running a
successful cloud application. Some of the important functions of the operations team include:
Deployment
Monitoring
Escalation
Incident response
Security auditing
Robust logging and tracing are particularly important in cloud applications. Involve the operations team in design
and planning, to ensure the application gives them the data and insight they need to be successful.

Recommendations
Make all things obser vable . Once a solution is deployed and running, logs and traces are your primary insight
into the system. Tracing records a path through the system, and is useful to pinpoint bottlenecks, performance
issues, and failure points. Logging captures individual events such as application state changes, errors, and
exceptions. Log in production, or else you lose insight at the very times when you need it the most.
Instrument for monitoring . Monitoring gives insight into how well (or poorly) an application is performing, in
terms of availability, performance, and system health. For example, monitoring tells you whether you are meeting
your SLA. Monitoring happens during the normal operation of the system. It should be as close to real-time as
possible, so that the operations staff can react to issues quickly. Ideally, monitoring can help avert problems before
they lead to a critical failure. For more information, see Monitoring and diagnostics.
Instrument for root cause analysis . Root cause analysis is the process of finding the underlying cause of
failures. It occurs after a failure has already happened.
Use distributed tracing . Use a distributed tracing system that is designed for concurrency, asynchrony, and cloud
scale. Traces should include a correlation ID that flows across service boundaries. A single operation may involve
calls to multiple application services. If an operation fails, the correlation ID helps to pinpoint the cause of the
failure.
Standardize logs and metrics . The operations team will need to aggregate logs from across the various services
in your solution. If every service uses its own logging format, it becomes difficult or impossible to get useful
information from them. Define a common schema that includes fields such as correlation ID, event name, IP
address of the sender, and so forth. Individual services can derive custom schemas that inherit the base schema,
and contain additional fields.
Automate management tasks , including provisioning, deployment, and monitoring. Automating a task makes it
repeatable and less prone to human errors.
Treat configuration as code . Check configuration files into a version control system, so that you can track and
version your changes, and roll back if needed.
 Use platform as a service (PaaS) options
12/18/2020 • 2 minutes to read • Edit Online

When possible, use platform as a service (PaaS) rather than

infrastructure as a service (IaaS)
IaaS is like having a box of parts. You can build anything, but you have to assemble it yourself. PaaS options are
easier to configure and administer. You don't need to provision VMs, set up VNets, manage patches and updates,
and all of the other overhead associated with running software on a VM.
For example, suppose your application needs a message queue. You could set up your own messaging service on a
VM, using something like RabbitMQ. But Azure Service Bus already provides reliable messaging as service, and it's
simpler to set up. Just create a Service Bus namespace (which can be done as part of a deployment script) and then
call Service Bus using the client SDK.
Of course, your application may have specific requirements that make an IaaS approach more suitable. However,
even if your application is based on IaaS, look for places where it may be natural to incorporate PaaS options.
These include cache, queues, and data storage.

IN ST EA D O F RUN N IN G. . . C O N SIDER USIN G. . .

Active Directory Azure Active Directory

Elasticsearch Azure Search

Hadoop HDInsight

IIS App Service

MongoDB Cosmos DB

Redis Azure Cache for Redis

SQL Server Azure SQL Database

File share Azure NetApp Files

Please note that this is not meant to be an exhaustive list, but a subset of equivalent options.
 Use the best data store for the job
12/18/2020 • 2 minutes to read • Edit Online

Pick the storage technology that is the best fit for your data and how it
will be used
Gone are the days when you would just stick all of your data into a big relational SQL database. Relational
databases are very good at what they do — providing ACID guarantees for transactions over relational data. But
they come with some costs:
Queries may require expensive joins.
Data must be normalized and conform to a predefined schema (schema on write).
Lock contention may impact performance.
In any large solution, it's likely that a single data store technology won't fill all your needs. Alternatives to relational
databases include key/value stores, document databases, search engine databases, time series databases, column
family databases, and graph databases. Each has pros and cons, and different types of data fit more naturally into
one or another.
For example, you might store a product catalog in a document database, such as Cosmos DB, which allows for a
flexible schema. In that case, each product description is a self-contained document. For queries over the entire
catalog, you might index the catalog and store the index in Azure Search. Product inventory might go into a SQL
database, because that data requires ACID guarantees.
Remember that data includes more than just the persisted application data. It also includes application logs, events,
messages, and caches.

Recommendations
Don't use a relational database for ever ything . Consider other data stores when appropriate. See Choose the
right data store.
Embrace polyglot persistence . In any large solution, it's likely that a single data store technology won't fill all
your needs.
Consider the type of data . For example, put transactional data into SQL, put JSON documents into a document
database, put telemetry data into a time series data base, put application logs in Elasticsearch, and put blobs in
Azure Blob Storage.
Prefer availability over (strong) consistency . The CAP theorem implies that a distributed system must make
trade-offs between availability and consistency. (Network partitions, the other leg of the CAP theorem, can never
be completely avoided.) Often, you can achieve higher availability by adopting an eventual consistency model.
Consider the skillset of the development team . There are advantages to using polyglot persistence, but it's
possible to go overboard. Adopting a new data storage technology requires a new set of skills. The development
team must understand how to get the most out of the technology. They must understand appropriate usage
patterns, how to optimize queries, tune for performance, and so on. Factor this in when considering storage
technologies.
Use compensating transactions . A side effect of polyglot persistence is that single transaction might write data
to multiple stores. If something fails, use compensating transactions to undo any steps that already completed.
Look at bounded contexts . Bounded context is a term from domain driven design. A bounded context is an
explicit boundary around a domain model, and defines which parts of the domain the model applies to. Ideally, a
bounded context maps to a subdomain of the business domain. The bounded contexts in your system are a natural
place to consider polyglot persistence. For example, "products" may appear in both the Product Catalog
subdomain and the Product Inventory subdomain, but it's very likely that these two subdomains have different
requirements for storing, updating, and querying products.
 Design for evolution
12/18/2020 • 3 minutes to read • Edit Online

An evolutionary design is key for continuous innovation

All successful applications change over time, whether to fix bugs, add new features, bring in new technologies, or
make existing systems more scalable and resilient. If all the parts of an application are tightly coupled, it becomes
very hard to introduce changes into the system. A change in one part of the application may break another part, or
cause changes to ripple through the entire codebase.
This problem is not limited to monolithic applications. An application can be decomposed into services, but still
exhibit the sort of tight coupling that leaves the system rigid and brittle. But when services are designed to evolve,
teams can innovate and continuously deliver new features.
Microservices are becoming a popular way to achieve an evolutionary design, because they address many of the
considerations listed here.

Recommendations
Enforce high cohesion and loose coupling . A service is cohesive if it provides functionality that logically
belongs together. Services are loosely coupled if you can change one service without changing the other. High
cohesion generally means that changes in one function will require changes in other related functions. If you find
that updating a service requires coordinated updates to other services, it may be a sign that your services are not
cohesive. One of the goals of domain-driven design (DDD) is to identify those boundaries.
Encapsulate domain knowledge . When a client consumes a service, the responsibility for enforcing the
business rules of the domain should not fall on the client. Instead, the service should encapsulate all of the domain
knowledge that falls under its responsibility. Otherwise, every client has to enforce the business rules, and you end
up with domain knowledge spread across different parts of the application.
Use asynchronous messaging . Asynchronous messaging is a way to decouple the message producer from the
consumer. The producer does not depend on the consumer responding to the message or taking any particular
action. With a pub/sub architecture, the producer may not even know who is consuming the message. New
services can easily consume the messages without any modifications to the producer.
Don't build domain knowledge into a gateway . Gateways can be useful in a microservices architecture, for
things like request routing, protocol translation, load balancing, or authentication. However, the gateway should be
restricted to this sort of infrastructure functionality. It should not implement any domain knowledge, to avoid
becoming a heavy dependency.
Expose open interfaces . Avoid creating custom translation layers that sit between services. Instead, a service
should expose an API with a well-defined API contract. The API should be versioned, so that you can evolve the API
while maintaining backward compatibility. That way, you can update a service without coordinating updates to all
of the upstream services that depend on it. Public facing services should expose a RESTful API over HTTP. Backend
services might use an RPC-style messaging protocol for performance reasons.
Design and test against ser vice contracts . When services expose well-defined APIs, you can develop and test
against those APIs. That way, you can develop and test an individual service without spinning up all of its
dependent services. (Of course, you would still perform integration and load testing against the real services.)
Abstract infrastructure away from domain logic . Don't let domain logic get mixed up with infrastructure-
related functionality, such as messaging or persistence. Otherwise, changes in the domain logic will require
updates to the infrastructure layers and vice versa.
Offload cross-cutting concerns to a separate ser vice . For example, if several services need to authenticate
requests, you could move this functionality into its own service. Then you could evolve the authentication service
— for example, by adding a new authentication flow — without touching any of the services that use it.
Deploy ser vices independently . When the DevOps team can deploy a single service independently of other
services in the application, updates can happen more quickly and safely. Bug fixes and new features can be rolled
out at a more regular cadence. Design both the application and the release process to support independent
updates.
 Build for the needs of the business
12/18/2020 • 2 minutes to read • Edit Online

Every design decision must be justified by a business requirement

This design principle may seem obvious, but it's crucial to keep in mind when designing a solution. Do you
anticipate millions of users, or a few thousand? Is a one-hour application outage acceptable? Do you expect large
bursts in traffic or a predictable workload? Ultimately, every design decision must be justified by a business
requirement.

Recommendations
Define business objectives , including the recovery time objective (RTO), recovery point objective (RPO), and
maximum tolerable outage (MTO). These numbers should inform decisions about the architecture. For example, to
achieve a low RTO, you might implement automated failover to a secondary region. But if your solution can tolerate
a higher RTO, that degree of redundancy might be unnecessary.
Document ser vice level agreements (SL A) and ser vice level objectives (SLO) , including availability and
performance metrics. You might build a solution that delivers 99.95% availability. Is that enough? The answer is a
business decision.
Model the application around the business domain . Start by analyzing the business requirements. Use these
requirements to model the application. Consider using a domain-driven design (DDD) approach to create domain
models that reflect the business processes and use cases.
Capture both functional and nonfunctional requirements . Functional requirements let you judge whether
the application does the right thing. Nonfunctional requirements let you judge whether the application does those
things well. In particular, make sure that you understand your requirements for scalability, availability, and latency.
These requirements will influence design decisions and choice of technology.
Decompose by workload . The term "workload" in this context means a discrete capability or computing task,
which can be logically separated from other tasks. Different workloads may have different requirements for
availability, scalability, data consistency, and disaster recovery.
Plan for growth . A solution might meet your current needs, in terms of number of users, volume of transactions,
data storage, and so forth. However, a robust application can handle growth without major architectural changes.
See Design to scale out and Partition around limits. Also consider that your business model and business
requirements will likely change over time. If an application's service model and data models are too rigid, it
becomes hard to evolve the application for new use cases and scenarios. See Design for evolution.
Manage costs . In a traditional on-premises application, you pay upfront for hardware as a capital expenditure. In a
cloud application, you pay for the resources that you consume. Make sure that you understand the pricing model
for the services that you consume. The total cost will include network bandwidth usage, storage, IP addresses,
service consumption, and other factors. For more information, see Azure pricing. Also consider your operations
costs. In the cloud, you don't have to manage the hardware or other infrastructure, but you still need to manage
your applications, including DevOps, incident response, disaster recovery, and so forth.
 Choose an Azure compute service for your
application
12/18/2020 • 6 minutes to read • Edit Online

Azure offers a number of ways to host your application code. The term compute refers to the hosting model for
the computing resources that your application runs on. The following flowchart will help you to choose a compute
service for your application.
If your application consists of multiple workloads, evaluate each workload separately. A complete solution may
incorporate two or more compute services.

Choose a candidate service

Use the following flowchart to select a candidate compute service.

Definitions:
"Lift and shift" is a strategy for migrating a workload to the cloud without redesigning the application or
making code changes. Also called rehosting. For more information, see Azure migration center.
Cloud optimized is a strategy for migrating to the cloud by refactoring an application to take advantage of
cloud-native features and capabilities.
The output from this flowchart is a star ting point for consideration. Next, perform a more detailed evaluation of
the service to see if it meets your needs.
This article includes several tables which may help you to make these tradeoff decisions. Based on this analysis,
you may find that the initial candidate isn't suitable for your particular application or workload. In that case, expand
your analysis to include other compute services.

Understand the basic features

If you're not familiar with the Azure service selected in the previous step, read the overview documentation to
understand the basics of the service.
App Service. A managed service for hosting web apps, mobile app back ends, RESTful APIs, or automated
business processes.
Azure Kubernetes Service (AKS). A managed Kubernetes service for running containerized applications.
Batch. A managed service for running large-scale parallel and high-performance computing (HPC) applications
Container Instances. The fastest and simplest way to run a container in Azure, without having to provision any
virtual machines and without having to adopt a higher-level service.
Functions. A managed FaaS service.
Service Fabric. A distributed systems platform that can run in many environments, including Azure or on
premises.
Virtual machines. Deploy and manage VMs inside an Azure virtual network.

Understand the hosting models

Cloud services, including Azure services, generally fall into three categories: IaaS, PaaS, or FaaS. (There is also SaaS,
software-as-a-service, which is out of scope for this article.) It's useful to understand the differences.
Infrastructure-as-a-Ser vice (IaaS) lets you provision individual VMs along with the associated networking and
storage components. Then you deploy whatever software and applications you want onto those VMs. This model is
the closest to a traditional on-premises environment, except that Microsoft manages the infrastructure. You still
manage the individual VMs.
Platform-as-a-Ser vice (PaaS) provides a managed hosting environment, where you can deploy your application
without needing to manage VMs or networking resources. Azure App Service is a PaaS service.
Functions-as-a-Ser vice (FaaS) goes even further in removing the need to worry about the hosting environment.
In a FaaS model, you simply deploy your code and the service automatically runs it. Azure Functions is a FaaS
service.

NOTE
Azure Functions is an Azure serverless compute offering. You may read Choose the right integration and automation
services in Azure to know how this service compares with other Azure serverless offerings, such as Logic Apps which
provides serverless workflows.

There is a spectrum from IaaS to pure PaaS. For example, Azure VMs can autoscale by using virtual machine scale
sets. This automatic scaling capability isn't strictly PaaS, but it's the type of management feature found in PaaS
services.
In general, there is a tradeoff between control and ease of management. IaaS gives the most control, flexibility, and
portability, but you have to provision, configure and manage the VMs and network components you create. FaaS
services automatically manage nearly all aspects of running an application. PaaS services fall somewhere in
between.

A Z URE
VIRT UA L APP SERVIC E A Z URE K UB ERN ET E C O N TA IN ER A Z URE
C RIT ERIA M A C H IN ES SERVIC E FA B RIC F UN C T IO N S S SERVIC E IN STA N C ES B ATC H

Application Agnostic Applications Services, Functions Containers Containers Scheduled

compositio , containers guest jobs
n executables,
containers

Density Agnostic Multiple Multiple Serverless 1 Multiple No Multiple

apps per services per containers dedicated apps per
instance via VM per node instances VM
app service
plans

Minimum 12 1 53 Serverless 1 33 No 14
number of dedicated
nodes nodes

State Stateless or Stateless Stateless or Stateless Stateless or Stateless Stateless

manageme Stateful stateful Stateful
nt

Web Agnostic Built in Agnostic Not Agnostic Agnostic No

hosting applicable
 A Z URE
VIRT UA L APP SERVIC E A Z URE K UB ERN ET E C O N TA IN ER A Z URE
C RIT ERIA M A C H IN ES SERVIC E FA B RIC F UN C T IO N S S SERVIC E IN STA N C ES B ATC H

Can be Supported Supported5 Supported Supported Supported Supported Supported

deployed to 5
dedicated
VNet?

Hybrid Supported Supported Supported Supported Supported Not Supported

connectivity 6 7 supported

Notes
1. If using Consumption plan. If using App Service plan, functions run on the VMs allocated for your App Service
plan. See Choose the correct service plan for Azure Functions.
2. Higher SLA with two or more instances.
3. Recommended for production environments.
4. Can scale down to zero after job completes.
5. Requires App Service Environment (ASE).
6. Use Azure App Service Hybrid Connections.
7. Requires App Service plan or Azure Functions Premium plan.

DevOps
A Z URE
VIRT UA L APP SERVIC E A Z URE K UB ERN ET E C O N TA IN ER A Z URE
C RIT ERIA M A C H IN ES SERVIC E FA B RIC F UN C T IO N S S SERVIC E IN STA N C ES B ATC H

Local Agnostic IIS Express, Local node Visual Minikube, Local Not
debugging others 1 cluster Studio or others container supported
Azure runtime
Functions
CLI

Programmi Agnostic Web and Guest Functions Agnostic Agnostic Command

ng model API executable, with line
applications Service triggers application
, WebJobs model,
for Actor
background model,
tasks Containers

Application No built-in Deploymen Rolling Deploymen Rolling Not

update support t slots upgrade t slots update applicable
(per service)

Notes
1. Options include IIS Express for ASP.NET or node.js (iisnode); PHP web server; Azure Toolkit for IntelliJ, Azure
Toolkit for Eclipse. App Service also supports remote debugging of deployed web app.
2. See Resource Manager providers, regions, API versions and schemas.

Scalability
 A Z URE
VIRT UA L APP SERVIC E A Z URE K UB ERN ET E C O N TA IN ER A Z URE
C RIT ERIA M A C H IN ES SERVIC E FA B RIC F UN C T IO N S S SERVIC E IN STA N C ES B ATC H

Autoscaling Virtual Built-in Virtual Built-in Pod auto- Not N/A

machine service machine service scaling1 , supported
scale sets scale sets cluster
auto-
scaling2

Load Azure Load Integrated Azure Load Integrated Azure Load No built-in Azure Load
balancer Balancer Balancer Balancer or support Balancer
Application
Gateway

Scale limit3 Platform 20 100 nodes 200 100 nodes 20 20 core

image: instances, per scale instances per cluster container limit
1000 nodes 100 with set per (default groups per (default
per scale App Service Function limit) subscription limit).
set, Custom Environmen app (default
image: 600 t limit).
nodes per
scale set

Notes
1. See Autoscale pods.
2. See Automatically scale a cluster to meet application demands on Azure Kubernetes Service (AKS).
3. See Azure subscription and service limits, quotas, and constraints.

Availability
A Z URE
VIRT UA L APP SERVIC E A Z URE K UB ERN ET E C O N TA IN ER A Z URE
C RIT ERIA M A C H IN ES SERVIC E FA B RIC F UN C T IO N S S SERVIC E IN STA N C ES B ATC H

SLA SLA for SLA for App SLA for SLA for SLA for AKS SLA for SLA for
Virtual Service Service Functions Container Azure Batch
Machines Fabric Instances

Multi Traffic Traffic Traffic Azure Front Traffic Not Not

region manager manager manager, Door manager supported Supported
failover Multi-
Region
Cluster

For guided learning on Service Guarantees, review Core Cloud Services - Azure architecture and service
guarantees.

Security
Review and understand the available security controls and visibility for each service
App Service
Azure Kubernetes Service
Batch
Container Instances
Functions
 Service Fabric
Virtual machine - Windows
Virtual machine - LINUX

Other criteria
A Z URE
VIRT UA L APP SERVIC E A Z URE K UB ERN ET E C O N TA IN ER A Z URE
C RIT ERIA M A C H IN ES SERVIC E FA B RIC F UN C T IO N S S SERVIC E IN STA N C ES B ATC H

SSL Configured Supported Supported Supported Ingress Use sidecar Supported

in VM controller container

Cost Windows, App Service Service Azure AKS pricing Container Azure Batch
Linux pricing Fabric Functions Instances pricing
pricing pricing pricing

Suitable N-Tier, Big Web- Microservic Microservic Microservic Microservic Big

architecture compute Queue- es, Event- es, Event- es, Event- es, task compute
styles (HPC) Worker, N- driven driven driven automation, (HPC)
Tier architecture architecture architecture batch jobs

The output from this flowchart is a star ting point for consideration. Next, perform a more detailed evaluation of
the service to see if it meets your needs.

Consider limits and cost

Perform a more detailed evaluation looking at the following aspects of the service:
Service limits
Cost
SLA
Regional availability
Compute comparison tables

Next steps
Core Cloud Services - Azure compute options. This Microsoft Learn module explores how compute services can
solve common business needs.
 Understand data store models
12/18/2020 • 12 minutes to read • Edit Online

Modern business systems manage increasingly large volumes of heterogeneous data. This heterogeneity means
that a single data store is usually not the best approach. Instead, it's often better to store different types of data in
different data stores, each focused toward a specific workload or usage pattern. The term polyglot persistence is
used to describe solutions that use a mix of data store technologies. Therefore, it's important to understand the
main storage models and their tradeoffs.
Selecting the right data store for your requirements is a key design decision. There are literally hundreds of
implementations to choose from among SQL and NoSQL databases. Data stores are often categorized by how they
structure data and the types of operations they support. This article describes several of the most common storage
models. Note that a particular data store technology may support multiple storage models. For example, a
relational database management systems (RDBMS) may also support key/value or graph storage. In fact, there is a
general trend for so-called multi-model support, where a single database system supports several models. But it's
still useful to understand the different models at a high level.
Not all data stores in a given category provide the same feature-set. Most data stores provide server-side
functionality to query and process data. Sometimes this functionality is built into the data storage engine. In other
cases, the data storage and processing capabilities are separated, and there may be several options for processing
and analysis. Data stores also support different programmatic and management interfaces.
Generally, you should start by considering which storage model is best suited for your requirements. Then consider
a particular data store within that category, based on factors such as feature set, cost, and ease of management.

Relational database management systems

Relational databases organize data as a series of two-dimensional tables with rows and columns. Most vendors
provide a dialect of the Structured Query Language (SQL) for retrieving and managing data. An RDBMS typically
implements a transactionally consistent mechanism that conforms to the ACID (Atomic, Consistent, Isolated,
Durable) model for updating information.
An RDBMS typically supports a schema-on-write model, where the data structure is defined ahead of time, and all
read or write operations must use the schema.
This model is very useful when strong consistency guarantees are important — where all changes are atomic, and
transactions always leave the data in a consistent state. However, an RDBMS generally can't scale out horizontally
without sharding the data in some way. Also, the data in an RDBMS must normalized, which isn't appropriate for
every data set.
Azure services
Azure SQL Database | (Security Baseline)
Azure Database for MySQL | (Security Baseline)
Azure Database for PostgreSQL | (Security Baseline)
Azure Database for MariaDB | (Security Baseline)
Workload
Records are frequently created and updated.
Multiple operations have to be completed in a single transaction.
Relationships are enforced using database constraints.
Indexes are used to optimize query performance.
Data type
Data is highly normalized.
Database schemas are required and enforced.
Many-to-many relationships between data entities in the database.
Constraints are defined in the schema and imposed on any data in the database.
Data requires high integrity. Indexes and relationships need to be maintained accurately.
Data requires strong consistency. Transactions operate in a way that ensures all data are 100% consistent for all
users and processes.
Size of individual data entries is small to medium-sized.
Examples
Inventory management
Order management
Reporting database
Accounting

Key/value stores
A key/value store associates each data value with a unique key. Most key/value stores only support simple query,
insert, and delete operations. To modify a value (either partially or completely), an application must overwrite the
existing data for the entire value. In most implementations, reading or writing a single value is an atomic operation.
An application can store arbitrary data as a set of values. Any schema information must be provided by the
application. The key/value store simply retrieves or stores the value by key.

Key/value stores are highly optimized for applications performing simple lookups, but are less suitable if you need
to query data across different key/value stores. Key/value stores are also not optimized for querying by value.
A single key/value store can be extremely scalable, as the data store can easily distribute data across multiple nodes
on separate machines.
Azure services
Azure Cosmos DB Table API, etcd API (preview), and SQL API | (Cosmos DB Security Baseline)
Azure Cache for Redis | (Security Baseline)
Azure Table Storage | (Security Baseline)
Workload
Data is accessed using a single key, like a dictionary.
No joins, lock, or unions are required.
No aggregation mechanisms are used.
Secondary indexes are generally not used.
Data type
Each key is associated with a single value.
 There is no schema enforcement.
No relationships between entities.
Examples
Data caching
Session management
User preference and profile management
Product recommendation and ad serving

Document databases
A document database stores a collection of documents, where each document consists of named fields and data.
The data can be simple values or complex elements such as lists and child collections. Documents are retrieved by
unique keys.
Typically, a document contains the data for single entity, such as a customer or an order. A document may contain
information that would be spread across several relational tables in an RDBMS. Documents don't need to have the
same structure. Applications can store different data in documents as business requirements change.

Azure service
Azure Cosmos DB SQL API | (Cosmos DB Security Baseline)
Workload
Insert and update operations are common.
No object-relational impedance mismatch. Documents can better match the object structures used in application
code.
Individual documents are retrieved and written as a single block.
Data requires index on multiple fields.
Data type
Data can be managed in de-normalized way.
Size of individual document data is relatively small.
Each document type can use its own schema.
Documents can include optional fields.
Document data is semi-structured, meaning that data types of each field are not strictly defined.
Examples
Product catalog
Content management
Inventory management

Graph databases
A graph database stores two types of information, nodes and edges. Edges specify relationships between nodes.
Nodes and edges can have properties that provide information about that node or edge, similar to columns in a
table. Edges can also have a direction indicating the nature of the relationship.
Graph databases can efficiently perform queries across the network of nodes and edges and analyze the
relationships between entities. The following diagram shows an organization's personnel database structured as a
graph. The entities are employees and departments, and the edges indicate reporting relationships and the
departments in which employees work.

This structure makes it straightforward to perform queries such as "Find all employees who report directly or
indirectly to Sarah" or "Who works in the same department as John?" For large graphs with lots of entities and
relationships, you can perform very complex analyses very quickly. Many graph databases provide a query
language that you can use to traverse a network of relationships efficiently.
Azure services
Azure Cosmos DB Gremlin API | (Security Baseline)
SQL Server | (Security Baseline)
Workload
Complex relationships between data items involving many hops between related data items.
The relationship between data items are dynamic and change over time.
Relationships between objects are first-class citizens, without requiring foreign-keys and joins to traverse.
Data type
Nodes and relationships.
Nodes are similar to table rows or JSON documents.
Relationships are just as important as nodes, and are exposed directly in the query language.
 Composite objects, such as a person with multiple phone numbers, tend to be broken into separate, smaller
nodes, combined with traversable relationships
Examples
Organization charts
Social graphs
Fraud detection
Recommendation engines

Data analytics
Data analytics stores provide massively parallel solutions for ingesting, storing, and analyzing data. The data is
distributed across multiple servers to maximize scalability. Large data file formats such as delimiter files (CSV),
parquet, and ORC are widely used in data analytics. Historical data is typically stored in data stores such as blob
storage or Azure Data Lake Storage Gen2, which are then accessed by Azure Synapse, Databricks, or HDInsight as
external tables. A typical scenario using data stored as parquet files for performance, is described in the article Use
external tables with Synapse SQL.
Azure services
Azure Synapse Analytics | (Security Baseline)
Azure Data Lake | (Security Baseline)
Azure Data Explorer | (Security Baseline)
Azure Analysis Services
HDInsight | (Security Baseline)
Azure Databricks | (Security Baseline)
Workload
Data analytics
Enterprise BI
Data type
Historical data from multiple sources.
Usually denormalized in a "star" or "snowflake" schema, consisting of fact and dimension tables.
Usually loaded with new data on a scheduled basis.
Dimension tables often include multiple historic versions of an entity, referred to as a slowly changing
dimension.
Examples
Enterprise data warehouse

Column-family databases
A column-family database organizes data into rows and columns. In its simplest form, a column-family database
can appear very similar to a relational database, at least conceptually. The real power of a column-family database
lies in its denormalized approach to structuring sparse data.
You can think of a column-family database as holding tabular data with rows and columns, but the columns are
divided into groups known as column families. Each column family holds a set of columns that are logically related
together and are typically retrieved or manipulated as a unit. Other data that is accessed separately can be stored in
separate column families. Within a column family, new columns can be added dynamically, and rows can be sparse
(that is, a row doesn't need to have a value for every column).
The following diagram shows an example with two column families, Identity and Contact Info . The data for a
single entity has the same row key in each column-family. This structure, where the rows for any given object in a
column family can vary dynamically, is an important benefit of the column-family approach, making this form of
data store highly suited for storing structured, volatile data.

Unlike a key/value store or a document database, most column-family databases store data in key order, rather than
by computing a hash. Many implementations allow you to create indexes over specific columns in a column-family.
Indexes let you retrieve data by columns value, rather than row key.
Read and write operations for a row are usually atomic with a single column-family, although some
implementations provide atomicity across the entire row, spanning multiple column-families.
Azure services
Azure Cosmos DB Cassandra API | (Security Baseline)
HBase in HDInsight | (Security Baseline)
Workload
Most column-family databases perform write operations extremely quickly.
Update and delete operations are rare.
Designed to provide high throughput and low-latency access.
Supports easy query access to a particular set of fields within a much larger record.
Massively scalable.
Data type
Data is stored in tables consisting of a key column and one or more column families.
Specific columns can vary by individual rows.
Individual cells are accessed via get and put commands
Multiple rows are returned using a scan command.
Examples
Recommendations
Personalization
Sensor data
Telemetry
Messaging
Social media analytics
Web analytics
Activity monitoring
Weather and other time-series data

Search Engine Databases

A search engine database allows applications to search for information held in external data stores. A search engine
database can index massive volumes of data and provide near real-time access to these indexes.
Indexes can be multi-dimensional and may support free-text searches across large volumes of text data. Indexing
can be performed using a pull model, triggered by the search engine database, or using a push model, initiated by
external application code.
Searching can be exact or fuzzy. A fuzzy search finds documents that match a set of terms and calculates how
closely they match. Some search engines also support linguistic analysis that can return matches based on
synonyms, genre expansions (for example, matching dogs to pets ), and stemming (matching words with the
same root).
Azure service
Azure Search | (Security Baseline)
Workload
Data indexes from multiple sources and services.
Queries are ad-hoc and can be complex.
Full text search is required.
Ad hoc self-service query is required.
Data type
Semi-structured or unstructured text
Text with reference to structured data
Examples
Product catalogs
Site search
Logging

Time series databases

Time series data is a set of values organized by time. Time series databases typically collect large amounts of data in
real time from a large number of sources. Updates are rare, and deletes are often done as bulk operations. Although
the records written to a time-series database are generally small, there are often a large number of records, and
total data size can grow rapidly.
Azure service
Azure Time Series Insights
Workload
Records are generally appended sequentially in time order.
An overwhelming proportion of operations (95-99%) are writes.
Updates are rare.
Deletes occur in bulk, and are made to contiguous blocks or records.
Data is read sequentially in either ascending or descending time order, often in parallel.
Data type
A timestamp is used as the primary key and sorting mechanism.
Tags may define additional information about the type, origin, and other information about the entry.
Examples
Monitoring and event telemetry.
Sensor or other IoT data.

Object storage
Object storage is optimized for storing and retrieving large binary objects (images, files, video and audio streams,
large application data objects and documents, virtual machine disk images). Large data files are also popularly used
in this model, for example, delimiter file (CSV), parquet, and ORC. Object stores can manage extremely large
amounts of unstructured data.
Azure service
Azure Blob Storage | (Security Baseline)
Azure Data Lake Storage Gen2 | (Security Baseline)
Workload
Identified by key.
Content is typically an asset such as a delimiter, image, or video file.
Content must be durable and external to any application tier.
Data type
Data size is large.
Value is opaque.
Examples
Images, videos, office documents, PDFs
Static HTML, JSON, CSS
Log and audit files
Database backups

Shared files
Sometimes, using simple flat files can be the most effective means of storing and retrieving information. Using file
shares enables files to be accessed across a network. Given appropriate security and concurrent access control
mechanisms, sharing data in this way can enable distributed services to provide highly scalable data access for
performing basic, low-level operations such as simple read and write requests.
Azure service
Azure Files | (Security Baseline)
Workload
Migration from existing apps that interact with the file system.
Requires SMB interface.
Data type
Files in a hierarchical set of folders.
Accessible with standard I/O libraries.
Examples
Legacy files
Shared content accessible among a number of VMs or app instances
Aided with this understanding of different data storage models, the next step is to evaluate your workload and
application, and decide which data store will meet your specific needs. Use the data storage decision tree to help
with this process.
 Select an Azure data store for your application
12/18/2020 • 2 minutes to read • Edit Online

Azure offers a number of managed data storage solutions, each providing different features and capabilities. This
article will help you to choose a data store for your application.
If your application consists of multiple workloads, evaluate each workload separately. A complete solution may
incorporate multiple data stores.
Use the following flowchart to select a candidate data store.
 START

Azure Database MySql

for MySQL Else... Yes
Require compatible Relational data?
format? SQL Database
Azure Database PostreSQL
for PostgreSQL No

MariaDB Semi-structured Yes CosmosDB SQL

Azure Database data, schema-on-
for MariaDB read? API

Cassandra No
CosmosDB
Cassandra API
Yes
Need SMB interface? Azure Files
CosmosDB MongoDB
MongoDB API No
Blob Storage
Yes
Archive? cool access tier
archive access tier

No

Yes
Search index data? Azure Search

No

Yes
Time series data?
Time Series
Insights

No

File system Yes Analyse with

File system or object Large data (TB/PB) Azure Synapse, Databricks,
Data Lake Store store? for analysis? Azure Analytics Services.

Object No

Yes Binary blobs

Blob Storage (images, PDF files,
etc.)?

No

Yes Cosmos DB
Graph data?
Graph API

No
Yes
Azure Cache for
Redis
Transient data?
No
Cosmos DB SQL
API

The output from this flowchart is a star ting point for consideration. Next, perform a more detailed evaluation of
the data store to see if it meets your needs. Refer to Criteria for choosing a data store to aid in this evaluation.
 Criteria for choosing a data store
12/18/2020 • 3 minutes to read • Edit Online

This article describes the comparison criteria you should use when evaluating a data store. The goal is to help you
determine which data storage types can meet your solution's requirements.

General considerations
Keep the following considerations in mind when making your selection.
Functional requirements
Data format . What type of data are you intending to store? Common types include transactional data,
JSON objects, telemetry, search indexes, or flat files.
Data size . How large are the entities you need to store? Will these entities need to be maintained as a
single document, or can they be split across multiple documents, tables, collections, and so forth?
Scale and structure . What is the overall amount of storage capacity you need? Do you anticipate
partitioning your data?
Data relationships . Will your data need to support one-to-many or many-to-many relationships? Are
relationships themselves an important part of the data? Will you need to join or otherwise combine data
from within the same dataset, or from external datasets?
Consistency model . How important is it for updates made in one node to appear in other nodes, before
further changes can be made? Can you accept eventual consistency? Do you need ACID guarantees for
transactions?
Schema flexibility . What kind of schemas will you apply to your data? Will you use a fixed schema, a
schema-on-write approach, or a schema-on-read approach?
Concurrency . What kind of concurrency mechanism do you want to use when updating and synchronizing
data? Will the application perform many updates that could potentially conflict. If so, you may require
record locking and pessimistic concurrency control. Alternatively, can you support optimistic concurrency
controls? If so, is simple timestamp-based concurrency control enough, or do you need the added
functionality of multi-version concurrency control?
Data movement . Will your solution need to perform ETL tasks to move data to other stores or data
warehouses?
Data lifecycle . Is the data write-once, read-many? Can it be moved into cool or cold storage?
Other suppor ted features . Do you need any other specific features, such as schema validation,
aggregation, indexing, full-text search, MapReduce, or other query capabilities?
Non-functional requirements
Performance and scalability . What are your data performance requirements? Do you have specific
requirements for data ingestion rates and data processing rates? What are the acceptable response times
for querying and aggregation of data once ingested? How large will you need the data store to scale up? Is
your workload more read-heavy or write-heavy?
Reliability . What overall SLA do you need to support? What level of fault-tolerance do you need to provide
for data consumers? What kind of backup and restore capabilities do you need?
 Replication . Will your data need to be distributed among multiple replicas or regions? What kind of data
replication capabilities do you require?
Limits . Will the limits of a particular data store support your requirements for scale, number of
connections, and throughput?
Management and cost
Managed ser vice . When possible, use a managed data service, unless you require specific capabilities that
can only be found in an IaaS-hosted data store.
Region availability . For managed services, is the service available in all Azure regions? Does your solution
need to be hosted in certain Azure regions?
Por tability . Will your data need to be migrated to on-premises, external datacenters, or other cloud hosting
environments?
Licensing . Do you have a preference of a proprietary versus OSS license type? Are there any other external
restrictions on what type of license you can use?
Overall cost . What is the overall cost of using the service within your solution? How many instances will
need to run, to support your uptime and throughput requirements? Consider operations costs in this
calculation. One reason to prefer managed services is the reduced operational cost.
Cost effectiveness . Can you partition your data, to store it more cost effectively? For example, can you
move large objects out of an expensive relational database into an object store?
Security
Security . What type of encryption do you require? Do you need encryption at rest? What authentication
mechanism do you want to use to connect to your data?
Auditing . What kind of audit log do you need to generate?
Networking requirements . Do you need to restrict or otherwise manage access to your data from other
network resources? Does data need to be accessible only from inside the Azure environment? Does the data
need to be accessible from specific IP addresses or subnets? Does it need to be accessible from applications
or services hosted on-premises or in other external datacenters?
DevOps
Skill set . Are there particular programming languages, operating systems, or other technology that your
team is particularly adept at using? Are there others that would be difficult for your team to work with?
Clients Is there good client support for your development languages?
 Overview of load-balancing options in Azure
12/18/2020 • 4 minutes to read • Edit Online

The term load balancing refers to the distribution of workloads across multiple computing resources. Load
balancing aims to optimize resource use, maximize throughput, minimize response time, and avoid overloading
any single resource. It can also improve availability by sharing a workload across redundant computing resources.

Overview
Azure load balancing services can be categorized along two dimensions: global versus regional, and HTTP(S) versus
non-HTTP(S).
Global versus regional
Global load-balancing services distribute traffic across regional backends, clouds, or hybrid on-premises
services. These services route end-user traffic to the closest available backend. They also react to changes in
service reliability or performance, in order to maximize availability and performance. You can think of them
as systems that load balance between application stamps, endpoints, or scale-units hosted across different
regions/geographies.
Regional load-balancing services distribute traffic within virtual networks across virtual machines (VMs) or
zonal and zone-redundant service endpoints within a region. You can think of them as systems that load
balance between VMs, containers, or clusters within a region in a virtual network.
HTTP(S ) versus non-HTTP(S )
HTTP(S) load-balancing services are Layer 7 load balancers that only accept HTTP(S) traffic. They are
intended for web applications or other HTTP(S) endpoints. They include features such as SSL offload, web
application firewall, path-based load balancing, and session affinity.
Non-HTTP/S load-balancing services can handle non-HTTP(S) traffic and are recommended for non-web
workloads.
The following table summarizes the Azure load balancing services by these categories:

SERVIC E GLO B A L / REGIO N A L REC O M M EN DED T RA F F IC

Azure Front Door Global HTTP(S)

Traffic Manager Global non-HTTP(S)

Application Gateway Regional HTTP(S)

Azure Load Balancer Global non-HTTP(S)

Azure load balancing services

Here are the main load-balancing services currently available in Azure:
Front Door is an application delivery network that provides global load balancing and site acceleration service for
web applications. It offers Layer 7 capabilities for your application like SSL offload, path-based routing, fast failover,
caching, etc. to improve performance and high-availability of your applications.
 NOTE
At this time, Azure Front Door does not support Web Sockets.

Traffic Manager is a DNS-based traffic load balancer that enables you to distribute traffic optimally to services
across global Azure regions, while providing high availability and responsiveness. Because Traffic Manager is a
DNS-based load-balancing service, it load balances only at the domain level. For that reason, it can't fail over as
quickly as Front Door, because of common challenges around DNS caching and systems not honoring DNS TTLs.
Application Gateway provides application delivery controller (ADC) as a service, offering various Layer 7 load-
balancing capabilities. Use it to optimize web farm productivity by offloading CPU-intensive SSL termination to the
gateway.
Azure Load Balancer is a high-performance, ultra low-latency Layer 4 load-balancing service (inbound and
outbound) for all UDP and TCP protocols. It is built to handle millions of requests per second while ensuring your
solution is highly available. Azure Load Balancer is zone-redundant, ensuring high availability across Availability
Zones.

Decision tree for load balancing in Azure

When selecting the load-balancing options, here are some factors to consider:
Traffic type . Is it a web (HTTP/HTTPS) application? Is it public facing or a private application?
Global versus. regional . Do you need to load balance VMs or containers within a virtual network, or load
balance scale unit/deployments across regions, or both?
Availability . What is the service SLA?
Cost . See Azure pricing. In addition to the cost of the service itself, consider the operations cost for managing a
solution built on that service.
Features and limits . What are the overall limitations of each service? See Service limits.
The following flowchart will help you to choose a load-balancing solution for your application. The flowchart
guides you through a set of key decision criteria to reach a recommendation.
Treat this flowchar t as a star ting point. Every application has unique requirements, so use the
recommendation as a starting point. Then perform a more detailed evaluation.
If your application consists of multiple workloads, evaluate each workload separately. A complete solution may
incorporate two or more load-balancing solutions.
Definitions
Internet facing . Applications that are publicly accessible from the internet. As a best practice, application
owners apply restrictive access policies or protect the application by setting up offerings like web application
firewall and DDoS protection.
Global . End users or clients are located beyond a small geographical area. For example, users across
multiple continents, across countries/regions within a continent, or even across multiple metropolitan areas
within a larger country/region.
PaaS . Platform as a service (PaaS) services provide a managed hosting environment, where you can deploy
your application without needing to manage VMs or networking resources. In this case, PaaS refers to
services that provide integrated load balancing within a region. See Choosing a compute service –
Scalability.
IaaS . Infrastructure as a service (IaaS) is a computing option where you provision the VMs that you need,
along with associated network and storage components. IaaS applications require internal load balancing
within a virtual network, using Azure Load Balancer.
Application-layer processing refers to special routing within a virtual network. For example, path-based
routing within the virtual network across VMs or virtual machine scale sets. For more information, see When
should we deploy an Application Gateway behind Front Door?.
 Asynchronous messaging options in Azure
12/18/2020 • 17 minutes to read • Edit Online

This article describes the different types of messages and the entities that participate in a messaging infrastructure.
Based on the requirements of each message type, the article recommends Azure messaging services. The options
include Azure Service Bus, Event Grid, and Event Hubs.
At an architectural level, a message is a datagram created by an entity (producer), to distribute information so that
other entities (consumers) can be aware and act accordingly. The producer and the consumer can communicate
directly or optionally through an intermediary entity (message broker). This article focuses on asynchronous
messaging using a message broker.

Messages can be classified into two main categories. If the producer expects an action from the consumer, that
message is a command. If the message informs the consumer that an action has taken place, then the message is
an event.

Commands
The producer sends a command with the intent that the consumer(s) will perform an operation within the scope of
a business transaction.
A command is a high-value message and must be delivered at least once. If a command is lost, the entire business
transaction might fail. Also, a command shouldn't be processed more than once. Doing so might cause an
erroneous transaction. A customer might get duplicate orders or billed twice.
Commands are often used to manage the workflow of a multistep business transaction. Depending on the
business logic, the producer may expect the consumer to acknowledge the message and report the results of the
operation. Based on that result, the producer may choose an appropriate course of action.

Events
An event is a type of message that a producer raises to announce facts.
The producer (known as the publisher in this context) has no expectations that the events will result in any action.
Interested consumer(s), can subscribe, listen for events, and take actions depending on their consumption scenario.
Events can have multiple subscribers or no subscribers at all. Two different subscribers can react to an event with
different actions and not be aware of one another.
The producer and consumer are loosely coupled and managed independently. The consumer isn't expected to
acknowledge the event back to the producer. A consumer that is no longer interested in the events, can
unsubscribe. The consumer is removed from the pipeline without affecting the producer or the overall functionality
of the system.
There are two categories of events:
 The producer raises events to announce discrete facts. A common use case is event notification. For
example, Azure Resource Manager raises events when it creates, modifies, or deletes resources. A subscriber
of those events could be a Logic App that sends alert emails.
The producer raises related events in a sequence, or a stream of events, over a period of time. Typically, a
stream is consumed for statistical evaluation. The evaluation can be done within a temporal window or as
events arrive. Telemetry is a common use case, for example, health and load monitoring of a system.
Another case is event streaming from IoT devices.
A common pattern for implementing event messaging is the Publisher-Subscriber pattern.

Role and benefits of a message broker

An intermediate message broker provides the functionality of moving messages from producer to consumer and
can offer additional benefits.
Decoupling
A message broker decouples the producer from the consumer in the logic that generates and uses the messages,
respectively. In a complex workflow, the broker can encourage business operations to be decoupled and help
coordinate the workflow.
For example, a single business transaction requires distinct operations that are performed in a business logic
sequence. The producer issues a command that signals a consumer to start an operation. The consumer
acknowledges the message in a separate queue reserved for lining up responses for the producer. Only after
receiving the response, the producer sends a new message to start the next operation in the sequence. A different
consumer processes that message and sends a completion message to the response queue. By using messaging,
the services coordinate the workflow of the transaction among themselves.
A message broker provides temporal decoupling. The producer and consumer don't have to run concurrently. A
producer can send a message to the message broker regardless of the availability of the consumer. Conversely, the
consumer isn't restricted by the producer's availability.
For example, the user interface of a web app generates messages and uses a queue as the message broker. When
ready, consumers can retrieve messages from the queue and perform the work. Temporal decoupling helps the
user interface to remain responsive. It's not blocked while the messages are handled asynchronously.
Certain operations can take long to complete. After issuing a command, the producer shouldn't have to wait until
the consumer completes it. A message broker helps asynchronous processing of messages.
Load balancing
Producers may post a large number of messages that are serviced by many consumers. Use a message broker to
distribute processing across servers and improve throughput. Consumers can run on different servers to spread
the load. Consumers can be added dynamically to scale out the system when needed or removed otherwise.

The Competing Consumers Pattern explains how to process multiple messages concurrently to optimize
throughput, to improve scalability and availability, and to balance the workload.
Load leveling
The volume of messages generated by the producer or a group of producers can be variable. At times there might
be a large volume causing spikes in messages. Instead of adding consumers to handle this work, a message broker
can act as a buffer, and consumers gradually drain messages at their own pace without stressing the system.

The Queue-based Load Leveling Pattern provides more information.

Reliable messaging
A message broker helps ensure that messages aren't lost even if communication fails between the producer and
consumer. The producer can post messages to the message broker and the consumer can retrieve them when
communication is reestablished. The producer isn't blocked unless it loses connectivity with the message broker.
Resilient messaging
A message broker can add resiliency to the consumers in your system. If a consumer fails while processing a
message, another instance of the consumer can process that message. The reprocessing is possible because the
message persists in the broker.

Technology choices for a message broker

Azure provides several message broker services, each with a range of features. Before choosing a service,
determine the intent and requirements of the message.
Azure Service Bus
Azure Service Bus queues are well suited for transferring commands from producers to consumers. Here are some
considerations.
Pull model
A consumer of a Service Bus queue constantly polls Service Bus to check if new messages are available. The client
SDKs and Azure Functions trigger for Service Bus abstract that model. When a new message is available, the
consumer's callback is invoked, and the message is sent to the consumer.
Guaranteed delivery
Service Bus allows a consumer to peek the queue and lock a message from other consumers.
It's the responsibility of the consumer to report the processing status of the message. Only when the consumer
marks the message as consumed, Service Bus removes the message from the queue. If a failure, timeout, or crash
occurs, Service Bus unlocks the message so that other consumers can retrieve it. This way messages aren't lost in
transfer.
A producer might accidentally send the same message twice. For instance, a producer instance fails after sending a
message. Another producer replaces the original instance and sends the message again. Azure Service Bus queues
provide a built-in de-duping capability that detects and removes duplicate messages. There's still a chance that a
message is delivered twice. For example, if a consumer fails while processing, the message is returned to the queue
and is retrieved by the same or another consumer. The message processing logic in the consumer should be
idempotent so that even if the work is repeated, the state of the system isn't changed. For more information about
idempotency, see Idempotency Patterns on Jonathon Oliver's blog.
Message ordering
If you want consumers to get the messages in the order they are sent, Service Bus queues guarantee first-in-first-
out (FIFO) ordered delivery by using sessions. A session can have one or more messages. The messages are
correlated with the SessionId property. Messages that are part of a session, never expire. A session can be locked
to a consumer to prevent its messages from being handled by a different consumer.
For more information, see Message Sessions.
Message persistence
Service bus queues support temporal decoupling. Even when a consumer isn't available or unable to process the
message, it remains in the queue.
Checkpoint long-running transactions
Business transactions can run for a long time. Each operation in the transaction can have multiple messages. Use
checkpointing to coordinate the workflow and provide resiliency in case a transaction fails.
Service Bus queues allow checkpointing through the session state capability. State information is incrementally
recorded in the queue (SetState ) for messages that belong to a session. For example, a consumer can track
progress by checking the state (GetState ) every now and then. If a consumer fails, another consumer can use state
information to determine the last known checkpoint to resume the session.
Dead-letter queue (DLQ)
A Service Bus queue has a default subqueue, called the dead-letter queue (DLQ) to hold messages that couldn't be
delivered or processed. Service Bus or the message processing logic in the consumer can add messages to the
DLQ. The DLQ keeps the messages until they are retrieved from the queue.
Here are examples when a message can end up being in the DLQ:
A poison message is a message that cannot be handled because it's malformed or contains unexpected
information. In Service Bus queues, you can detect poison messages by setting the MaxDeliver yCount
property of the queue. If number of times the same message is received exceeds that property value,
Service Bus moves the message to the DLQ.
A message might no longer be relevant if it isn't processed within a period. Service Bus queues allow the
producer to post messages with a time-to-live attribute. If this period expires before the message is
received, the message is placed in the DLQ.
Examine messages in the DLQ to determine the reason for failure.
Hybrid solution
Service Bus bridges on-premises systems and cloud solutions. On-premises systems are often difficult to reach
because of firewall restrictions. Both the producer and consumer (either can be on-premises or the cloud) can use
the Service Bus queue endpoint as the pickup and drop off location for messages.
Topics and subscriptions
Service Bus supports the Publisher-Subscriber pattern through Service Bus topics and subscriptions.
This feature provides a way for the producer to broadcast messages to multiple consumers. When a topic receives
a message, it's forwarded to all the subscribed consumers. Optionally, a subscription can have filter criteria that
allows the consumer to get a subset of messages. Each consumer retrieves messages from a subscription in a
similar way to a queue.
For more information, see Azure Service Bus topics.
Azure Event Grid
Azure Event Grid is recommended for discrete events. Event Grid follows the Publisher-Subscriber pattern. When
event sources trigger events, they are published to Event grid topics. Consumers of those events create Event Grid
subscriptions by specifying event types and event handler that will process the events. If there are no subscribers,
the events are discarded. Each event can have multiple subscriptions.
Push Model
Event Grid propagates messages to the subscribers in a push model. Suppose you have an event grid subscription
with a webhook. When a new event arrives, Event Grid posts the event to the webhook endpoint.
Integrated with Azure
Choose Event Grid if you want to get notifications about Azure resources. Many Azure services act as event sources
that have built-in Event Grid topics. Event Grid also supports various Azure services that can be configured as
event handlers. It's easy to subscribe to those topics to route events to event handlers of your choice. For example,
you can use Event Grid to invoke an Azure Function when a blob storage is created or deleted.
Custom topics
Create custom Event Grid topics, if you want to send events from your application or an Azure service that isn't
integrated with Event Grid.
For example, to see the progress of an entire business transaction, you want the participating services to raise
events as they are processing their individual business operations. A web app shows those events. One way is to
create a custom topic and add a subscription with your web app registered through an HTTP WebHook. As
business services send events to the custom topic, Event Grid pushes them to your web app.
Filtered events
You can specify filters in a subscription to instruct Event Grid to route only a subset of events to a specific event
handler. The filters are specified in the subscription schema. Any event sent to the topic with values that match the
filter are automatically forwarded to that subscription.
For example, content in various formats are uploaded to Blob Storage. Each time a file is added, an event is raised
and published to Event Grid. The event subscription might have a filter that only sends events for images so that an
event handler can generate thumbnails.
For more information about filtering, see Filter events for Event Grid.
High throughput
Event Grid can route 10,000,000 events per second per region. The first 100,000 operations per month are free.
For cost considerations, see How much does Event Grid cost?
Resilient delivery
Even though successful delivery for events isn't as crucial as commands, you might still want some guarantee
depending on the type of event. Event Grid offers features that you can enable and customize, such as retry
policies, expiration time, and dead lettering. For more information, see Delivery and retry.
Event Grid's retry process can help resiliency but it's not fail-safe. In the retry process, Event Grid might deliver the
message more than once, skip, or delay some retries if the endpoint is unresponsive for a long time. For more
information, see Retry schedule and duration.
You can persist undelivered events to a blob storage account by enabling dead-lettering. There's a delay in
delivering the message to the blob storage endpoint and if that endpoint is unresponsive, then Event Grid discards
the event. For more information, see Dead letter and retry policies.
Azure Event Hubs
When working with an event stream, Azure Event Hubs is the recommended message broker. Essentially, it's a
large buffer that's capable of receiving large volumes of data with low latency. The received data can be read
quickly through concurrent operations. You can transform the received data by using any real-time analytics
provider. Event Hubs also provides the capability to store events in a storage account.
Fast ingestion
Event Hubs is capable of ingesting millions of events per second. The events are only appended to the stream and
are ordered by time.
Pull model
Like Event Grid, Event Hubs also offers Publisher-Subscriber capabilities. A key difference between Event Grid and
Event Hubs is in the way event data is made available to the subscribers. Event Grid pushes the ingested data to the
subscribers whereas Event Hub makes the data available in a pull model. As events are received, Event Hubs
appends them to the stream. A subscriber manages its cursor and can move forward and back in the stream, select
a time offset, and replay a sequence at its pace.
Stream processors are subscribers that pull data from Event Hubs for the purposes of transformation and
statistical analysis. Use Azure Stream Analytics and Apache Spark for complex processing such as aggregation over
time windows or anomaly detection.
If you want to act on each event per partition, you can pull the data by using Event Processing Host or by using
built in connector such as Logic Apps to provide the transformation logic. Another option is to use Azure Functions.
Partitioning
A partition is a portion of the event stream. The events are divided by using a partition key. For example, several
IoT devices send device data to an event hub. The partition key is the device identifier. As events are ingested, Event
Hubs moves them to separate partitions. Within each partition, all events are ordered by time.
A consumer is an instance of code that processes the event data. Event Hubs follows a partitioned consumer
pattern. Each consumer only reads a specific partition. Having multiple partitions results in faster processing
because the stream can be read concurrently by multiple consumers.
Instances of the same consumer make up a single consumer group. Multiple consumer groups can read the same
stream with different intentions. Suppose an event stream has data from a temperature sensor. One consumer
group can read the stream to detect anomalies such as a spike in temperature. Another can read the same stream
to calculate a rolling average temperature in a temporal window.
Event Hubs supports the Publisher-Subscriber pattern by allowing multiple consumer groups. Each consumer
group is a subscriber.
For more information about Event Hub partitioning, see Partitions.
Event Hubs Capture
The Capture feature allows you to store the event stream to an Azure Blob storage or Data Lake Storage. This way
of storing events is reliable because even if the storage account isn't available, Capture keeps your data for a
period, and then writes to the storage after it's available.

Storage services can also offer additional features for analyzing events. For example, by taking advantage of
the access tiers of a blob storage account, you can store events in a hot tier for data that needs frequent access.
You might use that data for visualization. Alternately, you can store data in the archive tier and retrieve it
occasionally for auditing purposes.

Capture stores all events ingested by Event Hubs and is useful for batch processing. You can generate reports on
the data by using a MapReduce function. Captured data can also serve as the source of truth. If certain facts were
missed while aggregating the data, you can refer to the captured data.
For details about this feature, see Capture events through Azure Event Hubs in Azure Blob Storage or Azure Data
Lake Storage.
Support for Apache Kafka clients
Event Hubs provides an endpoint for Apache Kafka clients. Existing clients can update their configuration to point
to the endpoint and start sending events to Event Hubs. No code changes are required.
For more information, see Event Hubs for Apache Kafka.

Crossover scenarios
In some cases, it's advantageous to combine two messaging services.
Combining services can increase the efficiency of your messaging system. For instance, in your business
transaction, you use Azure Service Bus queues to handle messages. Queues that are mostly idle and receive
messages occasionally are inefficient because the consumer is constantly polling the queue for new messages. You
can set up an Event Grid subscription with an Azure Function as the event handler. Each time the queue receives a
message and there are no consumers listening, Event Grid sends a notification, which invokes the Azure Function
that drains the queue.

For details about connecting Service Bus to Event Grid, see Azure Service Bus to Event Grid integration overview.
The Enterprise integration on Azure using message queues and events reference architecture shows an
implementation of Service Bus to Event Grid integration.
Here's another example. Event Grid receives a set of events in which some events require a workflow while others
are for notification. The message metadata indicates the type of event. One way is to check the metadata by using
the filtering feature in the event subscription. If it requires a workflow, Event Grid sends it to Azure Service Bus
queue. The receivers of that queue can take necessary actions. The notification events are sent to Logic Apps to
send alert emails.

Related patterns
Consider these patterns when implementing asynchronous messaging:
Competing Consumers Pattern. Multiple consumers may need to compete to read messages from a queue. This
pattern explains how to process multiple messages concurrently to optimize throughput, to improve scalability
and availability, and to balance the workload.
Priority Queue Pattern. For cases where the business logic requires that some messages are processed before
others, this pattern describes how messages posted by a producer that have a higher priority can be received
and processed more quickly by a consumer than messages of a lower priority.
Queue-based Load Leveling Pattern. This pattern uses a message broker to act as a buffer between a producer
and a consumer to help to minimize the impact on availability and responsiveness of intermittent heavy loads
for both those entities.
Retry Pattern. A producer or consumer might be unable connect to a queue, but the reasons for this failure may
be temporary and quickly pass. This pattern describes how to handle this situation to add resiliency to an
application.
Scheduler Agent Supervisor Pattern. Messaging is often used as part of a workflow implementation. This
pattern demonstrates how messaging can coordinate a set of actions across a distributed set of services and
other remote resources, and enable a system to recover and retry actions that fail.
Choreography pattern. This pattern shows how services can use messaging to control the workflow of a
business transaction.
Claim-Check Pattern. This pattern shows how to split a large message into a claim check and a payload.
 Web API design
12/18/2020 • 28 minutes to read • Edit Online

Most modern web applications expose APIs that clients can use to interact with the application. A well-designed
web API should aim to support:
Platform independence . Any client should be able to call the API, regardless of how the API is
implemented internally. This requires using standard protocols, and having a mechanism whereby the client
and the web service can agree on the format of the data to exchange.
Ser vice evolution . The web API should be able to evolve and add functionality independently from client
applications. As the API evolves, existing client applications should continue to function without
modification. All functionality should be discoverable so that client applications can fully use it.
This guidance describes issues that you should consider when designing a web API.

Introduction to REST
In 2000, Roy Fielding proposed Representational State Transfer (REST) as an architectural approach to designing
web services. REST is an architectural style for building distributed systems based on hypermedia. REST is
independent of any underlying protocol and is not necessarily tied to HTTP. However, most common REST
implementations use HTTP as the application protocol, and this guide focuses on designing REST APIs for HTTP.
A primary advantage of REST over HTTP is that it uses open standards, and does not bind the implementation of
the API or the client applications to any specific implementation. For example, a REST web service could be written
in ASP.NET, and client applications can use any language or toolset that can generate HTTP requests and parse
HTTP responses.
Here are some of the main design principles of RESTful APIs using HTTP:
REST APIs are designed around resources, which are any kind of object, data, or service that can be
accessed by the client.
A resource has an identifier, which is a URI that uniquely identifies that resource. For example, the URI for a
particular customer order might be:

https://adventure-works.com/orders/1

Clients interact with a service by exchanging representations of resources. Many web APIs use JSON as the
exchange format. For example, a GET request to the URI listed above might return this response body:

{"orderId":1,"orderValue":99.90,"productId":1,"quantity":1}

REST APIs use a uniform interface, which helps to decouple the client and service implementations. For
REST APIs built on HTTP, the uniform interface includes using standard HTTP verbs to perform operations
on resources. The most common operations are GET, POST, PUT, PATCH, and DELETE.
REST APIs use a stateless request model. HTTP requests should be independent and may occur in any order,
so keeping transient state information between requests is not feasible. The only place where information is
stored is in the resources themselves, and each request should be an atomic operation. This constraint
enables web services to be highly scalable, because there is no need to retain any affinity between clients
 and specific servers. Any server can handle any request from any client. That said, other factors can limit
scalability. For example, many web services write to a backend data store, which may be hard to scale out.
For more information about strategies to scale out a data store, see Horizontal, vertical, and functional data
partitioning.
REST APIs are driven by hypermedia links that are contained in the representation. For example, the
following shows a JSON representation of an order. It contains links to get or update the customer
associated with the order.

{
"orderID":3,
"productID":2,
"quantity":4,
"orderValue":16.60,
"links": [
{"rel":"product","href":"https://adventure-works.com/customers/3", "action":"GET" },
{"rel":"product","href":"https://adventure-works.com/customers/3", "action":"PUT" }
]
}

In 2008, Leonard Richardson proposed the following maturity model for web APIs:
Level 0: Define one URI, and all operations are POST requests to this URI.
Level 1: Create separate URIs for individual resources.
Level 2: Use HTTP methods to define operations on resources.
Level 3: Use hypermedia (HATEOAS, described below).
Level 3 corresponds to a truly RESTful API according to Fielding's definition. In practice, many published web APIs
fall somewhere around level 2.

Organize the API around resources

Focus on the business entities that the web API exposes. For example, in an e-commerce system, the primary
entities might be customers and orders. Creating an order can be achieved by sending an HTTP POST request that
contains the order information. The HTTP response indicates whether the order was placed successfully or not.
When possible, resource URIs should be based on nouns (the resource) and not verbs (the operations on the
resource).

https://adventure-works.com/orders // Good

https://adventure-works.com/create-order // Avoid

A resource doesn't have to be based on a single physical data item. For example, an order resource might be
implemented internally as several tables in a relational database, but presented to the client as a single entity.
Avoid creating APIs that simply mirror the internal structure of a database. The purpose of REST is to model
entities and the operations that an application can perform on those entities. A client should not be exposed to the
internal implementation.
Entities are often grouped together into collections (orders, customers). A collection is a separate resource from
the item within the collection, and should have its own URI. For example, the following URI might represent the
collection of orders:

https://adventure-works.com/orders

Sending an HTTP GET request to the collection URI retrieves a list of items in the collection. Each item in the
collection also has its own unique URI. An HTTP GET request to the item's URI returns the details of that item.
Adopt a consistent naming convention in URIs. In general, it helps to use plural nouns for URIs that reference
collections. It's a good practice to organize URIs for collections and items into a hierarchy. For example,
/customers is the path to the customers collection, and /customers/5 is the path to the customer with ID equal to
5. This approach helps to keep the web API intuitive. Also, many web API frameworks can route requests based on
parameterized URI paths, so you could define a route for the path /customers/{id} .
Also consider the relationships between different types of resources and how you might expose these
associations. For example, the /customers/5/orders might represent all of the orders for customer 5. You could
also go in the other direction, and represent the association from an order back to a customer with a URI such as
/orders/99/customer . However, extending this model too far can become cumbersome to implement. A better
solution is to provide navigable links to associated resources in the body of the HTTP response message. This
mechanism is described in more detail in the section Use HATEOAS to enable navigation to related resources.
In more complex systems, it can be tempting to provide URIs that enable a client to navigate through several levels
of relationships, such as /customers/1/orders/99/products . However, this level of complexity can be difficult to
maintain and is inflexible if the relationships between resources change in the future. Instead, try to keep URIs
relatively simple. Once an application has a reference to a resource, it should be possible to use this reference to
find items related to that resource. The preceding query can be replaced with the URI /customers/1/orders to find
all the orders for customer 1, and then /orders/99/products to find the products in this order.

TIP
Avoid requiring resource URIs more complex than collection/item/collection.

Another factor is that all web requests impose a load on the web server. The more requests, the bigger the load.
Therefore, try to avoid "chatty" web APIs that expose a large number of small resources. Such an API may require
a client application to send multiple requests to find all of the data that it requires. Instead, you might want to
denormalize the data and combine related information into bigger resources that can be retrieved with a single
request. However, you need to balance this approach against the overhead of fetching data that the client doesn't
need. Retrieving large objects can increase the latency of a request and incur additional bandwidth costs. For more
information about these performance antipatterns, see Chatty I/O and Extraneous Fetching.
Avoid introducing dependencies between the web API and the underlying data sources. For example, if your data
is stored in a relational database, the web API doesn't need to expose each table as a collection of resources. In fact,
that's probably a poor design. Instead, think of the web API as an abstraction of the database. If necessary,
introduce a mapping layer between the database and the web API. That way, client applications are isolated from
changes to the underlying database scheme.
Finally, it might not be possible to map every operation implemented by a web API to a specific resource. You can
handle such non-resource scenarios through HTTP requests that invoke a function and return the results as an
HTTP response message. For example, a web API that implements simple calculator operations such as add and
subtract could provide URIs that expose these operations as pseudo resources and use the query string to specify
the parameters required. For example, a GET request to the URI /add?operand1=99&operand2=1 would return a
response message with the body containing the value 100. However, only use these forms of URIs sparingly.

Define operations in terms of HTTP methods

The HTTP protocol defines a number of methods that assign semantic meaning to a request. The common HTTP
methods used by most RESTful web APIs are:
GET retrieves a representation of the resource at the specified URI. The body of the response message contains
the details of the requested resource.
 POST creates a new resource at the specified URI. The body of the request message provides the details of the
new resource. Note that POST can also be used to trigger operations that don't actually create resources.
PUT either creates or replaces the resource at the specified URI. The body of the request message specifies the
resource to be created or updated.
PATCH performs a partial update of a resource. The request body specifies the set of changes to apply to the
resource.
DELETE removes the resource at the specified URI.
The effect of a specific request should depend on whether the resource is a collection or an individual item. The
following table summarizes the common conventions adopted by most RESTful implementations using the e-
commerce example. Not all of these requests might be implemented—it depends on the specific scenario.

RESO URC E P O ST GET P UT DEL ET E

/customers Create a new Retrieve all customers Bulk update of Remove all customers
customer customers

/customers/1 Error Retrieve the details Update the details of Remove customer 1
for customer 1 customer 1 if it exists

/customers/1/orders Create a new order Retrieve all orders for Bulk update of orders Remove all orders for
for customer 1 customer 1 for customer 1 customer 1

The differences between POST, PUT, and PATCH can be confusing.

A POST request creates a resource. The server assigns a URI for the new resource, and returns that URI to
the client. In the REST model, you frequently apply POST requests to collections. The new resource is added
to the collection. A POST request can also be used to submit data for processing to an existing resource,
without any new resource being created.
A PUT request creates a resource or updates an existing resource. The client specifies the URI for the
resource. The request body contains a complete representation of the resource. If a resource with this URI
already exists, it is replaced. Otherwise a new resource is created, if the server supports doing so. PUT
requests are most frequently applied to resources that are individual items, such as a specific customer,
rather than collections. A server might support updates but not creation via PUT. Whether to support
creation via PUT depends on whether the client can meaningfully assign a URI to a resource before it exists.
If not, then use POST to create resources and PUT or PATCH to update.
A PATCH request performs a partial update to an existing resource. The client specifies the URI for the
resource. The request body specifies a set of changes to apply to the resource. This can be more efficient
than using PUT, because the client only sends the changes, not the entire representation of the resource.
Technically PATCH can also create a new resource (by specifying a set of updates to a "null" resource), if the
server supports this.
PUT requests must be idempotent. If a client submits the same PUT request multiple times, the results should
always be the same (the same resource will be modified with the same values). POST and PATCH requests are not
guaranteed to be idempotent.

Conform to HTTP semantics

This section describes some typical considerations for designing an API that conforms to the HTTP specification.
However, it doesn't cover every possible detail or scenario. When in doubt, consult the HTTP specifications.
Media types
As mentioned earlier, clients and servers exchange representations of resources. For example, in a POST request,
the request body contains a representation of the resource to create. In a GET request, the response body contains
a representation of the fetched resource.
In the HTTP protocol, formats are specified through the use of media types, also called MIME types. For non-binary
data, most web APIs support JSON (media type = application/json) and possibly XML (media type =
application/xml).
The Content-Type header in a request or response specifies the format of the representation. Here is an example of
a POST request that includes JSON data:

POST https://adventure-works.com/orders HTTP/1.1

Content-Type: application/json; charset=utf-8
Content-Length: 57

{"Id":1,"Name":"Gizmo","Category":"Widgets","Price":1.99}

If the server doesn't support the media type, it should return HTTP status code 415 (Unsupported Media Type).
A client request can include an Accept header that contains a list of media types the client will accept from the
server in the response message. For example:

GET https://adventure-works.com/orders/2 HTTP/1.1

Accept: application/json

If the server cannot match any of the media type(s) listed, it should return HTTP status code 406 (Not Acceptable).
GET methods
A successful GET method typically returns HTTP status code 200 (OK). If the resource cannot be found, the method
should return 404 (Not Found).
POST methods
If a POST method creates a new resource, it returns HTTP status code 201 (Created). The URI of the new resource
is included in the Location header of the response. The response body contains a representation of the resource.
If the method does some processing but does not create a new resource, the method can return HTTP status code
200 and include the result of the operation in the response body. Alternatively, if there is no result to return, the
method can return HTTP status code 204 (No Content) with no response body.
If the client puts invalid data into the request, the server should return HTTP status code 400 (Bad Request). The
response body can contain additional information about the error or a link to a URI that provides more details.
PUT methods
If a PUT method creates a new resource, it returns HTTP status code 201 (Created), as with a POST method. If the
method updates an existing resource, it returns either 200 (OK) or 204 (No Content). In some cases, it might not
be possible to update an existing resource. In that case, consider returning HTTP status code 409 (Conflict).
Consider implementing bulk HTTP PUT operations that can batch updates to multiple resources in a collection. The
PUT request should specify the URI of the collection, and the request body should specify the details of the
resources to be modified. This approach can help to reduce chattiness and improve performance.
PATCH methods
With a PATCH request, the client sends a set of updates to an existing resource, in the form of a patch document.
The server processes the patch document to perform the update. The patch document doesn't describe the whole
resource, only a set of changes to apply. The specification for the PATCH method (RFC 5789) doesn't define a
particular format for patch documents. The format must be inferred from the media type in the request.
JSON is probably the most common data format for web APIs. There are two main JSON-based patch formats,
called JSON patch and JSON merge patch.
JSON merge patch is somewhat simpler. The patch document has the same structure as the original JSON
resource, but includes just the subset of fields that should be changed or added. In addition, a field can be deleted
by specifying null for the field value in the patch document. (That means merge patch is not suitable if the
original resource can have explicit null values.)
For example, suppose the original resource has the following JSON representation:

{
"name":"gizmo",
"category":"widgets",
"color":"blue",
"price":10
}

Here is a possible JSON merge patch for this resource:

{
"price":12,
"color":null,
"size":"small"
}

This tells the server to update price , delete color , and add size , while name and category are not modified.
For the exact details of JSON merge patch, see RFC 7396. The media type for JSON merge patch is
application/merge-patch+json .

Merge patch is not suitable if the original resource can contain explicit null values, due to the special meaning of
null in the patch document. Also, the patch document doesn't specify the order that the server should apply the
updates. That may or may not matter, depending on the data and the domain. JSON patch, defined in RFC 6902, is
more flexible. It specifies the changes as a sequence of operations to apply. Operations include add, remove,
replace, copy, and test (to validate values). The media type for JSON patch is application/json-patch+json .
Here are some typical error conditions that might be encountered when processing a PATCH request, along with
the appropriate HTTP status code.

ERRO R C O N DIT IO N H T T P STAT US C O DE

The patch document format isn't supported. 415 (Unsupported Media Type)

Malformed patch document. 400 (Bad Request)

The patch document is valid, but the changes can't be applied 409 (Conflict)
to the resource in its current state.

DELETE methods
If the delete operation is successful, the web server should respond with HTTP status code 204, indicating that the
process has been successfully handled, but that the response body contains no further information. If the resource
doesn't exist, the web server can return HTTP 404 (Not Found).
Asynchronous operations
Sometimes a POST, PUT, PATCH, or DELETE operation might require processing that takes a while to complete. If
you wait for completion before sending a response to the client, it may cause unacceptable latency. If so, consider
making the operation asynchronous. Return HTTP status code 202 (Accepted) to indicate the request was accepted
for processing but is not completed.
You should expose an endpoint that returns the status of an asynchronous request, so the client can monitor the
status by polling the status endpoint. Include the URI of the status endpoint in the Location header of the 202
response. For example:

HTTP/1.1 202 Accepted

Location: /api/status/12345

If the client sends a GET request to this endpoint, the response should contain the current status of the request.
Optionally, it could also include an estimated time to completion or a link to cancel the operation.

HTTP/1.1 200 OK
Content-Type: application/json

{
"status":"In progress",
"link": { "rel":"cancel", "method":"delete", "href":"/api/status/12345" }
}

If the asynchronous operation creates a new resource, the status endpoint should return status code 303 (See
Other) after the operation completes. In the 303 response, include a Location header that gives the URI of the new
resource:

HTTP/1.1 303 See Other

Location: /api/orders/12345

For more information, see Asynchronous Request-Reply pattern.

Filter and paginate data

Exposing a collection of resources through a single URI can lead to applications fetching large amounts of data
when only a subset of the information is required. For example, suppose a client application needs to find all
orders with a cost over a specific value. It might retrieve all orders from the /orders URI and then filter these
orders on the client side. Clearly this process is highly inefficient. It wastes network bandwidth and processing
power on the server hosting the web API.
Instead, the API can allow passing a filter in the query string of the URI, such as /orders?minCost=n. The web API is
then responsible for parsing and handling the minCost parameter in the query string and returning the filtered
results on the server side.
GET requests over collection resources can potentially return a large number of items. You should design a web
API to limit the amount of data returned by any single request. Consider supporting query strings that specify the
maximum number of items to retrieve and a starting offset into the collection. For example:

/orders?limit=25&offset=50

Also consider imposing an upper limit on the number of items returned, to help prevent Denial of Service attacks.
To assist client applications, GET requests that return paginated data should also include some form of metadata
that indicate the total number of resources available in the collection.
You can use a similar strategy to sort data as it is fetched, by providing a sort parameter that takes a field name as
the value, such as /orders?sort=ProductID. However, this approach can have a negative effect on caching, because
query string parameters form part of the resource identifier used by many cache implementations as the key to
cached data.
You can extend this approach to limit the fields returned for each item, if each item contains a large amount of
data. For example, you could use a query string parameter that accepts a comma-delimited list of fields, such as
/orders?fields=ProductID,Quantity.
Give all optional parameters in query strings meaningful defaults. For example, set the limit parameter to 10
and the offset parameter to 0 if you implement pagination, set the sort parameter to the key of the resource if
you implement ordering, and set the fields parameter to all fields in the resource if you support projections.

Support partial responses for large binary resources

A resource may contain large binary fields, such as files or images. To overcome problems caused by unreliable
and intermittent connections and to improve response times, consider enabling such resources to be retrieved in
chunks. To do this, the web API should support the Accept-Ranges header for GET requests for large resources.
This header indicates that the GET operation supports partial requests. The client application can submit GET
requests that return a subset of a resource, specified as a range of bytes.
Also, consider implementing HTTP HEAD requests for these resources. A HEAD request is similar to a GET request,
except that it only returns the HTTP headers that describe the resource, with an empty message body. A client
application can issue a HEAD request to determine whether to fetch a resource by using partial GET requests. For
example:

HEAD https://adventure-works.com/products/10?fields=productImage HTTP/1.1

Here is an example response message:

HTTP/1.1 200 OK

Accept-Ranges: bytes
Content-Type: image/jpeg
Content-Length: 4580

The Content-Length header gives the total size of the resource, and the Accept-Ranges header indicates that the
corresponding GET operation supports partial results. The client application can use this information to retrieve
the image in smaller chunks. The first request fetches the first 2500 bytes by using the Range header:

GET https://adventure-works.com/products/10?fields=productImage HTTP/1.1

Range: bytes=0-2499

The response message indicates that this is a partial response by returning HTTP status code 206. The Content-
Length header specifies the actual number of bytes returned in the message body (not the size of the resource),
and the Content-Range header indicates which part of the resource this is (bytes 0-2499 out of 4580):

HTTP/1.1 206 Partial Content

Accept-Ranges: bytes
Content-Type: image/jpeg
Content-Length: 2500
Content-Range: bytes 0-2499/4580

[...]

A subsequent request from the client application can retrieve the remainder of the resource.
Use HATEOAS to enable navigation to related resources
One of the primary motivations behind REST is that it should be possible to navigate the entire set of resources
without requiring prior knowledge of the URI scheme. Each HTTP GET request should return the information
necessary to find the resources related directly to the requested object through hyperlinks included in the
response, and it should also be provided with information that describes the operations available on each of these
resources. This principle is known as HATEOAS, or Hypertext as the Engine of Application State. The system is
effectively a finite state machine, and the response to each request contains the information necessary to move
from one state to another; no other information should be necessary.

NOTE
Currently there are no general-purpose standards that define how to model the HATEOAS principle. The examples shown in
this section illustrate one possible, proprietary solution.

For example, to handle the relationship between an order and a customer, the representation of an order could
include links that identify the available operations for the customer of the order. Here is a possible representation:

{
"orderID":3,
"productID":2,
"quantity":4,
"orderValue":16.60,
"links":[
{
"rel":"customer",
"href":"https://adventure-works.com/customers/3",
"action":"GET",
"types":["text/xml","application/json"]
},
{
"rel":"customer",
"href":"https://adventure-works.com/customers/3",
"action":"PUT",
"types":["application/x-www-form-urlencoded"]
},
{
"rel":"customer",
"href":"https://adventure-works.com/customers/3",
"action":"DELETE",
"types":[]
},
{
"rel":"self",
"href":"https://adventure-works.com/orders/3",
"action":"GET",
"types":["text/xml","application/json"]
},
{
"rel":"self",
"href":"https://adventure-works.com/orders/3",
"action":"PUT",
"types":["application/x-www-form-urlencoded"]
},
{
"rel":"self",
"href":"https://adventure-works.com/orders/3",
"action":"DELETE",
"types":[]
}]
}
In this example, the links array has a set of links. Each link represents an operation on a related entity. The data
for each link includes the relationship ("customer"), the URI ( https://adventure-works.com/customers/3 ), the HTTP
method, and the supported MIME types. This is all the information that a client application needs to be able to
invoke the operation.
The links array also includes self-referencing information about the resource itself that has been retrieved. These
have the relationship self.
The set of links that are returned may change, depending on the state of the resource. This is what is meant by
hypertext being the "engine of application state."

Versioning a RESTful web API

It is highly unlikely that a web API will remain static. As business requirements change new collections of
resources may be added, the relationships between resources might change, and the structure of the data in
resources might be amended. While updating a web API to handle new or differing requirements is a relatively
straightforward process, you must consider the effects that such changes will have on client applications
consuming the web API. The issue is that although the developer designing and implementing a web API has full
control over that API, the developer does not have the same degree of control over client applications, which may
be built by third-party organizations operating remotely. The primary imperative is to enable existing client
applications to continue functioning unchanged while allowing new client applications to take advantage of new
features and resources.
Versioning enables a web API to indicate the features and resources that it exposes, and a client application can
submit requests that are directed to a specific version of a feature or resource. The following sections describe
several different approaches, each of which has its own benefits and trade-offs.
No versioning
This is the simplest approach, and may be acceptable for some internal APIs. Significant changes could be
represented as new resources or new links. Adding content to existing resources might not present a breaking
change as client applications that are not expecting to see this content will ignore it.
For example, a request to the URI https://adventure-works.com/customers/3 should return the details of a single
customer containing id , name , and address fields expected by the client application:

HTTP/1.1 200 OK
Content-Type: application/json; charset=utf-8

{"id":3,"name":"Contoso LLC","address":"1 Microsoft Way Redmond WA 98053"}

NOTE
For simplicity, the example responses shown in this section do not include HATEOAS links.

If the DateCreated field is added to the schema of the customer resource, then the response would look like this:

HTTP/1.1 200 OK
Content-Type: application/json; charset=utf-8

{"id":3,"name":"Contoso LLC","dateCreated":"2014-09-04T12:11:38.0376089Z","address":"1 Microsoft Way Redmond

WA 98053"}

Existing client applications might continue functioning correctly if they are capable of ignoring unrecognized fields,
while new client applications can be designed to handle this new field. However, if more radical changes to the
schema of resources occur (such as removing or renaming fields) or the relationships between resources change
then these may constitute breaking changes that prevent existing client applications from functioning correctly. In
these situations, you should consider one of the following approaches.
URI versioning
Each time you modify the web API or change the schema of resources, you add a version number to the URI for
each resource. The previously existing URIs should continue to operate as before, returning resources that
conform to their original schema.
Extending the previous example, if the address field is restructured into subfields containing each constituent part
of the address (such as streetAddress , city , state , and zipCode ), this version of the resource could be exposed
through a URI containing a version number, such as https://adventure-works.com/v2/customers/3 :

HTTP/1.1 200 OK
Content-Type: application/json; charset=utf-8

{"id":3,"name":"Contoso LLC","dateCreated":"2014-09-04T12:11:38.0376089Z","address":{"streetAddress":"1
Microsoft Way","city":"Redmond","state":"WA","zipCode":98053}}

This versioning mechanism is very simple but depends on the server routing the request to the appropriate
endpoint. However, it can become unwieldy as the web API matures through several iterations and the server has
to support a number of different versions. Also, from a purist's point of view, in all cases the client applications are
fetching the same data (customer 3), so the URI should not really be different depending on the version. This
scheme also complicates implementation of HATEOAS as all links will need to include the version number in their
URIs.
Query string versioning
Rather than providing multiple URIs, you can specify the version of the resource by using a parameter within the
query string appended to the HTTP request, such as https://adventure-works.com/customers/3?version=2 . The
version parameter should default to a meaningful value such as 1 if it is omitted by older client applications.
This approach has the semantic advantage that the same resource is always retrieved from the same URI, but it
depends on the code that handles the request to parse the query string and send back the appropriate HTTP
response. This approach also suffers from the same complications for implementing HATEOAS as the URI
versioning mechanism.

NOTE
Some older web browsers and web proxies will not cache responses for requests that include a query string in the URI. This
can degrade performance for web applications that use a web API and that run from within such a web browser.

Header versioning
Rather than appending the version number as a query string parameter, you could implement a custom header
that indicates the version of the resource. This approach requires that the client application adds the appropriate
header to any requests, although the code handling the client request could use a default value (version 1) if the
version header is omitted. The following examples use a custom header named Custom-Header. The value of this
header indicates the version of web API.
Version 1:

GET https://adventure-works.com/customers/3 HTTP/1.1

Custom-Header: api-version=1
 HTTP/1.1 200 OK
Content-Type: application/json; charset=utf-8

{"id":3,"name":"Contoso LLC","address":"1 Microsoft Way Redmond WA 98053"}

Version 2:

GET https://adventure-works.com/customers/3 HTTP/1.1

Custom-Header: api-version=2

HTTP/1.1 200 OK
Content-Type: application/json; charset=utf-8

{"id":3,"name":"Contoso LLC","dateCreated":"2014-09-04T12:11:38.0376089Z","address":{"streetAddress":"1
Microsoft Way","city":"Redmond","state":"WA","zipCode":98053}}

As with the previous two approaches, implementing HATEOAS requires including the appropriate custom header
in any links.
Media type versioning
When a client application sends an HTTP GET request to a web server it should stipulate the format of the content
that it can handle by using an Accept header, as described earlier in this guidance. Frequently the purpose of the
Accept header is to allow the client application to specify whether the body of the response should be XML, JSON,
or some other common format that the client can parse. However, it is possible to define custom media types that
include information enabling the client application to indicate which version of a resource it is expecting. The
following example shows a request that specifies an Accept header with the value application/vnd.adventure-
works.v1+json. The vnd.adventure-works.v1 element indicates to the web server that it should return version 1 of
the resource, while the json element specifies that the format of the response body should be JSON:

GET https://adventure-works.com/customers/3 HTTP/1.1

Accept: application/vnd.adventure-works.v1+json

The code handling the request is responsible for processing the Accept header and honoring it as far as possible
(the client application may specify multiple formats in the Accept header, in which case the web server can choose
the most appropriate format for the response body). The web server confirms the format of the data in the
response body by using the Content-Type header:

HTTP/1.1 200 OK
Content-Type: application/vnd.adventure-works.v1+json; charset=utf-8

{"id":3,"name":"Contoso LLC","address":"1 Microsoft Way Redmond WA 98053"}

If the Accept header does not specify any known media types, the web server could generate an HTTP 406 (Not
Acceptable) response message or return a message with a default media type.
This approach is arguably the purest of the versioning mechanisms and lends itself naturally to HATEOAS, which
can include the MIME type of related data in resource links.
 NOTE
When you select a versioning strategy, you should also consider the implications on performance, especially caching on the
web server. The URI versioning and Query String versioning schemes are cache-friendly inasmuch as the same URI/query
string combination refers to the same data each time.
The Header versioning and Media Type versioning mechanisms typically require additional logic to examine the values in the
custom header or the Accept header. In a large-scale environment, many clients using different versions of a web API can
result in a significant amount of duplicated data in a server-side cache. This issue can become acute if a client application
communicates with a web server through a proxy that implements caching, and that only forwards a request to the web
server if it does not currently hold a copy of the requested data in its cache.

Open API Initiative

The Open API Initiative was created by an industry consortium to standardize REST API descriptions across
vendors. As part of this initiative, the Swagger 2.0 specification was renamed the OpenAPI Specification (OAS) and
brought under the Open API Initiative.
You may want to adopt OpenAPI for your web APIs. Some points to consider:
The OpenAPI Specification comes with a set of opinionated guidelines on how a REST API should be
designed. That has advantages for interoperability, but requires more care when designing your API to
conform to the specification.
OpenAPI promotes a contract-first approach, rather than an implementation-first approach. Contract-first
means you design the API contract (the interface) first and then write code that implements the contract.
Tools like Swagger can generate client libraries or documentation from API contracts. For example, see
ASP.NET Web API help pages using Swagger.

More information
Microsoft REST API guidelines. Detailed recommendations for designing public REST APIs.
Web API checklist. A useful list of items to consider when designing and implementing a web API.
Open API Initiative. Documentation and implementation details on Open API.
 Web API implementation
12/18/2020 • 46 minutes to read • Edit Online

A carefully designed RESTful web API defines the resources, relationships, and navigation schemes that are
accessible to client applications. When you implement and deploy a web API, you should consider the physical
requirements of the environment hosting the web API and the way in which the web API is constructed rather than
the logical structure of the data. This guidance focuses on best practices for implementing a web API and
publishing it to make it available to client applications. For detailed information about web API design, see Web API
design.

Processing requests
Consider the following points when you implement the code to handle requests.
GET, PUT, DELETE, HEAD, and PATCH actions should be idempotent
The code that implements these requests should not impose any side-effects. The same request repeated over the
same resource should result in the same state. For example, sending multiple DELETE requests to the same URI
should have the same effect, although the HTTP status code in the response messages may be different. The first
DELETE request might return status code 204 (No Content), while a subsequent DELETE request might return status
code 404 (Not Found).

NOTE
The article Idempotency Patterns on Jonathan Oliver's blog provides an overview of idempotency and how it relates to data
management operations.

POST actions that create new resources should not have unrelated side -effects
If a POST request is intended to create a new resource, the effects of the request should be limited to the new
resource (and possibly any directly related resources if there is some sort of linkage involved) For example, in an e-
commerce system, a POST request that creates a new order for a customer might also amend inventory levels and
generate billing information, but it should not modify information not directly related to the order or have any
other side-effects on the overall state of the system.
Avoid implementing chatty POST, PUT, and DELETE operations
Support POST, PUT and DELETE requests over resource collections. A POST request can contain the details for
multiple new resources and add them all to the same collection, a PUT request can replace the entire set of
resources in a collection, and a DELETE request can remove an entire collection.
The OData support included in ASP.NET Web API 2 provides the ability to batch requests. A client application can
package up several web API requests and send them to the server in a single HTTP request, and receive a single
HTTP response that contains the replies to each request. For more information, Introducing batch support in Web
API and Web API OData.
Follow the HTTP specification when sending a response
A web API must return messages that contain the correct HTTP status code to enable the client to determine how to
handle the result, the appropriate HTTP headers so that the client understands the nature of the result, and a
suitably formatted body to enable the client to parse the result.
For example, a POST operation should return status code 201 (Created) and the response message should include
the URI of the newly created resource in the Location header of the response message.
Support content negotiation
The body of a response message may contain data in a variety of formats. For example, an HTTP GET request could
return data in JSON, or XML format. When the client submits a request, it can include an Accept header that
specifies the data formats that it can handle. These formats are specified as media types. For example, a client that
issues a GET request that retrieves an image can specify an Accept header that lists the media types that the client
can handle, such as image/jpeg, image/gif, image/png . When the web API returns the result, it should format the
data by using one of these media types and specify the format in the Content-Type header of the response.
If the client does not specify an Accept header, then use a sensible default format for the response body. As an
example, the ASP.NET Web API framework defaults to JSON for text-based data.
Provide links to support HATEOAS -style navigation and discovery of resources
The HATEOAS approach enables a client to navigate and discover resources from an initial starting point. This is
achieved by using links containing URIs; when a client issues an HTTP GET request to obtain a resource, the
response should contain URIs that enable a client application to quickly locate any directly related resources. For
example, in a web API that supports an e-commerce solution, a customer may have placed many orders. When a
client application retrieves the details for a customer, the response should include links that enable the client
application to send HTTP GET requests that can retrieve these orders. Additionally, HATEOAS-style links should
describe the other operations (POST, PUT, DELETE, and so on) that each linked resource supports together with the
corresponding URI to perform each request. This approach is described in more detail in API design.
Currently there are no standards that govern the implementation of HATEOAS, but the following example illustrates
one possible approach. In this example, an HTTP GET request that finds the details for a customer returns a
response that includes HATEOAS links that reference the orders for that customer:

GET https://adventure-works.com/customers/2 HTTP/1.1

Accept: text/json
...

HTTP/1.1 200 OK
...
Content-Type: application/json; charset=utf-8
...
Content-Length: ...
{"CustomerID":2,"CustomerName":"Bert","Links":[
{"rel":"self",
"href":"https://adventure-works.com/customers/2",
"action":"GET",
"types":["text/xml","application/json"]},
{"rel":"self",
"href":"https://adventure-works.com/customers/2",
"action":"PUT",
"types":["application/x-www-form-urlencoded"]},
{"rel":"self",
"href":"https://adventure-works.com/customers/2",
"action":"DELETE",
"types":[]},
{"rel":"orders",
"href":"https://adventure-works.com/customers/2/orders",
"action":"GET",
"types":["text/xml","application/json"]},
{"rel":"orders",
"href":"https://adventure-works.com/customers/2/orders",
"action":"POST",
"types":["application/x-www-form-urlencoded"]}
]}

In this example, the customer data is represented by the Customer class shown in the following code snippet. The
HATEOAS links are held in the Links collection property:

public class Customer

{
public int CustomerID { get; set; }
public string CustomerName { get; set; }
public List<Link> Links { get; set; }
...
}

public class Link

{
public string Rel { get; set; }
public string Href { get; set; }
public string Action { get; set; }
public string [] Types { get; set; }
}

The HTTP GET operation retrieves the customer data from storage and constructs a Customer object, and then
populates the Links collection. The result is formatted as a JSON response message. Each link comprises the
following fields:
The relationship between the object being returned and the object described by the link. In this case self
indicates that the link is a reference back to the object itself (similar to a this pointer in many object-oriented
languages), and orders is the name of a collection containing the related order information.
The hyperlink ( Href ) for the object being described by the link in the form of a URI.
The type of HTTP request ( Action ) that can be sent to this URI.
The format of any data ( Types ) that should be provided in the HTTP request or that can be returned in the
response, depending on the type of the request.
The HATEOAS links shown in the example HTTP response indicate that a client application can perform the
following operations:
An HTTP GET request to the URI https://adventure-works.com/customers/2 to fetch the details of the customer
(again). The data can be returned as XML or JSON.
An HTTP PUT request to the URI https://adventure-works.com/customers/2 to modify the details of the customer.
The new data must be provided in the request message in x-www-form-urlencoded format.
An HTTP DELETE request to the URI https://adventure-works.com/customers/2 to delete the customer. The
request does not expect any additional information or return data in the response message body.
An HTTP GET request to the URI https://adventure-works.com/customers/2/orders to find all the orders for the
customer. The data can be returned as XML or JSON.
An HTTP POST request to the URI https://adventure-works.com/customers/2/orders to create a new order for this
customer. The data must be provided in the request message in x-www-form-urlencoded format.

Handling exceptions
Consider the following points if an operation throws an uncaught exception.
Capture exceptions and return a meaningful response to clients
The code that implements an HTTP operation should provide comprehensive exception handling rather than letting
uncaught exceptions propagate to the framework. If an exception makes it impossible to complete the operation
successfully, the exception can be passed back in the response message, but it should include a meaningful
description of the error that caused the exception. The exception should also include the appropriate HTTP status
code rather than simply returning status code 500 for every situation. For example, if a user request causes a
database update that violates a constraint (such as attempting to delete a customer that has outstanding orders),
you should return status code 409 (Conflict) and a message body indicating the reason for the conflict. If some
other condition renders the request unachievable, you can return status code 400 (Bad Request). You can find a full
list of HTTP status codes on the Status code definitions page on the W3C website.
The code example traps different conditions and returns an appropriate response.

[HttpDelete]
[Route("customers/{id:int}")]
public IHttpActionResult DeleteCustomer(int id)
{
try
{
// Find the customer to be deleted in the repository
var customerToDelete = repository.GetCustomer(id);

// If there is no such customer, return an error response

// with status code 404 (Not Found)
if (customerToDelete == null)
{
return NotFound();
}

// Remove the customer from the repository

// The DeleteCustomer method returns true if the customer
// was successfully deleted
if (repository.DeleteCustomer(id))
{
// Return a response message with status code 204 (No Content)
// To indicate that the operation was successful
return StatusCode(HttpStatusCode.NoContent);
}
else
{
// Otherwise return a 400 (Bad Request) error response
return BadRequest(Strings.CustomerNotDeleted);
}
}
catch
{
// If an uncaught exception occurs, return an error response
// with status code 500 (Internal Server Error)
return InternalServerError();
}
}

TIP
Do not include information that could be useful to an attacker attempting to penetrate your API.

Many web servers trap error conditions themselves before they reach the web API. For example, if you configure
authentication for a web site and the user fails to provide the correct authentication information, the web server
should respond with status code 401 (Unauthorized). Once a client has been authenticated, your code can perform
its own checks to verify that the client should be able access the requested resource. If this authorization fails, you
should return status code 403 (Forbidden).
Handle exceptions consistently and log information about errors
To handle exceptions in a consistent manner, consider implementing a global error handling strategy across the
entire web API. You should also incorporate error logging which captures the full details of each exception; this
error log can contain detailed information as long as it is not made accessible over the web to clients.
Distinguish between client-side errors and server-side errors
The HTTP protocol distinguishes between errors that occur due to the client application (the HTTP 4xx status codes),
and errors that are caused by a mishap on the server (the HTTP 5xx status codes). Make sure that you respect this
convention in any error response messages.

Optimizing client-side data access

In a distributed environment such as that involving a web server and client applications, one of the primary sources
of concern is the network. This can act as a considerable bottleneck, especially if a client application is frequently
sending requests or receiving data. Therefore you should aim to minimize the amount of traffic that flows across
the network. Consider the following points when you implement the code to retrieve and maintain data:
Support client-side caching
The HTTP 1.1 protocol supports caching in clients and intermediate servers through which a request is routed by
the use of the Cache-Control header. When a client application sends an HTTP GET request to the web API, the
response can include a Cache-Control header that indicates whether the data in the body of the response can be
safely cached by the client or an intermediate server through which the request has been routed, and for how long
before it should expire and be considered out-of-date. The following example shows an HTTP GET request and the
corresponding response that includes a Cache-Control header:

GET https://adventure-works.com/orders/2 HTTP/1.1

HTTP/1.1 200 OK
...
Cache-Control: max-age=600, private
Content-Type: text/json; charset=utf-8
Content-Length: ...
{"orderID":2,"productID":4,"quantity":2,"orderValue":10.00}

In this example, the Cache-Control header specifies that the data returned should be expired after 600 seconds, and
is only suitable for a single client and must not be stored in a shared cache used by other clients (it is private). The
Cache-Control header could specify public rather than private in which case the data can be stored in a shared
cache, or it could specify no-store in which case the data must not be cached by the client. The following code
example shows how to construct a Cache-Control header in a response message:
 public class OrdersController : ApiController
{
...
[Route("api/orders/{id:int:min(0)}")]
[HttpGet]
public IHttpActionResult FindOrderByID(int id)
{
// Find the matching order
Order order = ...;
...
// Create a Cache-Control header for the response
var cacheControlHeader = new CacheControlHeaderValue();
cacheControlHeader.Private = true;
cacheControlHeader.MaxAge = new TimeSpan(0, 10, 0);
...

// Return a response message containing the order and the cache control header
OkResultWithCaching<Order> response = new OkResultWithCaching<Order>(order, this)
{
CacheControlHeader = cacheControlHeader
};
return response;
}
...
}

This code uses a custom IHttpActionResult class named OkResultWithCaching . This class enables the controller to
set the cache header contents:

public class OkResultWithCaching<T> : OkNegotiatedContentResult<T>

{
public OkResultWithCaching(T content, ApiController controller)
: base(content, controller) { }

public OkResultWithCaching(T content, IContentNegotiator contentNegotiator, HttpRequestMessage request,

IEnumerable<MediaTypeFormatter> formatters)
: base(content, contentNegotiator, request, formatters) { }

public CacheControlHeaderValue CacheControlHeader { get; set; }

public EntityTagHeaderValue ETag { get; set; }

public override async Task<HttpResponseMessage> ExecuteAsync(CancellationToken cancellationToken)

{
HttpResponseMessage response;
try
{
response = await base.ExecuteAsync(cancellationToken);
response.Headers.CacheControl = this.CacheControlHeader;
response.Headers.ETag = ETag;
}
catch (OperationCanceledException)
{
response = new HttpResponseMessage(HttpStatusCode.Conflict) {ReasonPhrase = "Operation was
cancelled"};
}
return response;
}
}
 NOTE
The HTTP protocol also defines the no-cache directive for the Cache-Control header. Rather confusingly, this directive does
not mean "do not cache" but rather "revalidate the cached information with the server before returning it"; the data can still
be cached, but it is checked each time it is used to ensure that it is still current.

Cache management is the responsibility of the client application or intermediate server, but if properly
implemented it can save bandwidth and improve performance by removing the need to fetch data that has already
been recently retrieved.
The max-age value in the Cache-Control header is only a guide and not a guarantee that the corresponding data
won't change during the specified time. The web API should set the max-age to a suitable value depending on the
expected volatility of the data. When this period expires, the client should discard the object from the cache.

NOTE
Most modern web browsers support client-side caching by adding the appropriate cache-control headers to requests and
examining the headers of the results, as described. However, some older browsers will not cache the values returned from a
URL that includes a query string. This is not usually an issue for custom client applications which implement their own cache
management strategy based on the protocol discussed here.
Some older proxies exhibit the same behavior and might not cache requests based on URLs with query strings. This could be
an issue for custom client applications that connect to a web server through such a proxy.

Provide ETags to optimize query processing

When a client application retrieves an object, the response message can also include an ETag (Entity Tag). An ETag is
an opaque string that indicates the version of a resource; each time a resource changes the ETag is also modified.
This ETag should be cached as part of the data by the client application. The following code example shows how to
add an ETag as part of the response to an HTTP GET request. This code uses the GetHashCode method of an object
to generate a numeric value that identifies the object (you can override this method if necessary and generate your
own hash using an algorithm such as MD5) :

public class OrdersController : ApiController

{
...
public IHttpActionResult FindOrderByID(int id)
{
// Find the matching order
Order order = ...;
...

var hashedOrder = order.GetHashCode();

string hashedOrderEtag = $"\"{hashedOrder}\"";
var eTag = new EntityTagHeaderValue(hashedOrderEtag);

// Return a response message containing the order and the cache control header
OkResultWithCaching<Order> response = new OkResultWithCaching<Order>(order, this)
{
...,
ETag = eTag
};
return response;
}
...
}

The response message posted by the web API looks like this:
 HTTP/1.1 200 OK
...
Cache-Control: max-age=600, private
Content-Type: text/json; charset=utf-8
ETag: "2147483648"
Content-Length: ...
{"orderID":2,"productID":4,"quantity":2,"orderValue":10.00}

TIP
For security reasons, do not allow sensitive data or data returned over an authenticated (HTTPS) connection to be cached.

A client application can issue a subsequent GET request to retrieve the same resource at any time, and if the
resource has changed (it has a different ETag) the cached version should be discarded and the new version added
to the cache. If a resource is large and requires a significant amount of bandwidth to transmit back to the client,
repeated requests to fetch the same data can become inefficient. To combat this, the HTTP protocol defines the
following process for optimizing GET requests that you should support in a web API:
The client constructs a GET request containing the ETag for the currently cached version of the resource
referenced in an If-None-Match HTTP header:

GET https://adventure-works.com/orders/2 HTTP/1.1

If-None-Match: "2147483648"

The GET operation in the web API obtains the current ETag for the requested data (order 2 in the above
example), and compares it to the value in the If-None-Match header.
If the current ETag for the requested data matches the ETag provided by the request, the resource has not
changed and the web API should return an HTTP response with an empty message body and a status code
of 304 (Not Modified).
If the current ETag for the requested data does not match the ETag provided by the request, then the data has
changed and the web API should return an HTTP response with the new data in the message body and a
status code of 200 (OK).
If the requested data no longer exists then the web API should return an HTTP response with the status code
of 404 (Not Found).
The client uses the status code to maintain the cache. If the data has not changed (status code 304) then the
object can remain cached and the client application should continue to use this version of the object. If the
data has changed (status code 200) then the cached object should be discarded and the new one inserted. If
the data is no longer available (status code 404) then the object should be removed from the cache.

NOTE
If the response header contains the Cache-Control header no-store then the object should always be removed from the
cache regardless of the HTTP status code.

The code below shows the FindOrderByID method extended to support the If-None-Match header. Notice that if the
If-None-Match header is omitted, the specified order is always retrieved:
 public class OrdersController : ApiController
{
[Route("api/orders/{id:int:min(0)}")]
[HttpGet]
public IHttpActionResult FindOrderByID(int id)
{
try
{
// Find the matching order
Order order = ...;

// If there is no such order then return NotFound

if (order == null)
{
return NotFound();
}

// Generate the ETag for the order

var hashedOrder = order.GetHashCode();
string hashedOrderEtag = $"\"{hashedOrder}\"";

// Create the Cache-Control and ETag headers for the response

IHttpActionResult response;
var cacheControlHeader = new CacheControlHeaderValue();
cacheControlHeader.Public = true;
cacheControlHeader.MaxAge = new TimeSpan(0, 10, 0);
var eTag = new EntityTagHeaderValue(hashedOrderEtag);

// Retrieve the If-None-Match header from the request (if it exists)

var nonMatchEtags = Request.Headers.IfNoneMatch;

// If there is an ETag in the If-None-Match header and

// this ETag matches that of the order just retrieved,
// then create a Not Modified response message
if (nonMatchEtags.Count > 0 &&
String.CompareOrdinal(nonMatchEtags.First().Tag, hashedOrderEtag) == 0)
{
response = new EmptyResultWithCaching()
{
StatusCode = HttpStatusCode.NotModified,
CacheControlHeader = cacheControlHeader,
ETag = eTag
};
}
// Otherwise create a response message that contains the order details
else
{
response = new OkResultWithCaching<Order>(order, this)
{
CacheControlHeader = cacheControlHeader,
ETag = eTag
};
}

return response;
}
catch
{
return InternalServerError();
}
}
...
}

This example incorporates an additional custom IHttpActionResult class named EmptyResultWithCaching . This class
simply acts as a wrapper around an HttpResponseMessage object that does not contain a response body:
 public class EmptyResultWithCaching : IHttpActionResult
{
public CacheControlHeaderValue CacheControlHeader { get; set; }
public EntityTagHeaderValue ETag { get; set; }
public HttpStatusCode StatusCode { get; set; }
public Uri Location { get; set; }

public async Task<HttpResponseMessage> ExecuteAsync(CancellationToken cancellationToken)

{
HttpResponseMessage response = new HttpResponseMessage(StatusCode);
response.Headers.CacheControl = this.CacheControlHeader;
response.Headers.ETag = this.ETag;
response.Headers.Location = this.Location;
return response;
}
}

TIP
In this example, the ETag for the data is generated by hashing the data retrieved from the underlying data source. If the ETag
can be computed in some other way, then the process can be optimized further and the data only needs to be fetched from
the data source if it has changed. This approach is especially useful if the data is large or accessing the data source can result
in significant latency (for example, if the data source is a remote database).

Use ETags to Support Optimistic Concurrency

To enable updates over previously cached data, the HTTP protocol supports an optimistic concurrency strategy. If,
after fetching and caching a resource, the client application subsequently sends a PUT or DELETE request to change
or remove the resource, it should include in If-Match header that references the ETag. The web API can then use this
information to determine whether the resource has already been changed by another user since it was retrieved
and send an appropriate response back to the client application as follows:
The client constructs a PUT request containing the new details for the resource and the ETag for the currently
cached version of the resource referenced in an If-Match HTTP header. The following example shows a PUT
request that updates an order:

PUT https://adventure-works.com/orders/1 HTTP/1.1

If-Match: "2282343857"
Content-Type: application/x-www-form-urlencoded
Content-Length: ...
productID=3&quantity=5&orderValue=250

The PUT operation in the web API obtains the current ETag for the requested data (order 1 in the above
example), and compares it to the value in the If-Match header.
If the current ETag for the requested data matches the ETag provided by the request, the resource has not
changed and the web API should perform the update, returning a message with HTTP status code 204 (No
Content) if it is successful. The response can include Cache-Control and ETag headers for the updated
version of the resource. The response should always include the Location header that references the URI of
the newly updated resource.
If the current ETag for the requested data does not match the ETag provided by the request, then the data has
been changed by another user since it was fetched and the web API should return an HTTP response with an
empty message body and a status code of 412 (Precondition Failed).
If the resource to be updated no longer exists then the web API should return an HTTP response with the
status code of 404 (Not Found).
 The client uses the status code and response headers to maintain the cache. If the data has been updated
(status code 204) then the object can remain cached (as long as the Cache-Control header does not specify
no-store) but the ETag should be updated. If the data was changed by another user changed (status code
412) or not found (status code 404) then the cached object should be discarded.
The next code example shows an implementation of the PUT operation for the Orders controller:

public class OrdersController : ApiController

{
[HttpPut]
[Route("api/orders/{id:int}")]
public IHttpActionResult UpdateExistingOrder(int id, DTOOrder order)
{
try
{
var baseUri = Constants.GetUriFromConfig();
var orderToUpdate = this.ordersRepository.GetOrder(id);
if (orderToUpdate == null)
{
return NotFound();
}

var hashedOrder = orderToUpdate.GetHashCode();

string hashedOrderEtag = $"\"{hashedOrder}\"";

// Retrieve the If-Match header from the request (if it exists)

var matchEtags = Request.Headers.IfMatch;

// If there is an ETag in the If-Match header and

// this ETag matches that of the order just retrieved,
// or if there is no ETag, then update the Order
if (((matchEtags.Count > 0 &&
String.CompareOrdinal(matchEtags.First().Tag, hashedOrderEtag) == 0)) ||
matchEtags.Count == 0)
{
// Modify the order
orderToUpdate.OrderValue = order.OrderValue;
orderToUpdate.ProductID = order.ProductID;
orderToUpdate.Quantity = order.Quantity;

// Save the order back to the data store

// ...

// Create the No Content response with Cache-Control, ETag, and Location headers
var cacheControlHeader = new CacheControlHeaderValue();
cacheControlHeader.Private = true;
cacheControlHeader.MaxAge = new TimeSpan(0, 10, 0);

hashedOrder = order.GetHashCode();
hashedOrderEtag = $"\"{hashedOrder}\"";
var eTag = new EntityTagHeaderValue(hashedOrderEtag);

var location = new Uri($"{baseUri}/{Constants.ORDERS}/{id}");

var response = new EmptyResultWithCaching()
{
StatusCode = HttpStatusCode.NoContent,
CacheControlHeader = cacheControlHeader,
ETag = eTag,
Location = location
};

return response;
}

// Otherwise return a Precondition Failed response

return StatusCode(HttpStatusCode.PreconditionFailed);
}
 }
catch
{
return InternalServerError();
}
}
...
}

TIP
Use of the If-Match header is entirely optional, and if it is omitted the web API will always attempt to update the specified
order, possibly blindly overwriting an update made by another user. To avoid problems due to lost updates, always provide an
If-Match header.

Handling large requests and responses

There may be occasions when a client application needs to issue requests that send or receive data that may be
several megabytes (or bigger) in size. Waiting while this amount of data is transmitted could cause the client
application to become unresponsive. Consider the following points when you need to handle requests that include
significant amounts of data:
Optimize requests and responses that involve large objects
Some resources may be large objects or include large fields, such as graphics images or other types of binary data.
A web API should support streaming to enable optimized uploading and downloading of these resources.
The HTTP protocol provides the chunked transfer encoding mechanism to stream large data objects back to a client.
When the client sends an HTTP GET request for a large object, the web API can send the reply back in piecemeal
chunks over an HTTP connection. The length of the data in the reply may not be known initially (it might be
generated), so the server hosting the web API should send a response message with each chunk that specifies the
Transfer-Encoding: Chunked header rather than a Content-Length header. The client application can receive each
chunk in turn to build up the complete response. The data transfer completes when the server sends back a final
chunk with zero size.
A single request could conceivably result in a massive object that consumes considerable resources. If during the
streaming process the web API determines that the amount of data in a request has exceeded some acceptable
bounds, it can abort the operation and return a response message with status code 413 (Request Entity Too Large).
You can minimize the size of large objects transmitted over the network by using HTTP compression. This approach
helps to reduce the amount of network traffic and the associated network latency, but at the cost of requiring
additional processing at the client and the server hosting the web API. For example, a client application that expects
to receive compressed data can include an Accept-Encoding: gzip request header (other data compression
algorithms can also be specified). If the server supports compression it should respond with the content held in
gzip format in the message body and the Content-Encoding: gzip response header.
You can combine encoded compression with streaming; compress the data first before streaming it, and specify the
gzip content encoding and chunked transfer encoding in the message headers. Also note that some web servers
(such as Internet Information Server) can be configured to automatically compress HTTP responses regardless of
whether the web API compresses the data or not.
Implement partial responses for clients that do not support asynchronous operations
As an alternative to asynchronous streaming, a client application can explicitly request data for large objects in
chunks, known as partial responses. The client application sends an HTTP HEAD request to obtain information
about the object. If the web API supports partial responses if should respond to the HEAD request with a response
message that contains an Accept-Ranges header and a Content-Length header that indicates the total size of the
object, but the body of the message should be empty. The client application can use this information to construct a
series of GET requests that specify a range of bytes to receive. The web API should return a response message with
HTTP status 206 (Partial Content), a Content-Length header that specifies the actual amount of data included in the
body of the response message, and a Content-Range header that indicates which part (such as bytes 4000 to 8000)
of the object this data represents.
HTTP HEAD requests and partial responses are described in more detail in API design.
Avoid sending unnecessary 100-Continue status messages in client applications
A client application that is about to send a large amount of data to a server may determine first whether the server
is actually willing to accept the request. Prior to sending the data, the client application can submit an HTTP request
with an Expect: 100-Continue header, a Content-Length header that indicates the size of the data, but an empty
message body. If the server is willing to handle the request, it should respond with a message that specifies the
HTTP status 100 (Continue). The client application can then proceed and send the complete request including the
data in the message body.
If you are hosting a service by using IIS, the HTTP.sys driver automatically detects and handles Expect: 100-Continue
headers before passing requests to your web application. This means that you are unlikely to see these headers in
your application code, and you can assume that IIS has already filtered any messages that it deems to be unfit or
too large.
If you are building client applications by using the .NET Framework, then all POST and PUT messages will first send
messages with Expect: 100-Continue headers by default. As with the server-side, the process is handled
transparently by the .NET Framework. However, this process results in each POST and PUT request causing two
round-trips to the server, even for small requests. If your application is not sending requests with large amounts of
data, you can disable this feature by using the ServicePointManager class to create ServicePoint objects in the
client application. A ServicePoint object handles the connections that the client makes to a server based on the
scheme and host fragments of URIs that identify resources on the server. You can then set the Expect100Continue
property of the ServicePoint object to false. All subsequent POST and PUT requests made by the client through a
URI that matches the scheme and host fragments of the ServicePoint object will be sent without Expect: 100-
Continue headers. The following code shows how to configure a ServicePoint object that configures all requests
sent to URIs with a scheme of http and a host of www.contoso.com .

Uri uri = new Uri("https://www.contoso.com/");

ServicePoint sp = ServicePointManager.FindServicePoint(uri);
sp.Expect100Continue = false;

You can also set the static Expect100Continue property of the ServicePointManager class to specify the default value
of this property for all subsequently created ServicePoint objects.
Support pagination for requests that may return large numbers of objects
If a collection contains a large number of resources, issuing a GET request to the corresponding URI could result in
significant processing on the server hosting the web API affecting performance, and generate a significant amount
of network traffic resulting in increased latency.
To handle these cases, the web API should support query strings that enable the client application to refine requests
or fetch data in more manageable, discrete blocks (or pages). The code below shows the GetAllOrders method in
the Orders controller. This method retrieves the details of orders. If this method was unconstrained, it could
conceivably return a large amount of data. The limit and offset parameters are intended to reduce the volume
of data to a smaller subset, in this case only the first 10 orders by default:
 public class OrdersController : ApiController
{
...
[Route("api/orders")]
[HttpGet]
public IEnumerable<Order> GetAllOrders(int limit=10, int offset=0)
{
// Find the number of orders specified by the limit parameter
// starting with the order specified by the offset parameter
var orders = ...
return orders;
}
...
}

A client application can issue a request to retrieve 30 orders starting at offset 50 by using the URI
https://www.adventure-works.com/api/orders?limit=30&offset=50 .

TIP
Avoid enabling client applications to specify query strings that result in a URI that is more than 2000 characters long. Many
web clients and servers cannot handle URIs that are this long.

Maintaining responsiveness, scalability, and availability

The same web API might be used by many client applications running anywhere in the world. It is important to
ensure that the web API is implemented to maintain responsiveness under a heavy load, to be scalable to support a
highly varying workload, and to guarantee availability for clients that perform business-critical operations.
Consider the following points when determining how to meet these requirements:
Provide asynchronous support for long-running requests
A request that might take a long time to process should be performed without blocking the client that submitted
the request. The web API can perform some initial checking to validate the request, initiate a separate task to
perform the work, and then return a response message with HTTP code 202 (Accepted). The task could run
asynchronously as part of the web API processing, or it could be offloaded to a background task.
The web API should also provide a mechanism to return the results of the processing to the client application. You
can achieve this by providing a polling mechanism for client applications to periodically query whether the
processing has finished and obtain the result, or enabling the web API to send a notification when the operation
has completed.
You can implement a simple polling mechanism by providing a polling URI that acts as a virtual resource using the
following approach:
1. The client application sends the initial request to the web API.
2. The web API stores information about the request in a table held in table storage or Microsoft Azure Cache, and
generates a unique key for this entry, possibly in the form of a GUID.
3. The web API initiates the processing as a separate task. The web API records the state of the task in the table as
Running.
4. The web API returns a response message with HTTP status code 202 (Accepted), and the GUID of the table entry
in the body of the message.
5. When the task has completed, the web API stores the results in the table, and sets the state of the task to
Complete. Note that if the task fails, the web API could also store information about the failure and set the status
to Failed.
6. While the task is running, the client can continue performing its own processing. It can periodically send a
 request to the URI /polling/{guid} where {guid} is the GUID returned in the 202 response message by the web
API.
7. The web API at the /polling/{guid} URI queries the state of the corresponding task in the table and returns a
response message with HTTP status code 200 (OK) containing this state (Running, Complete, or Failed). If the
task has completed or failed, the response message can also include the results of the processing or any
information available about the reason for the failure.
Options for implementing notifications include:
Using a notification hub to push asynchronous responses to client applications. For more information, see Send
notifications to specific users by using Azure Notification Hubs.
Using the Comet model to retain a persistent network connection between the client and the server hosting the
web API, and using this connection to push messages from the server back to the client. The MSDN magazine
article Building a Simple Comet Application in the Microsoft .NET Framework describes an example solution.
Using SignalR to push data in real time from the web server to the client over a persistent network connection.
SignalR is available for ASP.NET web applications as a NuGet package. You can find more information on the
ASP.NET SignalR website.
Ensure that each request is stateless
Each request should be considered atomic. There should be no dependencies between one request made by a client
application and any subsequent requests submitted by the same client. This approach assists in scalability;
instances of the web service can be deployed on a number of servers. Client requests can be directed at any of
these instances and the results should always be the same. It also improves availability for a similar reason; if a web
server fails requests can be routed to another instance (by using Azure Traffic Manager) while the server is
restarted with no ill effects on client applications.
Track clients and implement throttling to reduce the chances of DOS attacks
If a specific client makes a large number of requests within a given period of time it might monopolize the service
and affect the performance of other clients. To mitigate this issue, a web API can monitor calls from client
applications either by tracking the IP address of all incoming requests or by logging each authenticated access. You
can use this information to limit resource access. If a client exceeds a defined limit, the web API can return a
response message with status 503 (Service Unavailable) and include a Retry-After header that specifies when the
client can send the next request without it being declined. This strategy can help to reduce the chances of a Denial
Of Service (DOS) attack from a set of clients stalling the system.
Manage persistent HTTP connections carefully
The HTTP protocol supports persistent HTTP connections where they are available. The HTTP 1.0 specification
added the Connection:Keep-Alive header that enables a client application to indicate to the server that it can use the
same connection to send subsequent requests rather than opening new ones. The connection closes automatically
if the client does not reuse the connection within a period defined by the host. This behavior is the default in HTTP
1.1 as used by Azure services, so there is no need to include Keep-Alive headers in messages.
Keeping a connection open can help to improve responsiveness by reducing latency and network congestion, but it
can be detrimental to scalability by keeping unnecessary connections open for longer than required, limiting the
ability of other concurrent clients to connect. It can also affect battery life if the client application is running on a
mobile device; if the application only makes occasional requests to the server, maintaining an open connection can
cause the battery to drain more quickly. To ensure that a connection is not made persistent with HTTP 1.1, the client
can include a Connection:Close header with messages to override the default behavior. Similarly, if a server is
handling a very large number of clients it can include a Connection:Close header in response messages which
should close the connection and save server resources.
 NOTE
Persistent HTTP connections are a purely optional feature to reduce the network overhead associated with repeatedly
establishing a communications channel. Neither the web API nor the client application should depend on a persistent HTTP
connection being available. Do not use persistent HTTP connections to implement Comet-style notification systems; instead
you should use sockets (or web sockets if available) at the TCP layer. Finally, note Keep-Alive headers are of limited use if a
client application communicates with a server via a proxy; only the connection with the client and the proxy will be persistent.

Publishing and managing a web API

To make a web API available for client applications, the web API must be deployed to a host environment. This
environment is typically a web server, although it may be some other type of host process. You should consider the
following points when publishing a web API:
All requests must be authenticated and authorized, and the appropriate level of access control must be enforced.
A commercial web API might be subject to various quality guarantees concerning response times. It is important
to ensure that host environment is scalable if the load can vary significantly over time.
It may be necessary to meter requests for monetization purposes.
It might be necessary to regulate the flow of traffic to the web API, and implement throttling for specific clients
that have exhausted their quotas.
Regulatory requirements might mandate logging and auditing of all requests and responses.
To ensure availability, it may be necessary to monitor the health of the server hosting the web API and restart it
if necessary.
It is useful to be able to decouple these issues from the technical issues concerning the implementation of the web
API. For this reason, consider creating a façade, running as a separate process and that routes requests to the web
API. The façade can provide the management operations and forward validated requests to the web API. Using a
façade can also bring many functional advantages, including:
Acting as an integration point for multiple web APIs.
Transforming messages and translating communications protocols for clients built by using varying
technologies.
Caching requests and responses to reduce load on the server hosting the web API.

Testing a web API

A web API should be tested as thoroughly as any other piece of software. You should consider creating unit tests to
validate the functionality.
The nature of a web API brings its own additional requirements to verify that it operates correctly. You should pay
particular attention to the following aspects:
Test all routes to verify that they invoke the correct operations. Be especially aware of HTTP status code 405
(Method Not Allowed) being returned unexpectedly as this can indicate a mismatch between a route and the
HTTP methods (GET, POST, PUT, DELETE) that can be dispatched to that route.
Send HTTP requests to routes that do not support them, such as submitting a POST request to a specific
resource (POST requests should only be sent to resource collections). In these cases, the only valid response
should be status code 405 (Not Allowed).
Verify that all routes are protected properly and are subject to the appropriate authentication and
authorization checks.
 NOTE
Some aspects of security such as user authentication are most likely to be the responsibility of the host environment
rather than the web API, but it is still necessary to include security tests as part of the deployment process.

Test the exception handling performed by each operation and verify that an appropriate and meaningful
HTTP response is passed back to the client application.
Verify that request and response messages are well-formed. For example, if an HTTP POST request contains
the data for a new resource in x-www-form-urlencoded format, confirm that the corresponding operation
correctly parses the data, creates the resources, and returns a response containing the details of the new
resource, including the correct Location header.
Verify all links and URIs in response messages. For example, an HTTP POST message should return the URI
of the newly created resource. All HATEOAS links should be valid.
Ensure that each operation returns the correct status codes for different combinations of input. For example:
If a query is successful, it should return status code 200 (OK)
If a resource is not found, the operation should return HTTP status code 404 (Not Found).
If the client sends a request that successfully deletes a resource, the status code should be 204 (No
Content).
If the client sends a request that creates a new resource, the status code should be 201 (Created).
Watch out for unexpected response status codes in the 5xx range. These messages are usually reported by the host
server to indicate that it was unable to fulfill a valid request.
Test the different request header combinations that a client application can specify and ensure that the web
API returns the expected information in response messages.
Test query strings. If an operation can take optional parameters (such as pagination requests), test the
different combinations and order of parameters.
Verify that asynchronous operations complete successfully. If the web API supports streaming for requests
that return large binary objects (such as video or audio), ensure that client requests are not blocked while
the data is streamed. If the web API implements polling for long-running data modification operations, verify
that the operations report their status correctly as they proceed.
You should also create and run performance tests to check that the web API operates satisfactorily under duress.
You can build a web performance and load test project by using Visual Studio Ultimate. For more information, see
Run performance tests on an application before a release.

Using Azure API Management

On Azure, consider using Azure API Management to publish and manage a web API. Using this facility, you can
generate a service that acts as a façade for one or more web APIs. The service is itself a scalable web service that
you can create and configure by using the Azure portal. You can use this service to publish and manage a web API
as follows:
1. Deploy the web API to a website, Azure cloud service, or Azure virtual machine.
2. Connect the API management service to the web API. Requests sent to the URL of the management API are
mapped to URIs in the web API. The same API management service can route requests to more than one
web API. This enables you to aggregate multiple web APIs into a single management service. Similarly, the
same web API can be referenced from more than one API management service if you need to restrict or
partition the functionality available to different applications.
 NOTE
The URIs in HATEOAS links generated as part of the response for HTTP GET requests should reference the URL of the
API management service and not the web server hosting the web API.

3. For each web API, specify the HTTP operations that the web API exposes together with any optional
parameters that an operation can take as input. You can also configure whether the API management service
should cache the response received from the web API to optimize repeated requests for the same data.
Record the details of the HTTP responses that each operation can generate. This information is used to
generate documentation for developers, so it is important that it is accurate and complete.
You can either define operations manually using the wizards provided by the Azure portal, or you can
import them from a file containing the definitions in WADL or Swagger format.
4. Configure the security settings for communications between the API management service and the web
server hosting the web API. The API management service currently supports Basic authentication and
mutual authentication using certificates, and OAuth 2.0 user authorization.
5. Create a product. A product is the unit of publication; you add the web APIs that you previously connected to
the management service to the product. When the product is published, the web APIs become available to
developers.

NOTE
Prior to publishing a product, you can also define user-groups that can access the product and add users to these
groups. This gives you control over the developers and applications that can use the web API. If a web API is subject
to approval, prior to being able to access it a developer must send a request to the product administrator. The
administrator can grant or deny access to the developer. Existing developers can also be blocked if circumstances
change.

6. Configure policies for each web API. Policies govern aspects such as whether cross-domain calls should be
allowed, how to authenticate clients, whether to convert between XML and JSON data formats transparently,
whether to restrict calls from a given IP range, usage quotas, and whether to limit the call rate. Policies can
be applied globally across the entire product, for a single web API in a product, or for individual operations
in a web API.
For more information, see the API Management documentation.

TIP
Azure provides the Azure Traffic Manager which enables you to implement failover and load-balancing, and reduce latency
across multiple instances of a web site hosted in different geographic locations. You can use Azure Traffic Manager in
conjunction with the API Management Service; the API Management Service can route requests to instances of a web site
through Azure Traffic Manager. For more information, see Traffic Manager routing methods.
In this structure, if you are using custom DNS names for your web sites, you should configure the appropriate CNAME record
for each web site to point to the DNS name of the Azure Traffic Manager web site.

Supporting client-side developers

Developers constructing client applications typically require information on how to access the web API, and
documentation concerning the parameters, data types, return types, and return codes that describe the different
requests and responses between the web service and the client application.
Document the REST operations for a web API
The Azure API Management Service includes a developer portal that describes the REST operations exposed by a
web API. When a product has been published it appears on this portal. Developers can use this portal to sign up for
access; the administrator can then approve or deny the request. If the developer is approved, they are assigned a
subscription key that is used to authenticate calls from the client applications that they develop. This key must be
provided with each web API call otherwise it will be rejected.
This portal also provides:
Documentation for the product, listing the operations that it exposes, the parameters required, and the different
responses that can be returned. Note that this information is generated from the details provided in step 3 in
the list in the Publishing a web API by using the Microsoft Azure API Management Service section.
Code snippets that show how to invoke operations from several languages, including JavaScript, C#, Java, Ruby,
Python, and PHP.
A developers' console that enables a developer to send an HTTP request to test each operation in the product
and view the results.
A page where the developer can report any issues or problems found.
The Azure portal enables you to customize the developer portal to change the styling and layout to match the
branding of your organization.
Implement a client SDK
Building a client application that invokes REST requests to access a web API requires writing a significant amount of
code to construct each request and format it appropriately, send the request to the server hosting the web service,
and parse the response to work out whether the request succeeded or failed and extract any data returned. To
insulate the client application from these concerns, you can provide an SDK that wraps the REST interface and
abstracts these low-level details inside a more functional set of methods. A client application uses these methods,
which transparently convert calls into REST requests and then convert the responses back into method return
values. This is a common technique that is implemented by many services, including the Azure SDK.
Creating a client-side SDK is a considerable undertaking as it has to be implemented consistently and tested
carefully. However, much of this process can be made mechanical, and many vendors supply tools that can
automate many of these tasks.

Monitoring a web API

Depending on how you have published and deployed your web API you can monitor the web API directly, or you
can gather usage and health information by analyzing the traffic that passes through the API Management service.
Monitoring a web API directly
If you have implemented your web API by using the ASP.NET Web API template (either as a Web API project or as a
Web role in an Azure cloud service) and Visual Studio 2013, you can gather availability, performance, and usage
data by using ASP.NET Application Insights. Application Insights is a package that transparently tracks and records
information about requests and responses when the web API is deployed to the cloud; once the package is installed
and configured, you don't need to amend any code in your web API to use it. When you deploy the web API to an
Azure web site, all traffic is examined and the following statistics are gathered:
Server response time.
Number of server requests and the details of each request.
The top slowest requests in terms of average response time.
The details of any failed requests.
The number of sessions initiated by different browsers and user agents.
The most frequently viewed pages (primarily useful for web applications rather than web APIs).
The different user roles accessing the web API.
You can view this data in real time in the Azure portal. You can also create web tests that monitor the health of the
web API. A web test sends a periodic request to a specified URI in the web API and captures the response. You can
specify the definition of a successful response (such as HTTP status code 200), and if the request does not return
this response you can arrange for an alert to be sent to an administrator. If necessary, the administrator can restart
the server hosting the web API if it has failed.
For more information, see Application Insights - Get started with ASP.NET.
Monitoring a web API through the API Management Service
If you have published your web API by using the API Management service, the API Management page on the Azure
portal contains a dashboard that enables you to view the overall performance of the service. The Analytics page
enables you to drill down into the details of how the product is being used. This page contains the following tabs:
Usage . This tab provides information about the number of API calls made and the bandwidth used to handle
these calls over time. You can filter usage details by product, API, and operation.
Health . This tab enables you to view the outcome of API requests (the HTTP status codes returned), the
effectiveness of the caching policy, the API response time, and the service response time. Again, you can filter
health data by product, API, and operation.
Activity . This tab provides a text summary of the numbers of successful calls, failed calls, blocked calls, average
response time, and response times for each product, web API, and operation. This page also lists the number of
calls made by each developer.
At a glance . This tab displays a summary of the performance data, including the developers responsible for
making the most API calls, and the products, web APIs, and operations that received these calls.
You can use this information to determine whether a particular web API or operation is causing a bottleneck, and if
necessary scale the host environment and add more servers. You can also ascertain whether one or more
applications are using a disproportionate volume of resources and apply the appropriate policies to set quotas and
limit call rates.

NOTE
You can change the details for a published product, and the changes are applied immediately. For example, you can add or
remove an operation from a web API without requiring that you republish the product that contains the web API.

More information
ASP.NET Web API OData contains examples and further information on implementing an OData web API by
using ASP.NET.
Introducing batch support in Web API and Web API OData describes how to implement batch operations in a
web API by using OData.
Idempotency patterns on Jonathan Oliver's blog provides an overview of idempotency and how it relates to
data management operations.
Status code definitions on the W3C website contains a full list of HTTP status codes and their descriptions.
Run background tasks with WebJobs provides information and examples on using WebJobs to perform
background operations.
Azure Notification Hubs notify users shows how to use an Azure Notification Hub to push asynchronous
responses to client applications.
API Management describes how to publish a product that provides controlled and secure access to a web API.
Azure API Management REST API reference describes how to use the API Management REST API to build custom
management applications.
Traffic Manager routing methods summarizes how Azure Traffic Manager can be used to load-balance requests
across multiple instances of a website hosting a web API.
Application Insights - Get started with ASP.NET provides detailed information on installing and configuring
Application Insights in an ASP.NET Web API project.
 Autoscaling
12/18/2020 • 15 minutes to read • Edit Online

Autoscaling is the process of dynamically allocating resources to match performance requirements. As the
volume of work grows, an application may need additional resources to maintain the desired performance levels
and satisfy service-level agreements (SLAs). As demand slackens and the additional resources are no longer
needed, they can be de-allocated to minimize costs.
Autoscaling takes advantage of the elasticity of cloud-hosted environments while easing management overhead.
It reduces the need for an operator to continually monitor the performance of a system and make decisions about
adding or removing resources.
There are two main ways that an application can scale:
Ver tical scaling , also called scaling up and down, means changing the capacity of a resource. For
example, you could move an application to a larger VM size. Vertical scaling often requires making the
system temporarily unavailable while it is being redeployed. Therefore, it's less common to automate
vertical scaling.
Horizontal scaling , also called scaling out and in, means adding or removing instances of a resource. The
application continues running without interruption as new resources are provisioned. When the
provisioning process is complete, the solution is deployed on these additional resources. If demand drops,
the additional resources can be shut down cleanly and deallocated.
Many cloud-based systems, including Microsoft Azure, support automatic horizontal scaling. The rest of this
article focuses on horizontal scaling.

NOTE
Autoscaling mostly applies to compute resources. While it's possible to horizontally scale a database or message queue, this
usually involves data partitioning, which is generally not automated.

Overview
An autoscaling strategy typically involves the following pieces:
Instrumentation and monitoring systems at the application, service, and infrastructure levels. These systems
capture key metrics, such as response times, queue lengths, CPU utilization, and memory usage.
Decision-making logic that evaluates these metrics against predefined thresholds or schedules, and decides
whether to scale.
Components that scale the system.
Testing, monitoring, and tuning of the autoscaling strategy to ensure that it functions as expected.
Azure provides built-in autoscaling mechanisms that address common scenarios. If a particular service or
technology does not have built-in autoscaling functionality, or if you have specific autoscaling requirements
beyond its capabilities, you might consider a custom implementation. A custom implementation would collect
operational and system metrics, analyze the metrics, and then scale resources accordingly.

Configure autoscaling for an Azure solution

Azure provides built-in autoscaling for most compute options.
 Azure Vir tual Machines autoscale via virtual machine scale sets, which manage a set of Azure virtual
machines as a group. See How to use automatic scaling and virtual machine scale sets.
Ser vice Fabric also supports autoscaling through virtual machine scale sets. Every node type in a Service
Fabric cluster is set up as a separate virtual machine scale set. That way, each node type can be scaled in or
out independently. See Scale a Service Fabric cluster in or out using autoscale rules.
Azure App Ser vice has built-in autoscaling. Autoscale settings apply to all of the apps within an App
Service. See Scale instance count manually or automatically.
Azure Cloud Ser vices has built-in autoscaling at the role level. See How to configure auto scaling for a
Cloud Service in the portal.
These compute options all use Azure Monitor autoscale to provide a common set of autoscaling functionality.
Azure Functions differs from the previous compute options, because you don't need to configure any
autoscale rules. Instead, Azure Functions automatically allocates compute power when your code is running,
scaling out as necessary to handle load. For more information, see Choose the correct hosting plan for Azure
Functions.
Finally, a custom autoscaling solution can sometimes be useful. For example, you could use Azure diagnostics and
application-based metrics, along with custom code to monitor and export the application metrics. Then you could
define custom rules based on these metrics, and use Resource Manager REST APIs to trigger autoscaling.
However, a custom solution is not simple to implement, and should be considered only if none of the previous
approaches can fulfill your requirements.
Use the built-in autoscaling features of the platform, if they meet your requirements. If not, carefully consider
whether you really need more complex scaling features. Examples of additional requirements may include more
granularity of control, different ways to detect trigger events for scaling, scaling across subscriptions, and scaling
other types of resources.

Use Azure Monitor autoscale

Azure Monitor autoscale provide a common set of autoscaling functionality for virtual machine scale sets, Azure
App Service, and Azure Cloud Service. Scaling can be performed on a schedule, or based on a runtime metric,
such as CPU or memory usage.
Examples:
Scale out to 10 instances on weekdays, and scale in to 4 instances on Saturday and Sunday.
Scale out by one instance if average CPU usage is above 70%, and scale in by one instance if CPU usage falls
below 50%.
Scale out by one instance if the number of messages in a queue exceeds a certain threshold.
Scale up the resource when load increases to ensure availability. Similarly, at times of low usage, scale down, so
you can optimize cost. Always use a scale-out and scale-in rule combination. Otherwise, the autoscaling takes
place only in one direction until it reaches the threshold (maximum or minimum instance counts) set in the
profile.
Select a default instance count that's safe for your workload. It's scaled based on that value if maximum or
minimum instance counts are not set.
For a list of built-in metrics, see Azure Monitor autoscaling common metrics. You can also implement custom
metrics by using Application Insights.
You can configure autoscaling by using PowerShell, the Azure CLI, an Azure Resource Manager template, or the
Azure portal. For more detailed control, use the Azure Resource Manager REST API. The Azure Monitoring Service
Management Library and the Microsoft Insights Library (in preview) are SDKs that allow collecting metrics from
different resources, and perform autoscaling by making use of the REST APIs. For resources where Azure
Resource Manager support isn't available, or if you are using Azure Cloud Services, the Service Management
REST API can be used for autoscaling. In all other cases, use Azure Resource Manager.
Consider the following points when using Azure autoscale:
Consider whether you can predict the load on the application accurately enough to use scheduled
autoscaling, adding and removing instances to meet anticipated peaks in demand. If this isn't possible, use
reactive autoscaling based on runtime metrics, in order to handle unpredictable changes in demand.
Typically, you can combine these approaches. For example, create a strategy that adds resources based on a
schedule of the times when you know the application is busiest. This helps to ensure that capacity is
available when required, without any delay from starting new instances. For each scheduled rule, define
metrics that allow reactive autoscaling during that period to ensure that the application can handle
sustained but unpredictable peaks in demand.
It's often difficult to understand the relationship between metrics and capacity requirements, especially
when an application is initially deployed. Provision a little extra capacity at the beginning, and then monitor
and tune the autoscaling rules to bring the capacity closer to the actual load.
Configure the autoscaling rules, and then monitor the performance of your application over time. Use the
results of this monitoring to adjust the way in which the system scales if necessary. However, keep in mind
that autoscaling is not an instantaneous process. It takes time to react to a metric such as average CPU
utilization exceeding (or falling below) a specified threshold.
Autoscaling rules that use a detection mechanism based on a measured trigger attribute (such as CPU
usage or queue length) use an aggregated value over time, rather than instantaneous values, to trigger an
autoscaling action. By default, the aggregate is an average of the values. This prevents the system from
reacting too quickly, or causing rapid oscillation. It also allows time for new instances that are automatically
started to settle into running mode, preventing additional autoscaling actions from occurring while the
new instances are starting up. For Azure Cloud Services and Azure Virtual Machines, the default period for
the aggregation is 45 minutes, so it can take up to this period of time for the metric to trigger autoscaling
in response to spikes in demand. You can change the aggregation period by using the SDK, but periods of
less than 25 minutes may cause unpredictable results. For Web Apps, the averaging period is much shorter,
allowing new instances to be available in about five minutes after a change to the average trigger measure.
Avoid flapping where scale-in and scale-out actions continually go back and forth. Suppose there are two
instances, and upper limit is 80% CPU, lower limit is 60%. When the load is at 85%, another instance is
added. After some time, the load decreases to 60%. Before scaling in, the autoscale service calculates the
distribution of total load (of three instances) when an instance is removed, taking it to 90%. This means it
would have to scale out again immediately. So, it skips scaling-in and you might never see the expected
scaling results.
The flapping situation can be controlled by choosing an adequate margin between the scale-out and scale-
in thresholds.
Manual scaling is reset by maximum and minimum number of instances used for autoscaling. If you
manually update the instance count to a value higher or lower than the maximum value, the autoscale
engine automatically scales back to the minimum (if lower) or the maximum (if higher). For example, you
set the range between 3 and 6. If you have one running instance, the autoscale engine scales to three
instances on its next run. Likewise, if you manually set the scale to eight instances, on the next run
autoscale will scale it back to six instances on its next run. Manual scaling is temporary unless you reset the
autoscale rules as well.
The autoscale engine processes only one profile at a time. If a condition is not met, then it checks for the
next profile. Keep key metrics out of the default profile because that profile is checked last. Within a profile,
 you can have multiple rules. On scale-out, autoscale runs if any rule is met. On scale-in, autoscale require
all rules to be met.
For details about how Azure Monitor scales, see Best practices for Autoscale.
If you configure autoscaling using the SDK rather than the portal, you can specify a more detailed schedule
during which the rules are active. You can also create your own metrics and use them with or without any
of the existing ones in your autoscaling rules. For example, you may wish to use alternative counters, such
as the number of requests per second or the average memory availability, or use custom counters to
measure specific business processes.
When autoscaling Service Fabric, the node types in your cluster are made of virtual machine scale sets at
the back end, so you need to set up autoscale rules for each node type. Take into account the number of
nodes that you must have before you set up autoscaling. The minimum number of nodes that you must
have for the primary node type is driven by the reliability level you have chosen. For more information, see
scale a Service Fabric cluster in or out using autoscale rules.
You can use the portal to link resources such as SQL Database instances and queues to a Cloud Service
instance. This allows you to more easily access the separate manual and automatic scaling configuration
options for each of the linked resources. For more information, see How to: Link a resource to a cloud
service.
When you configure multiple policies and rules, they could conflict with each other. Autoscale uses the
following conflict resolution rules to ensure that there is always a sufficient number of instances running:
Scale-out operations always take precedence over scale-in operations.
When scale-out operations conflict, the rule that initiates the largest increase in the number of instances
takes precedence.
When scale in operations conflict, the rule that initiates the smallest decrease in the number of instances
takes precedence.
In an App Service Environment, any worker pool or front-end metrics can be used to define autoscale rules.
For more information, see Autoscaling and App Service Environment.

Application design considerations

Autoscaling isn't an instant solution. Simply adding resources to a system or running more instances of a process
doesn't guarantee that the performance of the system will improve. Consider the following points when designing
an autoscaling strategy:
The system must be designed to be horizontally scalable. Avoid making assumptions about instance
affinity; do not design solutions that require that the code is always running in a specific instance of a
process. When scaling a cloud service or web site horizontally, don't assume that a series of requests from
the same source will always be routed to the same instance. For the same reason, design services to be
stateless to avoid requiring a series of requests from an application to always be routed to the same
instance of a service. When designing a service that reads messages from a queue and processes them,
don't make any assumptions about which instance of the service handles a specific message. Autoscaling
could start additional instances of a service as the queue length grows. The Competing Consumers pattern
describes how to handle this scenario.
If the solution implements a long-running task, design this task to support both scaling out and scaling in.
Without due care, such a task could prevent an instance of a process from being shut down cleanly when
the system scales in, or it could lose data if the process is forcibly terminated. Ideally, refactor a long-
running task and break up the processing that it performs into smaller, discrete chunks. The Pipes and
Filters pattern provides an example of how you can achieve this.
Alternatively, you can implement a checkpoint mechanism that records state information about the task at
 regular intervals, and save this state in durable storage that can be accessed by any instance of the process
running the task. In this way, if the process is shut down, the work that it was performing can be resumed
from the last checkpoint by using another instance.
When background tasks run on separate compute instances, such as in worker roles of a cloud-services–
hosted application, you may need to scale different parts of the application using different scaling policies.
For example, you may need to deploy additional user interface (UI) compute instances without increasing
the number of background compute instances, or the opposite of this. If you offer different levels of service
(such as basic and premium service packages), you may need to scale out the compute resources for
premium service packages more aggressively than those for basic service packages in order to meet SLAs.
Consider using the length of the queue over which UI and background compute instances communicate as
a criterion for your autoscaling strategy. This is the best indicator of an imbalance or difference between
the current load and the processing capacity of the background task.
If you base your autoscaling strategy on counters that measure business processes, such as the number of
orders placed per hour or the average execution time of a complex transaction, ensure that you fully
understand the relationship between the results from these types of counters and the actual compute
capacity requirements. It may be necessary to scale more than one component or compute unit in
response to changes in business process counters.
To prevent a system from attempting to scale out excessively, and to avoid the costs associated with
running many thousands of instances, consider limiting the maximum number of instances that can be
automatically added. Most autoscaling mechanisms allow you to specify the minimum and maximum
number of instances for a rule. In addition, consider gracefully degrading the functionality that the system
provides if the maximum number of instances have been deployed, and the system is still overloaded.
Keep in mind that autoscaling might not be the most appropriate mechanism to handle a sudden burst in
workload. It takes time to provision and start new instances of a service or add resources to a system, and
the peak demand may have passed by the time these additional resources have been made available. In
this scenario, it may be better to throttle the service. For more information, see the Throttling pattern.
Conversely, if you do need the capacity to process all requests when the volume fluctuates rapidly, and cost
isn't a major contributing factor, consider using an aggressive autoscaling strategy that starts additional
instances more quickly. You can also use a scheduled policy that starts a sufficient number of instances to
meet the maximum load before that load is expected.
The autoscaling mechanism should monitor the autoscaling process, and log the details of each
autoscaling event (what triggered it, what resources were added or removed, and when). If you create a
custom autoscaling mechanism, ensure that it incorporates this capability. Analyze the information to help
measure the effectiveness of the autoscaling strategy, and tune it if necessary. You can tune both in the
short term, as the usage patterns become more obvious, and over the long term, as the business expands
or the requirements of the application evolve. If an application reaches the upper limit defined for
autoscaling, the mechanism might also alert an operator who could manually start additional resources if
necessary. Note that under these circumstances the operator may also be responsible for manually
removing these resources after the workload eases.

Related patterns and guidance

The following patterns and guidance may also be relevant to your scenario when implementing autoscaling:
Throttling pattern. This pattern describes how an application can continue to function and meet SLAs when
an increase in demand places an extreme load on resources. Throttling can be used with autoscaling to
prevent a system from being overwhelmed while the system scales out.
Competing Consumers pattern. This pattern describes how to implement a pool of service instances that
can handle messages from any application instance. Autoscaling can be used to start and stop service
instances to match the anticipated workload. This approach enables a system to process multiple messages
concurrently to optimize throughput, improve scalability and availability, and balance the workload.
Monitoring and diagnostics. Instrumentation and telemetry are vital for gathering the information that can
drive the autoscaling process.
 Background jobs
12/18/2020 • 23 minutes to read • Edit Online

Many types of applications require background tasks that run independently of the user interface (UI). Examples
include batch jobs, intensive processing tasks, and long-running processes such as workflows. Background jobs
can be executed without requiring user interaction--the application can start the job and then continue to process
interactive requests from users. This can help to minimize the load on the application UI, which can improve
availability and reduce interactive response times.
For example, if an application is required to generate thumbnails of images that are uploaded by users, it can do
this as a background job and save the thumbnail to storage when it is complete--without the user needing to wait
for the process to be completed. In the same way, a user placing an order can initiate a background workflow that
processes the order, while the UI allows the user to continue browsing the web app. When the background job is
complete, it can update the stored orders data and send an email to the user that confirms the order.
When you consider whether to implement a task as a background job, the main criteria is whether the task can run
without user interaction and without the UI needing to wait for the job to be completed. Tasks that require the user
or the UI to wait while they are completed might not be appropriate as background jobs.

Types of background jobs

Background jobs typically include one or more of the following types of jobs:
CPU-intensive jobs, such as mathematical calculations or structural model analysis.
I/O-intensive jobs, such as executing a series of storage transactions or indexing files.
Batch jobs, such as nightly data updates or scheduled processing.
Long-running workflows, such as order fulfillment, or provisioning services and systems.
Sensitive-data processing where the task is handed off to a more secure location for processing. For example,
you might not want to process sensitive data within a web app. Instead, you might use a pattern such as the
Gatekeeper pattern to transfer the data to an isolated background process that has access to protected storage.

Triggers
Background jobs can be initiated in several different ways. They fall into one of the following categories:
Event-driven triggers . The task is started in response to an event, typically an action taken by a user or a step
in a workflow.
Schedule-driven triggers . The task is invoked on a schedule based on a timer. This might be a recurring
schedule or a one-off invocation that is specified for a later time.
Event-driven triggers
Event-driven invocation uses a trigger to start the background task. Examples of using event-driven triggers
include:
The UI or another job places a message in a queue. The message contains data about an action that has taken
place, such as the user placing an order. The background task listens on this queue and detects the arrival of a
new message. It reads the message and uses the data in it as the input to the background job.
The UI or another job saves or updates a value in storage. The background task monitors the storage and
detects changes. It reads the data and uses it as the input to the background job.
The UI or another job makes a request to an endpoint, such as an HTTPS URI, or an API that is exposed as a web
 service. It passes the data that is required to complete the background task as part of the request. The endpoint
or web service invokes the background task, which uses the data as its input.
Typical examples of tasks that are suited to event-driven invocation include image processing, workflows, sending
information to remote services, sending email messages, and provisioning new users in multitenant applications.
Schedule -driven triggers
Schedule-driven invocation uses a timer to start the background task. Examples of using schedule-driven triggers
include:
A timer that is running locally within the application or as part of the application's operating system invokes a
background task on a regular basis.
A timer that is running in a different application, such as Azure Logic Apps, sends a request to an API or web
service on a regular basis. The API or web service invokes the background task.
A separate process or application starts a timer that causes the background task to be invoked once after a
specified time delay, or at a specific time.
Typical examples of tasks that are suited to schedule-driven invocation include batch-processing routines (such as
updating related-products lists for users based on their recent behavior), routine data processing tasks (such as
updating indexes or generating accumulated results), data analysis for daily reports, data retention cleanup, and
data consistency checks.
If you use a schedule-driven task that must run as a single instance, be aware of the following:
If the compute instance that is running the scheduler (such as a virtual machine using Windows scheduled
tasks) is scaled, you will have multiple instances of the scheduler running. These could start multiple instances
of the task.
If tasks run for longer than the period between scheduler events, the scheduler may start another instance of
the task while the previous one is still running.

Returning results
Background jobs execute asynchronously in a separate process, or even in a separate location, from the UI or the
process that invoked the background task. Ideally, background tasks are "fire and forget" operations, and their
execution progress has no impact on the UI or the calling process. This means that the calling process does not
wait for completion of the tasks. Therefore, it cannot automatically detect when the task ends.
If you require a background task to communicate with the calling task to indicate progress or completion, you
must implement a mechanism for this. Some examples are:
Write a status indicator value to storage that is accessible to the UI or caller task, which can monitor or check
this value when required. Other data that the background task must return to the caller can be placed into the
same storage.
Establish a reply queue that the UI or caller listens on. The background task can send messages to the queue
that indicate status and completion. Data that the background task must return to the caller can be placed into
the messages. If you are using Azure Service Bus, you can use the ReplyTo and CorrelationId properties to
implement this capability.
Expose an API or endpoint from the background task that the UI or caller can access to obtain status
information. Data that the background task must return to the caller can be included in the response.
Have the background task call back to the UI or caller through an API to indicate status at predefined points or
on completion. This might be through events raised locally or through a publish-and-subscribe mechanism.
Data that the background task must return to the caller can be included in the request or event payload.

Hosting environment
You can host background tasks by using a range of different Azure platform services:
Azure Web Apps and WebJobs . You can use WebJobs to execute custom jobs based on a range of different
types of scripts or executable programs within the context of a web app.
Azure Vir tual Machines . If you have a Windows service or want to use the Windows Task Scheduler, it is
common to host your background tasks within a dedicated virtual machine.
Azure Batch . Batch is a platform service that schedules compute-intensive work to run on a managed
collection of virtual machines. It can automatically scale compute resources.
Azure Kubernetes Ser vice (AKS). Azure Kubernetes Service provides a managed hosting environment for
Kubernetes on Azure.
The following sections describe each of these options in more detail, and include considerations to help you
choose the appropriate option.
Azure Web Apps and WebJobs
You can use Azure WebJobs to execute custom jobs as background tasks within an Azure Web App. WebJobs run
within the context of your web app as a continuous process. WebJobs also run in response to a trigger event from
Azure Logic Apps or external factors, such as changes to storage blobs and message queues. Jobs can be started
and stopped on demand, and shut down gracefully. If a continuously running WebJob fails, it is automatically
restarted. Retry and error actions are configurable.
When you configure a WebJob:
If you want the job to respond to an event-driven trigger, you should configure it as Run continuously . The
script or program is stored in the folder named site/wwwroot/app_data/jobs/continuous.
If you want the job to respond to a schedule-driven trigger, you should configure it as Run on a schedule . The
script or program is stored in the folder named site/wwwroot/app_data/jobs/triggered.
If you choose the Run on demand option when you configure a job, it will execute the same code as the Run
on a schedule option when you start it.
Azure WebJobs run within the sandbox of the web app. This means that they can access environment variables and
share information, such as connection strings, with the web app. The job has access to the unique identifier of the
machine that is running the job. The connection string named AzureWebJobsStorage provides access to Azure
storage queues, blobs, and tables for application data, and access to Service Bus for messaging and
communication. The connection string named AzureWebJobsDashboard provides access to the job action log
files.
Azure WebJobs have the following characteristics:
Security : WebJobs are protected by the deployment credentials of the web app.
Suppor ted file types : You can define WebJobs by using command scripts (.cmd), batch files (.bat), PowerShell
scripts (.ps1), bash shell scripts (.sh), PHP scripts (.php), Python scripts (.py), JavaScript code (.js), and executable
programs (.exe, .jar, and more).
Deployment : You can deploy scripts and executables by using the Azure portal, by using Visual Studio, by
using the Azure WebJobs SDK, or by copying them directly to the following locations:
For triggered execution: site/wwwroot/app_data/jobs/triggered/{ job name}
For continuous execution: site/wwwroot/app_data/jobs/continuous/{ job name}
Logging : Console.Out is treated (marked) as INFO. Console.Error is treated as ERROR. You can access
monitoring and diagnostics information by using the Azure portal. You can download log files directly from the
site. They are saved in the following locations:
For triggered execution: Vfs/data/jobs/triggered/jobName
For continuous execution: Vfs/data/jobs/continuous/jobName
Configuration : You can configure WebJobs by using the portal, the REST API, and PowerShell. You can use a
configuration file named settings.job in the same root directory as the job script to provide configuration
 information for a job. For example:
{ "stopping_wait_time": 60 }
{ "is_singleton": true }
Considerations
By default, WebJobs scale with the web app. However, you can configure jobs to run on single instance by
setting the is_singleton configuration property to true . Single instance WebJobs are useful for tasks that you
do not want to scale or run as simultaneous multiple instances, such as reindexing, data analysis, and similar
tasks.
To minimize the impact of jobs on the performance of the web app, consider creating an empty Azure Web App
instance in a new App Service plan to host long-running or resource-intensive WebJobs.
Azure Virtual Machines
Background tasks might be implemented in a way that prevents them from being deployed to Azure Web Apps, or
these options might not be convenient. Typical examples are Windows services, and third-party utilities and
executable programs. Another example might be programs written for an execution environment that is different
than that hosting the application. For example, it might be a Unix or Linux program that you want to execute from
a Windows or .NET application. You can choose from a range of operating systems for an Azure virtual machine,
and run your service or executable on that virtual machine.
To help you choose when to use Virtual Machines, see Azure App Services, Cloud Services and Virtual Machines
comparison. For information about the options for Virtual Machines, see Sizes for Windows virtual machines in
Azure. For more information about the operating systems and prebuilt images that are available for Virtual
Machines, see Azure Virtual Machines Marketplace.
To initiate the background task in a separate virtual machine, you have a range of options:
You can execute the task on demand directly from your application by sending a request to an endpoint that the
task exposes. This passes in any data that the task requires. This endpoint invokes the task.
You can configure the task to run on a schedule by using a scheduler or timer that is available in your chosen
operating system. For example, on Windows you can use Windows Task Scheduler to execute scripts and tasks.
Or, if you have SQL Server installed on the virtual machine, you can use the SQL Server Agent to execute scripts
and tasks.
You can use Azure Logic Apps to initiate the task by adding a message to a queue that the task listens on, or by
sending a request to an API that the task exposes.
See the earlier section Triggers for more information about how you can initiate background tasks.
Considerations
Consider the following points when you are deciding whether to deploy background tasks in an Azure virtual
machine:
Hosting background tasks in a separate Azure virtual machine provides flexibility and allows precise control
over initiation, execution, scheduling, and resource allocation. However, it will increase runtime cost if a virtual
machine must be deployed just to run background tasks.
There is no facility to monitor the tasks in the Azure portal and no automated restart capability for failed tasks--
although you can monitor the basic status of the virtual machine and manage it by using the Azure Resource
Manager Cmdlets. However, there are no facilities to control processes and threads in compute nodes. Typically,
using a virtual machine will require additional effort to implement a mechanism that collects data from
instrumentation in the task, and from the operating system in the virtual machine. One solution that might be
appropriate is to use the System Center Management Pack for Azure.
You might consider creating monitoring probes that are exposed through HTTP endpoints. The code for these
probes could perform health checks, collect operational information and statistics--or collate error information
and return it to a management application. For more information, see the Health Endpoint Monitoring pattern.
For more information, see:
Virtual Machines
Azure Virtual Machines FAQ
Azure Batch
Consider Azure Batch if you need to run large, parallel high-performance computing (HPC) workloads across tens,
hundreds, or thousands of VMs.
The Batch service provisions the VMs, assign tasks to the VMs, runs the tasks, and monitors the progress. Batch
can automatically scale out the VMs in response to the workload. Batch also provides job scheduling. Azure Batch
supports both Linux and Windows VMs.
Considerations
Batch works well with intrinsically parallel workloads. It can also perform parallel calculations with a reduce step at
the end, or run Message Passing Interface (MPI) applications for parallel tasks that require message passing
between nodes.
An Azure Batch job runs on a pool of nodes (VMs). One approach is to allocate a pool only when needed and then
delete it after the job completes. This maximizes utilization, because nodes are not idle, but the job must wait for
nodes to be allocated. Alternatively, you can create a pool ahead of time. That approach minimizes the time that it
takes for a job to start, but can result in having nodes that sit idle. For more information, see Pool and compute
node lifetime.
For more information, see:
What is Azure Batch?
Develop large-scale parallel compute solutions with Batch
Batch and HPC solutions for large-scale computing workloads
Azure Kubernetes Service
Azure Kubernetes Service (AKS) manages your hosted Kubernetes environment, which makes it easy to deploy and
manage containerized applications.
Containers can be useful for running background jobs. Some of the benefits include:
Containers support high-density hosting. You can isolate a background task in a container, while placing
multiple containers in each VM.
The container orchestrator handles internal load balancing, configuring the internal network, and other
configuration tasks.
Containers can be started and stopped as needed.
Azure Container Registry allows you to register your containers inside Azure boundaries. This comes with
security, privacy, and proximity benefits.
Considerations
Requires an understanding of how to use a container orchestrator. Depending on the skill set of your DevOps
team, this may or may not be an issue.
For more information, see:
Overview of containers in Azure
Introduction to private Docker container registries

Partitioning
If you decide to include background tasks within an existing compute instance, you must consider how this will
affect the quality attributes of the compute instance and the background task itself. These factors will help you to
decide whether to colocate the tasks with the existing compute instance or separate them out into a separate
compute instance:
Availability : Background tasks might not need to have the same level of availability as other parts of the
application, in particular the UI and other parts that are directly involved in user interaction. Background
tasks might be more tolerant of latency, retried connection failures, and other factors that affect availability
because the operations can be queued. However, there must be sufficient capacity to prevent the backup of
requests that could block queues and affect the application as a whole.
Scalability : Background tasks are likely to have a different scalability requirement than the UI and the
interactive parts of the application. Scaling the UI might be necessary to meet peaks in demand, while
outstanding background tasks might be completed during less busy times by fewer compute instances.
Resiliency : Failure of a compute instance that just hosts background tasks might not fatally affect the
application as a whole if the requests for these tasks can be queued or postponed until the task is available
again. If the compute instance and/or tasks can be restarted within an appropriate interval, users of the
application might not be affected.
Security : Background tasks might have different security requirements or restrictions than the UI or other
parts of the application. By using a separate compute instance, you can specify a different security
environment for the tasks. You can also use patterns such as Gatekeeper to isolate the background compute
instances from the UI in order to maximize security and separation.
Performance : You can choose the type of compute instance for background tasks to specifically match the
performance requirements of the tasks. This might mean using a less expensive compute option if the tasks
do not require the same processing capabilities as the UI, or a larger instance if they require additional
capacity and resources.
Manageability : Background tasks might have a different development and deployment rhythm from the
main application code or the UI. Deploying them to a separate compute instance can simplify updates and
versioning.
Cost : Adding compute instances to execute background tasks increases hosting costs. You should carefully
consider the trade-off between additional capacity and these extra costs.
For more information, see the Leader Election pattern and the Competing Consumers pattern.

Conflicts
If you have multiple instances of a background job, it is possible that they will compete for access to resources and
services, such as databases and storage. This concurrent access can result in resource contention, which might
cause conflicts in availability of the services and in the integrity of data in storage. You can resolve resource
contention by using a pessimistic locking approach. This prevents competing instances of a task from concurrently
accessing a service or corrupting data.
Another approach to resolve conflicts is to define background tasks as a singleton, so that there is only ever one
instance running. However, this eliminates the reliability and performance benefits that a multiple-instance
configuration can provide. This is especially true if the UI can supply sufficient work to keep more than one
background task busy.
It is vital to ensure that the background task can automatically restart and that it has sufficient capacity to cope
with peaks in demand. You can achieve this by allocating a compute instance with sufficient resources, by
implementing a queueing mechanism that can store requests for later execution when demand decreases, or by
using a combination of these techniques.
Coordination
The background tasks might be complex and might require multiple individual tasks to execute to produce a result
or to fulfill all the requirements. It is common in these scenarios to divide the task into smaller discreet steps or
subtasks that can be executed by multiple consumers. Multistep jobs can be more efficient and more flexible
because individual steps might be reusable in multiple jobs. It is also easy to add, remove, or modify the order of
the steps.
Coordinating multiple tasks and steps can be challenging, but there are three common patterns that you can use to
guide your implementation of a solution:
Decomposing a task into multiple reusable steps . An application might be required to perform a
variety of tasks of varying complexity on the information that it processes. A straightforward but inflexible
approach to implementing this application might be to perform this processing as a monolithic module.
However, this approach is likely to reduce the opportunities for refactoring the code, optimizing it, or
reusing it if parts of the same processing are required elsewhere within the application. For more
information, see the Pipes and Filters pattern.
Managing execution of the steps for a task . An application might perform tasks that comprise a
number of steps (some of which might invoke remote services or access remote resources). The individual
steps might be independent of each other, but they are orchestrated by the application logic that
implements the task. For more information, see Scheduler Agent Supervisor pattern.
Managing recover y for task steps that fail . An application might need to undo the work that is
performed by a series of steps (which together define an eventually consistent operation) if one or more of
the steps fail. For more information, see the Compensating Transaction pattern.

Resiliency considerations
Background tasks must be resilient in order to provide reliable services to the application. When you are planning
and designing background tasks, consider the following points:
Background tasks must be able to gracefully handle restarts without corrupting data or introducing
inconsistency into the application. For long-running or multistep tasks, consider using check pointing by
saving the state of jobs in persistent storage, or as messages in a queue if this is appropriate. For example,
you can persist state information in a message in a queue and incrementally update this state information
with the task progress so that the task can be processed from the last known good checkpoint--instead of
restarting from the beginning. When using Azure Service Bus queues, you can use message sessions to
enable the same scenario. Sessions allow you to save and retrieve the application processing state by using
the SetState and GetState methods. For more information about designing reliable multistep processes and
workflows, see the Scheduler Agent Supervisor pattern.
When you use queues to communicate with background tasks, the queues can act as a buffer to store
requests that are sent to the tasks while the application is under higher than usual load. This allows the
tasks to catch up with the UI during less busy periods. It also means that restarts will not block the UI. For
more information, see the Queue-Based Load Leveling pattern. If some tasks are more important than
others, consider implementing the Priority Queue pattern to ensure that these tasks run before less
important ones.
Background tasks that are initiated by messages or process messages must be designed to handle
inconsistencies, such as messages arriving out of order, messages that repeatedly cause an error (often
referred to as poison messages), and messages that are delivered more than once. Consider the following:
Messages that must be processed in a specific order, such as those that change data based on the
existing data value (for example, adding a value to an existing value), might not arrive in the original
order in which they were sent. Alternatively, they might be handled by different instances of a
 background task in a different order due to varying loads on each instance. Messages that must be
processed in a specific order should include a sequence number, key, or some other indicator that
background tasks can use to ensure that they are processed in the correct order. If you are using
Azure Service Bus, you can use message sessions to guarantee the order of delivery. However, it is
usually more efficient, where possible, to design the process so that the message order is not
important.
Typically, a background task will peek at messages in the queue, which temporarily hides them from
other message consumers. Then it deletes the messages after they have been successfully processed.
If a background task fails when processing a message, that message will reappear on the queue after
the peek time-out expires. It will be processed by another instance of the task or during the next
processing cycle of this instance. If the message consistently causes an error in the consumer, it will
block the task, the queue, and eventually the application itself when the queue becomes full.
Therefore, it is vital to detect and remove poison messages from the queue. If you are using Azure
Service Bus, messages that cause an error can be moved automatically or manually to an associated
dead letter queue.
Queues are guaranteed at least once delivery mechanisms, but they might deliver the same message
more than once. In addition, if a background task fails after processing a message but before deleting
it from the queue, the message will become available for processing again. Background tasks should
be idempotent, which means that processing the same message more than once does not cause an
error or inconsistency in the application's data. Some operations are naturally idempotent, such as
setting a stored value to a specific new value. However, operations such as adding a value to an
existing stored value without checking that the stored value is still the same as when the message
was originally sent will cause inconsistencies. Azure Service Bus queues can be configured to
automatically remove duplicated messages.
Some messaging systems, such as Azure storage queues and Azure Service Bus queues, support a
de-queue count property that indicates the number of times a message has been read from the
queue. This can be useful in handling repeated and poison messages. For more information, see
Asynchronous Messaging Primer and Idempotency Patterns.

Scaling and performance considerations

Background tasks must offer sufficient performance to ensure they do not block the application, or cause
inconsistencies due to delayed operation when the system is under load. Typically, performance is improved by
scaling the compute instances that host the background tasks. When you are planning and designing background
tasks, consider the following points around scalability and performance:
Azure supports autoscaling (both scaling out and scaling back in) based on current demand and load or on
a predefined schedule, for Web Apps and Virtual Machines hosted deployments. Use this feature to ensure
that the application as a whole has sufficient performance capabilities while minimizing runtime costs.
Where background tasks have a different performance capability from the other parts of an application (for
example, the UI or components such as the data access layer), hosting the background tasks together in a
separate compute service allows the UI and background tasks to scale independently to manage the load. If
multiple background tasks have significantly different performance capabilities from each other, consider
dividing them and scaling each type independently. However, note that this might increase runtime costs.
Simply scaling the compute resources might not be sufficient to prevent loss of performance under load.
You might also need to scale storage queues and other resources to prevent a single point of the overall
processing chain from becoming a bottleneck. Also, consider other limitations, such as the maximum
throughput of storage and other services that the application and the background tasks rely on.
Background tasks must be designed for scaling. For example, they must be able to dynamically detect the
 number of storage queues in use in order to listen on or send messages to the appropriate queue.
By default, WebJobs scale with their associated Azure Web Apps instance. However, if you want a WebJob to
run as only a single instance, you can create a Settings.job file that contains the JSON data {
"is_singleton": true } . This forces Azure to only run one instance of the WebJob, even if there are multiple
instances of the associated web app. This can be a useful technique for scheduled jobs that must run as only
a single instance.

Related patterns
Compute Partitioning Guidance
 Caching
12/18/2020 • 55 minutes to read • Edit Online

Caching is a common technique that aims to improve the performance and scalability of a system. It does this by
temporarily copying frequently accessed data to fast storage that's located close to the application. If this fast data
storage is located closer to the application than the original source, then caching can significantly improve
response times for client applications by serving data more quickly.
Caching is most effective when a client instance repeatedly reads the same data, especially if all the following
conditions apply to the original data store:
It remains relatively static.
It's slow compared to the speed of the cache.
It's subject to a high level of contention.
It's far away when network latency can cause access to be slow.

Caching in distributed applications

Distributed applications typically implement either or both of the following strategies when caching data:
Using a private cache, where data is held locally on the computer that's running an instance of an application
or service.
Using a shared cache, serving as a common source that can be accessed by multiple processes and machines.
In both cases, caching can be performed client-side and server-side. Client-side caching is done by the process
that provides the user interface for a system, such as a web browser or desktop application. Server-side caching is
done by the process that provides the business services that are running remotely.
Private caching
The most basic type of cache is an in-memory store. It's held in the address space of a single process and accessed
directly by the code that runs in that process. This type of cache is quick to access. It can also provide an effective
means for storing modest amounts of static data, since the size of a cache is typically constrained by the amount
of memory available on the machine hosting the process.
If you need to cache more information than is physically possible in memory, you can write cached data to the
local file system. This will be slower to access than data held in memory, but should still be faster and more
reliable than retrieving data across a network.
If you have multiple instances of an application that uses this model running concurrently, each application
instance has its own independent cache holding its own copy of the data.
Think of a cache as a snapshot of the original data at some point in the past. If this data is not static, it is likely that
different application instances hold different versions of the data in their caches. Therefore, the same query
performed by these instances can return different results, as shown in Figure 1.
Figure 1: Using an in-memory cache in different instances of an application.
Shared caching
Using a shared cache can help alleviate concerns that data might differ in each cache, which can occur with in-
memory caching. Shared caching ensures that different application instances see the same view of cached data. It
does this by locating the cache in a separate location, typically hosted as part of a separate service, as shown in
Figure 2.

Figure 2: Using a shared cache.

An important benefit of the shared caching approach is the scalability it provides. Many shared cache services are
implemented by using a cluster of servers and use software to distribute the data across the cluster transparently.
An application instance simply sends a request to the cache service. The underlying infrastructure determines the
location of the cached data in the cluster. You can easily scale the cache by adding more servers.
There are two main disadvantages of the shared caching approach:
The cache is slower to access because it is no longer held locally to each application instance.
The requirement to implement a separate cache service might add complexity to the solution.

Considerations for using caching

The following sections describe in more detail the considerations for designing and using a cache.
Decide when to cache data
Caching can dramatically improve performance, scalability, and availability. The more data that you have and the
larger the number of users that need to access this data, the greater the benefits of caching become. That's
because caching reduces the latency and contention that's associated with handling large volumes of concurrent
requests in the original data store.
For example, a database might support a limited number of concurrent connections. Retrieving data from a
shared cache, however, rather than the underlying database, makes it possible for a client application to access
this data even if the number of available connections is currently exhausted. Additionally, if the database becomes
unavailable, client applications might be able to continue by using the data that's held in the cache.
Consider caching data that is read frequently but modified infrequently (for example, data that has a higher
proportion of read operations than write operations). However, we don't recommend that you use the cache as
the authoritative store of critical information. Instead, ensure that all changes that your application cannot afford
to lose are always saved to a persistent data store. This means that if the cache is unavailable, your application can
still continue to operate by using the data store, and you won't lose important information.
Determine how to cache data effectively
The key to using a cache effectively lies in determining the most appropriate data to cache, and caching it at the
appropriate time. The data can be added to the cache on demand the first time it is retrieved by an application.
This means that the application needs to fetch the data only once from the data store, and that subsequent access
can be satisfied by using the cache.
Alternatively, a cache can be partially or fully populated with data in advance, typically when the application starts
(an approach known as seeding). However, it might not be advisable to implement seeding for a large cache
because this approach can impose a sudden, high load on the original data store when the application starts
running.
Often an analysis of usage patterns can help you decide whether to fully or partially prepopulate a cache, and to
choose the data to cache. For example, it can be useful to seed the cache with the static user profile data for
customers who use the application regularly (perhaps every day), but not for customers who use the application
only once a week.
Caching typically works well with data that is immutable or that changes infrequently. Examples include reference
information such as product and pricing information in an e-commerce application, or shared static resources that
are costly to construct. Some or all of this data can be loaded into the cache at application startup to minimize
demand on resources and to improve performance. It might also be appropriate to have a background process
that periodically updates reference data in the cache to ensure it is up-to-date, or that refreshes the cache when
reference data changes.
Caching is less useful for dynamic data, although there are some exceptions to this consideration (see the section
Cache highly dynamic data later in this article for more information). When the original data changes regularly,
either the cached information becomes stale very quickly or the overhead of synchronizing the cache with the
original data store reduces the effectiveness of caching.
Note that a cache does not have to include the complete data for an entity. For example, if a data item represents a
multivalued object such as a bank customer with a name, address, and account balance, some of these elements
might remain static (such as the name and address), while others (such as the account balance) might be more
dynamic. In these situations, it can be useful to cache the static portions of the data and retrieve (or calculate) only
the remaining information when it is required.
We recommend that you carry out performance testing and usage analysis to determine whether prepopulating
or on-demand loading of the cache, or a combination of both, is appropriate. The decision should be based on the
volatility and usage pattern of the data. Cache utilization and performance analysis are particularly important in
applications that encounter heavy loads and must be highly scalable. For example, in highly scalable scenarios it
might make sense to seed the cache to reduce the load on the data store at peak times.
Caching can also be used to avoid repeating computations while the application is running. If an operation
transforms data or performs a complicated calculation, it can save the results of the operation in the cache. If the
same calculation is required afterward, the application can simply retrieve the results from the cache.
An application can modify data that's held in a cache. However, we recommend thinking of the cache as a transient
data store that could disappear at any time. Do not store valuable data in the cache only; make sure that you
maintain the information in the original data store as well. This means that if the cache becomes unavailable, you
minimize the chance of losing data.
Cache highly dynamic data
When you store rapidly changing information in a persistent data store, it can impose an overhead on the system.
For example, consider a device that continually reports status or some other measurement. If an application
chooses not to cache this data on the basis that the cached information will nearly always be outdated, then the
same consideration could be true when storing and retrieving this information from the data store. In the time it
takes to save and fetch this data, it might have changed.
In a situation such as this, consider the benefits of storing the dynamic information directly in the cache instead of
in the persistent data store. If the data is noncritical and does not require auditing, then it doesn't matter if the
occasional change is lost.
Manage data expiration in a cache
In most cases, data that's held in a cache is a copy of data that's held in the original data store. The data in the
original data store might change after it was cached, causing the cached data to become stale. Many caching
systems enable you to configure the cache to expire data and reduce the period for which data may be out of date.
When cached data expires, it's removed from the cache, and the application must retrieve the data from the
original data store (it can put the newly fetched information back into cache). You can set a default expiration
policy when you configure the cache. In many cache services, you can also stipulate the expiration period for
individual objects when you store them programmatically in the cache. Some caches enable you to specify the
expiration period as an absolute value, or as a sliding value that causes the item to be removed from the cache if it
is not accessed within the specified time. This setting overrides any cache-wide expiration policy, but only for the
specified objects.

NOTE
Consider the expiration period for the cache and the objects that it contains carefully. If you make it too short, objects will
expire too quickly and you will reduce the benefits of using the cache. If you make the period too long, you risk the data
becoming stale.

It's also possible that the cache might fill up if data is allowed to remain resident for a long time. In this case, any
requests to add new items to the cache might cause some items to be forcibly removed in a process known as
eviction. Cache services typically evict data on a least-recently-used (LRU) basis, but you can usually override this
policy and prevent items from being evicted. However, if you adopt this approach, you risk exceeding the memory
that's available in the cache. An application that attempts to add an item to the cache will fail with an exception.
Some caching implementations might provide additional eviction policies. There are several types of eviction
policies. These include:
A most-recently-used policy (in the expectation that the data will not be required again).
A first-in-first-out policy (oldest data is evicted first).
An explicit removal policy based on a triggered event (such as the data being modified).
Invalidate data in a client-side cache
Data that's held in a client-side cache is generally considered to be outside the auspices of the service that
provides the data to the client. A service cannot directly force a client to add or remove information from a client-
side cache.
This means that it's possible for a client that uses a poorly configured cache to continue using outdated
information. For example, if the expiration policies of the cache aren't properly implemented, a client might use
outdated information that's cached locally when the information in the original data source has changed.
If you are building a web application that serves data over an HTTP connection, you can implicitly force a web
client (such as a browser or web proxy) to fetch the most recent information. You can do this if a resource is
updated by a change in the URI of that resource. Web clients typically use the URI of a resource as the key in the
client-side cache, so if the URI changes, the web client ignores any previously cached versions of a resource and
fetches the new version instead.

Managing concurrency in a cache

Caches are often designed to be shared by multiple instances of an application. Each application instance can read
and modify data in the cache. Consequently, the same concurrency issues that arise with any shared data store
also apply to a cache. In a situation where an application needs to modify data that's held in the cache, you might
need to ensure that updates made by one instance of the application do not overwrite the changes made by
another instance.
Depending on the nature of the data and the likelihood of collisions, you can adopt one of two approaches to
concurrency:
Optimistic . Immediately prior to updating the data, the application checks to see whether the data in the
cache has changed since it was retrieved. If the data is still the same, the change can be made. Otherwise, the
application has to decide whether to update it. (The business logic that drives this decision will be application-
specific.) This approach is suitable for situations where updates are infrequent, or where collisions are unlikely
to occur.
Pessimistic . When it retrieves the data, the application locks it in the cache to prevent another instance from
changing it. This process ensures that collisions cannot occur, but they can also block other instances that need
to process the same data. Pessimistic concurrency can affect the scalability of a solution and is recommended
only for short-lived operations. This approach might be appropriate for situations where collisions are more
likely, especially if an application updates multiple items in the cache and must ensure that these changes are
applied consistently.
Implement high availability and scalability, and improve performance
Avoid using a cache as the primary repository of data; this is the role of the original data store from which the
cache is populated. The original data store is responsible for ensuring the persistence of the data.
Be careful not to introduce critical dependencies on the availability of a shared cache service into your solutions.
An application should be able to continue functioning if the service that provides the shared cache is unavailable.
The application should not become unresponsive or fail while waiting for the cache service to resume.
Therefore, the application must be prepared to detect the availability of the cache service and fall back to the
original data store if the cache is inaccessible. The Circuit-Breaker pattern is useful for handling this scenario. The
service that provides the cache can be recovered, and once it becomes available, the cache can be repopulated as
data is read from the original data store, following a strategy such as the Cache-aside pattern.
However, system scalability may be affected if the application falls back to the original data store when the cache
is temporarily unavailable. While the data store is being recovered, the original data store could be swamped with
requests for data, resulting in timeouts and failed connections.
Consider implementing a local, private cache in each instance of an application, together with the shared cache
that all application instances access. When the application retrieves an item, it can check first in its local cache,
then in the shared cache, and finally in the original data store. The local cache can be populated using the data in
either the shared cache, or in the database if the shared cache is unavailable.
This approach requires careful configuration to prevent the local cache from becoming too stale with respect to
the shared cache. However, the local cache acts as a buffer if the shared cache is unreachable. Figure 3 shows this
structure.

Figure 3: Using a local private cache with a shared cache.

To support large caches that hold relatively long-lived data, some cache services provide a high-availability option
that implements automatic failover if the cache becomes unavailable. This approach typically involves replicating
the cached data that's stored on a primary cache server to a secondary cache server, and switching to the
secondary server if the primary server fails or connectivity is lost.
To reduce the latency that's associated with writing to multiple destinations, the replication to the secondary
server might occur asynchronously when data is written to the cache on the primary server. This approach leads
to the possibility that some cached information might be lost in the event of a failure, but the proportion of this
data should be small compared to the overall size of the cache.
If a shared cache is large, it might be beneficial to partition the cached data across nodes to reduce the chances of
contention and improve scalability. Many shared caches support the ability to dynamically add (and remove)
nodes and rebalance the data across partitions. This approach might involve clustering, in which the collection of
nodes is presented to client applications as a seamless, single cache. Internally, however, the data is dispersed
between nodes following a predefined distribution strategy that balances the load evenly. For more information
about possible partitioning strategies, see Data partitioning guidance.
Clustering can also increase the availability of the cache. If a node fails, the remainder of the cache is still
accessible. Clustering is frequently used in conjunction with replication and failover. Each node can be replicated,
and the replica can be quickly brought online if the node fails.
Many read and write operations are likely to involve single data values or objects. However, at times it might be
necessary to store or retrieve large volumes of data quickly. For example, seeding a cache could involve writing
hundreds or thousands of items to the cache. An application might also need to retrieve a large number of related
items from the cache as part of the same request.
Many large-scale caches provide batch operations for these purposes. This enables a client application to package
up a large volume of items into a single request and reduces the overhead that's associated with performing a
large number of small requests.

Caching and eventual consistency

For the cache-aside pattern to work, the instance of the application that populates the cache must have access to
the most recent and consistent version of the data. In a system that implements eventual consistency (such as a
replicated data store) this might not be the case.
One instance of an application could modify a data item and invalidate the cached version of that item. Another
instance of the application might attempt to read this item from a cache, which causes a cache-miss, so it reads
the data from the data store and adds it to the cache. However, if the data store has not been fully synchronized
with the other replicas, the application instance could read and populate the cache with the old value.
For more information about handling data consistency, see the Data consistency primer.
Protect cached data
Irrespective of the cache service you use, consider how to protect the data that's held in the cache from
unauthorized access. There are two main concerns:
The privacy of the data in the cache.
The privacy of data as it flows between the cache and the application that's using the cache.
To protect data in the cache, the cache service might implement an authentication mechanism that requires that
applications specify the following:
Which identities can access data in the cache.
Which operations (read and write) that these identities are allowed to perform.
To reduce overhead that's associated with reading and writing data, after an identity has been granted write
and/or read access to the cache, that identity can use any data in the cache.
If you need to restrict access to subsets of the cached data, you can do one of the following:
Split the cache into partitions (by using different cache servers) and only grant access to identities for the
partitions that they should be allowed to use.
Encrypt the data in each subset by using different keys, and provide the encryption keys only to identities that
should have access to each subset. A client application might still be able to retrieve all of the data in the cache,
but it will only be able to decrypt the data for which it has the keys.
You must also protect the data as it flows in and out of the cache. To do this, you depend on the security features
provided by the network infrastructure that client applications use to connect to the cache. If the cache is
implemented using an on-site server within the same organization that hosts the client applications, then the
isolation of the network itself might not require you to take additional steps. If the cache is located remotely and
requires a TCP or HTTP connection over a public network (such as the Internet), consider implementing SSL.
Considerations for implementing caching in Azure
Azure Cache for Redis is an implementation of the open source Redis cache that runs as a service in an Azure
datacenter. It provides a caching service that can be accessed from any Azure application, whether the application
is implemented as a cloud service, a website, or inside an Azure virtual machine. Caches can be shared by client
applications that have the appropriate access key.
Azure Cache for Redis is a high-performance caching solution that provides availability, scalability and security. It
typically runs as a service spread across one or more dedicated machines. It attempts to store as much
information as it can in memory to ensure fast access. This architecture is intended to provide low latency and
high throughput by reducing the need to perform slow I/O operations.
Azure Cache for Redis is compatible with many of the various APIs that are used by client applications. If you have
existing applications that already use Azure Cache for Redis running on-premises, the Azure Cache for Redis
provides a quick migration path to caching in the cloud.
Features of Redis
Redis is more than a simple cache server. It provides a distributed in-memory database with an extensive
command set that supports many common scenarios. These are described later in this document, in the section
Using Redis caching. This section summarizes some of the key features that Redis provides.
Redis as an in-memory database
Redis supports both read and write operations. In Redis, writes can be protected from system failure either by
being stored periodically in a local snapshot file or in an append-only log file. This is not the case in many caches
(which should be considered transitory data stores).
All writes are asynchronous and do not block clients from reading and writing data. When Redis starts running, it
reads the data from the snapshot or log file and uses it to construct the in-memory cache. For more information,
see Redis persistence on the Redis website.

NOTE
Redis does not guarantee that all writes will be saved in the event of a catastrophic failure, but at worst you might lose only
a few seconds worth of data. Remember that a cache is not intended to act as an authoritative data source, and it is the
responsibility of the applications using the cache to ensure that critical data is saved successfully to an appropriate data
store. For more information, see the Cache-aside pattern.

Redis data types

Redis is a key-value store, where values can contain simple types or complex data structures such as hashes, lists,
and sets. It supports a set of atomic operations on these data types. Keys can be permanent or tagged with a
limited time-to-live, at which point the key and its corresponding value are automatically removed from the cache.
For more information about Redis keys and values, visit the page An introduction to Redis data types and
abstractions on the Redis website.
Redis replication and clustering
Redis supports primary/subordinate replication to help ensure availability and maintain throughput. Write
operations to a Redis primary node are replicated to one or more subordinate nodes. Read operations can be
served by the primary or any of the subordinates.
In the event of a network partition, subordinates can continue to serve data and then transparently resynchronize
with the primary when the connection is reestablished. For further details, visit the Replication page on the Redis
website.
Redis also provides clustering, which enables you to transparently partition data into shards across servers and
spread the load. This feature improves scalability, because new Redis servers can be added and the data
repartitioned as the size of the cache increases.
Furthermore, each server in the cluster can be replicated by using primary/subordinate replication. This ensures
availability across each node in the cluster. For more information about clustering and sharding, visit the Redis
cluster tutorial page on the Redis website.
Redis memory use
A Redis cache has a finite size that depends on the resources available on the host computer. When you configure
a Redis server, you can specify the maximum amount of memory it can use. You can also configure a key in a
Redis cache to have an expiration time, after which it is automatically removed from the cache. This feature can
help prevent the in-memory cache from filling with old or stale data.
As memory fills up, Redis can automatically evict keys and their values by following a number of policies. The
default is LRU (least recently used), but you can also select other policies such as evicting keys at random or
turning off eviction altogether (in which, case attempts to add items to the cache fail if it is full). The page Using
Redis as an LRU cache provides more information.
Redis transactions and batches
Redis enables a client application to submit a series of operations that read and write data in the cache as an
atomic transaction. All the commands in the transaction are guaranteed to run sequentially, and no commands
issued by other concurrent clients will be interwoven between them.
However, these are not true transactions as a relational database would perform them. Transaction processing
consists of two stages--the first is when the commands are queued, and the second is when the commands are
run. During the command queuing stage, the commands that comprise the transaction are submitted by the
client. If some sort of error occurs at this point (such as a syntax error, or the wrong number of parameters) then
Redis refuses to process the entire transaction and discards it.
During the run phase, Redis performs each queued command in sequence. If a command fails during this phase,
Redis continues with the next queued command and does not roll back the effects of any commands that have
already been run. This simplified form of transaction helps to maintain performance and avoid performance
problems that are caused by contention.
Redis does implement a form of optimistic locking to assist in maintaining consistency. For detailed information
about transactions and locking with Redis, visit the Transactions page on the Redis website.
Redis also supports nontransactional batching of requests. The Redis protocol that clients use to send commands
to a Redis server enables a client to send a series of operations as part of the same request. This can help to
reduce packet fragmentation on the network. When the batch is processed, each command is performed. If any of
these commands are malformed, they will be rejected (which doesn't happen with a transaction), but the
remaining commands will be performed. There is also no guarantee about the order in which the commands in
the batch will be processed.
Redis security
Redis is focused purely on providing fast access to data, and is designed to run inside a trusted environment that
can be accessed only by trusted clients. Redis supports a limited security model based on password
authentication. (It is possible to remove authentication completely, although we don't recommend this.)
All authenticated clients share the same global password and have access to the same resources. If you need more
comprehensive sign-in security, you must implement your own security layer in front of the Redis server, and all
client requests should pass through this additional layer. Redis should not be directly exposed to untrusted or
unauthenticated clients.
You can restrict access to commands by disabling them or renaming them (and by providing only privileged
clients with the new names).
Redis does not directly support any form of data encryption, so all encoding must be performed by client
applications. Additionally, Redis does not provide any form of transport security. If you need to protect data as it
flows across the network, we recommend implementing an SSL proxy.
For more information, visit the Redis security page on the Redis website.

NOTE
Azure Cache for Redis provides its own security layer through which clients connect. The underlying Redis servers are not
exposed to the public network.

Azure Redis cache

Azure Cache for Redis provides access to Redis servers that are hosted at an Azure datacenter. It acts as a façade
that provides access control and security. You can provision a cache by using the Azure portal.
The portal provides a number of predefined configurations. These range from a 53 GB cache running as a
dedicated service that supports SSL communications (for privacy) and master/subordinate replication with an
SLA of 99.9% availability, down to a 250 MB cache without replication (no availability guarantees) running on
shared hardware.
Using the Azure portal, you can also configure the eviction policy of the cache, and control access to the cache by
adding users to the roles provided. These roles, which define the operations that members can perform, include
Owner, Contributor, and Reader. For example, members of the Owner role have complete control over the cache
(including security) and its contents, members of the Contributor role can read and write information in the cache,
and members of the Reader role can only retrieve data from the cache.
Most administrative tasks are performed through the Azure portal. For this reason, many of the administrative
commands that are available in the standard version of Redis are not available, including the ability to modify the
configuration programmatically, shut down the Redis server, configure additional subordinates, or forcibly save
data to disk.
The Azure portal includes a convenient graphical display that enables you to monitor the performance of the
cache. For example, you can view the number of connections being made, the number of requests being
performed, the volume of reads and writes, and the number of cache hits versus cache misses. Using this
information, you can determine the effectiveness of the cache and if necessary, switch to a different configuration
or change the eviction policy.
Additionally, you can create alerts that send email messages to an administrator if one or more critical metrics fall
outside of an expected range. For example, you might want to alert an administrator if the number of cache
misses exceeds a specified value in the last hour, because it means the cache might be too small or data might be
being evicted too quickly.
You can also monitor the CPU, memory, and network usage for the cache.
For further information and examples showing how to create and configure an Azure Cache for Redis, visit the
page Lap around Azure Cache for Redis on the Azure blog.

Caching session state and HTML output

If you're building ASP.NET web applications that run by using Azure web roles, you can save session state
information and HTML output in an Azure Cache for Redis. The session state provider for Azure Cache for Redis
enables you to share session information between different instances of an ASP.NET web application, and is very
useful in web farm situations where client-server affinity is not available and caching session data in-memory
would not be appropriate.
Using the session state provider with Azure Cache for Redis delivers several benefits, including:
Sharing session state with a large number of instances of ASP.NET web applications.
 Providing improved scalability.
Supporting controlled, concurrent access to the same session state data for multiple readers and a single
writer.
Using compression to save memory and improve network performance.
For more information, see ASP.NET session state provider for Azure Cache for Redis.

NOTE
Do not use the session state provider for Azure Cache for Redis with ASP.NET applications that run outside of the Azure
environment. The latency of accessing the cache from outside of Azure can eliminate the performance benefits of caching
data.

Similarly, the output cache provider for Azure Cache for Redis enables you to save the HTTP responses generated
by an ASP.NET web application. Using the output cache provider with Azure Cache for Redis can improve the
response times of applications that render complex HTML output. Application instances that generate similar
responses can use the shared output fragments in the cache rather than generating this HTML output afresh. For
more information, see ASP.NET output cache provider for Azure Cache for Redis.

Building a custom Redis cache

Azure Cache for Redis acts as a façade to the underlying Redis servers. If you require an advanced configuration
that is not covered by the Azure Redis cache (such as a cache bigger than 53 GB) you can build and host your own
Redis servers by using Azure virtual machines.
This is a potentially complex process because you might need to create several VMs to act as primary and
subordinate nodes if you want to implement replication. Furthermore, if you wish to create a cluster, then you
need multiple primaries and subordinate servers. A minimal clustered replication topology that provides a high
degree of availability and scalability comprises at least six VMs organized as three pairs of primary/subordinate
servers (a cluster must contain at least three primary nodes).
Each primary/subordinate pair should be located close together to minimize latency. However, each set of pairs
can be running in different Azure datacenters located in different regions, if you wish to locate cached data close
to the applications that are most likely to use it. For an example of building and configuring a Redis node running
as an Azure VM, see Running Redis on a CentOS Linux VM in Azure.

NOTE
If you implement your own Redis cache in this way, you are responsible for monitoring, managing, and securing the service.

Partitioning a Redis cache

Partitioning the cache involves splitting the cache across multiple computers. This structure gives you several
advantages over using a single cache server, including:
Creating a cache that is much bigger than can be stored on a single server.
Distributing data across servers, improving availability. If one server fails or becomes inaccessible, the data
that it holds is unavailable, but the data on the remaining servers can still be accessed. For a cache, this is not
crucial because the cached data is only a transient copy of the data that's held in a database. Cached data on a
server that becomes inaccessible can be cached on a different server instead.
Spreading the load across servers, thereby improving performance and scalability.
Geolocating data close to the users that access it, thus reducing latency.
For a cache, the most common form of partitioning is sharding. In this strategy, each partition (or shard) is a Redis
cache in its own right. Data is directed to a specific partition by using sharding logic, which can use a variety of
approaches to distribute the data. The Sharding pattern provides more information about implementing sharding.
To implement partitioning in a Redis cache, you can take one of the following approaches:
Server-side query routing. In this technique, a client application sends a request to any of the Redis servers
that comprise the cache (probably the closest server). Each Redis server stores metadata that describes the
partition that it holds, and also contains information about which partitions are located on other servers. The
Redis server examines the client request. If it can be resolved locally, it will perform the requested operation.
Otherwise it will forward the request on to the appropriate server. This model is implemented by Redis
clustering, and is described in more detail on the Redis cluster tutorial page on the Redis website. Redis
clustering is transparent to client applications, and additional Redis servers can be added to the cluster (and
the data re-partitioned) without requiring that you reconfigure the clients.
Client-side partitioning. In this model, the client application contains logic (possibly in the form of a library)
that routes requests to the appropriate Redis server. This approach can be used with Azure Cache for Redis.
Create multiple Azure Cache for Redis (one for each data partition) and implement the client-side logic that
routes the requests to the correct cache. If the partitioning scheme changes (if additional Azure Cache for Redis
are created, for example), client applications might need to be reconfigured.
Proxy-assisted partitioning. In this scheme, client applications send requests to an intermediary proxy service
which understands how the data is partitioned and then routes the request to the appropriate Redis server.
This approach can also be used with Azure Cache for Redis; the proxy service can be implemented as an Azure
cloud service. This approach requires an additional level of complexity to implement the service, and requests
might take longer to perform than using client-side partitioning.
The page Partitioning: how to split data among multiple Redis instances on the Redis website provides further
information about implementing partitioning with Redis.
Implement Redis cache client applications
Redis supports client applications written in numerous programming languages. If you are building new
applications by using the .NET Framework, the recommended approach is to use the StackExchange.Redis client
library. This library provides a .NET Framework object model that abstracts the details for connecting to a Redis
server, sending commands, and receiving responses. It is available in Visual Studio as a NuGet package. You can
use this same library to connect to an Azure Cache for Redis, or a custom Redis cache hosted on a VM.
To connect to a Redis server you use the static Connect method of the ConnectionMultiplexer class. The
connection that this method creates is designed to be used throughout the lifetime of the client application, and
the same connection can be used by multiple concurrent threads. Do not reconnect and disconnect each time you
perform a Redis operation because this can degrade performance.
You can specify the connection parameters, such as the address of the Redis host and the password. If you are
using Azure Cache for Redis, the password is either the primary or secondary key that is generated for Azure
Cache for Redis by using the Azure portal.
After you have connected to the Redis server, you can obtain a handle on the Redis database that acts as the cache.
The Redis connection provides the GetDatabase method to do this. You can then retrieve items from the cache and
store data in the cache by using the StringGet and StringSet methods. These methods expect a key as a
parameter, and return the item either in the cache that has a matching value ( StringGet ) or add the item to the
cache with this key ( StringSet ).
Depending on the location of the Redis server, many operations might incur some latency while a request is
transmitted to the server and a response is returned to the client. The StackExchange library provides
asynchronous versions of many of the methods that it exposes to help client applications remain responsive.
These methods support the Task-based Asynchronous pattern in the .NET Framework.
The following code snippet shows a method named RetrieveItem . It illustrates an implementation of the cache-
aside pattern based on Redis and the StackExchange library. The method takes a string key value and attempts to
retrieve the corresponding item from the Redis cache by calling the StringGetAsync method (the asynchronous
version of StringGet ).
If the item is not found, it is fetched from the underlying data source using the GetItemFromDataSourceAsync
method (which is a local method and not part of the StackExchange library). It's then added to the cache by using
the StringSetAsync method so it can be retrieved more quickly next time.

// Connect to the Azure Redis cache

ConfigurationOptions config = new ConfigurationOptions();
config.EndPoints.Add("<your DNS name>.redis.cache.windows.net");
config.Password = "<Redis cache key from management portal>";
ConnectionMultiplexer redisHostConnection = ConnectionMultiplexer.Connect(config);
IDatabase cache = redisHostConnection.GetDatabase();
...
private async Task<string> RetrieveItem(string itemKey)
{
// Attempt to retrieve the item from the Redis cache
string itemValue = await cache.StringGetAsync(itemKey);

// If the value returned is null, the item was not found in the cache
// So retrieve the item from the data source and add it to the cache
if (itemValue == null)
{
itemValue = await GetItemFromDataSourceAsync(itemKey);
await cache.StringSetAsync(itemKey, itemValue);
}

// Return the item

return itemValue;
}

The StringGet and StringSet methods are not restricted to retrieving or storing string values. They can take any
item that is serialized as an array of bytes. If you need to save a .NET object, you can serialize it as a byte stream
and use the StringSet method to write it to the cache.
Similarly, you can read an object from the cache by using the StringGet method and deserializing it as a .NET
object. The following code shows a set of extension methods for the IDatabase interface (the GetDatabase method
of a Redis connection returns an IDatabase object), and some sample code that uses these methods to read and
write a BlogPost object to the cache:
 public static class RedisCacheExtensions
{
public static async Task<T> GetAsync<T>(this IDatabase cache, string key)
{
return Deserialize<T>(await cache.StringGetAsync(key));
}

public static async Task<object> GetAsync(this IDatabase cache, string key)

{
return Deserialize<object>(await cache.StringGetAsync(key));
}

public static async Task SetAsync(this IDatabase cache, string key, object value)
{
await cache.StringSetAsync(key, Serialize(value));
}

static byte[] Serialize(object o)

{
byte[] objectDataAsStream = null;

if (o != null)
{
BinaryFormatter binaryFormatter = new BinaryFormatter();
using (MemoryStream memoryStream = new MemoryStream())
{
binaryFormatter.Serialize(memoryStream, o);
objectDataAsStream = memoryStream.ToArray();
}
}

return objectDataAsStream;
}

static T Deserialize<T>(byte[] stream)

{
T result = default(T);

if (stream != null)
{
BinaryFormatter binaryFormatter = new BinaryFormatter();
using (MemoryStream memoryStream = new MemoryStream(stream))
{
result = (T)binaryFormatter.Deserialize(memoryStream);
}
}

return result;
}
}

The following code illustrates a method named RetrieveBlogPost that uses these extension methods to read and
write a serializable BlogPost object to the cache following the cache-aside pattern:
 // The BlogPost type
[Serializable]
public class BlogPost
{
private HashSet<string> tags;

public BlogPost(int id, string title, int score, IEnumerable<string> tags)

{
this.Id = id;
this.Title = title;
this.Score = score;
this.tags = new HashSet<string>(tags);
}

public int Id { get; set; }

public string Title { get; set; }
public int Score { get; set; }
public ICollection<string> Tags => this.tags;
}
...
private async Task<BlogPost> RetrieveBlogPost(string blogPostKey)
{
BlogPost blogPost = await cache.GetAsync<BlogPost>(blogPostKey);
if (blogPost == null)
{
blogPost = await GetBlogPostFromDataSourceAsync(blogPostKey);
await cache.SetAsync(blogPostKey, blogPost);
}

return blogPost;
}

Redis supports command pipelining if a client application sends multiple asynchronous requests. Redis can
multiplex the requests using the same connection rather than receiving and responding to commands in a strict
sequence.
This approach helps to reduce latency by making more efficient use of the network. The following code snippet
shows an example that retrieves the details of two customers concurrently. The code submits two requests and
then performs some other processing (not shown) before waiting to receive the results. The Wait method of the
cache object is similar to the .NET Framework Task.Wait method:

ConnectionMultiplexer redisHostConnection = ...;

IDatabase cache = redisHostConnection.GetDatabase();
...
var task1 = cache.StringGetAsync("customer:1");
var task2 = cache.StringGetAsync("customer:2");
...
var customer1 = cache.Wait(task1);
var customer2 = cache.Wait(task2);

For additional information on writing client applications that can the Azure Cache for Redis, see the Azure Cache
for Redis documentation. More information is also available at StackExchange.Redis.
The page Pipelines and multiplexers on the same website provides more information about asynchronous
operations and pipelining with Redis and the StackExchange library.

Using Redis caching

The simplest use of Redis for caching concerns is key-value pairs where the value is an uninterpreted string of
arbitrary length that can contain any binary data. (It is essentially an array of bytes that can be treated as a string).
This scenario was illustrated in the section Implement Redis Cache client applications earlier in this article.
Note that keys also contain uninterpreted data, so you can use any binary information as the key. The longer the
key is, however, the more space it will take to store, and the longer it will take to perform lookup operations. For
usability and ease of maintenance, design your keyspace carefully and use meaningful (but not verbose) keys.
For example, use structured keys such as "customer:100" to represent the key for the customer with ID 100 rather
than simply "100". This scheme enables you to easily distinguish between values that store different data types.
For example, you could also use the key "orders:100" to represent the key for the order with ID 100.
Apart from one-dimensional binary strings, a value in a Redis key-value pair can also hold more structured
information, including lists, sets (sorted and unsorted), and hashes. Redis provides a comprehensive command set
that can manipulate these types, and many of these commands are available to .NET Framework applications
through a client library such as StackExchange. The page An introduction to Redis data types and abstractions on
the Redis website provides a more detailed overview of these types and the commands that you can use to
manipulate them.
This section summarizes some common use cases for these data types and commands.
Perform atomic and batch operations
Redis supports a series of atomic get-and-set operations on string values. These operations remove the possible
race hazards that might occur when using separate GET and SET commands. The operations that are available
include:
INCR , INCRBY , DECR , and DECRBY , which perform atomic increment and decrement operations on integer
numeric data values. The StackExchange library provides overloaded versions of the
IDatabase.StringIncrementAsync and IDatabase.StringDecrementAsync methods to perform these
operations and return the resulting value that is stored in the cache. The following code snippet illustrates
how to use these methods:

ConnectionMultiplexer redisHostConnection = ...;

IDatabase cache = redisHostConnection.GetDatabase();
...
await cache.StringSetAsync("data:counter", 99);
...
long oldValue = await cache.StringIncrementAsync("data:counter");
// Increment by 1 (the default)
// oldValue should be 100

long newValue = await cache.StringDecrementAsync("data:counter", 50);

// Decrement by 50
// newValue should be 50

GETSET , which retrieves the value that's associated with a key and changes it to a new value. The
StackExchange library makes this operation available through the IDatabase.StringGetSetAsync method.
The code snippet below shows an example of this method. This code returns the current value that's
associated with the key "data:counter" from the previous example. Then it resets the value for this key back
to zero, all as part of the same operation:

ConnectionMultiplexer redisHostConnection = ...;

IDatabase cache = redisHostConnection.GetDatabase();
...
string oldValue = await cache.StringGetSetAsync("data:counter", 0);

MGET and MSET, which can return or change a set of string values as a single operation. The
IDatabase.StringGetAsync and IDatabase.StringSetAsync methods are overloaded to support this
functionality, as shown in the following example:
 ConnectionMultiplexer redisHostConnection = ...;
IDatabase cache = redisHostConnection.GetDatabase();
...
// Create a list of key-value pairs
var keysAndValues =
new List<KeyValuePair<RedisKey, RedisValue>>()
{
new KeyValuePair<RedisKey, RedisValue>("data:key1", "value1"),
new KeyValuePair<RedisKey, RedisValue>("data:key99", "value2"),
new KeyValuePair<RedisKey, RedisValue>("data:key322", "value3")
};

// Store the list of key-value pairs in the cache

cache.StringSet(keysAndValues.ToArray());
...
// Find all values that match a list of keys
RedisKey[] keys = { "data:key1", "data:key99", "data:key322"};
// values should contain { "value1", "value2", "value3" }
RedisValue[] values = cache.StringGet(keys);

You can also combine multiple operations into a single Redis transaction as described in the Redis transactions
and batches section earlier in this article. The StackExchange library provides support for transactions through the
ITransaction interface.

You create an ITransaction object by using the IDatabase.CreateTransaction method. You invoke commands to
the transaction by using the methods provided by the ITransaction object.
The ITransaction interface provides access to a set of methods that's similar to those accessed by the IDatabase
interface, except that all the methods are asynchronous. This means that they are only performed when the
ITransaction.Execute method is invoked. The value that's returned by the ITransaction.Execute method
indicates whether the transaction was created successfully (true) or if it failed (false).
The following code snippet shows an example that increments and decrements two counters as part of the same
transaction:

ConnectionMultiplexer redisHostConnection = ...;

IDatabase cache = redisHostConnection.GetDatabase();
...
ITransaction transaction = cache.CreateTransaction();
var tx1 = transaction.StringIncrementAsync("data:counter1");
var tx2 = transaction.StringDecrementAsync("data:counter2");
bool result = transaction.Execute();
Console.WriteLine("Transaction {0}", result ? "succeeded" : "failed");
Console.WriteLine("Result of increment: {0}", tx1.Result);
Console.WriteLine("Result of decrement: {0}", tx2.Result);

Remember that Redis transactions are unlike transactions in relational databases. The Execute method simply
queues all the commands that comprise the transaction to be run, and if any of them is malformed then the
transaction is stopped. If all the commands have been queued successfully, each command runs asynchronously.
If any command fails, the others still continue processing. If you need to verify that a command has completed
successfully, you must fetch the results of the command by using the Result property of the corresponding task,
as shown in the example above. Reading the Result property will block the calling thread until the task has
completed.
For more information, see Transactions in Redis.
When performing batch operations, you can use the IBatch interface of the StackExchange library. This interface
provides access to a set of methods similar to those accessed by the IDatabase interface, except that all the
methods are asynchronous.
You create an object by using the IDatabase.CreateBatch method, and then run the batch by using the
IBatch
IBatch.Execute method, as shown in the following example. This code simply sets a string value, increments and
decrements the same counters used in the previous example, and displays the results:

ConnectionMultiplexer redisHostConnection = ...;

IDatabase cache = redisHostConnection.GetDatabase();
...
IBatch batch = cache.CreateBatch();
batch.StringSetAsync("data:key1", 11);
var t1 = batch.StringIncrementAsync("data:counter1");
var t2 = batch.StringDecrementAsync("data:counter2");
batch.Execute();
Console.WriteLine("{0}", t1.Result);
Console.WriteLine("{0}", t2.Result);

It is important to understand that unlike a transaction, if a command in a batch fails because it is malformed, the
other commands might still run. The IBatch.Execute method does not return any indication of success or failure.
Perform fire and forget cache operations
Redis supports fire and forget operations by using command flags. In this situation, the client simply initiates an
operation but has no interest in the result and does not wait for the command to be completed. The example
below shows how to perform the INCR command as a fire and forget operation:

ConnectionMultiplexer redisHostConnection = ...;

IDatabase cache = redisHostConnection.GetDatabase();
...
await cache.StringSetAsync("data:key1", 99);
...
cache.StringIncrement("data:key1", flags: CommandFlags.FireAndForget);

Specify automatically expiring keys

When you store an item in a Redis cache, you can specify a timeout after which the item will be automatically
removed from the cache. You can also query how much more time a key has before it expires by using the TTL
command. This command is available to StackExchange applications by using the IDatabase.KeyTimeToLive
method.
The following code snippet shows how to set an expiration time of 20 seconds on a key, and query the remaining
lifetime of the key:

ConnectionMultiplexer redisHostConnection = ...;

IDatabase cache = redisHostConnection.GetDatabase();
...
// Add a key with an expiration time of 20 seconds
await cache.StringSetAsync("data:key1", 99, TimeSpan.FromSeconds(20));
...
// Query how much time a key has left to live
// If the key has already expired, the KeyTimeToLive function returns a null
TimeSpan? expiry = cache.KeyTimeToLive("data:key1");

You can also set the expiration time to a specific date and time by using the EXPIRE command, which is available in
the StackExchange library as the KeyExpireAsync method:
 ConnectionMultiplexer redisHostConnection = ...;
IDatabase cache = redisHostConnection.GetDatabase();
...
// Add a key with an expiration date of midnight on 1st January 2015
await cache.StringSetAsync("data:key1", 99);
await cache.KeyExpireAsync("data:key1",
new DateTime(2015, 1, 1, 0, 0, 0, DateTimeKind.Utc));
...

TIP
You can manually remove an item from the cache by using the DEL command, which is available through the StackExchange
library as the IDatabase.KeyDeleteAsync method.

Use tags to cross-correlate cached items

A Redis set is a collection of multiple items that share a single key. You can create a set by using the SADD
command. You can retrieve the items in a set by using the SMEMBERS command. The StackExchange library
implements the SADD command with the IDatabase.SetAddAsync method, and the SMEMBERS command with the
IDatabase.SetMembersAsync method.

You can also combine existing sets to create new sets by using the SDIFF (set difference), SINTER (set intersection),
and SUNION (set union) commands. The StackExchange library unifies these operations in the
IDatabase.SetCombineAsync method. The first parameter to this method specifies the set operation to perform.

The following code snippets show how sets can be useful for quickly storing and retrieving collections of related
items. This code uses the BlogPost type that was described in the section Implement Redis Cache Client
Applications earlier in this article.
A BlogPost object contains four fields—an ID, a title, a ranking score, and a collection of tags. The first code
snippet below shows the sample data that's used for populating a C# list of BlogPost objects:
 List<string[]> tags = new List<string[]>
{
new[] { "iot","csharp" },
new[] { "iot","azure","csharp" },
new[] { "csharp","git","big data" },
new[] { "iot","git","database" },
new[] { "database","git" },
new[] { "csharp","database" },
new[] { "iot" },
new[] { "iot","database","git" },
new[] { "azure","database","big data","git","csharp" },
new[] { "azure" }
};

List<BlogPost> posts = new List<BlogPost>();

int blogKey = 1;
int numberOfPosts = 20;
Random random = new Random();
for (int i = 0; i < numberOfPosts; i++)
{
blogKey++;
posts.Add(new BlogPost(
blogKey, // Blog post ID
string.Format(CultureInfo.InvariantCulture, "Blog Post #{0}",
blogKey), // Blog post title
random.Next(100, 10000), // Ranking score
tags[i % tags.Count])); // Tags--assigned from a collection
// in the tags list
}

You can store the tags for each BlogPost object as a set in a Redis cache and associate each set with the ID of the
BlogPost . This enables an application to quickly find all the tags that belong to a specific blog post. To enable
searching in the opposite direction and find all blog posts that share a specific tag, you can create another set that
holds the blog posts referencing the tag ID in the key:

ConnectionMultiplexer redisHostConnection = ...;

IDatabase cache = redisHostConnection.GetDatabase();
...
// Tags are easily represented as Redis Sets
foreach (BlogPost post in posts)
{
string redisKey = string.Format(CultureInfo.InvariantCulture,
"blog:posts:{0}:tags", post.Id);
// Add tags to the blog post in Redis
await cache.SetAddAsync(
redisKey, post.Tags.Select(s => (RedisValue)s).ToArray());

// Now do the inverse so we can figure out which blog posts have a given tag
foreach (var tag in post.Tags)
{
await cache.SetAddAsync(string.Format(CultureInfo.InvariantCulture,
"tag:{0}:blog:posts", tag), post.Id);
}
}

These structures enable you to perform many common queries very efficiently. For example, you can find and
display all of the tags for blog post 1 like this:
 // Show the tags for blog post #1
foreach (var value in await cache.SetMembersAsync("blog:posts:1:tags"))
{
Console.WriteLine(value);
}

You can find all tags that are common to blog post 1 and blog post 2 by performing a set intersection operation,
as follows:

// Show the tags in common for blog posts #1 and #2

foreach (var value in await cache.SetCombineAsync(SetOperation.Intersect, new RedisKey[]
{ "blog:posts:1:tags", "blog:posts:2:tags" }))
{
Console.WriteLine(value);
}

And you can find all blog posts that contain a specific tag:

// Show the ids of the blog posts that have the tag "iot".
foreach (var value in await cache.SetMembersAsync("tag:iot:blog:posts"))
{
Console.WriteLine(value);
}

Find recently accessed items

A common task required of many applications is to find the most recently accessed items. For example, a blogging
site might want to display information about the most recently read blog posts.
You can implement this functionality by using a Redis list. A Redis list contains multiple items that share the same
key. The list acts as a double-ended queue. You can push items to either end of the list by using the LPUSH (left
push) and RPUSH (right push) commands. You can retrieve items from either end of the list by using the LPOP
and RPOP commands. You can also return a set of elements by using the LRANGE and RRANGE commands.
The code snippets below show how you can perform these operations by using the StackExchange library. This
code uses the BlogPost type from the previous examples. As a blog post is read by a user, the
IDatabase.ListLeftPushAsync method pushes the title of the blog post onto a list that's associated with the key
"blog:recent_posts" in the Redis cache.

ConnectionMultiplexer redisHostConnection = ...;

IDatabase cache = redisHostConnection.GetDatabase();
...
string redisKey = "blog:recent_posts";
BlogPost blogPost = ...; // Reference to the blog post that has just been read
await cache.ListLeftPushAsync(
redisKey, blogPost.Title); // Push the blog post onto the list

As more blog posts are read, their titles are pushed onto the same list. The list is ordered by the sequence in which
the titles have been added. The most recently read blog posts are toward the left end of the list. (If the same blog
post is read more than once, it will have multiple entries in the list.)
You can display the titles of the most recently read posts by using the IDatabase.ListRange method. This method
takes the key that contains the list, a starting point, and an ending point. The following code retrieves the titles of
the 10 blog posts (items from 0 to 9) at the left-most end of the list:
 // Show latest ten posts
foreach (string postTitle in await cache.ListRangeAsync(redisKey, 0, 9))
{
Console.WriteLine(postTitle);
}

Note that the ListRangeAsync method does not remove items from the list. To do this, you can use the
IDatabase.ListLeftPopAsync and IDatabase.ListRightPopAsync methods.
To prevent the list from growing indefinitely, you can periodically cull items by trimming the list. The code snippet
below shows you how to remove all but the five left-most items from the list:

await cache.ListTrimAsync(redisKey, 0, 5);

Implement a leader board

By default, the items in a set are not held in any specific order. You can create an ordered set by using the ZADD
command (the IDatabase.SortedSetAdd method in the StackExchange library). The items are ordered by using a
numeric value called a score, which is provided as a parameter to the command.
The following code snippet adds the title of a blog post to an ordered list. In this example, each blog post also has
a score field that contains the ranking of the blog post.

ConnectionMultiplexer redisHostConnection = ...;

IDatabase cache = redisHostConnection.GetDatabase();
...
string redisKey = "blog:post_rankings";
BlogPost blogPost = ...; // Reference to a blog post that has just been rated
await cache.SortedSetAddAsync(redisKey, blogPost.Title, blogPost.Score);

You can retrieve the blog post titles and scores in ascending score order by using the
IDatabase.SortedSetRangeByRankWithScores method:

foreach (var post in await cache.SortedSetRangeByRankWithScoresAsync(redisKey))

{
Console.WriteLine(post);
}

NOTE
The StackExchange library also provides the IDatabase.SortedSetRangeByRankAsync method, which returns the data in
score order, but does not return the scores.

You can also retrieve items in descending order of scores, and limit the number of items that are returned by
providing additional parameters to the IDatabase.SortedSetRangeByRankWithScoresAsync method. The next example
displays the titles and scores of the top 10 ranked blog posts:

foreach (var post in await cache.SortedSetRangeByRankWithScoresAsync(

redisKey, 0, 9, Order.Descending))
{
Console.WriteLine(post);
}

The next example uses the IDatabase.SortedSetRangeByScoreWithScoresAsync method, which you can use to limit
the items that are returned to those that fall within a given score range:

// Blog posts with scores between 5000 and 100000

foreach (var post in await cache.SortedSetRangeByScoreWithScoresAsync(
redisKey, 5000, 100000))
{
Console.WriteLine(post);
}

Message by using channels

Apart from acting as a data cache, a Redis server provides messaging through a high-performance
publisher/subscriber mechanism. Client applications can subscribe to a channel, and other applications or services
can publish messages to the channel. Subscribing applications will then receive these messages and can process
them.
Redis provides the SUBSCRIBE command for client applications to use to subscribe to channels. This command
expects the name of one or more channels on which the application will accept messages. The StackExchange
library includes the ISubscription interface, which enables a .NET Framework application to subscribe and
publish to channels.
You create an ISubscription object by using the GetSubscriber method of the connection to the Redis server.
Then you listen for messages on a channel by using the SubscribeAsync method of this object. The following code
example shows how to subscribe to a channel named "messages:blogPosts":

ConnectionMultiplexer redisHostConnection = ...;

ISubscriber subscriber = redisHostConnection.GetSubscriber();
...
await subscriber.SubscribeAsync("messages:blogPosts", (channel, message) => Console.WriteLine("Title is:
{0}", message));

The first parameter to the Subscribe method is the name of the channel. This name follows the same conventions
that are used by keys in the cache. The name can contain any binary data, although it is advisable to use relatively
short, meaningful strings to help ensure good performance and maintainability.
Note also that the namespace used by channels is separate from that used by keys. This means you can have
channels and keys that have the same name, although this may make your application code more difficult to
maintain.
The second parameter is an Action delegate. This delegate runs asynchronously whenever a new message
appears on the channel. This example simply displays the message on the console (the message will contain the
title of a blog post).
To publish to a channel, an application can use the Redis PUBLISH command. The StackExchange library provides
the IServer.PublishAsync method to perform this operation. The next code snippet shows how to publish a
message to the "messages:blogPosts" channel:

ConnectionMultiplexer redisHostConnection = ...;

ISubscriber subscriber = redisHostConnection.GetSubscriber();
...
BlogPost blogPost = ...;
subscriber.PublishAsync("messages:blogPosts", blogPost.Title);

There are several points you should understand about the publish/subscribe mechanism:
Multiple subscribers can subscribe to the same channel, and they will all receive the messages that are
published to that channel.
 Subscribers only receive messages that have been published after they have subscribed. Channels are not
buffered, and once a message is published, the Redis infrastructure pushes the message to each subscriber and
then removes it.
By default, messages are received by subscribers in the order in which they are sent. In a highly active system
with a large number of messages and many subscribers and publishers, guaranteed sequential delivery of
messages can slow performance of the system. If each message is independent and the order is unimportant,
you can enable concurrent processing by the Redis system, which can help to improve responsiveness. You can
achieve this in a StackExchange client by setting the PreserveAsyncOrder of the connection used by the
subscriber to false:

ConnectionMultiplexer redisHostConnection = ...;

redisHostConnection.PreserveAsyncOrder = false;
ISubscriber subscriber = redisHostConnection.GetSubscriber();

Serialization considerations
When you choose a serialization format, consider tradeoffs between performance, interoperability, versioning,
compatibility with existing systems, data compression, and memory overhead. When you are evaluating
performance, remember that benchmarks are highly dependent on context. They may not reflect your actual
workload, and may not consider newer libraries or versions. There is no single "fastest" serializer for all scenarios.
Some options to consider include:
Protocol Buffers (also called protobuf) is a serialization format developed by Google for serializing
structured data efficiently. It uses strongly typed definition files to define message structures. These
definition files are then compiled to language-specific code for serializing and deserializing messages.
Protobuf can be used over existing RPC mechanisms, or it can generate an RPC service.
Apache Thrift uses a similar approach, with strongly typed definition files and a compilation step to
generate the serialization code and RPC services.
Apache Avro provides similar functionality to Protocol Buffers and Thrift, but there is no compilation step.
Instead, serialized data always includes a schema that describes the structure.
JSON is an open standard that uses human-readable text fields. It has broad cross-platform support. JSON
does not use message schemas. Being a text-based format, it is not very efficient over the wire. In some
cases, however, you may be returning cached items directly to a client via HTTP, in which case storing JSON
could save the cost of deserializing from another format and then serializing to JSON.
BSON is a binary serialization format that uses a structure similar to JSON. BSON was designed to be
lightweight, easy to scan, and fast to serialize and deserialize, relative to JSON. Payloads are comparable in
size to JSON. Depending on the data, a BSON payload may be smaller or larger than a JSON payload.
BSON has some additional data types that are not available in JSON, notably BinData (for byte arrays) and
Date.
MessagePack is a binary serialization format that is designed to be compact for transmission over the wire.
There are no message schemas or message type checking.
Bond is a cross-platform framework for working with schematized data. It supports cross-language
serialization and deserialization. Notable differences from other systems listed here are support for
inheritance, type aliases, and generics.
gRPC is an open-source RPC system developed by Google. By default, it uses Protocol Buffers as its
definition language and underlying message interchange format.

Related patterns and guidance

The following patterns might also be relevant to your scenario when you implement caching in your applications:
Cache-aside pattern: This pattern describes how to load data on demand into a cache from a data store.
This pattern also helps to maintain consistency between data that's held in the cache and the data in the
original data store.
The Sharding pattern provides information about implementing horizontal partitioning to help improve
scalability when storing and accessing large volumes of data.

More information
Azure Cache for Redis documentation
Azure Cache for Redis FAQ
Task-based Asynchronous pattern
Redis documentation
StackExchange.Redis
Data partitioning guide
 Best practices for using content delivery networks
(CDNs)
12/18/2020 • 8 minutes to read • Edit Online

A content delivery network (CDN) is a distributed network of servers that can efficiently deliver web content to
users. CDNs store cached content on edge servers that are close to end users to minimize latency.
CDNs are typically used to deliver static content such as images, style sheets, documents, client-side scripts, and
HTML pages. The major advantages of using a CDN are lower latency and faster delivery of content to users,
regardless of their geographical location in relation to the datacenter where the application is hosted. CDNs can
also help to reduce load on a web application, because the application does not have to service requests for the
content that is hosted in the CDN.

In Azure, the Azure Content Delivery Network is a global CDN solution for delivering high-bandwidth content that
is hosted in Azure or any other location. Using Azure CDN, you can cache publicly available objects loaded from
Azure blob storage, a web application, virtual machine, any publicly accessible web server.
This topic describes some general best practices and considerations when using a CDN. For more information, see
Azure CDN.

How and why a CDN is used

Typical uses for a CDN include:
Delivering static resources for client applications, often from a website. These resources can be images, style
sheets, documents, files, client-side scripts, HTML pages, HTML fragments, or any other content that the
server does not need to modify for each request. The application can create items at runtime and make
them available to the CDN (for example, by creating a list of current news headlines), but it does not do so
for each request.
Delivering public static and shared content to devices such as mobile phones and tablet computers. The
application itself is a web service that offers an API to clients running on the various devices. The CDN can
also deliver static datasets (via the web service) for the clients to use, perhaps to generate the client UI. For
example, the CDN could be used to distribute JSON or XML documents.
Serving entire websites that consist of only public static content to clients, without requiring any dedicated
compute resources.
Streaming video files to the client on demand. Video benefits from the low latency and reliable connectivity
 available from the globally located datacenters that offer CDN connections. Microsoft Azure Media Services
(AMS) integrates with Azure CDN to deliver content directly to the CDN for further distribution. For more
information, see Streaming endpoints overview.
Generally improving the experience for users, especially those located far from the datacenter hosting the
application. These users might otherwise suffer higher latency. A large proportion of the total size of the
content in a web application is often static, and using the CDN can help to maintain performance and
overall user experience while eliminating the requirement to deploy the application to multiple datacenters.
For a list of Azure CDN node locations, see Azure CDN POP Locations.
Supporting IoT (Internet of Things) solutions. The huge numbers of devices and appliances involved in an
IoT solution could easily overwhelm an application if it had to distribute firmware updates directly to each
device.
Coping with peaks and surges in demand without requiring the application to scale, avoiding the
consequent increase in running costs. For example, when an update to an operating system is released for a
hardware device such as a specific model of router, or for a consumer device such as a smart TV, there will
be a huge peak in demand as it is downloaded by millions of users and devices over a short period.

Challenges
There are several challenges to take into account when planning to use a CDN.
Deployment . Decide the origin from which the CDN fetches the content, and whether you need to deploy
the content in more than one storage system. Take into account the process for deploying static content and
resources. For example, you may need to implement a separate step to load content into Azure blob
storage.
Versioning and cache-control . Consider how you will update static content and deploy new versions.
Understand how the CDN performs caching and time-to-live (TTL). For Azure CDN, see How caching works.
Testing . It can be difficult to perform local testing of your CDN settings when developing and testing an
application locally or in a staging environment.
Search engine optimization (SEO) . Content such as images and documents are served from a different
domain when you use the CDN. This can have an effect on SEO for this content.
Content security . Not all CDNs offer any form of access control for the content. Some CDN services,
including Azure CDN, support token-based authentication to protect CDN content. For more information,
see Securing Azure Content Delivery Network assets with token authentication.
Client security . Clients might connect from an environment that does not allow access to resources on the
CDN. This could be a security-constrained environment that limits access to only a set of known sources, or
one that prevents loading of resources from anything other than the page origin. A fallback implementation
is required to handle these cases.
Resilience . The CDN is a potential single point of failure for an application.
Scenarios where a CDN may be less useful include:
If the content has a low hit rate, it might be accessed only few times while it is valid (determined by its time-
to-live setting).
If the data is private, such as for large enterprises or supply chain ecosystems.

General guidelines and good practices

Using a CDN is a good way to minimize the load on your application, and maximize availability and performance.
Consider adopting this strategy for all of the appropriate content and resources your application uses. Consider
the points in the following sections when designing your strategy to use a CDN.
Deployment
Static content may need to be provisioned and deployed independently from the application if you do not include
it in the application deployment package or process. Consider how this will affect the versioning approach you use
to manage both the application components and the static resource content.
Consider using bundling and minification techniques to reduce load times for clients. Bundling combines multiple
files into a single file. Minification removes unnecessary characters from scripts and CSS files without altering
functionality.
If you need to deploy the content to an additional location, this will be an extra step in the deployment process. If
the application updates the content for the CDN, perhaps at regular intervals or in response to an event, it must
store the updated content in any additional locations as well as the endpoint for the CDN.
Consider how you will handle local development and testing when some static content is expected to be served
from a CDN. For example, you could predeploy the content to the CDN as part of your build script. Alternatively,
use compile directives or flags to control how the application loads the resources. For example, in debug mode, the
application could load static resources from a local folder. In release mode, the application would use the CDN.
Consider the options for file compression, such as gzip (GNU zip). Compression may be performed on the origin
server by the web application hosting or directly on the edge servers by the CDN. For more information, see
Improve performance by compressing files in Azure CDN.
Routing and versioning
You may need to use different CDN instances at various times. For example, when you deploy a new version of the
application you may want to use a new CDN and retain the old CDN (holding content in an older format) for
previous versions. If you use Azure blob storage as the content origin, you can create a separate storage account
or a separate container and point the CDN endpoint to it.
Do not use the query string to denote different versions of the application in links to resources on the CDN
because, when retrieving content from Azure blob storage, the query string is part of the resource name (the blob
name). This approach can also affect how the client caches resources.
Deploying new versions of static content when you update an application can be a challenge if the previous
resources are cached on the CDN. For more information, see the section on cache control, below.
Consider restricting the CDN content access by country/region. Azure CDN allows you to filter requests based on
the country or region of origin and restrict the content delivered. For more information, see Restrict access to your
content by country/region.
Cache control
Consider how to manage caching within the system. For example, in Azure CDN, you can set global caching rules,
and then set custom caching for particular origin endpoints. You can also control how caching is performed in a
CDN by sending cache-directive headers at the origin.
For more information, see How caching works.
To prevent objects from being available on the CDN, you can delete them from the origin, remove or delete the
CDN endpoint, or in the case of blob storage, make the container or blob private. However, items are not removed
from the CDN until the time-to-live expires. You can also manually purge a CDN endpoint.
Security
The CDN can deliver content over HTTPS (SSL), by using the certificate provided by the CDN, as well as over
standard HTTP. To avoid browser warnings about mixed content, you might need to use HTTPS to request static
content that is displayed in pages loaded through HTTPS.
If you deliver static assets such as font files by using the CDN, you might encounter same-origin policy issues if
you use an XMLHttpRequest call to request these resources from a different domain. Many web browsers prevent
cross-origin resource sharing (CORS) unless the web server is configured to set the appropriate response headers.
You can configure the CDN to support CORS by using one of the following methods:
Configure the CDN to add CORS headers to the responses. For more information, see Using Azure CDN
with CORS.
If the origin is Azure blob storage, add CORS rules to the storage endpoint. For more information, see
Cross-Origin Resource Sharing (CORS) Support for the Azure Storage Services.
Configure the application to set the CORS headers. For example, see Enabling Cross-Origin Requests
(CORS) in the ASP.NET Core documentation.
CDN fallback
Consider how your application will cope with a failure or temporary unavailability of the CDN. Client applications
may be able to use copies of the resources that were cached locally (on the client) during previous requests, or you
can include code that detects failure and instead requests resources from the origin (the application folder or
Azure blob container that holds the resources) if the CDN is unavailable.
 Horizontal, vertical, and functional data partitioning
12/18/2020 • 17 minutes to read • Edit Online

In many large-scale solutions, data is divided into partitions that can be managed and accessed separately.
Partitioning can improve scalability, reduce contention, and optimize performance. It can also provide a
mechanism for dividing data by usage pattern. For example, you can archive older data in cheaper data storage.
However, the partitioning strategy must be chosen carefully to maximize the benefits while minimizing adverse
effects.

NOTE
In this article, the term partitioning means the process of physically dividing data into separate data stores. It is not the
same as SQL Server table partitioning.

Why partition data?

Improve scalability . When you scale up a single database system, it will eventually reach a physical
hardware limit. If you divide data across multiple partitions, each hosted on a separate server, you can
scale out the system almost indefinitely.
Improve performance . Data access operations on each partition take place over a smaller volume of
data. Correctly done, partitioning can make your system more efficient. Operations that affect more than
one partition can run in parallel.
Improve security . In some cases, you can separate sensitive and nonsensitive data into different
partitions and apply different security controls to the sensitive data.
Provide operational flexibility . Partitioning offers many opportunities for fine-tuning operations,
maximizing administrative efficiency, and minimizing cost. For example, you can define different strategies
for management, monitoring, backup and restore, and other administrative tasks based on the
importance of the data in each partition.
Match the data store to the pattern of use . Partitioning allows each partition to be deployed on a
different type of data store, based on cost and the built-in features that data store offers. For example,
large binary data can be stored in blob storage, while more structured data can be held in a document
database. See Choose the right data store.
Improve availability . Separating data across multiple servers avoids a single point of failure. If one
instance fails, only the data in that partition is unavailable. Operations on other partitions can continue.
For managed PaaS data stores, this consideration is less relevant, because these services are designed
with built-in redundancy.

Designing partitions
There are three typical strategies for partitioning data:
Horizontal par titioning (often called sharding). In this strategy, each partition is a separate data store,
but all partitions have the same schema. Each partition is known as a shard and holds a specific subset of
the data, such as all the orders for a specific set of customers.
Ver tical par titioning . In this strategy, each partition holds a subset of the fields for items in the data
 store. The fields are divided according to their pattern of use. For example, frequently accessed fields
might be placed in one vertical partition and less frequently accessed fields in another.
Functional par titioning . In this strategy, data is aggregated according to how it is used by each
bounded context in the system. For example, an e-commerce system might store invoice data in one
partition and product inventory data in another.
These strategies can be combined, and we recommend that you consider them all when you design a
partitioning scheme. For example, you might divide data into shards and then use vertical partitioning to further
subdivide the data in each shard.
Horizontal partitioning (sharding)
Figure 1 shows horizontal partitioning or sharding. In this example, product inventory data is divided into shards
based on the product key. Each shard holds the data for a contiguous range of shard keys (A-G and H-Z),
organized alphabetically. Sharding spreads the load over more computers, which reduces contention and
improves performance.

Figure 1 - Horizontally partitioning (sharding) data based on a partition key.

The most important factor is the choice of a sharding key. It can be difficult to change the key after the system is
in operation. The key must ensure that data is partitioned to spread the workload as evenly as possible across
the shards.
The shards don't have to be the same size. It's more important to balance the number of requests. Some shards
might be very large, but each item has a low number of access operations. Other shards might be smaller, but
each item is accessed much more frequently. It's also important to ensure that a single shard does not exceed the
scale limits (in terms of capacity and processing resources) of the data store.
Avoid creating "hot" partitions that can affect performance and availability. For example, using the first letter of a
customer's name causes an unbalanced distribution, because some letters are more common. Instead, use a hash
of a customer identifier to distribute data more evenly across partitions.
Choose a sharding key that minimizes any future requirements to split large shards, coalesce small shards into
larger partitions, or change the schema. These operations can be very time consuming, and might require taking
one or more shards offline while they are performed.
If shards are replicated, it might be possible to keep some of the replicas online while others are split, merged, or
reconfigured. However, the system might need to limit the operations that can be performed during the
reconfiguration. For example, the data in the replicas might be marked as read-only to prevent data
inconsistences.
For more information about horizontal partitioning, see sharding pattern.
Vertical partitioning
The most common use for vertical partitioning is to reduce the I/O and performance costs associated with
fetching items that are frequently accessed. Figure 2 shows an example of vertical partitioning. In this example,
different properties of an item are stored in different partitions. One partition holds data that is accessed more
frequently, including product name, description, and price. Another partition holds inventory data: the stock
count and last-ordered date.

Figure 2 - Vertically partitioning data by its pattern of use.

In this example, the application regularly queries the product name, description, and price when displaying the
product details to customers. Stock count and last- ordered date are held in a separate partition because these
two items are commonly used together.
Other advantages of vertical partitioning:
Relatively slow-moving data (product name, description, and price) can be separated from the more
dynamic data (stock level and last ordered date). Slow moving data is a good candidate for an application
to cache in memory.
Sensitive data can be stored in a separate partition with additional security controls.
Vertical partitioning can reduce the amount of concurrent access that's needed.
Vertical partitioning operates at the entity level within a data store, partially normalizing an entity to break it
down from a wide item to a set of narrow items. It is ideally suited for column-oriented data stores such as
HBase and Cassandra. If the data in a collection of columns is unlikely to change, you can also consider using
column stores in SQL Server.
Functional partitioning
When it's possible to identify a bounded context for each distinct business area in an application, functional
partitioning is a way to improve isolation and data access performance. Another common use for functional
partitioning is to separate read-write data from read-only data. Figure 3 shows an overview of functional
partitioning where inventory data is separated from customer data.
Figure 3 - Functionally partitioning data by bounded context or subdomain.
This partitioning strategy can help reduce data access contention across different parts of a system.

Designing partitions for scalability

It's vital to consider size and workload for each partition and balance them so that data is distributed to achieve
maximum scalability. However, you must also partition the data so that it does not exceed the scaling limits of a
single partition store.
Follow these steps when designing partitions for scalability:
1. Analyze the application to understand the data access patterns, such as the size of the result set returned by
each query, the frequency of access, the inherent latency, and the server-side compute processing
requirements. In many cases, a few major entities will demand most of the processing resources.
2. Use this analysis to determine the current and future scalability targets, such as data size and workload. Then
distribute the data across the partitions to meet the scalability target. For horizontal partitioning, choosing the
right shard key is important to make sure distribution is even. For more information, see the sharding pattern.
3. Make sure each partition has enough resources to handle the scalability requirements, in terms of data size
and throughput. Depending on the data store, there might be a limit on the amount of storage space,
processing power, or network bandwidth per partition. If the requirements are likely to exceed these limits,
you may need to refine your partitioning strategy or split data out further, possibly combining two or more
strategies.
4. Monitor the system to verify that data is distributed as expected and that the partitions can handle the load.
Actual usage does not always match what an analysis predicts. If so, it might be possible to rebalance the
partitions, or else redesign some parts of the system to gain the required balance.
Some cloud environments allocate resources in terms of infrastructure boundaries. Ensure that the limits of your
selected boundary provide enough room for any anticipated growth in the volume of data, in terms of data
storage, processing power, and bandwidth.
For example, if you use Azure table storage, there is a limit to the volume of requests that can be handled by a
single partition in a particular period of time. (For more information, see Azure storage scalability and
performance targets.) A busy shard might require more resources than a single partition can handle. If so, the
shard might need to be repartitioned to spread the load. If the total size or throughput of these tables exceeds
the capacity of a storage account, you might need to create additional storage accounts and spread the tables
across these accounts.
Designing partitions for query performance
Query performance can often be boosted by using smaller data sets and by running parallel queries. Each
partition should contain a small proportion of the entire data set. This reduction in volume can improve the
performance of queries. However, partitioning is not an alternative for designing and configuring a database
appropriately. For example, make sure that you have the necessary indexes in place.
Follow these steps when designing partitions for query performance:
1. Examine the application requirements and performance:
Use business requirements to determine the critical queries that must always perform quickly.
Monitor the system to identify any queries that perform slowly.
Find which queries are performed most frequently. Even if a single query has a minimal cost, the
cumulative resource consumption could be significant.
2. Partition the data that is causing slow performance:
Limit the size of each partition so that the query response time is within target.
If you use horizontal partitioning, design the shard key so that the application can easily select the right
partition. This prevents the query from having to scan through every partition.
Consider the location of a partition. If possible, try to keep data in partitions that are geographically
close to the applications and users that access it.
3. If an entity has throughput and query performance requirements, use functional partitioning based on
that entity. If this still doesn't satisfy the requirements, apply horizontal partitioning as well. In most cases,
a single partitioning strategy will suffice, but in some cases it is more efficient to combine both strategies.
4. Consider running queries in parallel across partitions to improve performance.

Designing partitions for availability

Partitioning data can improve the availability of applications by ensuring that the entire dataset does not
constitute a single point of failure and that individual subsets of the dataset can be managed independently.
Consider the following factors that affect availability:
How critical the data is to business operations . Identify which data is critical business information, such as
transactions, and which data is less critical operational data, such as log files.
Consider storing critical data in highly available partitions with an appropriate backup plan.
Establish separate management and monitoring procedures for the different datasets.
Place data that has the same level of criticality in the same partition so that it can be backed up together at
an appropriate frequency. For example, partitions that hold transaction data might need to be backed up
more frequently than partitions that hold logging or trace information.
How individual par titions can be managed . Designing partitions to support independent management and
maintenance provides several advantages. For example:
If a partition fails, it can be recovered independently without applications that access data in other
partitions.
Partitioning data by geographical area allows scheduled maintenance tasks to occur at off-peak hours for
each location. Ensure that partitions are not too large to prevent any planned maintenance from being
completed during this period.
Whether to replicate critical data across par titions . This strategy can improve availability and
performance, but can also introduce consistency issues. It takes time to synchronize changes with every replica.
During this period, different partitions will contain different data values.

Application design considerations

Partitioning adds complexity to the design and development of your system. Consider partitioning as a
fundamental part of system design even if the system initially only contains a single partition. If you address
partitioning as an afterthought, it will be more challenging because you already have a live system to maintain:
Data access logic will need to be modified.
Large quantities of existing data may need to be migrated, to distribute it across partitions.
Users expect to be able to continue using the system during the migration.
In some cases, partitioning is not considered important because the initial dataset is small and can be easily
handled by a single server. This might be true for some workloads, but many commercial systems need to
expand as the number of users increases.
Moreover, it's not only large data stores that benefit from partitioning. For example, a small data store might be
heavily accessed by hundreds of concurrent clients. Partitioning the data in this situation can help to reduce
contention and improve throughput.
Consider the following points when you design a data partitioning scheme:
Minimize cross-par tition data access operations . Where possible, keep data for the most common
database operations together in each partition to minimize cross-partition data access operations. Querying
across partitions can be more time-consuming than querying within a single partition, but optimizing partitions
for one set of queries might adversely affect other sets of queries. If you must query across partitions, minimize
query time by running parallel queries and aggregating the results within the application. (This approach might
not be possible in some cases, such as when the result from one query is used in the next query.)
Consider replicating static reference data. If queries use relatively static reference data, such as postal code
tables or product lists, consider replicating this data in all of the partitions to reduce separate lookup operations
in different partitions. This approach can also reduce the likelihood of the reference data becoming a "hot"
dataset, with heavy traffic from across the entire system. However, there is an additional cost associated with
synchronizing any changes to the reference data.
Minimize cross-par tition joins. Where possible, minimize requirements for referential integrity across
vertical and functional partitions. In these schemes, the application is responsible for maintaining referential
integrity across partitions. Queries that join data across multiple partitions are inefficient because the application
typically needs to perform consecutive queries based on a key and then a foreign key. Instead, consider
replicating or de-normalizing the relevant data. If cross-partition joins are necessary, run parallel queries over
the partitions and join the data within the application.
Embrace eventual consistency . Evaluate whether strong consistency is actually a requirement. A common
approach in distributed systems is to implement eventual consistency. The data in each partition is updated
separately, and the application logic ensures that the updates are all completed successfully. It also handles the
inconsistencies that can arise from querying data while an eventually consistent operation is running.
Consider how queries locate the correct par tition . If a query must scan all partitions to locate the required
data, there is a significant impact on performance, even when multiple parallel queries are running. With vertical
and functional partitioning, queries can naturally specify the partition. Horizontal partitioning, on the other hand,
can make locating an item difficult, because every shard has the same schema. A typical solution to maintain a
map that is used to look up the shard location for specific items. This map can be implemented in the sharding
logic of the application, or maintained by the data store if it supports transparent sharding.
Consider periodically rebalancing shards . With horizontal partitioning, rebalancing shards can help
distribute the data evenly by size and by workload to minimize hotspots, maximize query performance, and work
around physical storage limitations. However, this is a complex task that often requires the use of a custom tool
or process.
Replicate par titions. If you replicate each partition, it provides additional protection against failure. If a single
replica fails, queries can be directed toward a working copy.
If you reach the physical limits of a par titioning strategy, you might need to extend the scalability
to a different level . For example, if partitioning is at the database level, you might need to locate or replicate
partitions in multiple databases. If partitioning is already at the database level, and physical limitations are an
issue, it might mean that you need to locate or replicate partitions in multiple hosting accounts.
Avoid transactions that access data in multiple par titions . Some data stores implement transactional
consistency and integrity for operations that modify data, but only when the data is located in a single partition.
If you need transactional support across multiple partitions, you will probably need to implement this as part of
your application logic because most partitioning systems do not provide native support.
All data stores require some operational management and monitoring activity. The tasks can range from loading
data, backing up and restoring data, reorganizing data, and ensuring that the system is performing correctly and
efficiently.
Consider the following factors that affect operational management:
How to implement appropriate management and operational tasks when the data is
par titioned . These tasks might include backup and restore, archiving data, monitoring the system, and
other administrative tasks. For example, maintaining logical consistency during backup and restore
operations can be a challenge.
How to load the data into multiple par titions and add new data that's arriving from other
sources . Some tools and utilities might not support sharded data operations such as loading data into
the correct partition.
How to archive and delete the data on a regular basis . To prevent the excessive growth of
partitions, you need to archive and delete data on a regular basis (such as monthly). It might be necessary
to transform the data to match a different archive schema.
How to locate data integrity issues . Consider running a periodic process to locate any data integrity
issues, such as data in one partition that references missing information in another. The process can either
attempt to fix these issues automatically or generate a report for manual review.

Rebalancing partitions
As a system matures, you might have to adjust the partitioning scheme. For example, individual partitions might
start getting a disproportionate volume of traffic and become hot, leading to excessive contention. Or you might
have underestimated the volume of data in some partitions, causing some partitions to approach capacity limits.
Some data stores, such as Cosmos DB, can automatically rebalance partitions. In other cases, rebalancing is an
administrative task that consists of two stages:
1. Determine a new partitioning strategy.
Which partitions need to be split (or possibly combined)?
What is the new partition key?
2. Migrate data from the old partitioning scheme to the new set of partitions.
Depending on the data store, you might be able to migrate data between partitions while they are in use. This is
called online migration. If that's not possible, you might need to make partitions unavailable while the data is
relocated (offline migration).
Offline migration
Offline migration is typically simpler because it reduces the chances of contention occurring. Conceptually, offline
migration works as follows:
1. Mark the partition offline.
2. Split-merge and move the data to the new partitions.
3. Verify the data.
4. Bring the new partitions online.
5. Remove the old partition.
Optionally, you can mark a partition as read-only in step 1, so that applications can still read the data while it is
being moved.

Online migration
Online migration is more complex to perform but less disruptive. The process is similar to offline migration,
except the original partition is not marked offline. Depending on the granularity of the migration process (for
example, item by item versus shard by shard), the data access code in the client applications might have to
handle reading and writing data that's held in two locations, the original partition and the new partition.

Related patterns
The following design patterns might be relevant to your scenario:
The sharding pattern describes some common strategies for sharding data.
The index table pattern shows how to create secondary indexes over data. An application can quickly
retrieve data with this approach, by using queries that do not reference the primary key of a collection.
The materialized view pattern describes how to generate prepopulated views that summarize data to
support fast query operations. This approach can be useful in a partitioned data store if the partitions that
contain the data being summarized are distributed across multiple sites.

Next steps
Learn about partitioning strategies for specific Azure services. See Data partitioning strategies
 Data partitioning strategies
12/18/2020 • 31 minutes to read • Edit Online

This article describes some strategies for partitioning data in various Azure data stores. For general guidance about
when to partition data and best practices, see Data partitioning.

Partitioning Azure SQL Database

A single SQL database has a limit to the volume of data that it can contain. Throughput is constrained by
architectural factors and the number of concurrent connections that it supports.
Elastic pools support horizontal scaling for a SQL database. Using elastic pools, you can partition your data into
shards that are spread across multiple SQL databases. You can also add or remove shards as the volume of data
that you need to handle grows and shrinks. Elastic pools can also help reduce contention by distributing the load
across databases.
Each shard is implemented as a SQL database. A shard can hold more than one dataset (called a shardlet). Each
database maintains metadata that describes the shardlets that it contains. A shardlet can be a single data item, or a
group of items that share the same shardlet key. For example, in a multitenant application, the shardlet key can be
the tenant ID, and all data for a tenant can be held in the same shardlet.
Client applications are responsible for associating a dataset with a shardlet key. A separate SQL database acts as a
global shard map manager. This database has a list of all the shards and shardlets in the system. The application
connects to the shard map manager database to obtain a copy of the shard map. It caches the shard map locally,
and uses the map to route data requests to the appropriate shard. This functionality is hidden behind a series of
APIs that are contained in the Elastic Database client library, which is available for Java and .NET.
For more information about elastic pools, see Scaling out with Azure SQL Database.
To reduce latency and improve availability, you can replicate the global shard map manager database. With the
Premium pricing tiers, you can configure active geo-replication to continuously copy data to databases in different
regions.
Alternatively, use Azure SQL Data Sync or Azure Data Factory to replicate the shard map manager database across
regions. This form of replication runs periodically and is more suitable if the shard map changes infrequently, and
does not require Premium tier.
Elastic Database provides two schemes for mapping data to shardlets and storing them in shards:
A list shard map associates a single key to a shardlet. For example, in a multitenant system, the data for
each tenant can be associated with a unique key and stored in its own shardlet. To guarantee isolation, each
shardlet can be held within its own shard.
 A range shard map associates a set of contiguous key values to a shardlet. For example, you can group the
data for a set of tenants (each with their own key) within the same shardlet. This scheme is less expensive
than the first, because tenants share data storage, but has less isolation.

A single shard can contain the data for several shardlets. For example, you can use list shardlets to store data for
different non-contiguous tenants in the same shard. You can also mix range shardlets and list shardlets in the same
shard, although they will be addressed through different maps. The following diagram shows this approach:
Elastic pools make it possible to add and remove shards as the volume of data shrinks and grows. Client
applications can create and delete shards dynamically, and transparently update the shard map manager. However,
removing a shard is a destructive operation that also requires deleting all the data in that shard.
If an application needs to split a shard into two separate shards or combine shards, use the split-merge tool. This
tool runs as an Azure web service, and migrates data safely between shards.
The partitioning scheme can significantly affect the performance of your system. It can also affect the rate at which
shards have to be added or removed, or that data must be repartitioned across shards. Consider the following
points:
Group data that is used together in the same shard, and avoid operations that access data from multiple
shards. A shard is a SQL database in its own right, and cross-database joins must be performed on the client
side.
Although SQL Database does not support cross-database joins, you can use the Elastic Database tools to
perform multi-shard queries. A multi-shard query sends individual queries to each database and merges the
results.
Don't design a system that has dependencies between shards. Referential integrity constraints, triggers, and
 stored procedures in one database cannot reference objects in another.
If you have reference data that is frequently used by queries, consider replicating this data across shards.
This approach can remove the need to join data across databases. Ideally, such data should be static or slow-
moving, to minimize the replication effort and reduce the chances of it becoming stale.
Shardlets that belong to the same shard map should have the same schema. This rule is not enforced by
SQL Database, but data management and querying becomes very complex if each shardlet has a different
schema. Instead, create separate shard maps for each schema. Remember that data belonging to different
shardlets can be stored in the same shard.
Transactional operations are only supported for data within a shard, and not across shards. Transactions can
span shardlets as long as they are part of the same shard. Therefore, if your business logic needs to perform
transactions, either store the data in the same shard or implement eventual consistency.
Place shards close to the users that access the data in those shards. This strategy helps reduce latency.
Avoid having a mixture of highly active and relatively inactive shards. Try to spread the load evenly across
shards. This might require hashing the sharding keys. If you are geo-locating shards, make sure that the
hashed keys map to shardlets held in shards stored close to the users that access that data.

Partitioning Azure table storage

Azure table storage is a key-value store that's designed around partitioning. All entities are stored in a partition,
and partitions are managed internally by Azure table storage. Each entity stored in a table must provide a two-part
key that includes:
The par tition key . This is a string value that determines the partition where Azure table storage will place
the entity. All entities with the same partition key are stored in the same partition.
The row key . This is a string value that identifies the entity within the partition. All entities within a partition
are sorted lexically, in ascending order, by this key. The partition key/row key combination must be unique
for each entity and cannot exceed 1 KB in length.
If an entity is added to a table with a previously unused partition key, Azure table storage creates a new partition
for this entity. Other entities with the same partition key will be stored in the same partition.
This mechanism effectively implements an automatic scale-out strategy. Each partition is stored on the same server
in an Azure datacenter to help ensure that queries that retrieve data from a single partition run quickly.
Microsoft has published scalability targets for Azure Storage. If your system is likely to exceed these limits, consider
splitting entities into multiple tables. Use vertical partitioning to divide the fields into the groups that are most
likely to be accessed together.
The following diagram shows the logical structure of an example storage account. The storage account contains
three tables: Customer Info, Product Info, and Order Info.
Each table has multiple partitions.
In the Customer Info table, the data is partitioned according to the city where the customer is located. The row
key contains the customer ID.
In the Product Info table, products are partitioned by product category, and the row key contains the product
number.
In the Order Info table, the orders are partitioned by order date, and the row key specifies the time the order
was received. All data is ordered by the row key in each partition.
Consider the following points when you design your entities for Azure table storage:
Select a partition key and row key by how the data is accessed. Choose a partition key/row key combination
that supports the majority of your queries. The most efficient queries retrieve data by specifying the
partition key and the row key. Queries that specify a partition key and a range of row keys can be completed
by scanning a single partition. This is relatively fast because the data is held in row key order. If queries don't
specify which partition to scan, every partition must be scanned.
If an entity has one natural key, then use it as the partition key and specify an empty string as the row key. If
an entity has a composite key consisting of two properties, select the slowest changing property as the
partition key and the other as the row key. If an entity has more than two key properties, use a
concatenation of properties to provide the partition and row keys.
If you regularly perform queries that look up data by using fields other than the partition and row keys,
consider implementing the Index Table pattern, or consider using a different data store that supports
indexing, such as Cosmos DB.
If you generate partition keys by using a monotonic sequence (such as "0001", "0002", "0003") and each
partition only contains a limited amount of data, Azure table storage can physically group these partitions
together on the same server. Azure Storage assumes that the application is most likely to perform queries
 across a contiguous range of partitions (range queries) and is optimized for this case. However, this
approach can lead to hotspots, because all insertions of new entities are likely to be concentrated at one end
the contiguous range. It can also reduce scalability. To spread the load more evenly, consider hashing the
partition key.
Azure table storage supports transactional operations for entities that belong to the same partition. An
application can perform multiple insert, update, delete, replace, or merge operations as an atomic unit, as
long as the transaction doesn't include more than 100 entities and the payload of the request doesn't exceed
4 MB. Operations that span multiple partitions are not transactional, and might require you to implement
eventual consistency. For more information about table storage and transactions, see Performing entity
group transactions.
Consider the granularity of the partition key:
Using the same partition key for every entity results in a single partition that's held on one server.
This prevents the partition from scaling out and focuses the load on a single server. As a result, this
approach is only suitable for storing a small number of entities. However, it does ensure that all
entities can participate in entity group transactions.
Using a unique partition key for every entity causes the table storage service to create a separate
partition for each entity, possibly resulting in a large number of small partitions. This approach is
more scalable than using a single partition key, but entity group transactions are not possible. Also,
queries that fetch more than one entity might involve reading from more than one server. However, if
the application performs range queries, then using a monotonic sequence for the partition keys
might help to optimize these queries.
Sharing the partition key across a subset of entities makes it possible to group related entities in the
same partition. Operations that involve related entities can be performed by using entity group
transactions, and queries that fetch a set of related entities can be satisfied by accessing a single
server.
For more information, see Azure storage table design guide and Scalable partitioning strategy.

Partitioning Azure blob storage

Azure blob storage makes it possible to hold large binary objects. Use block blobs in scenarios when you need to
upload or download large volumes of data quickly. Use page blobs for applications that require random rather than
serial access to parts of the data.
Each blob (either block or page) is held in a container in an Azure storage account. You can use containers to group
related blobs that have the same security requirements. This grouping is logical rather than physical. Inside a
container, each blob has a unique name.
The partition key for a blob is account name + container name + blob name. The partition key is used to partition
data into ranges and these ranges are load-balanced across the system. Blobs can be distributed across many
servers in order to scale out access to them, but a single blob can only be served by a single server.
If your naming scheme uses timestamps or numerical identifiers, it can lead to excessive traffic going to one
partition, limiting the system from effectively load balancing. For instance, if you have daily operations that use a
blob object with a timestamp such as yyyy-mm-dd, all the traffic for that operation would go to a single partition
server. Instead, consider prefixing the name with a three-digit hash. For more information, see Partition Naming
Convention.
The actions of writing a single block or page are atomic, but operations that span blocks, pages, or blobs are not. If
you need to ensure consistency when performing write operations across blocks, pages, and blobs, take out a write
lock by using a blob lease.
Partitioning Azure storage queues
Azure storage queues enable you to implement asynchronous messaging between processes. An Azure storage
account can contain any number of queues, and each queue can contain any number of messages. The only
limitation is the space that's available in the storage account. The maximum size of an individual message is 64 KB.
If you require messages bigger than this, then consider using Azure Service Bus queues instead.
Each storage queue has a unique name within the storage account that contains it. Azure partitions queues based
on the name. All messages for the same queue are stored in the same partition, which is controlled by a single
server. Different queues can be managed by different servers to help balance the load. The allocation of queues to
servers is transparent to applications and users.
In a large-scale application, don't use the same storage queue for all instances of the application because this
approach might cause the server that's hosting the queue to become a hot spot. Instead, use different queues for
different functional areas of the application. Azure storage queues do not support transactions, so directing
messages to different queues should have little effect on messaging consistency.
An Azure storage queue can handle up to 2,000 messages per second. If you need to process messages at a greater
rate than this, consider creating multiple queues. For example, in a global application, create separate storage
queues in separate storage accounts to handle application instances that are running in each region.

Partitioning Azure Service Bus

Azure Service Bus uses a message broker to handle messages that are sent to a Service Bus queue or topic. By
default, all messages that are sent to a queue or topic are handled by the same message broker process. This
architecture can place a limitation on the overall throughput of the message queue. However, you can also partition
a queue or topic when it is created. You do this by setting the EnablePartitioning property of the queue or topic
description to true.
A partitioned queue or topic is divided into multiple fragments, each of which is backed by a separate message
store and message broker. Service Bus takes responsibility for creating and managing these fragments. When an
application posts a message to a partitioned queue or topic, Service Bus assigns the message to a fragment for that
queue or topic. When an application receives a message from a queue or subscription, Service Bus checks each
fragment for the next available message and then passes it to the application for processing.
This structure helps distribute the load across message brokers and message stores, increasing scalability and
improving availability. If the message broker or message store for one fragment is temporarily unavailable, Service
Bus can retrieve messages from one of the remaining available fragments.
Service Bus assigns a message to a fragment as follows:
If the message belongs to a session, all messages with the same value for the SessionId property are sent to
the same fragment.
If the message does not belong to a session, but the sender has specified a value for the PartitionKey
property, then all messages with the same PartitionKey value are sent to the same fragment.

NOTE
If the SessionId and PartitionKey properties are both specified, then they must be set to the same value or the
message will be rejected.

If the SessionId and PartitionKey properties for a message are not specified, but duplicate detection is
enabled, the MessageId property will be used. All messages with the same MessageId will be directed to the
same fragment.
 If messages do not include a SessionId, PartitionKey, or MessageId property, then Service Bus assigns
messages to fragments sequentially. If a fragment is unavailable, Service Bus will move on to the next. This
means that a temporary fault in the messaging infrastructure does not cause the message-send operation to
fail.
Consider the following points when deciding if or how to partition a Service Bus message queue or topic:
Service Bus queues and topics are created within the scope of a Service Bus namespace. Service Bus
currently allows up to 100 partitioned queues or topics per namespace.
Each Service Bus namespace imposes quotas on the available resources, such as the number of
subscriptions per topic, the number of concurrent send and receive requests per second, and the maximum
number of concurrent connections that can be established. These quotas are documented in Service Bus
quotas. If you expect to exceed these values, then create additional namespaces with their own queues and
topics, and spread the work across these namespaces. For example, in a global application, create separate
namespaces in each region and configure application instances to use the queues and topics in the nearest
namespace.
Messages that are sent as part of a transaction must specify a partition key. This can be a SessionId,
PartitionKey, or MessageId property. All messages that are sent as part of the same transaction must specify
the same partition key because they must be handled by the same message broker process. You cannot
send messages to different queues or topics within the same transaction.
Partitioned queues and topics can't be configured to be automatically deleted when they become idle.
Partitioned queues and topics can't currently be used with the Advanced Message Queuing Protocol (AMQP)
if you are building cross-platform or hybrid solutions.

Partitioning Cosmos DB
Azure Cosmos DB is a NoSQL database that can store JSON documents using the Azure Cosmos DB SQL API. A
document in a Cosmos DB database is a JSON-serialized representation of an object or other piece of data. No
fixed schemas are enforced except that every document must contain a unique ID.
Documents are organized into collections. You can group related documents together in a collection. For example,
in a system that maintains blog postings, you can store the contents of each blog post as a document in a
collection. You can also create collections for each subject type. Alternatively, in a multitenant application, such as a
system where different authors control and manage their own blog posts, you can partition blogs by author and
create separate collections for each author. The storage space that's allocated to collections is elastic and can shrink
or grow as needed.
Cosmos DB supports automatic partitioning of data based on an application-defined partition key. A logical
partition is a partition that stores all the data for a single partition key value. All documents that share the same
value for the partition key are placed within the same logical partition. Cosmos DB distributes values according to
hash of the partition key. A logical partition has a maximum size of 10 GB. Therefore, the choice of the partition key
is an important decision at design time. Choose a property with a wide range of values and even access patterns.
For more information, see Partition and scale in Azure Cosmos DB.

NOTE
Each Cosmos DB database has a performance level that determines the amount of resources it gets. A performance level is
associated with a request unit (RU) rate limit. The RU rate limit specifies the volume of resources that's reserved and available
for exclusive use by that collection. The cost of a collection depends on the performance level that's selected for that
collection. The higher the performance level (and RU rate limit) the higher the charge. You can adjust the performance level of
a collection by using the Azure portal. For more information, see Request Units in Azure Cosmos DB.
If the partitioning mechanism that Cosmos DB provides is not sufficient, you may need to shard the data at the
application level. Document collections provide a natural mechanism for partitioning data within a single database.
The simplest way to implement sharding is to create a collection for each shard. Containers are logical resources
and can span one or more servers. Fixed-size containers have a maximum limit of 10 GB and 10,000 RU/s
throughput. Unlimited containers do not have a maximum storage size, but must specify a partition key. With
application sharding, the client application must direct requests to the appropriate shard, usually by implementing
its own mapping mechanism based on some attributes of the data that define the shard key.
All databases are created in the context of a Cosmos DB database account. A single account can contain several
databases, and it specifies in which regions the databases are created. Each account also enforces its own access
control. You can use Cosmos DB accounts to geo-locate shards (collections within databases) close to the users
who need to access them, and enforce restrictions so that only those users can connect to them.
Consider the following points when deciding how to partition data with the Cosmos DB SQL API:
The resources available to a Cosmos DB database are subject to the quota limitations of the
account . Each database can hold a number of collections, and each collection is associated with a
performance level that governs the RU rate limit (reserved throughput) for that collection. For more
information, see Azure subscription and service limits, quotas, and constraints.
Each document must have an attribute that can be used to uniquely identify that document
within the collection in which it is held . This attribute is different from the shard key, which defines
which collection holds the document. A collection can contain a large number of documents. In theory, it's
limited only by the maximum length of the document ID. The document ID can be up to 255 characters.
All operations against a document are performed within the context of a transaction.
Transactions are scoped to the collection in which the document is contained. If an operation fails,
the work that it has performed is rolled back. While a document is subject to an operation, any changes that
are made are subject to snapshot-level isolation. This mechanism guarantees that if, for example, a request
to create a new document fails, another user who's querying the database simultaneously will not see a
partial document that is then removed.
Database queries are also scoped to the collection level . A single query can retrieve data from only
one collection. If you need to retrieve data from multiple collections, you must query each collection
individually and merge the results in your application code.
Cosmos DB suppor ts programmable items that can all be stored in a collection alongside
documents . These include stored procedures, user-defined functions, and triggers (written in JavaScript).
These items can access any document within the same collection. Furthermore, these items run either inside
the scope of the ambient transaction (in the case of a trigger that fires as the result of a create, delete, or
replace operation performed against a document), or by starting a new transaction (in the case of a stored
procedure that is run as the result of an explicit client request). If the code in a programmable item throws
an exception, the transaction is rolled back. You can use stored procedures and triggers to maintain integrity
and consistency between documents, but these documents must all be part of the same collection.
The collections that you intend to hold in the databases should be unlikely to exceed the
throughput limits defined by the performance levels of the collections . For more information, see
Request Units in Azure Cosmos DB. If you anticipate reaching these limits, consider splitting collections
across databases in different accounts to reduce the load per collection.

Partitioning Azure Search

The ability to search for data is often the primary method of navigation and exploration that's provided by many
web applications. It helps users find resources quickly (for example, products in an e-commerce application) based
on combinations of search criteria. The Azure Search service provides full-text search capabilities over web content,
and includes features such as type-ahead, suggested queries based on near matches, and faceted navigation. For
more information, see What is Azure Search?.
Azure Search stores searchable content as JSON documents in a database. You define indexes that specify the
searchable fields in these documents and provide these definitions to Azure Search. When a user submits a search
request, Azure Search uses the appropriate indexes to find matching items.
To reduce contention, the storage that's used by Azure Search can be divided into 1, 2, 3, 4, 6, or 12 partitions, and
each partition can be replicated up to 6 times. The product of the number of partitions multiplied by the number of
replicas is called the search unit (SU). A single instance of Azure Search can contain a maximum of 36 SUs (a
database with 12 partitions only supports a maximum of 3 replicas).
You are billed for each SU that is allocated to your service. As the volume of searchable content increases or the
rate of search requests grows, you can add SUs to an existing instance of Azure Search to handle the extra load.
Azure Search itself distributes the documents evenly across the partitions. No manual partitioning strategies are
currently supported.
Each partition can contain a maximum of 15 million documents or occupy 300 GB of storage space (whichever is
smaller). You can create up to 50 indexes. The performance of the service varies and depends on the complexity of
the documents, the available indexes, and the effects of network latency. On average, a single replica (1 SU) should
be able to handle 15 queries per second (QPS), although we recommend performing benchmarking with your own
data to obtain a more precise measure of throughput. For more information, see Service limits in Azure Search.

NOTE
You can store a limited set of data types in searchable documents, including strings, Booleans, numeric data, datetime data,
and some geographical data. For more information, see the page Supported data types (Azure Search) on the Microsoft
website.

You have limited control over how Azure Search partitions data for each instance of the service. However, in a
global environment you might be able to improve performance and reduce latency and contention further by
partitioning the service itself using either of the following strategies:
Create an instance of Azure Search in each geographic region, and ensure that client applications are
directed toward the nearest available instance. This strategy requires that any updates to searchable content
are replicated in a timely manner across all instances of the service.
Create two tiers of Azure Search:
A local service in each region that contains the data that's most frequently accessed by users in that
region. Users can direct requests here for fast but limited results.
A global service that encompasses all the data. Users can direct requests here for slower but more
complete results.
This approach is most suitable when there is a significant regional variation in the data that's being searched.

Partitioning Azure Cache for Redis

Azure Cache for Redis provides a shared caching service in the cloud that's based on the Redis key-value data store.
As its name implies, Azure Cache for Redis is intended as a caching solution. Use it only for holding transient data
and not as a permanent data store. Applications that use Azure Cache for Redis should be able to continue
functioning if the cache is unavailable. Azure Cache for Redis supports primary/secondary replication to provide
high availability, but currently limits the maximum cache size to 53 GB. If you need more space than this, you must
create additional caches. For more information, see Azure Cache for Redis.
Partitioning a Redis data store involves splitting the data across instances of the Redis service. Each instance
constitutes a single partition. Azure Cache for Redis abstracts the Redis services behind a façade and does not
expose them directly. The simplest way to implement partitioning is to create multiple Azure Cache for Redis
instances and spread the data across them.
You can associate each data item with an identifier (a partition key) that specifies which cache stores the data item.
The client application logic can then use this identifier to route requests to the appropriate partition. This scheme is
very simple, but if the partitioning scheme changes (for example, if additional Azure Cache for Redis instances are
created), client applications might need to be reconfigured.
Native Redis (not Azure Cache for Redis) supports server-side partitioning based on Redis clustering. In this
approach, you can divide the data evenly across servers by using a hashing mechanism. Each Redis server stores
metadata that describes the range of hash keys that the partition holds, and also contains information about which
hash keys are located in the partitions on other servers.
Client applications simply send requests to any of the participating Redis servers (probably the closest one). The
Redis server examines the client request. If it can be resolved locally, it performs the requested operation.
Otherwise it forwards the request on to the appropriate server.
This model is implemented by using Redis clustering, and is described in more detail on the Redis cluster tutorial
page on the Redis website. Redis clustering is transparent to client applications. Additional Redis servers can be
added to the cluster (and the data can be repartitioned) without requiring that you reconfigure the clients.

IMPORTANT
Azure Cache for Redis currently supports Redis clustering in premium tier only.

The page Partitioning: how to split data among multiple Redis instances on the Redis website provides more
information about implementing partitioning with Redis. The remainder of this section assumes that you are
implementing client-side or proxy-assisted partitioning.
Consider the following points when deciding how to partition data with Azure Cache for Redis:
Azure Cache for Redis is not intended to act as a permanent data store, so whatever partitioning scheme you
implement, your application code must be able to retrieve data from a location that's not the cache.
Data that is frequently accessed together should be kept in the same partition. Redis is a powerful key-value
store that provides several highly optimized mechanisms for structuring data. These mechanisms can be
one of the following:
Simple strings (binary data up to 512 MB in length)
Aggregate types such as lists (which can act as queues and stacks)
Sets (ordered and unordered)
Hashes (which can group related fields together, such as the items that represent the fields in an object)
The aggregate types enable you to associate many related values with the same key. A Redis key identifies a
list, set, or hash rather than the data items that it contains. These types are all available with Azure Cache for
Redis and are described by the Data types page on the Redis website. For example, in part of an e-commerce
system that tracks the orders that are placed by customers, the details of each customer can be stored in a
Redis hash that is keyed by using the customer ID. Each hash can hold a collection of order IDs for the
customer. A separate Redis set can hold the orders, again structured as hashes, and keyed by using the order
ID. Figure 8 shows this structure. Note that Redis does not implement any form of referential integrity, so it
is the developer's responsibility to maintain the relationships between customers and orders.
Figure 8. Suggested structure in Redis storage for recording customer orders and their details.

NOTE
In Redis, all keys are binary data values (like Redis strings) and can contain up to 512 MB of data. In theory, a key can contain
almost any information. However, we recommend adopting a consistent naming convention for keys that is descriptive of the
type of data and that identifies the entity, but is not excessively long. A common approach is to use keys of the form
"entity_type:ID". For example, you can use "customer:99" to indicate the key for a customer with the ID 99.

You can implement vertical partitioning by storing related information in different aggregations in the same
database. For example, in an e-commerce application, you can store commonly accessed information about
products in one Redis hash and less frequently used detailed information in another. Both hashes can use
the same product ID as part of the key. For example, you can use "product: nn" (where nn is the product ID)
for the product information and "product_details: nn" for the detailed data. This strategy can help reduce the
volume of data that most queries are likely to retrieve.
You can repartition a Redis data store, but keep in mind that it's a complex and time-consuming task. Redis
clustering can repartition data automatically, but this capability is not available with Azure Cache for Redis.
Therefore, when you design your partitioning scheme, try to leave sufficient free space in each partition to
allow for expected data growth over time. However, remember that Azure Cache for Redis is intended to
cache data temporarily, and that data held in the cache can have a limited lifetime specified as a time-to-live
(TTL) value. For relatively volatile data, the TTL can be short, but for static data the TTL can be a lot longer.
Avoid storing large amounts of long-lived data in the cache if the volume of this data is likely to fill the
cache. You can specify an eviction policy that causes Azure Cache for Redis to remove data if space is at a
premium.

NOTE
When you use Azure Cache for Redis, you specify the maximum size of the cache (from 250 MB to 53 GB) by
selecting the appropriate pricing tier. However, after an Azure Cache for Redis has been created, you cannot increase
(or decrease) its size.
 Redis batches and transactions cannot span multiple connections, so all data that is affected by a batch or
transaction should be held in the same database (shard).

NOTE
A sequence of operations in a Redis transaction is not necessarily atomic. The commands that compose a transaction
are verified and queued before they run. If an error occurs during this phase, the entire queue is discarded. However,
after the transaction has been successfully submitted, the queued commands run in sequence. If any command fails,
only that command stops running. All previous and subsequent commands in the queue are performed. For more
information, go to the Transactions page on the Redis website.

Redis supports a limited number of atomic operations. The only operations of this type that support
multiple keys and values are MGET and MSET operations. MGET operations return a collection of values for
a specified list of keys, and MSET operations store a collection of values for a specified list of keys. If you
need to use these operations, the key-value pairs that are referenced by the MSET and MGET commands
must be stored within the same database.

Partitioning Azure Service Fabric

Azure Service Fabric is a microservices platform that provides a runtime for distributed applications in the cloud.
Service Fabric supports .Net guest executables, stateful and stateless services, and containers. Stateful services
provide a reliable collection to persistently store data in a key-value collection within the Service Fabric cluster. For
more information about strategies for partitioning keys in a reliable collection, see guidelines and
recommendations for reliable collections in Azure Service Fabric.
More information
Overview of Azure Service Fabric is an introduction to Azure Service Fabric.
Partition Service Fabric reliable services provides more information about reliable services in Azure Service
Fabric.

Partitioning Azure Event Hubs

Azure Event Hubs is designed for data streaming at massive scale, and partitioning is built into the service to
enable horizontal scaling. Each consumer only reads a specific partition of the message stream.
The event publisher is only aware of its partition key, not the partition to which the events are published. This
decoupling of key and partition insulates the sender from needing to know too much about the downstream
processing. (It's also possible send events directly to a given partition, but generally that's not recommended.)
Consider long-term scale when you select the partition count. After an event hub is created, you can't change the
number of partitions.
For more information about using partitions in Event Hubs, see What is Event Hubs?.
For considerations about trade-offs between availability and consistency, see Availability and consistency in Event
Hubs.
 Message encoding considerations
12/18/2020 • 11 minutes to read • Edit Online

Many cloud applications use asynchronous messages to exchange information between components of the system.
An important aspect of messaging is the format used to encode the payload data. After you choose a messaging
technology, the next step is to define how the messages will be encoded. There are many options available, but the
right choice depends on your use case.
This article describes some of the considerations.

Overview
A message exchange between a producer and a consumer needs:
A shape or structure that defines the payload of the message.
An encoding format to represent the payload.
Serialization libraries to read and write the encoded payload.
The producer of the message defines the message shape based on the business logic and the information it wants
to send to the consumer(s). To structure the shape, divide the information into discrete or related subjects (fields).
Decide the characteristics of the values for those fields. Consider: What is the most efficient datatype? Will the
payload always have certain fields? Will the payload have a single record or a repeated set of values?
Then, choose an encoding format depending on your need. Certain factors include the ability to create highly
structured data if you need it, time taken to encode and transfer the message, and the ability to parse the payload.
Depending on the encoding format, choose a serialization library that is well supported.
A consumer of the message must be aware of those decisions so that it knows how to read incoming messages.
To transfer messages, the producer serializes the message to an encoding format. At the receiving end, the
consumer deserializes the payload to use the data. This way both entities share the model and as long as the shape
doesn't change, messaging continues without issues. When the contract changes, the encoding format should be
capable of handling the change without breaking the consumer.
Some encoding formats such as JSON are self-describing, meaning they can be parsed without referencing a
schema. However, such formats tend to yield larger messages. With other formats, the data may not be parsed as
easily but the messages are compact. This article highlights some factors that can help you choose a format.

Encoding format considerations

The encoding format defines how a set of structured data is represented as bytes. The type of message can
influence the format choice. Messages related to business transactions most likely will contain highly structured
data. Also, you may want to retrieve it later for auditing purposes. For a stream of events, you might want to read a
sequence of records as quickly as possible and store it for statistical analysis.
Here are some points to consider when choosing an encoding format.
Human readability
Message encoding can be broadly divided into text-based and binary formats.
With text-based encoding, the message payload is in plain text and therefore can be inspected by a person without
using any code libraries. Human readable formats are suitable for archival data. Also, because a human can read
the payload, text-based formats are easier to debug and send to logs for troubleshooting errors.
The downside is that the payload tends to be larger. A common text-based format is JSON.
Encoding size
Message size impacts network I/O performance across the wire. Binary formats are more compact than text-based
formats. Binary formats require serialization/deserialization libraries. The payload can't be read unless it's decoded.
Use a binary format if you want to reduce wire footprint and transfer messages faster. This category of format is
recommended in scenarios where storage or network bandwidth is a concern. Options for binary formats include
Apache Avro, Google Protocol Buffers (protobuf), MessagePack, and Concise Binary Object Representation (CBOR).
The pros and cons of those formats are described in this section.
The disadvantage is that the payload isn't human readable. Most binary formats use complex systems that can be
costly to maintain. Also, they need specialized libraries to decode, which may not be supported if you want to
retrieve archival data.
Understanding the payload
A message payload arrives as a sequence of bytes. To parse this sequence, the consumer must have access to
metadata that describes the data fields in the payload. There are two main approaches for storing and distributing
metadata:
Tagged metadata . In some encodings, notably JSON, fields are tagged with the data type and identifier, within the
body of the message. These formats are self-describing because they can be parsed into a dictionary of values
without referring to a schema. One way for the consumer to understand the fields is to query for expected values.
For example, the producer sends a payload in JSON. The consumer parses the JSON into a dictionary and checks
the existence of fields to understand the payload. Another way is for the consumer to apply a data model shared by
the producer. For example, if you are using a statically typed language, many JSON serialization libraries can parse
a JSON string into a typed class.
Schema . A schema formally defines the structure and data fields of a message. In this model, the producer and
consumer have a contract through a well-defined schema. The schema can define the data types, required/optional
fields, version information, and the structure of the payload. The producer sends the payload as per the writer
schema. The consumer receives the payload by applying a reader schema. The message is serialized/deserialized by
using the encoding-specific libraries. There are two ways to distribute schemas:
Store the schema as a preamble or header in the message but separate from the payload.
Store the schema externally.
Some encoding formats define the schema and use tools that generate classes from the schema. The producer and
consumer use those classes and libraries to serialize and deserialize the payload. The libraries also provide
compatibility checks between the writer and reader schema. Both protobuf and Apache Avro follow that approach.
The key difference is that protobuf has a language-agnostic schema definition but Avro uses compact JSON.
Another difference is in the way both formats provide compatibility checks between reader and writer schemas.
Another way to store the schema externally in a schema registry. The message contains a reference to the schema
and the payload. The producer sends the schema identifier in the message and the consumer retrieves the schema
by specifying that identifier from an external store. Both parties use format-specific library to read and write
messages. Apart from storing the schema a registry can provide compatibility checks to make sure the contract
between the producer and consumer isn't broken as the schema evolves.
Before choosing an approach, decide what is more important: the transfer data size or the ability to parse the
archived data later.
Storing the schema along with the payload yields a larger encoding size and is preferred for intermittent messages.
Choose this approach if transferring smaller chunks of bytes is crucial or you expect a sequence of records. The cost
of maintaining an external schema store can be high.
However, if on-demand decoding of the payload is more important than size, including the schema with the
payload or the tagged metadata approach guarantees decoding afterwards. There might be a significant increase in
message size and may impact the cost of storage.
Schema versioning
As business requirements change, the shape is expected to change, and the schema will evolve. Versioning allows
the producer to indicate schema updates that might include new features. There are two aspects to versioning:
The consumer should be aware of the changes.
One way is for the consumer to check all fields to determine whether the schema has changed. Another way
is for the producer to publish a schema version number with the message. When the schema evolves, the
producer increments the version.
Changes must not affect or break the business logic of consumers.
Suppose a field is added to an existing schema. If consumers using the new version get a payload as per the
old version, their logic might break if they are not able to overlook the lack of the new field. Considering the
reverse case, suppose a field is removed in the new schema. Consumers using the old schema might not be
able to read the data.
Encoding formats such as Avro offer the ability to define default values. In the preceding example, if the field
is added with a default value, the missing field will be populated with the default value. Other formats such
as protobuf provide similar functionality through required and optional fields.
Payload structure
Consider the way data is arranged in the payload. Is it a sequence of records or a discrete single payload? The
payload structure can be categorized into one of these models:
Array/dictionary/value: Defines entries that hold values in one or multi-dimensional arrays. Entries have
unique key-value pairs. It can be extended to represent the complex structures. Some examples include,
JSON, Apache Avro, and MessagePack.
This layout is suitable if messages are individual encoded with different schemas. If you have multiple
records, the payload can get overly redundant causing the payload to bloat.
Tabular data: Information is divided into rows and columns. Each column indicates a field, or the subject of
the information and each row contains values for those fields. This layout is efficient for a repeating set of
information, such as time series data.
CSV is one of the simplest text-based formats. It presents data as a sequence of records with a common
header. For binary encoding, Apache Avro has a preamble is similar to a CSV header but generate compact
encoding size.
Library support
Consider using well-known formats over a proprietary model.
Well-known formats are supported through libraries that are universally supported by the community. With
specialized formats, you need specific libraries. Your business logic might have to work around some of the API
design choices provided by the libraries.
For schema-based format, choose an encoding library that makes compatibility checks between the reader and
writer schema. Certain encoding libraries, such as Apache Avro, expect the consumer to specify both writer and the
reader schema before deserializing the message. This check ensures that the consumer is aware of the schema
versions.
Interoperability
Your choice of formats might depend on the particular workload or technology ecosystem.
For example:
Azure Stream Analytics has native support for JSON, CSV, and Avro. When using Stream Analytics, it makes
sense to choose one of these formats if possible. If not, you can provide a custom deserializer, but this adds
some additional complexity to your solution.
JSON is a standard interchange format for HTTP REST APIs. If your application receives JSON payloads from
clients and then places these onto a message queue for asynchronous processing, it might make sense to
use JSON for the messaging, rather than re-encode into a different format.
These are just two examples of interoperability considerations. In general, standardized formats will be more
interoperable than custom formats. In text-based options, JSON is one of the most interoperable.

Choices for encoding formats

Here are some popular encoding formats. Factor in the considerations before you choose a format.
JSON
JSON is an open standard (IETF RFC8259). It's a text-based format that follows the array/dictionary/value model.
JSON can be used for tagging metadata and you can parse the payload without a schema. JSON supports the
option to specify optional fields, which helps with forward and backward compatibility.
The biggest advantage is that its universally available. It's most interoperable and the default encoding format for
many messaging services.
Being a text-based format, it isn't efficient over the wire and not an ideal choice in cases where storage is a concern.
If you're returning cached items directly to a client via HTTP, storing JSON could save the cost of deserializing from
another format and then serializing to JSON.
Use for messages single record messages or a sequence of messages where each message has a different schema.
Avoid using JSON for a sequence of records, such as time-series data.
There are other variations of JSON such as BSON, which is a binary encoding aligned to work with MongoDB.
Comma-Separated Values (CSV )
CSV is a text-based tabular format. The header of the table indicates the fields. It's a preferred choice where the
message contains a set of records.
The disadvantage is lack of standardization. There are many ways of expressing separators, headers, and empty
fields.
Protocol Buffers (protobuf )
Google Protocol Buffers (or protobuf) is a serialization format that uses strongly typed definition files to define
schemas in key/value pairs. These definition files are then compiled to language-specific classes that are used for
serializing and deserializing messages.
The message contains a compressed binary small payload, which results is faster transfer. The downside is the
payload isn't human readable. Also, because the schema is external, it's not recommended for cases where you
have to retrieve archived data.
Apache Avro
Apache Avro is a binary serialization format that uses definition file similar to protobuf but there isn't a compilation
step. Instead, serialized data always includes a schema preamble.
The preamble can hold the header or a schema identifier. Because of the smaller encoding size, Avro is
recommended for streaming data. Also, because it has a header that applies to a set of records, it's a good choice
for tabular data.
MessagePack
MessagePack is a binary serialization format that is designed to be compact for transmission over the wire. There
are no message schemas or message type checking. This format isn't recommended for bulk storage.
CBOR
Concise Binary Object Representation (CBOR) (Specification) is a binary format that offers small encoding size. The
advantage of CBOR over MessagePack is that its compliant with IETF in RFC7049.

Next steps
Understand messaging design patterns for cloud applications.
 Best practices for monitoring cloud applications
12/18/2020 • 68 minutes to read • Edit Online

Distributed applications and services running in the cloud are, by their nature, complex pieces of software that
comprise many moving parts. In a production environment, it's important to be able to track the way in which
users use your system, trace resource utilization, and generally monitor the health and performance of your
system. You can use this information as a diagnostic aid to detect and correct issues, and also to help spot
potential problems and prevent them from occurring.

Monitoring and diagnostics scenarios

You can use monitoring to gain an insight into how well a system is functioning. Monitoring is a crucial part of
maintaining quality-of-service targets. Common scenarios for collecting monitoring data include:
Ensuring that the system remains healthy.
Tracking the availability of the system and its component elements.
Maintaining performance to ensure that the throughput of the system does not degrade unexpectedly as the
volume of work increases.
Guaranteeing that the system meets any service-level agreements (SLAs) established with customers.
Protecting the privacy and security of the system, users, and their data.
Tracking the operations that are performed for auditing or regulatory purposes.
Monitoring the day-to-day usage of the system and spotting trends that might lead to problems if they're not
addressed.
Tracking issues that occur, from initial report through to analysis of possible causes, rectification, consequent
software updates, and deployment.
Tracing operations and debugging software releases.

NOTE
This list is not intended to be comprehensive. This document focuses on these scenarios as the most common situations for
performing monitoring. There might be others that are less common or are specific to your environment.

The following sections describe these scenarios in more detail. The information for each scenario is discussed in
the following format:
1. A brief overview of the scenario.
2. The typical requirements of this scenario.
3. The raw instrumentation data that's required to support the scenario, and possible sources of this information.
4. How this raw data can be analyzed and combined to generate meaningful diagnostic information.

Health monitoring
A system is healthy if it is running and capable of processing requests. The purpose of health monitoring is to
generate a snapshot of the current health of the system so that you can verify that all components of the system
are functioning as expected.
Requirements for health monitoring
An operator should be alerted quickly (within a matter of seconds) if any part of the system is deemed to be
unhealthy. The operator should be able to ascertain which parts of the system are functioning normally, and which
parts are experiencing problems. System health can be highlighted through a traffic-light system:
Red for unhealthy (the system has stopped)
Yellow for partially healthy (the system is running with reduced functionality)
Green for completely healthy
A comprehensive health-monitoring system enables an operator to drill down through the system to view the
health status of subsystems and components. For example, if the overall system is depicted as partially healthy,
the operator should be able to zoom in and determine which functionality is currently unavailable.
Data sources, instrumentation, and data-collection requirements
The raw data that's required to support health monitoring can be generated as a result of:
Tracing execution of user requests. This information can be used to determine which requests have succeeded,
which have failed, and how long each request takes.
Synthetic user monitoring. This process simulates the steps performed by a user and follows a predefined
series of steps. The results of each step should be captured.
Logging exceptions, faults, and warnings. This information can be captured as a result of trace statements
embedded into the application code, as well as retrieving information from the event logs of any services that
the system references.
Monitoring the health of any third-party services that the system uses. This monitoring might require
retrieving and parsing health data that these services supply. This information might take a variety of formats.
Endpoint monitoring. This mechanism is described in more detail in the "Availability monitoring" section.
Collecting ambient performance information, such as background CPU utilization or I/O (including network)
activity.
Analyzing health data
The primary focus of health monitoring is to quickly indicate whether the system is running. Hot analysis of the
immediate data can trigger an alert if a critical component is detected as unhealthy. (It fails to respond to a
consecutive series of pings, for example.) The operator can then take the appropriate corrective action.
A more advanced system might include a predictive element that performs a cold analysis over recent and current
workloads. A cold analysis can spot trends and determine whether the system is likely to remain healthy or
whether the system will need additional resources. This predictive element should be based on critical
performance metrics, such as:
The rate of requests directed at each service or subsystem.
The response times of these requests.
The volume of data flowing into and out of each service.
If the value of any metric exceeds a defined threshold, the system can raise an alert to enable an operator or
autoscaling (if available) to take the preventative actions necessary to maintain system health. These actions might
involve adding resources, restarting one or more services that are failing, or applying throttling to lower-priority
requests.

Availability monitoring
A truly healthy system requires that the components and subsystems that compose the system are available.
Availability monitoring is closely related to health monitoring. But whereas health monitoring provides an
immediate view of the current health of the system, availability monitoring is concerned with tracking the
availability of the system and its components to generate statistics about the uptime of the system.
In many systems, some components (such as a database) are configured with built-in redundancy to permit rapid
failover in the event of a serious fault or loss of connectivity. Ideally, users should not be aware that such a failure
has occurred. But from an availability monitoring perspective, it's necessary to gather as much information as
possible about such failures to determine the cause and take corrective actions to prevent them from recurring.
The data that's required to track availability might depend on a number of lower-level factors. Many of these
factors might be specific to the application, system, and environment. An effective monitoring system captures the
availability data that corresponds to these low-level factors and then aggregates them to give an overall picture of
the system. For example, in an e-commerce system, the business functionality that enables a customer to place
orders might depend on the repository where order details are stored and the payment system that handles the
monetary transactions for paying for these orders. The availability of the order-placement part of the system is
therefore a function of the availability of the repository and the payment subsystem.
Requirements for availability monitoring
An operator should also be able to view the historical availability of each system and subsystem, and use this
information to spot any trends that might cause one or more subsystems to periodically fail. (Do services start to
fail at a particular time of day that corresponds to peak processing hours?)
A monitoring solution should provide an immediate and historical view of the availability or unavailability of each
subsystem. It should also be capable of quickly alerting an operator when one or more services fail or when users
can't connect to services. This is a matter of not only monitoring each service, but also examining the actions that
each user performs if these actions fail when they attempt to communicate with a service. To some extent, a
degree of connectivity failure is normal and might be due to transient errors. But it might be useful to allow the
system to raise an alert for the number of connectivity failures to a specified subsystem that occur during a
specific period.
Data sources, instrumentation, and data-collection requirements
As with health monitoring, the raw data that's required to support availability monitoring can be generated as a
result of synthetic user monitoring and logging any exceptions, faults, and warnings that might occur. In addition,
availability data can be obtained from performing endpoint monitoring. The application can expose one or more
health endpoints, each testing access to a functional area within the system. The monitoring system can ping each
endpoint by following a defined schedule and collect the results (success or fail).
All timeouts, network connectivity failures, and connection retry attempts must be recorded. All data should be
timestamped.
Analyzing availability data
The instrumentation data must be aggregated and correlated to support the following types of analysis:
The immediate availability of the system and subsystems.
The availability failure rates of the system and subsystems. Ideally, an operator should be able to correlate
failures with specific activities: what was happening when the system failed?
A historical view of failure rates of the system or any subsystems across any specified period, and the load on
the system (number of user requests, for example) when a failure occurred.
The reasons for unavailability of the system or any subsystems. For example, the reasons might be service not
running, connectivity lost, connected but timing out, and connected but returning errors.
You can calculate the percentage availability of a service over a period of time by using the following formula:

%Availability = ((Total Time – Total Downtime) / Total Time ) * 100

This is useful for SLA purposes. (SLA monitoring is described in more detail later in this guidance.) The definition
of downtime depends on the service. For example, Visual Studio Team Services Build Service defines downtime as
the period (total accumulated minutes) during which Build Service is unavailable. A minute is considered
unavailable if all continuous HTTP requests to Build Service to perform customer-initiated operations throughout
the minute either result in an error code or do not return a response.
Performance monitoring
As the system is placed under more and more stress (by increasing the volume of users), the size of the datasets
that these users access grows and the possibility of failure of one or more components becomes more likely.
Frequently, component failure is preceded by a decrease in performance. If you're able detect such a decrease, you
can take proactive steps to remedy the situation.
System performance depends on a number of factors. Each factor is typically measured through key performance
indicators (KPIs), such as the number of database transactions per second or the volume of network requests that
are successfully serviced in a specified time frame. Some of these KPIs might be available as specific performance
measures, whereas others might be derived from a combination of metrics.

NOTE
Determining poor or good performance requires that you understand the level of performance at which the system should
be capable of running. This requires observing the system while it's functioning under a typical load and capturing the data
for each KPI over a period of time. This might involve running the system under a simulated load in a test environment and
gathering the appropriate data before deploying the system to a production environment.
You should also ensure that monitoring for performance purposes does not become a burden on the system. You might be
able to dynamically adjust the level of detail for the data that the performance monitoring process gathers.

Requirements for performance monitoring

To examine system performance, an operator typically needs to see information that includes:
The response rates for user requests.
The number of concurrent user requests.
The volume of network traffic.
The rates at which business transactions are being completed.
The average processing time for requests.
It can also be helpful to provide tools that enable an operator to help spot correlations, such as:
The number of concurrent users versus request latency times (how long it takes to start processing a request
after the user has sent it).
The number of concurrent users versus the average response time (how long it takes to complete a request
after it has started processing).
The volume of requests versus the number of processing errors.
Along with this high-level functional information, an operator should be able to obtain a detailed view of the
performance for each component in the system. This data is typically provided through low-level performance
counters that track information such as:
Memory utilization.
Number of threads.
CPU processing time.
Request queue length.
Disk or network I/O rates and errors.
Number of bytes written or read.
Middleware indicators, such as queue length.
All visualizations should allow an operator to specify a time period. The displayed data might be a snapshot of the
current situation and/or a historical view of the performance.
An operator should be able to raise an alert based on any performance measure for any specified value during
any specified time interval.
Data sources, instrumentation, and data-collection requirements
You can gather high-level performance data (throughput, number of concurrent users, number of business
transactions, error rates, and so on) by monitoring the progress of users' requests as they arrive and pass through
the system. This involves incorporating tracing statements at key points in the application code, together with
timing information. All faults, exceptions, and warnings should be captured with sufficient data for correlating
them with the requests that caused them. The Internet Information Services (IIS) log is another useful source.
If possible, you should also capture performance data for any external systems that the application uses. These
external systems might provide their own performance counters or other features for requesting performance
data. If this is not possible, record information such as the start time and end time of each request made to an
external system, together with the status (success, fail, or warning) of the operation. For example, you can use a
stopwatch approach to time requests: start a timer when the request starts and then stop the timer when the
request finishes.
Low-level performance data for individual components in a system might be available through features and
services such as Windows performance counters and Azure Diagnostics.
Analyzing performance data
Much of the analysis work consists of aggregating performance data by user request type and/or the subsystem
or service to which each request is sent. An example of a user request is adding an item to a shopping cart or
performing the checkout process in an e-commerce system.
Another common requirement is summarizing performance data in selected percentiles. For example, an operator
might determine the response times for 99 percent of requests, 95 percent of requests, and 70 percent of
requests. There might be SLA targets or other goals set for each percentile. The ongoing results should be
reported in near real time to help detect immediate issues. The results should also be aggregated over the longer
time for statistical purposes.
In the case of latency issues affecting performance, an operator should be able to quickly identify the cause of the
bottleneck by examining the latency of each step that each request performs. The performance data must
therefore provide a means of correlating performance measures for each step to tie them to a specific request.
Depending on the visualization requirements, it might be useful to generate and store a data cube that contains
views of the raw data. This data cube can allow complex ad hoc querying and analysis of the performance
information.

Security monitoring
All commercial systems that include sensitive data must implement a security structure. The complexity of the
security mechanism is usually a function of the sensitivity of the data. In a system that requires users to be
authenticated, you should record:
All sign-in attempts, whether they fail or succeed.
All operations performed by—and the details of all resources accessed by—an authenticated user.
When a user ends a session and signs out.
Monitoring might be able to help detect attacks on the system. For example, a large number of failed sign-in
attempts might indicate a brute-force attack. An unexpected surge in requests might be the result of a distributed
denial-of-service (DDoS) attack. You must be prepared to monitor all requests to all resources regardless of the
source of these requests. A system that has a sign-in vulnerability might accidentally expose resources to the
outside world without requiring a user to actually sign in.
Requirements for security monitoring
The most critical aspects of security monitoring should enable an operator to quickly:
 Detect attempted intrusions by an unauthenticated entity.
Identify attempts by entities to perform operations on data for which they have not been granted access.
Determine whether the system, or some part of the system, is under attack from outside or inside. (For
example, a malicious authenticated user might be attempting to bring the system down.)
To support these requirements, an operator should be notified if:
One account makes repeated failed sign-in attempts within a specified period.
One authenticated account repeatedly tries to access a prohibited resource during a specified period.
A large number of unauthenticated or unauthorized requests occur during a specified period.
The information that's provided to an operator should include the host address of the source for each request. If
security violations regularly arise from a particular range of addresses, these hosts might be blocked.
A key part in maintaining the security of a system is being able to quickly detect actions that deviate from the
usual pattern. Information such as the number of failed and/or successful sign-in requests can be displayed
visually to help detect whether there is a spike in activity at an unusual time. (An example of this activity is users
signing in at 3:00 AM and performing a large number of operations when their working day starts at 9:00 AM).
This information can also be used to help configure time-based autoscaling. For example, if an operator observes
that a large number of users regularly sign in at a particular time of day, the operator can arrange to start
additional authentication services to handle the volume of work, and then shut down these additional services
when the peak has passed.
Data sources, instrumentation, and data-collection requirements
Security is an all-encompassing aspect of most distributed systems. The pertinent data is likely to be generated at
multiple points throughout a system. You should consider adopting a Security Information and Event
Management (SIEM) approach to gather the security-related information that results from events raised by the
application, network equipment, servers, firewalls, antivirus software, and other intrusion-prevention elements.
Security monitoring can incorporate data from tools that are not part of your application. These tools can include
utilities that identify port-scanning activities by external agencies, or network filters that detect attempts to gain
unauthenticated access to your application and data.
In all cases, the gathered data must enable an administrator to determine the nature of any attack and take the
appropriate countermeasures.
Analyzing security data
A feature of security monitoring is the variety of sources from which the data arises. The different formats and
level of detail often require complex analysis of the captured data to tie it together into a coherent thread of
information. Apart from the simplest of cases (such as detecting a large number of failed sign-ins, or repeated
attempts to gain unauthorized access to critical resources), it might not be possible to perform any complex
automated processing of security data. Instead, it might be preferable to write this data, timestamped but
otherwise in its original form, to a secure repository to allow for expert manual analysis.

SLA monitoring
Many commercial systems that support paying customers make guarantees about the performance of the system
in the form of SLAs. Essentially, SLAs state that the system can handle a defined volume of work within an agreed
time frame and without losing critical information. SLA monitoring is concerned with ensuring that the system can
meet measurable SLAs.
 NOTE
SLA monitoring is closely related to performance monitoring. But whereas performance monitoring is concerned with
ensuring that the system functions optimally, SLA monitoring is governed by a contractual obligation that defines what
optimally actually means.

SLAs are often defined in terms of:

Overall system availability. For example, an organization might guarantee that the system will be available for
99.9 percent of the time. This equates to no more than 9 hours of downtime per year, or approximately 10
minutes a week.
Operational throughput. This aspect is often expressed as one or more high-water marks, such as guaranteeing
that the system can support up to 100,000 concurrent user requests or handle 10,000 concurrent business
transactions.
Operational response time. The system might also make guarantees for the rate at which requests are
processed. An example is that 99 percent of all business transactions will finish within 2 seconds, and no single
transaction will take longer than 10 seconds.

NOTE
Some contracts for commercial systems might also include SLAs for customer support. An example is that all help-desk
requests will elicit a response within five minutes, and that 99 percent of all problems will be fully addressed within 1
working day. Effective issue tracking (described later in this section) is key to meeting SLAs such as these.

Requirements for SLA monitoring

At the highest level, an operator should be able to determine at a glance whether the system is meeting the
agreed SLAs or not. And if not, the operator should be able to drill down and examine the underlying factors to
determine the reasons for substandard performance.
Typical high-level indicators that can be depicted visually include:
The percentage of service uptime.
The application throughput (measured in terms of successful transactions and/or operations per second).
The number of successful/failing application requests.
The number of application and system faults, exceptions, and warnings.
All of these indicators should be capable of being filtered by a specified period of time.
A cloud application will likely comprise a number of subsystems and components. An operator should be able to
select a high-level indicator and see how it's composed from the health of the underlying elements. For example, if
the uptime of the overall system falls below an acceptable value, an operator should be able to zoom in and
determine which elements are contributing to this failure.

NOTE
System uptime needs to be defined carefully. In a system that uses redundancy to ensure maximum availability, individual
instances of elements might fail, but the system can remain functional. System uptime as presented by health monitoring
should indicate the aggregate uptime of each element and not necessarily whether the system has actually halted.
Additionally, failures might be isolated. So even if a specific system is unavailable, the remainder of the system might remain
available, although with decreased functionality. (In an e-commerce system, a failure in the system might prevent a
customer from placing orders, but the customer might still be able to browse the product catalog.)

For alerting purposes, the system should be able to raise an event if any of the high-level indicators exceed a
specified threshold. The lower-level details of the various factors that compose the high-level indicator should be
available as contextual data to the alerting system.
Data sources, instrumentation, and data-collection requirements
The raw data that's required to support SLA monitoring is similar to the raw data that's required for performance
monitoring, together with some aspects of health and availability monitoring. (See those sections for more
details.) You can capture this data by:
Performing endpoint monitoring.
Logging exceptions, faults, and warnings.
Tracing the execution of user requests.
Monitoring the availability of any third-party services that the system uses.
Using performance metrics and counters.
All data must be timed and timestamped.
Analyzing SLA data
The instrumentation data must be aggregated to generate a picture of the overall performance of the system.
Aggregated data must also support drill-down to enable examination of the performance of the underlying
subsystems. For example, you should be able to:
Calculate the total number of user requests during a specified period and determine the success and failure
rate of these requests.
Combine the response times of user requests to generate an overall view of system response times.
Analyze the progress of user requests to break down the overall response time of a request into the response
times of the individual work items in that request.
Determine the overall availability of the system as a percentage of uptime for any specific period.
Analyze the percentage time availability of the individual components and services in the system. This might
involve parsing logs that third-party services have generated.
Many commercial systems are required to report real performance figures against agreed SLAs for a specified
period, typically a month. This information can be used to calculate credits or other forms of repayments for
customers if the SLAs are not met during that period. You can calculate availability for a service by using the
technique described in the section Analyzing availability data.
For internal purposes, an organization might also track the number and nature of incidents that caused services to
fail. Learning how to resolve these issues quickly, or eliminate them completely, will help to reduce downtime and
meet SLAs.

Auditing
Depending on the nature of the application, there might be statutory or other legal regulations that specify
requirements for auditing users' operations and recording all data access. Auditing can provide evidence that links
customers to specific requests. Nonrepudiation is an important factor in many e-business systems to help
maintain trust be between a customer and the organization that's responsible for the application or service.
Requirements for auditing
An analyst must be able to trace the sequence of business operations that users are performing so that you can
reconstruct users' actions. This might be necessary simply as a matter of record, or as part of a forensic
investigation.
Audit information is highly sensitive. It will likely include data that identifies the users of the system, together with
the tasks that they're performing. For this reason, audit information will most likely take the form of reports that
are available only to trusted analysts rather than as an interactive system that supports drill-down of graphical
operations. An analyst should be able to generate a range of reports. For example, reports might list all users'
activities occurring during a specified time frame, detail the chronology of activity for a single user, or list the
sequence of operations performed against one or more resources.
Data sources, instrumentation, and data-collection requirements
The primary sources of information for auditing can include:
The security system that manages user authentication.
Trace logs that record user activity.
Security logs that track all identifiable and unidentifiable network requests.
The format of the audit data and the way in which it's stored might be driven by regulatory requirements. For
example, it might not be possible to clean the data in any way. (It must be recorded in its original format.) Access
to the repository where it's held must be protected to prevent tampering.
Analyzing audit data
An analyst must be able to access the raw data in its entirety, in its original form. Aside from the requirement to
generate common audit reports, the tools for analyzing this data are likely to be specialized and kept external to
the system.

Usage monitoring
Usage monitoring tracks how the features and components of an application are used. An operator can use the
gathered data to:
Determine which features are heavily used and determine any potential hotspots in the system. High-traffic
elements might benefit from functional partitioning or even replication to spread the load more evenly. An
operator can also use this information to ascertain which features are infrequently used and are possible
candidates for retirement or replacement in a future version of the system.
Obtain information about the operational events of the system under normal use. For example, in an e-
commerce site, you can record the statistical information about the number of transactions and the volume
of customers that are responsible for them. This information can be used for capacity planning as the
number of customers grows.
Detect (possibly indirectly) user satisfaction with the performance or functionality of the system. For
example, if a large number of customers in an e-commerce system regularly abandon their shopping carts,
this might be due to a problem with the checkout functionality.
Generate billing information. A commercial application or multitenant service might charge customers for
the resources that they use.
Enforce quotas. If a user in a multitenant system exceeds their paid quota of processing time or resource
usage during a specified period, their access can be limited or processing can be throttled.
Requirements for usage monitoring
To examine system usage, an operator typically needs to see information that includes:
The number of requests that are processed by each subsystem and directed to each resource.
The work that each user is performing.
The volume of data storage that each user occupies.
The resources that each user is accessing.
An operator should also be able to generate graphs. For example, a graph might display the most resource-
hungry users, or the most frequently accessed resources or system features.
Data sources, instrumentation, and data-collection requirements
Usage tracking can be performed at a relatively high level. It can note the start and end times of each request and
the nature of the request (read, write, and so on, depending on the resource in question). You can obtain this
information by:
Tracing user activity.
Capturing performance counters that measure the utilization for each resource.
Monitoring the resource consumption by each user.
For metering purposes, you also need to be able to identify which users are responsible for performing which
operations, and the resources that these operations use. The gathered information should be detailed enough to
enable accurate billing.

Issue tracking
Customers and other users might report issues if unexpected events or behavior occurs in the system. Issue
tracking is concerned with managing these issues, associating them with efforts to resolve any underlying
problems in the system, and informing customers of possible resolutions.
Requirements for issue tracking
Operators often perform issue tracking by using a separate system that enables them to record and report the
details of problems that users report. These details can include the tasks that the user was trying to perform,
symptoms of the problem, the sequence of events, and any error or warning messages that were issued.
Data sources, instrumentation, and data-collection requirements
The initial data source for issue-tracking data is the user who reported the issue in the first place. The user might
be able to provide additional data such as:
A crash dump (if the application includes a component that runs on the user's desktop).
A screen snapshot.
The date and time when the error occurred, together with any other environmental information such as the
user's location.
This information can be used to help the debugging effort and help construct a backlog for future releases of the
software.
Analyzing issue -tracking data
Different users might report the same problem. The issue-tracking system should associate common reports.
The progress of the debugging effort should be recorded against each issue report. When the problem is resolved,
the customer can be informed of the solution.
If a user reports an issue that has a known solution in the issue-tracking system, the operator should be able to
inform the user of the solution immediately.

Tracing operations and debugging software releases

When a user reports an issue, the user is often only aware of the immediate effect that it has on their operations.
The user can only report the results of their own experience back to an operator who is responsible for
maintaining the system. These experiences are usually just a visible symptom of one or more fundamental
problems. In many cases, an analyst will need to dig through the chronology of the underlying operations to
establish the root cause of the problem. This process is called root cause analysis.
 NOTE
Root cause analysis might uncover inefficiencies in the design of an application. In these situations, it might be possible to
rework the affected elements and deploy them as part of a subsequent release. This process requires careful control, and the
updated components should be monitored closely.

Requirements for tracing and debugging

For tracing unexpected events and other problems, it's vital that the monitoring data provides enough information
to enable an analyst to trace back to the origins of these issues and reconstruct the sequence of events that
occurred. This information must be sufficient to enable an analyst to diagnose the root cause of any problems. A
developer can then make the necessary modifications to prevent them from recurring.
Data sources, instrumentation, and data-collection requirements
Troubleshooting can involve tracing all the methods (and their parameters) invoked as part of an operation to
build up a tree that depicts the logical flow through the system when a customer makes a specific request.
Exceptions and warnings that the system generates as a result of this flow need to be captured and logged.
To support debugging, the system can provide hooks that enable an operator to capture state information at
crucial points in the system. Or, the system can deliver detailed step-by-step information as selected operations
progress. Capturing data at this level of detail can impose an additional load on the system and should be a
temporary process. An operator uses this process mainly when a highly unusual series of events occurs and is
difficult to replicate, or when a new release of one or more elements into a system requires careful monitoring to
ensure that the elements function as expected.

The monitoring and diagnostics pipeline

Monitoring a large-scale distributed system poses a significant challenge. Each of the scenarios described in the
previous section should not necessarily be considered in isolation. There is likely to be a significant overlap in the
monitoring and diagnostic data that's required for each situation, although this data might need to be processed
and presented in different ways. For these reasons, you should take a holistic view of monitoring and diagnostics.
You can envisage the entire monitoring and diagnostics process as a pipeline that comprises the stages shown in
Figure 1.

Figure 1 - The stages in the monitoring and diagnostics pipeline.

Figure 1 highlights how the data for monitoring and diagnostics can come from a variety of data sources. The
instrumentation and collection stages are concerned with identifying the sources from where the data needs to be
captured, determining which data to capture, how to capture it, and how to format this data so that it can be easily
examined. The analysis/diagnosis stage takes the raw data and uses it to generate meaningful information that an
operator can use to determine the state of the system. The operator can use this information to make decisions
about possible actions to take, and then feed the results back into the instrumentation and collection stages. The
visualization/alerting stage phase presents a consumable view of the system state. It can display information in
near real time by using a series of dashboards. And it can generate reports, graphs, and charts to provide a
historical view of the data that can help identify long-term trends. If information indicates that a KPI is likely to
exceed acceptable bounds, this stage can also trigger an alert to an operator. In some cases, an alert can also be
used to trigger an automated process that attempts to take corrective actions, such as autoscaling.
Note that these steps constitute a continuous-flow process where the stages are happening in parallel. Ideally, all
the phases should be dynamically configurable. At some points, especially when a system has been newly
deployed or is experiencing problems, it might be necessary to gather extended data on a more frequent basis. At
other times, it should be possible to revert to capturing a base level of essential information to verify that the
system is functioning properly.
Additionally, the entire monitoring process should be considered a live, ongoing solution that's subject to fine-
tuning and improvements as a result of feedback. For example, you might start with measuring many factors to
determine system health. Analysis over time might lead to a refinement as you discard measures that aren't
relevant, enabling you to more precisely focus on the data that you need while minimizing background noise.

Sources of monitoring and diagnostic data

The information that the monitoring process uses can come from several sources, as illustrated in Figure 1. At the
application level, information comes from trace logs incorporated into the code of the system. Developers should
follow a standard approach for tracking the flow of control through their code. For example, an entry to a method
can emit a trace message that specifies the name of the method, the current time, the value of each parameter, and
any other pertinent information. Recording the entry and exit times can also prove useful.
You should log all exceptions and warnings, and ensure that you retain a full trace of any nested exceptions and
warnings. Ideally, you should also capture information that identifies the user who is running the code, together
with activity correlation information (to track requests as they pass through the system). And you should log
attempts to access all resources such as message queues, databases, files, and other dependent services. This
information can be used for metering and auditing purposes.
Many applications use libraries and frameworks to perform common tasks such as accessing a data store or
communicating over a network. These frameworks might be configurable to provide their own trace messages
and raw diagnostic information, such as transaction rates and data transmission successes and failures.

NOTE
Many modern frameworks automatically publish performance and trace events. Capturing this information is simply a
matter of providing a means to retrieve and store it where it can be processed and analyzed.

The operating system where the application is running can be a source of low-level system-wide information, such
as performance counters that indicate I/O rates, memory utilization, and CPU usage. Operating system errors
(such as the failure to open a file correctly) might also be reported.
You should also consider the underlying infrastructure and components on which your system runs. Virtual
machines, virtual networks, and storage services can all be sources of important infrastructure-level performance
counters and other diagnostic data.
If your application uses other external services, such as a web server or database management system, these
services might publish their own trace information, logs, and performance counters. Examples include SQL Server
Dynamic Management Views for tracking operations performed against a SQL Server database, and IIS trace logs
for recording requests made to a web server.
As the components of a system are modified and new versions are deployed, it's important to be able to attribute
issues, events, and metrics to each version. This information should be tied back to the release pipeline so that
problems with a specific version of a component can be tracked quickly and rectified.
Security issues might occur at any point in the system. For example, a user might attempt to sign in with an invalid
user ID or password. An authenticated user might try to obtain unauthorized access to a resource. Or a user might
provide an invalid or outdated key to access encrypted information. Security-related information for successful
and failing requests should always be logged.
The section Instrumenting an application contains more guidance on the information that you should capture. But
you can use a variety of strategies to gather this information:
Application/system monitoring . This strategy uses internal sources within the application, application
frameworks, operating system, and infrastructure. The application code can generate its own monitoring
data at notable points during the lifecycle of a client request. The application can include tracing statements
that might be selectively enabled or disabled as circumstances dictate. It might also be possible to inject
diagnostics dynamically by using a diagnostics framework. These frameworks typically provide plug-ins
that can attach to various instrumentation points in your code and capture trace data at these points.
Additionally, your code and/or the underlying infrastructure might raise events at critical points. Monitoring
agents that are configured to listen for these events can record the event information.
Real user monitoring . This approach records the interactions between a user and the application and
observes the flow of each request and response. This information can have a two-fold purpose: it can be
used for metering usage by each user, and it can be used to determine whether users are receiving a
suitable quality of service (for example, fast response times, low latency, and minimal errors). You can use
the captured data to identify areas of concern where failures occur most often. You can also use the data to
identify elements where the system slows down, possibly due to hotspots in the application or some other
form of bottleneck. If you implement this approach carefully, it might be possible to reconstruct users'
flows through the application for debugging and testing purposes.

IMPORTANT
You should consider the data that's captured by monitoring real users to be highly sensitive because it might include
confidential material. If you save captured data, store it securely. If you want to use the data for performance
monitoring or debugging purposes, strip out all personally identifiable information first.

Synthetic user monitoring . In this approach, you write your own test client that simulates a user and
performs a configurable but typical series of operations. You can track the performance of the test client to
help determine the state of the system. You can also use multiple instances of the test client as part of a
load-testing operation to establish how the system responds under stress, and what sort of monitoring
output is generated under these conditions.

NOTE
You can implement real and synthetic user monitoring by including code that traces and times the execution of
method calls and other critical parts of an application.

Profiling . This approach is primarily targeted at monitoring and improving application performance.
Rather than operating at the functional level of real and synthetic user monitoring, it captures lower-level
information as the application runs. You can implement profiling by using periodic sampling of the
execution state of an application (determining which piece of code that the application is running at a given
point in time). You can also use instrumentation that inserts probes into the code at important junctures
(such as the start and end of a method call) and records which methods were invoked, at what time, and
how long each call took. You can then analyze this data to determine which parts of the application might
cause performance problems.
Endpoint monitoring . This technique uses one or more diagnostic endpoints that the application exposes
 specifically to enable monitoring. An endpoint provides a pathway into the application code and can return
information about the health of the system. Different endpoints can focus on various aspects of the
functionality. You can write your own diagnostics client that sends periodic requests to these endpoints and
assimilate the responses. For more information, see the Health Endpoint Monitoring pattern.
For maximum coverage, you should use a combination of these techniques.

Instrumenting an application
Instrumentation is a critical part of the monitoring process. You can make meaningful decisions about the
performance and health of a system only if you first capture the data that enables you to make these decisions.
The information that you gather by using instrumentation should be sufficient to enable you to assess
performance, diagnose problems, and make decisions without requiring you to sign in to a remote production
server to perform tracing (and debugging) manually. Instrumentation data typically comprises metrics and
information that's written to trace logs.
The contents of a trace log can be the result of textual data that's written by the application or binary data that's
created as the result of a trace event (if the application is using Event Tracing for Windows--ETW). They can also be
generated from system logs that record events arising from parts of the infrastructure, such as a web server.
Textual log messages are often designed to be human-readable, but they should also be written in a format that
enables an automated system to parse them easily.
You should also categorize logs. Don't write all trace data to a single log, but use separate logs to record the trace
output from different operational aspects of the system. You can then quickly filter log messages by reading from
the appropriate log rather than having to process a single lengthy file. Never write information that has different
security requirements (such as audit information and debugging data) to the same log.

NOTE
A log might be implemented as a file on the file system, or it might be held in some other format, such as a blob in blob
storage. Log information might also be held in more structured storage, such as rows in a table.

Metrics will generally be a measure or count of some aspect or resource in the system at a specific time, with one
or more associated tags or dimensions (sometimes called a sample). A single instance of a metric is usually not
useful in isolation. Instead, metrics have to be captured over time. The key issue to consider is which metrics you
should record and how frequently. Generating data for metrics too often can impose a significant additional load
on the system, whereas capturing metrics infrequently might cause you to miss the circumstances that lead to a
significant event. The considerations will vary from metric to metric. For example, CPU utilization on a server
might vary significantly from second to second, but high utilization becomes an issue only if it's long-lived over a
number of minutes.
Information for correlating data
You can easily monitor individual system-level performance counters, capture metrics for resources, and obtain
application trace information from various log files. But some forms of monitoring require the analysis and
diagnostics stage in the monitoring pipeline to correlate the data that's retrieved from several sources. This data
might take several forms in the raw data, and the analysis process must be provided with sufficient
instrumentation data to be able to map these different forms. For example, at the application framework level, a
task might be identified by a thread ID. Within an application, the same work might be associated with the user ID
for the user who is performing that task.
Also, there's unlikely to be a 1:1 mapping between threads and user requests, because asynchronous operations
might reuse the same threads to perform operations on behalf of more than one user. To complicate matters
further, a single request might be handled by more than one thread as execution flows through the system. If
possible, associate each request with a unique activity ID that's propagated through the system as part of the
request context. (The technique for generating and including activity IDs in trace information depends on the
technology that's used to capture the trace data.)
All monitoring data should be timestamped in the same way. For consistency, record all dates and times by using
Coordinated Universal Time. This will help you more easily trace sequences of events.

NOTE
Computers operating in different time zones and networks might not be synchronized. Don't depend on using timestamps
alone for correlating instrumentation data that spans multiple machines.

Information to include in the instrumentation data

Consider the following points when you're deciding which instrumentation data you need to collect:
Make sure that information captured by trace events is machine and human readable. Adopt well-defined
schemas for this information to facilitate automated processing of log data across systems, and to provide
consistency to operations and engineering staff reading the logs. Include environmental information, such
as the deployment environment, the machine on which the process is running, the details of the process,
and the call stack.
Enable profiling only when necessary because it can impose a significant overhead on the system. Profiling
by using instrumentation records an event (such as a method call) every time it occurs, whereas sampling
records only selected events. The selection can be time-based (once every n seconds), or frequency-based
(once every n requests). If events occur very frequently, profiling by instrumentation might cause too much
of a burden and itself affect overall performance. In this case, the sampling approach might be preferable.
However, if the frequency of events is low, sampling might miss them. In this case, instrumentation might
be the better approach.
Provide sufficient context to enable a developer or administrator to determine the source of each request.
This might include some form of activity ID that identifies a specific instance of a request. It might also
include information that can be used to correlate this activity with the computational work performed and
the resources used. Note that this work might cross process and machine boundaries. For metering, the
context should also include (either directly or indirectly via other correlated information) a reference to the
customer who caused the request to be made. This context provides valuable information about the
application state at the time that the monitoring data was captured.
Record all requests, and the locations or regions from which these requests are made. This information can
assist in determining whether there are any location-specific hotspots. This information can also be useful
in determining whether to repartition an application or the data that it uses.
Record and capture the details of exceptions carefully. Often, critical debug information is lost as a result of
poor exception handling. Capture the full details of exceptions that the application throws, including any
inner exceptions and other context information. Include the call stack if possible.
Be consistent in the data that the different elements of your application capture, because this can assist in
analyzing events and correlating them with user requests. Consider using a comprehensive and
configurable logging package to gather information, rather than depending on developers to adopt the
same approach as they implement different parts of the system. Gather data from key performance
counters, such as the volume of I/O being performed, network utilization, number of requests, memory use,
and CPU utilization. Some infrastructure services might provide their own specific performance counters,
such as the number of connections to a database, the rate at which transactions are being performed, and
the number of transactions that succeed or fail. Applications might also define their own specific
performance counters.
Log all calls made to external services, such as database systems, web services, or other system-level
 services that are part of the infrastructure. Record information about the time taken to perform each call,
and the success or failure of the call. If possible, capture information about all retry attempts and failures
for any transient errors that occur.
Ensuring compatibility with telemetry systems
In many cases, the information that instrumentation produces is generated as a series of events and passed to a
separate telemetry system for processing and analysis. A telemetry system is typically independent of any specific
application or technology, but it expects information to follow a specific format that's usually defined by a schema.
The schema effectively specifies a contract that defines the data fields and types that the telemetry system can
ingest. The schema should be generalized to allow for data arriving from a range of platforms and devices.
A common schema should include fields that are common to all instrumentation events, such as the event name,
the event time, the IP address of the sender, and the details that are required for correlating with other events
(such as a user ID, a device ID, and an application ID). Remember that any number of devices might raise events, so
the schema should not depend on the device type. Additionally, various devices might raise events for the same
application; the application might support roaming or some other form of cross-device distribution.
The schema might also include domain fields that are relevant to a particular scenario that's common across
different applications. This might be information about exceptions, application start and end events, and success
and/or failure of web service API calls. All applications that use the same set of domain fields should emit the
same set of events, enabling a set of common reports and analytics to be built.
Finally, a schema might contain custom fields for capturing the details of application-specific events.
Best practices for instrumenting applications
The following list summarizes best practices for instrumenting a distributed application running in the cloud.
Make logs easy to read and easy to parse. Use structured logging where possible. Be concise and
descriptive in log messages.
In all logs, identify the source and provide context and timing information as each log record is written.
Use the same time zone and format for all timestamps. This will help to correlate events for operations that
span hardware and services running in different geographic regions.
Categorize logs and write messages to the appropriate log file.
Do not disclose sensitive information about the system or personal information about users. Scrub this
information before it's logged, but ensure that the relevant details are retained. For example, remove the ID
and password from any database connection strings, but write the remaining information to the log so that
an analyst can determine that the system is accessing the correct database. Log all critical exceptions, but
enable the administrator to turn logging on and off for lower levels of exceptions and warnings. Also,
capture and log all retry logic information. This data can be useful in monitoring the transient health of the
system.
Trace out of process calls, such as requests to external web services or databases.
Don't mix log messages with different security requirements in the same log file. For example, don't write
debug and audit information to the same log.
With the exception of auditing events, make sure that all logging calls are fire-and-forget operations that do
not block the progress of business operations. Auditing events are exceptional because they are critical to
the business and can be classified as a fundamental part of business operations.
Make sure that logging is extensible and does not have any direct dependencies on a concrete target. For
example, rather than writing information by using System.Diagnostics.Trace, define an abstract interface
(such as ILogger) that exposes logging methods and that can be implemented through any appropriate
means.
 Make sure that all logging is fail-safe and never triggers any cascading errors. Logging must not throw any
exceptions.
Treat instrumentation as an ongoing iterative process and review logs regularly, not just when there is a
problem.

Collecting and storing data

The collection stage of the monitoring process is concerned with retrieving the information that instrumentation
generates, formatting this data to make it easier for the analysis/diagnosis stage to consume, and saving the
transformed data in reliable storage. The instrumentation data that you gather from different parts of a distributed
system can be held in a variety of locations and with varying formats. For example, your application code might
generate trace log files and generate application event log data, whereas performance counters that monitor key
aspects of the infrastructure that your application uses can be captured through other technologies. Any third-
party components and services that your application uses might provide instrumentation information in different
formats, by using separate trace files, blob storage, or even a custom data store.
Data collection is often performed through a collection service that can run autonomously from the application
that generates the instrumentation data. Figure 2 illustrates an example of this architecture, highlighting the
instrumentation data-collection subsystem.

Figure 2 - Collecting instrumentation data.

Note that this is a simplified view. The collection service is not necessarily a single process and might comprise
many constituent parts running on different machines, as described in the following sections. Additionally, if the
analysis of some telemetry data must be performed quickly (hot analysis, as described in the section Supporting
hot, warm, and cold analysis later in this document), local components that operate outside the collection service
might perform the analysis tasks immediately. Figure 2 depicts this situation for selected events. After analytical
processing, the results can be sent directly to the visualization and alerting subsystem. Data that's subjected to
warm or cold analysis is held in storage while it awaits processing.
For Azure applications and services, Azure Diagnostics provides one possible solution for capturing data. Azure
Diagnostics gathers data from the following sources for each compute node, aggregates it, and then uploads it to
Azure Storage:
IIS logs
IIS Failed Request logs
Windows event logs
Performance counters
Crash dumps
 Azure Diagnostics infrastructure logs
Custom error logs
.NET EventSource
Manifest-based ETW
For more information, see the article Azure: Telemetry Basics and Troubleshooting.
Strategies for collecting instrumentation data
Considering the elastic nature of the cloud, and to avoid the necessity of manually retrieving telemetry data from
every node in the system, you should arrange for the data to be transferred to a central location and consolidated.
In a system that spans multiple datacenters, it might be useful to first collect, consolidate, and store data on a
region-by-region basis, and then aggregate the regional data into a single central system.
To optimize the use of bandwidth, you can elect to transfer less urgent data in chunks, as batches. However, the
data must not be delayed indefinitely, especially if it contains time-sensitive information.
Pulling and pushing instrumentation data
The instrumentation data-collection subsystem can actively retrieve instrumentation data from the various logs
and other sources for each instance of the application (the pull model). Or, it can act as a passive receiver that
waits for the data to be sent from the components that constitute each instance of the application (the push
model).
One approach to implementing the pull model is to use monitoring agents that run locally with each instance of
the application. A monitoring agent is a separate process that periodically retrieves (pulls) telemetry data collected
at the local node and writes this information directly to centralized storage that all instances of the application
share. This is the mechanism that Azure Diagnostics implements. Each instance of an Azure web or worker role
can be configured to capture diagnostic and other trace information that's stored locally. The monitoring agent
that runs alongside each instance copies the specified data to Azure Storage. The article Enabling Diagnostics in
Azure Cloud Services and Virtual Machines provides more details on this process. Some elements, such as IIS logs,
crash dumps, and custom error logs, are written to blob storage. Data from the Windows event log, ETW events,
and performance counters is recorded in table storage. Figure 3 illustrates this mechanism.

Figure 3 - Using a monitoring agent to pull information and write to shared storage.
 NOTE
Using a monitoring agent is ideally suited to capturing instrumentation data that's naturally pulled from a data source. An
example is information from SQL Server Dynamic Management Views or the length of an Azure Service Bus queue.

It's feasible to use the approach just described to store telemetry data for a small-scale application running on a
limited number of nodes in a single location. However, a complex, highly scalable, global cloud application might
generate huge volumes of data from hundreds of web and worker roles, database shards, and other services. This
flood of data can easily overwhelm the I/O bandwidth available with a single, central location. Therefore, your
telemetry solution must be scalable to prevent it from acting as a bottleneck as the system expands. Ideally, your
solution should incorporate a degree of redundancy to reduce the risks of losing important monitoring
information (such as auditing or billing data) if part of the system fails.
To address these issues, you can implement queuing, as shown in Figure 4. In this architecture, the local
monitoring agent (if it can be configured appropriately) or custom data-collection service (if not) posts data to a
queue. A separate process running asynchronously (the storage writing service in Figure 4) takes the data in this
queue and writes it to shared storage. A message queue is suitable for this scenario because it provides "at least
once" semantics that help ensure that queued data will not be lost after it's posted. You can implement the storage
writing service by using a separate worker role.

Figure 4 - Using a queue to buffer instrumentation data.

The local data-collection service can add data to a queue immediately after it's received. The queue acts as a buffer,
and the storage writing service can retrieve and write the data at its own pace. By default, a queue operates on a
first-in, first-out basis. But you can prioritize messages to accelerate them through the queue if they contain data
that must be handled more quickly. For more information, see the Priority Queue pattern. Alternatively, you can
use different channels (such as Service Bus topics) to direct data to different destinations depending on the form
of analytical processing that's required.
For scalability, you can run multiple instances of the storage writing service. If there is a high volume of events,
you can use an event hub to dispatch the data to different compute resources for processing and storage.
Consolidating instrumentation data
The instrumentation data that the data-collection service retrieves from a single instance of an application gives a
localized view of the health and performance of that instance. To assess the overall health of the system, it's
necessary to consolidate some aspects of the data in the local views. You can perform this after the data has been
stored, but in some cases, you can also achieve it as the data is collected. Rather than being written directly to
shared storage, the instrumentation data can pass through a separate data consolidation service that combines
data and acts as a filter and cleanup process. For example, instrumentation data that includes the same correlation
information such as an activity ID can be amalgamated. (It's possible that a user starts performing a business
operation on one node and then gets transferred to another node in the event of node failure, or depending on
how load balancing is configured.) This process can also detect and remove any duplicated data (always a
possibility if the telemetry service uses message queues to push instrumentation data out to storage). Figure 5
illustrates an example of this structure.

Figure 5 - Using a separate service to consolidate and clean up instrumentation data.

Storing instrumentation data
The previous discussions have depicted a rather simplistic view of the way in which instrumentation data is stored.
In reality, it can make sense to store the different types of information by using technologies that are most
appropriate to the way in which each type is likely to be used.
For example, Azure blob and table storage have some similarities in the way in which they're accessed. But they
have limitations in the operations that you can perform by using them, and the granularity of the data that they
hold is quite different. If you need to perform more analytical operations or require full-text search capabilities on
the data, it might be more appropriate to use data storage that provides capabilities that are optimized for specific
types of queries and data access. For example:
Performance counter data can be stored in a SQL database to enable ad hoc analysis.
Trace logs might be better stored in Azure Cosmos DB.
Security information can be written to HDFS.
Information that requires full-text search can be stored through Elasticsearch (which can also speed searches
by using rich indexing).
You can implement an additional service that periodically retrieves the data from shared storage, partitions and
filters the data according to its purpose, and then writes it to an appropriate set of data stores as shown in Figure
6. An alternative approach is to include this functionality in the consolidation and cleanup process and write the
data directly to these stores as it's retrieved rather than saving it in an intermediate shared storage area. Each
approach has its advantages and disadvantages. Implementing a separate partitioning service lessens the load on
the consolidation and cleanup service, and it enables at least some of the partitioned data to be regenerated if
necessary (depending on how much data is retained in shared storage). However, it consumes additional
resources. Also, there might be a delay between the receipt of instrumentation data from each application instance
and the conversion of this data into actionable information.

Figure 6 - Partitioning data according to analytical and storage requirements.

The same instrumentation data might be required for more than one purpose. For example, performance counters
can be used to provide a historical view of system performance over time. This information might be combined
with other usage data to generate customer billing information. In these situations, the same data might be sent to
more than one destination, such as a document database that can act as a long-term store for holding billing
information, and a multidimensional store for handling complex performance analytics.
You should also consider how urgently the data is required. Data that provides information for alerting must be
accessed quickly, so it should be held in fast data storage and indexed or structured to optimize the queries that
the alerting system performs. In some cases, it might be necessary for the telemetry service that gathers the data
on each node to format and save data locally so that a local instance of the alerting system can quickly notify you
of any issues. The same data can be dispatched to the storage writing service shown in the previous diagrams and
stored centrally if it's also required for other purposes.
Information that's used for more considered analysis, for reporting, and for spotting historical trends is less urgent
and can be stored in a manner that supports data mining and ad hoc queries. For more information, see the
section Supporting hot, warm, and cold analysis later in this document.
Log rotation and data retention
Instrumentation can generate considerable volumes of data. This data can be held in several places, starting with
the raw log files, trace files, and other information captured at each node to the consolidated, cleaned, and
partitioned view of this data held in shared storage. In some cases, after the data has been processed and
transferred, the original raw source data can be removed from each node. In other cases, it might be necessary or
simply useful to save the raw information. For example, data that's generated for debugging purposes might be
best left available in its raw form but can then be discarded quickly after any bugs have been rectified.
Performance data often has a longer life so that it can be used for spotting performance trends and for capacity
planning. The consolidated view of this data is usually kept online for a finite period to enable fast access. After
that, it can be archived or discarded. Data gathered for metering and billing customers might need to be saved
indefinitely. Additionally, regulatory requirements might dictate that information collected for auditing and
security purposes also needs to be archived and saved. This data is also sensitive and might need to be encrypted
or otherwise protected to prevent tampering. You should never record users' passwords or other information that
might be used to commit identity fraud. Such details should be scrubbed from the data before it's stored.
Down-sampling
It's useful to store historical data so you can spot long-term trends. Rather than saving old data in its entirety, it
might be possible to down-sample the data to reduce its resolution and save storage costs. As an example, rather
than saving minute-by-minute performance indicators, you can consolidate data that's more than a month old to
form an hour-by-hour view.
Best practices for collecting and storing logging information
The following list summarizes best practices for capturing and storing logging information:
The monitoring agent or data-collection service should run as an out-of-process service and should be
simple to deploy.
All output from the monitoring agent or data-collection service should be an agnostic format that's
independent of the machine, operating system, or network protocol. For example, emit information in a
self-describing format such as JSON, MessagePack, or Protobuf rather than ETL/ETW. Using a standard
format enables the system to construct processing pipelines; components that read, transform, and send
data in the agreed format can be easily integrated.
The monitoring and data-collection process must be fail-safe and must not trigger any cascading error
conditions.
In the event of a transient failure in sending information to a data sink, the monitoring agent or data-
collection service should be prepared to reorder telemetry data so that the newest information is sent first.
 (The monitoring agent/data-collection service might elect to drop the older data, or save it locally and
transmit it later to catch up, at its own discretion.)

Analyzing data and diagnosing issues

An important part of the monitoring and diagnostics process is analyzing the gathered data to obtain a picture of
the overall well-being of the system. You should have defined your own KPIs and performance metrics, and it's
important to understand how you can structure the data that has been gathered to meet your analysis
requirements. It's also important to understand how the data that's captured in different metrics and log files is
correlated, because this information can be key to tracking a sequence of events and help diagnose problems that
arise.
As described in the section Consolidating instrumentation data, the data for each part of the system is typically
captured locally, but it generally needs to be combined with data generated at other sites that participate in the
system. This information requires careful correlation to ensure that data is combined accurately. For example, the
usage data for an operation might span a node that hosts a website to which a user connects, a node that runs a
separate service accessed as part of this operation, and data storage held on another node. This information needs
to be tied together to provide an overall view of the resource and processing usage for the operation. Some
preprocessing and filtering of data might occur on the node on which the data is captured, whereas aggregation
and formatting are more likely to occur on a central node.
Supporting hot, warm, and cold analysis
Analyzing and reformatting data for visualization, reporting, and alerting purposes can be a complex process that
consumes its own set of resources. Some forms of monitoring are time-critical and require immediate analysis of
data to be effective. This is known as hot analysis. Examples include the analyses that are required for alerting and
some aspects of security monitoring (such as detecting an attack on the system). Data that's required for these
purposes must be quickly available and structured for efficient processing. In some cases, it might be necessary to
move the analysis processing to the individual nodes where the data is held.
Other forms of analysis are less time-critical and might require some computation and aggregation after the raw
data has been received. This is called warm analysis. Performance analysis often falls into this category. In this
case, an isolated, single performance event is unlikely to be statistically significant. (It might be caused by a sudden
spike or glitch.) The data from a series of events should provide a more reliable picture of system performance.
Warm analysis can also be used to help diagnose health issues. A health event is typically processed through hot
analysis and can raise an alert immediately. An operator should be able to drill into the reasons for the health
event by examining the data from the warm path. This data should contain information about the events leading
up to the issue that caused the health event.
Some types of monitoring generate more long-term data. This analysis can be performed at a later date, possibly
according to a predefined schedule. In some cases, the analysis might need to perform complex filtering of large
volumes of data captured over a period of time. This is called cold analysis. The key requirement is that the data is
stored safely after it has been captured. For example, usage monitoring and auditing require an accurate picture of
the state of the system at regular points in time, but this state information does not have to be available for
processing immediately after it has been gathered.
An operator can also use cold analysis to provide the data for predictive health analysis. The operator can gather
historical information over a specified period and use it in conjunction with the current health data (retrieved from
the hot path) to spot trends that might soon cause health issues. In these cases, it might be necessary to raise an
alert so that corrective action can be taken.
Correlating data
The data that instrumentation captures can provide a snapshot of the system state, but the purpose of analysis is
to make this data actionable. For example:
 What has caused an intense I/O loading at the system level at a specific time?
Is it the result of a large number of database operations?
Is this reflected in the database response times, the number of transactions per second, and application
response times at the same juncture?
If so, one remedial action that might reduce the load might be to shard the data over more servers. In addition,
exceptions can arise as a result of a fault in any level of the system. An exception in one level often triggers
another fault in the level above.
For these reasons, you need to be able to correlate the different types of monitoring data at each level to produce
an overall view of the state of the system and the applications that are running on it. You can then use this
information to make decisions about whether the system is functioning acceptably or not, and determine what can
be done to improve the quality of the system.
As described in the section Information for correlating data, you must ensure that the raw instrumentation data
includes sufficient context and activity ID information to support the required aggregations for correlating events.
Additionally, this data might be held in different formats, and it might be necessary to parse this information to
convert it into a standardized format for analysis.
Troubleshooting and diagnosing issues
Diagnosis requires the ability to determine the cause of faults or unexpected behavior, including performing root
cause analysis. The information that's required typically includes:
Detailed information from event logs and traces, either for the entire system or for a specified subsystem
during a specified time window.
Complete stack traces resulting from exceptions and faults of any specified level that occur within the system
or a specified subsystem during a specified period.
Crash dumps for any failed processes either anywhere in the system or for a specified subsystem during a
specified time window.
Activity logs recording the operations that are performed either by all users or for selected users during a
specified period.
Analyzing data for troubleshooting purposes often requires a deep technical understanding of the system
architecture and the various components that compose the solution. As a result, a large degree of manual
intervention is often required to interpret the data, establish the cause of problems, and recommend an
appropriate strategy to correct them. It might be appropriate simply to store a copy of this information in its
original format and make it available for cold analysis by an expert.

Visualizing data and raising alerts

An important aspect of any monitoring system is the ability to present the data in such a way that an operator can
quickly spot any trends or problems. Also important is the ability to quickly inform an operator if a significant
event has occurred that might require attention.
Data presentation can take several forms, including visualization by using dashboards, alerting, and reporting.
Visualization by using dashboards
The most common way to visualize data is to use dashboards that can display information as a series of charts,
graphs, or some other illustration. These items can be parameterized, and an analyst should be able to select the
important parameters (such as the time period) for any specific situation.
Dashboards can be organized hierarchically. Top-level dashboards can give an overall view of each aspect of the
system but enable an operator to drill down to the details. For example, a dashboard that depicts the overall disk
I/O for the system should allow an analyst to view the I/O rates for each individual disk to ascertain whether one
or more specific devices account for a disproportionate volume of traffic. Ideally, the dashboard should also
display related information, such as the source of each request (the user or activity) that's generating this I/O. This
information can then be used to determine whether (and how) to spread the load more evenly across devices, and
whether the system would perform better if more devices were added.
A dashboard might also use color-coding or some other visual cues to indicate values that appear anomalous or
that are outside an expected range. Using the previous example:
A disk with an I/O rate that's approaching its maximum capacity over an extended period (a hot disk) can be
highlighted in red.
A disk with an I/O rate that periodically runs at its maximum limit over short periods (a warm disk) can be
highlighted in yellow.
A disk that's exhibiting normal usage can be displayed in green.
Note that for a dashboard system to work effectively, it must have the raw data to work with. If you are building
your own dashboard system, or using a dashboard developed by another organization, you must understand
which instrumentation data you need to collect, at what levels of granularity, and how it should be formatted for
the dashboard to consume.
A good dashboard does not only display information, it also enables an analyst to pose ad hoc questions about
that information. Some systems provide management tools that an operator can use to perform these tasks and
explore the underlying data. Alternatively, depending on the repository that's used to hold this information, it
might be possible to query this data directly, or import it into tools such as Microsoft Excel for further analysis and
reporting.

NOTE
You should restrict access to dashboards to authorized personnel, because this information might be commercially sensitive.
You should also protect the underlying data for dashboards to prevent users from changing it.

Raising alerts
Alerting is the process of analyzing the monitoring and instrumentation data and generating a notification if a
significant event is detected.
Alerting helps ensure that the system remains healthy, responsive, and secure. It's an important part of any
system that makes performance, availability, and privacy guarantees to the users where the data might need to be
acted on immediately. An operator might need to be notified of the event that triggered the alert. Alerting can also
be used to invoke system functions such as autoscaling.
Alerting usually depends on the following instrumentation data:
Security events . If the event logs indicate that repeated authentication and/or authorization failures are
occurring, the system might be under attack and an operator should be informed.
Performance metrics . The system must quickly respond if a particular performance metric exceeds a
specified threshold.
Availability information . If a fault is detected, it might be necessary to quickly restart one or more
subsystems, or fail over to a backup resource. Repeated faults in a subsystem might indicate more serious
concerns.
Operators might receive alert information by using many delivery channels such as email, a pager device, or an
SMS text message. An alert might also include an indication of how critical a situation is. Many alerting systems
support subscriber groups, and all operators who are members of the same group can receive the same set of
alerts.
An alerting system should be customizable, and the appropriate values from the underlying instrumentation data
can be provided as parameters. This approach enables an operator to filter data and focus on those thresholds or
combinations of values that are of interest. Note that in some cases, the raw instrumentation data can be provided
to the alerting system. In other situations, it might be more appropriate to supply aggregated data. (For example,
an alert can be triggered if the CPU utilization for a node has exceeded 90 percent over the last 10 minutes). The
details provided to the alerting system should also include any appropriate summary and context information.
This data can help reduce the possibility that false-positive events will trip an alert.
Reporting
Reporting is used to generate an overall view of the system. It might incorporate historical data in addition to
current information. Reporting requirements themselves fall into two broad categories: operational reporting and
security reporting.
Operational reporting typically includes the following aspects:
Aggregating statistics that you can use to understand resource utilization of the overall system or specified
subsystems during a specified time window.
Identifying trends in resource usage for the overall system or specified subsystems during a specified period.
Monitoring the exceptions that have occurred throughout the system or in specified subsystems during a
specified period.
Determining the efficiency of the application in terms of the deployed resources, and understanding whether
the volume of resources (and their associated cost) can be reduced without affecting performance
unnecessarily.
Security reporting is concerned with tracking customers' use of the system. It can include:
Auditing user operations . This requires recording the individual requests that each user performs, together
with dates and times. The data should be structured to enable an administrator to quickly reconstruct the
sequence of operations that a user performs over a specified period.
Tracking resource use by user . This requires recording how each request for a user accesses the various
resources that compose the system, and for how long. An administrator must be able to use this data to
generate a utilization report by user over a specified period, possibly for billing purposes.
In many cases, batch processes can generate reports according to a defined schedule. (Latency is not normally an
issue.) But they should also be available for generation on an ad hoc basis if needed. As an example, if you are
storing data in a relational database such as Azure SQL Database, you can use a tool such as SQL Server
Reporting Services to extract and format data and present it as a set of reports.

Related patterns and guidance

Autoscaling guidance describes how to decrease management overhead by reducing the need for an operator
to continually monitor the performance of a system and make decisions about adding or removing resources.
Health Endpoint Monitoring pattern describes how to implement functional checks within an application that
external tools can access through exposed endpoints at regular intervals.
Priority Queue pattern shows how to prioritize queued messages so that urgent requests are received and can
be processed before less urgent messages.

Next steps
Azure Monitor overview
Monitor, diagnose, and troubleshoot Microsoft Azure Storage
Overview of alerts in Microsoft Azure
View service health notifications by using the Azure portal
What is Application Insights?
Performance diagnostics for Azure virtual machines
Download and install SQL Server Data Tools (SSDT) for Visual Studio
 Retry guidance for Azure services
12/18/2020 • 41 minutes to read • Edit Online

Most Azure services and client SDKs include a retry mechanism. However, these differ because each service has
different characteristics and requirements, and so each retry mechanism is tuned to a specific service. This guide
summarizes the retry mechanism features for the majority of Azure services, and includes information to help you
use, adapt, or extend the retry mechanism for that service.
For general guidance on handling transient faults, and retrying connections and operations against services and
resources, see Retry guidance.
The following table summarizes the retry features for the Azure services described in this guidance.

P O L IC Y T EL EM ET RY
SERVIC E RET RY C A PA B IL IT IES C O N F IGURAT IO N SC O P E F EAT URES

Azure Active Native in ADAL Embedded into ADAL Internal None

Director y library library

Cosmos DB Native in service Non-configurable Global TraceSource

Data Lake Store Native in client Non-configurable Individual operations None

Event Hubs Native in client Programmatic Client None

IoT Hub Native in client SDK Programmatic Client None

Azure Cache for Native in client Programmatic Client TextWriter

Redis

Search Native in client Programmatic Client ETW or Custom

Ser vice Bus Native in client Programmatic Namespace Manager, ETW

Messaging Factory,
and Client

Ser vice Fabric Native in client Programmatic Client None

SQL Database with Polly Declarative and Single statements or Custom

ADO.NET programmatic blocks of code

SQL Database with Native in client Programmatic Global per None

Entity Framework AppDomain

SQL Database with Native in client Programmatic Global per None

Entity Framework AppDomain
Core

Storage Native in client Programmatic Client and individual TraceSource

operations
 NOTE
For most of the Azure built-in retry mechanisms, there is currently no way apply a different retry policy for different types of
error or exception. You should configure a policy that provides the optimum average performance and availability. One way
to fine-tune the policy is to analyze log files to determine the type of transient faults that are occurring.

Azure Active Directory

Azure Active Directory (Azure AD) is a comprehensive identity and access management cloud solution that
combines core directory services, advanced identity governance, security, and application access management.
Azure AD also offers developers an identity management platform to deliver access control to their applications,
based on centralized policy and rules.

NOTE
For retry guidance on Managed Service Identity endpoints, see How to use an Azure VM Managed Service Identity (MSI) for
token acquisition.

Retry mechanism
There is a built-in retry mechanism for Azure Active Directory in the Active Directory Authentication Library
(ADAL). To avoid unexpected lockouts, we recommend that third-party libraries and application code do not retry
failed connections, but allow ADAL to handle retries.
Retry usage guidance
Consider the following guidelines when using Azure Active Directory:
When possible, use the ADAL library and the built-in support for retries.
If you are using the REST API for Azure Active Directory, retry the operation if the result code is 429 (Too Many
Requests) or an error in the 5xx range. Do not retry for any other errors.
For 429 errors, only retry after the time indicated in the Retr y-After header.
For 5xx errors, use exponential back-off, with the first retry at least 5 seconds after the response.
Do not retry on errors other than 429 and 5xx.
More information
Azure Active Directory Authentication Libraries

Cosmos DB
Cosmos DB is a fully managed multi-model database that supports schemaless JSON data. It offers configurable
and reliable performance, native JavaScript transactional processing, and is built for the cloud with elastic scale.
Retry mechanism
The DocumentClient class automatically retries failed attempts. To set the number of retries and the maximum wait
time, configure ConnectionPolicy.RetryOptions. Exceptions that the client raises are either beyond the retry policy
or are not transient errors.
If Cosmos DB throttles the client, it returns an HTTP 429 error. Check the status code in the
DocumentClientException .

Policy configuration
The following table shows the default settings for the RetryOptions class.
 SET T IN G DEFA ULT VA L UE DESC RIP T IO N

MaxRetryAttemptsOnThrottledRequests 9 The maximum number of retries if the

request fails because Cosmos DB
applied rate limiting on the client.

MaxRetryWaitTimeInSeconds 30 The maximum retry time in seconds.

Example

DocumentClient client = new DocumentClient(new Uri(endpoint), authKey);

var options = client.ConnectionPolicy.RetryOptions;
options.MaxRetryAttemptsOnThrottledRequests = 5;
options.MaxRetryWaitTimeInSeconds = 15;

Telemetry
Retry attempts are logged as unstructured trace messages through a .NET TraceSource . You must configure a
TraceListener to capture the events and write them to a suitable destination log.
For example, if you add the following to your App.config file, traces will be generated in a text file in the same
location as the executable:

<configuration>
<system.diagnostics>
<switches>
<add name="SourceSwitch" value="Verbose"/>
</switches>
<sources>
<source name="DocDBTrace" switchName="SourceSwitch" switchType="System.Diagnostics.SourceSwitch" >
<listeners>
<add name="MyTextListener" type="System.Diagnostics.TextWriterTraceListener"
traceOutputOptions="DateTime,ProcessId,ThreadId" initializeData="CosmosDBTrace.txt"></add>
</listeners>
</source>
</sources>
</system.diagnostics>
</configuration>

Event Hubs
Azure Event Hubs is a hyperscale telemetry ingestion service that collects, transforms, and stores millions of
events.
Retry mechanism
Retry behavior in the Azure Event Hubs Client Library is controlled by the RetryPolicy property on the
EventHubClient class. The default policy retries with exponential backoff when Azure Event Hub returns a transient
EventHubsException or an OperationCanceledException . Default retry policy for Event Hubs is to retry up to 9 times
with an exponential back-off time of up to 30 seconds .
Example

EventHubClient client = EventHubClient.CreateFromConnectionString("[event_hub_connection_string]");

client.RetryPolicy = RetryPolicy.Default;

More information
.NET Standard client library for Azure Event Hubs
IoT Hub
Azure IoT Hub is a service for connecting, monitoring, and managing devices to develop Internet of Things (IoT)
applications.
Retry mechanism
The Azure IoT device SDK can detect errors in the network, protocol, or application. Based on the error type, the
SDK checks whether a retry needs to be performed. If the error is recoverable, the SDK begins to retry using the
configured retry policy.
The default retry policy is exponential back-off with random jitter, but it can be configured.
Policy configuration
Policy configuration differs by language. For more details, see IoT Hub retry policy configuration.
More information
IoT Hub retry policy
Troubleshoot IoT Hub device disconnection

Azure Cache for Redis

Azure Cache for Redis is a fast data access and low latency cache service based on the popular open-source Redis
cache. It is secure, managed by Microsoft, and is accessible from any application in Azure.
The guidance in this section is based on using the StackExchange.Redis client to access the cache. A list of other
suitable clients can be found on the Redis website, and these may have different retry mechanisms.
Note that the StackExchange.Redis client uses multiplexing through a single connection. The recommended usage
is to create an instance of the client at application startup and use this instance for all operations against the cache.
For this reason, the connection to the cache is made only once, and so all of the guidance in this section is related
to the retry policy for this initial connection—and not for each operation that accesses the cache.
Retry mechanism
The StackExchange.Redis client uses a connection manager class that is configured through a set of options,
including:
ConnectRetr y . The number of times a failed connection to the cache will be retried.
ReconnectRetr yPolicy . The retry strategy to use.
ConnectTimeout . The maximum waiting time in milliseconds.
Policy configuration
Retry policies are configured programmatically by setting the options for the client before connecting to the cache.
This can be done by creating an instance of the ConfigurationOptions class, populating its properties, and
passing it to the Connect method.
The built-in classes support linear (constant) delay and exponential backoff with randomized retry intervals. You
can also create a custom retry policy by implementing the IReconnectRetr yPolicy interface.
The following example configures a retry strategy using exponential backoff.
 var deltaBackOffInMilliseconds = TimeSpan.FromSeconds(5).Milliseconds;
var maxDeltaBackOffInMilliseconds = TimeSpan.FromSeconds(20).Milliseconds;
var options = new ConfigurationOptions
{
EndPoints = {"localhost"},
ConnectRetry = 3,
ReconnectRetryPolicy = new ExponentialRetry(deltaBackOffInMilliseconds, maxDeltaBackOffInMilliseconds),
ConnectTimeout = 2000
};
ConnectionMultiplexer redis = ConnectionMultiplexer.Connect(options, writer);

Alternatively, you can specify the options as a string, and pass this to the Connect method. The
ReconnectRetr yPolicy property cannot be set this way, only through code.

var options = "localhost,connectRetry=3,connectTimeout=2000";

ConnectionMultiplexer redis = ConnectionMultiplexer.Connect(options, writer);

You can also specify options directly when you connect to the cache.

var conn = ConnectionMultiplexer.Connect("redis0:6380,redis1:6380,connectRetry=3");

For more information, see Stack Exchange Redis Configuration in the StackExchange.Redis documentation.
The following table shows the default settings for the built-in retry policy.

DEFA ULT VA L UE
C O N T EXT SET T IN G ( V 1. 2. 2) M EA N IN G

ConfigurationOptions ConnectRetry 3 The number of times to

repeat connect attempts
ConnectTimeout Maximum 5000 ms plus during the initial connection
SyncTimeout operation.
SyncTimeout 1000 Timeout (ms) for connect
operations. Not a delay
ReconnectRetryPolicy LinearRetry 5000 ms between retry attempts.
Time (ms) to allow for
synchronous operations.

Retry every 5000 ms.

NOTE
For synchronous operations, SyncTimeout can add to the end-to-end latency, but setting the value too low can cause
excessive timeouts. See How to troubleshoot Azure Cache for Redis. In general, avoid using synchronous operations, and use
asynchronous operations instead. For more information, see Pipelines and Multiplexers.

Retry usage guidance

Consider the following guidelines when using Azure Cache for Redis:
The StackExchange Redis client manages its own retries, but only when establishing a connection to the cache
when the application first starts. You can configure the connection timeout, the number of retry attempts, and
the time between retries to establish this connection, but the retry policy does not apply to operations against
the cache.
Instead of using a large number of retry attempts, consider falling back by accessing the original data source
instead.
Telemetry
You can collect information about connections (but not other operations) using a TextWriter .

var writer = new StringWriter();

ConnectionMultiplexer redis = ConnectionMultiplexer.Connect(options, writer);

An example of the output this generates is shown below.

localhost:6379,connectTimeout=2000,connectRetry=3
1 unique nodes specified
Requesting tie-break from localhost:6379 > __Booksleeve_TieBreak...
Allowing endpoints 00:00:02 to respond...
localhost:6379 faulted: SocketFailure on PING
localhost:6379 failed to nominate (Faulted)
> UnableToResolvePhysicalConnection on GET
No masters detected
localhost:6379: Standalone v2.0.0, master; keep-alive: 00:01:00; int: Connecting; sub: Connecting; not in use:
DidNotRespond
localhost:6379: int ops=0, qu=0, qs=0, qc=1, wr=0, sync=1, socks=2; sub ops=0, qu=0, qs=0, qc=0, wr=0, socks=2
Circular op-count snapshot; int: 0 (0.00 ops/s; spans 10s); sub: 0 (0.00 ops/s; spans 10s)
Sync timeouts: 0; fire and forget: 0; last heartbeat: -1s ago
resetting failing connections to retry...
retrying; attempts left: 2...
...

Examples
The following code example configures a constant (linear) delay between retries when initializing the
StackExchange.Redis client. This example shows how to set the configuration using a ConfigurationOptions
instance.
 using System;
using System.Collections.Generic;
using System.IO;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
using StackExchange.Redis;

namespace RetryCodeSamples
{
class CacheRedisCodeSamples
{
public async static Task Samples()
{
var writer = new StringWriter();
{
try
{
var retryTimeInMilliseconds = TimeSpan.FromSeconds(4).Milliseconds; // delay between
retries

// Using object-based configuration.

var options = new ConfigurationOptions
{
EndPoints = { "localhost" },
ConnectRetry = 3,
ReconnectRetryPolicy = new LinearRetry(retryTimeInMilliseconds)
};
ConnectionMultiplexer redis = ConnectionMultiplexer.Connect(options, writer);

// Store a reference to the multiplexer for use in the application.

}
catch
{
Console.WriteLine(writer.ToString());
throw;
}
}
}
}
}

The next example sets the configuration by specifying the options as a string. The connection timeout is the
maximum period of time to wait for a connection to the cache, not the delay between retry attempts. Note that the
ReconnectRetr yPolicy property can only be set by code.
 using System.Collections.Generic;
using System.IO;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
using StackExchange.Redis;

namespace RetryCodeSamples
{
class CacheRedisCodeSamples
{
public async static Task Samples()
{
var writer = new StringWriter();
{
try
{
// Using string-based configuration.
var options = "localhost,connectRetry=3,connectTimeout=2000";
ConnectionMultiplexer redis = ConnectionMultiplexer.Connect(options, writer);

// Store a reference to the multiplexer for use in the application.

}
catch
{
Console.WriteLine(writer.ToString());
throw;
}
}
}
}
}

For more examples, see Configuration on the project website.

More information
Redis website

Azure Search
Azure Search can be used to add powerful and sophisticated search capabilities to a website or application, quickly
and easily tune search results, and construct rich and fine-tuned ranking models.
Retry mechanism
Retry behavior in the Azure Search SDK is controlled by the SetRetryPolicy method on the SearchServiceClient
and SearchIndexClient classes. The default policy retries with exponential backoff when Azure Search returns a 5xx
or 408 (Request Timeout) response.
Telemetry
Trace with ETW or by registering a custom trace provider. For more information, see the AutoRest documentation.

Service Bus
Service Bus is a cloud messaging platform that provides loosely coupled message exchange with improved scale
and resiliency for components of an application, whether hosted in the cloud or on-premises.
Retry mechanism
Service Bus implements retries using implementations of the abstract Retr yPolicy class. The namespace and
some of the configuration details depend on which Service Bus client SDK package is used:
 PA C K A GE DESC RIP T IO N N A M ESPA C E

Microsoft.Azure.ServiceBus Azure Service Bus client library for .NET Microsoft.ServiceBus

Standard.

WindowsAzure.ServiceBus This package is the older Service Bus Microsoft.Azure.ServiceBus

client library. It requires .Net Framework
4.5.2.

Both versions of the client library provide the following built-in implementations of RetryPolicy :
RetryExponential. Implements exponential backoff.
NoRetry. Does not perform retries. Use this class when you don't need retries at the Service Bus API level,
for example when another process manages retries as part of a batch or multistep operation.
The RetryPolicy.Default property returns a default policy of type RetryExponential . This policy object has the
following settings:

SET T IN G DEFA ULT VA L UE M EA N IN G

MinimalBackoff 0 Minimum back-off interval. Added to

the retry interval computed from
deltaBackoff .

MaximumBackoff 30 seconds Maximum back-off interval.

DeltaBackoff 3 seconds Back-off interval between retries.

Multiples of this timespan are used for
subsequent retry attempts.

MaxRetryCount 5 The maximum number of retries.

(Default value is 10 in the
WindowsAzure.ServiceBus package.)

In addition, the following property is defined in the older WindowsAzure.ServiceBus package:

SET T IN G DEFA ULT VA L UE M EA N IN G

TerminationTimeBuffer 5 seconds Retry attempts will be abandoned if the

remaining time is less than this value.

Service Bus actions can return a range of exceptions, listed in Service Bus messaging exceptions. Exceptions
returned from Service Bus expose the IsTransient property that indicates whether the client should retry the
operation. The built-in Retr yExponential policy checks this property before retrying.
If the last exception encountered was Ser verBusyException , the Retr yExponential policy adds 10 seconds to
the computed retry interval. This value cannot be changed.
Custom implementations could use a combination of the exception type and the IsTransient property to provide
more fine-grained control over retry actions. For example, you could detect a QuotaExceededException and
take action to drain the queue before retrying sending a message to it.
The following code sets the retry policy on a Service Bus client using the Microsoft.Azure.ServiceBus library:
 const string QueueName = "queue1";
const string ServiceBusConnectionString = "<your_connection_string>";

var policy = new RetryExponential(

minimumBackoff: TimeSpan.FromSeconds(10),
maximumBackoff: TimeSpan.FromSeconds(30),
maximumRetryCount: 3);
var queueClient = new QueueClient(ServiceBusConnectionString, QueueName, ReceiveMode.PeekLock, policy);

The retry policy cannot be set at the individual operation level. It applies to all operations for the client.
Retry usage guidance
Consider the following guidelines when using Service Bus:
When using the built-in Retr yExponential implementation, do not implement a fallback operation as the
policy reacts to Server Busy exceptions and automatically switches to an appropriate retry mode.
Service Bus supports a feature called Paired Namespaces that implements automatic failover to a backup
queue in a separate namespace if the queue in the primary namespace fails. Messages from the secondary
queue can be sent back to the primary queue when it recovers. This feature helps to address transient failures.
For more information, see Asynchronous Messaging Patterns and High Availability.
Consider starting with the following settings for retrying operations. These settings are general purpose, and you
should monitor the operations and fine-tune the values to suit your own scenario.

EXA M P L E M A XIM UM
C O N T EXT L AT EN C Y RET RY P O L IC Y SET T IN GS H O W IT W O RK S

Interactive, UI, or 2 seconds* Exponential MinimumBackoff = 0 Attempt 1: Delay 0

foreground MaximumBackoff = sec.
30 sec. Attempt 2: Delay
DeltaBackoff = 300 ~300 msec.
msec. Attempt 3: Delay
TimeBuffer = 300 ~900 msec.
msec.
MaxRetryCount = 2

Background or batch 30 seconds Exponential MinimumBackoff = 1 Attempt 1: Delay ~1

MaximumBackoff = sec.
30 sec. Attempt 2: Delay ~3
DeltaBackoff = 1.75 sec.
sec. Attempt 3: Delay ~6
TimeBuffer = 5 sec. msec.
MaxRetryCount = 3 Attempt 4: Delay ~13
msec.

* Not including additional delay that is added if a Server Busy response is received.
Telemetry
Service Bus logs retries as ETW events using an EventSource . You must attach an EventListener to the event
source to capture the events and view them in Performance Viewer, or write them to a suitable destination log. The
retry events are of the following form:
 Microsoft-ServiceBus-Client/RetryPolicyIteration
ThreadID="14,500"
FormattedMessage="[TrackingId:] RetryExponential: Operation Get:https://retry-
tests.servicebus.windows.net/TestQueue/?api-version=2014-05 at iteration 0 is retrying after 00:00:00.1000000
sleep because of Microsoft.ServiceBus.Messaging.MessagingCommunicationException: The remote name could not be
resolved: 'retry-tests.servicebus.windows.net'.TrackingId:6a26f99c-dc6d-422e-8565-f89fdd0d4fe3,
TimeStamp:9/5/2014 10:00:13 PM."
trackingId=""
policyType="RetryExponential"
operation="Get:https://retry-tests.servicebus.windows.net/TestQueue/?api-version=2014-05"
iteration="0"
iterationSleep="00:00:00.1000000"
lastExceptionType="Microsoft.ServiceBus.Messaging.MessagingCommunicationException"
exceptionMessage="The remote name could not be resolved: 'retry-
tests.servicebus.windows.net'.TrackingId:6a26f99c-dc6d-422e-8565-f89fdd0d4fe3,TimeStamp:9/5/2014 10:00:13 PM"

Examples
The following code example shows how to set the retry policy for:
A namespace manager. The policy applies to all operations on that manager, and cannot be overridden for
individual operations.
A messaging factory. The policy applies to all clients created from that factory, and cannot be overridden when
creating individual clients.
An individual messaging client. After a client has been created, you can set the retry policy for that client. The
policy applies to all operations on that client.

using System;
using System.Threading.Tasks;
using Microsoft.ServiceBus;
using Microsoft.ServiceBus.Messaging;

namespace RetryCodeSamples
{
class ServiceBusCodeSamples
{
private const string connectionString =
@"Endpoint=sb://[my-namespace].servicebus.windows.net/;
SharedAccessKeyName=RootManageSharedAccessKey;
SharedAccessKey=C99..........Mk=";

public async static Task Samples()

{
const string QueueName = "TestQueue";

ServiceBusEnvironment.SystemConnectivity.Mode = ConnectivityMode.Http;

var namespaceManager = NamespaceManager.CreateFromConnectionString(connectionString);

// The namespace manager will have a default exponential policy with 10 retry attempts
// and a 3 second delay delta.
// Retry delays will be approximately 0 sec, 3 sec, 9 sec, 25 sec and the fixed 30 sec,
// with an extra 10 sec added when receiving a ServiceBusyException.

{
// Set different values for the retry policy, used for all operations on the namespace
manager.
namespaceManager.Settings.RetryPolicy =
new RetryExponential(
minBackoff: TimeSpan.FromSeconds(0),
maxBackoff: TimeSpan.FromSeconds(30),
maxRetryCount: 3);

// Policies cannot be specified on a per-operation basis.

if (!await namespaceManager.QueueExistsAsync(QueueName))
 if (!await namespaceManager.QueueExistsAsync(QueueName))
{
await namespaceManager.CreateQueueAsync(QueueName);
}
}

var messagingFactory = MessagingFactory.Create(

namespaceManager.Address, namespaceManager.Settings.TokenProvider);
// The messaging factory will have a default exponential policy with 10 retry attempts
// and a 3 second delay delta.
// Retry delays will be approximately 0 sec, 3 sec, 9 sec, 25 sec and the fixed 30 sec,
// with an extra 10 sec added when receiving a ServiceBusyException.

{
// Set different values for the retry policy, used for clients created from it.
messagingFactory.RetryPolicy =
new RetryExponential(
minBackoff: TimeSpan.FromSeconds(1),
maxBackoff: TimeSpan.FromSeconds(30),
maxRetryCount: 3);

// Policies cannot be specified on a per-operation basis.

var session = await messagingFactory.AcceptMessageSessionAsync();
}

{
var client = messagingFactory.CreateQueueClient(QueueName);
// The client inherits the policy from the factory that created it.

// Set different values for the retry policy on the client.

client.RetryPolicy =
new RetryExponential(
minBackoff: TimeSpan.FromSeconds(0.1),
maxBackoff: TimeSpan.FromSeconds(30),
maxRetryCount: 3);

// Policies cannot be specified on a per-operation basis.

var session = await client.AcceptMessageSessionAsync();
}
}
}
}

More information
Asynchronous Messaging Patterns and High Availability

Service Fabric
Distributing reliable services in a Service Fabric cluster guards against most of the potential transient faults
discussed in this article. Some transient faults are still possible, however. For example, the naming service might be
in the middle of a routing change when it gets a request, causing it to throw an exception. If the same request
comes 100 milliseconds later, it will probably succeed.
Internally, Service Fabric manages this kind of transient fault. You can configure some settings by using the
OperationRetrySettings class while setting up your services. The following code shows an example. In most cases,
this should not be necessary, and the default settings will be fine.
 FabricTransportRemotingSettings transportSettings = new FabricTransportRemotingSettings
{
OperationTimeout = TimeSpan.FromSeconds(30)
};

var retrySettings = new OperationRetrySettings(TimeSpan.FromSeconds(15), TimeSpan.FromSeconds(1), 5);

var clientFactory = new FabricTransportServiceRemotingClientFactory(transportSettings);

var serviceProxyFactory = new ServiceProxyFactory((c) => clientFactory, retrySettings);

var client = serviceProxyFactory.CreateServiceProxy<ISomeService>(

new Uri("fabric:/SomeApp/SomeStatefulReliableService"),
new ServicePartitionKey(0));

More information
Remote exception handling

SQL Database using ADO.NET

SQL Database is a hosted SQL database available in a range of sizes and as both a standard (shared) and premium
(non-shared) service.
Retry mechanism
SQL Database has no built-in support for retries when accessed using ADO.NET. However, the return codes from
requests can be used to determine why a request failed. For more information about SQL Database throttling, see
Azure SQL Database resource limits. For a list of relevant error codes, see SQL error codes for SQL Database client
applications.
You can use the Polly library to implement retries for SQL Database. See Transient fault handling with Polly.
Retry usage guidance
Consider the following guidelines when accessing SQL Database using ADO.NET:
Choose the appropriate service option (shared or premium). A shared instance may suffer longer than usual
connection delays and throttling due to the usage by other tenants of the shared server. If more predictable
performance and reliable low latency operations are required, consider choosing the premium option.
Ensure that you perform retries at the appropriate level or scope to avoid non-idempotent operations causing
inconsistency in the data. Ideally, all operations should be idempotent so that they can be repeated without
causing inconsistency. Where this is not the case, the retry should be performed at a level or scope that allows
all related changes to be undone if one operation fails; for example, from within a transactional scope. For more
information, see Cloud Service Fundamentals Data Access Layer – Transient Fault Handling.
A fixed interval strategy is not recommended for use with Azure SQL Database except for interactive scenarios
where there are only a few retries at very short intervals. Instead, consider using an exponential back-off
strategy for the majority of scenarios.
Choose a suitable value for the connection and command timeouts when defining connections. Too short a
timeout may result in premature failures of connections when the database is busy. Too long a timeout may
prevent the retry logic working correctly by waiting too long before detecting a failed connection. The value of
the timeout is a component of the end-to-end latency; it is effectively added to the retry delay specified in the
retry policy for every retry attempt.
Close the connection after a certain number of retries, even when using an exponential back off retry logic, and
retry the operation on a new connection. Retrying the same operation multiple times on the same connection
can be a factor that contributes to connection problems. For an example of this technique, see Cloud Service
Fundamentals Data Access Layer – Transient Fault Handling.
When connection pooling is in use (the default) there is a chance that the same connection will be chosen from
 the pool, even after closing and reopening a connection. If this is the case, a technique to resolve it is to call the
ClearPool method of the SqlConnection class to mark the connection as not reusable. However, you should
do this only after several connection attempts have failed, and only when encountering the specific class of
transient failures such as SQL timeouts (error code -2) related to faulty connections.
If the data access code uses transactions initiated as TransactionScope instances, the retry logic should
reopen the connection and initiate a new transaction scope. For this reason, the retryable code block should
encompass the entire scope of the transaction.
Consider starting with the following settings for retrying operations. These settings are general purpose, and you
should monitor the operations and fine-tune the values to suit your own scenario.

SA M P L E TA RGET
E2E
C O N T EXT M A X L AT EN C Y RET RY ST RAT EGY SET T IN GS VA L UES H O W IT W O RK S

Interactive, UI, 2 sec FixedInterval Retry count 3 Attempt 1 -

or foreground Retry interval 500 ms delay 0 sec
First fast retry true Attempt 2 -
delay 500 ms
Attempt 3 -
delay 500 ms

Background 30 sec ExponentialBacko Retry count 5 Attempt 1 -

or batch ff Min back-off 0 sec delay 0 sec
Max back-off 60 sec Attempt 2 -
Delta back-off 2 sec delay ~2 sec
First fast retry false Attempt 3 -
delay ~6 sec
Attempt 4 -
delay ~14 sec
Attempt 5 -
delay ~30 sec

NOTE
The end-to-end latency targets assume the default timeout for connections to the service. If you specify longer connection
timeouts, the end-to-end latency will be extended by this additional time for every retry attempt.

Examples
This section shows how you can use Polly to access Azure SQL Database using a set of retry policies configured in
the Policy class.
The following code shows an extension method on the SqlCommand class that calls ExecuteAsync with exponential
backoff.
 public async static Task<SqlDataReader> ExecuteReaderWithRetryAsync(this SqlCommand command)
{
GuardConnectionIsNotNull(command);

var policy = Policy.Handle<Exception>().WaitAndRetryAsync(

retryCount: 3, // Retry 3 times
sleepDurationProvider: attempt => TimeSpan.FromMilliseconds(200 * Math.Pow(2, attempt - 1)), //
Exponential backoff based on an initial 200 ms delay.
onRetry: (exception, attempt) =>
{
// Capture some information for logging/telemetry.
logger.LogWarn($"ExecuteReaderWithRetryAsync: Retry {attempt} due to {exception}.");
});

// Retry the following call according to the policy.

await policy.ExecuteAsync<SqlDataReader>(async token =>
{
// This code is executed within the Policy

if (conn.State != System.Data.ConnectionState.Open) await conn.OpenAsync(token);

return await command.ExecuteReaderAsync(System.Data.CommandBehavior.Default, token);

}, cancellationToken);
}

This asynchronous extension method can be used as follows.

var sqlCommand = sqlConnection.CreateCommand();

sqlCommand.CommandText = "[some query]";

using (var reader = await sqlCommand.ExecuteReaderWithRetryAsync())

{
// Do something with the values
}

More information
Cloud Service Fundamentals Data Access Layer – Transient Fault Handling
For general guidance on getting the most from SQL Database, see Azure SQL Database performance and elasticity
guide.

SQL Database using Entity Framework 6

SQL Database is a hosted SQL database available in a range of sizes and as both a standard (shared) and premium
(non-shared) service. Entity Framework is an object-relational mapper that enables .NET developers to work with
relational data using domain-specific objects. It eliminates the need for most of the data-access code that
developers usually need to write.
Retry mechanism
Retry support is provided when accessing SQL Database using Entity Framework 6.0 and higher through a
mechanism called Connection resiliency / retry logic. The main features of the retry mechanism are:
The primary abstraction is the IDbExecutionStrategy interface. This interface:
Defines synchronous and asynchronous Execute methods.
Defines classes that can be used directly or can be configured on a database context as a default strategy,
mapped to provider name, or mapped to a provider name and server name. When configured on a
context, retries occur at the level of individual database operations, of which there might be several for a
given context operation.
 Defines when to retry a failed connection, and how.
It includes several built-in implementations of the IDbExecutionStrategy interface:
Default: no retrying.
Default for SQL Database (automatic): no retrying, but inspects exceptions and wraps them with
suggestion to use the SQL Database strategy.
Default for SQL Database: exponential (inherited from base class) plus SQL Database detection logic.
It implements an exponential back-off strategy that includes randomization.
The built-in retry classes are stateful and are not thread-safe. However, they can be reused after the current
operation is completed.
If the specified retry count is exceeded, the results are wrapped in a new exception. It does not bubble up the
current exception.
Policy configuration
Retry support is provided when accessing SQL Database using Entity Framework 6.0 and higher. Retry policies are
configured programmatically. The configuration cannot be changed on a per-operation basis.
When configuring a strategy on the context as the default, you specify a function that creates a new strategy on
demand. The following code shows how you can create a retry configuration class that extends the
DbConfiguration base class.

public class BloggingContextConfiguration : DbConfiguration

{
public BlogConfiguration()
{
// Set up the execution strategy for SQL Database (exponential) with 5 retries and 4 sec delay
this.SetExecutionStrategy(
"System.Data.SqlClient", () => new SqlAzureExecutionStrategy(5, TimeSpan.FromSeconds(4)));
}
}

You can then specify this as the default retry strategy for all operations using the SetConfiguration method of
the DbConfiguration instance when the application starts. By default, EF will automatically discover and use the
configuration class.

DbConfiguration.SetConfiguration(new BloggingContextConfiguration());

You can specify the retry configuration class for a context by annotating the context class with a
DbConfigurationType attribute. However, if you have only one configuration class, EF will use it without the need
to annotate the context.

[DbConfigurationType(typeof(BloggingContextConfiguration))]
public class BloggingContext : DbContext

If you need to use different retry strategies for specific operations, or disable retries for specific operations, you
can create a configuration class that allows you to suspend or swap strategies by setting a flag in the CallContext .
The configuration class can use this flag to switch strategies, or disable the strategy you provide and use a default
strategy. For more information, see Suspend Execution Strategy (EF6 onwards).
Another technique for using specific retry strategies for individual operations is to create an instance of the
required strategy class and supply the desired settings through parameters. You then invoke its ExecuteAsync
method.
 var executionStrategy = new SqlAzureExecutionStrategy(5, TimeSpan.FromSeconds(4));
var blogs = await executionStrategy.ExecuteAsync(
async () =>
{
using (var db = new BloggingContext("Blogs"))
{
// Acquire some values asynchronously and return them
}
},
new CancellationToken()
);

The simplest way to use a DbConfiguration class is to locate it in the same assembly as the DbContext class.
However, this is not appropriate when the same context is required in different scenarios, such as different
interactive and background retry strategies. If the different contexts execute in separate AppDomains, you can use
the built-in support for specifying configuration classes in the configuration file or set it explicitly using code. If the
different contexts must execute in the same AppDomain, a custom solution will be required.
For more information, see Code-Based Configuration (EF6 onwards).
The following table shows the default settings for the built-in retry policy when using EF6.

SET T IN G DEFA ULT VA L UE M EA N IN G

Policy Exponential Exponential back-off.

MaxRetryCount 5 The maximum number of retries.

MaxDelay 30 seconds The maximum delay between retries.

This value does not affect how the
series of delays are computed. It only
defines an upper bound.

DefaultCoefficient 1 second The coefficient for the exponential back-

off computation. This value cannot be
changed.

DefaultRandomFactor 1.1 The multiplier used to add a random

delay for each entry. This value cannot
be changed.

DefaultExponentialBase 2 The multiplier used to calculate the next

delay. This value cannot be changed.

Retry usage guidance

Consider the following guidelines when accessing SQL Database using EF6:
Choose the appropriate service option (shared or premium). A shared instance may suffer longer than
usual connection delays and throttling due to the usage by other tenants of the shared server. If predictable
performance and reliable low latency operations are required, consider choosing the premium option.
A fixed interval strategy is not recommended for use with Azure SQL Database. Instead, use an exponential
back-off strategy because the service may be overloaded, and longer delays allow more time for it to
recover.
Choose a suitable value for the connection and command timeouts when defining connections. Base the
timeout on both your business logic design and through testing. You may need to modify this value over
 time as the volumes of data or the business processes change. Too short a timeout may result in premature
failures of connections when the database is busy. Too long a timeout may prevent the retry logic working
correctly by waiting too long before detecting a failed connection. The value of the timeout is a component
of the end-to-end latency, although you cannot easily determine how many commands will execute when
saving the context. You can change the default timeout by setting the CommandTimeout property of the
DbContext instance.
Entity Framework supports retry configurations defined in configuration files. However, for maximum
flexibility on Azure you should consider creating the configuration programmatically within the application.
The specific parameters for the retry policies, such as the number of retries and the retry intervals, can be
stored in the service configuration file and used at runtime to create the appropriate policies. This allows
the settings to be changed without requiring the application to be restarted.
Consider starting with the following settings for retrying operations. You cannot specify the delay between retry
attempts (it is fixed and generated as an exponential sequence). You can specify only the maximum values, as
shown here; unless you create a custom retry strategy. These settings are general purpose, and you should
monitor the operations and fine-tune the values to suit your own scenario.

SA M P L E TA RGET
E2E
C O N T EXT M A X L AT EN C Y RET RY P O L IC Y SET T IN GS VA L UES H O W IT W O RK S

Interactive, UI, 2 seconds Exponential MaxRetryCount 3 Attempt 1 -

or foreground MaxDelay 750 ms delay 0 sec
Attempt 2 -
delay 750 ms
Attempt 3 –
delay 750 ms

Background 30 seconds Exponential MaxRetryCount 5 Attempt 1 -

or batch MaxDelay 12 seconds delay 0 sec
Attempt 2 -
delay ~1 sec
Attempt 3 -
delay ~3 sec
Attempt 4 -
delay ~7 sec
Attempt 5 -
delay 12 sec

NOTE
The end-to-end latency targets assume the default timeout for connections to the service. If you specify longer connection
timeouts, the end-to-end latency will be extended by this additional time for every retry attempt.

Examples
The following code example defines a simple data access solution that uses Entity Framework. It sets a specific
retry strategy by defining an instance of a class named BlogConfiguration that extends DbConfiguration .
 using System;
using System.Collections.Generic;
using System.Data.Entity;
using System.Data.Entity.SqlServer;
using System.Threading.Tasks;

namespace RetryCodeSamples
{
public class BlogConfiguration : DbConfiguration
{
public BlogConfiguration()
{
// Set up the execution strategy for SQL Database (exponential) with 5 retries and 12 sec delay.
// These values could be loaded from configuration rather than being hard-coded.
this.SetExecutionStrategy(
"System.Data.SqlClient", () => new SqlAzureExecutionStrategy(5,
TimeSpan.FromSeconds(12)));
}
}

// Specify the configuration type if more than one has been defined.
// [DbConfigurationType(typeof(BlogConfiguration))]
public class BloggingContext : DbContext
{
// Definition of content goes here.
}

class EF6CodeSamples
{
public async static Task Samples()
{
// Execution strategy configured by DbConfiguration subclass, discovered automatically or
// or explicitly indicated through configuration or with an attribute. Default is no retries.
using (var db = new BloggingContext("Blogs"))
{
// Add, edit, delete blog items here, then:
await db.SaveChangesAsync();
}
}
}
}

More examples of using the Entity Framework retry mechanism can be found in Connection resiliency / retry logic.
More information
Azure SQL Database performance and elasticity guide

SQL Database using Entity Framework Core

Entity Framework Core is an object-relational mapper that enables .NET Core developers to work with data using
domain-specific objects. It eliminates the need for most of the data-access code that developers usually need to
write. This version of Entity Framework was written from the ground up, and doesn't automatically inherit all the
features from EF6.x.
Retry mechanism
Retry support is provided when accessing SQL Database using Entity Framework Core through a mechanism
called connection resiliency. Connection resiliency was introduced in EF Core 1.1.0.
The primary abstraction is the IExecutionStrategy interface. The execution strategy for SQL Server, including SQL
Azure, is aware of the exception types that can be retried and has sensible defaults for maximum retries, delay
between retries, and so on.
Examples
The following code enables automatic retries when configuring the DbContext object, which represents a session
with the database.

protected override void OnConfiguring(DbContextOptionsBuilder optionsBuilder)

{
optionsBuilder
.UseSqlServer(
@"Server=
(localdb)\mssqllocaldb;Database=EFMiscellaneous.ConnectionResiliency;Trusted_Connection=True;",
options => options.EnableRetryOnFailure());
}

The following code shows how to execute a transaction with automatic retries, by using an execution strategy. The
transaction is defined in a delegate. If a transient failure occurs, the execution strategy will invoke the delegate
again.

using (var db = new BloggingContext())

{
var strategy = db.Database.CreateExecutionStrategy();

strategy.Execute(() =>
{
using (var transaction = db.Database.BeginTransaction())
{
db.Blogs.Add(new Blog { Url = "https://blogs.msdn.com/dotnet" });
db.SaveChanges();

db.Blogs.Add(new Blog { Url = "https://blogs.msdn.com/visualstudio" });

db.SaveChanges();

transaction.Commit();
}
});
}

Azure Storage
Azure Storage services include blob storage, files, and storage queues.
Blobs, Queues and Files
The ClientOptions Class is the base type for all client option types and exposes various common client options like
Diagnostics, Retry, Transport. To provide the client configuration options for connecting to Azure Queue, Blob, and
File Storage you must use the corresponding derived type. In the next example, you use the QueueClientOptions
class (derived from ClientOptions) to configure a client to connect to Azure Queue Service. The Retry property is
the set of options that can be specified to influence how retry attempts are made, and how a failure is eligible to be
retried.
 using System;
using System.Threading;
using Azure.Core;
using Azure.Identity;
using Azure.Storage;
using Azure.Storage.Queues;
using Azure.Storage.Queues.Models;

namespace RetryCodeSamples
{
class AzureStorageCodeSamples {

public async static Task Samples() {

// Provide the client configuration options for connecting to Azure Queue Storage
QueueClientOptions queueClientOptions = new QueueClientOptions()
{
Retry = {
Delay = TimeSpan.FromSeconds(2), //The delay between retry attempts for a fixed
approach or the delay on which to base
//calculations for a backoff-based approach
MaxRetries = 5, //The maximum number of retry attempts before giving
up
Mode = RetryMode.Exponential, //The approach to use for calculating retry delays
MaxDelay = TimeSpan.FromSeconds(10) //The maximum permissible delay between retry
attempts
},

GeoRedundantSecondaryUri = new Uri("https://...")

// If the GeoRedundantSecondaryUri property is set, the secondary Uri will be used for GET
or HEAD requests during retries.
// If the status of the response from the secondary Uri is a 404, then subsequent retries
for the request will not use the
// secondary Uri again, as this indicates that the resource may not have propagated there
yet.
// Otherwise, subsequent retries will alternate back and forth between primary and
secondary Uri.
};

Uri queueServiceUri = new Uri("https://storageaccount.queue.core.windows.net/");

string accountName = "Storage account name";
string accountKey = "storage account key";

// Create a client object for the Queue service, including QueueClientOptions.

QueueServiceClient serviceClient = new QueueServiceClient(queueServiceUri, new
DefaultAzureCredential(), queueClientOptions);

CancellationTokenSource source = new CancellationTokenSource();

CancellationToken cancellationToken = source.Token;

// Return an async collection of queues in the storage account.

var queues = serviceClient.GetQueuesAsync(QueueTraits.None, null, cancellationToken);

Table Support

NOTE
WindowsAzure.Storage Nuget Package has been deprecated. For Azure table support, see Microsoft.Azure.Cosmos.Table
Nuget Package

Retry mechanism
Retries occur at the individual REST operation level and are an integral part of the client API implementation. The
client storage SDK uses classes that implement the IExtendedRetryPolicy Interface.
The built-in classes provide support for linear (constant delay) and exponential with randomization retry intervals.
There is also a no retry policy for use when another process is handling retries at a higher level. However, you can
implement your own retry classes if you have specific requirements not provided by the built-in classes.
Alternate retries switch between primary and secondary storage service location if you are using read access geo-
redundant storage (RA-GRS) and the result of the request is a retryable error. See Azure Storage Redundancy
Options for more information.
Policy configuration
Retry policies are configured programmatically. A typical procedure is to create and populate a
TableRequestOptions , BlobRequestOptions , FileRequestOptions , or QueueRequestOptions instance.

TableRequestOptions interactiveRequestOption = new TableRequestOptions()

{
RetryPolicy = new LinearRetry(TimeSpan.FromMilliseconds(500), 3),
// For Read-access geo-redundant storage, use PrimaryThenSecondary.
// Otherwise set this to PrimaryOnly.
LocationMode = LocationMode.PrimaryThenSecondary,
// Maximum execution time based on the business use case.
MaximumExecutionTime = TimeSpan.FromSeconds(2)
};

The request options instance can then be set on the client, and all operations with the client will use the specified
request options.

client.DefaultRequestOptions = interactiveRequestOption;
var stats = await client.GetServiceStatsAsync();

You can override the client request options by passing a populated instance of the request options class as a
parameter to operation methods.

var stats = await client.GetServiceStatsAsync(interactiveRequestOption, operationContext: null);

You use an OperationContext instance to specify the code to execute when a retry occurs and when an operation
has completed. This code can collect information about the operation for use in logs and telemetry.

// Set up notifications for an operation

var context = new OperationContext();
context.ClientRequestID = "some request id";
context.Retrying += (sender, args) =>
{
// Collect retry information
};
context.RequestCompleted += (sender, args) =>
{
// Collect operation completion information
};
var stats = await client.GetServiceStatsAsync(null, context);

In addition to indicating whether a failure is suitable for retry, the extended retry policies return a Retr yContext
object that indicates the number of retries, the results of the last request, whether the next retry will happen in the
primary or secondary location (see table below for details). The properties of the Retr yContext object can be
used to decide if and when to attempt a retry. For more information, see IExtendedRetryPolicy.Evaluate Method.
The following tables show the default settings for the built-in retry policies.
Request options:
 SET T IN G DEFA ULT VA L UE M EA N IN G

MaximumExecutionTime None Maximum execution time for the

request, including all potential retry
attempts. If it is not specified, then the
amount of time that a request is
permitted to take is unlimited. In other
words, the request might stop
responding.

ServerTimeout None Server timeout interval for the request

(value is rounded to seconds). If not
specified, it will use the default value for
all requests to the server. Usually, the
best option is to omit this setting so
that the server default is used.

LocationMode None If the storage account is created with

the Read access geo-redundant storage
(RA-GRS) replication option, you can
use the location mode to indicate which
location should receive the request. For
example, if Primar yThenSecondar y is
specified, requests are always sent to
the primary location first. If a request
fails, it is sent to the secondary location.

RetryPolicy ExponentialPolicy See below for details of each option.

Exponential policy:

SET T IN G DEFA ULT VA L UE M EA N IN G

maxAttempt 3 Number of retry attempts.

deltaBackoff 4 seconds Back-off interval between retries.

Multiples of this timespan, including a
random element, will be used for
subsequent retry attempts.

MinBackoff 3 seconds Added to all retry intervals computed

from deltaBackoff. This value cannot be
changed.

MaxBackoff 120 seconds MaxBackoff is used if the computed

retry interval is greater than
MaxBackoff. This value cannot be
changed.

Linear policy:

SET T IN G DEFA ULT VA L UE M EA N IN G

maxAttempt 3 Number of retry attempts.

deltaBackoff 30 seconds Back-off interval between retries.

Retry usage guidance

Consider the following guidelines when accessing Azure storage services using the storage client API:
Use the built-in retry policies from the Microsoft.Azure.Storage.RetryPolicies namespace where they are
appropriate for your requirements. In most cases, these policies will be sufficient.
Use the ExponentialRetr y policy in batch operations, background tasks, or non-interactive scenarios. In
these scenarios, you can typically allow more time for the service to recover—with a consequently
increased chance of the operation eventually succeeding.
Consider specifying the MaximumExecutionTime property of the RequestOptions parameter to limit
the total execution time, but take into account the type and size of the operation when choosing a timeout
value.
If you need to implement a custom retry, avoid creating wrappers around the storage client classes. Instead,
use the capabilities to extend the existing policies through the IExtendedRetr yPolicy interface.
If you are using read access geo-redundant storage (RA-GRS) you can use the LocationMode to specify
that retry attempts will access the secondary read-only copy of the store should the primary access fail.
However, when using this option you must ensure that your application can work successfully with data that
may be stale if the replication from the primary store has not yet completed.
Consider starting with the following settings for retrying operations. These settings are general purpose, and you
should monitor the operations and fine-tune the values to suit your own scenario.

SA M P L E TA RGET
E2E
C O N T EXT M A X L AT EN C Y RET RY P O L IC Y SET T IN GS VA L UES H O W IT W O RK S

Interactive, UI, 2 seconds Linear maxAttempt 3 Attempt 1 -

or foreground deltaBackoff 500 ms delay 500 ms
Attempt 2 -
delay 500 ms
Attempt 3 -
delay 500 ms

Background 30 seconds Exponential maxAttempt 5 Attempt 1 -

or batch deltaBackoff 4 seconds delay ~3 sec
Attempt 2 -
delay ~7 sec
Attempt 3 -
delay ~15 sec

Telemetry
Retry attempts are logged to a TraceSource . You must configure a TraceListener to capture the events and write
them to a suitable destination log. You can use the TextWriterTraceListener or XmlWriterTraceListener to
write the data to a log file, the EventLogTraceListener to write to the Windows Event Log, or the
EventProviderTraceListener to write trace data to the ETW subsystem. You can also configure autoflushing of
the buffer, and the verbosity of events that will be logged (for example, Error, Warning, Informational, and Verbose).
For more information, see Client-side Logging with the .NET Storage Client Library.
Operations can receive an OperationContext instance, which exposes a Retr ying event that can be used to
attach custom telemetry logic. For more information, see OperationContext.Retrying Event.
Examples
The following code example shows how to create two TableRequestOptions instances with different retry
settings; one for interactive requests and one for background requests. The example then sets these two retry
policies on the client so that they apply for all requests, and also sets the interactive strategy on a specific request
so that it overrides the default settings applied to the client.
using System;
using System.Threading.Tasks;
using Microsoft.Azure.Cosmos.Table;

namespace RetryCodeSamples
{
class AzureStorageCodeSamples
{
private const string connectionString = "UseDevelopmentStorage=true";

public async static Task Samples()

{
var storageAccount = CloudStorageAccount.Parse(connectionString);

TableRequestOptions interactiveRequestOption = new TableRequestOptions()

{
RetryPolicy = new LinearRetry(TimeSpan.FromMilliseconds(500), 3),
// For Read-access geo-redundant storage, use PrimaryThenSecondary.
// Otherwise set this to PrimaryOnly.
LocationMode = LocationMode.PrimaryThenSecondary,
// Maximum execution time based on the business use case.
MaximumExecutionTime = TimeSpan.FromSeconds(2)
};

TableRequestOptions backgroundRequestOption = new TableRequestOptions()

{
// Client has a default exponential retry policy with 4 sec delay and 3 retry attempts
// Retry delays will be approximately 3 sec, 7 sec, and 15 sec
MaximumExecutionTime = TimeSpan.FromSeconds(30),
// PrimaryThenSecondary in case of Read-access geo-redundant storage, else set this to
PrimaryOnly
LocationMode = LocationMode.PrimaryThenSecondary
};

var client = storageAccount.CreateCloudTableClient();

// Client has a default exponential retry policy with 4 sec delay and 3 retry attempts
// Retry delays will be approximately 3 sec, 7 sec, and 15 sec
// ServerTimeout and MaximumExecutionTime are not set

{
// Set properties for the client (used on all requests unless overridden)
// Different exponential policy parameters for background scenarios
client.DefaultRequestOptions = backgroundRequestOption;
// Linear policy for interactive scenarios
client.DefaultRequestOptions = interactiveRequestOption;
}

{
// set properties for a specific request
var stats = await client.GetServiceStatsAsync(interactiveRequestOption, operationContext:
null);
}

{
// Set up notifications for an operation
var context = new OperationContext();
context.ClientRequestID = "some request id";
context.Retrying += (sender, args) =>
{
// Collect retry information
};
context.RequestCompleted += (sender, args) =>
{
// Collect operation completion information
};
var stats = await client.GetServiceStatsAsync(null, context);
}
}
 }
}
}

More information
Azure Storage client Library retry policy recommendations
Storage Client Library 2.0 – Implementing retry policies

General REST and retry guidelines

Consider the following when accessing Azure or third-party services:
Use a systematic approach to managing retries, perhaps as reusable code, so that you can apply a
consistent methodology across all clients and all solutions.
Consider using a retry framework such as Polly to manage retries if the target service or client has no built-
in retry mechanism. This will help you implement a consistent retry behavior, and it may provide a suitable
default retry strategy for the target service. However, you may need to create custom retry code for services
that have nonstandard behavior, that do not rely on exceptions to indicate transient failures, or if you want
to use a Retr y-Response reply to manage retry behavior.
The transient detection logic will depend on the actual client API you use to invoke the REST calls. Some
clients, such as the newer HttpClient class, will not throw exceptions for completed requests with a non-
success HTTP status code.
The HTTP status code returned from the service can help to indicate whether the failure is transient. You
may need to examine the exceptions generated by a client or the retry framework to access the status code
or to determine the equivalent exception type. The following HTTP codes typically indicate that a retry is
appropriate:
408 Request Timeout
429 Too Many Requests
500 Internal Server Error
502 Bad Gateway
503 Service Unavailable
504 Gateway Timeout
If you base your retry logic on exceptions, the following typically indicate a transient failure where no
connection could be established:
WebExceptionStatus.ConnectionClosed
WebExceptionStatus.ConnectFailure
WebExceptionStatus.Timeout
WebExceptionStatus.RequestCanceled
In the case of a service unavailable status, the service might indicate the appropriate delay before retrying
in the Retr y-After response header or a different custom header. Services might also send additional
information as custom headers, or embedded in the content of the response.
Do not retry for status codes representing client errors (errors in the 4xx range) except for a 408 Request
Timeout and 429 Too Many Requests.
Thoroughly test your retry strategies and mechanisms under a range of conditions, such as different
network states and varying system loadings.
Retry strategies
The following are the typical types of retry strategy intervals:
 Exponential . A retry policy that performs a specified number of retries, using a randomized exponential
back off approach to determine the interval between retries. For example:

var random = new Random();

var delta = (int)((Math.Pow(2.0, currentRetryCount) - 1.0) *

random.Next((int)(this.deltaBackoff.TotalMilliseconds * 0.8),
(int)(this.deltaBackoff.TotalMilliseconds * 1.2)));
var interval = (int)Math.Min(checked(this.minBackoff.TotalMilliseconds + delta),
this.maxBackoff.TotalMilliseconds);
retryInterval = TimeSpan.FromMilliseconds(interval);

Incremental . A retry strategy with a specified number of retry attempts and an incremental time interval
between retries. For example:

retryInterval = TimeSpan.FromMilliseconds(this.initialInterval.TotalMilliseconds +
(this.increment.TotalMilliseconds * currentRetryCount));

LinearRetr y . A retry policy that performs a specified number of retries, using a specified fixed time interval
between retries. For example:

retryInterval = this.deltaBackoff;

Transient fault handling with Polly

Polly is a library to programmatically handle retries and circuit breaker strategies. The Polly project is a member of
the .NET Foundation. For services where the client does not natively support retries, Polly is a valid alternative and
avoids the need to write custom retry code, which can be hard to implement correctly. Polly also provides a way to
trace errors when they occur, so that you can log retries.
More information
connection resiliency
Data Points - EF Core 1.1
 Transient fault handling
12/18/2020 • 19 minutes to read • Edit Online

All applications that communicate with remote services and resources must be sensitive to transient faults. This is
especially the case for applications that run in the cloud, where the nature of the environment and connectivity
over the Internet means these types of faults are likely to be encountered more often. Transient faults include the
momentary loss of network connectivity to components and services, the temporary unavailability of a service, or
timeouts that arise when a service is busy. These faults are often self-correcting, and if the action is repeated after a
suitable delay it is likely to succeed.
This document covers general guidance for transient fault handling. For information about handling transient
faults when using Microsoft Azure services, see Azure service-specific retry guidelines.

Why do transient faults occur in the cloud?

Transient faults can occur in any environment, on any platform or operating system, and in any kind of application.
In solutions that run on local on-premises infrastructure, the performance and availability of the application and its
components is typically maintained through expensive and often underused hardware redundancy, and
components and resources are located close to each other. While this approach makes a failure less likely, it can
still result in transient faults - and even an outage through unforeseen events such as external power supply or
network issues, or other disaster scenarios.
Cloud hosting, including private cloud systems, can offer higher overall availability by using shared resources,
redundancy, automatic failover, and dynamic resource allocation across many commodity compute nodes.
However, the nature of these environments can mean that transient faults are more likely to occur. There are
several reasons for this:
Many resources in a cloud environment are shared, and access to these resources is subject to throttling in
order to protect the resource. Some services will refuse connections when the load rises to a specific level,
or a maximum throughput rate is reached, in order to allow processing of existing requests and to maintain
performance of the service for all users. Throttling helps to maintain the quality of service for neighbors
and other tenants using the shared resource.
Cloud environments are built using vast numbers of commodity hardware units. They deliver performance
by dynamically distributing the load across multiple computing units and infrastructure components, and
deliver reliability by automatically recycling or replacing failed units. This dynamic nature means that
transient faults and temporary connection failures may occasionally occur.
There are often more hardware components, including network infrastructure such as routers and load
balancers, between the application and the resources and services it uses. This additional infrastructure can
occasionally introduce additional connection latency and transient connection faults.
Network conditions between the client and the server may be variable, especially when communication
crosses the Internet. Even in on-premises locations, heavy traffic loads may slow communication and cause
intermittent connection failures.

Challenges
Transient faults can have a huge effect on the perceived availability of an application, even if it has been thoroughly
tested under all foreseeable circumstances. To ensure that cloud-hosted applications operate reliably, they must be
able to respond to the following challenges:
 The application must be able to detect faults when they occur, and determine if these faults are likely to be
transient, more long-lasting, or are terminal failures. Different resources are likely to return different
responses when a fault occurs, and these responses may also vary depending on the context of the
operation; for example, the response for an error when reading from storage may be different from
response for an error when writing to storage. Many resources and services have well-documented
transient failure contracts. However, where such information is not available, it may be difficult to discover
the nature of the fault and whether it is likely to be transient.
The application must be able to retry the operation if it determines that the fault is likely to be transient and
keep track of the number of times the operation was retried.
The application must use an appropriate strategy for the retries. This strategy specifies the number of times
it should retry, the delay between each attempt, and the actions to take after a failed attempt. The
appropriate number of attempts and the delay between each one are often difficult to determine, and vary
based on the type of resource as well as the current operating conditions of the resource and the
application itself.

General guidelines
The following guidelines will help you to design a suitable transient fault handling mechanism for your
applications:
Determine if there is a built-in retr y mechanism:
Many services provide an SDK or client library that contains a transient fault handling mechanism.
The retry policy it uses is typically tailored to the nature and requirements of the target service.
Alternatively, REST interfaces for services may return information that is useful in determining
whether a retry is appropriate, and how long to wait before the next retry attempt.
Use the built-in retry mechanism where available, unless you have specific and well-understood
requirements that make a different retry behavior more appropriate.
Determine if the operation is suitable for retr ying :
You should only retry operations where the faults are transient (typically indicated by the nature of
the error), and if there is at least some likelihood that the operation will succeed when reattempted.
There is no point in reattempting operations that indicate an invalid operation such as a database
update to an item that does not exist, or requests to a service or resource that has suffered a fatal
error.
In general, you should implement retries only where the full impact of this can be determined, and
the conditions are well understood and can be validated. If not, leave it to the calling code to
implement retries. Remember that the errors returned from resources and services outside your
control may evolve over time, and you may need to revisit your transient fault detection logic.
When you create services or components, consider implementing error codes and messages that will
help clients determine whether they should retry failed operations. In particular, indicate if the client
should retry the operation (perhaps by returning an isTransient value) and suggest a suitable delay
before the next retry attempt. If you build a web service, consider returning custom errors defined
within your service contracts. Even though generic clients may not be able to read these, they will be
useful when building custom clients.
Determine an appropriate retr y count and inter val:
It is vital to optimize the retry count and the interval to the type of use case. If you do not retry a
sufficient number of times, the application will be unable to complete the operation and is likely to
experience a failure. If you retry too many times, or with too short an interval between tries, the
application can potentially hold resources such as threads, connections, and memory for long
periods, which will adversely affect the health of the application.
The appropriate values for the time interval and the number of retry attempts depend on the type of
operation being attempted. For example, if the operation is part of a user interaction, the interval
should be short and only a few retries attempted to avoid making users wait for a response (which
holds open connections and can reduce availability for other users). If the operation is part of a long
running or critical workflow, where canceling and restarting the process is expensive or time-
consuming, it is appropriate to wait longer between attempts and retry more times.
Determining the appropriate intervals between retries is the most difficult part of designing a
successful strategy. Typical strategies use the following types of retry interval:
Exponential back-off . The application waits a short time before the first retry, and then
exponentially increasing times between each subsequent retry. For example, it may retry the
operation after 3 seconds, 12 seconds, 30 seconds, and so on.
Incremental inter vals . The application waits a short time before the first retry, and then
incrementally increasing times between each subsequent retry. For example, it may retry the
operation after 3 seconds, 7 seconds, 13 seconds, and so on.
Regular inter vals . The application waits for the same period of time between each attempt.
For example, it may retry the operation every 3 seconds.
Immediate retr y . Sometimes a transient fault is brief, perhaps due to an event such as a
network packet collision or a spike in a hardware component. In this case, retrying the
operation immediately is appropriate because it may succeed if the fault has cleared in the
time it takes the application to assemble and send the next request. However, there should
never be more than one immediate retry attempt, and you should switch to alternative
strategies, such as exponential back-off or fallback actions, if the immediate retry fails.
Randomization . Any of the retry strategies listed above may include a randomization to
prevent multiple instances of the client sending subsequent retry attempts at the same time.
For example, one instance may retry the operation after 3 seconds, 11 seconds, 28 seconds,
and so on, while another instance may retry the operation after 4 seconds, 12 seconds, 26
seconds, and so on. Randomization is a useful technique that may be combined with other
strategies.
As a general guideline, use an exponential back-off strategy for background operations, and
immediate or regular interval retry strategies for interactive operations. In both cases, you should
choose the delay and the retry count so that the maximum latency for all retry attempts is within the
required end-to-end latency requirement.
Take into account the combination of all the factors that contribute to the overall maximum timeout
for a retried operation. These factors include the time taken for a failed connection to produce a
response (typically set by a timeout value in the client) as well as the delay between retry attempts
and the maximum number of retries. The total of all these times can result in long overall operation
times, especially when using an exponential delay strategy where the interval between retries grows
rapidly after each failure. If a process must meet a specific service level agreement (SLA), the overall
operation time, including all timeouts and delays, must be within the limits defined in the SLA.
Overly aggressive retry strategies, which have intervals that are too short or retries that are too
frequent, can have an adverse effect on the target resource or service. This may prevent the resource
or service from recovering from its overloaded state, and it will continue to block or refuse requests.
This results in a vicious circle where more and more requests are sent to the resource or service, and
consequently its ability to recover is further reduced.
 Take into account the timeout of the operations when choosing the retry intervals to avoid launching
a subsequent attempt immediately (for example, if the timeout period is similar to the retry interval).
Also consider if you need to keep the total possible period (the timeout plus the retry intervals) to
below a specific total time. Operations that have unusually short or very long timeouts may influence
how long to wait, and how often to retry the operation.
Use the type of the exception and any data it contains, or the error codes and messages returned
from the service, to optimize the interval and the number of retries. For example, some exceptions or
error codes (such as the HTTP code 503 Service Unavailable with a Retry-After header in the
response) may indicate how long the error might last, or that the service has failed and will not
respond to any subsequent attempt.
Avoid anti-patterns :
In the vast majority of cases, you should avoid implementations that include duplicated layers of
retry code. Avoid designs that include cascading retry mechanisms, or that implement retry at every
stage of an operation that involves a hierarchy of requests, unless you have specific requirements
that demand this. In these exceptional circumstances, use policies that prevent excessive numbers of
retries and delay periods, and make sure you understand the consequences. For example, if one
component makes a request to another, which then accesses the target service, and you implement
retry with a count of three on both calls there will be nine retry attempts in total against the service.
Many services and resources implement a built-in retry mechanism and you should investigate how
you can disable or modify this if you need to implement retries at a higher level.
Never implement an endless retry mechanism. This is likely to prevent the resource or service
recovering from overload situations, and cause throttling and refused connections to continue for a
longer period. Use a finite number or retries, or implement a pattern such as Circuit Breaker to allow
the service to recover.
Never perform an immediate retry more than once.
Avoid using a regular retry interval, especially when you have a large number of retry attempts,
when accessing services and resources in Azure. The optimum approach is this scenario is an
exponential back-off strategy with a circuit-breaking capability.
Prevent multiple instances of the same client, or multiple instances of different clients, from sending
retries at the same times. If this is likely to occur, introduce randomization into the retry intervals.
Test your retr y strategy and implementation:
Ensure you fully test your retry strategy implementation under as wide a set of circumstances as
possible, especially when both the application and the target resources or services it uses are under
extreme load. To check behavior during testing, you can:
Inject transient and nontransient faults into the service. For example, send invalid requests or
add code that detects test requests and responds with different types of errors. For an
example using TestApi, see Fault Injection Testing with TestApi and Introduction to TestApi –
Part 5: Managed Code Fault Injection APIs.
Create a mock of the resource or service that returns a range of errors that the real service
may return. Ensure you cover all the types of error that your retry strategy is designed to
detect.
Force transient errors to occur by temporarily disabling or overloading the service if it is a
custom service that you created and deployed (of course, you should not attempt to overload
any shared resources or shared services within Azure).
For HTTP-based APIs, consider using the FiddlerCore library in your automated tests to
 change the outcome of HTTP requests, either by adding extra roundtrip times or by changing
the response (such as the HTTP status code, headers, body, or other factors). This enables
deterministic testing of a subset of the failure conditions, whether transient faults or other
types of failure. For more information, see FiddlerCore. For examples of how to use the library,
particularly the HttpMangler class, examine the source code for the Azure Storage SDK.
Perform high load factor and concurrent tests to ensure that the retry mechanism and
strategy works correctly under these conditions, and does not have an adverse effect on the
operation of the client or cause cross-contamination between requests.
Manage retr y policy configurations:
A retry policy is a combination of all of the elements of your retry strategy. It defines the detection
mechanism that determines whether a fault is likely to be transient, the type of interval to use (such
as regular, exponential back-off, and randomization), the actual interval value(s), and the number of
times to retry.
Retries must be implemented in many places within even the simplest application, and in every layer
of more complex applications. Rather than hard-coding the elements of each policy at multiple
locations, consider using a central point for storing all the policies. For example, store the values such
as the interval and retry count in application configuration files, read them at runtime, and
programmatically build the retry policies. This makes it easier to manage the settings, and to modify
and fine-tune the values in order to respond to changing requirements and scenarios. However,
design the system to store the values rather than rereading a configuration file every time, and
ensure suitable defaults are used if the values cannot be obtained from configuration.
In an Azure Cloud Services application, consider storing the values that are used to build the retry
policies at runtime in the service configuration file so that they can be changed without needing to
restart the application.
Take advantage of built-in or default retry strategies available in the client APIs you use, but only
where they are appropriate for your scenario. These strategies are typically general purpose. In some
scenarios they may be all that is required, but in other scenarios they may not offer the full range of
options to suit your specific requirements. You must understand how the settings will affect your
application through testing to determine the most appropriate values.
Log and track transient and nontransient faults:
As part of your retry strategy, include exception handling and other instrumentation that logs when
retry attempts are made. While an occasional transient failure and retry are to be expected, and do
not indicate a problem, regular and increasing numbers of retries are often an indicator of an issue
that may cause a failure, or is currently degrading application performance and availability.
Log transient faults as Warning entries rather than Error entries so that monitoring systems do not
detect them as application errors that may trigger false alerts.
Consider storing a value in your log entries that indicates if the retries were caused by throttling in
the service, or by other types of faults such as connection failures, so that you can differentiate them
during analysis of the data. An increase in the number of throttling errors is often an indicator of a
design flaw in the application or the need to switch to a premium service that offers dedicated
hardware.
Consider measuring and logging the overall time taken for operations that include a retry
mechanism. This is a good indicator of the overall effect of transient faults on user response times,
process latency, and the efficiency of the application use cases. Also log the number of retries
occurred in order to understand the factors that contributed to the response time.
 Consider implementing a telemetry and monitoring system that can raise alerts when the number
and rate of failures, the average number of retries, or the overall times taken for operations to
succeed, is increasing.
Manage operations that continually fail:
There will be circumstances where the operation continues to fail at every attempt, and it is vital to
consider how you will handle this situation:
Although a retry strategy will define the maximum number of times that an operation should
be retried, it does not prevent the application repeating the operation again, with the same
number of retries. For example, if an order processing service fails with a fatal error that puts
it out of action permanently, the retry strategy may detect a connection timeout and consider
it to be a transient fault. The code will retry the operation a specified number of times and
then give up. However, when another customer places an order, the operation will be
attempted again - even though it is sure to fail every time.
To prevent continual retries for operations that continually fail, consider implementing the
Circuit Breaker pattern. In this pattern, if the number of failures within a specified time window
exceeds the threshold, requests are returned to the caller immediately as errors, without
attempting to access the failed resource or service.
The application can periodically test the service, on an intermittent basis and with long
intervals between requests, to detect when it becomes available. An appropriate interval will
depend on the scenario, such as the criticality of the operation and the nature of the service,
and might be anything between a few minutes and several hours. At the point where the test
succeeds, the application can resume normal operations and pass requests to the newly
recovered service.
In the meantime, it may be possible to fall back to another instance of the service (perhaps in
a different datacenter or application), use a similar service that offers compatible (perhaps
simpler) functionality, or perform some alternative operations in the hope that the service will
become available soon. For example, it may be appropriate to store requests for the service in
a queue or data store and replay them later. Otherwise you might be able to redirect the user
to an alternative instance of the application, degrade the performance of the application but
still offer acceptable functionality, or just return a message to the user indicating that the
application is not available at present.
Other considerations
When deciding on the values for the number of retries and the retry intervals for a policy, consider if
the operation on the service or resource is part of a long-running or multistep operation. It may be
difficult or expensive to compensate all the other operational steps that have already succeeded
when one fails. In this case, a very long interval and a large number of retries may be acceptable as
long as it does not block other operations by holding or locking scarce resources.
Consider if retrying the same operation may cause inconsistencies in data. If some parts of a
multistep process are repeated, and the operations are not idempotent, it may result in an
inconsistency. For example, an operation that increments a value, if repeated, will produce an invalid
result. Repeating an operation that sends a message to a queue may cause an inconsistency in the
message consumer if it cannot detect duplicate messages. To prevent this, ensure that you design
each step as an idempotent operation. For more information about idempotency, see Idempotency
patterns.
Consider the scope of the operations that will be retried. For example, it may be easier to implement
retry code at a level that encompasses several operations, and retry them all if one fails. However,
 doing this may result in idempotency issues or unnecessary rollback operations.
If you choose a retry scope that encompasses several operations, take into account the total latency
of all of them when determining the retry intervals, when monitoring the time taken, and before
raising alerts for failures.
Consider how your retry strategy may affect neighbors and other tenants in a shared application, or
when using shared resources and services. Aggressive retry policies can cause an increasing number
of transient faults to occur for these other users and for applications that share the resources and
services. Likewise, your application may be affected by the retry policies implemented by other users
of the resources and services. For mission-critical applications, you may decide to use premium
services that are not shared. This provides you with much more control over the load and
consequent throttling of these resources and services, which can help to justify the additional cost.

More information
Azure service-specific retry guidelines
Circuit Breaker pattern
Compensating Transaction pattern
Idempotency patterns
 Performance tuning a distributed application
11/2/2020 • 2 minutes to read • Edit Online

In this series, we walk through several cloud application scenarios, showing how a development team used load
tests and metrics to diagnose performance issues. These articles are based on actual load testing that we
performed when developing example applications. The code for each scenario is available on GitHub.
Scenarios:
Distributed business transaction
Calling multiple backend services
Event stream processing

What is performance?
Performance is frequently measured in terms of throughput, response time, and availability. Performance targets
should be based on business operations. Customer-facing tasks may have more stringent requirements than
operational tasks such as generating reports.
Define a service level objective (SLO) that defines performance targets for each workload. You typically achieve this
by breaking a performance target into a set of Key Performance Indicators (KPIs), such as:
Latency or response time of specific requests
The number of requests performed per second
The rate at which the system generates exceptions.
Performance targets should explicitly include a target load. Also, not all users will receive exactly the same level of
performance, even when accessing the system simultaneously and performing the same work. So an SLO should
be framed in terms of percentiles.
An example SLO for might be: "Client requests will have a response within 500 ms @ P90, at loads up to 25 K
requests/second."

Challenges of performance tuning a distributed system

It can be especially challenging to diagnose performance issues in a distributed application. Some of the
challenges are:
A single business transaction or operation typically involves multiple components of the system. It can be
hard to get a holistic end-to-end view of a single operation.
Resource consumption is distributed across multiple nodes. To get a consistent view, you need to aggregate
logs and metrics in one place.
The cloud offers elastic scale. Autoscaling is an important technique for handling spikes in load, but it can
also mask underlying issues. Also, it can be hard to know which components need to scale and when.
Cascading failures can cause failures upstream of the root problem. As a result, the first signal of the
problem may appear in a different component than the root cause.

General best practices

Performance tuning is both an art and a science, but it can be made closer to science by taking a systematic
approach. Here are some best practices:
Enable telemetry to collect metrics. Instrument your code. Follow best practices for monitoring. Use
correlated tracing so that you can view all the steps in a transaction.
Monitor the 90/95/99 percentiles, not just average. The average can mask outliers. The sampling rate for
metrics also matters. If the sampling rate is too low, it can hide spikes or outliers that might indicate
problems.
Attack one bottleneck at a time. Form a hypothesis and test it by changing one variable at a time. Removing
one bottleneck will often uncover another bottleneck further upstream or downstream.
Errors and retries can have a large impact on performance. If you see that you are being throttled by
backend services, scale out or try to optimize usage (for example by tuning database queries).
Look for common performance anti-patterns.
Look for opportunities to parallelize. Two common sources of bottlenecks are message queues and
databases. In both cases, sharding can help. For more information, see Horizontal, vertical, and functional
data partitioning. Look for hot partitions that might indicate imbalanced read or write loads.

Next steps
Read the performance tuning scenarios
Distributed business transaction
Calling multiple backend services
Event stream processing
 Performance tuning scenario: Distributed business
transactions
11/2/2020 • 9 minutes to read • Edit Online

This article describes how a development team used metrics to find bottlenecks and improve the performance of a
distributed system. The article is based on actual load testing that we did for a sample application. The application
code is available on GitHub.
This article is part of a series. Read the first part here.
Scenario : A client application initiates a business transaction that involves multiple steps.
This scenario involves a drone delivery application that runs on Azure Kubernetes Service (AKS). Customers use a
web app to schedule deliveries by drone. Each transaction requires multiple steps that are performed by separate
microservices on the back end:
The Delivery service manages deliveries.
The Drone Scheduler service schedules drones for pickup.
The Package service manages packages.
There are two other services: An Ingestion service that accepts client requests and puts them on a queue for
processing, and a Workflow service that coordinates the steps in the workflow.

For more information about this scenario, see Designing a microservices architecture.

Test 1: Baseline
For the first load test, the team created a 6-node AKS cluster and deployed three replicas of each microservice. The
load test was a step-load test, starting at two simulated users and ramping up to 40 simulated users.

SET T IN G VA L UE

Cluster nodes 6

Pods 3 per service

The following graph shows the results of the load test, as shown in Visual Studio. The purple line plots user load,
and the orange line plots total requests.
The first thing to realize about this scenario is that client requests per second is not a useful metric of performance.
That's because the application processes requests asynchronously, so the client gets a response right away. The
response code is always HTTP 202 (Accepted), meaning the request was accepted but processing is not complete.
What we really want to know is whether the backend is keeping up with the request rate. The Service Bus queue
can absorb spikes, but if the backend cannot handle a sustained load, processing will fall further and further
behind.
Here's a more informative graph. It plots the number incoming and outgoing messages on the Service Bus queue.
Incoming messages are shown in light blue, and outgoing messages are shown in dark blue:

This chart is showing that the rate of incoming messages increases, reaching a peak and then dropping back to
zero at the end of the load test. But the number of outgoing messages peaks early in the test and then actually
drops. That means the Workflow service, which handles the requests, isn't keeping up. Even after the load test ends
(around 9:22 on the graph), messages are still being processed as the Workflow service continues to drain the
queue.
What's slowing down the processing? The first thing to look for is errors or exceptions that might indicate a
systematic issue. The Application Map in Azure Monitor shows the graph of calls between components, and is a
quick way to spot issues and then click through to get more details.
Sure enough, the Application Map shows that the Workflow service is getting errors from the Delivery service:
To see more details, you can select a node in the graph and click into an end-to-end transaction view. In this case, it
shows that the Delivery service is returning HTTP 500 errors. The error messages indicate that an exception is
being thrown due to memory limits in Azure Cache for Redis.

You may notice that these calls to Redis don't appear in the Application Map. That's because the .NET library for
Application Insights doesn't have built-in support for tracking Redis as a dependency. (For a list of what's
supported out of the box, see Dependency auto-collection.) As a fallback, you can use the TrackDependency API to
track any dependency. Load testing often reveals these kinds of gaps in the telemetry, which can be remediated.

Test 2: Increased cache size

For the second load test, the development team increased the cache size in Azure Cache for Redis. (See How to
Scale Azure Cache for Redis.) This change resolved the out-of-memory exceptions, and now the Application Map
shows zero errors:

However, there is still a dramatic lag in processing messages. At the peak of the load test, the incoming message
rate is more than 5× the outgoing rate:

The following graph measures throughput in terms of message completion — that is, the rate at which the
Workflow service marks the Service Bus messages as completed. Each point on the graph represents 5 seconds of
data, showing ~16/sec maximum throughput.
This graph was generated by running a query in the Log Analytics workspace, using the Kusto query language:

let start=datetime("2019-07-31T22:30:00.000Z");
let end=datetime("2019-07-31T22:45:00.000Z");
dependencies
| where cloud_RoleName == 'fabrikam-workflow'
| where timestamp > start and timestamp < end
| where type == 'Azure Service Bus'
| where target has 'https://dev-i-iuosnlbwkzkau.servicebus.windows.net'
| where client_Type == "PC"
| where name == "Complete"
| summarize succeeded=sumif(itemCount, success == true), failed=sumif(itemCount, success == false) by
bin(timestamp, 5s)
| render timechart

Test 3: Scale out the backend services

It appears the back end is the bottleneck. An easy next step is to scale out the business services (Package, Delivery,
and Drone Scheduler), and see if throughput improves. For the next load test, the team scaled these services out
from three replicas to six replicas.

SET T IN G VA L UE

Cluster nodes 6

Ingestion service 3 replicas

Workflow service 3 replicas

Package, Delivery, Drone Scheduler services 6 replicas each

Unfortunately this load test shows only modest improvement. Outgoing messages are still not keeping up with
incoming messages:
Throughput is more consistent, but the maximum achieved is about the same as the previous test:

Moreover, looking at Azure Monitor for containers, it appears the problem is not caused by resource exhaustion
within the cluster. First, the node-level metrics show that CPU utilization remains under 40% even at the 95th
percentile, and memory utilization is about 20%.

In a Kubernetes environment, it's possible for individual pods to be resource-constrained even when the nodes
aren't. But the pod-level view shows that all pods are healthy.
From this test, it seems that just adding more pods to the back end won't help. The next step is to look more closely
at the Workflow service to understand what's happening when it processes messages. Application Insights shows
that the average duration of the Workflow service's Process operation is 246 ms.

We can also run a query to get metrics on the individual operations within each transaction:

TA RGET P ERC EN T IL E_DURAT IO N _50 P ERC EN T IL E_DURAT IO N _95

https://dev-i- 86.66950203 283.4255578

iuosnlbwkzkau.servicebus.windows.net/
| dev-i-iuosnlbwkzkau

delivery 37 57

package 12 17

dronescheduler 21 41

The first row in this table represents the Service Bus queue. The other rows are the calls to the backend services.
For reference, here's the Log Analytics query for this table:

let start=datetime("2019-07-31T22:30:00.000Z");
let end=datetime("2019-07-31T22:45:00.000Z");
let dataset=dependencies
| where timestamp > start and timestamp < end
| where (cloud_RoleName == 'fabrikam-workflow')
| where name == 'Complete' or target in ('package', 'delivery', 'dronescheduler');
dataset
| summarize percentiles(duration, 50, 95) by target
These latencies look reasonable. But here is the key insight: If the total operation time is ~250 ms, that puts a strict
upper bound on how fast messages can be processed in serial. The key to improving throughput, therefore, is
greater parallelism.
That should be possible in this scenario, for two reasons:
These are network calls, so most of the time is spent waiting for I/O completion
The messages are independent, and don't need to be processed in order.

Test 4: Increase parallelism

For this test, the team focused on increasing parallelism. To do so, they adjusted two settings on the Service Bus
client used by the Workflow service:

SET T IN G DESC RIP T IO N DEFA ULT N EW VA L UE

MaxConcurrentCalls The maximum number of 1 20

messages to process
concurrently.

PrefetchCount How many messages the 0 3000

client will fetch ahead of
time into its local cache.

For more information about these settings, see Best Practices for performance improvements using Service Bus
Messaging. Running the test with these settings produced the following graph:
Recall that incoming messages are shown in light blue, and outgoing messages are shown in dark blue.
At first glance, this is a very weird graph. For a while, the outgoing message rate exactly tracks the incoming rate.
But then, at about the 2:03 mark, the rate of incoming messages levels off, while the number of outgoing messages
continues to rise, actually exceeding the total number of incoming messages. That seems impossible.
The clue to this mystery can be found in the Dependencies view in Application Insights. This chart summarizes all
of the calls that the Workflow service made to Service Bus:

Notice that entry for DeadLetter . That calls indicates messages are going into the Service Bus dead-letter queue.
To understand what's happening, you need to understand Peek-Lock semantics in Service Bus. When a client uses
Peek-Lock, Service Bus atomically retrieves and locks a message. While the lock is held, the message is guaranteed
not to be delivered to other receivers. If the lock expires, the message becomes available to other receivers. After a
maximum number of delivery attempts (which is configurable), Service Bus will put the messages onto a dead-
letter queue, where it can be examined later.
Remember that the Workflow service is prefetching large batches of messages — 3000 messages at a time). That
means the total time to process each message is longer, which results in messages timing out, going back onto the
queue, and eventually going into the dead-letter queue.
You can also see this behavior in the exceptions, where numerous MessageLostLockException exceptions are
recorded:

Test 5: Increase lock duration

For this load test, the message lock duration was set to 5 minutes, to prevent lock timeouts. The graph of incoming
and outgoing messages now shows that the system is keeping up with the rate of incoming messages:

Over the total duration of the 8-minute load test, the application completed 25 K operations, with a peak
throughput of 72 operations/sec, representing a 400% increase in maximum throughput.
However, running the same test with a longer duration showed that the application could not sustain this rate:

The container metrics show that maximum CPU utilization was close to 100%. At this point, the application appears
to be CPU-bound. Scaling the cluster might improve performance now, unlike the previous attempt at scaling out.

Test 6: Scale out the backend services (again)

For the final load test in the series, the team scaled out the Kubernetes cluster and pods as follows:

SET T IN G VA L UE

Cluster nodes 12

Ingestion service 3 replicas

Workflow service 6 replicas

Package, Delivery, Drone Scheduler services 9 replicas each

This test resulted in a higher sustained throughput, with no significant lags in processing messages. Moreover,
node CPU utilization stayed below 80%.
Summary
For this scenario, the following bottlenecks were identified:
Out-of-memory exceptions in Azure Cache for Redis.
Lack of parallelism in message processing.
Insufficient message lock duration, leading to lock timeouts and messages being placed in the dead letter
queue.
CPU exhaustion.
To diagnose these issues, the development team relied on the following metrics:
The rate of incoming and outgoing Service Bus messages.
Application Map in Application Insights.
Errors and exceptions.
Custom Log Analytics queries.
CPU and memory utilization in Azure Monitor for containers.

Next steps
For more information about the design of this scenario, see Designing a microservices architecture.
 Performance tuning scenario: Multiple backend
services
11/2/2020 • 10 minutes to read • Edit Online

This article describes how a development team used metrics to find bottlenecks and improve the performance of a
distributed system. The article is based on actual load testing that we did for a sample application. The application
code is available on GitHub, along with the Visual Studio load test project used to generate the results.
This article is part of a series. Read the first part here.
Scenario : Call multiple backend services to retrieve information and then aggregate the results.
This scenario involves a drone delivery application. Clients can query a REST API to get their latest invoice
information. The invoice includes a summary of the customer's deliveries, packages, and total drone utilization.
This application uses a microservices architecture running on Azure Kubernetes Service (AKS), and the information
needed for the invoice is spread across several microservices.
Rather than the client calling each service directly, the application implements the Gateway Aggregation pattern.
Using this pattern, the client makes a single request to a gateway service. The gateway in turn calls the backend
services in parallel, and then aggregates the results into a single response payload.

Test 1: Baseline performance

To establish a baseline, the development team started with a step-load test, ramping the load from one simulated
user up to 40 users, over a total duration of 8 minutes. The following chart, taken from Visual Studio, shows the
results. The purple line shows the user load, and the orange line shows throughput (average requests per second).
The red line along the bottom of the chart shows that no errors were returned to the client, which is encouraging.
However, the average throughput peaks about half way through the test, and then drops off for the remainder,
even while the load continues to increase. That indicates the back end is not able to keep up. The pattern seen here
is common when a system starts to reach resource limits — after reaching a maximum, throughput actually falls
significantly. Resource contention, transient errors, or an increase in the rate of exceptions can all contribute to this
pattern.
Let's dig into the monitoring data to learn what's happening inside the system. The next chart is taken from
Application Insights. It shows the average durations of the HTTP calls from the gateway to the backend services.

This chart shows that one operation in particular, GetDroneUtilization , takes much longer on average — by an
order of magnitude. The gateway makes these calls in parallel, so the slowest operation determines how long it
takes for the entire request to complete.
Clearly the next step is dig into the GetDroneUtilization operation and look for any bottlenecks. One possibility is
resource exhaustion. Perhaps this particular backend service is running out of CPU or memory. For an AKS cluster,
this information is available in the Azure portal through the Azure Monitor for containers feature. The following
graphs show resource utilization at the cluster level:
In this screenshot, both the average and maximum values are shown. It's important to look at more than just the
average, because the average can hide spikes in the data. Here, the average CPU utilization stays below 50%, but
there are a couple of spikes to 80%. That's close to capacity but still within tolerances. Something else is causing
the bottleneck.
The next chart reveals the true culprit. This chart shows HTTP response codes from the Delivery service's backend
database, which in this case is Cosmos DB. The blue line represents success codes (HTTP 2xx), while the green line
represents HTTP 429 errors. An HTTP 429 return code means that Cosmos DB is temporarily throttling requests,
because the caller is consuming more resource units (RU) than provisioned.

To get further insight, the development team used Application Insights to view the end-to-end telemetry for a
representative sample of requests. Here is one instance:
This view shows the calls related to a single client request, along with timing information and response codes. The
top-level calls are from the gateway to the backend services. The call to GetDroneUtilization is expanded to show
calls to external dependencies — in this case, to Cosmos DB. The call in red returned an HTTP 429 error.
Note the large gap between the HTTP 429 error and the next call. When the Cosmos DB client library receives an
HTTP 429 error, it automatically backs off and waits to retry the operation. What this view shows is that during the
672 ms this operation took, most of that time was spent waiting to retry Cosmos DB.
Here's another interesting graph for this analysis. It shows RU consumption per physical partition versus
provisioned RUs per physical partition:

To make sense of this graph, you need to understand how Cosmos DB manages partitions. Collections in Cosmos
DB can have a partition key. Each possible key value defines a logical partition of the data within the collection.
Cosmos DB distributes these logical partitions across one or more physical partitions. The management of physical
partitions is handled automatically by Cosmos DB. As you store more data, Cosmos DB might move logical
partitions into new physical partitions, in order to spread load across the physical partitions.
For this load test, the Cosmos DB collection was provisioned with 900 RUs. The chart shows 100 RU per physical
partition, which implies a total of nine physical partitions. Although Cosmos DB automatically handles the sharding
of physical partitions, knowing the partition count can give insight into performance. The development team will
use this information later, as they continue to optimize. Where the blue line crosses the purple horizontal line, RU
consumption has exceeded the provisioned RUs. That's the point where Cosmos DB will begin to throttle calls.

Test 2: Increase resource units

For the second load test, the team scaled out the Cosmos DB collection from 900 RU to 2500 RU. Throughput
increased from 19 requests/second to 23 requests/second, and average latency dropped from 669 ms to 569 ms.

M ET RIC T EST 1 T EST 2

Throughput (req/sec) 19 23

Average latency (ms) 669 569

Successful requests 9.8 K 11 K

These aren't huge gains, but looking at the graph over time shows a more complete picture:

Whereas the previous test showed an initial spike followed by a sharp drop, this test shows more consistent
throughput. However, the maximum throughput is not significantly higher.
All requests to Cosmos DB returned a 2xx status, and the HTTP 429 errors went away:
The graph of RU consumption versus provisioned RUs shows there is plenty of headroom. There are about 275
RUs per physical partition, and the load test peaked at about 100 RUs consumed per second.

Another interesting metric is the number of calls to Cosmos DB per successful operation:

M ET RIC T EST 1 T EST 2

Calls per operation 11 9

Assuming no errors, the number of calls should match the actual query plan. In this case, the operation involves a
cross-partition query that hits all nine physical partitions. The higher value in the first load test reflects the number
of calls that returned a 429 error.
This metric was calculated by running a custom Log Analytics query:

let start=datetime("2019-06-18T20:59:00.000Z");
let end=datetime("2019-07-24T21:10:00.000Z");
let operationNameToEval="GET DroneDeliveries/GetDroneUtilization";
let dependencyType="Azure DocumentDB";
let dataset=requests
| where timestamp > start and timestamp < end
| where success == true
| where name == operationNameToEval;
dataset
| project reqOk=itemCount
| summarize
SuccessRequests=sum(reqOk),
TotalNumberOfDepCalls=(toscalar(dependencies
| where timestamp > start and timestamp < end
| where type == dependencyType
| summarize sum(itemCount)))
| project
OperationName=operationNameToEval,
DependencyName=dependencyType,
SuccessRequests,
AverageNumberOfDepCallsPerOperation=(TotalNumberOfDepCalls/SuccessRequests)

To summarize, the second load test shows improvement. However, the GetDroneUtilization operation still takes
about an order of magnitude longer than the next-slowest operation. Looking at the end-to-end transactions helps
to explain why:

As mentioned earlier, the GetDroneUtilization operation involves a cross-partition query to Cosmos DB. This
means the Cosmos DB client has to fan out the query to each physical partition and collect the results. As the end-
to-end transaction view shows, these queries are being performed in serial. The operation takes as long as the sum
of all the queries — and this problem will only get worse as the size of the data grows and more physical partitions
are added.
Test 3: Parallel queries
Based on the previous results, an obvious way to reduce latency is to issue the queries in parallel. The Cosmos DB
client SDK has a setting that controls the maximum degree of parallelism.

VA L UE DESC RIP T IO N

0 No parallelism (default)

>0 Maximum number of parallel calls

-1 The client SDK selects an optimal degree of parallelism

For the third load test, this setting was changed from 0 to -1. The following table summarizes the results:

M ET RIC T EST 1 T EST 2 T EST 3

Throughput (req/sec) 19 23 42

Average latency (ms) 669 569 215

Successful requests 9.8 K 11 K 20 K

Throttled requests 2.72 K 0 0

From the load test graph, not only is the overall throughput much higher (the orange line), but throughput also
keeps pace with the load (the purple line).

We can verify that the Cosmos DB client is making queries in parallel by looking at the end-to-end transaction
view:
Interestingly, a side effect of increasing the throughput is that the number of RUs consumed per second also
increases. Although Cosmos DB did not throttle any requests during this test, the consumption was close to the
provisioned RU limit:

This graph might be a signal to further scale out the database. However, it turns out that we can optimize the query
instead.

Step 4: Optimize the query

The previous load test showed better performance in terms of latency and throughput. Average request latency
was reduced by 68% and throughput increased 220%. However, the cross-partition query is a concern.
The problem with cross-partition queries is that you pay for RU across every partition. If the query is only run
occasionally — say, once an hour — it might not matter. But whenever you see a read-heavy workload that
involves a cross-partition query, you should see whether the query can be optimized by including a partition key.
(You might need to redesign the collection to use a different partition key.)
Here's the query for this particular scenario:

SELECT * FROM c
WHERE c.ownerId = <ownerIdValue> and
c.year = <yearValue> and
c.month = <monthValue>

This query selects records that match a particular owner ID and month/year. In the original design, none of these
properties is the partition key. That requires the client to fan out the query to each physical partition and gather
the results. To improve query performance, the development team changed the design so that owner ID is the
partition key for the collection. That way, the query can target a specific physical partition. (Cosmos DB handles this
automatically; you don't have to manage the mapping between partition key values and physical partitions.)
After switching the collection to the new partition key, there was a dramatic improvement in RU consumption,
which translates directly into lower costs.

M ET RIC T EST 1 T EST 2 T EST 3 T EST 4

RUs per operation 29 29 29 3.4

Calls per operation 11 9 10 1

The end-to-end transaction view shows that as predicted, the query reads only one physical partition:

The load test shows improved throughput and latency:

M ET RIC T EST 1 T EST 2 T EST 3 T EST 4

Throughput (req/sec) 19 23 42 59

Average latency (ms) 669 569 215 176

Successful requests 9.8 K 11 K 20 K 29 K

Throttled requests 2.72 K 0 0 0

A consequence of the improved performance is that node CPU utilization becomes very high:
Toward the end of the load test, average CPU reached about 90%, and maximum CPU reached 100%. This metric
indicates that CPU is the next bottleneck in the system. If higher throughput is needed, the next step might be
scaling out the Delivery service to more instances.

Summary
For this scenario, the following bottlenecks were identified:
Cosmos DB throttling requests due to insufficient RUs provisioned.
High latency caused by querying multiple database partitions in serial.
Inefficient cross-partition query, because the query did not include the partition key.
In addition, CPU utilization was identified as a potential bottleneck at higher scale. To diagnose these issues, the
development team looked at:
Latency and throughput from the load test.
Cosmos DB errors and RU consumption.
The end-to-end transaction view in Application Insight.
CPU and memory utilization in Azure Monitor for containers.

Next steps
Review performance antipatterns
 Performance tuning scenario: Event streaming with
Azure Functions
11/2/2020 • 7 minutes to read • Edit Online

This article describes how a development team used metrics to find bottlenecks and improve the performance of a
distributed system. The article is based on actual load testing that we did for a sample application. The application
code is available on GitHub.
This article is part of a series. Read the first part here.
Scenario : Process a stream of events using Azure Functions.

In this scenario, a fleet of drones sends position data in real time to Azure IoT Hub. A Functions app receives the
events, transforms the data into GeoJSON format, and writes the transformed data to Cosmos DB. Cosmos DB has
native support for geospatial data, and Cosmos DB collections can be indexed for efficient spatial queries. For
example, a client application could query for all drones within 1 km of a given location, or find all drones within a
certain area.
These processing requirements are simple enough that they don't require a full-fledged stream processing engine.
In particular, the processing doesn't join streams, aggregate data, or process across time windows. Based on these
requirements, Azure Functions is a good fit for processing the messages. Cosmos DB can also scale to support
very high write throughput.

Monitoring throughput
This scenario presents an interesting performance challenge. The data rate per device is known, but the number of
devices may fluctuate. For this business scenario, the latency requirements are not particularly stringent. The
reported position of a drone only needs to be accurate within a minute. That said, the function app must keep up
with the average ingestion rate over time.
IoT Hub stores messages in a log stream. Incoming messages are appended to the tail of the stream. A reader of
the stream — in this case, the function app — controls its own rate of traversing the stream. This decoupling of the
read and write paths makes IoT Hub very efficient, but also means that a slow reader can fall behind. To detect this
condition, the development team added a custom metric to measure message lateness. This metric records the
delta between when a message arrives at IoT Hub, and when the function receives the message for processing.

var ticksUTCNow = DateTimeOffset.UtcNow;

// Track whether messages are arriving at the function late.

DateTime? firstMsgEnqueuedTicksUtc = messages[0]?.EnqueuedTimeUtc;
if (firstMsgEnqueuedTicksUtc.HasValue)
{
CustomTelemetry.TrackMetric(
context,
"IoTHubMessagesReceivedFreshnessMsec",
(ticksUTCNow - firstMsgEnqueuedTicksUtc.Value).TotalMilliseconds);
}
The TrackMetric method writes a custom metric to Application Insights. For information about using TrackMetric
inside an Azure Function, see Custom telemetry in C# function.
If the function keeps up with the volume of messages, this metric should stay at a low steady state. Some latency is
unavoidable, so the value will never be zero. But if the function falls behind, the delta between enqueued time and
processing time will start to go up.

Test 1: Baseline
The first load test showed an immediate problem: The Function app consistently received HTTP 429 errors from
Cosmos DB, indicating that Cosmos DB was throttling the write requests.

In response, the team scaled Cosmos DB by increasing the number RUs allocated for the collection, but the errors
continued. This seemed strange, because their back-of-envelope calculation showed that Cosmos DB should have
no problem keeping up with the volume of write requests.
Later that day, one of the developers sent the following email to the team:

I looked at Cosmos DB for the warm path. There's one thing I don't understand. The partition key is deliveryId,
however we don't send deliveryId to Cosmos DB. Am I missing something?

That was the clue. Looking at the partition heat map, it turned out that all of the documents were landing on the
same partition.
What you want to see in the heat map is an even distribution across all of the partitions. In this case, because every
document was getting written to the same partition, adding RUs didn't help. The problem turned out to be a bug in
the code. Although the Cosmos DB collection had a partition key, the Azure Function didn't actually include the
partition key in the document. For more information about the partition heat map, see Determine the throughput
distribution across partitions.

Test 2: Fix partitioning issue

When the team deployed a code fix and re-ran the test, Cosmos DB stopped throttling. For a while, everything
looked good. But at a certain load, telemetry showed that the function was writing fewer documents that it should.
The following graph shows messages receives from IoT Hub versus documents written to Cosmos DB. The yellow
line is number of messages received per batch, and the green is the number of documents written per batch. These
should be proportional. Instead, the number of database write operations per batch drops significantly at about
07:30.
The next graph shows the latency between when a message arrives at IoT Hub from a device, and when the
function app processes that message. You can see that at the same point in time, the lateness spikes dramatically,
levels off, and the declines.

The reason the value peaks at 5 minutes and then drops to zero is because the function app discards messages
that are more than 5 minutes late:

foreach (var message in messages)

{
// Drop stale messages,
if (message.EnqueuedTimeUtc < cutoffTime)
{
log.Info($"Dropping late message batch. Enqueued time = {message.EnqueuedTimeUtc}, Cutoff =
{cutoffTime}");
droppedMessages++;
continue;
}

You can see this in the graph when the lateness metric drops back to zero. In the meantime, data has been lost,
because the function was throwing away messages.
What was happening? For this particular load test, the Cosmos DB collection had RUs to spare, so the bottleneck
was not at the database. Rather, the problem was in the message processing loop. Simply put, the function was not
writing documents quickly enough to keep up with the incoming volume of messages. Over time, it fell further and
further behind.
Test 3: Parallel writes
If the time to process a message is the bottleneck, one solution is to process more messages in parallel. In this
scenario:
Increase the number of IoT Hub partitions. Each IoT Hub partition gets assigned one function instance at a time,
so we would expect throughput to scale linearly with the number of partitions.
Parallelize the document writes within the function.
To explore the second option, the team modified the function to support parallel writes. The original version of the
function used the Cosmos DB output binding. The optimized version calls the Cosmos DB client directly and
performs the writes in parallel using Task.WhenAll:

private async Task<(long documentsUpserted,

long droppedMessages,
long cosmosDbTotalMilliseconds)>
ProcessMessagesFromEventHub(
int taskCount,
int numberOfDocumentsToUpsertPerTask,
EventData[] messages,
TraceWriter log)
{
DateTimeOffset cutoffTime = DateTimeOffset.UtcNow.AddMinutes(-5);

var tasks = new List<Task>();

for (var i = 0; i < taskCount; i++)

{
var docsToUpsert = messages
.Skip(i * numberOfDocumentsToUpsertPerTask)
.Take(numberOfDocumentsToUpsertPerTask);
// client will attempt to create connections to the data
// nodes on Cosmos db's clusters on a range of port numbers
tasks.Add(UpsertDocuments(i, docsToUpsert, cutoffTime, log));
}

await Task.WhenAll(tasks);

return (this.UpsertedDocuments,
this.DroppedMessages,
this.CosmosDbTotalMilliseconds);
}

Note that race conditions are possible with approach. Suppose that two messages from the same drone happen to
arrive in the same batch of messages. By writing them in parallel, the earlier message could overwrite the later
message. For this particular scenario, the application can tolerate losing an occasional message. Drones send new
position data every 5 seconds, so the data in Cosmos DB is updated continually. In other scenarios, however, it may
be important to process messages strictly in order.
After deploying this code change, the application was able to ingest more than 2500 requests/sec, using an IoT
Hub with 32 partitions.

Client-side Consideration
Overall client experience may be diminished by aggressive parallelization on server side. Consider leveraging
Azure Cosmos DB bulk executor library (not shown in this implementation) which significantly reduces the client-
side compute resources needed to saturate the throughput allocated to a Cosmos DB container. A single threaded
application that writes data using the bulk import API achieves nearly ten times greater write throughput when
compared to a multi-threaded application that writes data in parallel while saturating the client machine's CPU.
Summary
For this scenario, the following bottlenecks were identified:
Hot write partition, due to a missing partition key value in the documents being written.
Writing documents in serial per IoT Hub partition.
To diagnose these issues, the development team relied on the following metrics:
Throttled requests in Cosmos DB.
Partition heat map — Maximum consumed RUs per partition.
Messages received versus documents created.
Message lateness.

Next steps
Review performance antipatterns
 Performance antipatterns for cloud applications
12/18/2020 • 2 minutes to read • Edit Online

A performance antipattern is a common practice that is likely to cause scalability problems when an application is
under pressure.
Here is a common scenario: An application behaves well during performance testing. It's released to production,
and begins to handle real workloads. At that point, it starts to perform poorly—rejecting user requests, stalling, or
throwing exceptions. The development team is then faced with two questions:
Why didn't this behavior show up during testing?
How do we fix it?
The answer to the first question is straightforward. It's difficult to simulate real users in a test environment, along
with their behavior patterns and the volumes of work they might perform. The only completely sure way to
understand how a system behaves under load is to observe it in production. To be clear, we aren't suggesting that
you should skip performance testing. Performance tests are crucial for getting baseline performance metrics. But
you must be prepared to observe and correct performance issues when they arise in the live system.
The answer to the second question, how to fix the problem, is less straightforward. Any number of factors might
contribute, and sometimes the problem only manifests under certain circumstances. Instrumentation and logging
are key to finding the root cause, but you also have to know what to look for.
Based on our engagements with Microsoft Azure customers, we've identified some of the most common
performance issues that customers see in production. For each antipattern, we describe why the antipattern
typically occurs, symptoms of the antipattern, and techniques for resolving the problem. We also provide sample
code that illustrates both the antipattern and a suggested solution.
Some of these antipatterns may seem obvious when you read the descriptions, but they occur more often than
you might think. Sometimes an application inherits a design that worked on-premises, but doesn't scale in the
cloud. Or an application might start with a very clean design, but as new features are added, one or more of these
antipatterns creeps in. Regardless, this guide will help you to identify and fix these antipatterns.

Catalog of antipatterns
Here is the list of the antipatterns that we've identified:

A N T IPAT T ERN DESC RIP T IO N

Busy Database Offloading too much processing to a data store.

Busy Front End Moving resource-intensive tasks onto background threads.

Chatty I/O Continually sending many small network requests.

Extraneous Fetching Retrieving more data than is needed, resulting in unnecessary

I/O.

Improper Instantiation Repeatedly creating and destroying objects that are designed
to be shared and reused.
 A N T IPAT T ERN DESC RIP T IO N

Monolithic Persistence Using the same data store for data with very different usage
patterns.

No Caching Failing to cache data.

Synchronous I/O Blocking the calling thread while I/O completes.

Next steps
For more about performance tuning, see Performance tuning a distributed application
 Busy Database antipattern
12/18/2020 • 7 minutes to read • Edit Online

Offloading processing to a database server can cause it to spend a significant proportion of time running code,
rather than responding to requests to store and retrieve data.

Problem description
Many database systems can run code. Examples include stored procedures and triggers. Often, it's more efficient to
perform this processing close to the data, rather than transmitting the data to a client application for processing.
However, overusing these features can hurt performance, for several reasons:
The database server may spend too much time processing, rather than accepting new client requests and
fetching data.
A database is usually a shared resource, so it can become a bottleneck during periods of high use.
Runtime costs may be excessive if the data store is metered. That's particularly true of managed database
services. For example, Azure SQL Database charges for Database Transaction Units (DTUs).
Databases have finite capacity to scale up, and it's not trivial to scale a database horizontally. Therefore, it may
be better to move processing into a compute resource, such as a VM or App Service app, that can easily scale
out.
This antipattern typically occurs because:
The database is viewed as a service rather than a repository. An application might use the database server to
format data (for example, converting to XML), manipulate string data, or perform complex calculations.
Developers try to write queries whose results can be displayed directly to users. For example, a query might
combine fields or format dates, times, and currency according to locale.
Developers are trying to correct the Extraneous Fetching antipattern by pushing computations to the database.
Stored procedures are used to encapsulate business logic, perhaps because they are considered easier to
maintain and update.
The following example retrieves the 20 most valuable orders for a specified sales territory and formats the results
as XML. It uses Transact-SQL functions to parse the data and convert the results to XML. You can find the complete
sample here.
 SELECT TOP 20
soh.[SalesOrderNumber] AS '@OrderNumber',
soh.[Status] AS '@Status',
soh.[ShipDate] AS '@ShipDate',
YEAR(soh.[OrderDate]) AS '@OrderDateYear',
MONTH(soh.[OrderDate]) AS '@OrderDateMonth',
soh.[DueDate] AS '@DueDate',
FORMAT(ROUND(soh.[SubTotal],2),'C')
AS '@SubTotal',
FORMAT(ROUND(soh.[TaxAmt],2),'C')
AS '@TaxAmt',
FORMAT(ROUND(soh.[TotalDue],2),'C')
AS '@TotalDue',
CASE WHEN soh.[TotalDue] > 5000 THEN 'Y' ELSE 'N' END
AS '@ReviewRequired',
(
SELECT
c.[AccountNumber] AS '@AccountNumber',
UPPER(LTRIM(RTRIM(REPLACE(
CONCAT( p.[Title], ' ', p.[FirstName], ' ', p.[MiddleName], ' ', p.[LastName], ' ', p.[Suffix]),
' ', ' ')))) AS '@FullName'
FROM [Sales].[Customer] c
INNER JOIN [Person].[Person] p
ON c.[PersonID] = p.[BusinessEntityID]
WHERE c.[CustomerID] = soh.[CustomerID]
FOR XML PATH ('Customer'), TYPE
),

(
SELECT
sod.[OrderQty] AS '@Quantity',
FORMAT(sod.[UnitPrice],'C')
AS '@UnitPrice',
FORMAT(ROUND(sod.[LineTotal],2),'C')
AS '@LineTotal',
sod.[ProductID] AS '@ProductId',
CASE WHEN (sod.[ProductID] >= 710) AND (sod.[ProductID] <= 720) AND (sod.[OrderQty] >= 5) THEN 'Y' ELSE
'N' END
AS '@InventoryCheckRequired'

FROM [Sales].[SalesOrderDetail] sod

WHERE sod.[SalesOrderID] = soh.[SalesOrderID]
ORDER BY sod.[SalesOrderDetailID]
FOR XML PATH ('LineItem'), TYPE, ROOT('OrderLineItems')
)

FROM [Sales].[SalesOrderHeader] soh

WHERE soh.[TerritoryId] = @TerritoryId
ORDER BY soh.[TotalDue] DESC
FOR XML PATH ('Order'), ROOT('Orders')

Clearly, this is complex query. As we'll see later, it turns out to use significant processing resources on the database
server.

How to fix the problem

Move processing from the database server into other application tiers. Ideally, you should limit the database to
performing data access operations, using only the capabilities that the database is optimized for, such as
aggregation in an RDBMS.
For example, the previous Transact-SQL code can be replaced with a statement that simply retrieves the data to be
processed.
 SELECT
soh.[SalesOrderNumber] AS [OrderNumber],
soh.[Status] AS [Status],
soh.[OrderDate] AS [OrderDate],
soh.[DueDate] AS [DueDate],
soh.[ShipDate] AS [ShipDate],
soh.[SubTotal] AS [SubTotal],
soh.[TaxAmt] AS [TaxAmt],
soh.[TotalDue] AS [TotalDue],
c.[AccountNumber] AS [AccountNumber],
p.[Title] AS [CustomerTitle],
p.[FirstName] AS [CustomerFirstName],
p.[MiddleName] AS [CustomerMiddleName],
p.[LastName] AS [CustomerLastName],
p.[Suffix] AS [CustomerSuffix],
sod.[OrderQty] AS [Quantity],
sod.[UnitPrice] AS [UnitPrice],
sod.[LineTotal] AS [LineTotal],
sod.[ProductID] AS [ProductId]
FROM [Sales].[SalesOrderHeader] soh
INNER JOIN [Sales].[Customer] c ON soh.[CustomerID] = c.[CustomerID]
INNER JOIN [Person].[Person] p ON c.[PersonID] = p.[BusinessEntityID]
INNER JOIN [Sales].[SalesOrderDetail] sod ON soh.[SalesOrderID] = sod.[SalesOrderID]
WHERE soh.[TerritoryId] = @TerritoryId
AND soh.[SalesOrderId] IN (
SELECT TOP 20 SalesOrderId
FROM [Sales].[SalesOrderHeader] soh
WHERE soh.[TerritoryId] = @TerritoryId
ORDER BY soh.[TotalDue] DESC)
ORDER BY soh.[TotalDue] DESC, sod.[SalesOrderDetailID]

The application then uses the .NET Framework System.Xml.Linq APIs to format the results as XML.

// Create a new SqlCommand to run the Transact-SQL query

using (var command = new SqlCommand(...))
{
command.Parameters.AddWithValue("@TerritoryId", id);

// Run the query and create the initial XML document

using (var reader = await command.ExecuteReaderAsync())
{
var lastOrderNumber = string.Empty;
var doc = new XDocument();
var orders = new XElement("Orders");
doc.Add(orders);

XElement lineItems = null;

// Fetch each row in turn, format the results as XML, and add them to the XML document
while (await reader.ReadAsync())
{
var orderNumber = reader["OrderNumber"].ToString();
if (orderNumber != lastOrderNumber)
{
lastOrderNumber = orderNumber;

var order = new XElement("Order");

orders.Add(order);
var customer = new XElement("Customer");
lineItems = new XElement("OrderLineItems");
order.Add(customer, lineItems);

var orderDate = (DateTime)reader["OrderDate"];

var totalDue = (Decimal)reader["TotalDue"];
var reviewRequired = totalDue > 5000 ? 'Y' : 'N';

order.Add(
 new XAttribute("OrderNumber", orderNumber),
new XAttribute("Status", reader["Status"]),
new XAttribute("ShipDate", reader["ShipDate"]),
... // More attributes, not shown.

var fullName = string.Join(" ",

reader["CustomerTitle"],
reader["CustomerFirstName"],
reader["CustomerMiddleName"],
reader["CustomerLastName"],
reader["CustomerSuffix"]
)
.Replace(" ", " ") //remove double spaces
.Trim()
.ToUpper();

customer.Add(
new XAttribute("AccountNumber", reader["AccountNumber"]),
new XAttribute("FullName", fullName));
}

var productId = (int)reader["ProductID"];

var quantity = (short)reader["Quantity"];
var inventoryCheckRequired = (productId >= 710 && productId <= 720 && quantity >= 5) ? 'Y' : 'N';

lineItems.Add(
new XElement("LineItem",
new XAttribute("Quantity", quantity),
new XAttribute("UnitPrice", ((Decimal)reader["UnitPrice"]).ToString("C")),
new XAttribute("LineTotal", RoundAndFormat(reader["LineTotal"])),
new XAttribute("ProductId", productId),
new XAttribute("InventoryCheckRequired", inventoryCheckRequired)
));
}
// Match the exact formatting of the XML returned from SQL
var xml = doc
.ToString(SaveOptions.DisableFormatting)
.Replace(" />", "/>");
}
}

NOTE
This code is somewhat complex. For a new application, you might prefer to use a serialization library. However, the
assumption here is that the development team is refactoring an existing application, so the method needs to return the
exact same format as the original code.

Considerations
Many database systems are highly optimized to perform certain types of data processing, such as
calculating aggregate values over large datasets. Don't move those types of processing out of the database.
Do not relocate processing if doing so causes the database to transfer far more data over the network. See
the Extraneous Fetching antipattern.
If you move processing to an application tier, that tier may need to scale out to handle the additional work.

How to detect the problem

Symptoms of a busy database include a disproportionate decline in throughput and response times in operations
that access the database.
You can perform the following steps to help identify this problem:
1. Use performance monitoring to identify how much time the production system spends performing
database activity.
2. Examine the work performed by the database during these periods.
3. If you suspect that particular operations might cause too much database activity, perform load testing in a
controlled environment. Each test should run a mixture of the suspect operations with a variable user load.
Examine the telemetry from the load tests to observe how the database is used.
4. If the database activity reveals significant processing but little data traffic, review the source code to
determine whether the processing can better be performed elsewhere.
If the volume of database activity is low or response times are relatively fast, then a busy database is unlikely to be
a performance problem.

Example diagnosis
The following sections apply these steps to the sample application described earlier.
Monitor the volume of database activity
The following graph shows the results of running a load test against the sample application, using a step load of up
to 50 concurrent users. The volume of requests quickly reaches a limit and stays at that level, while the average
response time steadily increases. A logarithmic scale is used for those two metrics.

This line graph shows user load, requests per second, and average response time. The graph shows that response
time increases as load increases.
The next graph shows CPU utilization and DTUs as a percentage of service quota. DTUs provide a measure of how
much processing the database performs. The graph shows that CPU and DTU utilization both quickly reached
100%.
This line graph shows CPU percentage and DTU percentage over time. The graph shows that both quickly reach
100%.
Examine the work performed by the database
It could be that the tasks performed by the database are genuine data access operations, rather than processing, so
it is important to understand the SQL statements being run while the database is busy. Monitor the system to
capture the SQL traffic and correlate the SQL operations with application requests.
If the database operations are purely data access operations, without a lot of processing, then the problem might
be Extraneous Fetching.
Implement the solution and verify the result
The following graph shows a load test using the updated code. Throughput is significantly higher, over 400
requests per second versus 12 earlier. The average response time is also much lower, just above 0.1 seconds
compared to over 4 seconds.
This line graph shows user load, requests per second, and average response time. The graph shows that response
time remains roughly constant throughout the load test.
CPU and DTU utilization shows that the system took longer to reach saturation, despite the increased throughput.

This line graph shows CPU percentage and DTU percentage over time. The graph shows that CPU and DTU take
longer to reach 100% than previously.

Related resources
Extraneous Fetching antipattern
 Busy Front End antipattern
12/18/2020 • 8 minutes to read • Edit Online

Performing asynchronous work on a large number of background threads can starve other concurrent foreground
tasks of resources, decreasing response times to unacceptable levels.

Problem description
Resource-intensive tasks can increase the response times for user requests and cause high latency. One way to
improve response times is to offload a resource-intensive task to a separate thread. This approach lets the
application stay responsive while processing happens in the background. However, tasks that run on a background
thread still consume resources. If there are too many of them, they can starve the threads that are handling
requests.

NOTE
The term resource can encompass many things, such as CPU utilization, memory occupancy, and network or disk I/O.

This problem typically occurs when an application is developed as monolithic piece of code, with all of the business
logic combined into a single tier shared with the presentation layer.
Here's an example using ASP.NET that demonstrates the problem. You can find the complete sample here.

public class WorkInFrontEndController : ApiController

{
[HttpPost]
[Route("api/workinfrontend")]
public HttpResponseMessage Post()
{
new Thread(() =>
{
//Simulate processing
Thread.SpinWait(Int32.MaxValue / 100);
}).Start();

return Request.CreateResponse(HttpStatusCode.Accepted);
}
}

public class UserProfileController : ApiController

{
[HttpGet]
[Route("api/userprofile/{id}")]
public UserProfile Get(int id)
{
//Simulate processing
return new UserProfile() { FirstName = "Alton", LastName = "Hudgens" };
}
}

The Post method in the WorkInFrontEnd controller implements an HTTP POST operation. This operation
simulates a long-running, CPU-intensive task. The work is performed on a separate thread, in an attempt to
enable the POST operation to complete quickly.
The Get method in the UserProfile controller implements an HTTP GET operation. This method is much
 less CPU intensive.
The primary concern is the resource requirements of the Post method. Although it puts the work onto a
background thread, the work can still consume considerable CPU resources. These resources are shared with other
operations being performed by other concurrent users. If a moderate number of users send this request at the
same time, overall performance is likely to suffer, slowing down all operations. Users might experience significant
latency in the Get method, for example.

How to fix the problem

Move processes that consume significant resources to a separate back end.
With this approach, the front end puts resource-intensive tasks onto a message queue. The back end picks up the
tasks for asynchronous processing. The queue also acts as a load leveler, buffering requests for the back end. If the
queue length becomes too long, you can configure autoscaling to scale out the back end.
Here is a revised version of the previous code. In this version, the Post method puts a message on a Service Bus
queue.

public class WorkInBackgroundController : ApiController

{
private static readonly QueueClient QueueClient;
private static readonly string QueueName;
private static readonly ServiceBusQueueHandler ServiceBusQueueHandler;

public WorkInBackgroundController()
{
var serviceBusConnectionString = ...;
QueueName = ...;
ServiceBusQueueHandler = new ServiceBusQueueHandler(serviceBusConnectionString);
QueueClient = ServiceBusQueueHandler.GetQueueClientAsync(QueueName).Result;
}

[HttpPost]
[Route("api/workinbackground")]
public async Task<long> Post()
{
return await ServiceBusQueueHandler.AddWorkLoadToQueueAsync(QueueClient, QueueName, 0);
}
}

The back end pulls messages from the Service Bus queue and does the processing.
 public async Task RunAsync(CancellationToken cancellationToken)
{
this._queueClient.OnMessageAsync(
// This lambda is invoked for each message received.
async (receivedMessage) =>
{
try
{
// Simulate processing of message
Thread.SpinWait(Int32.MaxValue / 1000);

await receivedMessage.CompleteAsync();
}
catch
{
receivedMessage.Abandon();
}
});
}

Considerations
This approach adds some additional complexity to the application. You must handle queuing and dequeuing
safely to avoid losing requests in the event of a failure.
The application takes a dependency on an additional service for the message queue.
The processing environment must be sufficiently scalable to handle the expected workload and meet the
required throughput targets.
While this approach should improve overall responsiveness, the tasks that are moved to the back end may take
longer to complete.

How to detect the problem

Symptoms of a busy front end include high latency when resource-intensive tasks are being performed. End users
are likely to report extended response times or failures caused by services timing out. These failures could also
return HTTP 500 (Internal Server) errors or HTTP 503 (Service Unavailable) errors. Examine the event logs for the
web server, which are likely to contain more detailed information about the causes and circumstances of the errors.
You can perform the following steps to help identify this problem:
1. Perform process monitoring of the production system, to identify points when response times slow down.
2. Examine the telemetry data captured at these points to determine the mix of operations being performed and
the resources being used.
3. Find any correlations between poor response times and the volumes and combinations of operations that were
happening at those times.
4. Load test each suspected operation to identify which operations are consuming resources and starving other
operations.
5. Review the source code for those operations to determine why they might cause excessive resource
consumption.

Example diagnosis
The following sections apply these steps to the sample application described earlier.
Identify points of slowdown
Instrument each method to track the duration and resources consumed by each request. Then monitor the
application in production. This can provide an overall view of how requests compete with each other. During
periods of stress, slow-running resource-hungry requests will likely affect other operations, and this behavior can
be observed by monitoring the system and noting the drop off in performance.
The following image shows a monitoring dashboard. (We used AppDynamics for our tests.) Initially, the system has
light load. Then users start requesting the UserProfile GET method. The performance is reasonably good until
other users start issuing requests to the WorkInFrontEnd POST method. At that point, response times increase
dramatically (first arrow). Response times only improve after the volume of requests to the WorkInFrontEnd
controller diminishes (second arrow).

Examine telemetry data and find correlations

The next image shows some of the metrics gathered to monitor resource utilization during the same interval. At
first, few users are accessing the system. As more users connect, CPU utilization becomes very high (100%). Also
notice that the network I/O rate initially goes up as CPU usage rises. But once CPU usage peaks, network I/O
actually goes down. That's because the system can only handle a relatively small number of requests once the CPU
is at capacity. As users disconnect, the CPU load tails off.

At this point, it appears the Post method in the WorkInFrontEnd controller is a prime candidate for closer
examination. Further work in a controlled environment is needed to confirm the hypothesis.
Perform load testing
The next step is to perform tests in a controlled environment. For example, run a series of load tests that include
and then omit each request in turn to see the effects.
The graph below shows the results of a load test performed against an identical deployment of the cloud service
used in the previous tests. The test used a constant load of 500 users performing the Get operation in the
UserProfile controller, along with a step load of users performing the Post operation in the WorkInFrontEnd
controller.

Initially, the step load is 0, so the only active users are performing the UserProfile requests. The system is able to
respond to approximately 500 requests per second. After 60 seconds, a load of 100 additional users starts sending
POST requests to the WorkInFrontEnd controller. Almost immediately, the workload sent to the UserProfile
controller drops to about 150 requests per second. This is due to the way the load-test runner functions. It waits for
a response before sending the next request, so the longer it takes to receive a response, the lower the request rate.
As more users send POST requests to the WorkInFrontEnd controller, the response rate of the UserProfile
controller continues to drop. But note that the volume of requests handled by the WorkInFrontEnd controller
remains relatively constant. The saturation of the system becomes apparent as the overall rate of both requests
tends toward a steady but low limit.
Review the source code
The final step is to look at the source code. The development team was aware that the Post method could take a
considerable amount of time, which is why the original implementation used a separate thread. That solved the
immediate problem, because the Post method did not block waiting for a long-running task to complete.
However, the work performed by this method still consumes CPU, memory, and other resources. Enabling this
process to run asynchronously might actually damage performance, as users can trigger a large number of these
operations simultaneously, in an uncontrolled manner. There is a limit to the number of threads that a server can
run. Past this limit, the application is likely to get an exception when it tries to start a new thread.
 NOTE
This doesn't mean you should avoid asynchronous operations. Performing an asynchronous await on a network call is a
recommended practice. (See the Synchronous I/O antipattern.) The problem here is that CPU-intensive work was spawned on
another thread.

Implement the solution and verify the result

The following image shows performance monitoring after the solution was implemented. The load was similar to
that shown earlier, but the response times for the UserProfile controller are now much faster. The volume of
requests increased over the same duration, from 2,759 to 23,565.

Note that the WorkInBackground controller also handled a much larger volume of requests. However, you can't
make a direct comparison in this case, because the work being performed in this controller is very different from
the original code. The new version simply queues a request, rather than performing a time consuming calculation.
The main point is that this method no longer drags down the entire system under load.
CPU and network utilization also show the improved performance. The CPU utilization never reached 100%, and
the volume of handled network requests was far greater than earlier, and did not tail off until the workload
dropped.

The following graph shows the results of a load test. The overall volume of requests serviced is greatly improved
compared to the earlier tests.
Related guidance
Autoscaling best practices
Background jobs best practices
Queue-Based Load Leveling pattern
Web Queue Worker architecture style
 Chatty I/O antipattern
12/18/2020 • 9 minutes to read • Edit Online

The cumulative effect of a large number of I/O requests can have a significant impact on performance and
responsiveness.

Problem description
Network calls and other I/O operations are inherently slow compared to compute tasks. Each I/O request typically
has significant overhead, and the cumulative effect of numerous I/O operations can slow down the system. Here
are some common causes of chatty I/O.
Reading and writing individual records to a database as distinct requests
The following example reads from a database of products. There are three tables, Product , ProductSubcategory ,
and ProductPriceListHistory . The code retrieves all of the products in a subcategory, along with the pricing
information, by executing a series of queries:
1. Query the subcategory from the ProductSubcategory table.
2. Find all products in that subcategory by querying the Product table.
3. For each product, query the pricing data from the ProductPriceListHistory table.
The application uses Entity Framework to query the database. You can find the complete sample here.

public async Task<IHttpActionResult> GetProductsInSubCategoryAsync(int subcategoryId)

{
using (var context = GetContext())
{
// Get product subcategory.
var productSubcategory = await context.ProductSubcategories
.Where(psc => psc.ProductSubcategoryId == subcategoryId)
.FirstOrDefaultAsync();

// Find products in that category.

productSubcategory.Product = await context.Products
.Where(p => subcategoryId == p.ProductSubcategoryId)
.ToListAsync();

// Find price history for each product.

foreach (var prod in productSubcategory.Product)
{
int productId = prod.ProductId;
var productListPriceHistory = await context.ProductListPriceHistory
.Where(pl => pl.ProductId == productId)
.ToListAsync();
prod.ProductListPriceHistory = productListPriceHistory;
}
return Ok(productSubcategory);
}
}

This example shows the problem explicitly, but sometimes an O/RM can mask the problem, if it implicitly fetches
child records one at a time. This is known as the "N+1 problem".
Implementing a single logical operation as a series of HTTP requests
This often happens when developers try to follow an object-oriented paradigm, and treat remote objects as if they
were local objects in memory. This can result in too many network round trips. For example, the following web API
exposes the individual properties of User objects through individual HTTP GET methods.

public class UserController : ApiController

{
[HttpGet]
[Route("users/{id:int}/username")]
public HttpResponseMessage GetUserName(int id)
{
...
}

[HttpGet]
[Route("users/{id:int}/gender")]
public HttpResponseMessage GetGender(int id)
{
...
}

[HttpGet]
[Route("users/{id:int}/dateofbirth")]
public HttpResponseMessage GetDateOfBirth(int id)
{
...
}
}

While there's nothing technically wrong with this approach, most clients will probably need to get several
properties for each User , resulting in client code like the following.

HttpResponseMessage response = await client.GetAsync("users/1/username");

response.EnsureSuccessStatusCode();
var userName = await response.Content.ReadAsStringAsync();

response = await client.GetAsync("users/1/gender");

response.EnsureSuccessStatusCode();
var gender = await response.Content.ReadAsStringAsync();

response = await client.GetAsync("users/1/dateofbirth");

response.EnsureSuccessStatusCode();
var dob = await response.Content.ReadAsStringAsync();

Reading and writing to a file on disk

File I/O involves opening a file and moving to the appropriate point before reading or writing data. When the
operation is complete, the file might be closed to save operating system resources. An application that continually
reads and writes small amounts of information to a file will generate significant I/O overhead. Small write requests
can also lead to file fragmentation, slowing subsequent I/O operations still further.
The following example uses a FileStream to write a Customer object to a file. Creating the FileStream opens the
file, and disposing it closes the file. (The using statement automatically disposes the FileStream object.) If the
application calls this method repeatedly as new customers are added, the I/O overhead can accumulate quickly.
 private async Task SaveCustomerToFileAsync(Customer customer)
{
using (Stream fileStream = new FileStream(CustomersFileName, FileMode.Append))
{
BinaryFormatter formatter = new BinaryFormatter();
byte [] data = null;
using (MemoryStream memStream = new MemoryStream())
{
formatter.Serialize(memStream, customer);
data = memStream.ToArray();
}
await fileStream.WriteAsync(data, 0, data.Length);
}
}

How to fix the problem

Reduce the number of I/O requests by packaging the data into larger, fewer requests.
Fetch data from a database as a single query, instead of several smaller queries. Here's a revised version of the
code that retrieves product information.

public async Task<IHttpActionResult> GetProductCategoryDetailsAsync(int subCategoryId)

{
using (var context = GetContext())
{
var subCategory = await context.ProductSubcategories
.Where(psc => psc.ProductSubcategoryId == subCategoryId)
.Include("Product.ProductListPriceHistory")
.FirstOrDefaultAsync();

if (subCategory == null)
return NotFound();

return Ok(subCategory);
}
}

Follow REST design principles for web APIs. Here's a revised version of the web API from the earlier example.
Instead of separate GET methods for each property, there is a single GET method that returns the User . This
results in a larger response body per request, but each client is likely to make fewer API calls.

public class UserController : ApiController

{
[HttpGet]
[Route("users/{id:int}")]
public HttpResponseMessage GetUser(int id)
{
...
}
}

// Client code
HttpResponseMessage response = await client.GetAsync("users/1");
response.EnsureSuccessStatusCode();
var user = await response.Content.ReadAsStringAsync();

For file I/O, consider buffering data in memory and then writing the buffered data to a file as a single operation.
This approach reduces the overhead from repeatedly opening and closing the file, and helps to reduce
fragmentation of the file on disk.
 // Save a list of customer objects to a file
private async Task SaveCustomerListToFileAsync(List<Customer> customers)
{
using (Stream fileStream = new FileStream(CustomersFileName, FileMode.Append))
{
BinaryFormatter formatter = new BinaryFormatter();
foreach (var customer in customers)
{
byte[] data = null;
using (MemoryStream memStream = new MemoryStream())
{
formatter.Serialize(memStream, customer);
data = memStream.ToArray();
}
await fileStream.WriteAsync(data, 0, data.Length);
}
}
}

// In-memory buffer for customers.

List<Customer> customers = new List<Customers>();

// Create a new customer and add it to the buffer

var customer = new Customer(...);
customers.Add(customer);

// Add more customers to the list as they are created

...

// Save the contents of the list, writing all customers in a single operation
await SaveCustomerListToFileAsync(customers);

Considerations
The first two examples make fewer I/O calls, but each one retrieves more information. You must consider
the tradeoff between these two factors. The right answer will depend on the actual usage patterns. For
example, in the web API example, it might turn out that clients often need just the user name. In that case, it
might make sense to expose it as a separate API call. For more information, see the Extraneous Fetching
antipattern.
When reading data, do not make your I/O requests too large. An application should only retrieve the
information that it is likely to use.
Sometimes it helps to partition the information for an object into two chunks, frequently accessed data that
accounts for most requests, and less frequently accessed data that is used rarely. Often the most frequently
accessed data is a relatively small portion of the total data for an object, so returning just that portion can
save significant I/O overhead.
When writing data, avoid locking resources for longer than necessary, to reduce the chances of contention
during a lengthy operation. If a write operation spans multiple data stores, files, or services, then adopt an
eventually consistent approach. See Data Consistency guidance.
If you buffer data in memory before writing it, the data is vulnerable if the process crashes. If the data rate
typically has bursts or is relatively sparse, it may be safer to buffer the data in an external durable queue
such as Event Hubs.
Consider caching data that you retrieve from a service or a database. This can help to reduce the volume of
I/O by avoiding repeated requests for the same data. For more information, see Caching best practices.

How to detect the problem

Symptoms of chatty I/O include high latency and low throughput. End users are likely to report extended response
times or failures caused by services timing out, due to increased contention for I/O resources.
You can perform the following steps to help identify the causes of any problems:
1. Perform process monitoring of the production system to identify operations with poor response times.
2. Perform load testing of each operation identified in the previous step.
3. During the load tests, gather telemetry data about the data access requests made by each operation.
4. Gather detailed statistics for each request sent to a data store.
5. Profile the application in the test environment to establish where possible I/O bottlenecks might be occurring.
Look for any of these symptoms:
A large number of small I/O requests made to the same file.
A large number of small network requests made by an application instance to the same service.
A large number of small requests made by an application instance to the same data store.
Applications and services becoming I/O bound.

Example diagnosis
The following sections apply these steps to the example shown earlier that queries a database.
Load test the application
This graph shows the results of load testing. Median response time is measured in tens of seconds per request.
The graph shows very high latency. With a load of 1000 users, a user might have to wait for nearly a minute to see
the results of a query.

NOTE
The application was deployed as an Azure App Service web app, using Azure SQL Database. The load test used a simulated
step workload of up to 1000 concurrent users. The database was configured with a connection pool supporting up to 1000
concurrent connections, to reduce the chance that contention for connections would affect the results.

Monitor the application

You can use an application performance monitoring (APM) package to capture and analyze the key metrics that
might identify chatty I/O. Which metrics are important will depend on the I/O workload. For this example, the
interesting I/O requests were the database queries.
The following image shows results generated using New Relic APM. The average database response time peaked
at approximately 5.6 seconds per request during the maximum workload. The system was able to support an
average of 410 requests per minute throughout the test.

Gather detailed data access information

Digging deeper into the monitoring data shows the application executes three different SQL SELECT statements.
These correspond to the requests generated by Entity Framework to fetch data from the ProductListPriceHistory ,
Product , and ProductSubcategory tables. Furthermore, the query that retrieves data from the
ProductListPriceHistory table is by far the most frequently executed SELECT statement, by an order of magnitude.
It turns out that the GetProductsInSubCategoryAsync method, shown earlier, performs 45 SELECT queries. Each
query causes the application to open a new SQL connection.
 NOTE
This image shows trace information for the slowest instance of the GetProductsInSubCategoryAsync operation in the load
test. In a production environment, it's useful to examine traces of the slowest instances, to see if there is a pattern that
suggests a problem. If you just look at the average values, you might overlook problems that will get dramatically worse
under load.

The next image shows the actual SQL statements that were issued. The query that fetches price information is run
for each individual product in the product subcategory. Using a join would considerably reduce the number of
database calls.

If you are using an O/RM, such as Entity Framework, tracing the SQL queries can provide insight into how the
O/RM translates programmatic calls into SQL statements, and indicate areas where data access might be
optimized.
Implement the solution and verify the result
Rewriting the call to Entity Framework produced the following results.
This load test was performed on the same deployment, using the same load profile. This time the graph shows
much lower latency. The average request time at 1000 users is between 5 and 6 seconds, down from nearly a
minute.
This time the system supported an average of 3,970 requests per minute, compared to 410 for the earlier test.

Tracing the SQL statement shows that all the data is fetched in a single SELECT statement. Although this query is
considerably more complex, it is performed only once per operation. And while complex joins can become
expensive, relational database systems are optimized for this type of query.
Related resources
API Design best practices
Caching best practices
Data Consistency Primer
Extraneous Fetching antipattern
No Caching antipattern
 Extraneous Fetching antipattern
12/18/2020 • 10 minutes to read • Edit Online

Retrieving more data than needed for a business operation can result in unnecessary I/O overhead and reduce
responsiveness.

Problem description
This antipattern can occur if the application tries to minimize I/O requests by retrieving all of the data that it
might need. This is often a result of overcompensating for the Chatty I/O antipattern. For example, an application
might fetch the details for every product in a database. But the user may need just a subset of the details (some
may not be relevant to customers), and probably doesn't need to see all of the products at once. Even if the user
is browsing the entire catalog, it would make sense to paginate the results—showing 20 at a time, for example.
Another source of this problem is following poor programming or design practices. For example, the following
code uses Entity Framework to fetch the complete details for every product. Then it filters the results to return
only a subset of the fields, discarding the rest. You can find the complete sample here.

public async Task<IHttpActionResult> GetAllFieldsAsync()

{
using (var context = new AdventureWorksContext())
{
// Execute the query. This happens at the database.
var products = await context.Products.ToListAsync();

// Project fields from the query results. This happens in application memory.
var result = products.Select(p => new ProductInfo { Id = p.ProductId, Name = p.Name });
return Ok(result);
}
}

In the next example, the application retrieves data to perform an aggregation that could be done by the database
instead. The application calculates total sales by getting every record for all orders sold, and then computing the
sum over those records. You can find the complete sample here.

public async Task<IHttpActionResult> AggregateOnClientAsync()

{
using (var context = new AdventureWorksContext())
{
// Fetch all order totals from the database.
var orderAmounts = await context.SalesOrderHeaders.Select(soh => soh.TotalDue).ToListAsync();

// Sum the order totals in memory.

var total = orderAmounts.Sum();
return Ok(total);
}
}

The next example shows a subtle problem caused by the way Entity Framework uses LINQ to Entities.
 var query = from p in context.Products.AsEnumerable()
where p.SellStartDate < DateTime.Now.AddDays(-7) // AddDays cannot be mapped by LINQ to Entities
select ...;

List<Product> products = query.ToList();

The application is trying to find products with a SellStartDate more than a week old. In most cases, LINQ to
Entities would translate a where clause to a SQL statement that is executed by the database. In this case, however,
LINQ to Entities cannot map the AddDays method to SQL. Instead, every row from the Product table is returned,
and the results are filtered in memory.
The call to AsEnumerable is a hint that there is a problem. This method converts the results to an IEnumerable
interface. Although IEnumerable supports filtering, the filtering is done on the client side, not the database. By
default, LINQ to Entities uses IQueryable , which passes the responsibility for filtering to the data source.

How to fix the problem

Avoid fetching large volumes of data that may quickly become outdated or might be discarded, and only fetch
the data needed for the operation being performed.
Instead of getting every column from a table and then filtering them, select the columns that you need from the
database.

public async Task<IHttpActionResult> GetRequiredFieldsAsync()

{
using (var context = new AdventureWorksContext())
{
// Project fields as part of the query itself
var result = await context.Products
.Select(p => new ProductInfo {Id = p.ProductId, Name = p.Name})
.ToListAsync();
return Ok(result);
}
}

Similarly, perform aggregation in the database and not in application memory.

public async Task<IHttpActionResult> AggregateOnDatabaseAsync()

{
using (var context = new AdventureWorksContext())
{
// Sum the order totals as part of the database query.
var total = await context.SalesOrderHeaders.SumAsync(soh => soh.TotalDue);
return Ok(total);
}
}

When using Entity Framework, ensure that LINQ queries are resolved using the IQueryable interface and not
IEnumerable . You may need to adjust the query to use only functions that can be mapped to the data source. The
earlier example can be refactored to remove the AddDays method from the query, allowing filtering to be done
by the database.
 DateTime dateSince = DateTime.Now.AddDays(-7); // AddDays has been factored out.
var query = from p in context.Products
where p.SellStartDate < dateSince // This criterion can be passed to the database by LINQ to
Entities
select ...;

List<Product> products = query.ToList();

Considerations
In some cases, you can improve performance by partitioning data horizontally. If different operations
access different attributes of the data, horizontal partitioning may reduce contention. Often, most
operations are run against a small subset of the data, so spreading this load may improve performance.
See Data partitioning.
For operations that have to support unbounded queries, implement pagination and only fetch a limited
number of entities at a time. For example, if a customer is browsing a product catalog, you can show one
page of results at a time.
When possible, take advantage of features built into the data store. For example, SQL databases typically
provide aggregate functions.
If you're using a data store that doesn't support a particular function, such as aggregation, you could store
the calculated result elsewhere, updating the value as records are added or updated, so the application
doesn't have to recalculate the value each time it's needed.
If you see that requests are retrieving a large number of fields, examine the source code to determine
whether all of these fields are necessary. Sometimes these requests are the result of poorly designed
SELECT * query.

Similarly, requests that retrieve a large number of entities may be sign that the application is not filtering
data correctly. Verify that all of these entities are needed. Use database-side filtering if possible, for
example, by using WHERE clauses in SQL.
Offloading processing to the database is not always the best option. Only use this strategy when the
database is designed or optimized to do so. Most database systems are highly optimized for certain
functions, but are not designed to act as general-purpose application engines. For more information, see
the Busy Database antipattern.

How to detect the problem

Symptoms of extraneous fetching include high latency and low throughput. If the data is retrieved from a data
store, increased contention is also probable. End users are likely to report extended response times or failures
caused by services timing out. These failures could return HTTP 500 (Internal Server) errors or HTTP 503 (Service
Unavailable) errors. Examine the event logs for the web server, which likely contain more detailed information
about the causes and circumstances of the errors.
The symptoms of this antipattern and some of the telemetry obtained might be very similar to those of the
Monolithic Persistence antipattern.
You can perform the following steps to help identify the cause:
1. Identify slow workloads or transactions by performing load-testing, process monitoring, or other methods of
capturing instrumentation data.
2. Observe any behavioral patterns exhibited by the system. Are there particular limits in terms of transactions
per second or volume of users?
3. Correlate the instances of slow workloads with behavioral patterns.
4. Identify the data stores being used. For each data source, run lower-level telemetry to observe the behavior of
operations.
5. Identify any slow-running queries that reference these data sources.
6. Perform a resource-specific analysis of the slow-running queries and ascertain how the data is used and
consumed.
Look for any of these symptoms:
Frequent, large I/O requests made to the same resource or data store.
Contention in a shared resource or data store.
An operation that frequently receives large volumes of data over the network.
Applications and services spending significant time waiting for I/O to complete.

Example diagnosis
The following sections apply these steps to the previous examples.
Identify slow workloads
This graph shows performance results from a load test that simulated up to 400 concurrent users running the
GetAllFieldsAsync method shown earlier. Throughput diminishes slowly as the load increases. Average response
time goes up as the workload increases.

A load test for the AggregateOnClientAsync operation shows a similar pattern. The volume of requests is
reasonably stable. The average response time increases with the workload, although more slowly than the
previous graph.
Correlate slow workloads with behavioral patterns
Any correlation between regular periods of high usage and slowing performance can indicate areas of concern.
Closely examine the performance profile of functionality that is suspected to be slow running, to determine
whether it matches the load testing performed earlier.
Load test the same functionality using step-based user loads, to find the point where performance drops
significantly or fails completely. If that point falls within the bounds of your expected real-world usage, examine
how the functionality is implemented.
A slow operation is not necessarily a problem, if it is not being performed when the system is under stress, is not
time critical, and does not negatively affect the performance of other important operations. For example,
generating monthly operational statistics might be a long-running operation, but it can probably be performed
as a batch process and run as a low-priority job. On the other hand, customers querying the product catalog is a
critical business operation. Focus on the telemetry generated by these critical operations to see how the
performance varies during periods of high usage.
Identify data sources in slow workloads
If you suspect that a service is performing poorly because of the way it retrieves data, investigate how the
application interacts with the repositories it uses. Monitor the live system to see which sources are accessed
during periods of poor performance.
For each data source, instrument the system to capture the following:
The frequency that each data store is accessed.
The volume of data entering and exiting the data store.
The timing of these operations, especially the latency of requests.
The nature and rate of any errors that occur while accessing each data store under typical load.
Compare this information against the volume of data being returned by the application to the client. Track the
ratio of the volume of data returned by the data store against the volume of data returned to the client. If there is
any large disparity, investigate to determine whether the application is fetching data that it doesn't need.
You may be able to capture this data by observing the live system and tracing the lifecycle of each user request,
or you can model a series of synthetic workloads and run them against a test system.
The following graphs show telemetry captured using New Relic APM during a load test of the GetAllFieldsAsync
method. Note the difference between the volumes of data received from the database and the corresponding
HTTP responses.

For each request, the database returned 80,503 bytes, but the response to the client only contained 19,855 bytes,
about 25% of the size of the database response. The size of the data returned to the client can vary depending on
the format. For this load test, the client requested JSON data. Separate testing using XML (not shown) had a
response size of 35,655 bytes, or 44% of the size of the database response.
The load test for the AggregateOnClientAsync method shows more extreme results. In this case, each test
performed a query that retrieved over 280 Kb of data from the database, but the JSON response was a mere 14
bytes. The wide disparity is because the method calculates an aggregated result from a large volume of data.
Identify and analyze slow queries
Look for database queries that consume the most resources and take the most time to execute. You can add
instrumentation to find the start and completion times for many database operations. Many data stores also
provide in-depth information on how queries are performed and optimized. For example, the Query Performance
pane in the Azure SQL Database management portal lets you select a query and view detailed runtime
performance information. Here is the query generated by the GetAllFieldsAsync operation:
Implement the solution and verify the result
After changing the GetRequiredFieldsAsync method to use a SELECT statement on the database side, load testing
showed the following results.
This load test used the same deployment and the same simulated workload of 400 concurrent users as before.
The graph shows much lower latency. Response time rises with load to approximately 1.3 seconds, compared to 4
seconds in the previous case. The throughput is also higher at 350 requests per second compared to 100 earlier.
The volume of data retrieved from the database now closely matches the size of the HTTP response messages.

Load testing using the AggregateOnDatabaseAsync method generates the following results:
The average response time is now minimal. This is an order of magnitude improvement in performance, caused
primarily by the large reduction in I/O from the database.
Here is the corresponding telemetry for the AggregateOnDatabaseAsync method. The amount of data retrieved
from the database was vastly reduced, from over 280 Kb per transaction to 53 bytes. As a result, the maximum
sustained number of requests per minute was raised from around 2,000 to over 25,000.
Related resources
Busy Database antipattern
Chatty I/O antipattern
Data partitioning best practices
 Improper Instantiation antipattern
12/18/2020 • 6 minutes to read • Edit Online

It can hurt performance to continually create new instances of an object that is meant to be created once and then
shared.

Problem description
Many libraries provide abstractions of external resources. Internally, these classes typically manage their own
connections to the resource, acting as brokers that clients can use to access the resource. Here are some examples
of broker classes that are relevant to Azure applications:
System.Net.Http.HttpClient . Communicates with a web service using HTTP.
Microsoft.ServiceBus.Messaging.QueueClient . Posts and receives messages to a Service Bus queue.
Microsoft.Azure.Documents.Client.DocumentClient . Connects to a Cosmos DB instance.
StackExchange.Redis.ConnectionMultiplexer . Connects to Redis, including Azure Cache for Redis.

These classes are intended to be instantiated once and reused throughout the lifetime of an application. However,
it's a common misunderstanding that these classes should be acquired only as necessary and released quickly. (The
ones listed here happen to be .NET libraries, but the pattern is not unique to .NET.) The following ASP.NET example
creates an instance of HttpClient to communicate with a remote service. You can find the complete sample here.

public class NewHttpClientInstancePerRequestController : ApiController

{
// This method creates a new instance of HttpClient and disposes it for every call to GetProductAsync.
public async Task<Product> GetProductAsync(string id)
{
using (var httpClient = new HttpClient())
{
var hostName = HttpContext.Current.Request.Url.Host;
var result = await httpClient.GetStringAsync(string.Format("http://{0}:8080/api/...", hostName));
return new Product { Name = result };
}
}
}

In a web application, this technique is not scalable. A new HttpClient object is created for each user request. Under
heavy load, the web server may exhaust the number of available sockets, resulting in SocketException errors.
This problem is not restricted to the HttpClient class. Other classes that wrap resources or are expensive to create
might cause similar issues. The following example creates an instance of the ExpensiveToCreateService class. Here
the issue is not necessarily socket exhaustion, but simply how long it takes to create each instance. Continually
creating and destroying instances of this class might adversely affect the scalability of the system.
 public class NewServiceInstancePerRequestController : ApiController
{
public async Task<Product> GetProductAsync(string id)
{
var expensiveToCreateService = new ExpensiveToCreateService();
return await expensiveToCreateService.GetProductByIdAsync(id);
}
}

public class ExpensiveToCreateService

{
public ExpensiveToCreateService()
{
// Simulate delay due to setup and configuration of ExpensiveToCreateService
Thread.SpinWait(Int32.MaxValue / 100);
}
...
}

How to fix the problem

If the class that wraps the external resource is shareable and thread-safe, create a shared singleton instance or a
pool of reusable instances of the class.
The following example uses a static HttpClient instance, thus sharing the connection across all requests.

public class SingleHttpClientInstanceController : ApiController

{
private static readonly HttpClient httpClient;

static SingleHttpClientInstanceController()
{
httpClient = new HttpClient();
}

// This method uses the shared instance of HttpClient for every call to GetProductAsync.
public async Task<Product> GetProductAsync(string id)
{
var hostName = HttpContext.Current.Request.Url.Host;
var result = await httpClient.GetStringAsync(string.Format("http://{0}:8080/api/...", hostName));
return new Product { Name = result };
}
}

Considerations
The key element of this antipattern is repeatedly creating and destroying instances of a shareable object. If a
class is not shareable (not thread-safe), then this antipattern does not apply.
The type of shared resource might dictate whether you should use a singleton or create a pool. The
HttpClient class is designed to be shared rather than pooled. Other objects might support pooling,
enabling the system to spread the workload across multiple instances.
Objects that you share across multiple requests must be thread-safe. The HttpClient class is designed to be
used in this manner, but other classes might not support concurrent requests, so check the available
documentation.
Be careful about setting properties on shared objects, as this can lead to race conditions. For example,
setting DefaultRequestHeaders on the HttpClient class before each request can create a race condition. Set
such properties once (for example, during startup), and create separate instances if you need to configure
 different settings.
Some resource types are scarce and should not be held onto. Database connections are an example. Holding
an open database connection that is not required may prevent other concurrent users from gaining access
to the database.
In the .NET Framework, many objects that establish connections to external resources are created by using
static factory methods of other classes that manage these connections. These objects are intended to be
saved and reused, rather than disposed and re-created. For example, in Azure Service Bus, the QueueClient
object is created through a MessagingFactory object. Internally, the MessagingFactory manages connections.
For more information, see Best Practices for performance improvements using Service Bus Messaging.

How to detect the problem

Symptoms of this problem include a drop in throughput or an increased error rate, along with one or more of the
following:
An increase in exceptions that indicate exhaustion of resources such as sockets, database connections, file
handles, and so on.
Increased memory use and garbage collection.
An increase in network, disk, or database activity.
You can perform the following steps to help identify this problem:
1. Performing process monitoring of the production system, to identify points when response times slow down or
the system fails due to lack of resources.
2. Examine the telemetry data captured at these points to determine which operations might be creating and
destroying resource-consuming objects.
3. Load test each suspected operation, in a controlled test environment rather than the production system.
4. Review the source code and examine the how broker objects are managed.
Look at stack traces for operations that are slow-running or that generate exceptions when the system is under
load. This information can help to identify how these operations are using resources. Exceptions can help to
determine whether errors are caused by shared resources being exhausted.

Example diagnosis
The following sections apply these steps to the sample application described earlier.
Identify points of slowdown or failure
The following image shows results generated using New Relic APM, showing operations that have a poor response
time. In this case, the GetProductAsync method in the NewHttpClientInstancePerRequest controller is worth
investigating further. Notice that the error rate also increases when these operations are running.
Examine telemetry data and find correlations
The next image shows data captured using thread profiling, over the same period corresponding as the previous
image. The system spends a significant time opening socket connections, and even more time closing them and
handling socket exceptions.

Performing load testing

Use load testing to simulate the typical operations that users might perform. This can help to identify which parts
of a system suffer from resource exhaustion under varying loads. Perform these tests in a controlled environment
rather than the production system. The following graph shows the throughput of requests handled by the
NewHttpClientInstancePerRequest controller as the user load increases to 100 concurrent users.
At first, the volume of requests handled per second increases as the workload increases. At about 30 users,
however, the volume of successful requests reaches a limit, and the system starts to generate exceptions. From
then on, the volume of exceptions gradually increases with the user load.
The load test reported these failures as HTTP 500 (Internal Server) errors. Reviewing the telemetry showed that
these errors were caused by the system running out of socket resources, as more and more HttpClient objects
were created.
The next graph shows a similar test for a controller that creates the custom ExpensiveToCreateService object.
This time, the controller does not generate any exceptions, but throughput still reaches a plateau, while the average
response time increases by a factor of 20. (The graph uses a logarithmic scale for response time and throughput.)
Telemetry showed that creating new instances of the ExpensiveToCreateService was the main cause of the problem.
Implement the solution and verify the result
After switching the GetProductAsync method to share a single HttpClient instance, a second load test showed
improved performance. No errors were reported, and the system was able to handle an increasing load of up to
500 requests per second. The average response time was cut in half, compared with the previous test.

For comparison, the following image shows the stack trace telemetry. This time, the system spends most of its time
performing real work, rather than opening and closing sockets.

The next graph shows a similar load test using a shared instance of the ExpensiveToCreateService object. Again, the
volume of handled requests increases in line with the user load, while the average response time remains low.
 Monolithic Persistence antipattern
12/18/2020 • 6 minutes to read • Edit Online

Putting all of an application's data into a single data store can hurt performance, either because it leads to resource
contention, or because the data store is not a good fit for some of the data.

Problem description
Historically, applications have often used a single data store, regardless of the different types of data that the
application might need to store. Usually this was done to simplify the application design, or else to match the
existing skill set of the development team.
Modern cloud-based systems often have additional functional and nonfunctional requirements, and need to store
many heterogeneous types of data, such as documents, images, cached data, queued messages, application logs,
and telemetry. Following the traditional approach and putting all of this information into the same data store can
hurt performance, for two main reasons:
Storing and retrieving large amounts of unrelated data in the same data store can cause contention, which in
turn leads to slow response times and connection failures.
Whichever data store is chosen, it might not be the best fit for all of the different types of data, or it might not
be optimized for the operations that the application performs.
The following example shows an ASP.NET Web API controller that adds a new record to a database and also
records the result to a log. The log is held in the same database as the business data. You can find the complete
sample here.

public class MonoController : ApiController

{
private static readonly string ProductionDb = ...;

public async Task<IHttpActionResult> PostAsync([FromBody]string value)

{
await DataAccess.InsertPurchaseOrderHeaderAsync(ProductionDb);
await DataAccess.LogAsync(ProductionDb, LogTableName);
return Ok();
}
}

The rate at which log records are generated will probably affect the performance of the business operations. And if
another component, such as an application process monitor, regularly reads and processes the log data, that can
also affect the business operations.

How to fix the problem

Separate data according to its use. For each data set, select a data store that best matches how that data set will be
used. In the previous example, the application should be logging to a separate store from the database that holds
business data:
 public class PolyController : ApiController
{
private static readonly string ProductionDb = ...;
private static readonly string LogDb = ...;

public async Task<IHttpActionResult> PostAsync([FromBody]string value)

{
await DataAccess.InsertPurchaseOrderHeaderAsync(ProductionDb);
// Log to a different data store.
await DataAccess.LogAsync(LogDb, LogTableName);
return Ok();
}
}

Considerations
Separate data by the way it is used and how it is accessed. For example, don't store log information and
business data in the same data store. These types of data have significantly different requirements and
patterns of access. Log records are inherently sequential, while business data is more likely to require
random access, and is often relational.
Consider the data access pattern for each type of data. For example, store formatted reports and documents
in a document database such as Cosmos DB, but use Azure Cache for Redis to cache temporary data.
If you follow this guidance but still reach the limits of the database, you may need to scale up the database.
Also consider scaling horizontally and partitioning the load across database servers. However, partitioning
may require redesigning the application. For more information, see Data partitioning.

How to detect the problem

The system will likely slow down dramatically and eventually fail, as the system runs out of resources such as
database connections.
You can perform the following steps to help identify the cause.
1. Instrument the system to record the key performance statistics. Capture timing information for each operation,
as well as the points where the application reads and writes data.
2. If possible, monitor the system running for a few days in a production environment to get a real-world view of
how the system is used. If this is not possible, run scripted load tests with a realistic volume of virtual users
performing a typical series of operations.
3. Use the telemetry data to identify periods of poor performance.
4. Identify which data stores were accessed during those periods.
5. Identify data storage resources that might be experiencing contention.

Example diagnosis
The following sections apply these steps to the sample application described earlier.
Instrument and monitor the system
The following graph shows the results of load testing the sample application described earlier. The test used a step
load of up to 1000 concurrent users.
As the load increases to 700 users, so does the throughput. But at that point, throughput levels off, and the system
appears to be running at its maximum capacity. The average response gradually increases with user load, showing
that the system can't keep up with demand.
Identify periods of poor performance
If you are monitoring the production system, you might notice patterns. For example, response times might drop
off significantly at the same time each day. This could be caused by a regular workload or scheduled batch job, or
just because the system has more users at certain times. You should focus on the telemetry data for these events.
Look for correlations between increased response times and increased database activity or I/O to shared resources.
If there are correlations, it means the database might be a bottleneck.
Identify which data stores are accessed during those periods
The next graph shows the utilization of database throughput units (DTU) during the load test. (A DTU is a measure
of available capacity, and is a combination of CPU utilization, memory allocation, I/O rate.) Utilization of DTUs
quickly reached 100%. This is roughly the point where throughput peaked in the previous graph. Database
utilization remained very high until the test finished. There is a slight drop toward the end, which could be caused
by throttling, competition for database connections, or other factors.
Examine the telemetry for the data stores
Instrument the data stores to capture the low-level details of the activity. In the sample application, the data access
statistics showed a high volume of insert operations performed against both the PurchaseOrderHeader table and
the MonoLog table.

Identify resource contention

At this point, you can review the source code, focusing on the points where contended resources are accessed by
the application. Look for situations such as:
Data that is logically separate being written to the same store. Data such as logs, reports, and queued messages
should not be held in the same database as business information.
A mismatch between the choice of data store and the type of data, such as large blobs or XML documents in a
relational database.
Data with significantly different usage patterns that share the same store, such as high-write/low-read data
being stored with low-write/high-read data.
Implement the solution and verify the result
The application was changed to write logs to a separate data store. Here are the load test results:

The pattern of throughput is similar to the earlier graph, but the point at which performance peaks is
approximately 500 requests per second higher. The average response time is marginally lower. However, these
statistics don't tell the full story. Telemetry for the business database shows that DTU utilization peaks at around
75%, rather than 100%.

Similarly, the maximum DTU utilization of the log database only reaches about 70%. The databases are no longer
the limiting factor in the performance of the system.
Related resources
Choose the right data store
Criteria for choosing a data store
Data Access for Highly Scalable Solutions: Using SQL, NoSQL, and Polyglot Persistence
Data partitioning
 No Caching antipattern
12/18/2020 • 8 minutes to read • Edit Online

In a cloud application that handles many concurrent requests, repeatedly fetching the same data can reduce
performance and scalability.

Problem description
When data is not cached, it can cause a number of undesirable behaviors, including:
Repeatedly fetching the same information from a resource that is expensive to access, in terms of I/O overhead
or latency.
Repeatedly constructing the same objects or data structures for multiple requests.
Making excessive calls to a remote service that has a service quota and throttles clients past a certain limit.
In turn, these problems can lead to poor response times, increased contention in the data store, and poor
scalability.
The following example uses Entity Framework to connect to a database. Every client request results in a call to the
database, even if multiple requests are fetching exactly the same data. The cost of repeated requests, in terms of
I/O overhead and data access charges, can accumulate quickly.

public class PersonRepository : IPersonRepository

{
public async Task<Person> GetAsync(int id)
{
using (var context = new AdventureWorksContext())
{
return await context.People
.Where(p => p.Id == id)
.FirstOrDefaultAsync()
.ConfigureAwait(false);
}
}
}

You can find the complete sample here.

This antipattern typically occurs because:
Not using a cache is simpler to implement, and it works fine under low loads. Caching makes the code more
complicated.
The benefits and drawbacks of using a cache are not clearly understood.
There is concern about the overhead of maintaining the accuracy and freshness of cached data.
An application was migrated from an on-premises system, where network latency was not an issue, and the
system ran on expensive high-performance hardware, so caching wasn't considered in the original design.
Developers aren't aware that caching is a possibility in a given scenario. For example, developers may not think
of using ETags when implementing a web API.

How to fix the problem

The most popular caching strategy is the on-demand or cache-aside strategy.
 On read, the application tries to read the data from the cache. If the data isn't in the cache, the application
retrieves it from the data source and adds it to the cache.
On write, the application writes the change directly to the data source and removes the old value from the
cache. It will be retrieved and added to the cache the next time it is required.
This approach is suitable for data that changes frequently. Here is the previous example updated to use the Cache-
Aside pattern.

public class CachedPersonRepository : IPersonRepository

{
private readonly PersonRepository _innerRepository;

public CachedPersonRepository(PersonRepository innerRepository)

{
_innerRepository = innerRepository;
}

public async Task<Person> GetAsync(int id)

{
return await CacheService.GetAsync<Person>("p:" + id, () =>
_innerRepository.GetAsync(id)).ConfigureAwait(false);
}
}

public class CacheService

{
private static ConnectionMultiplexer _connection;

public static async Task<T> GetAsync<T>(string key, Func<Task<T>> loadCache, double

expirationTimeInMinutes)
{
IDatabase cache = Connection.GetDatabase();
T value = await GetAsync<T>(cache, key).ConfigureAwait(false);
if (value == null)
{
// Value was not found in the cache. Call the lambda to get the value from the database.
value = await loadCache().ConfigureAwait(false);
if (value != null)
{
// Add the value to the cache.
await SetAsync(cache, key, value, expirationTimeInMinutes).ConfigureAwait(false);
}
}
return value;
}
}

Notice that the GetAsync method now calls the CacheService class, rather than calling the database directly. The
CacheService class first tries to get the item from Azure Cache for Redis. If the value isn't found in the cache, the
CacheService invokes a lambda function that was passed to it by the caller. The lambda function is responsible for
fetching the data from the database. This implementation decouples the repository from the particular caching
solution, and decouples the CacheService from the database.

Considerations
If the cache is unavailable, perhaps because of a transient failure, don't return an error to the client. Instead,
fetch the data from the original data source. However, be aware that while the cache is being recovered, the
original data store could be swamped with requests, resulting in timeouts and failed connections. (After all,
this is one of the motivations for using a cache in the first place.) Use a technique such as the Circuit Breaker
pattern to avoid overwhelming the data source.
 Applications that cache dynamic data should be designed to support eventual consistency.
For web APIs, you can support client-side caching by including a Cache-Control header in request and
response messages, and using ETags to identify versions of objects. For more information, see API
implementation.
You don't have to cache entire entities. If most of an entity is static but only a small piece changes frequently,
cache the static elements and retrieve the dynamic elements from the data source. This approach can help
to reduce the volume of I/O being performed against the data source.
In some cases, if volatile data is short-lived, it can be useful to cache it. For example, consider a device that
continually sends status updates. It might make sense to cache this information as it arrives, and not write it
to a persistent store at all.
To prevent data from becoming stale, many caching solutions support configurable expiration periods, so
that data is automatically removed from the cache after a specified interval. You may need to tune the
expiration time for your scenario. Data that is highly static can stay in the cache for longer periods than
volatile data that may become stale quickly.
If the caching solution doesn't provide built-in expiration, you may need to implement a background
process that occasionally sweeps the cache, to prevent it from growing without limits.
Besides caching data from an external data source, you can use caching to save the results of complex
computations. Before you do that, however, instrument the application to determine whether the application
is really CPU bound.
It might be useful to prime the cache when the application starts. Populate the cache with the data that is
most likely to be used.
Always include instrumentation that detects cache hits and cache misses. Use this information to tune
caching policies, such what data to cache, and how long to hold data in the cache before it expires.
If the lack of caching is a bottleneck, then adding caching may increase the volume of requests so much that
the web front end becomes overloaded. Clients may start to receive HTTP 503 (Service Unavailable) errors.
These are an indication that you should scale out the front end.

How to detect the problem

You can perform the following steps to help identify whether lack of caching is causing performance problems:
1. Review the application design. Take an inventory of all the data stores that the application uses. For each,
determine whether the application is using a cache. If possible, determine how frequently the data changes.
Good initial candidates for caching include data that changes slowly, and static reference data that is read
frequently.
2. Instrument the application and monitor the live system to find out how frequently the application retrieves
data or calculates information.
3. Profile the application in a test environment to capture low-level metrics about the overhead associated
with data access operations or other frequently performed calculations.
4. Perform load testing in a test environment to identify how the system responds under a normal workload
and under heavy load. Load testing should simulate the pattern of data access observed in the production
environment using realistic workloads.
5. Examine the data access statistics for the underlying data stores and review how often the same data
requests are repeated.
Example diagnosis
The following sections apply these steps to the sample application described earlier.
Instrument the application and monitor the live system
Instrument the application and monitor it to get information about the specific requests that users make while the
application is in production.
The following image shows monitoring data captured by New Relic during a load test. In this case, the only HTTP
GET operation performed is Person/GetAsync . But in a live production environment, knowing the relative frequency
that each request is performed can give you insight into which resources should be cached.

If you need a deeper analysis, you can use a profiler to capture low-level performance data in a test environment
(not the production system). Look at metrics such as I/O request rates, memory usage, and CPU utilization. These
metrics may show a large number of requests to a data store or service, or repeated processing that performs the
same calculation.
Load test the application
The following graph shows the results of load testing the sample application. The load test simulates a step load of
up to 800 users performing a typical series of operations.
The number of successful tests performed each second reaches a plateau, and additional requests are slowed as a
result. The average test time steadily increases with the workload. The response time levels off once the user load
peaks.
Examine data access statistics
Data access statistics and other information provided by a data store can give useful information, such as which
queries are repeated most frequently. For example, in Microsoft SQL Server, the sys.dm_exec_query_stats
management view has statistical information for recently executed queries. The text for each query is available in
the sys.dm_exec-query_plan view. You can use a tool such as SQL Server Management Studio to run the following
SQL query and determine how frequently queries are performed.

SELECT UseCounts, Text, Query_Plan

FROM sys.dm_exec_cached_plans
CROSS APPLY sys.dm_exec_sql_text(plan_handle)
CROSS APPLY sys.dm_exec_query_plan(plan_handle)

The UseCount column in the results indicates how frequently each query is run. The following image shows that
the third query was run more than 250,000 times, significantly more than any other query.
Here is the SQL query that is causing so many database requests:

(@p__linq__0 int)SELECT TOP (2)

[Extent1].[BusinessEntityId] AS [BusinessEntityId],
[Extent1].[FirstName] AS [FirstName],
[Extent1].[LastName] AS [LastName]
FROM [Person].[Person] AS [Extent1]
WHERE [Extent1].[BusinessEntityId] = @p__linq__0

This is the query that Entity Framework generates in GetByIdAsync method shown earlier.
Implement the solution and verify the result
After you incorporate a cache, repeat the load tests and compare the results to the earlier load tests without a
cache. Here are the load test results after adding a cache to the sample application.
The volume of successful tests still reaches a plateau, but at a higher user load. The request rate at this load is
significantly higher than earlier. Average test time still increases with load, but the maximum response time is 0.05
ms, compared with 1 ms earlier—a 20× improvement.

Related resources
API implementation best practices
Cache-Aside pattern
Caching best practices
Circuit Breaker pattern
 Synchronous I/O antipattern
12/18/2020 • 6 minutes to read • Edit Online

Blocking the calling thread while I/O completes can reduce performance and affect vertical scalability.

Problem description
A synchronous I/O operation blocks the calling thread while the I/O completes. The calling thread enters a wait
state and is unable to perform useful work during this interval, wasting processing resources.
Common examples of I/O include:
Retrieving or persisting data to a database or any type of persistent storage.
Sending a request to a web service.
Posting a message or retrieving a message from a queue.
Writing to or reading from a local file.
This antipattern typically occurs because:
It appears to be the most intuitive way to perform an operation.
The application requires a response from a request.
The application uses a library that only provides synchronous methods for I/O.
An external library performs synchronous I/O operations internally. A single synchronous I/O call can block an
entire call chain.
The following code uploads a file to Azure blob storage. There are two places where the code blocks waiting for
synchronous I/O, the CreateIfNotExists method and the UploadFromStream method.

var blobClient = storageAccount.CreateCloudBlobClient();

var container = blobClient.GetContainerReference("uploadedfiles");

container.CreateIfNotExists();
var blockBlob = container.GetBlockBlobReference("myblob");

// Create or overwrite the "myblob" blob with contents from a local file.
using (var fileStream = File.OpenRead(HostingEnvironment.MapPath("~/FileToUpload.txt")))
{
blockBlob.UploadFromStream(fileStream);
}

Here's an example of waiting for a response from an external service. The GetUserProfile method calls a remote
service that returns a UserProfile .
 public interface IUserProfileService
{
UserProfile GetUserProfile();
}

public class SyncController : ApiController

{
private readonly IUserProfileService _userProfileService;

public SyncController()
{
_userProfileService = new FakeUserProfileService();
}

// This is a synchronous method that calls the synchronous GetUserProfile method.

public UserProfile GetUserProfile()
{
return _userProfileService.GetUserProfile();
}
}

You can find the complete code for both of these examples here.

How to fix the problem

Replace synchronous I/O operations with asynchronous operations. This frees the current thread to continue
performing meaningful work rather than blocking, and helps improve the utilization of compute resources.
Performing I/O asynchronously is particularly efficient for handling an unexpected surge in requests from client
applications.
Many libraries provide both synchronous and asynchronous versions of methods. Whenever possible, use the
asynchronous versions. Here is the asynchronous version of the previous example that uploads a file to Azure blob
storage.

var blobClient = storageAccount.CreateCloudBlobClient();

var container = blobClient.GetContainerReference("uploadedfiles");

await container.CreateIfNotExistsAsync();

var blockBlob = container.GetBlockBlobReference("myblob");

// Create or overwrite the "myblob" blob with contents from a local file.
using (var fileStream = File.OpenRead(HostingEnvironment.MapPath("~/FileToUpload.txt")))
{
await blockBlob.UploadFromStreamAsync(fileStream);
}

The await operator returns control to the calling environment while the asynchronous operation is performed.
The code after this statement acts as a continuation that runs when the asynchronous operation has completed.
A well designed service should also provide asynchronous operations. Here is an asynchronous version of the web
service that returns user profiles. The GetUserProfileAsync method depends on having an asynchronous version
of the User Profile service.
 public interface IUserProfileService
{
Task<UserProfile> GetUserProfileAsync();
}

public class AsyncController : ApiController

{
private readonly IUserProfileService _userProfileService;

public AsyncController()
{
_userProfileService = new FakeUserProfileService();
}

// This is an synchronous method that calls the Task based GetUserProfileAsync method.
public Task<UserProfile> GetUserProfileAsync()
{
return _userProfileService.GetUserProfileAsync();
}
}

For libraries that don't provide asynchronous versions of operations, it may be possible to create asynchronous
wrappers around selected synchronous methods. Follow this approach with caution. While it may improve
responsiveness on the thread that invokes the asynchronous wrapper, it actually consumes more resources. An
extra thread may be created, and there is overhead associated with synchronizing the work done by this thread.
Some tradeoffs are discussed in this blog post: Should I expose asynchronous wrappers for synchronous methods?
Here is an example of an asynchronous wrapper around a synchronous method.

// Asynchronous wrapper around synchronous library method

private async Task<int> LibraryIOOperationAsync()
{
return await Task.Run(() => LibraryIOOperation());
}

Now the calling code can await on the wrapper:

// Invoke the asynchronous wrapper using a task

await LibraryIOOperationAsync();

Considerations
I/O operations that are expected to be very short lived and are unlikely to cause contention might be more
performant as synchronous operations. An example might be reading small files on an SSD drive. The
overhead of dispatching a task to another thread, and synchronizing with that thread when the task
completes, might outweigh the benefits of asynchronous I/O. However, these cases are relatively rare, and
most I/O operations should be done asynchronously.
Improving I/O performance may cause other parts of the system to become bottlenecks. For example,
unblocking threads might result in a higher volume of concurrent requests to shared resources, leading in
turn to resource starvation or throttling. If that becomes a problem, you might need to scale out the number
of web servers or partition data stores to reduce contention.

How to detect the problem

For users, the application may seem unresponsive periodically. The application might fail with timeout exceptions.
These failures could also return HTTP 500 (Internal Server) errors. On the server, incoming client requests might be
blocked until a thread becomes available, resulting in excessive request queue lengths, manifested as HTTP 503
(Service Unavailable) errors.
You can perform the following steps to help identify the problem:
1. Monitor the production system and determine whether blocked worker threads are constraining
throughput.
2. If requests are being blocked due to lack of threads, review the application to determine which operations
may be performing I/O synchronously.
3. Perform controlled load testing of each operation that is performing synchronous I/O, to find out whether
those operations are affecting system performance.

Example diagnosis
The following sections apply these steps to the sample application described earlier.
Monitor web server performance
For Azure web applications and web roles, it's worth monitoring the performance of the IIS web server. In
particular, pay attention to the request queue length to establish whether requests are being blocked waiting for
available threads during periods of high activity. You can gather this information by enabling Azure diagnostics. For
more information, see:
Monitor Apps in Azure App Service
Create and use performance counters in an Azure application
Instrument the application to see how requests are handled once they have been accepted. Tracing the flow of a
request can help to identify whether it is performing slow-running calls and blocking the current thread. Thread
profiling can also highlight requests that are being blocked.
Load test the application
The following graph shows the performance of the synchronous GetUserProfile method shown earlier, under
varying loads of up to 4000 concurrent users. The application is an ASP.NET application running in an Azure Cloud
Service web role.

The synchronous operation is hard-coded to sleep for 2 seconds, to simulate synchronous I/O, so the minimum
response time is slightly over 2 seconds. When the load reaches approximately 2500 concurrent users, the average
response time reaches a plateau, although the volume of requests per second continues to increase. Note that the
scale for these two measures is logarithmic. The number of requests per second doubles between this point and
the end of the test.
In isolation, it's not necessarily clear from this test whether the synchronous I/O is a problem. Under heavier load,
the application may reach a tipping point where the web server can no longer process requests in a timely manner,
causing client applications to receive time-out exceptions.
Incoming requests are queued by the IIS web server and handed to a thread running in the ASP.NET thread pool.
Because each operation performs I/O synchronously, the thread is blocked until the operation completes. As the
workload increases, eventually all of the ASP.NET threads in the thread pool are allocated and blocked. At that
point, any further incoming requests must wait in the queue for an available thread. As the queue length grows,
requests start to time out.
Implement the solution and verify the result
The next graph shows the results from load testing the asynchronous version of the code.

Throughput is far higher. Over the same duration as the previous test, the system successfully handles a nearly
tenfold increase in throughput, as measured in requests per second. Moreover, the average response time is
relatively constant and remains approximately 25 times smaller than the previous test.
 Responsible Innovation: A Best Practices Toolkit
12/18/2020 • 2 minutes to read • Edit Online

This toolkit provides developers with a set of practices in development, for anticipating and addressing the
potential negative impacts of technology on people. We are sharing this as an early stage practice for feedback and
learning.

Judgment Call
Judgment Call is an award-winning game and team-based activity that puts Microsoft’s AI principles of fairness,
privacy and security, reliability and safety, transparency, inclusion, and accountability into action. The game provides
an easy-to-use method for cultivating stakeholder empathy through scenario-imagining. Game participants write
product reviews from the perspective of a particular stakeholder, describing what kind of impact and harms the
technology could produce from their point of view.

Harms Modeling
Harms Modeling is a framework for product teams, grounded in four core pillars that examine how people's lives
can be negatively impacted by technology: injuries, denial of consequential services, infringement on human rights,
and erosion of democratic & societal structures. Similar to Security Threat Modeling, Harms Modeling enables
product teams to anticipate potential real-world impacts of technology, which is a cornerstone of responsible
development.

Community Jury
Community Jury is a technique that brings together diverse stakeholders impacted by a technology. It is an
adaptation of the citizen jury. The stakeholders are provided an opportunity to learn from experts about a project,
deliberate together, and give feedback on use cases and product design. This technique allows project teams to
collaborate with researchers to identify stakeholder values, and understand the perceptions and concerns of
impacted stakeholders.
 Judgment Call
12/18/2020 • 5 minutes to read • Edit Online

Judgment Call is an award-winning game and team-based activity that puts Microsoft’s AI principles of fairness,
privacy and security, reliability and safety, transparency, inclusion, and accountability into action. The game
provides an easy-to-use method for cultivating stakeholder empathy by imagining their scenarios. Game
participants write product reviews from the perspective of a particular stakeholder, describing what kind of impact
and harms the technology could produce from their point of view.

Benefits
Technology builders need practical methods to incorporate ethics in product development, by considering the
values of diverse stakeholders and how technology may uphold or not uphold those values. The goal of the game
is to imagine potential outcomes of a product or platform by gaining a better understanding of stakeholders, and
what they need and expect.
The game helps people discuss challenging topics in a safe space within the context of gameplay, and gives
technology creator a vocabulary to facilitate ethics discussions during product development. It gives managers and
designers an interactive tool to lead ethical dialogue with their teams to incorporate ethical design in their
products.
The theoretical basis and initial outcomes of the Judgment Call game were presented at the 2019 ACM Design
Interactive Systems conference and the game was a finalist in the Fast Company 2019 Innovation by Design
Awards (Social Goods category). The game has been presented to thousands of people in the US and
internationally. While the largest group facilitated in game play has been over 100, each card deck can involve 1-10
players. This game is not intended to be a substitute for performing robust user research. Rather, it's an exercise to
help tech builders learn how to exercise their moral imagination.

Preparation
Judgment Call uses role play to surface ethical concerns in product development so players will anticipate societal
impact, write product reviews, and explore mitigations. Players think of what kind of harms and impacts the
technology might produce by writing product reviews from the point of view of a stakeholder.
To prepare for this game, download the printable Judgment Call game kit.

During the Game

The players are expected to go through the following steps in this game:
1. Identify the product
Start by identifying a product that your team is building, or a scenario you are working on. For example, if
your team is developing synthetic voice technology, your scenario could be developing a voice assistant for
online learning.
2. Pass the cards
Once you have decided on the scenario, pass out the cards. Each player receives a stakeholder card, an
ethical principle card, a rating card, and a review form. The following is a description of these cards:
 Stakeholder - This represents the group of people impacted by the technology or scenario you’ve
chosen. Stakeholders can also include the makers (those proposing and working on the technology)
as well as the internal project sponsors (managers and executives who are supportive of the project).
Principle - This includes Microsoft’s ethical principles of fairness, privacy and security, reliability and
safety, transparency, inclusion, and accountability.
Rating - Star cards give a rating: 1-star is poor, 3-star is mediocre, 5-star is excellent.
3. Brainstorm and identify stakeholders
As a group, brainstorm all the stakeholders that could be directly or indirectly affected by the technology.
Then assign each player a stakeholder from the list. In some cases, the Judgment Call decks may have pre-
populated stakeholder cards, but players can add additional cards relevant to their situation.
The deck provides following categories of stakeholders:
Direct - These use the technology and/or the technology is used on them.
Indirect - These feel the effects of the technology.
Special populations - A wide range of categories that includes those who cannot use the
technology or choose not to use it, and organizations that may be interested in its deployment like
advocacy groups or the government.
There are a range of harms that can be felt or perpetuated onto each group of stakeholder. Types of
stakeholders can include end users of the product, administrators, internal teams, advocacy groups,
manufacturers, someone who might be excluded from using the product. Stakeholders can also include the
makers who propose and work on the technology, as well as the internal project sponsors, that is, the
managers and executives who are supportive of the project.
4. Moderator selection
The moderator describes scenarios for gameplay, guides discussion, and collects comment cards at the end.
They will also give an overview of the ethical principles considered in the deck.
5. Presentation of a product
The moderator introduces a product, real or imaginary, for the players to role play. They can start with
questions about the product with prompts such as:
Who will use the product?
What features will it have?
When will it be used?
Where will it be used?
Why is it useful?
How will it be used?
6. Writing reviews
Players each write a review of the product or scenario based on the cards they’ve been dealt. Players take
turns reading their reviews aloud. Listen carefully for any issues or themes that might come up. Take note of
any changes you might make to your product based on insights discussed.
The stakeholder card tells them who they are for the round.
The ethical principle card is one of the six Microsoft AI principles; it tells them what area they’re
 focusing on in their review.
The rating card tells them how much their stakeholder likes the technology (1, 3, or 5 stars).
The player then writes a review of the technology, from the perspective of their stakeholder, of why they like
or dislike the product from the perspective of the ethical principle they received.

Discussion
The moderator has each player read their reviews. Everyone is invited to discuss the different perspectives and
other considerations that may not have come up.
Potential moderator questions include:
Who is the most impacted?
What features are problematic?
What are the potential harms?

Harms mitigation
Finally, the group picks a thread from the discussion to begin identifying design or process changes that can
mitigate a particular harm. At the end of each round, the decks are shuffled, and another round can begin with the
same or a different scenario.

Next steps
Once you have enough understanding of potential harms or negative impact your product or scenario may cause,
proceed to learn how to model these harms so you can devise effective mitigations.
 Foundations of assessing harm
12/18/2020 • 3 minutes to read • Edit Online

Harms Modeling is a practice designed to help you anticipate the potential for harm, identify gaps in product that
could put people at risk, and ultimately create approaches that proactively address harm.

Why harms modeling?

As technology builders, your work is global. Designing AI to be trustworthy requires creating solutions that reflect
ethical principles deeply rooted in important and timeless values. In the process of designing and building your
technologies, it is essential to evaluate not only ideal outcomes, but possible negative ones as well.
Technology and human rights
It is as important as ever to be aware of how digital technology could impact human rights. In addition to
continuing to protect privacy and security, we must address the risks of AI and other emerging technologies, such
as facial recognition. History teaches us that human rights violations not only result from the nefarious use of
technology, but also from a lack of awareness amongst those with good intentions. As a part of our company's
dedication to the protection of human rights, Microsoft forged a partnership with important stakeholders outside
of our industry, including the United Nations (UN).
An important set of UN principles that our company firmly supports, which was ratified by over 250 nations, is the
Universal Declaration of Human Rights (UDHR). The UDHR is a milestone document. Drafted by a diverse global
group of legal and cultural experts, the Declaration was proclaimed by the United Nations General Assembly in
1948 as a common standard of achievements for all peoples and all nations. It sets out, for the first time,
fundamental human rights to be universally protected. It has been translated into over 500 languages. Additionally,
Microsoft is one of 4,700 corporate signatories to the UN Global Compact, an international business initiative
designed to promote responsible corporate citizenship.

Download in Microsoft Word

Human understanding
In addition to appreciating the importance of human rights, building trustworthy systems requires us to consider
many people's perspectives. Asking who the stakeholders are, what they value, how they could benefit, and how
they could be hurt by our technology, is a powerful step that allows us to design and build better products.
Who does the technology impact?
Who are the customers?
What do they value?
How should they benefit?
How could tech harm them?
Who are the non-customer stakeholders?
What do they value?
How should they benefit?
How could tech harm them?
Asking these questions is a practice in Value Sensitive Design and is the beginning to better understanding what is
important to stakeholders, and how it plays into their relationship with the product.
Types of stakeholders
Project sponsors
Backers, decision makers, and owners make up this category. Their values are articulated in the project strategy
and goals.
Tech builders
Designers, developers, project managers, and those working directly on designing systems make up this group.
They bring their own ethical standards and profession-specific values to the system.
Direct & indirect stakeholders
These stakeholders are significantly impacted by the system. This includes end users, software staff, clients,
bystanders, interfacing institutions, and even past or future generations. Non-human factors such as places, for
example, historic buildings or sacred spaces, may also be included.
Marginalized populations
This category is made up of the population frequently considered a minority, vulnerable, or stigmatized. This
includes children, the elderly, members of the LGBTQ+ community, ethnic minorities, and other populations who
often experience unique and disproportionate consequences.

Assessing Harm
Once you have defined the technology purpose, use cases, and stakeholders, conduct a Harms Modeling exercise
to evaluate potential ways the use of a technology you are building could result in negative outcomes for people
and society.

Download in Microsoft Word

The diagram above is an example of a harms evaluation. This is a qualitative approach used to understand
potential magnitude of harm.
You can complete this ideation activity individually, but ideally it is conducted as collaboration between developers,
data scientists, designers, user researcher, business decision-makers, and other disciplines that are involved in
building the technology.
Suggestions for harm description statements:
Intended use: If [feature] was used for [use case], [stakeholder] could experience [harm description].
Unintended use: If [user] tried to use [feature] for [use case], [stakeholder] could experience [harm description].
System error: If [feature] failed to function properly when used for [use case], [stakeholder] could experience
[harm description].
Misuse: [Malicious actor] could potentially use [feature], to cause [harm description] to [stakeholder].
Use the categories, questions, and examples described in the Types of harm to generate specific ideas for how
harm could occur. The article lists categories of harms, that are based upon common negative impact areas. Adapt
and adopt additional categories that are relevant to you.

Next Steps
Read Types of harm for further harms analysis.
 Types of harm
12/18/2020 • 13 minutes to read • Edit Online

This article creates awareness for the different types of harms, so that appropriate mitigation steps can be
implemented.

Risk of injury
Physical injury
Consider how technology could hurt people or create dangerous environments.

H A RM DESC RIP T IO N C O N SIDERAT IO N ( S) EXA M P L E

Overreliance on safety This points out the How might people rely on A healthcare agent could
features dependence on technology this technology to keep misdiagnose illness, leading
to make decisions without them safe? How could this to unnecessary treatment.
adequate human oversight. technology reduce
appropriate human
oversight?

Inadequate fail-safes Real-world testing may If this technology fails or is If an automatic door failed
insufficiently consider a misused, how would people to detect a wheelchair
diverse set of users and be impacted? At what point during an emergency
scenarios. could a human intervene? evacuation, a person could
Are there alternative uses be trapped if there isn't an
that have not been tested accessible override button.
for? How would users be
impacted by a system
failure?

Exposure to unhealthy Manufacturing, as well as What negative outcomes Inadequate safety

agents disposal of technology can could come from the measures could cause
jeopardize the health and manufacturing of your workers to be exposed to
well-being of workers and components or devices? toxins during digital
nearby inhabitants. component manufacturing.

Emotional or psychological injury

Misused technology can lead to serious emotional and psychological distress.

H A RM DESC RIP T IO N C O N SIDERAT IO N ( S) EXA M P L E

Overreliance on Misguided beliefs can lead How could this technology A chat bot could be relied
automation users to trust the reliability reduce direct interpersonal upon for relationship advice
of a digital agent over that feedback? How might this or mental health counseling
of a human. technology interface with instead of a trained
trusted sources of professional.
information? How could
sole dependence on an
artificial agent impact a
person?
 H A RM DESC RIP T IO N C O N SIDERAT IO N ( S) EXA M P L E

Distor tion of reality or When intentionally Could this be used to An IoT device could enable
gaslighting misused, technology can modify digital media or monitoring and controlling
undermine trust and physical environments? of an ex-intimate partner
distort someone's sense of from afar.
reality.

Reduced self- Some shared content can How could this technology Synthetic media "revenge
esteem/reputation be harmful, false, be used to inappropriately porn" can swap faces,
damage misleading, or denigrating. share personal information? creating the illusion of a
How could it be person participating in a
manipulated to misuse video, who did not.
information and
misrepresent people?

Addiction/attention Technology could be In what ways might this Variable drop rates in video
hijacking designed for prolonged technology reward or game loot boxes could
interaction, without regard encourage continued cause players to keep
for well-being. interaction beyond playing and neglect self-
delivering user value? care.

Identity theft Identity theft may lead to How might an individual be Synthetic voice font could
loss of control over impersonated with this mimic the sound of a
personal credentials, technology? How might person's voice and be used
reputation, and/or this technology mistakenly to access a bank account.
representation. recognize the wrong
individual as an authentic
user?

Misattribution This includes crediting a In what ways might this Facial recognition can
person with an action or technology attribute an misidentify an individual
content that they are not action to an individual or during a police
responsible for. group? How could investigation.
someone be affected if an
action was incorrectly
attributed to them?

Denial of consequential services

Opportunity loss
Automated decisions could limit access to resources, services, and opportunities essential to well-being.

H A RM DESC RIP T IO N C O N SIDERAT IO N ( S) EXA M P L E

Employment Some people may be Are there ways in which Hiring AI could recommend
discrimination denied access to apply for this technology could fewer candidates with
or secure a job based on impact recommendations female-sounding names for
characteristics unrelated to or decisions related to interviews.
merit. employment?

Housing discrimination This includes denying How could this technology Public housing queuing
people access to housing or impact recommendations algorithm could cause
the ability to apply for or decisions related to people with international-
housing. housing? sounding names to wait
longer for vouchers.
 H A RM DESC RIP T IO N C O N SIDERAT IO N ( S) EXA M P L E

Insurance and benefit This includes denying Could this technology be Insurance company might
discrimination people insurance, social used to determine access, charge higher rates for
assistance, or access to a cost, allocation of insurance drivers working night shifts
medical trial due to biased or social benefits? due to algorithmic
standards. predictions suggesting
increased drunk driving
risk.

Educational Access to education may be How might this technology Emotion classifier could
discrimination denied because of an be used to determine incorrectly report that
unchangeable access, cost, students of color are less
characteristic. accommodations, or other engaged than their white
outcomes related to counterparts, leading to
education? lower grades.

Digital Disproportionate access to What prerequisite skills, Content throttling could

divide/technological the benefits of technology, equipment, or connectivity prevent rural students from
discrimination may leave some people less are necessary to get the accessing classroom
informed or less equipped most out of this instruction video feed.
to participate in society. technology? What might be
the impact of select people
gaining earlier access to
this technology than
others, in terms of
equipment, connectivity, or
other product functionality?

Loss of choice/network Presenting people with only How might this technology News feed could only
and filter bubble information that conforms affect which choices and present information that
to and reinforces their own information are made confirms existing beliefs.
beliefs. available to people? What
past behaviors or
preferences might this
technology rely on to
predict future behaviors or
preferences?

Economic loss
Automating decisions related to financial instruments, economic opportunity, and resources can amplify existing
societal inequities and obstruct well-being.

H A RM DESC RIP T IO N C O N SIDERAT IO N ( S) EXA M P L E

Credit discrimination People may be denied How might this technology Higher introductory rate
access to financial rely on existing credit offers could be sent only to
instruments based on structures to make homes in lower
characteristics unrelated to decisions? How might this socioeconomic postal
economic merit. technology affect the ability codes.
of an individual or group to
obtain or maintain a credit
score?
 H A RM DESC RIP T IO N C O N SIDERAT IO N ( S) EXA M P L E

Differential pricing of Goods or services may be How could this technology More could be charged for
goods and ser vices offered at different prices be used to determine products based on
for reasons unrelated to pricing of goods or designation for men or
the cost of production or services? What are the women.
delivery. criteria for determining the
cost to people for use of
this tech?

Economic exploitation People might be compelled What role did human labor Paying financially destitute
or misled to work on play in producing training people for their biometric
something that impacts data for this technology? data to train AI systems.
their dignity or wellbeing. How was this workforce
acquired? What role does
human labor play in
supporting this
technology? Where is this
workforce expected to
come from?

Devaluation of Technology may supplant How might this technology AI agents replace
individual exper tise the use of paid human impact the need to employ doctors/radiographers for
expertise or labor. an existing workforce? evaluation of medical
imaging.

Infringement on human rights

Dignity loss
Technology can influence how people perceive the world, and how they recognize, engage, and value one another.
The exchange of honor and respect between people can be interfered with.

H A RM DESC RIP T IO N C O N SIDERAT IO N ( S) EXA M P L E

Dehumanization Removing, reducing, or How might this technology Entity recognition and
obscuring visibility of a be used to simplify or virtual overlays in drone
person's humanity. abstract the way a person surveillance could reduce
is represented? How might the perceived accountability
this technology reduce the of human actions.
distinction between
humans and the digital
world?

Public shaming This may mean exposing How might movements or A fitness app could reveal a
people's private, sensitive, actions be revealed user's GPS location on
or socially inappropriate through data aggregation? social media, indicating
material. attendance at an Alcoholics
Anonymous meeting.

Liberty loss
Automation of legal, judicial, and social systems can reinforce biases and lead to detrimental consequences.

H A RM DESC RIP T IO N C O N SIDERAT IO N ( S) EXA M P L E

 H A RM DESC RIP T IO N C O N SIDERAT IO N ( S) EXA M P L E

Predictive policing Inferring suspicious How could this support or An algorithm can predict a
behavior and/or criminal replace human policing or number of area arrests, so
intent based on historical criminal justice decision- police make sure they
records. making? match, or exceed, that
number.

Social control Conformity may be What types of personal or Authoritarian government

reinforced or encouraged behavioral data might feed uses social media and e-
by publicly designating this technology? How commerce data to
human behaviors as would it be obtained? What determine a "trustworthy"
positive or negative. outputs would be derived score based on where
from this data? Is this people shop and who they
technology likely to be spend time with.
used to encourage or
discourage certain
behaviors?

Loss of effective This means an inability to How might people Automated prison sentence
remedy explain the rationale or lack understand the reasoning or pre-trial release decision
of opportunity to contest a for decisions made by this is not explained to the
decision. technology? How might an accused.
individual that relies on this
technology explain the
decisions it makes? How
could people contest or
question a decision this
technology makes?

Privacy loss
Information generated by our use of technology can be used to determine facts or make assumptions about
someone without their knowledge.

H A RM DESC RIP T IO N C O N SIDERAT IO N ( S) EXA M P L E

Interference with Revealing information a How could this technology Task-tracking AI could
private life person has not chosen to use information to infer monitor personal patterns
share. portions of a private life? from which it infers an
How could decisions based extramarital affair.
upon these inferences
expose things that a person
does not want made
public?

Forced association Requiring participation in How might use of this Biometric enrollment in a
the use of technology or technology be required for company's meeting room
surveillance to take part in participation in society or transcription AI is a
society. organization membership? stipulated requirement in
job offer letter.
 H A RM DESC RIP T IO N C O N SIDERAT IO N ( S) EXA M P L E

Inability to freely and This may mean restriction In what way does the Intelligent meeting system
fully develop of one's ability to truthfully system or product ascribe could record all discussions
personality express themselves or positive vs negative between colleagues
explore external avenues connotations toward including personal coaching
for self-development. particular personality traits? and mentorship sessions.
In what way can using the
product or system reveal
information to entities such
as the government or
employer that inhibits free
expression?

Never forgiven Digital files or records may What and where is data A teenager's social media
never be deleted. being stored from this history could remain
product, and who can searchable long after they
access it? How long is user have outgrown the
data stored after platform.
technology interaction?
How is user data updated
or deleted?

Loss of freedom of This means an inability to In what ways might this A real name could be
movement or assembly navigate the physical or technology monitor people required in order to sign up
virtual world with desired across physical and virtual for a video game enabling
anonymity. space? real-world stalking.

Environmental impact
The environment can be impacted by every decision in a system or product life cycle, from the amount of cloud
computing needed to retail packaging. Environmental changes can impact entire communities.

H A RM DESC RIP T IO N C O N SIDERAT IO N ( S) EXA M P L E

Exploitation or Obtaining the raw materials What materials are needed A local community could be
depletion of resources for a technology, including to build or run this displaced due to the
how it's powered, leads to technology? What energy harvesting of rare earth
negative consequences to requirements are needed to minerals and metals
the environment and its build or run this required for some
inhabitants. technology? electronic manufacturing.

Electronic waste Reduced quality of How might this technology Toxic materials inside
collective well-being reduce electronic waste by discarded electronic devices
because of the inability to recycling materials or could leach into the water
repair, recycle, or otherwise allowing users to self- supply, making local
responsibly dispose of repair? How might this populations ill.
electronics. technology contribute to
electronic waste when new
versions are released or
when current/past versions
stop working?

Erosion of social & democratic structures

Manipulation
The ability for technology to be used to create highly personalized and manipulative experiences can undermine an
informed citizenry and trust in societal structures.
 H A RM DESC RIP T IO N C O N SIDERAT IO N ( S) EXA M P L E

Misinformation Disguising fake information How might this technology Generation of synthetic
as legitimate or credible be used to generate speech of a political leader
information. misinformation? How could sways an election.
it be used to spread
misinformation that
appears credible?

Behavioral exploitation This means exploiting How might this technology Monitoring shopping
personal preferences or be used to observe habits in connected retail
patterns of behavior to patterns of behavior? How environment leads to
induce a desired reaction. could this technology be personalized incentives for
used to encourage impulse shoppers and
dysfunctional or hoarders.
maladaptive behaviors?

Social detriment
At scale, the way technology impacts people shapes social and economic structures within communities. It can
further ingrain elements that include or benefit some, at the exclusion of others.

H A RM DESC RIP T IO N C O N SIDERAT IO N ( S) EXA M P L E

Amplification of power This may perpetuate How might this technology Requiring a residential
inequality existing class or privilege be used in contexts where address & phone number
disparities. there are existing social, to register on a job website
economic, or class could prevent a homeless
disparities? How might person from applying for
people with more power or jobs.
privilege disproportionately
influence the technology?

Stereotype This may perpetuate How might this technology Results of an image search
reinforcement uninformed "conventional be used to reinforce or for "CEO" could primarily
wisdom" about historically amplify existing social show photos of Caucasian
or statistically norms or cultural men.
underrepresented people. stereotypes? How might
the data used by this
technology cause it to
reflect biases or
stereotypes?

Loss of individuality This may be an inability to How might this technology Limited customization
express a unique amplify majority opinions options in designing a
perspective. or "group-think"? video game avatar inhibits
Conversely, how might self-expression of a player's
unique forms of expression diversity.
be suppressed. In what
ways might the data
gathered by this
technology be used in
feedback to people?

Loss of representation Broad categories of How could this technology Automatic photo caption
generalization obscure, constrain identity options? assigns incorrect gender
diminish, or erase real Could it be used to identity and age to the
identities. automatically label or subject.
categorize people?
 H A RM DESC RIP T IO N C O N SIDERAT IO N ( S) EXA M P L E

Skill degradation and Overreliance on In what ways might this Overreliance on

complacency automation leads to technology reduce the automation could lead to
atrophy of manual skills. accessibility and ability to an inability to gauge the
use manual controls? airplane's true orientation
because the pilots have
been trained to rely on
instruments only.

Evaluate harms
Once you have generated a broad list of potential harms, you should complete your Harms Model by evaluating
the potential magnitude for each category of harm. This will allow you to prioritize your areas of focus. See the
following example harms model for reference:

C O N T RIB UT IN G FA C TO R DEF IN IT IO N

Severity How acutely could an individual or group's well-being be

impacted by the technology?

Scale How broadly could the impact to well-being be experienced

across populations or groups?

Probability How likely is it that individual or group's well-being will be

impacted by the technology?

Frequency How often would an individual or group experience an impact

to their well-being from the technology?

Next Steps
Use the Harms Model you developed to guide your product development work:
Seek more information from stakeholders that you identified as potentially experiencing harm.
Develop and validate hypothesis for addressing the areas you identified as having the highest potential for
harm.
Integrate the insights into your decisions throughout the technology development process: data collectionand
model training, system architecture, user experience design, product documentation, feedback loops, and
communication capabilities and limitationsof the technology.
Explore Community Jury.
Assess and mitigate unfairness using Azure Machine Learning and the open-source FairLearn package.
Other Responsible AI tools:
Responsible AI resource center
Guidelines for Human AI interaction
Conversational AI guidelines
Inclusive Design guidelines
AI Fairness checklist
Additional references:
Downloadable booklet for assessing harms
Value Sensitive Design
 Community Jury
12/18/2020 • 7 minutes to read • Edit Online

Community jury, an adaptation of the citizen jury, is a technique where diverse stakeholders impacted by a
technology are provided an opportunity to learn about a project, deliberate together, and give feedback on use
cases and product design. This technique allows project teams to understand the perceptions and concerns of
impacted stakeholders.
A community jury is different from a focus group or market research; it allows the impacted stakeholders to hear
directly from the subject matter experts in the product team, and cocreate solutions to challenging problems with
them.

Benefits
This section discusses some important benefits of community jury.
Expert Testimony
Members of a community jury hear details about the technology under consideration, directly from experts on the
product team. These experts help highlight the capabilities in particular applications.
Proximity
A community jury allows decision-makers to hear directly from the community, and to learn about their values,
concerns, and ideas regarding a particular issue or problem. It also provides a valuable opportunity to better
understand the reasons for their conclusions.
Consensus
By bringing impacted individuals together and providing an opportunity for them to learn and discuss key aspects
of the technology, a community jury can identify areas of agreement and build common ground solutions to
challenging problems.

Community jury contributors

Product Team
This group comprises owners who will bring the product to market, representative project managers, engineers,
designers, data scientists, and others involved in the making of a product. Additionally, subject matter experts are
included who can answer in-depth questions about the product.
Preparation
Effective deliberation and cocreation require ready-to-use answers to technical questions. As product owners, it is
important to ensure technical details are collected prior to the start of the jury session.
Relevant artifacts could include:
Documentation of existing protections, planned or in place.
Data Flows, for example, what data types collected, who will have access, for how long, with what protections,
and so on.
Mockups or prototypes of proposed solutions.
During the Session
Along with providing an accessible presentation of technical details, product team witnesses should describe
certain applications of the technology.
Relevant artifacts could include:
Customer benefits
Harms assessment and socio-technical impact
Storyboards
Competitive analyses
Academic or popular media reports
Moderator
Bring on a neutral user researcher to ensure everyone is heard, avoiding domination of the communications by
any one member. The moderator will facilitate brainstorms and deliberations, as well as educate jury members in
uncovering bias, and ways to ask difficult questions. If a user researcher is not available, choose a moderator who
is skilled at facilitating group discussions. Following the session, the moderator is responsible for the following:
ensure that the agreed-upon compensation is provided to the jury members;
produce a report that describes key insights, concerns, and recommendations, and
share key insights and next steps with the jury, and thank them for their participation.
Preparation
Structure the sessions so that there is ample time for learning, deliberation, and cocreation. This could mean
having multiple sessions that go in-depth on different topics or having longer sessions.
Pilot the jury with a smaller sample of community members to work out the procedural and content issues
prior to the actual sessions.
During the Session
Ensure that all perspectives are heard, including those of the project team and those of the jury members. This
minimizes group-thinking as well as one or two individuals dominating the discussion.
Reinforce the value of the juror participation by communicating the plan for integrating jury feedback into the
product planning process. Ensure that the project team is prepared to hear criticisms from the jury.
Jury members
These are direct and indirect stakeholders impacted by the technology, representative of the diverse community in
which the technology will be deployed.
Sample size
A jury should be large enough to represent the diversity and collective will of the community, but not so large that
the voices of quieter jurors are drowned out. It is ideal to get feedback from at least 16-20 individuals total who
meet the criteria below. You can split the community jury over multiple sessions so that no more than 8-10
individuals are part of one session.
Recruitment criteria
Stratify a randomly selected participant pool so that the jury includes the demographic diversity of the community.
Based on the project objectives, relevant recruitment criteria may include balancing across gender identity, age,
privacy index. Recruitment should follow good user research recruitment procedures and reach a wider
community.

Session structure
Sessions typically last 2-3 hours. Add more or longer deep dive sessions, as needed, if aspects of the project
require more learning, deliberation, and cocreation.
1. Over view, introduction, and Q&A - The moderator provides a session overview, then introduces the
project team and explains the product's purpose, along with potential use cases, benefits, and harms.
Questions are then accepted from community members. This session should be between 30 to 60 minutes
long.
2. Discussion of key themes - Jury members ask in-depth questions about aspects of the project, fielded by
the moderator. This session should also be between 30 to 60 minutes in length.
3. This step can be any one of the following options:
Deliberation and cocreation - This option is preferable. Jury members deliberate and co-create
solutions with the project team. This is typically 30 to 60 minutes long.
Individual anonymous sur vey - Alternatively, conduct an anonymous survey of jury members.
Anonymity may allow issues to be raised that would not otherwise be expressed. Use this 30-minute
session if there are time constraints.
4. Following the session - The moderator produces a study report that describes key insights, concerns, and
potential solutions to the concerns.
If the values of different stakeholders were in conflict with each other during the session and the value tensions
were left unresolved, the product team would need to brainstorm solutions, and conduct a follow-up session with
the jury to determine if the solutions adequately resolve their concerns.

Tips to run a successful jury

Ensure alignment of goals and outcomes with the project team before planning begins, including deliverables,
recruitment strategy, and timeline. Consider including additional subject-matter experts relevant to the project,
to address open questions/concerns.
The consent to audio and video recording of the jury should follow your company's standard procedures for
non-disclosure and consent that is obtained from participants during user research studies.
Provide fair compensation to participants for the time they devote to participation. Whenever possible,
participants should also be reimbursed for costs incurred as a result of study participation, for example, parking
and transportation costs. Experienced user research recruiters can help determine fair gratuities for various
participant profiles.
Ensure that all perspectives are heard, minimizing group-thinking as well as one or two individuals dominating
the discussion. This should include those of the project team and the jury members.
Structure the sessions so that there is ample time for learning, deliberation, and cocreation. This could mean
having multiple sessions going in-depth on different topics or having longer sessions.
Pilot the jury with a smaller sample of community members to work out the procedural and content issues
prior to the actual sessions.
If the topic being discussed is related to personal data privacy, aim to balance the composition of the
community jury to include individuals with different privacy indices.
Consider including session breaks and providing snacks/refreshments for sessions that are two hours or longer.
Reinforce the value of the juror participation by communicating the plan for integrating jury feedback into the
product planning process. Also ensure that the project team is prepared to hear criticisms from the jury.
After the jury, in addition to publishing the research report, send out a thank you note to the volunteer
participants of the jury, along with an executive summary of the key insights.

Additional information
Privacy Index
The Privacy Index is an approximate measure for an individual's concern about personal data privacy, and is
gauged using the following:
1. Consumers have lost all control over how personal information is collected and used by companies.
2. Most businesses handle the personal information they collect about consumers in a proper and confidential
way.
3. Existing laws and organizational practices provide a reasonable level of protection for consumer privacy today.
Participants are asked to provide responses to the above concerns using the scale of: 1 - Strongly Agree, 2 -
Somewhat Agree, 3 - Somewhat Disagree, 4- Strongly Disagree, and classified into the categories below.
High/Fundamentalist => IF A = 1 or 2 AND B & C = 3 or 4
Low/unconcerned => IF A = 3 or 4 AND B & C = 1 or 2
Medium/pragmatist => All other responses

Next steps
Explore the following references for detailed information on this method:
Jefferson center: creator of the Citizen's Jury method https://jefferson-center.org/about-us/how-we-work/
Citizen's jury handbook http://www.rachel.org/files/document/Citizens_Jury_Handbook.pdf
Case study: Connected Health Cities (UK)
Project page
Final report
Jury specification
Juror's report
Case study: Community Jury at Microsoft
 Azure for AWS Professionals
11/2/2020 • 2 minutes to read • Edit Online

This article helps Amazon Web Services (AWS) experts understand the basics of Microsoft Azure accounts,
platform, and services. It also covers key similarities and differences between the AWS and Azure platforms.
You'll learn:
How accounts and resources are organized in Azure.
How available solutions are structured in Azure.
How the major Azure services differ from AWS services.
Azure and AWS built their capabilities independently over time so that each has important implementation and
design differences.

Overview
Like AWS, Microsoft Azure is built around a core set of compute, storage, database, and networking services. In
many cases, both platforms offer a basic equivalence between the products and services they offer. Both AWS and
Azure allow you to build highly available solutions based on Windows or Linux hosts. So, if you're used to
development using Linux and OSS technology, both platforms can do the job.
While the capabilities of both platforms are similar, the resources that provide those capabilities are often
organized differently. Exact one-to-one relationships between the services required to build a solution are not
always clear. In other cases, a particular service might be offered on one platform, but not the other. See charts of
comparable Azure and AWS services.

Services
For a listing of how services map between platforms, see AWS to Azure services comparison.
Not all Azure products and services are available in all regions. Consult the Products by Region page for details. You
can find the uptime guarantees and downtime credit policies for each Azure product or service on the Service Level
Agreements page.

Components
A number of core components on Azure and AWS have similar functionality. To review the differences, visit the
component page for the topic you're interested in:
Accounts
Compute
Databases
Messaging
Networking
Regions and Zones
Resources
Security & Identity
Storage
 Azure and AWS accounts and subscriptions
11/2/2020 • 2 minutes to read • Edit Online

Azure services can be purchased using several pricing options, depending on your organization's size and needs.
See the pricing overview page for details.
Azure subscriptions are a grouping of resources with an assigned owner responsible for billing and permissions
management. Unlike AWS, where any resources created under the AWS account are tied to that account,
subscriptions exist independently of their owner accounts, and can be reassigned to new owners as needed.

Comparison of structure and ownership of AWS accounts and Azure subscriptions

Subscriptions are assigned three types of administrator accounts:

Account Administrator . The subscription owner and the account billed for the resources used in the
subscription. The account administrator can only be changed by transferring ownership of the subscription.
Ser vice Administrator . This account has rights to create and manage resources in the subscription, but is
not responsible for billing. By default, the account administrator and service administrator are assigned to
the same account. The account administrator can assign a separate user to the service administrator account
for managing the technical and operational aspects of a subscription. Only one service administrator is
assigned per subscription.
Co-administrator . There can be multiple co-administrator accounts assigned to a subscription. Co-
administrators cannot change the service administrator, but otherwise have full control over subscription
resources and users.
Below the subscription level user roles and individual permissions can also be assigned to specific resources,
similarly to how permissions are granted to IAM users and groups in AWS. In Azure, all user accounts are
associated with either a Microsoft Account or Organizational Account (an account managed through an Azure
Active Directory).
Like AWS accounts, subscriptions have default service quotas and limits. For a full list of these limits, see Azure
subscription and service limits, quotas, and constraints. These limits can be increased up to the maximum by filing
a support request in the management portal.
See also
How to add or change Azure administrator roles
How to download your Azure billing invoice and daily usage data
 Compute services on Azure and AWS
12/18/2020 • 4 minutes to read • Edit Online

EC2 Instances and Azure virtual machines

Although AWS instance types and Azure virtual machine sizes are categorized similarly, the RAM, CPU, and storage
capabilities differ.
Amazon EC2 Instance Types
Sizes for virtual machines in Azure (Windows)
Sizes for virtual machines in Azure (Linux)
Similar to AWS' per second billing, Azure on-demand VMs are billed per second.

EBS and Azure Storage for VM disks

Durable data storage for Azure VMs is provided by data disks residing in blob storage. This is similar to how EC2
instances store disk volumes on Elastic Block Store (EBS). Azure temporary storage also provides VMs the same
low-latency temporary read-write storage as EC2 Instance Storage (also called ephemeral storage).
Higher performance disk I/O is supported using Azure premium storage. This is similar to the Provisioned IOPS
storage options provided by AWS.

Lambda, Azure Functions, Azure Web-Jobs, and Azure Logic Apps

Azure Functions is the primary equivalent of AWS Lambda in providing serverless, on-demand code. However,
Lambda functionality also overlaps with other Azure services:
WebJobs allow you to create scheduled or continuously running background tasks.
Logic Apps provides communications, integration, and business rule management services.

Autoscaling, Azure VM scaling, and Azure App Service Autoscale

Autoscaling in Azure is handled by two services:
Virtual machine scale sets allow you to deploy and manage an identical set of VMs. The number of instances
can autoscale based on performance needs.
App Service Autoscale provides the capability to autoscale Azure App Service solutions.

Container Service
The Azure Kubernetes Service supports Docker containers managed through Kubernetes.

Distributed Systems Platform

Service Fabric is a platform for developing and hosting scalable microservices-based solutions.

Batch Processing
Azure Batch allows you to manage compute-intensive work across a scalable collection of virtual machines.
Service Comparison
Virtual servers
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Elastic Compute Cloud (EC2) Instances Virtual Machines Virtual servers allow users to deploy,
manage, and maintain OS and server
software. Instance types provide
combinations of CPU/RAM. Users pay
for what they use with the flexibility to
change sizes.

Batch Batch Run large-scale parallel and high-

performance computing applications
efficiently in the cloud.

Auto Scaling Virtual Machine Scale Sets Allows you to automatically change the
number of VM instances. You set
defined metric and thresholds that
determine if the platform adds or
removes instances.

VMware Cloud on AWS Azure VMware Solution Seamlessly move VMware-based

workloads from your datacenter to
Azure and integrate your VMware
environment with Azure. Keep
managing your existing environments
with the same VMware tools you
already know while you modernize your
applications with Azure native services.
Azure VMware Solution is a Microsoft
service, verified by VMware, that runs
on Azure infrastructure.

Parallel Cluster CycleCloud Create, manage, operate, and optimize

HPC and big compute clusters of any
scale

Containers and container orchestrators

AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Elastic Container Service (ECS) Container Instances Azure Container Instances is the fastest
and simplest way to run a container in
Fargate Azure, without having to provision any
virtual machines or adopt a higher-level
orchestration service.

Elastic Container Registry Container Registry Allows customers to store Docker

formatted images. Used to create all
types of container deployments on
Azure.
 AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Elastic Kubernetes Service (EKS) Kubernetes Service (AKS) Deploy orchestrated containerized
applications with Kubernetes. Simplify
monitoring and cluster management
through auto upgrades and a built-in
operations console. See AKS solution
journey.

App Mesh Service Fabric Mesh Fully managed service that enables
developers to deploy microservices
applications without managing virtual
machines, storage, or networking.

Container architectures

Baseline architecture on Azure Kubernetes Service (AKS)

07/20/2020
37 min read
Deploy a baseline infrastructure that deploys an AKS cluster with focus on security.

Microservices architecture on Azure Kubernetes Service (AKS)

5/07/2020
17 min read
Deploy a microservices architecture on Azure Kubernetes Service (AKS)

CI/CD pipeline for container-based workloads

7/05/2018
7 min read
Build a DevOps pipeline for a Node.js web app with Jenkins, Azure Container Registry, Azure Kubernetes
Service, Cosmos DB, and Grafana.
view all
Serverless
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Lambda Functions Integrate systems and run backend

processes in response to events or
schedules without provisioning or
managing servers.

Serverless architectures
 Social App for Mobile and Web with Authentication
12/16/2019
3 min read
View a detailed, step-by-step diagram depicting the build process and implementation of the mobile client
app architecture that offers social image sharing with a companion web app and authentication abilities,
even while offline.

HIPAA and HITRUST compliant health data AI

12/16/2019
2 min read
Manage HIPAA and HITRUST compliant health data and medical records with the highest level of built-in
security.

Cross Cloud Scaling Architecture

12/16/2019
1 min read
Learn how to improve cross cloud scalability with solution architecture that includes Azure Stack. A step-by-
step flowchart details instructions for implementation.

See also
Create a Linux VM on Azure using the portal
Azure Reference Architecture: Running a Linux VM on Azure
Get started with Node.js web apps in Azure App Service
Azure Reference Architecture: Basic web application
Create your first Azure Function
 Database technologies on Azure and AWS
11/2/2020 • 2 minutes to read • Edit Online

RDS and Azure relational database services

Azure provides several different relational database services that are the equivalent of AWS' Relational Database
Service (RDS).
SQL Database
Azure Database for MySQL
Azure Database for PostgreSQL
Other database engines such as SQL Server, Oracle, and MySQL can be deployed using Azure VM Instances.
Costs for AWS RDS are determined by the amount of hardware resources that your instance uses, like CPU, RAM,
storage, and network bandwidth. In the Azure database services, cost depends on your database size, concurrent
connections, and throughput levels.
See also
Azure SQL Database Tutorials
Configure geo-replication for Azure SQL Database with the Azure portal
Introduction to Cosmos DB: A NoSQL JSON Database
How to use Azure Table storage from Node.js

Analytics and big data

The Cortana Intelligence Suite is Azure's package of products and services designed to capture, organize, analyze,
and visualize large amounts of data. The Cortana suite consists of the following services:
HDInsight: managed Apache distribution that includes Hadoop, Spark, Storm, or HBase.
Data Factory: provides data orchestration and data pipeline functionality.
SQL Data Warehouse: large-scale relational data storage.
Data Lake Store: large-scale storage optimized for big data analytics workloads.
Machine Learning: used to build and apply predictive analytics on data.
Stream Analytics: real-time data analysis.
Data Lake Analytics: large-scale analytics service optimized to work with Data Lake Store
Power BI: used to power data visualization.

Service comparison
TYPE AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N
 TYPE AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Relational database RDS SQL Database Managed relational database

service where resiliency,
Database for MySQL scale, and maintenance are
primarily handled by the
Database for PostgreSQL platform.

NoSQL / Document DynamoDB Cosmos DB A globally distributed, multi-

model database that natively
SimpleDB supports multiple data
models: key-value,
Amazon DocumentDB documents, graphs, and
columnar.

Caching ElastiCache Cache for Redis An in-memory–based,

distributed caching service
that provides a high-
performance store typically
used to offload
nontransactional work from
a database.

Database migration Database Migration Service Database Migration Service Migration of database
schema and data from one
database format to a specific
database technology in the
cloud.

Database architectures

Gaming using Cosmos DB

12/16/2019
1 min read
Elastically scale your database to accommodate unpredictable bursts of traffic and deliver low-latency multi-
player experiences on a global scale.

Oracle Database Migration to Azure

12/16/2019
2 min read
Oracle DB migrations can be accomplished in multiple ways. This architecture covers one of these options
wherein Oracle Active Data Guard is used to migrate the Database.

Retail and e-commerce using Azure MySQL

12/16/2019
1 min read
 Build secure and scalable e-commerce solutions that meet the demands of both customers and business
using Azure Database for MySQL.
view all
See also
Cortana Intelligence Gallery
Understanding Microsoft big data solutions
Azure Data Lake and Azure HDInsight Blog
 Messaging services on Azure and AWS
11/2/2020 • 2 minutes to read • Edit Online

Simple Email Service

AWS provides the Simple Email Service (SES) for sending notification, transactional, or marketing emails. In Azure,
third-party solutions like SendGrid provide email services.

Simple Queueing Service

AWS Simple Queueing Service (SQS) provides a messaging system for connecting applications, services, and
devices within the AWS platform. Azure has two services that provide similar functionality:
Queue storage: a cloud messaging service that allows communication between application components
within Azure.
Service Bus: a more robust messaging system for connecting applications, services, and devices. Using the
related Service Bus relay, Service Bus can also connect to remotely hosted applications and services.

Messaging components
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Simple Queue Service (SQS) Queue Storage Provides a managed message queueing
service for communicating between
decoupled application components.

Simple Queue Service (SQS) Service Bus Supports a set of cloud-based,

message-oriented middleware
technologies including reliable message
queuing and durable publish/subscribe
messaging.

Amazon EventBridge Event Grid A fully managed event routing service

that allows for uniform event
consumption using a publish/subscribe
model.

Messaging architectures

Anomaly Detector Process

12/16/2019
1 min read
Learn more about Anomaly Detector with a step-by-step flowchart that details the process. See how
anomaly detection models are selected with time-series data.
 Networking on Azure and AWS
11/2/2020 • 2 minutes to read • Edit Online

Elastic Load Balancing, Azure Load Balancer, and Azure Application

Gateway
The Azure equivalents of the two Elastic Load Balancing services are:
Load Balancer: provides the same capabilities as the AWS Classic Load Balancer, allowing you to distribute
traffic for multiple VMs at the network level. It also provides failover capability.
Application Gateway: offers application-level rule-based routing comparable to the AWS Application Load
Balancer.

Route 53, Azure DNS, and Azure Traffic Manager

In AWS, Route 53 provides both DNS name management and DNS-level traffic routing and failover services. In
Azure this is handled through two services:
Azure DNS provides domain and DNS management.
Traffic Manager provides DNS level traffic routing, load balancing, and failover capabilities.

Direct Connect and Azure ExpressRoute

Azure provides similar site-to-site dedicated connections through its ExpressRoute service. ExpressRoute allows
you to connect your local network directly to Azure resources using a dedicated private network connection. Azure
also offers more conventional site-to-site VPN connections at a lower cost.

Network service comparison

A REA AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Cloud virtual networking Virtual Private Cloud (VPC) Virtual Network Provides an isolated, private
environment in the cloud.
Users have control over their
virtual networking
environment, including
selection of their own IP
address range, creation of
subnets, and configuration
of route tables and network
gateways.

Cross-premises connectivity VPN Gateway VPN Gateway Connects Azure virtual

networks to other Azure
virtual networks, or
customer on-premises
networks (Site To Site).
Allows end users to connect
to Azure services through
VPN tunneling (Point To
Site).
 A REA AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

DNS management Route 53 DNS Manage your DNS records

using the same credentials
and billing and support
contract as your other Azure
services

53 Traffic Manager A service that hosts domain

names, plus routes users to
Internet applications,
connects user requests to
datacenters, manages traffic
to apps, and improves app
availability with automatic
failover.

Dedicated network Direct Connect ExpressRoute Establishes a dedicated,

private network connection
from a location to the cloud
provider (not over the
Internet).

Load balancing Network Load Balancer Load Balancer Azure Load Balancer load
balances traffic at layer 4
(TCP or UDP). Standard Load
Balancer also supports
cross-region or global load
balancing.

Application Load Balancer Application Gateway Application Gateway is a

layer 7 load balancer. It
supports SSL termination,
cookie-based session affinity,
and round robin for load-
balancing traffic.

Networking architectures

Deploy highly available NVAs

12/08/2018
7 min read
Learn how to deploy network virtual appliances for high availability in Azure. This article includes example
architectures for ingress, egress, and both.

Hub-spoke network topology in Azure

9/30/2020
7 min read
Learn how to implement a hub-spoke topology in Azure, where the hub is a virtual network and the spokes
are virtual networks that peer with the hub.
 Implement a secure hybrid network
1/07/2020
9 min read
See a secure hybrid network that extends an on-premises network to Azure with a perimeter network
between the on-premises network and an Azure virtual network.
view all

See also
Create a virtual network using the Azure portal
Plan and design Azure Virtual Networks
Azure Network Security Best Practices
 Regions and zones on Azure and AWS
12/18/2020 • 3 minutes to read • Edit Online

Failures can vary in the scope of their impact. Some hardware failures, such as a failed disk, may affect a single host
machine. A failed network switch could affect a whole server rack. Less common are failures that disrupt a whole
datacenter, such as loss of power in a datacenter. Rarely, an entire region could become unavailable.
One of the main ways to make an application resilient is through redundancy. But you need to plan for this
redundancy when you design the application. Also, the level of redundancy that you need depends on your
business requirements—not every application needs redundancy across regions to guard against a regional
outage. In general, a tradeoff exists between greater redundancy and reliability versus higher cost and complexity.
In Azure, a region is divided into two or more Availability Zones. An Availability Zone corresponds with a physically
isolated datacenter in the geographic region. Azure has numerous features for providing application redundancy at
every level of potential failure, including availability sets , availability zones , and paired regions .

The diagram has three parts. The first part shows VMs in an availability set in a virtual network. The second part
shows an availability zone with two availability sets in a virtual network. The third part shows regional pairs with
resources in each region.
The following table summarizes each option.

AVA IL A B IL IT Y SET AVA IL A B IL IT Y Z O N E PA IRED REGIO N

Scope of failure Rack Datacenter Region

Request routing Load Balancer Cross-zone Load Balancer Traffic Manager

Network latency Very low Low Mid to high

Virtual networking VNet VNet Cross-region VNet peering

Availability sets
To protect against localized hardware failures, such as a disk or network switch failing, deploy two or more VMs in
an availability set. An availability set consists of two or more fault domains that share a common power source and
network switch. VMs in an availability set are distributed across the fault domains, so if a hardware failure affects
one fault domain, network traffic can still be routed to the VMs in the other fault domains. For more information
about Availability Sets, see Manage the availability of Windows virtual machines in Azure.
When VM instances are added to availability sets, they are also assigned an update domain. An update domain is a
group of VMs that are set for planned maintenance events at the same time. Distributing VMs across multiple
update domains ensures that planned update and patching events affect only a subset of these VMs at any given
time.
Availability sets should be organized by the instance's role in your application to ensure one instance in each role is
operational. For example, in a three-tier web application, create separate availability sets for the front-end,
application, and data tiers.
Availability zones
An Availability Zone is a physically separate zone within an Azure region. Each Availability Zone has a distinct
power source, network, and cooling. Deploying VMs across availability zones helps to protect an application
against datacenter-wide failures.

Paired regions
To protect an application against a regional outage, you can deploy the application across multiple regions, using
Azure Traffic Manager to distribute internet traffic to the different regions. Each Azure region is paired with another
region. Together, these form a regional pair. With the exception of Brazil South, regional pairs are located within the
same geography in order to meet data residency requirements for tax and law enforcement jurisdiction purposes.
Unlike Availability Zones, which are physically separate datacenters but may be in relatively nearby geographic
areas, paired regions are typically separated by at least 300 miles. This design ensures that large-scale disasters
only affect one of the regions in the pair. Neighboring pairs can be set to sync database and storage service data,
and are configured so that platform updates are rolled out to only one region in the pair at a time.
Azure geo-redundant storage is automatically backed up to the appropriate paired region. For all other resources,
creating a fully redundant solution using paired regions means creating a full copy of your solution in both regions.

See also
Regions for virtual machines in Azure
Availability options for virtual machines in Azure
High availability for Azure applications
Failure and disaster recovery for Azure applications
Planned maintenance for Linux virtual machines in Azure
 Resource management on Azure and AWS
11/2/2020 • 2 minutes to read • Edit Online

The term "resource" in Azure is used in the same way as in AWS, meaning any compute instance, storage object,
networking device, or other entity you can create or configure within the platform.
Azure resources are deployed and managed using one of two models: Azure Resource Manager, or the older Azure
classic deployment model. Any new resources are created using the Resource Manager model.

Resource groups
Both Azure and AWS have entities called "resource groups" that organize resources such as VMs, storage, and
virtual networking devices. However, Azure resource groups are not directly comparable to AWS resource groups.
While AWS allows a resource to be tagged into multiple resource groups, an Azure resource is always associated
with one resource group. A resource created in one resource group can be moved to another group, but can only
be in one resource group at a time. Resource groups are the fundamental grouping used by Azure Resource
Manager.
Resources can also be organized using tags. Tags are key-value pairs that allow you to group resources across your
subscription irrespective of resource group membership.

Management interfaces
Azure offers several ways to manage your resources:
Web interface. Like the AWS Dashboard, the Azure portal provides a full web-based management interface
for Azure resources.
REST API. The Azure Resource Manager REST API provides programmatic access to most of the features
available in the Azure portal.
Command Line. The Azure CLI provides a command-line interface capable of creating and managing Azure
resources. The Azure CLI is available for Windows, Linux, and Mac OS.
PowerShell. The Azure modules for PowerShell allow you to execute automated management tasks using a
script. PowerShell is available for Windows, Linux, and Mac OS.
Templates. Azure Resource Manager templates provide similar JSON template-based resource management
capabilities to the AWS CloudFormation service.
In each of these interfaces, the resource group is central to how Azure resources get created, deployed, or modified.
This is similar to the role a "stack" plays in grouping AWS resources during CloudFormation deployments.
The syntax and structure of these interfaces are different from their AWS equivalents, but they provide comparable
capabilities. In addition, many third-party management tools used on AWS, like Hashicorp's Terraform and Netflix
Spinnaker, are also available on Azure.

See also
Azure resource group guidelines
 Security and identity on Azure and AWS
11/2/2020 • 2 minutes to read • Edit Online

Directory service and Azure Active Directory

Azure splits up directory services into the following offerings:
Azure Active Directory: cloud-based directory and identity management service.
Azure Active Directory B2B: enables access to your corporate applications from partner-managed identities.
Azure Active Directory B2C: service offering support for single sign-on and user management for
consumer-facing applications.
Azure Active Directory Domain Services: hosted domain controller service, allowing Active Directory
compatible domain join and user management functionality.

Web application firewall

In addition to the Application Gateway Web Application Firewall, you can also use web application firewalls from
third-party vendors like Barracuda Networks.

See also
Getting started with Microsoft Azure security
Azure Identity Management and access control security best practices
 Comparing storage on Azure and AWS
11/2/2020 • 3 minutes to read • Edit Online

S3/EBS/EFS and Azure Storage

In the AWS platform, cloud storage is primarily broken down into three services:
Simple Storage Ser vice (S3) . Basic object storage that makes data available through an Internet
accessible API.
Elastic Block Storage (EBS) . Block level storage intended for access by a single VM.
Elastic File System (EFS) . File storage meant for use as shared storage for up to thousands of EC2
instances.
In Azure Storage, subscription-bound storage accounts allow you to create and manage the following storage
services:
Blob storage stores any type of text or binary data, such as a document, media file, or application installer.
You can set Blob storage for private access or share contents publicly to the Internet. Blob storage serves the
same purpose as both AWS S3 and EBS.
Table storage stores structured datasets. Table storage is a NoSQL key-attribute data store that allows for
rapid development and fast access to large quantities of data. Similar to AWS' SimpleDB and DynamoDB
services.
Queue storage provides messaging for workflow processing and for communication between components
of cloud services.
File storage offers shared storage for legacy applications using the standard server message block (SMB)
protocol. File storage is used in a similar manner to EFS in the AWS platform.

Glacier and Azure Storage

Azure Archive Blob Storage is comparable to AWS Glacier storage service. It is intended for rarely accessed data
that is stored for at least 180 days and can tolerate several hours of retrieval latency.
For data that is infrequently accessed but must be available immediately when accessed, Azure Cool Blob Storage
tier provides cheaper storage than standard blob storage. This storage tier is comparable to AWS S3 - Infrequent
Access storage service.

Storage comparison
Object storage
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Simple Storage Services (S3) Blob storage Object storage service, for use cases
including cloud applications, content
distribution, backup, archiving, disaster
recovery, and big data analytics.

Virtual server disks

 AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Elastic Block Store (EBS) managed disks SSD storage optimized for I/O intensive
read/write operations. For use as high-
performance Azure virtual machine
storage.

Shared files
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Elastic File System Files Provides a simple interface to create

and configure file systems quickly, and
share common files. Can be used with
traditional protocols that access files
over a network.

Archiving and backup

AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

S3 Infrequent Access (IA) Storage cool tier Cool storage is a lower-cost tier for
storing data that is infrequently
accessed and long-lived.

S3 Glacier, Deep Archive Storage archive access tier Archive storage has the lowest storage
cost and higher data retrieval costs
compared to hot and cool storage.

Backup Backup Back up and recover files and folders

from the cloud, and provide offsite
protection against data loss.

Hybrid storage
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Storage Gateway StorSimple Integrates on-premises IT environments

with cloud storage. Automates data
management and storage, plus
supports disaster recovery.

DataSync File Sync Azure Files can be deployed in two main

ways: by directly mounting the
serverless Azure file shares or by
caching Azure file shares on-premises
using Azure File Sync.

Bulk data transfer

AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Import/Export Disk Import/Export A data transport solution that uses

secure disks and appliances to transfer
large amounts of data. Also offers data
protection during transit.
 AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Import/Export Snowball, Snowball Edge, Data Box Petabyte- to exabyte-scale data

Snowmobile transport solution that uses secure data
storage devices to transfer large
amounts of data to and from Azure.

Storage architectures

HIPAA and HITRUST compliant health data AI

12/16/2019
2 min read
Manage HIPAA and HITRUST compliant health data and medical records with the highest level of built-in
security.

HPC Media Rendering

11/04/2020
2 min read
Optimize the media rendering process with a step-by-step HPC solution architecture from Azure that
combines Azure CycleCloud and HPC Cache.

Medical Data Storage Solutions

12/16/2019
2 min read
Store healthcare data effectively and affordably with cloud-based solutions from Azure. Manage medical
records with the highest level of built-in security.
view all

See also
Microsoft Azure Storage Performance and Scalability Checklist
Azure Storage security guide
Best practices for using content delivery networks (CDNs)
 AWS to Azure services comparison
12/18/2020 • 22 minutes to read • Edit Online

This article helps you understand how Microsoft Azure services compare to Amazon Web Services (AWS). Whether
you are planning a multicloud solution with Azure and AWS, or migrating to Azure, you can compare the IT
capabilities of Azure and AWS services in all categories.
This article compares services that are roughly comparable. Not every AWS service or Azure service is listed, and
not every matched service has exact feature-for-feature parity.

Azure and AWS for multicloud solutions

As the leading public cloud platforms, Azure and AWS each offer a broad and deep set of capabilities with global
coverage. Yet many organizations choose to use both platforms together for greater choice and flexibility, as well
as to spread their risk and dependencies with a multicloud approach. Consulting companies and software vendors
might also build on and use both Azure and AWS, as these platforms represent most of the cloud market demand.
For an overview of Azure for AWS users, see Introduction to Azure for AWS professionals.

Marketplace
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

AWS Marketplace Azure Marketplace Easy-to-deploy and automatically

configured third-party applications,
including single virtual machine or
multiple virtual machine solutions.

AI and machine learning

AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

SageMaker Machine Learning A cloud service to train, deploy,

automate, and manage machine
learning models.

Alexa Skills Kit Bot Framework Build and connect intelligent bots that
interact with your users using text/SMS,
Skype, Teams, Slack, Microsoft 365 mail,
Twitter, and other popular services.

Lex Speech Services API capable of converting speech to

text, understanding intent, and
converting text back to speech for
natural responsiveness.

Lex Language Understanding (LUIS) Allows your applications to understand

user commands contextually.

Polly, Transcribe Speech Services Enables both Speech to Text, and Text
into Speech capabilities.
 AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Rekognition Cognitive Services Computer Vision: Extract information

from images to categorize and process
visual data.

Face: Detect, identify, and analyze faces

in photos.

Emotions: Recognize emotions in

images.

Skills Kit Virtual Assistant The Virtual Assistant Template brings

together a number of best practices
we've identified through the building of
conversational experiences and
automates integration of components
that we've found to be highly beneficial
to Bot Framework developers.

AI and machine learning architectures

Image classification on Azure

7/05/2018
4 min read
Learn how to build image processing into your applications by using Azure services such as the Computer
Vision API and Azure Functions.

Predictive Marketing with Machine Learning

12/16/2019
2 min read
Learn how to build a machine-learning model with Microsoft R Server on Azure HDInsight Spark clusters to
recommend actions to maximize the purchase rate.

Scalable personalization on Azure

5/31/2019
6 min read
Use machine learning to automate content-based personalization for customers.
view all

Big data and analytics

Data warehouse
 AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Redshift Synapse Analytics Cloud-based Enterprise Data

Warehouse (EDW) that uses Massively
Parallel Processing (MPP) to quickly run
complex queries across petabytes of
data.

Lake Formation Data Share A simple and safe service for sharing
big data

Data warehouse architectures

Modern Data Warehouse Architecture

12/16/2019
2 min read
Explore a cloud data warehouse that uses big data. Modern data warehouse brings together all your data
and scales easily as your data grows.

Automated enterprise BI
6/03/2020
13 min read
Automate an extract, load, and transform (ELT) workflow in Azure using Azure Data Factory with Azure
Synapse Analytics.
view all
Big data processing
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

EMR Azure Data Explorer Fully managed, low latency, distributed

big data analytics platform to run
complex queries across petabytes of
data.

EMR Databricks Apache Spark-based analytics platform.

EMR HDInsight Managed Hadoop service. Deploy and

manage Hadoop clusters in Azure.

EMR Data Lake Storage Massively scalable, secure data lake

functionality built on Azure Blob
Storage.

Big data architectures

 Azure data platform end-to-end
1/31/2020
7 min read
Use Azure services to ingest, process, store, serve, and visualize data from different sources.

Campaign Optimization with Azure HDInsight Spark Clusters

12/16/2019
4 min read
This solution demonstrates how to build and deploy a machine learning model with Microsoft R Server on
Azure HDInsight Spark clusters to recommend actions to maximize the purchase rate of leads targeted by a
campaign. This solution enables efficient handling of big data on Spark with Microsoft R Server.
view all
Data orchestration / ETL
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Data Pipeline, Glue Data Factory Processes and moves data between
different compute and storage services,
as well as on-premises data sources at
specified intervals. Create, schedule,
orchestrate, and manage data pipelines.

Glue Data Catalog A fully managed service that serves as a

system of registration and system of
discovery for enterprise data sources

Dynamo DB Table Storage, Cosmos DB NoSQL key-value store for rapid

development using massive semi-
structured datasets.

Analytics and visualization

AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Kinesis Analytics Stream Analytics Storage and analysis platforms that

create insights from large quantities of
Azure Data Explorer data, or data that originates from many
sources.
Data Lake Analytics

Data Lake Store

QuickSight Power BI Business intelligence tools that build

visualizations, perform ad hoc analysis,
and develop business insights from
data.

CloudSearch Cognitive Search Delivers full-text search and related

search analytics and capabilities.
 AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Athena Data Lake Analytics Provides a serverless interactive query

service that uses standard SQL for
analyzing databases.

Analytics architectures

Advanced Analytics Architecture

12/16/2019
2 min read
Get near real-time data analytics on streaming services. This big data architecture allows you to combine
any data at any scale with custom machine learning.

Automated enterprise BI
6/03/2020
13 min read
Automate an extract, load, and transform (ELT) workflow in Azure using Azure Data Factory with Azure
Synapse Analytics.

Mass ingestion and analysis of news feeds on Azure

2/01/2019
5 min read
Create a pipeline for ingesting and analyzing text, images, sentiment, and other data from RSS news feeds
using only Azure services, including Azure Cosmos DB and Azure Cognitive Services.
view all

Compute
Virtual servers
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Elastic Compute Cloud (EC2) Instances Virtual Machines Virtual servers allow users to deploy,
manage, and maintain OS and server
software. Instance types provide
combinations of CPU/RAM. Users pay
for what they use with the flexibility to
change sizes.

Batch Batch Run large-scale parallel and high-

performance computing applications
efficiently in the cloud.
 AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Auto Scaling Virtual Machine Scale Sets Allows you to automatically change the
number of VM instances. You set
defined metric and thresholds that
determine if the platform adds or
removes instances.

VMware Cloud on AWS Azure VMware Solution Seamlessly move VMware-based

workloads from your datacenter to
Azure and integrate your VMware
environment with Azure. Keep
managing your existing environments
with the same VMware tools you
already know while you modernize your
applications with Azure native services.
Azure VMware Solution is a Microsoft
service, verified by VMware, that runs
on Azure infrastructure.

Parallel Cluster CycleCloud Create, manage, operate, and optimize

HPC and big compute clusters of any
scale

Containers and container orchestrators

AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Elastic Container Service (ECS) Container Instances Azure Container Instances is the fastest
and simplest way to run a container in
Fargate Azure, without having to provision any
virtual machines or adopt a higher-level
orchestration service.

Elastic Container Registry Container Registry Allows customers to store Docker

formatted images. Used to create all
types of container deployments on
Azure.

Elastic Kubernetes Service (EKS) Kubernetes Service (AKS) Deploy orchestrated containerized
applications with Kubernetes. Simplify
monitoring and cluster management
through auto upgrades and a built-in
operations console. See AKS solution
journey.

App Mesh Service Fabric Mesh Fully managed service that enables
developers to deploy microservices
applications without managing virtual
machines, storage, or networking.

Container architectures

Baseline architecture on Azure Kubernetes Service (AKS)

07/20/2020
37 min read
 Deploy a baseline infrastructure that deploys an AKS cluster with focus on security.

Microservices architecture on Azure Kubernetes Service (AKS)

5/07/2020
17 min read
Deploy a microservices architecture on Azure Kubernetes Service (AKS)

CI/CD pipeline for container-based workloads

7/05/2018
7 min read
Build a DevOps pipeline for a Node.js web app with Jenkins, Azure Container Registry, Azure Kubernetes
Service, Cosmos DB, and Grafana.
view all
Serverless
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Lambda Functions Integrate systems and run backend

processes in response to events or
schedules without provisioning or
managing servers.

Serverless architectures

Social App for Mobile and Web with Authentication

12/16/2019
3 min read
View a detailed, step-by-step diagram depicting the build process and implementation of the mobile client
app architecture that offers social image sharing with a companion web app and authentication abilities,
even while offline.

HIPAA and HITRUST compliant health data AI

12/16/2019
2 min read
Manage HIPAA and HITRUST compliant health data and medical records with the highest level of built-in
security.
 Cross Cloud Scaling Architecture
12/16/2019
1 min read
Learn how to improve cross cloud scalability with solution architecture that includes Azure Stack. A step-by-
step flowchart details instructions for implementation.

Database
TYPE AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Relational database RDS SQL Database Managed relational database

service where resiliency,
Database for MySQL scale, and maintenance are
primarily handled by the
Database for PostgreSQL platform.

NoSQL / Document DynamoDB Cosmos DB A globally distributed, multi-

model database that natively
SimpleDB supports multiple data
models: key-value,
Amazon DocumentDB documents, graphs, and
columnar.

Caching ElastiCache Cache for Redis An in-memory–based,

distributed caching service
that provides a high-
performance store typically
used to offload
nontransactional work from
a database.

Database migration Database Migration Service Database Migration Service Migration of database
schema and data from one
database format to a specific
database technology in the
cloud.

Database architectures

Gaming using Cosmos DB

12/16/2019
1 min read
Elastically scale your database to accommodate unpredictable bursts of traffic and deliver low-latency multi-
player experiences on a global scale.

Oracle Database Migration to Azure

12/16/2019
2 min read
 Oracle DB migrations can be accomplished in multiple ways. This architecture covers one of these options
wherein Oracle Active Data Guard is used to migrate the Database.

Retail and e-commerce using Azure MySQL

12/16/2019
1 min read
Build secure and scalable e-commerce solutions that meet the demands of both customers and business
using Azure Database for MySQL.
view all

DevOps and application monitoring

AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

CloudWatch, X-Ray Monitor Comprehensive solution for collecting,

analyzing, and acting on telemetry from
your cloud and on-premises
environments.

CodeDeploy DevOps A cloud service for collaborating on

code development.
CodeCommit

CodePipeline

Developer Tools Developer Tools Collection of tools for building,

debugging, deploying, diagnosing, and
managing multiplatform scalable apps
and services.

CodeBuild DevOps Fully managed build service that

supports continuous integration and
deployment.

Command Line Interface CLI Built on top of the native REST API
across all cloud services, various
PowerShell programming language-specific
wrappers provide easier ways to create
solutions.

OpsWorks (Chef-based) Automation Configures and operates applications of

all shapes and sizes, and provides
templates to create and manage a
collection of resources.

CloudFormation Resource Manager Provides a way for users to automate

the manual, long-running, error-prone,
VM extensions and frequently repeated IT tasks.

Azure Automation

Devops architectures
 Container CI/CD using Jenkins and Kubernetes on Azure Kubernetes Service (AKS)
12/16/2019
2 min read
Containers make it easy for you to continuously build and deploy applications. By orchestrating the
deployment of those containers using Azure Kubernetes Service (AKS), you can achieve replicable,
manageable clusters of containers.

Run a Jenkins server on Azure

11/19/2020
6 min read
Recommended architecture that shows how to deploy and operate a scalable, enterprise-grade Jenkins
server on Azure secured with single sign-on (SSO).

DevOps in a hybrid environment

12/16/2019
3 min read
The tools provided in Azure allow for the implementation of a DevOps strategy that capably manages both
cloud and on-premises environments in tandem.
view all

Internet of things (IoT)

AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

IoT IoT Hub A cloud gateway for managing

bidirectional communication with
billions of IoT devices, securely and at
scale.

Greengrass IoT Edge Deploy cloud intelligence directly on IoT

devices to run in on-premises scenarios.

Kinesis Firehose, Kinesis Streams Event Hubs Services that allow the mass ingestion
of small data inputs, typically from
devices and sensors, to process and
route the data.
 AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

IoT Things Graph Digital Twins Azure Digital Twins is an IoT service that
helps you create comprehensive models
of physical environments. Create spatial
intelligence graphs to model the
relationships and interactions between
people, places, and devices. Query data
from a physical space rather than
disparate sensors.

IOT architectures

IoT Architecture � Azure IoT Subsystems

12/16/2019
1 min read
Learn about our recommended IoT application architecture that supports hybrid cloud and edge computing.
A flowchart details how the subsystems function within the IoT application.

Azure IoT reference architecture

9/10/2020
12 min read
Recommended architecture for IoT applications on Azure using PaaS (platform-as-a-service) components

Process real-time vehicle data using IoT

11/17/2020
5 min read
This example builds a real-time data ingestion/processing pipeline to ingest and process messages from IoT
devices into a big data analytic platform in Azure.
view all

Management
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Trusted Advisor Advisor Provides analysis of cloud resource

configuration and security so
subscribers can ensure they're making
use of best practices and optimum
configurations.
 AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Usage and Billing Report Billing API Services to help generate, monitor,
forecast, and share billing data for
resource usage by time, organization, or
product resources.

Management Console Portal A unified management console that

simplifies building, deploying, and
operating your cloud resources.

Application Discovery Service Migrate Assesses on-premises workloads for

migration to Azure, performs
performance-based sizing, and provides
cost estimations.

EC2 Systems Manager Monitor Comprehensive solution for collecting,

analyzing, and acting on telemetry from
your cloud and on-premises
environments.

Personal Health Dashboard Resource Health Provides detailed information about the
health of resources as well as
recommended actions for maintaining
resource health.

CloudTrail Monitor Comprehensive solution for collecting,

analyzing, and acting on telemetry from
your cloud and on-premises
environments.

CloudWatch Application Insights Application Insights, is an extensible

Application Performance Management
(APM) service for developers and
DevOps professionals.

Cost Explorer Cost Management Optimize cloud costs while maximizing

cloud potential.

Messaging and eventing

AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Simple Queue Service (SQS) Queue Storage Provides a managed message queueing
service for communicating between
decoupled application components.

Simple Queue Service (SQS) Service Bus Supports a set of cloud-based,

message-oriented middleware
technologies including reliable message
queuing and durable publish/subscribe
messaging.

Amazon EventBridge Event Grid A fully managed event routing service

that allows for uniform event
consumption using a publish/subscribe
model.
Messaging architectures

Anomaly Detector Process

12/16/2019
1 min read
Learn more about Anomaly Detector with a step-by-step flowchart that details the process. See how
anomaly detection models are selected with time-series data.

Mobile services
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Mobile Hub App Center Provides backend mobile services for

rapid development of mobile solutions,
Xamarin Apps identity management, data
synchronization, and storage and
notifications across devices.

Mobile SDK App Center Provides the technology to rapidly build

cross-platform and native apps for
mobile devices.

Cognito App Center Provides authentication capabilities for

mobile applications.

Device Farm App Center Provides services to support testing

mobile applications.

Mobile Analytics App Center Supports monitoring, and feedback

collection for the debugging and
analysis of a mobile application service
quality.

Device Farm
The AWS Device Farm provides cross-device testing services. In Azure, Visual Studio App Center provides similar
cross-device front-end testing for mobile devices.
In addition to front-end testing, the Azure DevTest Labs provides back-end testing resources for Linux and
Windows environments.
Mobile architectures

Scalable web and mobile applications using Azure Database for PostgreSQL
12/16/2019
1 min read
Use Azure Database for PostgreSQL to rapidly build engaging, performant, and scalable cross-platform and
native apps for iOS, Android, Windows, or Mac.
 Social App for Mobile and Web with Authentication
12/16/2019
3 min read
View a detailed, step-by-step diagram depicting the build process and implementation of the mobile client
app architecture that offers social image sharing with a companion web app and authentication abilities,
even while offline.

Task-Based Consumer Mobile App

12/16/2019
3 min read
Learn how the task-based consumer mobile app architecture is created with a step-by-step flow chart that
shows the integration with Azure App Service Mobile Apps, Visual Studio, and Xamarin to simplify the build
process.
view all

Networking
A REA AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Cloud virtual networking Virtual Private Cloud (VPC) Virtual Network Provides an isolated, private
environment in the cloud.
Users have control over their
virtual networking
environment, including
selection of their own IP
address range, creation of
subnets, and configuration
of route tables and network
gateways.

Cross-premises connectivity VPN Gateway VPN Gateway Connects Azure virtual

networks to other Azure
virtual networks, or
customer on-premises
networks (Site To Site).
Allows end users to connect
to Azure services through
VPN tunneling (Point To
Site).

DNS management Route 53 DNS Manage your DNS records

using the same credentials
and billing and support
contract as your other Azure
services
 A REA AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

53 Traffic Manager A service that hosts domain

names, plus routes users to
Internet applications,
connects user requests to
datacenters, manages traffic
to apps, and improves app
availability with automatic
failover.

Dedicated network Direct Connect ExpressRoute Establishes a dedicated,

private network connection
from a location to the cloud
provider (not over the
Internet).

Load balancing Network Load Balancer Load Balancer Azure Load Balancer load
balances traffic at layer 4
(TCP or UDP). Standard Load
Balancer also supports
cross-region or global load
balancing.

Application Load Balancer Application Gateway Application Gateway is a

layer 7 load balancer. It
supports SSL termination,
cookie-based session affinity,
and round robin for load-
balancing traffic.

Networking architectures

Deploy highly available NVAs

12/08/2018
7 min read
Learn how to deploy network virtual appliances for high availability in Azure. This article includes example
architectures for ingress, egress, and both.

Hub-spoke network topology in Azure

9/30/2020
7 min read
Learn how to implement a hub-spoke topology in Azure, where the hub is a virtual network and the spokes
are virtual networks that peer with the hub.

Implement a secure hybrid network

1/07/2020
 9 min read
See a secure hybrid network that extends an on-premises network to Azure with a perimeter network
between the on-premises network and an Azure virtual network.
view all

Security, identity, and access

Authentication and authorization
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Identity and Access Management (IAM) Azure Active Directory Allows users to securely control access
to services and resources while offering
data security and protection. Create
and manage users and groups, and use
permissions to allow and deny access to
resources.

Identity and Access Management (IAM) Role Based Access Control Role-based access control (RBAC) helps
you manage who has access to Azure
resources, what they can do with those
resources, and what areas they have
access to.

Organizations Subscription Management + RBAC Security policy and role management

for working with multiple accounts.

Multi-Factor Authentication Multi-Factor Authentication Safeguard access to data and

applications while meeting user demand
for a simple sign-in process.

Directory Service Azure Active Directory Domain Services Provides managed domain services
such as domain join, group policy, LDAP,
and Kerberos/NTLM authentication that
are fully compatible with Windows
Server Active Directory.

Cognito Azure Active Directory B2C A highly available, global, identity

management service for consumer-
facing applications that scales to
hundreds of millions of identities.

Organizations Policy Azure Policy is a service in Azure that

you use to create, assign, and manage
policies. These policies enforce different
rules and effects over your resources, so
those resources stay compliant with
your corporate standards and service
level agreements.
 AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Organizations Management Groups Azure management groups provide a

level of scope above subscriptions. You
organize subscriptions into containers
called "management groups" and apply
your governance conditions to the
management groups. All subscriptions
within a management group
automatically inherit the conditions
applied to the management group.
Management groups give you
enterprise-grade management at a
large scale, no matter what type of
subscriptions you have.

Encryption
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Server-side encryption with Amazon S3 Azure Storage Service Encryption Helps you protect and safeguard your
Key Management Service data and meet your organizational
security and compliance commitments.

Key Management Service (KMS), Key Vault Provides security solution and works
CloudHSM with other services by providing a way
to manage, create, and control
encryption keys stored in hardware
security modules (HSM).

Firewall
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Web Application Firewall Web Application Firewall A firewall that protects web applications
from common web exploits.

Web Application Firewall Firewall Provides inbound protection for non-

HTTP/S protocols, outbound network-
level protection for all ports and
protocols, and application-level
protection for outbound HTTP/S.

Security
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Inspector Security Center An automated security assessment

service that improves the security and
compliance of applications.
Automatically assess applications for
vulnerabilities or deviations from best
practices.

Certificate Manager App Service Certificates available on the Service that allows customers to create,
Portal manage, and consume certificates
seamlessly in the cloud.
 AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

GuardDuty Advanced Threat Protection Detect and investigate advanced

attacks on-premises and in the cloud.

Artifact Service Trust Portal Provides access to audit reports,

compliance guides, and trust
documents from across cloud services.

Shield DDos Protection Service Provides cloud services with protection

from distributed denial of services
(DDoS) attacks.

Security architectures

Real-time fraud detection

7/05/2018
4 min read
Detect fraudulent activity in real-time using Azure Event Hubs and Stream Analytics.

Securely managed web applications

5/09/2019
8 min read
Learn about deploying secure applications using the Azure App Service Environment, the Azure Application
Gateway service, and Web Application Firewall.

Threat indicators for cyber threat intelligence in Azure Sentinel

4/13/2020
13 min read
Import threat indicators, view logs, create rules to generate security alerts and incidents, and visualize threat
intelligence data with Azure Sentinel.
view all

Storage
Object storage
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Simple Storage Services (S3) Blob storage Object storage service, for use cases
including cloud applications, content
distribution, backup, archiving, disaster
recovery, and big data analytics.
Virtual server disks
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Elastic Block Store (EBS) managed disks SSD storage optimized for I/O intensive
read/write operations. For use as high-
performance Azure virtual machine
storage.

Shared files
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Elastic File System Files Provides a simple interface to create

and configure file systems quickly, and
share common files. Can be used with
traditional protocols that access files
over a network.

Archiving and backup

AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

S3 Infrequent Access (IA) Storage cool tier Cool storage is a lower-cost tier for
storing data that is infrequently
accessed and long-lived.

S3 Glacier, Deep Archive Storage archive access tier Archive storage has the lowest storage
cost and higher data retrieval costs
compared to hot and cool storage.

Backup Backup Back up and recover files and folders

from the cloud, and provide offsite
protection against data loss.

Hybrid storage
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Storage Gateway StorSimple Integrates on-premises IT environments

with cloud storage. Automates data
management and storage, plus
supports disaster recovery.

DataSync File Sync Azure Files can be deployed in two main

ways: by directly mounting the
serverless Azure file shares or by
caching Azure file shares on-premises
using Azure File Sync.

Bulk data transfer

AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Import/Export Disk Import/Export A data transport solution that uses

secure disks and appliances to transfer
large amounts of data. Also offers data
protection during transit.
 AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Import/Export Snowball, Snowball Edge, Data Box Petabyte- to exabyte-scale data

Snowmobile transport solution that uses secure data
storage devices to transfer large
amounts of data to and from Azure.

Storage architectures

HIPAA and HITRUST compliant health data AI

12/16/2019
2 min read
Manage HIPAA and HITRUST compliant health data and medical records with the highest level of built-in
security.

HPC Media Rendering

11/04/2020
2 min read
Optimize the media rendering process with a step-by-step HPC solution architecture from Azure that
combines Azure CycleCloud and HPC Cache.

Medical Data Storage Solutions

12/16/2019
2 min read
Store healthcare data effectively and affordably with cloud-based solutions from Azure. Manage medical
records with the highest level of built-in security.
view all

Web applications
AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Elastic Beanstalk App Service Managed hosting platform providing

easy to use services for deploying and
scaling web applications and services.

API Gateway API Management A turnkey solution for publishing APIs

to external and internal consumers.

CloudFront Content Delivery Network A global content delivery network that

delivers audio, video, applications,
images, and other files.
 AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Global Accelerator Front Door Easily join your distributed

microservices architectures into a single
global application using HTTP load
balancing and path-based routing rules.
Automate turning up new regions and
scale-out with API-driven global
actions, and independent fault-
tolerance to your back end
microservices in Azure-or anywhere.

Global Accelerator Cross-regional load balancer Distribute and load balance traffic
across multiple Azure regions via a
single, static, global anycast public IP
address.

LightSail App Service Build, deploy, and scale web apps on a

fully managed platform.

Web architectures

Architect scalable e-commerce web app

12/16/2019
1 min read
The e-commerce website includes simple order processing workflows with the help of Azure services. Using
Azure Functions and Web Apps, developers can focus on building personalized experiences and let Azure
take care of the infrastructure.

Multi-region N-tier application

6/18/2019
10 min read
Deploy an application on Azure virtual machines in multiple regions for high availability and resiliency.

Serverless web application

5/28/2019
16 min read
This reference architecture shows a serverless web application, which serves static content from Azure Blob
Storage and implements an API using Azure Functions.
view all

Miscellaneous
 A REA AW S SERVIC E A Z URE SERVIC E DESC RIP T IO N

Backend process logic Step Functions Logic Apps Cloud technology to build
distributed applications
using out-of-the-box
connectors to reduce
integration challenges.
Connect apps, data and
devices on-premises or in
the cloud.

Enterprise application WorkMail, WorkDocs Microsoft 365 Fully integrated Cloud

services service providing
communications, email,
document management in
the cloud and available on a
wide variety of devices.

Gaming GameLift, GameSparks PlayFab Managed services for

hosting dedicated game
servers.

Media transcoding Elastic Transcoder Media Services Services that offer

broadcast-quality video
streaming services, including
various transcoding
technologies.

Workflow Simple Workflow Service Logic Apps Serverless technology for

(SWF) connecting apps, data and
devices anywhere, whether
on-premises or in the cloud
for large ecosystems of SaaS
and cloud-based connectors.

Hybrid Outposts Stack Azure Stack is a hybrid cloud

platform that enables you to
run Azure services in your
company's or service
provider's datacenter. As a
developer, you can build
apps on Azure Stack. You
can then deploy them to
either Azure Stack or Azure,
or you can build truly hybrid
apps that take advantage of
connectivity between an
Azure Stack cloud and Azure.

Media Elemental MediaConvert Media Services Cloud-based media

workflow platform to index,
package, protect, and stream
video at scale.

More learning
If you are new to Azure, review the interactive Core Cloud Services - Introduction to Azure module on Microsoft
Learn.
 Azure for GCP Professionals
12/18/2020 • 8 minutes to read • Edit Online

This article helps Google Cloud Platform (GCP) experts understand the basics of Microsoft Azure accounts,
platform, and services. It also covers key similarities and differences between the GCP and Azure platforms.
You'll learn:
How accounts and resources are organized in Azure.
How available solutions are structured in Azure.
How the major Azure services differ from GCP services.
Azure and GCP built their capabilities independently over time so that each has important implementation and
design differences.

Overview
Like GCP, Microsoft Azure is built around a core set of compute, storage, database, and networking services. In
many cases, both platforms offer a basic equivalence between the products and services they offer. Both GCP and
Azure allow you to build highly available solutions based on Linux or Windows hosts. So, if you're used to
development using Linux and OSS technology, both platforms can do the job.
While the capabilities of both platforms are similar, the resources that provide those capabilities are often
organized differently. Exact one-to-one relationships between the services required to build a solution are not
always clear. In other cases, a particular service might be offered on one platform, but not the other.

Managing accounts and subscription

Azure has a hierarchy of Management group and subscriptions and resource groups to manage resources
effectively. This is similar to the Folders and Project hierarchy for resources in Google Cloud.

Azure levels of management scope

Management groups: These groups are containers that help you manage access, policy, and compliance for
multiple subscriptions. All subscriptions in a management group automatically inherit the conditions applied to
the management group.
Subscriptions: A subscription logically associates user accounts and the resources that were created by those
user accounts. Each subscription has limits or quotas on the amount of resources you can create and use.
 Organizations can use subscriptions to manage costs and the resources that are created by users, teams, or
projects.
Resource groups: A resource group is a logical container into which Azure resources like web apps, databases,
and storage accounts are deployed and managed.
Resources: Resources are instances of services that you create, like virtual machines, storage, or SQL
databases.
Azure services can be purchased using several pricing options, depending on your organization's size and needs.
See the pricing overview page for details.
Azure subscriptions are a grouping of resources with an assigned owner responsible for billing and permissions
management.
Subscriptions are assigned three types of administrator accounts:
Account Administrator . The subscription owner and the account billed for the resources used in the
subscription. The account administrator can only be changed by transferring ownership of the subscription.
Ser vice Administrator . This account has rights to create and manage resources in the subscription but is not
responsible for billing. By default, the account administrator and service administrator are assigned to the same
account. The account administrator can assign a separate user to the service administrator account for
managing the technical and operational aspects of a subscription. Only one service administrator is assigned
per subscription.
Co-administrator . There can be multiple co-administrator accounts assigned to a subscription. Co-
administrators cannot change the service administrator, but otherwise have full control over subscription
resources and users.
Below the subscription level user roles and individual permissions can also be assigned to specific resources. In
Azure, all user accounts are associated with either a Microsoft Account or Organizational Account (an account
managed through an Azure Active Directory).
Subscriptions have default service quotas and limits. For a full list of these limits, see Azure subscription and
service limits, quotas, and constraints. These limits can be increased up to the maximum by filing a support request
in the management portal.
See also
How to add or change Azure administrator roles
How to download your Azure billing invoice and daily usage data

Resource management
The term "resource" in Azure means any compute instance, storage object, networking device, or other entity you
can create or configure within the platform.
Azure resources are deployed and managed using one of two models: Azure Resource Manager, or the older Azure
classic deployment model. Any new resources are created using the Resource Manager model.
Resource groups
Azure additionally has an entity called "resource groups" that organize resources such as VMs, storage, and virtual
networking devices. An Azure resource is always associated with one resource group. A resource created in one
resource group can be moved to another group but can only be in one resource group at a time. Resource groups
are the fundamental grouping used by Azure Resource Manager.
Resources can also be organized using tags. Tags are key-value pairs that allow you to group resources across your
subscription irrespective of resource group membership.
Management interfaces
Azure offers several ways to manage your resources:
Web interface. The Azure portal provides a full web-based management interface for Azure resources.
REST API. The Azure Resource Manager REST API provides programmatic access to most of the features
available in the Azure portal.
Command Line. The Azure CLI provides a command-line interface capable of creating and managing Azure
resources. The Azure CLI is available for Windows, Linux, and Mac OS.
PowerShell. The Azure modules for PowerShell allow you to execute automated management tasks using a
script. PowerShell is available for Windows, Linux, and Mac OS.
Templates. Azure Resource Manager templates provide JSON template-based resource management
capabilities.
In each of these interfaces, the resource group is central to how Azure resources get created, deployed, or modified.
In addition, many third-party management tools like Hashicorp's Terraform and Netflix Spinnaker, are also
available on Azure.
See also
Azure resource group guidelines

Regions and zones (high availability)

Failures can vary in the scope of their impact. Some hardware failures, such as a failed disk, may affect a single host
machine. A failed network switch could affect a whole server rack. Less common are failures that disrupt a whole
datacenter, such as loss of power in a datacenter. Rarely, an entire region could become unavailable.
One of the main ways to make an application resilient is through redundancy. But you need to plan for this
redundancy when you design the application. Also, the level of redundancy that you need depends on your
business requirements. Not every application needs redundancy across regions to guard against a regional outage.
In general, a tradeoff exists between greater redundancy and reliability versus higher cost and complexity.
In GCP, a region has two or more Availability Zones. An Availability Zone corresponds with a physically isolated
datacenter in the geographic region. Azure has numerous features for providing application redundancy at every
level of potential failure, including availability sets , availability zones , and paired regions .
The following table summarizes each option.

AVA IL A B IL IT Y SET AVA IL A B IL IT Y Z O N E PA IRED REGIO N

Scope of failure Rack Datacenter Region

Request routing Load Balancer Cross-zone Load Balancer Traffic Manager

Network latency Very low Low Mid to high

Vir tual networking VNet VNet Cross-region VNet peering

Availability sets
To protect against localized hardware failures, such as a disk or network switch failing, deploy two or more VMs in
an availability set. An availability set consists of two or more fault domains that share a common power source and
network switch. VMs in an availability set are distributed across the fault domains, so if a hardware failure affects
one fault domain, network traffic can still be routed the VMs in the other fault domains. For more information
about Availability Sets, see Manage the availability of Windows virtual machines in Azure.
When VM instances are added to availability sets, they are also assigned an update domain. An update domain is a
group of VMs that are set for planned maintenance events at the same time. Distributing VMs across multiple
update domains ensures that planned update and patching events affect only a subset of these VMs at any given
time.
Availability sets should be organized by the instance's role in your application to ensure one instance in each role is
operational. For example, in a three-tier web application, create separate availability sets for the front-end,
application, and data tiers.

Availability sets
Availability zones
Like GCP, Azure regions can have Availability zones. An Availability Zone is a physically separate zone within an
Azure region. Each Availability Zone has a distinct power source, network, and cooling. Deploying VMs across
availability zones helps to protect an application against datacenter-wide failures.

Zone redundant VM deployment on Azure

Paired regions
To protect an application against a regional outage, you can deploy the application across multiple regions, using
Azure Traffic Manager to distribute internet traffic to the different regions. Each Azure region is paired with another
region. Together, these form a regional pair. With the exception of Brazil South, regional pairs are located within the
same geography in order to meet data residency requirements for tax and law enforcement jurisdiction purposes.
Unlike Availability Zones, which are physically separate datacenters but may be in relatively nearby geographic
areas, paired regions are typically separated by at least 300 miles. This design ensures that large-scale disasters
only affect one of the regions in the pair. Neighboring pairs can be set to sync database and storage service data,
and are configured so that platform updates are rolled out to only one region in the pair at a time.
Azure geo-redundant storage is automatically backed up to the appropriate paired region. For all other resources,
creating a fully redundant solution using paired regions means creating a full copy of your solution in both regions.

Region Pairs in Azure

See also
Regions for virtual machines in Azure
Availability options for virtual machines in Azure
High availability for Azure applications
Failure and disaster recovery for Azure applications
Planned maintenance for Linux virtual machines in Azure

Services
For a listing of how services map between platforms, see GCP to Azure services comparison.
Not all Azure products and services are available in all regions. Consult the Products by Region page for details. You
can find the uptime guarantees and downtime credit policies for each Azure product or service on the Service Level
Agreements page.

Next steps
Get started with Azure
Azure solution architectures
Azure Reference Architectures
 GCP to Azure services comparison
12/18/2020 • 20 minutes to read • Edit Online

This article helps you understand how Microsoft Azure services compare to Google Cloud Platform (GCP). Whether
you are planning a multi-cloud solution with Azure and GCP, or migrating to Azure, you can compare the IT
capabilities of Azure and GCP services in all categories.
This article compares services that are roughly comparable. Not every GCP service or Azure service is listed, and
not every matched service has exact feature-for-feature parity.
For an overview of Azure for GCP users, see Introduction to Azure for GCP Professionals.

Marketplace
GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

GCP Marketplace Azure Marketplace Easy-to-deploy and automatically

configured third-party applications,
including single virtual machine or
multiple virtual machine solutions.

AI and Machine Learning

GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

AI Hub Azure Machine Learning A cloud service to train, deploy,

automate, and manage machine
learning models.

TensorFlow ML.NET ML.NET is an open source and cross-

platform machine learning framework
for both machine learning & AI.

TensorFlow ONNX (Open Neural Network ONNX is an open format built to

Exchange) represent machine learning models that
facilitates maximum compatibility and
increased inference performance.

AI Building blocks - Sight Azure Cognitive Services Computer Use visual data processing to label
Vision content, from objects to concepts,
extract printed and handwritten text,
recognize familiar subjects like brands
and landmarks, and moderate content.
No machine learning expertise is
required.

AI Building blocks - Language Azure Cognitive Services Text Analytics Cloud-based services that provides
advanced natural language processing
over raw text, and includes four main
functions: sentiment analysis, key
phrase extraction, language detection,
and named entity recognition.
GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

AI Building blocks - Language Azure Cognitive Services Language A machine learning-based service to
Understanding (LUIS) build natural language understanding
into apps, bots, and IoT devices. Quickly
create enterprise-ready, custom models
that continuously improve.

AI Building blocks - Conversation Azure Cognitive Services Speech To Text Swiftly convert audio into text from a
variety of sources. Customize models to
overcome common speech recognition
barriers, such as unique vocabularies,
speaking styles, or background noise.

AI Building blocks – Structured Data Azure ML - Automated Machine Empower professional and non-
Learning professional data scientists to build
machine learning models rapidly.
Automate time-consuming and iterative
tasks of model development using
breakthrough research-and accelerate
time to market. Available in Azure
Machine learning, Power BI, ML.NET &
Visual Studio.

AI Building blocks – Structured Data ML.NET Model Builder ML.NET Model Builder provides an easy
to understand visual interface to build,
train, and deploy custom machine
learning models. Prior machine learning
expertise is not required. Model Builder
supports AutoML, which automatically
explores different machine learning
algorithms and settings to help you find
the one that best suits your scenario.

AI Building blocks – Cloud AutoML Azure Cognitive Services Custom Vision Customize and embed state-of-the-art
computer vision for specific domains.
Build frictionless customer experiences,
optimize manufacturing processes,
accelerate digital marketing campaigns-
and more. No machine learning
expertise is required.

AI Building blocks – Cloud AutoML Video Indexer Easily extract insights from your videos
and quickly enrich your applications to
enhance discovery and engagement.

AI Building blocks – Cloud AutoML Azure Cognitive Services QnA Maker Build, train and publish a sophisticated
bot using FAQ pages, support websites,
product manuals, SharePoint
documents or editorial content through
an easy-to-use UI or via REST APIs.

AI Platform Notebooks Azure Notebooks Develop and run code from anywhere
with Jupyter notebooks on Azure.

Deep Learning VM Image Data Science Virtual Machines Pre-Configured environments in the
cloud for Data Science and AI
Development.
GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Deep Learning Containers GPU support on Azure Kubernetes Graphical processing units (GPUs) are
Service (AKS) often used for compute-intensive
workloads such as graphics and
visualization workloads. AKS supports
the creation of GPU-enabled node
pools to run these compute-intensive
workloads in Kubernetes.

Data Labeling Service Azure ML - Data Labeling A central place to create, manage, and
monitor labeling projects (public
preview). Use it to coordinate data,
labels, and team members to efficiently
manage labeling tasks. Machine
Learning supports image classification,
either multi-label or multi-class, and
object identification with bounded
boxes.

AI Platform Training Azure ML – Compute Targets Designated compute

resource/environment where you run
your training script or host your service
deployment. This location may be your
local machine or a cloud-based
compute resource. Using compute
targets make it easy for you to later
change your compute environment
without having to change your code.

AI Platform Predictions Azure ML - Deployments Deploy your machine learning model as

a web service in the Azure cloud or to
Azure IoT Edge devices. Leverage
serverless Azure Functions for model
inference for dynamic scale.

Continuous Evaluation Azure ML – Data Drift Monitor for data drift between the
training dataset and inference data of a
deployed model. In the context of
machine learning, trained machine
learning models may experience
degraded prediction performance
because of drift. With Azure Machine
Learning, you can monitor data drift
and the service can send an email alert
to you when drift is detected.

What-If Tool Azure ML – Model Interpretability Ensure machine learning model

compliance with company policies,
industry standards, and government
regulations.

Cloud TPU Azure ML – FPGA (Field Programmable FPGAs contain an array of

Gate Arrays) programmable logic blocks, and a
hierarchy of reconfigurable
interconnects. The interconnects allow
these blocks to be configured in various
ways after manufacturing. Compared to
other chips, FPGAs provide a
combination of programmability and
performance.
 GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Kubeflow Machine Learning Operations (MLOps) MLOps, or DevOps for machine

learning, enables data science and IT
teams to collaborate and increase the
pace of model development and
deployment via monitoring, validation,
and governance of machine learning
models.

Dialogflow Microsoft Bot Framework Build and connect intelligent bots that
interact with your users using text/SMS,
Skype, Teams, Slack, Microsoft 365 mail,
Twitter, and other popular services.

Big data and analytics

Data warehouse
GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

BigQuery Azure Synapse Analytics Cloud-based Enterprise Data

Warehouse (EDW) that uses Massively
Parallel Processing (MPP) to quickly run
complex queries across petabytes of
data.

BigQuery SQL Server Big Data Clusters Allow you to deploy scalable clusters of
SQL Server, Spark, and HDFS containers
running on Kubernetes. These
components are running side by side to
enable you to read, write, and process
big data from Transact-SQL or Spark,
allowing you to easily combine and
analyze your high-value relational data
with high-volume big data.

Big data processing

GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Dataproc Azure HDInsight Managed Apache Spark-based analytics

platform.

Data Orchestration and ETL

GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Cloud Data Fusion Azure Data Factory Processes and moves data between
different compute and storage services,
as well as on-premises data sources at
specified intervals. Create, schedule,
orchestrate, and manage data pipelines.

Cloud Data Catalog Azure Data Catalog A fully managed service that serves as a
system of registration and system of
discovery for enterprise data sources
Analytics and visualization
GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Cloud Dataflow Azure Databricks Managed platform for streaming batch

data based on Open Source Apache
products.

Datastudio Power BI Business intelligence tools that build

visualizations, perform ad hoc analysis,
Looker and develop business insights from
data.

Cloud Search Azure Search Delivers full-text search and related

search analytics and capabilities.

BigQuery SQL Server – ML Services Provides a serverless non-cloud

interactive query service that uses
Big Data Clusters (Spark) standard SQL for analyzing databases.

SQL Server Analysis Services

Compute
Virtual servers
GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Compute Engine Azure Virtual Machines Virtual servers allow users to deploy,
manage, and maintain OS and server
software. Instance types provide
combinations of CPU/RAM. Users pay
for what they use with the flexibility to
change sizes.

Batch Azure Batch Run large-scale parallel and high-

performance computing applications
efficiently in the cloud.

Compute Engine Managed Instance Azure Virtual Machine Scale Sets Allows you to automatically change the
Groups number of VM instances. You set
defined metric and thresholds that
determine if the platform adds or
removes instances.

VMware as a service Azure VMware by CloudSimple Redeploy and extend your VMware-
based enterprise workloads to Azure
with Azure VMware Solution by
CloudSimple. Keep using the VMware
tools you already know to manage
workloads on Azure without disrupting
network, security, or data protection
policies.

Containers and container orchestrators

 GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Cloud Run Azure Container Instances Azure Container Instances is the fastest
and simplest way to run a container in
Azure, without having to provision any
virtual machines or adopt a higher-level
orchestration service.

Artifact Registry (beta) Azure Container Registry Allows customers to store Docker
formatted images. Used to create all
Container Registry types of container deployments on
Azure.

Kubernetes Engine (GKE) Azure Kubernetes Service (AKS) Deploy orchestrated containerized
applications with Kubernetes. Simplify
cluster management and monitoring
through automatic upgrades and a
built-in operations console. See AKS
solution journey.

Kubernetes Engine Monitoring Azure Monitor for containers Azure Monitor for containers is a
feature designed to monitor the
performance of container workloads
deployed to: Managed Kubernetes
clusters hosted on Azure Kubernetes
Service (AKS); Self-managed Kubernetes
clusters hosted on Azure using AKS
Engine; Azure Container Instances, Self-
managed Kubernetes clusters hosted on
Azure Stack or on-premises; or Azure
Red Hat OpenShift.

Anthos Service Mesh Service Fabric Mesh Fully managed service that enables
developers to deploy microservices
applications without managing virtual
machines, storage, or networking.

Container architectures
Here are some architectures that use AKS as the orchestrator.

Baseline architecture on Azure Kubernetes Service (AKS)

07/20/2020
37 min read
Deploy a baseline infrastructure that deploys an AKS cluster with focus on security.

Microservices architecture on Azure Kubernetes Service (AKS)

5/07/2020
17 min read
Deploy a microservices architecture on Azure Kubernetes Service (AKS)
 CI/CD pipeline for container-based workloads
7/05/2018
7 min read
Build a DevOps pipeline for a Node.js web app with Jenkins, Azure Container Registry, Azure Kubernetes
Service, Cosmos DB, and Grafana.
view all
Functions
GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Cloud Functions Azure Functions Integrate systems and run backend

processes in response to events or
schedules without provisioning or
managing servers.

Database
TYPE GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Relational database (PaaS) Cloud SQL ( SQL Server, SQL Server Options Managed relational database
MySQL, PostgreSQL ) service where resiliency,
Azure Database for MySQL security, scale, and
maintenance are primarily
Azure Database for handled by the platform.
PostgreSQL

Cloud Spanner Azure Cosmos DB Managed relational database

service where resiliency,
security, scale, and
maintenance are primarily
handled by the platform.

NoSQL (PaaS) Cloud Bigtable Azure Cosmos DB Globally distributed, multi-

model database that natively
Cloud Firestore supports multiple data
models: key-value,
Firebase Realtime Database documents, graphs, and
columnar.

Caching Cloud Memorystore Azure Cache for Redis An in-memory–based,

distributed caching service
Redis Enterprise Cloud that provides a high-
performance store typically
used to offload non-
transactional work from a
database.

DevOps and application monitoring

GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N
GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Operations (formerly Stackdriver) Azure Monitor Maximizes the availability and

performance of your applications and
services by delivering a comprehensive
solution for collecting, analyzing, and
acting on telemetry from your cloud
and on-premises environments. It helps
you understand how your applications
are performing and proactively identifies
issues affecting them and the resources
on which they depend.

Cloud Trace Azure Monitor Maximizes the availability and

performance of your applications and
services by delivering a comprehensive
solution for collecting, analyzing, and
acting on telemetry from your cloud
and on-premises environments. It helps
you understand how your applications
are performing and proactively identifies
issues affecting them and the resources
on which they depend.

Cloud Debugger Azure Monitor Maximizes the availability and

performance of your applications and
services by delivering a comprehensive
solution for collecting, analyzing, and
acting on telemetry from your cloud
and on-premises environments. It helps
you understand how your applications
are performing and proactively identifies
issues affecting them and the resources
on which they depend.

Cloud Profiler Azure Monitor Maximizes the availability and

performance of your applications and
services by delivering a comprehensive
solution for collecting, analyzing, and
acting on telemetry from your cloud
and on-premises environments. It helps
you understand how your applications
are performing and proactively identifies
issues affecting them and the resources
on which they depend.

Cloud Source Repositories Azure DevOps Repos, GitHub Repos A cloud service for collaborating on
code development.

Cloud Build Azure DevOps Pipelines, GitHub Actions Fully managed build service that
supports continuous integration and
deployment.

Artifact Registry Azure DevOps Artifacts, GitHub Add fully integrated package
Packages management to your continuous
integration/continuous delivery (CI/CD)
pipelines with a single click. Create and
share Maven, npm, NuGet, and Python
package feeds from public and private
sources with teams of any size.
 GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Cloud Developer Tools (including Cloud Azure Developer Tools Collection of tools for building,
Code) debugging, deploying, diagnosing, and
managing multiplatform scalable apps
and services.

Gcloud SDK Azure CLI The Azure command-line interface

(Azure CLI) is a set of commands used
to create and manage Azure resources.
The Azure CLI is available across Azure
services and is designed to get you
working quickly with Azure, with an
emphasis on automation.

Cloud Shell Azure Cloud Shell Azure Cloud Shell is an interactive,

authenticated, browser-accessible shell
for managing Azure resources. It
provides the flexibility of choosing the
shell experience that best suits the way
you work, either Bash or PowerShell.

PowerShell on GCP Azure PowerShell Azure PowerShell is a set of cmdlets for

managing Azure resources directly from
the PowerShell command line. Azure
PowerShell is designed to make it easy
to learn and get started with, but
provides powerful features for
automation. Written in .NET Standard,
Azure PowerShell works with PowerShell
5.1 on Windows, and PowerShell 6.x and
higher on all platforms.

Cloud Deployment Manager Azure Automation Delivers a cloud-based automation and

configuration service that supports
consistent management across your
Azure and non-Azure environments. It
comprises process automation,
configuration management, update
management, shared capabilities, and
heterogeneous features. Automation
gives you complete control during
deployment, operations, and
decommissioning of workloads and
resources.

Cloud Deployment Manager Azure Resource Manager Provides a way for users to automate
the manual, long-running, error-prone,
and frequently repeated IT tasks.

Internet of things (IoT)

GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Cloud IoT Core Azure IoT Hub,Azure Event Hubs A cloud gateway for managing
bidirectional communication with
billions of IoT devices, securely and at
scale.
 GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Cloud Pub/Sub Azure Stream Analytics,HDInsight Kafka Process and route streaming data to
subsequent processing engine or
storage or database platform.

Edge Tpu Azure IoT Edge Deploy cloud intelligence directly on IoT
devices to run in on-premises scenarios.

Management
GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Cloud Billing Azure Billing API Services to help generate, monitor,

forecast, and share billing data for
resource usage by time, organization, or
product resources.

Cloud Console Azure portal A unified management console that

simplifies building, deploying, and
operating your cloud resources.

Operations (formerly Stackdriver) Azure Monitor Comprehensive solution for collecting,

analyzing, and acting on telemetry from
your cloud and on-premises
environments.

Messaging and eventing

GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Cloud Pub/Sub Azure Service Bus Supports a set of cloud-based,

message-oriented middleware
technologies including reliable message
queuing and durable publish/subscribe
messaging.

Cloud Pub/Sub Azure Event Grid A fully managed event routing service
that allows for uniform event
consumption using a publish/subscribe
model.

Networking
A REA GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N
A REA GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Cloud virtual networking Virtual Private Network Azure Virtual Network (Vnet) Provides an isolated, private
(VPC) environment in the cloud.
Users have control over their
virtual networking
environment, including
selection of their own IP
address range,
adding/updating address
ranges, creation of subnets,
and configuration of route
tables and network
gateways.

DNS management Cloud DNS Azure DNS Manage your DNS records
using the same credentials
and billing and support
contract as your other Azure
services

Cloud DNS Azure Traffic Manager Azure Traffic Manager is a

DNS-based load balancer
that enables you to
distribute traffic optimally to
services across global Azure
regions, while providing high
availability and
responsiveness.

Hybrid Connectivity Cloud Interconnect Azure ExpressRoute Establishes a private network

connection from a location
to the cloud provider (not
over the Internet).

Cloud VPN Gateway Azure Virtual Network Connects Azure virtual

Gateway networks to other Azure
virtual networks, or
customer on-premises
networks (site-to-site).
Allows end users to connect
to Azure services through
VPN tunneling (point-to-
site).

Cloud VPN Gateway Azure Virtual WAN Azure virtual WAN simplifies
large scale branch
connectivity with VPN and
ExpressRoute.

Cloud router Azure Virtual Network Enables dynamic routes

Gateway exchange using BGP.

Load balancing Network Load Balancing Azure Load Balancer Azure Load Balancer load-
balances traffic at layer 4 (all
TCP or UDP).
A REA GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Global load balancing Azure Front door Azure front door enables
global load balancing across
regions using a single
anycast IP.

Global load balancing Azure Application Gateway Application Gateway is a

layer 7 load balancer. IT
takes backends with any IP
that is reachable. It supports
SSL termination, cookie-
based session affinity, and
round robin for load-
balancing traffic.

Global load balancing Azure Traffic Manager Azure Traffic Manager is a

DNS-based load balancer
that enables you to
distribute traffic optimally to
services across global Azure
regions, while providing high
availability and
responsiveness.

Content delivery network Cloud CDN Azure CDN A content delivery network
(CDN) is a distributed
network of servers that can
efficiently deliver web
content to users.

Firewall Firewall rules Application security groups Azure Application security

groups allows you to group
virtual machines and define
network security policies
based on those groups.

Firewall rules Network Security groups Azure network security

group filters network traffic
to and from Azure resources
in an Azure virtual network.

Firewall rules Azure Firewall Azure Firewall is a managed,

cloud-based network
security service that protects
your Azure Virtual Network
resources. It's a fully stateful
firewall as a service with
built-in high availability and
unrestricted cloud scalability.

Web Application Firewall Cloud Armor Application Gateway - Web Azure Web Application
Application Firewall Firewall (WAF) provides
centralized protection of
your web applications from
common exploits and
vulnerabilities.
A REA GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Cloud Armor Front door – Azure Web Azure Web Application

Application Firewall Firewall (WAF) on Azure
Front Door provides
centralized protection for
your web applications.

Cloud Armor CDN – Azure Web Azure Web Application

Application Firewall Firewall (WAF) on Azure
Content Delivery Network
(CDN) from Microsoft
provides centralized
protection for your web
content.

NAT Gateway Cloud NAT Azure Virtual Network NAT Virtual Network NAT
(network address
translation) provides
outbound NAT translations
for internet connectivity for
virtual networks.

Private Connectivity to PaaS VPC Service controls Azure Private Link Azure Private Link enables
you to access Azure PaaS
Services and Azure hosted
customer-owned/partner
services over a private
endpoint in your virtual
network.

Telemetry VPC Flow logs NSG Flow logs Network security group
(NSG) flow logs are a feature
of Network Watcher that
allows you to view
information about ingress
and egress IP traffic through
an NSG.

Firewall Rules Logging NSG Flow logs Network security group

(NSG) flow logs are a feature
of Network Watcher that
allows you to view
information about ingress
and egress IP traffic through
an NSG.

Operations (formerly Azure Monitor Azure Monitor delivers a

Stackdriver) comprehensive solution for
collecting, analyzing, and
acting on telemetry from
your cloud and on-premises
environments. Log queries
help you to fully leverage
the value of the data
collected in Azure Monitor
Logs.
 A REA GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Network Intelligence Center Azure Network Watcher Azure Network Watcher

provides tools to monitor,
diagnose, view metrics, and
enable or disable logs for
resources in an Azure virtual
network.

Other Connectivity Options S2S,P2S Direct Interconnect,Partner Point to Site lets you create
Interconnect,Carrier Peering a secure connection to your
virtual network from an
individual client computer.
Site to Site is a connection
between two or more
networks, such as a
corporate network and a
branch office network.

Security
A REA GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Authentication and Cloud IAM Azure Active Directory Allows users to securely
Authorization control access to services
and resources while offering
data security and protection.
Create and manage users
and groups and use
permissions to allow and
deny access to resources.

Cloud IAM Azure Role Based Access Role-based access control

Control (RBAC) helps you manage
who has access to Azure
resources, what they can do
with those resources, and
what areas they have access
to.

Resource Manager Azure Subscription Structure to organize and

Management manage assets in Azure.

Multi-factor Authentication Azure Active Directory Safeguard access to data

Multi-factor Authentication and applications while
meeting user demand for a
simple sign-in process.

Firebase Authentication Azure Active Directory B2C A highly available, global,

identity management service
for consumer-facing
applications that scales to
hundreds of millions of
identities.
 A REA GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Encryption Cloud KMS,Secret Manager Azure Key Vault Provides security solution
and works with other
services by providing a way
to manage, create, and
control encryption keys
stored in hardware security
modules (HSM).

Data at rest encryption Encryption by default Azure Storage Service Azure Storage Service
Encryption - encryption by Encryption helps you protect
default and safeguard your data and
meet your organizational
security and compliance
commitments.

Security Security Command Center Azure Security Center An automated security

assessment service that
improves the security and
compliance of applications.
Automatically assess
applications for
vulnerabilities or deviations
from best practices.

Web Security Scanner Azure Security Center An automated security

assessment service that
improves the security and
compliance of applications.
Automatically assess
applications for
vulnerabilities or deviations
from best practices.

Event Threat Detection Azure Advanced Threat Detect and investigate

Protection advanced attacks on-
premises and in the cloud.

Storage
Object storage
GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Cloud Storage Azure Blob storage Object storage service, for use cases
including cloud applications, content
Cloud Storage for Firebase distribution, backup, archiving, disaster
recovery, and big data analytics.

Virtual server disks

GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Persistant Disk Azure managed disks SSD storage optimized for I/O intensive
read/write operations. For use as high-
Local SSD performance Azure virtual machine
storage.
File Storage
GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Filestore Azure Files,Azure NetApp Files File based storage and hosted NetApp
Appliance Storage.

Drive Enterprise OneDrive For business Shared files from from personal devices
to cloud

Bulk data transfer

GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Transfer Appliance Import/Export A data transport solution that uses

secure disks and appliances to transfer
large amounts of data. Also offers data
protection during transit.

Transfer Appliance Azure Data Box Petabyte- to exabyte-scale data

transport solution that uses secure data
storage devices to transfer large
amounts of data to and from Azure.

Application services
GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

App Engine Azure App Service Managed hosting platform providing

easy to use services for deploying and
scaling web applications and services.

Apigee Azure API Management A turnkey solution for publishing APIs

to external and internal consumers.

Miscellaneous
A REA GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Workflow Composer Azure Logic Apps Serverless technology for

connecting apps, data and
devices anywhere, whether
on-premises or in the cloud
for large ecosystems of SaaS
and cloud-based connectors.

Enterprise application G Suite Microsoft 365 Fully integrated Cloud

services service providing
communications, email,
document management in
the cloud and available on a
wide variety of devices.

Gaming Game Servers Azure PlayFab Managed services for

hosting dedicated game
servers.
 A REA GC P SERVIC E A Z URE SERVIC E DESC RIP T IO N

Hybrid Anthos Azure Arc For customers who want to

simplify complex and
distributed environments
across on-premises, edge
and multi-cloud, Azure Arc
enables deployment of
Azure services anywhere and
extends Azure management
to any infrastructure.

Blockchain Digital Asset Azure Blockchain Service Azure Blockchain Service is a

fully managed ledger service
that enables users the ability
to grow and operate
blockchain networks at scale
in Azure.

Monitoring Cloud Monitoring Application Insights Service which provides

visibility into the
performance, uptime, and
overall health of cloud-
powered applications.

Logging Cloud Logging Log Analytics Service for real-time log

management and analysis.

Azure Migration Tools

TYPE A Z URE SERVIC E DESC RIP T IO N

Generic Database migrationGuide (SQL Database migration guide Migration of database schema and data
or Open Source Database) from one database format to a specific
database technology in the cloud.

SQL Server Database Migration Data MigrationAssistant Assessment of SQL Server database
Assessment Tool before migration

SQL Server Database Migration Tool Data Migration Service Actual migration of database to Azure

Open source Database Migration Tool Database Native Tool Actual Migration of open source
likemysqldump,pg\dump database to Azure

Assessment and migration tool Azure Migrate Assesses on-premises workloads for
migration to Azure, performs
performance-based sizing, and provides
cost estimations.

Migration tool Movere Movere is a discovery solution that

provides the data and insights needed
to plan cloud migrations and
continuously optimize, monitor and
analyze IT environments with
confidence.

More learning
If you are new to Azure, review the interactive Core Cloud Services - Introduction to Azure module on Microsoft
Learn.
 Microsoft Azure Well-Architected Framework
12/18/2020 • 9 minutes to read • Edit Online

The Azure Well-Architected Framework is a set of guiding tenets that can be used to improve the quality of a
workload. The framework consists of five pillars of architecture excellence: Cost Optimization, Operational
Excellence, Performance Efficiency, Reliability, and Security.
To assess your workload using the tenets found in the Microsoft Azure Well-Architected Framework, see the
Microsoft Azure Well-Architected Review.

P IL L A R DESC RIP T IO N

Cost Optimization Managing costs to maximize the value delivered.

Operational Excellence Operations processes that keep a system running in

production.

Performance Efficiency The ability of a system to adapt to changes in load.

Reliability The ability of a system to recover from failures and continue

to function.

Security Protecting applications and data from threats.

Cost Optimization
When you are designing a cloud solution, focus on generating incremental value early. Apply the principles of
Build-Measure-Learn , to accelerate your time to market while avoiding capital-intensive solutions. Use the pay-
as-you-go strategy for your architecture, and invest in scaling out, rather than delivering a large investment first
version. Consider opportunity costs in your architecture, and the balance between first mover advantage versus
"fast follow". Use the cost calculators to estimate the initial cost and operational costs. Finally, establish policies,
budgets, and controls that set cost limits for your solution.
Cost guidance
Review cost principles
Develop a cost model
Create budgets and alerts
Review the cost optimization checklist

Operational Excellence
This pillar covers the operations processes that keep an application running in production. Deployments must be
reliable and predictable. They should be automated to reduce the chance of human error. They should be a fast and
routine process, so they don't slow down the release of new features or bug fixes. Equally important, you must be
able to quickly roll back or roll forward if an update has problems.
Monitoring and diagnostics are crucial. Cloud applications run in a remote data-center where you do not have full
control of the infrastructure or, in some cases, the operating system. In a large application, it's not practical to log
into VMs to troubleshoot an issue or sift through log files. With PaaS services, there may not even be a dedicated
VM to log into. Monitoring and diagnostics give insight into the system, so that you know when and where failures
occur. All systems must be observable. Use a common and consistent logging schema that lets you correlate
events across systems.
The monitoring and diagnostics process has several distinct phases:
Instrumentation. Generating the raw data, from application logs, web server logs, diagnostics built into the
Azure platform, and other sources.
Collection and storage. Consolidating the data into one place.
Analysis and diagnosis. To troubleshoot issues and see the overall health.
Visualization and alerts. Using telemetry data to spot trends or alert the operations team.
Use the DevOps checklist to review your design from a management and DevOps standpoint.
Operational excellence guidance
Design patterns for management and monitoring
Best practices: Monitoring and diagnostics

Performance efficiency
Performance efficiency is the ability of your workload to scale to meet the demands placed on it by users in an
efficient manner. The main ways to achieve this are by using scaling appropriately and implementing PaaS
offerings that have scaling built in.
There are two main ways that an application can scale. Vertical scaling (scaling up) means increasing the capacity
of a resource, for example by using a larger VM size. Horizontal scaling (scaling out) is adding new instances of a
resource, such as VMs or database replicas.
Horizontal scaling has significant advantages over vertical scaling:
True cloud scale. Applications can be designed to run on hundreds or even thousands of nodes, reaching scales
that are not possible on a single node.
Horizontal scale is elastic. You can add more instances if load increases, or remove them during quieter periods.
Scaling out can be triggered automatically, either on a schedule or in response to changes in load.
Scaling out may be cheaper than scaling up. Running several small VMs can cost less than a single large VM.
Horizontal scaling can also improve resiliency, by adding redundancy. If an instance goes down, the application
keeps running.
An advantage of vertical scaling is that you can do it without making any changes to the application. But at some
point you'll hit a limit, where you can't scale any up any more. At that point, any further scaling must be horizontal.
Horizontal scale must be designed into the system. For example, you can scale out VMs by placing them behind a
load balancer. But each VM in the pool must be able to handle any client request, so the application must be
stateless or store state externally (say, in a distributed cache). Managed PaaS services often have horizontal scaling
and autoscaling built in. The ease of scaling these services is a major advantage of using PaaS services.
Just adding more instances doesn't mean an application will scale, however. It might simply push the bottleneck
somewhere else. For example, if you scale a web front end to handle more client requests, that might trigger lock
contentions in the database. You would then need to consider additional measures, such as optimistic concurrency
or data partitioning, to enable more throughput to the database.
Always conduct performance and load testing to find these potential bottlenecks. The stateful parts of a system,
such as databases, are the most common cause of bottlenecks, and require careful design to scale horizontally.
Resolving one bottleneck may reveal other bottlenecks elsewhere.
Use the Performance efficiency checklist to review your design from a scalability standpoint.
Performance efficiency guidance
 Design patterns for scalability and performance
Best practices: Autoscaling, Background jobs, Caching, CDN, Data partitioning

Reliability
A reliable workload is one that is both resilient and available. Resiliency is the ability of the system to recover from
failures and continue to function. The goal of resiliency is to return the application to a fully functioning state after
a failure occurs. Availability is whether your users can access your workload when they need to.
In traditional application development, there has been a focus on increasing the mean time between failures
(MTBF). Effort was spent trying to prevent the system from failing. In cloud computing, a different mindset is
required, due to several factors:
Distributed systems are complex, and a failure at one point can potentially cascade throughout the system.
Costs for cloud environments are kept low through the use of commodity hardware, so occasional hardware
failures must be expected.
Applications often depend on external services, which may become temporarily unavailable or throttle high-
volume users.
Today's users expect an application to be available 24/7 without ever going offline.
All of these factors mean that cloud applications must be designed to expect occasional failures and recover from
them. Azure has many resiliency features already built into the platform. For example:
Azure Storage, SQL Database, and Cosmos DB all provide built-in data replication, both within a region and
across regions.
Azure managed disks are automatically placed in different storage scale units to limit the effects of hardware
failures.
VMs in an availability set are spread across several fault domains. A fault domain is a group of VMs that share a
common power source and network switch. Spreading VMs across fault domains limits the impact of physical
hardware failures, network outages, or power interruptions.
That said, you still need to build resiliency into your application. Resiliency strategies can be applied at all levels of
the architecture. Some mitigations are more tactical in nature — for example, retrying a remote call after a
transient network failure. Other mitigations are more strategic, such as failing over the entire application to a
secondary region. Tactical mitigations can make a big difference. While it's rare for an entire region to experience a
disruption, transient problems such as network congestion are more common — so target these first. Having the
right monitoring and diagnostics is also important, both to detect failures when they happen, and to find the root
causes.
When designing an application to be resilient, you must understand your availability requirements. How much
downtime is acceptable? This is partly a function of cost. How much will potential downtime cost your business?
How much should you invest in making the application highly available?
Reliability guidance
Designing reliable Azure applications
Design patterns for resiliency
Best practices: Transient fault handling, Retry guidance for specific services

Security
Think about security throughout the entire lifecycle of an application, from design and implementation to
deployment and operations. The Azure platform provides protections against a variety of threats, such as network
intrusion and DDoS attacks. But you still need to build security into your application and into your DevOps
processes.
Here are some broad security areas to consider.
Identity management
Consider using Azure Active Directory (Azure AD) to authenticate and authorize users. Azure AD is a fully managed
identity and access management service. You can use it to create domains that exist purely on Azure, or integrate
with your on-premises Active Directory identities. Azure AD also integrates with Office365, Dynamics CRM Online,
and many third-party SaaS applications. For consumer-facing applications, Azure Active Directory B2C lets users
authenticate with their existing social accounts (such as Facebook, Google, or LinkedIn), or create a new user
account that is managed by Azure AD.
If you want to integrate an on-premises Active Directory environment with an Azure network, several approaches
are possible, depending on your requirements. For more information, see our Identity Management reference
architectures.
Protecting your infrastructure
Control access to the Azure resources that you deploy. Every Azure subscription has a trust relationship with an
Azure AD tenant. Use role-based access control (RBAC) to grant users within your organization the correct
permissions to Azure resources. Grant access by assigning RBAC role to users or groups at a certain scope. The
scope can be a subscription, a resource group, or a single resource. Audit all changes to infrastructure.
Application security
In general, the security best practices for application development still apply in the cloud. These include things like
using SSL everywhere, protecting against CSRF and XSS attacks, preventing SQL injection attacks, and so on.
Cloud applications often use managed services that have access keys. Never check these into source control.
Consider storing application secrets in Azure Key Vault.
Data sovereignty and encryption
Make sure that your data remains in the correct geopolitical zone when using Azure data services. Azure's geo-
replicated storage uses the concept of a paired region in the same geopolitical region.
Use Key Vault to safeguard cryptographic keys and secrets. By using Key Vault, you can encrypt keys and secrets
by using keys that are protected by hardware security modules (HSMs). Many Azure storage and DB services
support data encryption at rest, including Azure Storage, Azure SQL Database, Azure Synapse Analytics, and
Cosmos DB.
Security resources
Azure Security Center provides integrated security monitoring and policy management across your Azure
subscriptions.
Azure Security Documentation
Microsoft Trust Center
 Principles of cost optimization
11/2/2020 • 2 minutes to read • Edit Online

A cost-effective workload is driven by business goals and the return on investment (ROI) while staying within a
given budget. The principles of cost optimization are a series of important considerations that can help achieve
both business objectives and cost justification.
To assess your workload using the tenets found in the Azure Well-Architected Framework, see the Microsoft Azure
Well-Architected Review.
Keep within the cost constraints
Every design choice has cost implications. Before choosing an architectural pattern, Azure service, or a price model
for the service, consider the budget constraints set by the company. As part of design, identify acceptable
boundaries on scale, redundancy, and performance against cost. After estimating the initial cost, set budgets and
alerts at different scopes to measure the cost. One of cost drivers can be unrestricted resources. These resources
typically need to scale and consume more cost to meet demand.
Aim for scalable costs
A key benefit of the cloud is the ability to scale dynamically. The workload cost should scale linearly with demand.
You can save cost through automatic scaling. Consider the usage metrics and performance to determine the
number of instances. Choose smaller instances for a highly variable workload and scale out to get the required
level of performance, rather than up. This choice will enable you to make your cost calculations and estimates
granular.
Pay for consumption
Adopt a leasing model instead of owning infrastructure. Azure offers many SaaS and PaaS resources that simplify
overall architecture. The cost of hardware, software, development, operations, security, and data center space
included in the pricing model.
Also, choose pay-as-you-go over fixed pricing. That way, as a consumer, you're charged for only what you use.
Right resources, right size
Choose the right resources that are aligned with business goals and can handle the performance of the workload.
An inappropriate or misconfigured service can impact cost. For example, building a multi-region service when the
service levels don't require high-availability or geo-redundancy will increase cost without any reasonable business
justification.
Certain infrastructure resources are delivered as fix-sized building blocks. Ensure that these blocks are adequately
sized to meet capacity demand, deliver expected performance without wasting resources.
Monitor and optimize
Treat cost monitoring and optimization as a process, rather than a point-in-time activity. Conduct regular cost
reviews and measure and forecast the capacity needs so that you can provision resources dynamically and scale
with demand. Review the cost management recommendations and take action.
If you're just starting in this process review enable success during a cloud adoption journey .
 Checklist - Design for cost
11/2/2020 • 2 minutes to read • Edit Online

Use this checklist when designing a cost-effective workload.

Cost model
Capture clear requirements . Gather detailed information about the business workflow, regulatory,
security, and availability.
Capture requirements
Estimate the initial cost . Use tools such as Azure pricing calculator to assess cost of the services you plan
to use in the workload. Use Azure Migrate and Microsoft Azure Total Cost of Ownership (TCO) Calculator for
migration projects. Accurately reflect the cost associated with right storage type. Add hidden costs, such as
networking cost for large data download.
Estimate the initial cost
Define policies for the cost constraints defined by the organization . Understand the constraints and
define acceptable boundaries for quality pillars of scale, availability, security.
Consider the cost constraints
Identify shared assets . Evaluate the business areas where you can use shared resources. Review the
billing meters build chargeback reports per consumer to identify metered costs for shared cloud services.
Create a structured view of the organization in the cloud
Plan a governance strategy . Plan for cost controls through Azure Policy. Use resource tags so that custom
cost report can be created. Define budgets and alerts to send notifications when certain thresholds are
reached.
Governance

Architecture
Check the cost of resources in various Azure geographic regions . Check your egress and ingress
cost, within regions and across regions. Only deploy to multiple regions if your service levels require it for
either availability or geo-distribution.
Azure regions
Choose a subscription that is appropriate for the workload . Azure Dev/Test subscription types are
suitable for experimental or non-production workloads and have lower prices on some Azure services such
as specific VM sizes. If you can commit to one or three years, consider subscriptions and offer types that
support Azure Reservations.
Subscription and offer type
Choose the right resources to handle the performance . Understand the usage meters and the
number of meters for each resource in the workload. Consider tradeoffs over time. For example, cheaper
virtual machines may initially indicate a lower cost but can be more expensive over time to maintain a
certain performance level. Be clear about the billing model of third-party services.
Azure resources
Use cost alerts to monitor usage and spending
Compare consumption-based pricing with pre-provisioned cost . Establish baseline cost by
considering the peaks and the frequency of peaks when analyzing performance.
Consumption and fixed cost models
Use proof-of-concept deployments . The Azure Architecture Center has many reference architectures and
implementations that can serve as a starting point. The Azure Tech Community has architecture and services
forums.
Choose managed ser vices when possible . With PaaS and SaaS options, the cost of running and
maintaining the infrastructure is included in the service price.
Managed services
 Develop a cost model
12/18/2020 • 6 minutes to read • Edit Online

Cost modeling is an exercise where you create logical groups of cloud resources that are mapped to the
organization's hierarchy and then estimate costs for those groups. The goal of cost modeling is to estimate the
overall cost of the organization in the cloud.
1. Understand how your responsibilities align with your organization
Map the organization's needs to logical groupings offered by cloud services. This way the business leaders
of the company get a clear view of the cloud services and how they're controlled.
2. Capture clear requirements
Start your planning with a careful enumeration of requirements. From the high-level requirements, narrow
down each requirement before starting on the design of the solution.
3. Consider the cost constraints
Evaluate the budget constraints on each business unit and determine the governance policies in Azure to
lower cost by reducing wastage, overprovisioning, or expensive provisioning of resources.
4. Consider tradeoffs
Optimal design doesn't equate to a lowest-cost design.
As requirements are prioritized, cost can be adjusted. Expect a series of tradeoffs in the areas that you want
to optimize, such as security, scalability, resilience, and operability. If the cost to address the challenges in
those areas is high, stakeholders will look for alternate options to reduce cost. There might be risky choices
made in favor or a cheaper solution.
5. Derive functional requirements from high-level goals
Break down the high-level goals into functional requirements for the components of the solution. Each
requirement must be based on realistic metrics to estimate the actual cost of the workload.
6. Consider the billing model for Azure resources
Azure services are offered with consumption-based price where you're charged for only what you use.
There's also options for fixed price where you're charged for provisioned resources
Most services are priced based on units of size, amount of data, or operations. Understand the meters that
are used to track usage. For more information, see Azure resources.
At the end of this exercise, you should have identified the lower and upper limits on cost and set budgets for the
workload. Azure allows you to create and manage budgets in Azure Cost Management. For information, see
Quickstart: Create a budget with an Azure Resource Manager template.
 H AVE F REQ UEN T A N D C L EA R C O M M UN IC AT IO N W IT H T H E STA K EH O L DERS

In the initial stages, communication between stakeholders is vital. The overall team must align on the requirements so that
overall business goals are met. If not, the entire solution might be at risk.
For instance, the development team indicates that the resilience of a monthly batch-processing job is low. They might request
the job to work as a single node without scaling capabilities. This request opposes the architect’s recommendation to
automatically scale out, and route requests to worker nodes.
This type of disagreement can introduce a point of failure into the system, risking the Service Level Agreement, and cause an
increase in operational cost.

Organization structure
Map the organization's needs to logical groupings offered by cloud services. This way the business leaders of the
company get a clear view of the cloud services and how they're controlled.
1. Understand how your workload fits into cost optimization across the portfolio of cloud workloads.
If you are working on a workload that fits into a broader portfolio of workloads, see CAF to get started guide to
document foundational decisions. That guide will help your team capture the broader portfolio view of business
units, resources organizations, responsibilities, and a view of the long term portfolio.
If cost optimization is being executed by a central team across the portfolio, see CAF to get started managing
enterprise costs.
2. Encourage a culture of democratized cost optimization decisions
As a workload owner, you can have a measurable impact on cost optimization. There are other roles in the
organization which can help improve cost management activites. To help embed the pillar of cost optimization into
your organization beyond your workload team, see the CAF article: Build a cost-conscious organization.
3. Reduce costs through shared cloud services and landing zones
If your workload has depedencies on shared assets like Active Directory, Network connectivity, security devices, or
other services that are also used by other workloads, encourage your central IT organization to provide those
services through a centrally managed landing zone to reduce duplicate costs. See the CAF article: get started with
centralized design & configuration to get started with the development of landing zones.
4. Calculate the ROI by understanding what is included in each grouping and what isn't.
Which aspects of the hierarchy are covered by cloud ser vices?

Azure pricing model is based on expenses incurred for the service. Expenses include hardware, software,
development, operations, security, and data center space to name a few. Evaluate the cost benefit of shifting
away from owned technology infrastructure to leased technology solutions.
5. Identify scenarios where you can use shared cloud services to lower cost.
Can some ser vices be shared by other consumers?

Identify areas where a service or an application environment can be shared with other business units.
Identify resources that can be used as shared services and review their billing meters. Examples include a
virtual network and its hybrid connectivity or a shared app service environment (ASE). If the meter data isn't
able to be split across consumers, decide on custom solutions to allocate proportional costs. Move shared
services to dedicated resources for consumers for cost reporting.
 Build chargeback reports per consumer to identify metered costs for shared cloud services. Aim for
granular reports to understand which workload is consuming what amount of the shared cloud service.

Next step
Capture cost requirements

Cost constraints
Here are some considerations for determining the governance policies that can assist with cost management.
What are the budget constraints set by the company for each business unit?
What are policies for the budget alert levels and associated actions?
Identify acceptable boundaries for scale, redundancy, and performance against cost.
Assess the limits for security. Don't compromise on security. Premium cloud security features can drive the cost
up. It’s not necessary to overinvest. Instead use the cost profile to drive a realistic threat profile.
Identify unrestricted resources. These resources typically need to scale and consume more cost to meet
demand.
Next step
Consider tradeoffs

Functional requirements
Break down high-level goals into functional requirements. For each of those requirements, define metrics to
calculate cost estimates accurately. Cloud services are priced based on performance, features, and locations. When
defining these metrics, identify acceptable boundaries of performance, scale, resilience, and security. Start by
expressing your goals in number of business transactions over time, breaking them down to fine-grain
requirements.
What resources are needed for a single transaction, and how many transactions are done per second,
day, year?

Start with a fixed cost of operations and a rough estimate of transaction volume to work out a cost-per-
transaction to establish a baseline. Consider the difference between cost models based on fixed, static
provisioning of services, more variable costs based upon autoscaling such as serverless technologies.

Use T -shirt sizes for choosing SKUs

When choosing options for services, start with an abstract representation of size. For example, if you choose a T-
shirt size approach, small, medium, large sizes, can represent an on-demand virtual machine instead of picking
specific virtual machines or managed disks SKU sizes.
Abstract sizes give you an idea of the expected performance and cost requirements. It sets the tone for various
consumption units that are used to measure compute resources for performance. Also, it helps in understanding
the on-demand consumption model of the services.
For more information, see Estimate the initial cost.
Next steps
Estimate the initial cost
 Capture cost requirements
12/18/2020 • 4 minutes to read • Edit Online

Start your planning with a careful enumeration of requirements. Make sure the needs of the stakeholders are
addressed. For strong alignment with business goals, those areas must be defined by the stakeholders and
shouldn’t be collected from a vendor.
Capture requirements at these levels:
Business workflow
Compliance and regulatory
Security
Availability
What do you aim to achieve by building your architecture in the cloud?

Here are some common answers.

Take advantage of features only available in the cloud, such as intelligent security systems, regions footprint, or
resiliency features.
Use the on-demand nature of the cloud to meet peak or seasonal requirements, then releasing that cost
investment when it is no longer needed.
Consolidate physical systems.
Retire on-premises infrastructure.
Reduce hardware or data center management costs.
Increase performance or processing capabilities, through services like big data and machine learning.
Meet regulatory considerations, including taking advantage of certified infrastructure.
Narrow down each requirement before you start the design of the workload. Expect the requirements to change
over time as the solution is deployed and optimized.

Landing zone
Consider the cost implications of the geographic region to which the landing zone is deployed.
The landing zone consists of the subscription and resource group, in which your cloud infrastructure components
exist. This zone impacts the overall cost. Consider the tradeoffs. For example, there are additional costs for network
ingress and egress for cross-zonal traffic. For more information, see Azure regions and Azure resources.
For information about landing zone for the entire organization, see CAF: Implement landing zone best practices.

Security
Security is one of the most important aspects of any architecture. Security measures protect the valuable data of
the organization. It provides confidentiality, integrity, and availability assurances against attacks and misuse of the
systems.
Factor in the cost of security controls, such as authentication, MFA, conditional access, information protection,
JIT/PIM, and premium Azure AD features. Those options will drive up the cost.
For security considerations, see the Security Pillar.
Business continuity
Does the application have a Ser vice Level Agreement that it must meet?

Factor in the cost when you create high availability and disaster recovery strategies.
Overall Service Level Agreement (SLA), Recovery Time Objective (RTO), and Recovery Point Objective (RPO) may
drive towards expensive design choices in order to support higher availability requirements. For example, a choice
might be to host the application across regions, which is costlier than single region but supports high availability.
If your service SLAs, RTOs and RPOs allow, then consider cheaper options. For instance, pre-build automation
scripts and packages that would redeploy the disaster recovery components of the solution from the ground-up in
case a disaster occurs. Alternatively, use Azure platform-managed replication. Both options can lower cost because
fewer cloud services are pre-deployed and managed, reducing wastage.
In general, if the cost of high availability exceeds the cost of application downtime, then you could be over
engineering the high availability strategy. Conversely, if the cost of high availability is less than the cost of a
reasonable period of downtime, you may need to invest more.
Suppose the downtime costs are relatively low, you can save by using recovery from your backup and disaster
recovery processes. If the downtime is likely to cost a significant amount per hour, then invest more in the high
availability and disaster recovery of the service. It's a three-way tradeoff between cost of service provision, the
availability requirements, and the organization's response to risk.

Application lifespan
Does your ser vice run seasonally or follow long-term patterns?

For long running applications, consider using Azure Reservations if you can commit to one-year or three-year
term. VM reservations can reduce cost by 60% or more when compared to pay-as-you-go prices.
Reservation is still an operational expense with all the corresponding benefits. Monitor the cost on workloads that
have been running in the cloud for an extended period to forecast the reserved instance sizes that are needed. For
information about optimization, see Reserved instances.
If your application runs intermittently, consider using Azure Functions in a consumption plan so you only pay for
compute resources you use.

Automation opportunities
Is it a business requirement to have the ser vice be available 24x7?

You may not have a business goal to leave the service running all the time. Doing so will incur a consistent cost.
Can you save by shutting down the service or scaling it down outside normal business hours? If you can,
Azure has a rich set of APIs, SDKs, and automation technology that utilizes DevOps and traditional automation
principles. Those technologies ensure that the workload is available at an appropriate level of scale as needed.
Repurpose some compute and data resources for other tasks that run out of regular business hours. See the
Compute Resource Consolidation pattern and consider containers or elastic pools for more compute and data
cost flexibility.

Budget for staff education

Keep the technical staff up to date in cloud management skills so that the invested services are optimally used.
Consider using resources such as FastTrack for Azure and Microsoft Learn to onboard the staff. Those resources
 provide engineering investments at no cost to customers.
Identify training requirements and costs for cloud migration projects, application development, and
architecture refinement.
Invest in key areas, such as identity management, security configuration, systems monitoring, and automation.
Give the staff access to training and relevant announcements. This way, they can be aware of new cloud
capabilities and updates.
Provide opportunities to get real-world experience of customers across the globe through conferences, specific
cloud training, and passing dedicated Microsoft Exams (AZ, MS, MB, etc.).

Standardization
Ensure that your cloud environments are integrated into any IT operations processes. Those operations include
user or application access provisioning, incident response, and disaster recovery. That mapping may uncover areas
where additional cloud cost is needed.

Next step
Determine the cost constraints
 Azure regions
11/2/2020 • 2 minutes to read • Edit Online

Cost of an Azure service can vary between locations based on demand and local infrastructure costs. Consider all
these geographical areas when choosing the location of your resources to estimate costs.

T ERM IN O LO GY DESC RIP T IO N

Azure region A set of datacenters deployed within a latency-defined

perimeter and connected through a dedicated regional low-
latency network.

Availability zone A unique physical location within a region with independent

power, network, and cooling to be tolerant to datacenter
failures through redundancy and logical isolation of services.

Billing zone A geographic collection of regions that is used for billing.

Location A region or a zone within a region. Azure has datacenters all

over the world and each datacenter is placed in a location.

Landing zone The ultimate location of your cloud solution or the landing
zone, typically consisting of logical containers such as a
subscription and resource group, in which your cloud
infrastructure components exist.

The complete list of Azure geographies, regions, and locations, is shown in Azure global infrastructure.
To see availability of a product by region, see Products available by region.

Tradeoff
Locating resources in a cheaper region should not negate the cost of network ingress and egress or by
degraded application performance because of increased latency.
An application hosted in a single region may cost less than an application hosted across regions because of
replication costs or the need for extra nodes.

Compliance and regulatory

Azure also offers differentiated cloud regions for specific security and compliance requirements.
Does your solution need specific levels of security and compliance?

If your solution needs to follow certain government regulations, the cost will be higher. Otherwise you can meet
less rigid compliance, through Azure Policy, which is free.
Certain Azure regions are built specifically for high compliance and security needs. For example, with Azure
Government (USA) you're given an isolated instance of Azure. Azure Germany has datacenters that meet privacy
certifications. These specialized regions have higher cost.
Regulatory requirements can dictate restrictions on data residency. These requirements may impact your data
replication options for resiliency and redundancy.
Traffic across billing zones and regions
Cross-regional traffic and cross-zonal traffic incur additional costs.
Is the application critical enough to have the footprint of the resources cross zones and,or cross
regions?

Bandwidth refers to data moving in and out of Azure datacenters. Inbound data transfers (data going into Azure
datacenters) are free for most services. For outbound data transfers, (data going out of Azure datacenters) the data
transfer pricing is based on billing zones. For more information, see Bandwidth Pricing Details.
Suppose, you want to build a cost-effective solution by provisioning resources in locations that offer the lowest
prices. The dependent resources and their users are located in different parts of the world. In this case, data
transfer between locations will add cost if there are meters tracking the volume of data moving across locations.
Any savings from choosing the cheapest location could be offset by the additional cost of transferring data.
The cross-regional and cross-zone additional costs do not apply to global services, such as Azure Active
Directory.
Not all Azure services support zones and not all regions in Azure support zones.

Before choosing a location, consider how important is the application to justify the cost of having
resources cross zones and/or cross regions. For non-mission critical applications such as, developer or test,
consider keeping the solution and its dependencies in a single region or single zone to leverage the
advantages of choosing the lower-cost region.
 Azure resources
12/18/2020 • 3 minutes to read • Edit Online

Just like on-premises equipment, there are several elements that affect monthly costs when using Azure services.

Usage meters for resources

Most services are priced based on units of size, amount of data, or operations. When you provision an Azure
resource, Azure creates metered instances for that resource. The meters track the resources' usage and generate a
usage record that is used to calculate your bill.
For example, you provision a virtual machine in Azure. Some meters that track its usage include: Compute Hours,
IP Address Hours, Data Transfer In, Data Transfer Out, Standard Managed Disk, Standard Managed Disk Operations,
Standard IO-Disk, Standard IO-Block Blob Read, Standard IO-Block Blob Write, Standard IO-Block Blob Delete.
How is the usage tracked for each resource in the workload?

For each Azure resource, have a clear understanding of the meters that track usage and the number of meters
associated with the resource tier. The meters correlate to several billable units. Those units are charged to the
account for each billing period. The rate per billable unit depends on the resource tier.
A resource tier impacts pricing because each tier offers levels of features such as performance or availability. For
example, a Standard HDD hard disk is cheaper than a Premium SSD hard disk.

Start with a lower resource tier then scale the resource up as needed. Growing a service with little to no
downtime is easier when compared to downscaling a service. Downscaling usually requires deprovisioning or
downtime. In general, choose scaling out instead of scaling up.

As part of the requirements, consider the metrics for each resource and build your alerts on baseline thresholds
for each metric. The alerts can be used to fine-tune the resources. For more information, see Respond to cost alert.

Allocated usage for the resource

Another way to look at pricing is the allocated usage.
Suppose, you de-allocate the virtual machine. You'll not be billed for Compute Hours, I/O reads or writes, and
other compute meters because the virtual machine is not running and has no given compute resources. However,
you'll be charged for storage costs for the disks.
Here are some considerations:
The meters and pricing vary per product and often have different pricing tiers based on the location, size, or
capacity of the resource.
Cost for associated with infrastructure is kept low with commodity hardware.
Failures cannot be prevented but the effects of failure can be minimized through design choices. The resiliency
features are factored in the price of a service and its features.
Here are some examples:
Azure Disk
Start with a small size in GB for a managed disk instead of pay-per-GB model. It’s cost effective because cost is
incurred on the allocated storage.
ExpressRoute
Start with a smaller bandwidth circuit and scale up as needed.
Compute infrastructure
Deploy an additional smaller instance of compute alongside a smaller unit in parallel. That approach is more cost
effective in comparison to restarting an instance to scale up.
For details about how billing works, see Azure Cost Management + Billing documentation.

Subscription and offer type

What is the subscription and offer type in which resources are created?

Azure usage rates and billing periods can vary depending on the subscription and offer type. Some subscription
types also include usage allowances or lower prices. For example, Azure Dev/Test subscription offers lower prices
on Azure services such as specific VM sizes, PaaS web apps, and VM images with pre-installed software. Visual
Studio subscribers obtain as part of their benefits access to Azure subscriptions with monthly allowances.
For information about the subscription offers, see Microsoft Azure Offer Details.
As you decide the offer type, consider the types that support Azure Reservations. With reservations, you prepay
for reserved capacity at a discount. Reservations are suitable for workloads that have a long-term usage pattern.
Combining the offer type with reservations can significantly lower the cost. For information about subscription
and offer types that are eligible for reservations, see Discounted subscription and offer types.

Billing structure for services in Azure Marketplace

Are you considering third-par ty ser vices through Azure Marketplace?

Azure Marketplace offers both the Azure products and services from third-party vendors. Different billing
structures apply to each of those categories. The billing structures can range from free, pay-as-you-go, one-time
purchase fee, or a managed offering with support and licensing monthly costs.
 Governance
12/18/2020 • 2 minutes to read • Edit Online

Governance can assist with cost management. This work will benefit your ongoing cost review process and will
offer a level of protection for new resources.

Understand how centralized governance functions can support your

team
Centralized governance can relieve some of the burden related to on-going cost monitoring and optimization.
However, that is an augmentation of the workload team's responsibilities, not a replacement. For an understanding
of how centralized cloud governance teams operate, see the Cloud Adoption Framework's Govern methodology.
For more detailed information on cost optimization, see the section on Cost Management discipline.
For an example of the types of guardrails provided by a governance team, see the narrative for improving the
cost management discipline, which includes examples of suggested tags and policies for improving cost
governance.
If your team isn't supported by centralized governance teams, see Cloud governance function to better
understand the types of activities your team may need to consider including in each sprint.

Follow organizational policies that define cost boundaries

Use policies to ensure compliance to the identified cost boundaries. Also, it eliminates the need for manual
resource approval and speeds up the total provisioning time.
Azure Policy can set rules on management groups, subscriptions, and resources groups. The policies control clouds
service resource SKU size, replication features, and locations that are allowed. Use policies to prevent provisioning
of expensive resources. Identify the built-in Azure policies that can help lower cost. For additional control, create
custom policies.
For more information, see Create management groups for resource organization and management.
Control the group who can manage resources in the subscription, see Built-in roles for Azure resources.
Set limits or quotas to prevent unexpected costs, For more information, see Azure subscription and service limits,
quotas, and constraints.

Enforce resource tagging

Use tags on resources and resource groups to track the incurred costs. Identify the service meters that can't be
tagged or viewed in the cost analysis tool in Azure portal.
The advantages of tagging include:
The cost can be reported to an owner, an application, a business department or a project initiative. This feature
is useful because the overall cost can span multiple resources, locations, and subscriptions.
Filter information. This filtering can be used in cost analysis tool in Azure portal allowing you to get granular
reports.
There are some limitations:
Tags can’t be inherited. If you tag a resource group, the resources under that group won’t inherit that tag. This
aspect requires, additional efforts in looking up the parent resource group to get tagging information if you
want to tag related resource.
Not all Azure resources can be tagged and not all taggable resources in Azure are accounted for in the Azure
cost analysis tool.
 Estimate the initial cost
12/18/2020 • 6 minutes to read • Edit Online

It’s difficult to attribute costs before deploying a workload to the cloud. If you use methods for on-premises
estimation or directly map on-premises assets to cloud resources, estimate will be inaccurate. For example, if you
build your own datacenter your costs may appear comparable to cloud. Most on-premises estimates don't account
for costs like cooling, electricity, IT and facilities labor, security, and disaster recovery.
Here are some best practices:
Use proof-of-concept deployments to help refine cost estimates.
Choose the right resources that can handle the performance of the workload. For example, cheaper virtual
machines may initially indicate a lower cost but can be more expensive eventually to maintain a certain
performance level.
Accurately reflect the cost associated with right storage type.
Add hidden costs, such as networking cost for large data download.

Migration workloads
Quantify the cost of running your business in Azure by calculating Total Cost Ownership (TCO) and the Return on
Investment (ROI). Compare those metrics to existing on-premises equivalents.
It’s difficult to attribute costs before migrating to the cloud.
Using on-premises calculation may not accurately reflect the cost of cloud resources. Here are some challenges:
On-premises TCO may not accurately account for hidden expenses. These expenses include under-utilization of
purchased hardware or network maintenance costs including labor and equipment failure.
Cloud TCO may not accurately account for a drop in the organization’s operational labor hours. Cloud
provider’s infrastructure, platform management services, and additional operational efficiencies are included in
the cloud service pricing. Especially true at a smaller scale, the cloud provider’s services don't result in
reduction of IT labor head count.
ROI may not accurately account for new organizational benefits because of cloud capabilities. It's hard to
quantify improved collaboration, reduced time to service customers, and fast scaling with minimal or no
downtime.
ROI may not accurately account for business process re-engineering needed to fully adopt cloud benefits. In
some cases, this re-engineering may not occur at all, leaving an organization in a state where new technology is
used inefficiently.
Azure provides these tools to determine cost.
Microsoft Azure Total Cost of Ownership (TCO) Calculator to reflect all costs.

For migration projects, the TCO Calculator may assist, as it pre-populates some common cost but allows
you to modify the cost assumptions.

Azure pricing calculator to assess cost of the services you plan to use in your solution.
Azure Migrate to evaluate your organization's current workloads in on-premises datacenters. It suggests Azure
replacement solution, such virtual machine sizes based on your workload. It also provides a cost estimate.
Example estimate for a microservices workload
Let's consider this scenario as an example. We'll use the Azure Pricing calculator to estimate the initial cost before
the workload is deployed. The cost is calculated per month or for 730 hours.
In this example, we've chosen the microservices pattern. As the container orchestrator, one of the options could be
Azure Kubernetes Service (AKS) that manages a cluster of pods. We choose NGINX ingress controller because it's
well-known controller for such workloads.

The example is based on the current price and is subject to change. The calculation shown is for information
purposes only.

Compute
For AKS, there's no charge for cluster management.
For AKS agent nodes, there are many instance sizes and SKUs options. Our example workload is expected to follow
a long running pattern and we can commit to three years. So, an instance that is eligible for reserved instances
would be a good choice. We can lower the cost by choosing the 3-year reser ved plan .
The workload needs two virtual machines. One is for the backend services and the other for the utility services.
The B12MS instance with 2 virtual machines is sufficient for this initial estimation. We can lower cost by choosing
reserved instances.
Estimated Total: $327.17 per month with upfront payment of $11,778.17 .
Application gateway
For this scenario, we consider the Standard_v2 Tier of Azure Application Gateway because of the autoscaling
capabilities and performance benefits. We also choose consumption-based pricing, which is calculated by capacity
units (CU). Each capacity unit is calculated based on compute, persistent connections, or throughput. For
Standard_v2 SKU - Each compute unit can handle approximately 50 connections per second with RSA 2048-bit key
TLS certificate. For this workload, we estimate 10 capacity units.
Estimated Total: $248.64 per month.
Load balancer
NGINX ingress controller deploys a load balancer that routes internet traffic to the ingress. Approximately 15 load
balancer rules are needed. NAT rules are free. The main cost driver is the amount of data processed inbound and
outbound independent of rules. We estimate traffic of 1 TB (inbound and outbound).
Estimated Total: $96.37 per month.
Bandwidth
We estimate 2-TB outbound traffic. The first 5 GB/month are free in Zone 1 (Zone 1 includes North America,
Europe, and Australia). Between 5 GB - 10 TB /month is charged $0.087 per GB.
Estimated Total: $177.74 per month
External data source
Because the schema-on read nature of the data handled by the workload, we choose Azure Cosmos DB as the
external data store. By using the Cosmos DB capacity calculator, we can calculate the throughput to reserve.
Cost variables
For lower latency, in this scenario we enable geo-replication by using the Multi-regions writes feature. By
default Cosmos DB uses one region for writes and the rest for reads.
Default choices in consistency model of Session and indexing policy to Automatic . Automatic indexing
makes Cosmos DB to index all properties in all items for flexible and efficient queries. Custom indexing
policy allows you to include/exclude properties from the index for lower write RU and storage size. So
uploading a custom indexing policy can make you reduce costs.
Total data store isn't a significant cost driver, here it's set to 500 GB.
The throughput is variable indicating peaks. The percentage of time at peak is set to 10%.
The item size is an average of 90k. By using the capacity calculator you can upload sample json files with
your document's data structure, the average document's size, the number of reads/writes per second. These
variables have the largest impact on cost because they're used to calculate the throughput. The throughput
values are shown in the image.
Now, we use those values in the Azure Pricing calculator.
The average throughput based on these settings is 20,000 RUs/sec which is the minimum throughput required for
a 3-year reser ved capacity plan.
Here is the total cost for three years using the reserved plan:
$1,635.20 Average per month ($58,867.20 charged upfront)
You save $700.00 by choosing the 3-year reserved capacity over the pay-as-you-go price.
CI/CD pipelines
With Azure DevOps, Basic Plan is included for Visual Studio Enterprise, Professional, Test Professional, and MSDN
Platforms subscribers. And no charge for adding or editing work items and bugs, viewing dashboards, backlogs,
and Kanban boards for stakeholders.
Basic plan license for five users is free.
Additional services
For Microsoft Hosted Pipelines, the Free tier includes one parallel CI/CD job with 1,800 minutes (30 hours) per
month. However you can select the Paid tier and have one CI/CD parallel job ($40.00), in this tier, each parallel
CI/CD job includes unlimited minutes.

For this stage of cost estimation, the Self Hosted Pipelines is not required because the workload doesn't have
custom software that runs in your build process which isn't included in the Microsoft-hosted option.

Azure Artifacts is a service where you can create package feeds to publish and consume Maven, npm, NuGet,
Python, and universal packages. Azure Artifacts is billed on a consumption basis, and is free up until 2 GB of
storage. For this scenario, we estimate 56 GB in artifacts ($56.00)
Azure DevOps offers a cloud-based solution for load testing your apps. Load tests are measured and billed in
virtual user minutes (VUMs). For this scenario, we estimate a 200,000 VUMs ($72.00).
Estimated Total: $168.00 per month
 Managed services
12/18/2020 • 2 minutes to read • Edit Online

Look for areas in the architecture where it may be natural to incorporate platform-as-a-service (PaaS) options.
These include caching, queues, and data storage. PaaS reduces time and cost of managing servers, storage,
networking, and other application infrastructure.
With PaaS, the infrastructure cost is included in the pricing model of the service. For example, you can provision a
lower SKU virtual machine as a jumpbox. There are additional costs for storage and managing a separate server.
You also need to configure a public IP on the virtual machine, which is not recommended. A managed service such
as Azure Bastion takes into consideration all those costs and offers better security.
Azure provides a wide range of PaaS resources. Here are some examples of when you might consider PaaS options:

TA SK USE

Host a web server Azure App Service instead of setting up IIS servers.

Indexing and querying heterogenous data Azure Cognitive Search instead of ElasticSearch.

Host a database server Azure offers many SQL and no-SQL options such as Azure
SQL Database and Azure Cosmos DB.

Secure access to virtual machine Azure Bastion instead of virtual machines as jump boxes.

Network security Azure Firewall instead of virtual network appliances.

For more information, see Use platform as a service (PaaS) options.

Reference architecture
To see an implementation that provides better security and lowers cost through PaaS services, see Network DMZ
between Azure and an on-premises datacenter.
 Consumption and fixed cost models
11/2/2020 • 2 minutes to read • Edit Online

The common pricing options for Azure services are:

Consumption-based price - You are charged for only what you use. This model is also known as the Pay-As-
You-Go rate.
Fixed price - You provision resources and are charged for those instances whether or not they are used.
A common way to estimate cost is by considering workloads on a peak throughput. Under consistently high
utilization, consumption-based pricing can be less efficient for estimating baseline costs when compared to the
equivalent provisioned pricing. PaaS and serverless technologies can help you understand the economy cutoff
point for consumption-based pricing.
Observe the difference between cost models based on fixed, static provisioning of services, more variable costs
based on autoscaling of serverless technologies.

Start with a fixed minimum level of performance and then use architectural patterns (such as Queue Based Load
Leveling) and autoscaling of services. With this approach the peaks can be smoothed out into a more consistent
flow of compute and data. This approach should temporarily extend your burst performance when the service is
under sustained load. If cost is an important factor but you need to maintain service availability under burst
workload use the Throttling pattern to maintain quality of service under load.
Compare and contrast the options and understand how to provision workloads that can potentially switch between
the two models. The model will be a tradeoff between scalability and predictability. Ideally in the architecture, blend
the two aspects.
 Provisioning cloud resources to optimize cost
11/2/2020 • 2 minutes to read • Edit Online

Deployment of cloud resources of a workload is known as provisioning.

Use the Azure Pricing calculator to estimate the cost of your SKU choices. This section describes some
considerations. This list is not meant to be an exhaustive list, but a subset of options.
AI + Machine Learning
Big data analytics
Compute
Networking
Data stores
 AI + Machine Learning cost estimates
12/18/2020 • 3 minutes to read • Edit Online

The main cost driver for machine learning workloads is the compute cost. Those resources are needed to run the
training model and host the deployment. For information about choosing a compute target, see What are compute
targets in Azure Machine Learning?.
The compute cost depends on the cluster size, node type, and number of nodes. Billing starts while the cluster
nodes are starting, running, or shutting down.
With services such as Azure Machine Learning, you have the option of creating fix-sized clusters or use autoscaling.

If the amount of compute is not known, start with a zero-node cluster. The cluster will scale up when it
detects jobs in the queue. A zero-node cluster is not charged.

Fix-sized clusters are appropriate for jobs that run at a constant rate and the amount of compute is known and
measured beforehand. The time taken to spin up or down a cluster incurs additional cost.

If you don’t need retraining frequently, turn off the cluster when not in use.

To lower the cost for experimental or development workloads, choose Spot VMs. They aren't recommended for
production workloads because they might be evicted by Azure at any time. For more information, see Use Spot
VMs in Azure.
For more information about the services that make up a machine learning workload, see What are the machine
learning products at Microsoft?.
This article provides cost considerations for some technology choices. This is not meant to be an exhaustive list, but
a subset of options.

Azure Machine Learning

There are two editions, Enterprise and Basic . Training models don't incur the machine learning service surcharge.
You're charged for these factors.
The cost is driven by compute choices, such as, the virtual machine sizes and the region in which they are
available. If you can commit to one or three years, choosing reserved instances can lower cost. For more
information, see Reserved instances.
As part of provisioning Machine Learning resources, additional resource are deployed such as Azure
Container Registry, Azure Block Blob Storage, and Key Vault. You're charged for as per the pricing of those
individual services.
If you deploy models to a Kubernetes Service cluster, Machine Learning adds a surcharge on top of the
Kubernetes Service compute cost. This cost can be lowered through autoscaling.
For more information, see these articles:
Machine Learning pricing calculator
Azure Machine Learning
Reference architecture
 Training of Python scikit-learn and deep learning models on Azure
Distributed training of deep learning models on Azure
Batch scoring of Python machine learning models on Azure
Batch scoring of deep learning models on Azure
Real-time scoring of Python scikit-learn and deep learning models on Azure
Machine learning operationalization (MLOps) for Python models using Azure MachineLearning
Batch scoring of R machine learning models on Azure
Real-time scoring of R machine learning models on Azure
Batch scoring of Spark machine learning models on Azure Databricks
Enterprise-grade conversational bot
Build a real-time recommendation API on Azure

Azure Cognitive services

The billing depends on the type of service. The charges are based on the number of transactions for each type of
operation specific to a service. Certain number of transactions are free. If you need additional transactions, choose
from the Standard instances. For more information, see
Cognitive services pricing calculator
Cognitive services pricing
Reference architecture
Build an enterprise-grade conversational bot

Azure Bot Service

The Azure Bot Service is a managed service purpose-built for enterprise-grade bot development. Billing is based
on the number of messages. Certain number of messages are free. If you need to create custom channels, choose
Premium channels , which can drive up the cost of the workload.
For a Web App Bot, an Azure App Service is provisioned to host the bot. Also, an instance of Application Insights is
provisioned. You're charged for as per the pricing of those individual services.
Reference architecture
Enterprise-grade conversational bot
 Big data analytics cost estimates
12/18/2020 • 6 minutes to read • Edit Online

Most big data workloads are designed to do:

Batch processing of big data sources at rest.
Stream processing of data in motion.
Those workloads have different needs. Batch processing is done with long-running batch jobs. For stream
processing, the data ingestion component should be able to capture and in some cases buffer, store real-time
messages. Both workloads also have the requirement to store large volume of data. Then, filter, aggregate, and
prepare that data for analysis.
For information about choosing technologies for each workload, see these articles:
Batch processing
Technology choices for batch processing
Capability matrix
Stream processing
What are your options when choosing a technology for real-time processing?
Capability matrix
This article provides cost considerations for some of those choices. This is not meant to be an exhaustive list, but a
subset of options.

Azure Synapse Analytics

The analytics resources are measured in Data Warehouse Units (DWUs), which tracks CPU, memory, and IO. DWU
also indicates the required level of performance. If you need higher performance, add more DWU blocks.
You can provision the resources in one of two service levels.
Compute Optimized Gen1 tracks usage in DWUs and is offered in a pay-as-you-go model.
Compute Optimized Gen2 tracks the compute DWUs (cDWUs) which allows you to scale the compute nodes.
This level is intended for intensive workloads with higher query performance and compute scalability. You can
choose the pay-as-you-go model or save 37% to 65% by using reserved instances if you can commit to one or
three years. For more information, see Reserved instances.

Start with smaller DWUs and measure performance for resource intensive operations, such as heavy data
loading or transformation. This will help you determine the number of units you need to increase or decrease.
Measure usage during the peak business hours so you can assess the number of concurrent queries and
accordingly add units to increase the parallelism. Conversely, measure off-peak usage so that you can pause
compute when needed.

In Azure Synapse Analytics, you can import or export data from an external data store, such as Azure Blob Storage
and Azure Data Lake Store. Storage and analytics resources aren't included in the price. There is additional
bandwidth cost for moving data in and out of the data warehouse.
For more information, see these articles:
 Azure Synapse Pricing
Manage compute in Azure Synapse Analytics data warehouse
Reference architecture
Automated enterprise BI with Azure Synapse Analytics and Azure Data Factory
Enterprise BI in Azure with Azure Synapse Analytics

Azure Databricks
Azure Databricks offers two SKUs Standard and Premium , each with these options, listed in the order of least to
most expensive.
Data Engineering Light is for data engineers to build and execute jobs in automated Spark clusters.
Data Engineering includes autoscaling and has features for machine learning flows.
Data Analytics includes the preceding set of features and is intended for data scientists to explore, visualize,
manipulate, and share data and insights interactively.
Choose a SKU depending on your workload. If you need features like log audit, which is available in Premium , the
overall cost can increase. If you need autoscaling of clusters to handle larger workloads or interactive Databricks
dashboards, choose an option higher than Data Engineering Light .
Here are factors that impact Databricks billing:
Azure location where the resource is provisioned.
The virtual machine instance tier and number of hours the instances were running.
Databricks units (DBU), which is a unit of processing capability per hour, billed on per-second usage.

The example is based on the current price and is subject to change. The calculation shown is for information
purposes only.

Suppose you run a Premium cluster for 100 hours in East US 2 with 10 DS13v2 instances.

IT EM EXA M P L E EST IM AT E

Cost for 10 DS13v2 instances 100 hours x 10 instances x $0.741/hour = $741.00

DBU cost for Data Analytics workload 100 hours x 10 instances x 2 DBU per node x $0.55/DBU =
$1,100

Total $1,841

For more information, see Azure Databricks Pricing.

If you can commit to one or three years, opt for reserved instances, which can save 38% - 59%. For more
information, see Reserved instances.

Turning off the Spark cluster when not in use to prevent unnecessary charges.

Reference architecture
Stream processing with Azure Databricks
Build a Real-time Recommendation API on Azure
Batch scoring of Spark models on Azure Databricks
Azure Stream Analytics
Stream analytics uses streaming units (SUs) to measure the amount of compute, memory, and throughput required
to process data. When provisioning a stream processing job, you're expected to specify an initial number of SUs.
Higher streaming units mean higher cost because more resources are used.
Stream processing with low latency requires a significant amount of memory. This resource is tracked by the SU%
utilization metric. Lower utilization indicates that the workload requires more compute resources. You can set an
alert on 80% SU Utilization metric to prevent resource exhaustion.

To evaluate the number of units you need, process an amount of data that is realistic for your production
level workload, observe the SU% Utilization metric, and accordingly adjust the SU value.

You can create stream processing jobs in Azure Stream Analytics and deploy them to devices running Azure IoT
Edge through Azure IoT Hub. The number of devices impacts over all cost. Billing starts when a job is deployed to
devices, regardless of the job status (running, failed, stopped).
SUs are based on the partition configuration for the inputs and the query that's defined within the job. For more
information, see Calculate the max streaming units for a job and Understand and adjust Streaming Units.
For pricing details, see Azure Stream Analytics pricing.
Reference architecture
Batch scoring of Python machine learning models on Azure
Azure IoT reference architecture

Azure Analysis Services

Big data solutions need to store data that can be used for reporting. Azure Analysis Services supports the creation
of tabular models to meet this need.
Here are the tiers:
Developer is recommended for evaluation, development, and test scenarios.
Basic is recommended for small production environment.
Standard is recommended for mission critical workloads.
Analysis Services uses Query Processing Units (QPUs) to determine the processing power. A QPU is an abstracted
measure of compute and data processing resources that impact performance. Higher the QPU results in higher
performance.
Each tier offers one or more instances. The main cost drivers are the QPUs and memory allocated for the tier
instance. Start with a smaller instance, monitor the QPU usage, and scale up or down by selecting a higher or lower
instance within the tier. Also, monitor usage during off-peak hours. You can pause the server when not in use. No
charges apply when you pause your instance. For more information, see these articles:
The right tier when you need it
Monitor server metrics
Azure Analysis Services pricing
Reference architecture
Enterprise business intelligence - Azure Reference Architectures
Automated enterprise BI - Azure Architecture Center

Azure Data Factory V2

Azure Data Factory is a big data orchestrator. The service transfers data to and from diverse types of data stores. It
transforms the data by using other compute services. It creates workflows that automate data movement and
transformation. You are only charged for consumption. Consumption is measured by these factors:
Pipeline activities that take the actions on the data. Those actions include copying the data from various
sources, transforming it, and controlling the flow. For more information, see data movement activities, data
transformation activities, and control activities.
You are charged for the total activities in thousands.
Executions measured in data integration units. Each unit tracks CPU, memory, and network resource
allocation. This measurement applies to Azure Integration Runtime.
You are also charged for execution activities, such as copying data, lookups, and external activities. Each activity is
individually priced. You are also charged for pipelines with no associated triggers or runs within the month. All
activities are prorated by the minute and rounded up.
Reference architecture
Automated enterprise BI - Azure Architecture Center
Enterprise business intelligence - Azure Reference Architectures
Build an enterprise-grade conversational bot - Azure Architecture Center
 Compute cost estimates
12/18/2020 • 8 minutes to read • Edit Online

Compute refers to the hosting model for the computing resources that your application runs on. Whether you’re
hosting model is Infrastructure as a Service (IaaS), Platform as a Service (PaaS), or Function as a service (FaaS),
each resource requires your evaluation to understand the tradeoffs that can be made that impact your cost. To
learn more about hosting models, read Understand the hosting models.
Infrastructure-as-a-Ser vice (IaaS) lets you provision individual virtual machines (VMs) along with the
associated networking and storage components. Then you deploy whatever software and applications you
want onto those VMs. This model is the closest to a traditional on-premises environment, except that
Microsoft manages the infrastructure. You still manage the individual VMs.
Platform-as-a-Ser vice (PaaS) provides a managed hosting environment, where you can deploy your
application without needing to manage VMs or networking resources. Azure App Service is a PaaS service.
Functions-as-a-Ser vice (FaaS) goes even further in removing the need to worry about the hosting
environment. In a FaaS model, you simply deploy your code and the service automatically runs it. Azure
Functions are a FaaS service.
What are the cost implications to consider for choosing a hosting model?

If your application can be broken down into short pieces of code, a FaaS hosting model might be the best choice.
You’re charged only for the time it takes to execute your code. For example, Azure Functions is a FaaS service that
processes events with serverless code. Azure Functions allows you to run short pieces of code (called functions)
without worrying about application infrastructure. Use one of the three Azure Functions pricing plans to fit your
need. To learn more about the pricing plans, see How much does Functions cost?
If you want to deploy a larger or more complex application, PaaS may be the better choice. With PaaS, your
application is always running, as opposed to FaaS, where your code is executed only when needed. Since more
resources are used with PaaS, the price increases.
If you are migrating your infrastructure from on-premises to Azure, IaaS will greatly reduce and optimize
infrastructure costs and salaries for IT staff who are no longer needed to manage the infrastructure. Since IaaS uses
more resources than PaaS and FaaS, your cost could be highest.
What are the main cost drivers for Azure ser vices?

You will be charged differently for each service depending on your region, licensing plan (e.g., Azure Hybrid Benefit
for Windows Server), number and type of instances you need, operating system, lifespan, and other parameters
required by the service. Assess the need for each compute service by using the flowchart in Choose a candidate
service. Consider the tradeoffs that will impact your cost by creating different estimates using the Pricing
Calculator. If your application consists of multiple workloads, we recommend that you evaluate each workload
separately. See Consider limits and costs to perform a more detailed evaluation on service limits, cost, SLAs, and
regional availability.
Are there payment options for Vir tual Machines (VMs) to help meet my budget?

The best choice is driven by the business requirements of your workload. If you have higher SLA requirements, it
will increase overall costs. You will likely need more VMs to ensure uptime and connectivity. Other factors that will
impact cost are region, operating system, and the number of instances you choose. Your cost also depends on the
workload life span. See Virtual machines and Use Spot VMs in Azure for more details.
 Pay as you go lets you pay for compute capacity by the second, with no long-term commitment or upfront
payments. This option allows you to increase or decrease compute capacity on demand. It is appropriate for
applications with short-term, spiky, or unpredictable workloads that cannot be interrupted. You can also start
or stop usage at any time, resulting in paying only for what you use.
Reser ved Vir tual Machine Instances lets you purchase a VM for one or three years in a specified region
in advance. It is appropriate for applications with steady-state usage. You may save more money compared
to pay-as-you-go pricing.
Spot pricing lets you purchase unused compute capacity at major discounts. If your workload can tolerate
interruptions, and its execution time is flexible, then using spot pricing for VMs can significantly reduce the
cost of running your workload in Azure.
Dev-Test pricing offers discounted rates on Azure to support your ongoing development and testing. Dev-
Test allows you to quickly create consistent development and test environments through a scalable, on-
demand infrastructure. This will allow you to spin up what you need, when you need it, and explore
scenarios before going into production. To learn more about Azure Dev-Test reduced rates, see Azure
Dev/Test Pricing.
For details on available sizes and options for the Azure VMs you can use to run your apps and workloads, see Sizes
for virtual machines in Azure. For details on specific Azure VM types, see Virtual Machine series.
Do I pay extra to run large-scale parallel and high-performance computing (HPC) batch jobs?

Use Azure Batch to run large-scale parallel and HPC batch jobs in Azure. You can save on VM cost because multiple
apps can run on one VM. Configure your workload with either the Low-priority tier (the cheapest option) or the
Standard tier (provides better CPU performance). There is no cost for the Azure Batch service. Charges accrue for
the underlying resources that run your batch workloads.

Use PaaS as an alternative to buying VMs

When you use the IaaS model, you do have final control over the VMs. It may appear to be a cheaper option at first,
but when you add operational and maintenance costs, the cost increases. When you use the PaaS model, these
extra costs are included in the pricing. In some cases, this means that PaaS services can be a cheaper than
managing VMs on your own. For help finding areas in the architecture where it may be natural to incorporate PaaS
options, see Managed services.
How can I cut costs with hosting my web apps in PaaS?

If you host you web apps in PaaS, you’ll need to choose an App Service plan to run your apps. The plan will define
the set of compute resources on which your app will run. If you have more than one app, they will run using the
same resources. This is where you will see the most significant cost savings, as you don’t incur cost for VMs.
If your apps are event-driven with a short-lived process using a microservices architecture style, we recommend
using Azure Functions. Your cost is determined by execution time and memory for a single function execution. For
pricing details, see Azure Functions pricing.
Is it more cost-effective to deploy development testing (dev-test) on a PaaS or IaaS hosting model?

If your dev-test is built on Azure managed services such as Azure DevOps, Azure SQL Database, Azure Cache for
Redis, and Application Insights, the cheapest solution might be using the PaaS hosting model. You won’t incur the
cost and maintenance of hardware. If your dev-test is built on Azure managed services such as Azure DevOps,
Azure DevTest Labs, VMs, and Application Insights, you need to add the cost of the VMs, which can greatly increase
your cost. For details on evaluating, see Azure Dev/Test Pricing.
A special case of PaaS - Containers
How can I get the best cost savings for a containerized workload that requires full orchestration?

Your business requirements may necessitate that you store container images so that you have fast, scalable
retrieval, and network-close deployment of container workloads. Although there are choices as to how you will run
them, we recommend that you use AKS to set up instances with a minimum of three (3) nodes. AKS reduces the
complexity and operational overhead of managing Kubernetes by offloading much of that responsibility to Azure.
There is no charge for AKS Cluster Management. Any additional costs are minimal. The containers themselves have
no impact on cost. You pay only for per-second billing and custom machine sizes.
Can I save money if my containerized workload does not need full orchestration?

Your business requirements may not necessitate full orchestration. If this is the case and you are using App Service
containers, we recommend that you use one of the App Service plans. Choose the appropriate plan based on your
environment and workload.
There is no charge to use SNI-based SSL. Standard and Premium service plans include the right to use one IP SSL
at no additional charge. Free and shared service plans do not support SSL. You can purchase the right to use
additional SSL connections for a fee. In all cases the SSL certificate itself must be purchased separately. To learn
more about each plan, see App services plans.
Where’s the savings if my workload is event driven with a shor t-lived process?

In this example, Service Fabric may be a better choice than AKS. The biggest difference between the two is that AKS
works only with Docker applications using Kubernetes. Service Fabric works with microservices and supports
many different runtime strategies. Service Fabric can deploy Docker and Windows Server containers. Like AKS,
Service Fabric is built for the microservice and event-driven architectures. AKS is strictly a service orchestrator and
handles deployments, whereas Service Fabric also offers a development framework that allows building
stateful/stateless applications.

Predict cost estimates using the Pricing Calculator

Use the Pricing Calculator to create your total cost estimate. After you run your initial scenario, you may find that
your plan is beyond the scope of your budget. You can adjust the overall cost and create various cost scenarios to
make sure your needs are met before you commit to purchasing.

NOTE
The costs in this example are based on the current price and are subject to change. The calculation is for information
purposes only. It shows the Collapsed view of the cost in this estimate.
 TIP
You can start building your cost estimate at any time and re-visit it later. The changes will be saved until you modify or delete
your estimate.

Next steps
Provisioning cloud resources to optimize cost
Virtual Machines documentation
VM payment options
Pricing Calculator
 Data store cost estimates
11/2/2020 • 9 minutes to read • Edit Online

Most cloud workloads adopt the polyglot persistence approach. Instead of using one data store service, a mix of
technologies is used. You can achieve optimal cost benefit from using this approach.
Each Azure data store has a different billing model. To establish a total cost estimate, first, identify the business
transactions and their requirements. Then, break each transaction into operations. Lastly, identify a data store
appropriate for the type of data. Do this for each workload separately.
Let's take an example of an e-commerce application. It needs to store data for transactions such as orders,
payments, and billing. The data structure is predetermined and not expected to change frequently. Data integrity
and consistency are crucial. There's a need to store product catalogs, social media posts, and product reviews. In
some cases, the data is unstructured, and is likely to change over time. Media files must be stored and data must be
stored for auditing purposes.
Learn about data stores in Understand data store models.

Guidelines: Identify the business transactions and their requirements

The following list of questions address the requirements that can have the greatest impact your cost estimate. For
example, your monthly bill may be within your budget now, but if you scale up or add storage space later, your cost
may increase well over your budget.
Will your data need to be migrated to on-premises, external data centers, or other cloud hosting environments?
What type of data are you intending to store?
How large are the entities you need to store?
What is the overall amount of storage capacity you need?
What kind of schemas will you apply to your data (e.g., fixed schema, schema-on-write, or schema-on-read)?
What are your data performance requirements (e.g., acceptable response times for querying and aggregation of
data once ingested)?
What level of fault-tolerance do you need to provide for data consumers?
What kind of data replication capabilities do you require?
Will the limits of a particular data store support your requirements for scale, number of connections, and
throughput?
How many instances will need to run to support your uptime and throughput requirements? (Consider
operations costs in this calculation.)
Can you partition your data to store it more cost effectively (e.g., can you move large objects out of an
expensive relational database into an object store)?
There are other requirements that may not have as great of an impact on your cost. For example, the US East
region is only slightly lower than Canada Central. See Criteria for choosing a data store for additional business
requirements.
Use the Pricing Calculator to determine different cost scenarios.
What are the network requirements that may impact cost?

Do you need to restrict or otherwise manage access to your data from other network resources?
Does data need to be accessible only from inside the Azure environment?
 Does the data need to be accessible from specific IP addresses or subnets?
Does the data need to be accessible from applications or services hosted on-premises or in other external data
centers?

Guidelines: Break each transaction into operations

For example, one business transaction can be processed by these distinct operations:
10-15 database calls
three append operations on blob, and
two list operations on two separate file shares
How does data type in each operation effect cost?

Consider Azure Blob Storage Block Blobs instead of storing binary image data in Azure SQL Database. Blob
storage is cheaper than Azure SQL Database.
If your design requires SQL, store a lookup table in SQL Database and retrieve the document when needed to
serve it to the user in your application middle tier. SQL Database is highly targeted for high-speed data lookups
and set-based operations.
The hot access tier of Azure Block Blob Storage cost is cheaper than the equivalent size of the Premium SSD
volume that has the database.
Read this decision chart to make your choices.

Guidelines: Identify a data store appropriate for the type of data

NOTE
An inappropriate data store or one that is mis-configured can have a huge cost impact on your design.

Relational database management systems (RDBMS) cost

RDBMS is the recommended choice when you need strong consistency guarantees. An RDBMS typically supports a
schema-on-write model, where the data structure is defined ahead of time, and all read or write operations must
use the schema.
How can I save money if my data is on-premises and already on SQL ser ver?

If the on-premises data is already on a SQL server, it might be a natural choice. The on-premises license with
Software Assurance can be used to bring down the cost if the workload is eligible for Azure Hybrid Benefit. This
option applies for Azure SQL Database (PaaS) and SQL Server on Azure Virtual Machines (IaaS).
For open-source databases such as MySQL, MariaDB, or PostGreSQL, Azure provides managed services that are
easy to provision.
What are some design considerations that will affect cost?

If the SLAs don't allow for downtime, can a read-only replica in a different region enable business continuity?
If the database in the other region must be read/write, how will the data be replicated?
Does the data need to be synchronous, or could consistency allow for asynchronous replication?
How important is it for updates made in one node to appear in other nodes, before further changes can be
made?
Azure storage has several options to make sure data is copied and available when needed. Locally redundant
storage (LRS) synchronously replicates data in the primary region. If the entire primary center is unavailable the
replicated data is lost. At the expensive end, Geo-zone-redundant storage (preview) (GZRS) replicates data in
availability zones within the primary region and asynchronously copied to another paired region. Databases that
offer geo-redundant storage, such as SQL Database, are more expensive. Most other OSS RDBMS database use
LRS storage, which contributes to the lower price range.
For more information, see automated backups.
Cosmos DB offers five consistency levels: strong, bounded staleness, session, consistent prefix, and eventual
consistency. Each level provides availability and performance tradeoffs and is backed by comprehensive SLAs. The
consistency level itself does not affect cost.
How can I minimize compute cost?

Higher throughput and IOPS require higher compute, memory, I/O and storage limits. These limits are expressed in
a vCore model. With higher vCore number, you buy more resources and consequently the cost is higher. Azure SQL
Database has more vCores and allows you to scale in smaller increments. Azure Database for MySQL, PostgreSQL,
and MariaDB have fewer vCores and scaling up to a higher vCore can cost more. MySQL provides in-memory
tables in the Memor y Optimized tier, which can also increase the cost.
All options offer a consumption and provisioned pricing models. With pre-provisioned instances, you save more if
you can commit to one or three years.
How is primar y and backup storage cost calculated?

With Azure SQL Database the initial 32 GB of storage is included in the price. For the other listed options, you need
to buy storage separately and might increase the cost depending on your storage needs.
For most databases, there is no charge for the price of backup storage that is equal in size to primary storage. If
you need more backup storage, you will incur an additional cost.

Key/value and document databases cost

Azure Cosmos DB is recommended for key/value stores and document databases. Azure Cache for Redis is also
recommended for key/value stores.
For Cosmos DB, here are some considerations that will affect cost:
Can storage and item size be adjusted?
Can index policies be reduced or customized to bring down the extra cost for writes and eliminate the
requirement for additional throughput capacity?
If data is no longer needed, can it be deleted from the Azure Cosmos account? As an alternative, you can
migrate the old data to another data store such as Azure blob storage or Azure data warehouse.
For Azure Cache for Redis, there is no upfront cost, no termination fees, you pay only for what you use, and billing
is per-hour.

Graphic databases cost

Cost considerations for graphic database stores include storage required by data and indexes used across all the
regions. For example, graphs need to scale beyond the capacity of a single server, and make it easy to lookup the
index when processing a query.
The Azure service that supports this is Gremlin API in Azure Cosmos DB. Cost is limited to the Azure Cosmos DB
usage. You pay for the operations you perform against the database and for the storage consumed by your data.
Charges are determined by the number of provisioned containers, the number of hours the containers were online,
and the provisioned throughput for each container.
If the on-premises data is already on an SQL server on Azure Virtual Machines (IaaS), the license with Software
Assurance can be used to bring down the cost if the workload is eligible for Azure Hybrid Benefit.

Data analytics cost

Cost considerations for data analytics stores include data storage, multiple servers to maximize scalability, and the
ability to access large data as external tables.
Azure has many services that support data analytics stores: Azure Synapse Analytics, Azure Data Lake, Azure Data
Explorer, Azure Analysis Services, HDInsight, and Azure Databricks. As an example of use, historical data is typically
stored in data stores such as blob storage or Azure Data Lake Storage Gen2, which are then accessed by Azure
Synapse, Databricks, or HDInsights as external tables.
When using Azure Synapse, you will only pay for the capabilities you opt in to use. During public preview, there will
not be a cost for provisioning an Azure Synapse workspace. Enabling a managed VNET and Customer Managed
Keys may incur a workspace fee after public preview. Pricing of workspaces with additional capabilities will be
announced at a future date.

Column family database cost

The main cost consideration for a column family database is that it needs to be massively scalable.
The Azure services that support column family databases are Azure Cosmos DB Cassandra API and HBase in
HDInsight.
For Azure Cosmos DB Cassandra API, you pay the cost of Cosmos DB, which includes database operations and
consumed storage. Cassandra API is open-source.
For HBase in HDInsight, you pay the cost of HDInsight, which includes instance size and number. HBase is open-
source.

Search engine database cost

Cost incurs for a search engine database when applications need to search for information held in external data
stores. It also needs to index massive volumes of data and provide near real-time access to these indexes.
Azure Cognitive Search is a search service that uses AI capabilities to identify and explore relevant content at scale.

Time series database cost

The main cost considerations for a time series database is the need to collect large amounts of data in real time
from a large number of sources. Although the records written to a time series database are generally small, there
are often a large number of records, and total data size can grow rapidly which will drive up cost.
Azure Time Series Insights may be the best option to minimize cost.

Object storage cost

Cost considerations include storing large binary objects such as images, text files, video and audio streams, large
application data objects and documents, and virtual machine disk images.
Azure services that support object storage are Azure Blob Storage and Azure Data Lake Storage Gen2.

Shared files cost

The main cost consideration is having the ability to access files across a network. For example, given appropriate
security and concurrent access control mechanisms, sharing data in this way can enable distributed services to
provide highly scalable data access for performing basic, low-level operations such as simple read and write
requests.
The Azure service that supports shared files is Azure Files.
 Azure messaging cost estimates
12/18/2020 • 2 minutes to read • Edit Online

The messaging services in this article have no up-front cost or termination fees, and you pay only for what you use.
In some cases, it's advantageous to combine two messaging services to increase the efficiency of your messaging
system. See Crossover scenarios for examples.
Cost is based on the number of operations or throughput units used depending on the message service. Using the
wrong messaging service for the intent can lead to higher costs. Before choosing a service, first, determine the
intent and requirements of the messages. Then, consider the tradeoffs between cost and operations/throughput
units. For tradeoff examples, see Technology choices for a message broker.
Use the Azure Pricing calculator for help creating various cost scenarios.

Service Bus cost

Connect on-premises and cloud-based applications and services to implement highly secure messaging workflows
using Service Bus. Cost is based on messaging operations and number of connections. The Basic tier is the
cheapest. If you want more operations and features, choose the Standard or Premium tier. For example, Service Bus
Premium runs in dedicated resources to provide higher throughput and more consistent performance.
For pricing details, see Service Bus pricing.

Event Grid cost

Manage routing of all events from any source to any destination to simplify event-based app development using
Event Grid. Event Grid can route a massive number of events per second per region. Cost is based on number of
operations performed. Examples of some operations are event ingress, subscription delivery attempts,
management calls, and filtering by subject suffix.
For pricing details, see Event Grid pricing.

Event Hubs
Stream millions of events per second from any source to build dynamic data pipelines and immediately respond to
business challenges using Event Hubs. Cost is based on throughput units. A key difference between Event Grid and
Event Hubs is in the way event data is made available to subscribers. For more information, see Pull model.
For questions and answers on pricing, see Pricing.
For pricing, see Event Hubs pricing.
 Networking cost estimates
12/18/2020 • 5 minutes to read • Edit Online

Traffic routing and load balancing

Most workloads have a load balancing service to route and distribute traffic in a way that a single resource isn't
overloaded, and the response time is minimum with maximum throughput.
Do you need to distribute traffic within a region or across regions?

Azure Front Door and Traffic Manager can distribute traffic to backends, clouds, or hybrid on-premises services
that reside in multiple regions. Otherwise, for regional traffic that moves within virtual networks or zonal and
zone-redundant service endpoints within a region, choose Application Gateway or Azure Load Balancer.
What is the type of traffic?

Load balancers such as Azure Application Gateway and Azure Front Door are designed to route and distribute
traffic to web applications or HTTP(S) endpoints. Those services support Layer 7 features such as SSL offload, web
application firewall, path-based load balancing, and session affinity. For non-web traffic, choose Traffic Manager or
Azure Load Balancer. Traffic Manager uses DNS to route traffic. Load Balancer supports Layer 4 features for all UDP
and TCP protocols.
Here’s the matrix for choosing load balancers by considering both dimensions.

SERVIC E GLO B A L / REGIO N A L REC O M M EN DED T RA F F IC

Azure Front Door Global HTTP(S)

Traffic Manager Global non-HTTP(S)

Application Gateway Regional HTTP(S)

Azure Load Balancer Regional non-HTTP(S)

For more information, see Choose a load balancing service.

Example cost analysis
Consider a web application that receives traffic from users across regions over the internet. To minimize the
request response time, the load balancer can delegate the responsibility from the web server to a different
machine. The application is expected to consume 10 TB of data. Routing rules are required to route incoming
requests to various paths. Also, we want to use Web Application Firewall (WAF) to secure the application through
policies.
By using the Azure Pricing Calculator, we can estimate the cost for these two services.

The example is based on the current price and is subject to change. The calculation shown is for information
purposes only.

Azure Front Door (West US), Zone 1.

Application Gateway (West US)
Consider a similar example where the type of traffic is changed. Instead the application is a UDP streaming service
that is deployed across regions and traffic goes over the internet. We can use a combination of Traffic Manager and
Azure Load Balancer. Traffic Manager is a simple routing service that uses DNS to direct clients to specific service.
Here are the cost estimates:
Azure Traffic Manager

IT EM EXA M P L E EST IM AT E

DNS queries 100 million x $0.54/per million = $54.00

Basic health checks 20 endpoints x $0.36/month = $7.20

Fast interval health checks 10 endpoints x $1.00/month = $10.00

External endpoints hosted outside Azure 10 endpoints x $0.54/month = $5.40

Real User Measurements 100 million user measurements x $2.00/month = $200.00

Traffic view 100 million data points processed x $2.00/month = $200.00

Total $486.60

Azure Load Balancer

IT EM EXA M P L E EST IM AT E

Rules First 5 rules: $0.032/rule/hour x 730 = $18.25

Additional 5 rules: $0.013/rule/hour x 730 = $36.50

Data processed 10 TB x $0.005/GB = 51.20

 IT EM EXA M P L E EST IM AT E

Total $105.95

Peering
Do you need to connect vir tual networks?

Peering technology is a service used for Azure virtual networks to connect with other virtual networks in the same
or different Azure region. Peering technology is used often in hub and spoke architectures.
An important consideration is the additional costs incurred by peering connections on both egress and ingress
traffic traversing the peering connections.

Keeping the top talking services of a workload within the same virtual network, zone and/or region
unless otherwise required. Use virtual networks as shared resources for multiple workloads against a single
virtual network per workload approach. This approach will localize traffic to a single virtual network and avoid
the additional costs on peering charges.

Example cost analysis

The example is based on the current price and is subject to change. The calculation shown is for information
purposes only.

Peering within the same region is cheaper than peering between regions or Global regions. For instance, you
consume 50 TB per month by connecting two VNETs in Central US. Using the current price here is the incurred cost.

IT EM EXA M P L E EST IM AT E

Ingress traffic 50 TB x $0.0100/GB = $512.50

Egress traffic 50 TB x $0.0100/GB = $512.50

Total $1,025.00

Let’s compare that cost for cross-region peering between Central US and East US.

IT EM EXA M P L E EST IM AT E

Ingress traffic 50 TB x $0.0350/GB = $1,792.00

Egress traffic 50 TB x $0.0350/GB = $1,792.00

Total $3,584.00

Hybrid connectivity
There are two main options for connecting an on-premises datacenter to Azure datacenters:
Azure VPN Gateway can be used to connect a virtual network to an on-premises network through a VPN
appliance or to Azure Stack through a site-to-site VPN tunnel.
Azure ExpressRoute creates private connections between Azure datacenters and infrastructure that’s on premise
 or in a colocation environment. What is the required throughput for cross-premises connectivity?

VPN gateway is recommended for development/test cases or small-scale production workloads where throughput
is less than 100 Mbps. Use ExpressRoute for enterprise and mission-critical workloads that access most Azure
services. You can choose bandwidth from 50 Mbps to 10 Gbps.
Another consideration is security. Unlike VPN Gateway traffic, ExpressRoute connections don't go over the public
internet. VPN Gateway traffic is secured by industry standard IPsec.
For both services, inbound transfers are free and outbound transfers are billed per the billing zone.
For more information, see Choose a solution for connecting an on-premises network to Azure.
This blog post provides a comparison of the two services.
Example cost analysis
This example compares the pricing details for VPN Gateway and ExpressRoute.

The example is based on the current price and is subject to change. The calculation shown is for information
purposes only.

Azure VPN Gateway

Suppose, you choose the VpnGw2AZ tier, which supports availability zones; 1 GB/s bandwidth. The workload
needs 15 site-to-site tunnels and 1 TB of outbound transfer. The gateway is provisioned and available for 720
hours.

IT EM EXA M P L E EST IM AT E

VPN Hours 720 x $0.564 = $406.08

Site to Site (S2S) Tunnels The first 10 tunnels are free. For additional tunnels, the cost is
5 tunnels x 720 hours x $0.015 per hour per tunnel = $54.00

Outbound transfer 1024 GB x 0.086 = $88.65

Total cost per month $548.73

ExpressRoute
Let's choose the Metered Data plan in billing zone of Zone 1 .

IT EM EXA M P L E EST IM AT E

Circuit bandwidth 1 GB/s speed has a fixed rate of $436.00 for the standard
price.

Outbound transfer 1 TB x $0.0025 = $25.60

Total cost per month $461.00

The main cost driver is outbound data transfer. ExpressRoute is more cost-effective than VPN Gateway when
consuming large amounts of outbound data. If you consume more than 200 TB per month, consider ExpressRoute
with the Unlimited Data plan where you're charged a flat rate.
For pricing details, see:
Azure VPN Gateway pricing
Azure ExpressRoute pricing
 Networking resources provisioning
12/18/2020 • 7 minutes to read • Edit Online

For design considerations, see Networking resource choices.

Azure Front Door

Azure Front Door billing is affected by outbound data transfers, inbound data transfers, and routing rules. The
pricing chart doesn't include the cost of accessing data from the backend services and transferring to Front Door.
Those costs are billed based on data transfer charges, described in Bandwidth Pricing Details.
Another consideration is Web Application Firewall (WAF) settings. Adding policies will drive up the cost.
For more information, see Azure Front Door Pricing.
Reference architecture
Highly available multi-region web application uses Front Door to route incoming requests to the primary region. If
the application running that region becomes unavailable, Front Door fails over to the secondary region.

Azure Application Gateway

There are two main pricing models:
Fixed price
You're charged for the time that the gateway is provisioned and available and the amount of data processed
by the gateway. For more information, see Application Gateway pricing.
Consumption price
This model applies to v2 SKUs that offer additional features such as autoscaling, Azure Kubernetes Service
Ingress Controller, zone redundancy, and others. You're charged based on the consumed capacity units. The
capacity units measure the compute resources, persistent connections, and throughput. Consumption price
is charged in addition to the fixed price.
For more information, see:
Application Gateway v2 pricing
Application Gateway pricing
Reference architecture
Microservices architecture on Azure Kubernetes Service (AKS) uses Application Gateway as the ingress
controller.
Securely managed web applications uses Application Gateway as a web traffic load balancer operating at
Layer 7 that manages traffic to the web application. Web Application Firewall (WAF) is enabled to enhance
security.

Azure ExpressRoute
There are two main pricing models:
Metered Data plan
 There are two pricing tiers: Standard and Premium , which is priced higher. The tier pricing is based on the
circuit bandwidth.
If you don't need to access the services globally, choose Standard . With this tier, you can connect to regions
within the same zone at no additional cost. Outbound cross-zonal traffic incurs more cost.
Unlimited Data plan
All inbound and outbound data transfer is included in the flat rate. There are two pricing tiers: Standard and
Premium , which is priced higher.
Calculate your utilization and choose a billing plan. The Unlimited Data plan is recommended if you exceed
about 68% of utilization.
For more information, see Azure ExpressRoute pricing.
Reference architecture
Connect an on-premises network to Azure using ExpressRoute connects an Azure virtual network and an on-
premises network connected using with VPN gateway failover.

Azure Firewall
Azure Firewall usage can be charged at a fixed rate per deployment hour. There's additional cost for the amount of
data transferred.
There aren't additional cost for a firewall deployed in an availability zone. There are additional costs for inbound
and outbound data transfers associated with availability zones.
When compared to network virtual appliances (NVAs), with Azure Firewall you can save up to 30-50%. For more
information see Azure Firewall vs NVA.
Reference architecture
Hub-spoke network topology in Azure
Deploy highly available NVAs

Azure Load Balancer

The service distributes inbound traffic according to the configured rules.
There are two tiers: Basic and Standard .
The Basic tier is free.
For the Standard tier, you are charged only for the number of configured load-balancing and outbound rules.
Inbound NAT rules are free. There's no hourly charge for the load balancer when no rules are configured.
See Azure Load Balancer Pricing for more information.
Reference architecture
Connect an on-premises network to Azure using ExpressRoute: Multiple subnets are connected through
Azure load balancers.
SAP S/4HANA in Linux on Azure: Distribute traffic to virtual machines in the application-tier subnet.
Extend an on-premises network using VPN Internal load balancer. Network traffic from the VPN gateway is
routed to the cloud application through an internal load balancer. The load balancer is located in the front-
end subnet of the application.
Azure VPN Gateway
When provisioning a VPN Gateway resource, choose between two gateway types:
VPN gateway to send encrypted traffic across the public internet. Site-to-Site, Point-to-Site, and VNet-to-VNet
connections all use a VPN gateway.
ExpressRoute gateway - To send network traffic on a private connection. This is used when configuring Azure
ExpressRoute.
For VPN gateway, select Route-based or Policy-based based on your VPN device and the kind of VPN
connection you want to create. Route-based gateway allows point-to-site, inter-virtual network, or multiple site-to-
site connections. Policy based only allows one site-to-site tunnel. Point-to-site isn't supported. So, route-based VPN
is more expensive.
Next, you need to choose the SKU for Route-based VPN. For developer/test workloads, use Basic . For production
workloads, an appropriate Generation1 or Generation2 SKU. Each SKU has a range and pricing depends on the
type of VPN gateway because each type offers different levels of bandwidth, site-to-site, and point-to-site tunnel
options. Some of those types also offer availability zones, which are more expensive. If you need higher bandwidth,
consider Azure ExpressRoute.
VPN gateway can be the cost driver in a workload because charges are based on the amount of time that the
gateway is provisioned and available.
All inbound traffic is free, all outbound traffic is charged as per the bandwidth of the VPN type. Bandwidth also
varies depending on the billing zone.
For more information, see
Hybrid connectivity
VPN Gateway Pricing.
Traffic across zones
Bandwidth Pricing Details.
Reference architecture
Extend an on-premises network using VPN connects the virtual network to the on-premises network through a
VPN device.

Traffic Manager
Traffic manager uses DNS to route and load balance traffic to service endpoints in different Azure regions. So, an
important use case is disaster recovery. In a workload, you can use Traffic Manager to route incoming requests to
the primary region. If that region becomes unavailable, Traffic Manager fails over to the secondary region. There
are other features that can make the application highly responsive and available. Those features cost money.
Determine the best web app to handle request based on geographic location.
Configure caching to reduce the response time.
Traffic Manager isn't charged for bandwidth consumption. Billing is based on the number of DNS queries received,
with a discount for services receiving more than 1 billion monthly queries. You're also charged for each monitored
endpoint.
Reference architecture
Multi-region N-tier application uses Traffic Manager to route incoming requests to the primary region. If that
region becomes unavailable, Traffic Manager fails over to the secondary region. For more information, see the
section Traffic Manager configuration.
DNS query charges
Traffic Manager uses DNS to direct clients to specific service.
Only DNS queries that reach Traffic Manager are charged in million query units. For 100 million DNS queries
month, the charges will be $54.00 a month based on the current Traffic Manager pricing.
Not all DNS queries reach Traffic Manager. Recursive DNS servers run by enterprises and ISPs first attempt to
resolve the query by using cached DNS responses. Those servers query Traffic Manager at a regular interval to get
updated DNS entries. That interval value or TTL is configurable in seconds. TTL can impact cost. Longer TTL
increases the amount of caching and reduces DNS query charges. Conversely, shorter TTL results in more queries.
However, there is a tradeoff. Increased caching also impacts how often the endpoint status is refreshed. For
example, the user failover times, for an endpoint failure, will become longer.
Health monitoring charges
When Traffic Manager receives a DNS request, it chooses an available endpoint based on configured state and
health of the endpoint. To do this, Traffic Manager continually monitors the health of each service endpoint.
The number of monitored endpoints are charged. You can add endpoints for services hosted in Azure and then add
on endpoints for services hosted on-premises or with a different hosting provider. The external endpoints are more
expensive, but health checks can provide high-availability applications that are resilient to endpoint failure,
including Azure region failures.
Real User Measurement charges
Real User Measurements evaluates network latency from the client applications to Azure regions. That influences
Traffic Manager to select the best Azure region in which the application is hosted. The number of measurements
sent to Traffic Manager is billed.
Traffic view charges
By using Traffic View, you can get insight into the traffic patterns where you have endpoints. The charges are based
on the number of data points used to create the insights presented.

Virtual Network
Azure Virtual Network is free. You can create up to 50 virtual networks across all regions within a subscription.
Here are a few considerations:
Inbound and outbound data transfers are charged per the billing zone. Traffic that moves across regions and
billing zones are more expensive. For more information, see:
Traffic across zones
Bandwidth Pricing Details.
VNET Peering has additional cost. Peering within the same region is cheaper than peering between regions
or Global regions. Inbound and outbound traffic is charged at both ends of the peered networks. For more
information, see Peering
Managed services (PaaS) don't always need a virtual network. The cost of networking is included in the
service cost.
 Web application cost estimates
11/2/2020 • 4 minutes to read • Edit Online

All web applications (apps) have no up-front cost or termination fees. Some charge only for what you use and
others charge per-second or per-hour. In addition, all web apps run in App Service plans. Together these costs can
help determine the total cost of a web app.
Use the Azure Pricing calculator to help create various cost scenarios.

App Service plans

The App Service plans include the set of compute resources needed for the web app to run. Except for Free tier, an
App Service plan carries a charge on the compute resources it uses. For a description of App Service plans, see
Azure App Service plan overview.
You can potentially save money by choosing one or more App Service plans. To help find the solution that meets
your business requirements, see Should I put an app in a new plan or an existing plan?
The pricing tier of an App Service plan determines what App Service features you get and how much you pay for
the plan. The higher the tier, the higher the cost. For details on the pricing tiers, see App Service Pricing.
Are there exceptions that impact App Ser vices plan cost?

You don't get charged for using the App Service features that are available to you (e.g., configuring custom
domains, TLS/SSL certificates, deployment slots, backups, etc.). For a list of exceptions, see How much does my App
Service plan cost?

App Service cost

App Service plans are billed on a per second basis. If your solution includes several App Service apps, consider
deploying them to separate App Service plans. This approach enables you to scale them independently because
they run on separate instances, which saves unnecessary cost.
Azure App Service supports two types of SSL connections: Server Name Indication (SNI) SSL Connections and IP
Address SSL Connections. SNI-based SSL works on modern browsers while IP-based SSL works on all browsers. If
your business requirements allow, use SNI-based SSL instead of IP-based SSL. There is no charge for SNI-based. IP-
based SSL incurs a cost per connection.
See the SSL Connections section in App Service pricing for details.

API Management cost

If you want to publish APIs hosted on Azure, on-premises, and in other clouds more securely, reliably, and at scale,
use API Management. The pricing tier you choose depends on the features needed based on the business
requirements. For example, depending on your requirements for scalability, the number of units required to scale
out range from 1-10. As the number of units increases, the cost of the pricing tier increases.
Self-hosted gateways are available in the Developer and Premium pricing tiers. There is an additional cost for this
service using the Premium tier. There is no cost for self-hosted gateways if using the Developer tier.
For pricing details, see API Management Pricing.
Content Delivery Network (CDN) cost
If you want to offer users optimal online content delivery, use CDN in your workload. The data size being used has
the most significant impact on cost, so you'll want to choose based on your business requirements. Purchasing data
in increments of TB is significantly higher than purchasing data in increments of GB.
The provider you choose also can impact cost. Choose Microsoft as the provider to get the lowest cost.
Some data, such as shopping carts, search results, and other dynamic content, is not cacheable. CDN offers
Acceleration Data Transfers, also called Dynamic Site Acceleration (DSA), to accelerate this type of web content. The
price for DSA is the same across Standard and Premium profiles.
For pricing details, see Content Delivery Network pricing.

Azure Cognitive Search cost

To set up and scale a search experience quickly and cost-effectively, use Azure Cognitive Search. When you create
an Azure Cognitive Search service, a resource is created at a pricing tier (or SKU) that's fixed for the lifetime of the
service. Billing depends on the type of service. See Search API for a list of services.
The charges are based on the number of transactions for each type of operation specific to a service. A certain
number of transactions are free. If you need additional transactions, choose from the Standard instances.
For pricing details, see Azure Cognitive Search pricing.

Azure SignalR cost

Use SignalR for any scenario that requires pushing data from server to client in real time. For example, Azure
SignalR provides secure and simplified communication between client and app one-to-one (e.g., chat window) or
one-to-many (instant broadcasting, IoT dashboards, or notification to social network). To learn more about Azure
SignalR, see What is Azure SignalR Service?
The Free tier is not recommended for a production environment. With the Standard tier, you pay only for what you
use. A Standard tier is recommended as an enterprise solution because it offers a large number of concurrent
connections and messages.
For pricing details, see Azure SignalR Service pricing.

Notification Hubs cost

To broadcast push notifications to millions of users at once, or tailor notifications to individual users, use
Notification Hubs. Pricing is based on number of pushes. The Free tier is a good starting point for exploring push
capabilities but is not recommended for production apps. If you require more pushes and features such as
scheduled pushes or multi-tenancy, you can buy additional pushes per million for a fee.
See Push notifications with Azure Notification Hubs: Frequently asked questions for more information. For pricing
details, see Notification Hubs pricing.
 Checklist - Monitor cost
12/18/2020 • 2 minutes to read • Edit Online

Use this checklist to monitor the cost of the workload.

Gather cost data from diverse sources to create repor ts . Start with tools like Azure Advisor and
Azure Cost Management. Build custom reports relevant for the business by using Consumption APIs.
Cost reports
Review costs in cost analysis
Use resource tag policies to build repor ts . Tags can be used to identify the owners of systems or
applications and create custom reports.
Follow a consistent tagging standard
Video: How to review tag policies with Azure Cost Management
Use RBAC built-in roles for cost . Only give access to users who are intended to view and analyze cost
reports. The roles are defined per scope. For example, use the Cost Management Reader role to enable
users to view costs for their resources in subscriptions or resource groups.
Provide the right level of cost access
Azure RBAC scopes
Respond to aler ts and have a response plan according to the constraints. Respond to alerts quickly
and identify possible causes and any required action.
Budget and alerts
Use cost alerts to monitor usage and spending
Adopt both proactive and reactive approaches for cost reviews . Conduct cost reviews at a regular
cadence to determine the cost trend. Also review reports that are created because of alerts.
Conduct cost reviews
Participate in central governance cost reviews
Analyze the cost at all scopes by using Cost analysis. Identify services that are driving the cost through
different dimensions, such as location, usage meters, and so on. Review whether certain optimizations are
bringing results. For example, analyze costs associated with reserved instances and Spot VMs against
business goals.
Quickstart: Explore and analyze costs with cost analysis
Detect anomalies and identify changes in business or applications that might have contributed changes in
cost. Focus on these factors:
Traffic pattern as the application scales.
Budget for the usage meters on resources.
Performance bottle necks.
CPU utilization and network throughput.
Storage footprint for blobs, backups, archiving.
Use Visualization tools to analyze cost information.
Create visuals and reports with the Azure Cost Management connector in Power BI Desktop
Cost Management App
 Set budgets and alerts
11/2/2020 • 2 minutes to read • Edit Online

Azure Cost Management has an alert feature. Alerts are generated when consumption reaches a threshold.
Consider the metrics that each resource in the workload. For each metric, build alerts on baseline thresholds. This
way, the admins can be alerted when the workload is using the services at capacity. The admins can then tune the
resources to target SKUs based on current load.
You can also set alerts on allowed budgets at the resource group or management groups scopes. Both cloud
services performance and budget requirements can be balanced through alerts on metrics and budgets.
Over time, the workload can be optimized to autoheal itself when alerts are triggered. For information about using
alerts, see Use cost alerts to monitor usage and spending.

Respond to alerts
When you receive an alert, check the current consumption data. Budget alerts aren't triggered in real time. There
may be a delay between the alert and the current actual cost. Look for significant difference between cost values
when the alert happened and the current cost. Next, conduct a cost review to discuss the cost trend, possible
causes, and any required action. For information about stakeholders in a cost review, see Cost reviews.
Determine short and long-term actions justified by business value. Can a temporary increase in the alert threshold
be a feasible fix? Does the budget need to be increased longer-term? Any increase in budget must be approved.
If the alert was caused because of unnecessary or expensive resources, you can implement additional Azure Policy
controls. You can also add budget automation to trigger resource scaling or shutdowns.

Revise budgets
After you identify and analyze your spending patterns, you can set budget limits for applications or business units.
You'll want to assign access to view or manage each budget to the appropriate groups. Setting several alert
thresholds for each budget can help track your burn down rate.
 Generate cost reports
11/2/2020 • 3 minutes to read • Edit Online

To monitor the cost of the workload, use Azure cost tools or custom reports. The reports can be scoped to business
units, applications, IT infrastructure shared services, and so on. Make sure that the information is consistently
shared with the stakeholders.

Azure cost tools

Azure provides cost tools that can help track cloud spend and make recommendations.
Azure Advisor
Azure Cost Management
Cost analysis
Cost analysis is a tool in Azure Cost Management that allows you to view aggregated costs over a period. This
view can help you understand your spending trends.
View costs at different scopes, such as for a resource group or specific resource tags. Cost Analysis provides built-in
charts and custom views. You can also download the cost data in CSV format to analyze with other tools.
For more information, see: Quickstart: Explore and analyze costs with cost analysis.

NOTE
There are many ways of purchasing Azure services. Not all are supported by Azure Cost Management. For example, detailed
billing information for services purchased through a Cloud Solution Provider (CSP) must be obtained directly from the CSP.
For more information about the supported cost data, see Understand cost management data.

Advisor recommendations
Azure Advisor recommendations for cost can highlight the over-provisioned services and ways to lower cost. For
example, the virtual machines that should be resized to a lower SKU, unprovisioned ExpressRoute circuits, and idle
virtual network gateways.
For more information, see Advisor cost management recommendations.

Consumption APIs
Granular and custom reports can help track cost over time. Azure provides a set of Consumption APIs to generate
such reports. These APIs allow you to query and create various cost data. Data includes usage data for Azure
services and third-party services through Marketplace, balances, budgets, recommendations on reserved
instances, among others. You can configure Azure Role-based Access control policies to allow only a certain set of
users or applications access the data.
For example, you want to determine the cost of all resources used in your workload for a given period. One way of
getting this data is by querying usage meters and the rate of those meters. You also need to know the billing period
of the usage. By combining these APIs, you can estimate the consumption cost.
Billing account API: To get your billing account to manage your invoices, payments, and track costs.
Billing Periods API: To get billing periods that have consumption data.
Usage Detail API: To get the breakdown of consumed quantities and estimated charges.
 Marketplace Store Charge API: To get usage-based marketplace charges for third-party services.
Price Sheet API: To get the applicable rate for each meter.
The result of the APIs can be imported into analysis tools.

NOTE
Consumption APIs are supported for Enterprise Enrollments and Web Direct Subscriptions (with exceptions). Check
Consumption APIs for updates to determine support for other types of Azure subscriptions.

For more information about common cost scenarios, see Billing automation scenarios.

Custom scripts
Use Azure APIs to schedule custom scripts that identify orphaned or empty resources. For example, unattached
managed disks, load balancers, application gateways, or Azure SQL Servers with no databases. These resources
incur a flat monthly charge while unused. Other resources may be stale, for example VM diagnostics data in blob or
table storage. To determine if the item should be deleted, check its last use and modification timestamps.

Analyze and visualize

Start with the usage details in the invoice. Review that information against relevant business data and events. If
there are anomalies, evaluate the significant changes in business or applications that might have contributed those
changes.
Power BI Desktop has a connector that allows you to connect billing data from Azure Cost Management. You can
create custom dashboards and reports, ask questions of your data and publish, and share your work.

NOTE
Sharing requires Power BI Premium licenses.

For more information, see Create visuals and reports with the Azure Cost Management connector in Power BI
Desktop.
You can also use the Cost Management App. The app uses Azure Cost Management Template app for Power BI. You
can import and analyze usage data and incurred cost within Power BI.
 Conduct cost reviews
12/18/2020 • 2 minutes to read • Edit Online

The goal of cost monitoring is to review cloud spend with the intent of establishing cost controls and preventing
any misuse. The organization should adopt both proactive and reactive review approaches for monitoring cost.
Effective cost reviews must be conducted by accountable stakeholders at a regular cadence. The reviews must be
complemented with reactive cost reviews, for example when a budget limit causes an alert.
Who should be included in a cost review?

The financial stakeholders must understand cloud billing to derive business benefits using financial metrics to
make effective decisions.
Also include members of the technical team. Application owners, systems administrators who monitor and back-up
cloud systems, and business unit representatives must be aware of the decisions. They can provide insights
because they understand cloud cost metering and cloud architecture. One way of identifying owners of systems or
applications is through resource tags.
What should be the cadence of cost reviews?

Cost reviews can be conducted as part of the regular business reviews. It’s recommended that such reviews are
scheduled,
During the billing period. This review is to create an awareness of the estimated pending billing. These reports
can be based on Azure Advisor and Azure Cost Management – Cost analysis.
After the billing period. This review shows the actual cost with activity for that month. Use Balance APIs to
generate monthly reports. The APIs can query data that gets information on balances, new purchases, Azure
Marketplace service charges, adjustments, and overage charges.
Because of a budget alert or Azure Advisor recommendations.
Web Direct (pay-as-you-go) and Cloud Solution Provider (CSP) billing occurs monthly. While Enterprise Agreement
(EA) billing occurs annually, costs should still be reviewed monthly.
 Checklist - Optimize cost
11/2/2020 • 2 minutes to read • Edit Online

Continue to monitor and optimize the workload by using the right resources and sizes. Use this checklist to
optimize a workload.
Review the underutilized resources . Evaluate CPU utilization and network throughput over time to
check if the resources are used adequately. Azure Advisor identifies underutilized virtual machines. You can
choose to decommission, resize, or shut down the machine to meet the cost requirements.
Resize virtual machines
Shutdown the under utilized instances
Continuously take action on the cost reviews . Treat cost optimization as a process, rather than a point-
in-time activity. Use tooling in Azure that provides recommendations on usage or cost optimization. Review
the cost management recommendations and take action. Make sure that all stakeholders are in agreement
about the implementation and timing of the change.
Recommended tab in the Azure portal
Recommendations in the Cost Management Power BI app
Recommendations in Azure Advisor
Recommendations using Reservation REST APIs
Use reser ved instances on long running workloads . Reserve a prepaid capacity for a period, generally
one or three years. With reserved instances, there’s a significant discount when compared with pay-as-you-
go pricing.
Reserved instances
Use discount prices . These methods of buying Azure resources can lower costs.
Azure Hybrid Use Benefit
Azure Reservations
There are also payment plans offered at a lower cost:
Microsoft Azure Enterprise Agreement
Enterprise Dev Test Subscription
Cloud Service Provider (Partner Program)
Have a scale-in and scale-out policy . In a cost-optimized architecture, costs scale linearly with demand.
Increasing customer base shouldn't require more investment in infrastructure. Conversely, if demand drops,
scale-down of unused resources. Autoscale Azure resources when possible.
Autoscale instances
Reevaluate design choices . Analyze the cost reports and forecast the capacity needs. You might need to
change some design choices.
Choose the right storage tier . Consider using hot, cold, archive tier for storage account data.
Storage accounts can provide automated tiering and lifecycle management. For more information,
see Review your storage options
Choose the right data store . Instead of using one data store service, use a mix of data store
depending on the type of data you need to store for each workload. For more information, see
Choose the right data store.
Choose Spot VMs for low priority workloads . Spot VMs are ideal for workloads that can be
interrupted, such as highly parallel batch processing jobs.
Spot VMs
Optimize data transfer . Only deploy to multiple regions if your service levels require it for either
availability or geo-distribution. Data going out of Azure datacenters can add cost because pricing is
based on Billing Zones.
Traffic across billing zones and regions
Reduce load on ser vers . Use Azure Content Delivery Network (CDN) and caching service to
reduce load on front-end servers. Caching is suitable for servers that are continually rendering
dynamic content that doesn't change frequently.
Use managed ser vices . Measure the cost of maintaining infrastructure and replace it with Azure
PaaS or SaaS services.
Managed services
 Autoscale instances
11/2/2020 • 2 minutes to read • Edit Online

In Azure, it’s easier to grow a service with little to no downtime against downscaling a service, which usually
requires deprovisioning or downtime. In general, opt for scale-out instead of scale up.
For certain application, capacity requirements may swing over time. Autoscaling policies allow for less error-prone
operations and cost savings through robust automation.

Virtual machine instances

For auto scaling, consider the choice of instance size. The size can significantly change the cost of your workload.

Choose smaller instances where workload is highly variable and scale out to get the desired level of
performance, rather than up. This choice will enable you to make your cost calculations and estimates granular.

Stateless applications
Many Azure services can be used to improve the application's ability to scale dynamically. Even where they may not
have been originally designed to do so.
For example, many ASP.NET stateful web applications can be made stateless. Then they can be auto scaled, which
results in cost benefit. You store state using Azure Redis Cache, or Cosmos DB as a back-end session state store
through a Session State Provider.
 Reserved instances
11/2/2020 • 2 minutes to read • Edit Online

Azure Reservations are offered on many services as a way to lower cost. You reserve a prepaid capacity for a
period, generally one or three years. With reserved instances, there’s a significant discount when compared with
pay-as-you-go pricing. You can pay up front or monthly, price wise both are same.

Usage pattern
Before opting for reserved instances, analyze the usage data with pay-as-you-go prices over a time duration.
Do the ser vices in the workload run intermittently or follow long-term patterns?

Azure provides several options that can help analyze usage and make recommendations by comparing
reservations prices with pay-as-you-go price. An easy way is to view the Recommended tab in the Azure portal.
Azure Advisor also provides recommendations that are applicable to an entire subscription. If you need a
programmatic way, use the Reservation Recommendations REST APIs.
Reserved instances can lower cost for long running workloads. For intermittent workloads, prepaid capacity might
go unused and it doesn't carry over to the next billing period. Usage exceeding the reserved quantity is charged
using more expensive pay-as-you-go rates. But there are some flexible options. You can exchange or refund a
reservation within limits. For more information, see Self-service exchanges and refunds for Azure Reservations.

Scope
Reservations can be applied to a specific scope—subscription, resource group, or a single resource. Suppose you
set the scope as the resource group, the reservation savings will apply to the applicable resources provisioned in
that group. For more information, see Scope reservations.

Services, subscription, and offer types

Many services are eligible for reservations. This range covers virtual machines and a wide variety of managed
services. For information about the services that are eligible for reservations, see Charges covered by reservation.
Certain subscriptions and offer types support reservations. For a list of such subscriptions, see Discounted
subscription and offer types.
 Virtual machines
11/2/2020 • 3 minutes to read • Edit Online

Virtual machines can be deployed in fix-sized blocks. These VMs must be adequately sized to meet capacity
demand and reduce waste.
For example, look at a VM running the SAP on Azure project can show you how initially the VM was sized based on
the size of existing hardware server (with cost around €1 K per month), but the real utilization of VM was not more
than 25% - but simple choosing the right VM size in the cloud we can achieve 75% saving (resize saving). And by
applying the snoozing you can get additional 14% of economy:

It is easy to handle cost comparison when you are well equipped and for this Microsoft provides the set of specific
services and tools that help you to understand and plan costs. These include the TCO Calculator, Azure Pricing
Calculator, Azure Cost Management (Cloudyn), Azure Migrate, Cosmos DB Sizing Calculator, and the Azure Site
Recovery Deployment Planner.
Here are some strategies that you can use to lower cost for virtual machines.

Resize virtual machines

You can lower cost by managing the size and the number of VMs.

Determine the load by analyzing the CPU utilization to make sure that the instance is adequately utilized.

Ideally, with the right size, the current load should fit in a lower SKU of the same tier. Another way is to lower the
number instances and still keep the load below a reasonable utilization. Azure Advisor recommends load less than
80% utilization for non-user facing workloads and 40% when user-facing workload. It also provides current and
target SKU information.
You can identify underutilized machines by adjusting the CPU utilization rule on each subscription.
Resizing a virtual machine does require the machine to be shut down and restarted. There might be a period of
time when it will be unavailable. Make sure you carefully time this action for minimal business impact.

Shut down the under utilized instances

Use the Star t/stop VMs during off-hours feature of virtual machines to minimize waste. There are many
configuration options to schedule start the stop times. The feature is suitable as a low-cost automation option. For
information, see Start/stop VMs during off-hours solution in Azure Automation.
Azure Advisor evaluates virtual machines based on CPU and network utilization over a time period. Then, the
recommended actions are shut down or resize instances and cost saving with both actions.

Spot VMs
Some workloads don't have a business requirement to complete a job within a period.
Can the workload be interrupted?

Spot VMs are ideal for workloads that can be interrupted, such as highly parallel batch processing jobs. These VMs
take advantage of the surplus capacity in Azure at a lower cost. They're also well suited for experimental,
development, and testing of large-scale solutions.
For more information, see Use Spot VMs in Azure.

Reserved VMs
Virtual machines are eligible for Azure Reservations. You can prepay for VM instances if you can commit to one or
three years. Reserved instances are appropriate for workloads that have a long-term usage pattern.
The discount only applies to compute and not the other meters used to measure usage for VMs. The discount can
be extended to other services that emit VM usage, such as Virtual machine scale sets and Container services, to
name a few. For more information, see Software costs not included with Azure Reserved VM Instances and Services
that get VM reservation discounts.
With reserved instances, you need to determine the VM size to buy. Analyze usage data using Reser vations
Consumption APIs and follow the recommendations of Azure portal and Azure Advisor to determine the size.
Reservations also apply to dedicated hosts. The discount is applied to all running hosts that match the reservation
scope and attributes. An important consideration is the SKU for the host. When selecting a SKU, choose the VM
series and type eligible to be a dedicated host. For more information, see Azure Dedicated Hosts pricing.
For information about discounts on virtual machines, see How the Azure reservation discount is applied to virtual
machines.
 Caching data
11/2/2020 • 2 minutes to read • Edit Online

Caching is a strategy where you store a copy of the data in front of the main data store. The cache store is typically
located closer to the consuming client than the main store. Advantages of caching include faster response times
and the ability to serve data quickly. In doing so, you can save on the overall cost. Be sure to assess the built-in
caching features of Azure services used in your architecture. Azure also offers caching services, such as Azure
Cache for Redis or Azure CDN.
For information about what type of data is suitable for caching, see Caching.

Lower costs associated with reliability and latency

Caching can be a cheaper way of storing data, providing reliability, and reducing network latency.
Depending on the type of the data, determine if you need the complex capabilities of the backend data store,
such as data consistency. If the data is fully static, you can choose to store it only in a caching store. If the data
doesn't change frequently, consider placing a copy of the data in a caching store and refresh it from time to time.
For example, an application stores images in blob storage. Every time it requests an image, the business logic
generates a thumbnail from the main image and returns it to the caller. If the main image doesn't change too
often, then return the previously generated thumbnails stored in a cache. This way you can save on resources
required to process the image and lower the request response rate.
If the backend is unavailable, the cache continues to handle requests by using the copy until the backend fails
over to the backup data store.
Caching can also be an effective way of reducing network latency. Instead of reaching the central server, the
response is sent by using the cache. That way, the client can receive the response almost instantaneously. If you
need higher network performance and the ability to support more client connections, choose a higher tier of the
caching service. However, higher tiers will incur more costs.

Caching can be expensive

Incorrect use of caching can result in severe business outcomes and higher costs.
Adding a cache will lead to multiple data sources in your architecture. There are added costs to keeping them
in sync. You may need to fill the cache before putting it in production. Filling the cache on the application's
first access can introduce latency or seeding the cache can have impact on application's start time. If you
don't refresh the cache, your customers can get stale data.
Invalidate the cache at the right time when there's latest information in the source system. Use strategies to
age out the cache when appropriate.
To make sure the caching layer is working optimally, add instrumentation. That feature will add complexity
and implementation cost.
Caching services such as Azure Cache for Redis are billed on the tier you choose. Pricing is primarily determined on
the cache size and network performance and they're dependent. A smaller cache will increase latency. Before
choosing a tier, estimate a baseline. An approach can be by load testing the number of users and cache size.
 Tradeoffs for costs
11/2/2020 • 3 minutes to read • Edit Online

As you design the workload, consider tradeoffs between cost optimization and other aspects of the design, such as
security, scalability, resilience, and operability.
What is most impor tant for the business: lowest cost, no downtime, high throughput?

An optimal design doesn’t equate to a low-cost design. There might be risky choices made in favor of a cheaper
solution.

Cost versus reliability

Cost has a direct correlation with reliability.
Does the cost of high availability components exceed the acceptable downtime?

Overall Service Level Agreement (SLA), Recovery Time Objective (RTO), and Recovery Point Objective (RPO) may
lead to expensive design choices. If your service SLAs, RTOs, and RPOs times are short, then higher investment is
inevitable for high availability and disaster recovery options.
For example, to support high availability, you choose to host the application across regions. This choice is costlier
than single region because of the replication costs or the need provisioning extra nodes. Data transfer between
regions will also add cost.
If the cost of high availability exceeds the cost of downtime, you can save by using Azure platform-managed
replication and recover data from the backup storage.
For resiliency, availability, and reliability considerations, see the Reliability pillar.

Cost versus performance efficiency

Boosting performance will lead to higher cost.
Many factors impact performance.
Fixed or consumption-based provisioning . Avoid cost estimation of a workload at consistently high
utilization. Consumption-based pricing will be more expensive that the equivalent provisioned pricing. Smooth out
the peaks to get a consistent flow of compute and data. Ideally, use manual and autoscaling to find the right
balance. Scaling up is more expensive than scaling out.
Azure regions . Cost scales directly with number of regions. Locating resources in cheaper regions shouldn't
negate the cost of network ingress and egress or by degraded application performance because of increased
latency.
Caching . Every render cycle of a payload consumes both compute and memory. You can use caching to reduce
load on servers and save with pre-canned storage and bandwidth costs, and the savings can be dramatic,
especially for static content services.
While caching can reduce cost, there are some performance tradeoffs. For example, Azure Traffic Manager pricing
is based on the number of DNS queries that reach the service. You can reduce that number through caching and
configure how often the cache is refreshed. Relying on the cache that isn't frequently updated will cause longer
user failover times if an endpoint is unavailable.
Batch or real-time processing . Using dedicated resources for batch processing long running jobs will increase
the cost. You can lower cost by provisioning Spot VMs but be prepared for the job to be interrupted every time
Azure evicts the VM.
For performance considerations, see the Performance Efficiency pillar.

Cost versus security

Increasing security of the workload will increase cost.
As a rule, don't compromise on security. For certain workloads, you can't avoid security costs. For example, for
specific security and compliance requirements, deploying to differentiated regions will be more expensive.
Premium security features can also increase the cost. There are areas you can reduce cost by using native security
features. For example, avoid implementing custom RBAC roles if you can use built-in roles.
For security considerations, see the Security Pillar.

Cost versus operational excellence

Investing in systems monitoring and automation might increase the cost initially but over time will reduce cost.
IT operations processes like user or application access provisioning, incident response, and disaster recovery
should be integrated with the workload.
Cost of maintaining infrastructure is more expensive. With PaaS or SaaS services, infrastructure, platform
management services, and additional operational efficiencies are included in the service pricing.
For operational considerations, see the Operational Excellence pillar.
 Overview of the operational excellence pillar
12/18/2020 • 2 minutes to read • Edit Online

This pillar covers the operations processes that keep an application running in production. Deployments must be
reliable and predictable. They should be automated to reduce the chance of human error. They should be a fast and
routine process, so they don't slow down the release of new features or bug fixes. Equally important, you must be
able to quickly roll back or roll forward if an update has problems.
To assess your workload using the tenets found in the Microsoft Azure Well-Architected Framework, see the
Microsoft Azure Well-Architected Review.
These are the disciplines we group in the operational excellence pillar:

O P ERAT IO N A L EXC EL L EN C E DISC IP L IN ES DESC RIP T IO N

Application design Provides guidance on how to design, build, and orchestrate

workloads with DevOps principles in mind

Monitoring Something that enterprises have been doing for years,

enriched with some specifics for applications running in the
cloud

Application performance management The monitoring and management of performance and

availability of software applications through DevOps

Code deployment How you deploy your application code is going to be one of
the key factors that will determine your application stability

Infrastructure provisioning Frequently known as "Automation" or "Infrastructure as code",

this discipline refers to best practices for deploying the
platform where your application will run on

Testing Testing is fundamental to be prepared for the unexpected and

to catch mistakes before they impact users
 Operational excellence principles
12/18/2020 • 2 minutes to read • Edit Online

Considering and improving how software is developed, deployed, operated, and maintained is one part of
achieving a higher competency in operations. Equally important is providing a team culture of experimentation and
growth, solutions for rationalizing the current state of operations, and incident response plans. The principles of
operational excellence are a series of considerations that can help achieve excellent operational practices.
To assess your workload using the tenets found in the Azure Well-Architected Framework, see the Microsoft Azure
Well-Architected Review.
Optimize build and release processes
From provisioning with Infrastructure as Code, building and releasing with CI/CD pipelines, automated testing, and
embracing software engineering disciplines across your entire environment. This approach ensures the creation
and management of environments throughout the software development lifecycle is consistent, repeatable, and
enables early detection of issues.
Monitor the entire system and understand operational health
Implement systems and processes to monitor build and release processes, infrastructure health, and application
health. Telemetry is critical to understanding the health of a workload and whether the service is meeting the
business goals.
Rehearse recover y and practice failure
Run DR drills on a regular cadence and use engineering practices to identify and remediate weak points in
application reliability. Regular rehearsal of failure will validate the effectiveness of recovery processes and ensure
teams are familiar with their responsibilities.
Embrace operational improvement
Continuously evaluate and refine operational procedures and tasks while striving to reduce complexity and
ambiguity. This approach enables an organization to evolve processes over time, optimizing inefficiencies, and
learning from failures.
Use loosely coupled architecture
Enable teams to independently test, deploy, and update their systems on demand without depending on other
teams for support, services, resources, or approvals.
Incident management
When incidents occur, have well thought out plans and solutions for incident management, incident
communication, and feedback loops. Take the lessons learned from each incident and build telemetry and
monitoring elements to prevent future occurrences.
 Automation overview: Goals, best practices, and
types
12/18/2020 • 4 minutes to read • Edit Online

Automation has revolutionized how businesses operate and this trend continues to evolve. Businesses have moved
to automating manual processes so that engineers can focus attention on tasks that add business value.
Automating business solutions allow you to:
Activate resources on demand.
Deploy solutions rapidly.
Minimize human error in configuring repetitive tasks.
Produce consistent and repeatable results.
To learn more, see Deployment considerations for automation.

Goals of automation
When automating technical processes and configurations, a common approach for some organizations is to
automate what they can and leave the more difficult processes for humans to perform manually.
A goal of automation is to make tools that do what humans can do, only better. For example, a human can perform
any given task once. But when the task requires repetitive runs, especially over long time periods, an automated
system is better equipped to do this with more predictable results that are error-free. Increasing speed is another
goal in automation. When you practice these automation goals, you can build systems that are faster, repetitive, and
can run on a daily basis.

Automate toil to improve efficiency

Most automation involves a percentage of toil. Toil is the operational work that is related to a process that is
manual, repetitive, can be automated, and has minimal value. It is counterproductive to automation but in many
organizations, a small amount of toil is unavoidable. It becomes an issue when too much toil slows progress. A
project's production velocity will decrease if engineers are continuously interrupted by manual tasks attributed to
toil, either planned or unplanned. Too much toil can impact job satisfaction. Engineers become dissatisfied when
they find themselves spending too much time on operational toil rather than on other projects.
Automation should be developed, and increased, so engineers can eliminate future toil. By reducing toil, engineers
can concentrate on innovating business solutions.
For more information, see Toil automation.

Automation best practices

Ensure consistency – The more manual processes are involved, the more prone to human error. This can lead
to mistakes, oversights, reduction of data quality, and reliability problems.
Centralize mistakes – Choose a platform to allow you to fix bugs in one place in order to fix them everywhere.
This eliminates the need for communicating with other humans which increases chance of error and the
possibility of the bug being re-introduced.
Identify issues quickly – Complex issue may not always be able to be identified in a timely manner. However,
with good automation, detection of these issues should occur quickly.
Maximize employee productivity – Automation leads to more innovative solutions, additional automation,
 and in general provide more value to the business. This in turn raises morale and job satisfaction. Automation
does not replace jobs with technology. For example, once a process is automated, training and maintenance can
be greatly reduced or eliminated. This frees engineers to spend less time on manual processes and more time
on automating business solutions.

Types of automation
Three types of automation are described in this article: infrastructure deployment, infrastructure configuration, and
operational tasks. These categories share the same goals and best practices mentioned previously. They differ in
areas where Azure provides solutions that help achieve optimal automation. Other types of automation such as
continuous deployment and continuous delivery are described further in the Operational Excellence pillar.
Infrastructure deployment
As businesses move to the cloud, they need to repeatedly deploy their solutions and know that their infrastructure
is in a reliable state. To meet these challenges, you can automate deployments using a practice referred to as
infrastructure as code. In code, you define the infrastructure that needs to be deployed.
Although there are many deployment technologies you can use with Azure, this article describes two of the more
popular technologies:
Azure Resource Manager (ARM) templates
Terraform
These technologies use a declarative approach. This lets you state what you intend to deploy without having to
write the sequence of programming commands to create it. You can deploy not only virtual machines, but also the
network infrastructure, storage systems, and any other resources you may need.
Infrastructure configuration
If you don't manage configuration carefully, your business could encounter disruptions such as systems outages
and security issues. Optimal configuration can enable you to quickly detect and correct configurations that could
interrupt or slow performance.
Configurations such as installing software on a virtual machine, adding data to a database, or starting pods in an
Azure Kubernetes Service cluster need to access Azure through an endpoint that is specific to your instance, outside
of the exposed REST APIs. This enables the configuration tools to use agents, networking, or other access methods
to provide resource-specific configuration options.
For example, when deploying to Azure, you may need to run post-deployment virtual machine configuration or run
other arbitrary code to bootstrap the deployed Azure resources. Another example is configuration tools that can be
used to configure and manage the ongoing configuration and state of Azure virtual machines.
Operational tasks
As the demand for speed in performing operational tasks increases over time, you are expected to deliver things
faster and faster. The old way of manually performing operational tasks won't scale to that type of demand now.
This is where automation can help. To meet on-demand delivery using an automation platform, you need to
develop automation components (such runbooks and configurations), create integrations to systems that are
already in place efficiently, and operate and troubleshoot.
Advantages of automating operational tasks include:
Optimize and extend existing processes (for example, using a PowerShell module or REST API).
Deliver flexible and reliable services.
Lower costs.
Improve predictability.
Two popular options for automating operational tasks are:
 Azure functions - Run code without managing the underlying infrastructure on where the code is run.
Azure automation - Uses a programming language to automate operational tasks in code and executed on
demand.
For more information, see Automation. To see a Microsoft Ignite video, see Automating Operational and
Management Tasks.

Next steps
Automate Repeatable Infrastructure
 Repeatable Infrastructure
12/18/2020 • 4 minutes to read • Edit Online

Historically, deploying a new service or application involves manual work such as procuring and preparing
hardware, configuring operating environments, and enabling monitoring solutions. Ideally, an organization would
have multiple environments in which to test deployments. These test environments should be similar enough to
production that deployment and run time issues are detected before deployment to production. This manual work
takes time, is error-prone, and can produce inconsistencies between the environments if not done well.
Cloud computing changes the way we procure infrastructure. No longer are we unboxing, racking, and cabling
physical infrastructure. We have internet accessible management portals and REST interfaces to help us. We can
now provision virtual machines, databases, and other cloud services on demand and globally. When we no longer
need cloud services, they can be easily deleted. However, cloud computing alone does not remove the effort and
risk in provisioning infrastructure. When using a cloud portal to build systems, many of the same manual
configuration tasks remain. Application servers require configuration, databases need networking, and firewalls
need firewalling.

Deploy infrastructure with code

To fully realize deployment optimization, reduce configuration effort, and automate full environments'
deployment, something more is required. One option is referred to as infrastructure as code.
Cloud computing changes so much about deploying and provisioning infrastructure. Not only can we procure
compute, data, and so many other service types on demand, we have APIs for doing so. Because of cloud service's
API-driven nature, programmatically deploying and configuring cloud services makes sense. The concept known
as infrastructure as code involves using a declarative framework to describe your desired service configuration.
Infrastructure as code solutions translate the declared configuration into the proper cloud provider API requests,
which, once deployed result in usable cloud services. Benefits of using infrastructure as code include:
Deploy similarly configured infrastructure across multiple environments e.g., test and production.
Deploy all required components as a single unit (infrastructure, monitoring solutions, and configured alerts).
Version control infrastructure in a source control solution.
Use continuous integration solutions to manage and test infrastructure deployments.
You can use many declarative infrastructure deployment technologies with Azure, here we detail two of the most
common.

Automate deployments with ARM Templates

Azure Resource Manager (ARM) Templates provide an Azure native infrastructure as code solution. ARM Templates
are written in a language derived from JavaScript Object Notation (JSON), and they define the infrastructure and
configurations for Azure deployments. An ARM template is declarative, you state what intend to deploy, provide
configuration values, and the Azure engine takes care of making the necessary Azure REST API put requests.
Additional benefits of using ARM templates for infrastructure deployments include:
Parallel resource deployment - the Azure deployment engine sequences resource deployments based on
defined dependencies. If dependencies do not exist between two resources, they are deployed at the same time.
Modular deployments - ARM templates can be broken up into multiple template files for reusability and
modularization.
Day one resource suppor t - ARM templates support all Azure resources and resource properties as they are
 released.
Extensibility - Azure deployments can be extended using deployment scripts and other automation solutions.
Validation - Azure deployments are evaluated against a validation API to catch configuration mistakes.
Testing - the ARM template test toolkit provides a static code analysis framework for testing ARM templates.
Change preview - ARM template what-if allows you to see what will be changed before deploying an ARM
template.
Tooling - Language service extensions are available for both Visual Studio Code and Visual Studio to assist in
authoring ARM templates.
The following example demonstrates a simple ARM template that deploys a single Azure Storage account. In this
example, a single parameter is defined to take in a name for the storage account. Under the resources section, a
storage account is defined, the storageName parameter is used to provide a name, and the storage account details
are defined. See the included documentation for an in-depth explanation of the different sections and
configurations for ARM templates.

{
"$schema": "https://schema.management.azure.com/schemas/2019-04-01/deploymentTemplate.json#",
"contentVersion": "1.0.0.0",
"parameters": {
"storageName": {
"type": "string",
"defaultValue": "newStorageAccount"
}
},
"resources": [
{
"name": "[parameters('storageName')]",
"type": "Microsoft.Storage/storageAccounts",
"apiVersion": "2019-06-01",
"location": "[resourceGroup().location]",
"kind": "StorageV2",
"sku": {
"name": "Premium_LRS",
"tier": "Premium"
}
}
]
}

Learn more
Documentation: What are ARM templates
Learn module: Deploy consistent infrastructure with ARM Templates
Code Samples: ARM templates

Automate deployments with Terraform

Terraform is a declarative framework for deploying and configuring infrastructure that supports many private and
public clouds, Azure being one of them. It has the main advantage of offering a cloud-agnostic framework. While
Terraform configurations are specific to each cloud, the framework itself is the same for all of them. Terraform
configurations are written in a domain-specific language (DSL) called Hashicorp Configuration Language.
The following example demonstrates a simple Terraform configuration that deploys an Azure resource group and
a single Azure Storage account.
 resource "azurerm_resource_group" "example" {
name = "newStorageAccount"
location = "eastus"
}

resource "azurerm_storage_account" "example" {

name = "storageaccountname"
resource_group_name = azurerm_resource_group.example.name
location = azurerm_resource_group.example.location
account_tier = "Standard"
account_replication_type = "GRS"
}

Take note, the Terraform provider for Azure is an abstraction on top of Azure APIs. This abstraction is beneficial
because the API complexities are obfuscated. This abstraction comes at a cost; the Terraform provider for Azure
does not always provide parity with the Azure APIs' capabilities.
Learn more
Documentation: Using Terraform on Azure

Next steps
Automate infrastructure configuration
 Configure infrastructure
12/18/2020 • 6 minutes to read • Edit Online

When working with Azure, many services can be created and configured programmatically using automation or
infrastructure as code tooling. These tools access Azure through the exposed REST APIs or what we refer to as the
Azure control plane. For example, an Azure Network Security Group can be deployed, and security group rules
created using an Azure Resource Manager template. The Network Security Group and its configuration are exposed
through the Azure control plane and natively accessible.
Other configurations, such as installing software on a virtual machine, adding data to a database, or starting pods
in an Azure Kubernetes Service cluster cannot be accessed through the Azure control plane. These actions require a
different set of configuration tools. We consider these configurations as being on the Azure data plane side, or not
exposed through Azure REST APIs. These data plane enables tools to use agents, networking, or other access
methods to provide resource-specific configuration options.
For example, when deploying a set of virtual machined to Azure, you may also want to install and configure a web
server, stage content, and then make the content available on the internet. Furthermore, if the virtual machine
configuration changes and no longer aligns with the configuration definition, you may want a configuration
management system to remediate the configuration. Many options are available for these data plane
configurations. This document details several and provides links for in-depth information.

Bootstrap automation
When deploying to Azure, you may need to run post-deployment virtual machine configuration or run other
arbitrary code to bootstrap the deployed Azure resources. Several options are available for these bootstrapping
tasks and are detailed in the following sections of this document.
Azure VM extensions
Azure virtual machine extensions are small packages that run post-deployment configuration and automation on
Azure virtual machines. Several extensions are available for many different configuration tasks, such as running
scripts, configuring antimalware solutions, and configuring logging solutions. These extensions can be installed and
run on virtual machines using an ARM template, the Azure CLI, Azure PowerShell module, or the Azure portal. Each
Azure VM has a VM Agent installed, and this agent manages the lifecycle of the extension.
A typical VM extension use case would be to use a custom script extension to install software, run commands, and
perform configurations on a virtual machine or virtual machine scale set. The custom script extension uses the
Azure virtual machine agent to download and execute a script. The custom script extensions can be configured to
run as part of infrastructure as code deployments such that the VM is created, and then the script extension is run
on the VM. Extensions can also be run outside of an Azure deployment using the Azure CLI, PowerShell module, or
the Azure portal.
In the following example, the Azure CLI is used to deploy a custom script extension to an existing virtual machine,
which installs a Nginx webserver.

az vm extension set \
--resource-group myResourceGroup \
--vm-name myVM --name customScript \
--publisher Microsoft.Azure.Extensions \
--settings '{"commandToExecute": "apt-get install -y nginx"}'

Learn more
Use the included code sample to deploy a virtual machine and configure a web server on that machine with the
custom script extension.
Documentation: Azure virtual machine extensions
Code Samples: Configure VM with script extension during Azure deployment
cloud-init
cloud-init is a known industry tool for configuring Linux virtual machines on first boot. Much like the Azure custom
script extension, cloud-init allows you to install packages and run commands on Linux virtual machines. cloud-init
can be used for things like software installation, system configurations, and content staging. Azure includes many
cloud-init enable Marketplace virtual machine images across many of the most well-known Linux distributions. For
a full list, see cloud-init support for virtual machines in Azure.
To use cloud-init, create a text file named cloud-init.txt and enter your cloud-init configuration. In this example, the
Nginx package is added to the cloud-init configuration.

#cloud-config
package_upgrade: true
packages:
- nginx

Create a resource group for the virtual machine.

az group create --name myResourceGroupAutomate --location eastus

Create the virtual machine, specifying the --custom-data property with the cloud-inti configuration name.

az vm create \
--resource-group myResourceGroupAutomate \
--name myAutomatedVM \
--image UbuntuLTS \
--admin-username azureuser \
--generate-ssh-keys \
--custom-data cloud-init.txt

On boot, cloud-init will use the systems native package management tool to install Nginx.
Learn more
Documentation: cloud-init support for virtual machines in Azure
Azure deployment script resource
When performing Azure deployments, you may need to run arbitrary code for bootstrapping things like managing
user accounts, Kubernetes pods, or querying data from a non-Azure system. Because none of these operations are
accessible through the Azure control plane, some other mechanism is required for performing this automation. To
run arbitrary code with an Azure deployment, check out the Microsoft.Resources/deploymentScripts Azure
resource.
The deployment script resource behaves similar to any other Azure resource:
Can be used in an ARM template.
Contain ARM template dependencies on other resources.
Consume input, produce output.
Use a user-assigned managed identity for authentication.
When deployed, the deployment script runs PowerShell or Azure CLI commands and scripts. Script execution and
logging can be observed in the Azure portal or with the Azure CLI and PowerShell module. Many options can be
configured like environment variables for the execution environment, timeout options, and what to do with the
resource after a script failure.
The following example shows an ARM template snippet with the deployment script resource configured to run a
PowerShell script.

{
"type": "Microsoft.Resources/deploymentScripts",
"apiVersion": "2019-10-01-preview",
"name": "runPowerShellScript",
"location": "[resourceGroup().location]",
"kind": "AzurePowerShell",
"identity": {
"type": "UserAssigned",
"userAssignedIdentities": {"[parameters('identity')]": {}}
},
"properties": {
"forceUpdateTag": "1",
"azPowerShellVersion": "3.0",
"arguments": "[concat('-sqlServer ', parameters('sqlServer'))]",
"primaryScriptUri": "[variables('script')]",
"timeout": "PT30M",
"cleanupPreference": "OnSuccess",
"retentionInterval": "P1D"
}
}

Learn more
Documentation: Use deployment scripts in templates

Configuration Management
Configuration management tools can be used to configure and manage the ongoing configuration and state of
Azure virtual machines. Three popular options are Azure Automation State Configuration, Chef, and Puppet.
Azure Automation State Configuration
Azure Automation State Configuration is a configuration management solution built on top of PowerShell Desired
State Configuration (DSC). State configuration works with Azure virtual machines, on-premises machines, and
machines in a cloud other than Azure. Using state configuration, you can import PowerShell DSC resources and
assign them to many virtual from a central location. Once each endpoint has evaluated and / or applied the desired
state, state compliance is reported to Azure and can be seen on a built-in dashboard.
The following example uses PowerShell DSC to ensure the NGINX has been installed on Linux systems.

configuration linuxpackage {

Import-DSCResource -Module nx

Node "localhost" {
nxPackage nginx {
Name = "nginx"
Ensure = "Present"
}
}
}

Once imported into Azure State Configuration and assigned to nodes, the state configuration dashboard provides
compliance results.
Learn more
Use the included code sample to deploy Azure Automation State Configuration and several Azure virtual machines.
The virtual machines are also onboarded to state configuration, and a configuration applied.
Documentation: Get started with Azure Automation State Configuration
Code Sample: Deploy DSC and VMs with an ARM template
Chef
Chef is an automation platform that helps define how your infrastructure is configured, deployed, and managed.
Additional components included Chef Habitat for application lifecycle automation rather than the infrastructure
and Chef InSpec that helps automate compliance with security and policy requirements. Chef Clients are installed
on target machines, with one or more central Chef Servers that store and manage the configurations.
Learn more
Documentation: An Overview of Chef
Puppet
Puppet is an enterprise-ready automation platform that handles the application delivery and deployment process.
Agents are installed on target machines to allow Puppet Master to run manifests that define the desired
configuration of the Azure infrastructure and virtual machines. Puppet can integrate with other solutions such as
Jenkins and GitHub for an improved DevOps workflow.
Learn more
Documentation: How Puppet works

Next steps
Automate operational tasks
 Automate operational tasks
12/18/2020 • 7 minutes to read • Edit Online

Operational tasks can include any action or activity you may perform while managing systems, system access, and
processes. Some examples include rebooting servers, creating accounts, and shipping logs to a data store. These
tasks may occur on a schedule, as a response to an event or monitoring alert, or ad-hock based on external factors.
Like many other activities related to managing computer systems, these activities are often performed manually,
which takes time, and are error-prone.
Many of these operational tasks can and should be automated. Using scripting technologies and related solutions,
you can shift effort from manually performing operational tasks towards building automation for these tasks. In
doing so, you achieve so much, including:
Reduce time to perform an action.
Reduce risk in performing the action.
Automated response to events and alerts.
Increased human capacity for further innovation.
When working in Azure, you have many options for automating operational tasks. This document details some of
the more popular.

Azure Functions
Azure Functions allows you to run code without managing the underlying infrastructure on where the code is run.
Functions provide a cost-effective, scalable, and event-driven platform for building applications and running
operational tasks. Functions support running code written in C#, Java, JavaScript, Python, and PowerShell.
When creating a Function, a hosting plan is selected. Hosting plans controls how a function app is scaled, resource
availability, and availability of advanced features such as virtual network connectivity and startup time. The hosting
plan also influences the cost.
Functions hosting plans:
Consumption - Default hosting plan, pay only for Function execution time, configurable timeout period,
automatic scale.
Premium - Faster start, VNet connectivity, unlimited execution duration, premium instance sizes, more
predictable pricing.
App Ser vice Plan - Functions run on dedicated virtual machines and can use custom images.
For full details on consumption plans, see Azure Functions scale and hosting.
Functions provide event-driven automation; each function has a trigger associated with it. These triggers are what
run the functions. Common triggers include:
HTTP / Webhook - Function is run when an HTTP request is received.
Queue - Function is run when a new message arrives in a message queue.
Blob storage - Function is run when a new blob is created in a storage container.
Timer - Function is run on a schedule.
Below are example triggers seen in the Azure portal when creating a new function
Once an event has occurred that initiates a Function, you may also want to consume data from this event or from
another source. Once a Function has been completed, you may want to publish or push data to an Azure service
such as a Blob Storage. Input and output are achieved using input and output bindings. For more information
about triggers and bindings, see Azure Functions triggers and binding concepts.
Both PowerShell and Python are common languages for automating everyday operational tasks. Because Azure
Functions supports both of these languages, it is an excellent platform for hosting, running, and auditing
automated operational tasks. For example, let's assume that you would like to build a solution to facilitate self-
service account creation. An Azure PowerShell Function could be used to create the account in Azure Active
Directory. An HTTP trigger can be used to initiate the Function, and an input binding configured to pull the account
details from the HTTP request body. The only remaining piece would be a solution that consumes the account
details and creates the HTTP requests against the Function.
Learn more
Documentation: Azure Functions PowerShell developer guide
Documentation: Azure Functions Python developer guide

Azure Automation
PowerShell and Python are popular programming languages for automating operational tasks. Using these
languages, performing operations like restarting services, moving logs between data stores, and scaling
infrastructure to meet demand can be expressed in code and executed on demand. Alone, these languages do not
offer a platform for centralized management, version control, or execution history. The languages also lack a native
mechanism for responding to events like monitoring driven alerts. To provide these capabilities, an automation
platform is needed.
Azure Automation provides an Azure-hosted platform for hosting and running PowerShell and Python code across
Azure, non-Azure cloud, and on-premises environments. PowerShell and Python code is stored in an Azure
Automation Runbook, which has the following attributes:
Execute Runbooks on demand, on a schedule, or through a webhook.
Execution history and logging.
Integrated secrets store.
 Source Control integration.
As seen in the following image, Azure Automation provides a portal experience for managing Azure Automation
Runbooks. Use the included code sample (ARM template) to deploy an Azure automation account, automation
runbook, and explore Azure Automation for yourself.

Learn more
Documentation: What is Azure Automation?

Scale operations
So far, this document has detailed options for scripting operational tasks; however, many Azure services come with
built-in automation, particularly in scale operations. As application demand increases (transactions, memory
consumption, and compute availability), you may need to scale the application hosting platform so that requests
continue to be served. As demand decreases, scaling back not only appropriately resizes your application but also
reduces operational cost.
In cloud computing, scale activities are classified into two buckets:
Scale-up - Adding additional resources to an existing system to meet demand.
Scale-out - Adding additional infrastructure to meet demand.
Many Azure services can be scaled up by changing the pricing tier of that service. Generally, this operation would
need to be performed manually or using detection logic and custom automation.
Some Azure services support automatic scale-out, which is the focus of this section.
Azure Monitor autoscale
Azure Monitor autoscale can be used to autoscale Virtual Machine Scale Sets, Cloud Services, App Service Web
Apps, and API Management service. To configure scale-out operations for these services, while in the Azure portal,
select the service, and then scale-out under the resource settings. Select Custom To configure autoscaling rules.
Automatic scale operations can also be configured using an Azure Resource Manager Template, the Azure
PowerShell module, and the Azure CLI.
When creating the autoscale rules, configure minimum and maximum instance counts. These settings prevent
inadvertent costly scale operations. Next, configure autoscale rules, at minimum one to add more instances, and
one to remove instances when no longer needed. Azure Monitor autoscale rules give you fine-grain control over
when a scale operation is initiated. See the Learn more section below for more information on configuring these
rules.

Learn more
Documentation: Azure Monitor autoscale overview
Azure Kubernetes Service
Azure Kubernetes Service (AKS) offers an Azure managed Kubernetes cluster. When considering scale operations in
Kubernetes, there are two components:
Pod scaling - Increase or decrease the number of load balanced pods to meet application demand.
Node scaling - Increase or decrease the number of cluster nodes to meet cluster demand.
Azure Kubernetes Service includes automation to facilitate both of these scale operation types.
Horizontal pod autoscaler
Horizontal pod autoscaler (HPA) monitors resource demand and automatically scales pod replicas. When
configuring horizontal pod autoscaling, you provide the minimum and maximum pod replicas that a cluster can
run and the metrics and thresholds that initiate the scale operation. To use horizontal pod autoscaling, each pod
must be configured with resource requests and limits, and a HorizontalPodAutoscaler Kubernetes object must
be created.
The following Kubernetes manifest demonstrates resource requests on a Kubernetes pod and also the definition of
a horizontal pod autoscaler object.

apiVersion: apps/v1
kind: Deployment
metadata:
name: azure-vote-back
spec:
replicas: 1
selector:
matchLabels:
app: azure-vote-back
template:
metadata:
labels:
app: azure-vote-back
spec:
nodeSelector:
"beta.kubernetes.io/os": linux
containers:
- name: azure-vote-back
image: redis
# Resource requests and limits
resources:
requests:
cpu: 100m
memory: 128Mi
limits:
cpu: 250m
memory: 256Mi
ports:
- containerPort: 6379
name: redis
---
apiVersion: autoscaling/v1
kind: HorizontalPodAutoscaler
metadata:
name: azure-vote-back-hpa
spec:
maxReplicas: 10 # define max replica count
minReplicas: 3 # define min replica count
scaleTargetRef:
apiVersion: apps/v1
kind: Deployment
name: azure-vote-back
targetCPUUtilizationPercentage: 50 # target CPU utilization

Cluster autoscaler
Where horizontal pod autoscaling is a response to demand on a specific application of service running in a
Kubernetes cluster, cluster autoscaling responds to demand on the entire Kubernetes cluster itself. If a Kubernetes
cluster does not have enough compute resources or nodes to facilitate all requested pods' resource requests, some
of these pods will enter a non-scheduled or pending state. In response to this situation, additional nodes can be
automatically added to the cluster. Conversely, once compute resources have been freed up, the cluster nodes can
automatically be removed to match steady-state demand.
Cluster autoscaler can be configured when creating an AKS cluster. The following example demonstrates this
operation with the Azure CLI. This operation can also be completed with an Azure Resource Manager template.
 az aks create \
--resource-group myResourceGroup \
--name myAKSCluster \
--node-count 1 \
--vm-set-type VirtualMachineScaleSets \
--load-balancer-sku standard \
--enable-cluster-autoscaler \
--min-count 1 \
--max-count 3

Cluster autoscaler can also be configured on an existing cluster using the following Azure CLI command.

az aks update \
--resource-group myResourceGroup \
--name myAKSCluster \
--enable-cluster-autoscaler \
--min-count 1 \
--max-count 3

See the included documentation for information on fine-grain control for cluster autoscale operations.
Learn more
Documentation: Horizontal pod autoscaling
Documentation: AKS cluster autoscaler
 Release Engineering: Application Development
12/18/2020 • 5 minutes to read • Edit Online

One of the primary goals of adopting modern release management strategies is to build systems that allow your
teams to turn ideas into production delivered software with as little friction as possible. Throughout this section of
the Well-Architected Framework, methods and tools for quickly and reliably delivering software are examined. You
will learn about things like continuous deployment software, integration strategies, and deployment environments.
Samples are provided to help you quickly get hands-on with this technology.
However, release engineering does not start with fancy deployment software, multiple deployment environments,
or Kubernetes clusters. Before examining how we can quickly and reliably release software, we need to first look at
how software is developed. Not only has the introduction of cloud computing had a significant impact on how
software is delivered and run, but it's also had a huge downstream impact on how software is developed. For
example, the introduction of container technology has changed how we can host, scale, and deprecate software.
That said, containers have also impacted things like dependency management, host environment, and tooling as we
develop software.
This article details many practices that you may want to consider when building strategies for developing for the
cloud. Topics include:
Development environments, or where you write your code.
Source control and branching strategies, how you manage, collaborate on, and eventually deploy your code.

Development environments
When developing software for the cloud, or any environment, care needs to be taken to ensure that the
development environment is set up for success. When setting up a development environment, you may consider
questions like the following:
How do I ensure that all dependencies are in place?
How can I best configure my development environment to emulate a production environment?
How do I develop code where service dependencies may exist with code already in production?
The following sections briefly detail technology that aids during the local or what is often referred to as "inner-loop"
development process.
Docker Desktop
Docker Desktop is an application that provides a Docker environment on your development system. Docker
Desktop includes not only the Docker runtime but application development tools and local Kubernetes
environment. Using Docker Desktop, you can develop in any language, create and test container images, and when
ready, push application ready container images to a container registry for production use.
Learn more
Docker Desktop
Windows Subsystem for Linux
Many applications and solutions are built on Linux. Windows Subsystem for Linux provides a Linux environment on
your Windows machines, including many command-line tools, utilities, and Linux applications. Multiple GNU/Linux
distributions are available and can be found in the Microsoft Store.
Learn more
Documentation: Windows Subsystem for Linux Documentation
Bridge to Kubernetes
Bridge to Kubernetes allows you to run and debug code on your development system while connected to a
Kubernetes cluster. This configuration can be helpful when working on microservice type architectures. Using
Bridge, you can work locally on one service, which has a dependency on other services. Rather than needing to run
all dependant services on your development system or deploy your in-development code to the cluster, Bridge
manages the communication between your development system and running services in your Kubernetes cluster.
Essentially, the code running on your development system behaves as if it's running in the Kubernetes cluster.
Some features of Bridge:
Works with Azure Kubernetes Service (AKS) and non-AKS clusters (in preview).
Redirects traffic between your Kubernetes cluster and code running on your development system.
Support for isolated traffic routing, which is important in shared clusters.
Replicates environment variables and mounted volumes from your cluster to your development system.
Cluster diagnostic logs are made available on your development system.
Learn more
Documentation: Use Bridge to Kubernetes with Visual Studio Code
Documentation: Use Bridge to Kubernetes with Visual Studio

Source control
Source control management (SCM) systems provide a way to control, collaborate, and peer review software
changes. As software is merged into source control, the system helps manage code conflicts. Ultimately, source
control provides a running history of the software, modification, and contributors. Whether a piece of software is
open-sourced or private, using source control software has become a standardized method of managing software
development. As detailed in later sections of the Well-Architected Framework, source control systems can also be
enlightened with integrated testing, security, and release practices. As cloud practices are adopted and because so
much of the cloud infrastructure is managed through code, version control systems are also becoming an integral
part of infrastructure management.
Many source control systems are powered by Git. Git is a distributed version control system with related tools that
allow you and your team to track source code changes during the software development lifecycle. Using Git, you
can create a copy of the software, make changes, propose the changes, and receive peer review on your proposal.
During peer review, Git makes it easy to see precisely the changes being proposed. Once the proposed changes
have been approved, Git helps merge the changes into the source, including conflict resolution. If, at any point, the
changes need to be reverted, Git can also manage rollback.
Let's examine a few aspects of version controlling software and infrastructure configurations.
Version Control and code changes
Beyond providing us with a place to store code, source control systems allow us to understand what version of the
software is current and identify changes between the present and past versions. Version control solutions should
also provide a method for reverting to the previous version when needed.
The following image demonstrates how Git and GitHub are used to see the proposed code changes.
Forking and Pull Requests
Using source control systems, you can create your own copies of the software source called forks. With your fork in
place, changes can be made to the software without the risk that the changes will impact the software's current
version. Once you are happy with the changes made in your fork, you can suggest that your changes be merged
into the main source control through what is known as a pull request.
Peer Review
As updates are made to software and infrastructure configurations, version control software allows us to propose
these changes before merging them into the source. During the proposal, peers can review the changes,
recommend updates, and approve the changes. Source control solutions provide an excellent platform for
collaboration on changes to the software.
To learn more about using Git, visit the DevOps Resource Center.
GitHub
GitHub is a popular source control system that uses Git. In addition to core Git functionality, GitHub includes several
other features such as access control, collaboration features such as code issues, project boards, wikis, and an
automation platform called GitHub actions.
Learn more
GitHub
Azure Repos
Azure DevOps is a collection of services for building, collaborating on, testing, and delivering software to any
environment. Azure DevOps services include Azure Repos, which is a source control system. Using Azure Repos
includes unlimited free private Git repositories. Standard Git powers Azure Repos, and you can use clients and tools
of your choice for working with them.
Learn more
Documentation: Azure Repos

Next steps
Release Engineering: Continuous integration
 Release Engineering: Continuous integration
12/18/2020 • 4 minutes to read • Edit Online

As code is developed, updated, or even removed, having a friction-free and safe method to integrate these changes
into the main code branch is paramount to enabling developers to provide value fast. As a developer, making small
code changes, pushing these to a code repository, and getting almost instantaneous feedback on the quality, test
coverage, and introduced bugs allows you to work faster, with more confidence, and less risk. Continuous
integration (CI) is a practice where source control systems and software deployment pipelines are integrated to
provide automated build, test, and feedback mechanisms for software development teams.
Continuous integration is about ensuring that software is ready for deployment but does not include the
deployment itself. This article covers the basics of continuous integration and offers links and examples for more
in-depth content.

Continious integration
Continuous integration is a software development practice under which developers integrate software updates into
a source control system on a regular cadence. The continuous integration process starts when an engineer creates
a pull request signaling to the CI system that code changes are ready to be integrated. Ideally, integration validates
the code against several baselines and tests and provides quick feedback to the requesting engineer on the status
of these tests. Assuming baseline checks and testing have gone well, the integration process produces and stages
assets such as compiled code and container images that will eventually deploy the updated software.
As a software engineer, continuous integration can help you deliver quality software more quickly by performing
the following:
Run automated tests against the code, providing early detection of breaking changes.
Run code analysis to ensure code standards, quality, and configuration.
Run compliance and security checks ensuring no known vulnerabilities.
Run acceptance or functional tested to ensure that the software operates as expected.
To provide quick feedback on detected issues.
Where applicable, produce deployable assets or packages that include the updated code.
To achieve continuous integration, use software solutions to manage, integrate, and automate the process. A
common practice is to use a continuous integration pipeline, detailed in this article's next section.

Continuous integration pipelines

A continuous integration pipeline involves a piece of software, in many cases, cloud-hosted, that provides a
platform for running automated tests, compliance scans, reporting, and all additional components that make up
the continuous integration process. In most cases, the pipeline software is attached to source control such that
when pull requests are created or software is merged into a specific branch, and the continuous integration
pipeline is run. Source control integration also provides the opportunity for providing CI feedback directly on the
pull request.
Many solutions provide continuous integration pipeline capabilities. This article touches on both Azure DevOps
Pipelines and GitHub actions and provides links to find more information.
Learn more
To learn how to create a continuous integration pipeline, see these articles:
 Azure DevOps: Create your first pipeline
GitHub Actions: Setting up continuous integration using workflow tempaltes
Source Control Integration
The integration of your continuous integration pipeline with your source control system is key to enabling fast, self-
service, and friction-free code contributions. Source control integration enables these things:
As pull requests are created, the CI pipeline is run, including all tests, security assessments, and other checks.
CI test results are provided to the pull request initiator directly in the pull request, allowing for almost real-time
feedback on quality.
Another popular practice is building small reports or badges that can be presented in source control to make
visible the current builds states
The following image shows the integration between GitHub and an Azure DevOps pipeline. In this example, a pull
request has been created, which in turn has triggered an Azure DevOps pipeline. The pipeline status can be seen
directly in the pull request.

Test Integration
A key element of continuous integration is the continual building and testing of code as code contributions are
created. Testing pull requests as they are created gives quick feedback that the commit has not introduced breaking
changes. The advantage is that the tests that are run by the continuous integration pipeline can be the same tests
run during test-driven development.
The following code snippet shows a test step from an Azure DevOps pipeline. There are two actions occurring:
The first task is using a popular Python testing framework to run CI tests. These tests reside in source control
alongside the Python code. The test results are output to a file named test-results.xml.
The second task consumes the test results and publishing them to the Azure DevOps pipeline as an integrated
report.

- script: |
pip3 install pytest
pytest azure-vote/azure-vote/tests/ --junitxml=junit/test-results.xml
continueOnError: true

- task: PublishTestResults@2
displayName: 'Publish Test Results'
inputs:
testResultsFormat: 'JUnit'
testResultsFiles: '**/test-results.xml'
failTaskOnFailedTests: true
testRunTitle: 'Python $(python.version)'

The following image shows the test results as seen in the Azure DevOps portal.
CI Result Badges
Many developers like to show that they're keeping their code quality high by displaying a status badge in their
repo. The following image shows an Azure Pipelines badge as displayed on the Readme file for an open-source
project on GitHub.

Learn more
To learn how to display badges in your repositories, see these articles:
Add an Azure Pipeline status badge to your repository.
Add a GitHub workflow status badge to your repository.

Next steps
Release Engineering: Release testing
 Testing your application and Azure environment
12/18/2020 • 6 minutes to read • Edit Online

Testing is one of the fundamental components and DevOps and agile development in general. If automation gives
DevOps the required speed and agility to deploy software quickly, only through extensive testing those
deployments will achieve the required reliability that customers demand.
A main tenet of a DevOps practice to achieve system reliability is the "Shift Left" principle. If developing and
deploying an application is a process depicted as a series of steps going from left to right, testing should not be
limited to being performed at the very end of the process (at the right). It should be shifted as much to the
beginning (to the left) as possible. Errors are cheaper to repair when caught early. They can be expensive or
impossible to fix later in the application life cycle.
Another aspect to consider is that testing should occur on both application code as well as infrastructure code and
they should be subject to the same quality controls. As described in [Infrastructure as Code][iac], the environment
where applications are running should be version-controlled and deployed through the same mechanisms as
application code, and hence can be tested and validated using DevOps testing paradigms too.
You can use your favorite testing tool to run your tests, including Azure Pipelines for automated testing and Azure
Testing Plans for manual testing.
There are multiple stages at which tests can be performed in the life cycle of code, and each of them has some
particularities that is important to understand. In this guide, you can find a summary of the different tests that you
should consider while developing and deploying applications.

Automated Testing
Automating tests is the best way to make sure that they are executed. Depending on how frequently tests are
performed, they are typically limited in duration and scope, as the different types of automated tests will show:
Unit Testing
Unit tests are tests typically run by each new version of code committed into version control. Unit Tests should be
extensive (should cover ideally 100% of the code) and quick (typically under 30 seconds, although this number is
not a rule set in stone). Unit testing could verify things like the syntax correctness of application code, Resource
Manager templates or Terraform configurations, that the code is following best practices, or that they produce the
expected results when provided certain inputs.
Unit tests should be applied both to application code and infrastructure code.
Smoke Testing
Smoke tests are more exhaustive than unit tests, but still not as much as integration tests. They normally run in less
than 15 minutes. Still not verifying the interoperability of the different components with each other, smoke tests
verify that each of them can be correctly built and offers the expected functionality and performance.
Smoke tests usually involve building the application code, and if infrastructure, possibly testing the deployment in
a test environment.
Integration Testing
After making sure that the different application components operate correctly individually, integration testing has
as goal determine whether they can interact with each other as they should. Integration tests usually take longer
than smoke testing, and as a consequence they are sometimes executed not as frequently. For example, running
integration tests every night still offers a good compromise, detecting interoperability issues between application
components no later than one day after they were introduced.

Application Manual Testing

Manual testing is much more expensive than automated testing, and as a consequence it is run much less
frequently. However, manual testing is fundamental for the correct functioning of the DevOps feedback loop, to
correct errors before they become too expensive to repair, or cause customer dissatisfaction.
Acceptance Testing
There are many different ways of confirming that the application is doing what it should.
Blue/Green deployments : when deploying a new application version, you can deploy it in parallel to the
existing one. This way you can start redirecting clients to the new version, and if everything goes well you will
decommission the old version. If there is any problem with the new deployment, you can always redirect the
users back to the old one.
Canar y releases : you can expose new functionality of your application (ideally using feature flags) to a select
group of users. If users are satisfied with the new functionality, you can extend it to the rest of the user
community. In this case we are talking about releasing functionality, and not necessarily about deploying a new
version of the application.
A/B testing : A/B testing is similar to canary release-testing, but while canary releases focus on mitigate risk,
A/B testing focus on evaluating the effectiveness of two similar ways of achieving different goals. For example, if
you have two versions of the layout of a certain area of your application, you could send half of your users to
one, the other half to the other, and use some metrics to see which layout works better for the application goals.
An important aspect to consider is how to measure the effectiveness of new features in the application. A way to do
that is through Application Insights User Behavior Analytic, with which you can determine how people are using
your application. This way you can decide whether a new feature has improved your applications without bringing
effects such as decreasing usability.
Certain services in Azure offer functionality that can help with this kind of tests, such as the slot functionality in the
Azure App Service, that allows having two different versions of the same application running at the same time, and
redirect part of the users to one or the other.
Stress tests
As other sections of this framework have explained, designing your application code and infrastructure for
scalability is of paramount importance. Testing that you can increase the application load and that both the code
and the infrastructure will react to it is critical, so that your environment will adapt to changing load conditions.
During these stress tests, it is critical monitoring all the components of the system to identify whether there is any
bottleneck. Every component of the system not able to scale out can turn into a scale limitation, such as
active/passive network components or databases. It is hence important knowing their limits so that you can
mitigate their impact into the application scale. This exercise might drive you to changing some of those
components for more scalable counterparts.
It is equally important verifying that after the stress test is concluded, the infrastructure scales back down to its
normal condition in order to keep costs under control.
Business Continuity Drills
Certain infrastructure test scenarios can be considered under the category of acceptance testing, such as Business
continuity drills. In particular Disaster Recovery scenarios are difficult to test on-premises, but the public cloud
makes this kind of tests easier. Tools such as Azure Site Recovery make it possible to start an isolated copy of the
primary location in a secondary environment, so that it can be verified that the applications have come up as they
should.
In case there is any problem, the Disaster Recovery procedure can be optimized, and the infrastructure in the
secondary environment can be deleted.
Exploratory testing
Experts explore the application in its entirety trying to find faults or suboptimal implementations of functionality.
These experts could be developers, UX specialists, product owners, actual users, and other profiles. Test plans are
typically not used, since testing is left to the ability of the individual tester.
Fault injection
The same concept can be applied to the infrastructure. If the application should be resilient to infrastructure
failures, introducing faults in the underlying infrastructure and observing how the application behaves is
fundamental for increasing the trust in your redundancy mechanisms. Shutting down infrastructure components,
pourposley degrading performance, or introducing faults are ways of verifying that the application is going to react
as expected when these situations occur.
Most companies use a controlled way of injecting faults in the system, although if confident with the application
resiliency, automated frameworks could be used. A new science has been developed around fault injection, called
Chaos Engineering.

Summary
In order to deploy software quickly and reliably, testing is a fundamental component of the development and
deployment life cycle. Not only application code should be tested, but infrastructure automation and resiliency
should equally be put to the test, to make sure that the application is going to perform as expected in every
situation.

Next steps
Release Engineering: Performance
 Performance considerations for your deployment
infrastructure
12/18/2020 • 3 minutes to read • Edit Online

Build status shows if your product is in a deployable state, so builds are the heartbeat of your continuous delivery
system. It's important to have a build process up and running from the first day of your product development.
Since builds provide such a crucial information about the status of your product, you should always strive for fast
builds.
It will be hard to fix build problem if it takes longer to build, and the team will suffer from a broken window
disorder. Eventually, nobody cares if the build is broken since it's always broken and it takes lot of effort to fix it.

Build times
Here are few ways you can achieve faster builds.
Selecting right size VMs: Speeding up your builds starts with selecting the right size VMs. Fast machines
can make the difference between hours and minutes. If your pipelines are in Azure Pipelines, then you've got
a convenient option to run your jobs using a Microsoft-hosted agent. With Microsoft-hosted agents,
maintenance and upgrades are taken care of for you. For more info see, Microsoft-hosted agents.
Build ser ver location: When you're building your code, a lot of bits are moved across the wire, so inputs
to the builds are fetched from a source control repository and the artifact repository, such as source code,
NuGet packages, etc. At the end, the output from the build process needs to be copied, not only the
compiled artifacts, but also test reports, code coverage results, and debug symbols. So it is important that
these copy actions are fast. If you are using your own build server, ensure that the build server is located
near the sources and a target location, and it can reduce the duration of your build considerably.
Scaling out build ser vers: A single build server may be sufficient for a small product, but as the size and
the scope of the product and the number of teams working on the product increases, the single server may
not be enough. Scale your infrastructure horizontally over multiple machines when you reach the limit. For
more info see, how you can leverage Azure DevOps Agent Pools.
Optimizing the build:
Add parallel build execution so we can speed up the build process. For more info see, Azure DevOps
parallel jobs.
Enable parallel execution of test suites, which is often a huge time saver, especially when executing
integration and UI tests. For more info see, Run tests in parallel using Azure Pipeline.
Use the notion of a multiplier, where you can scale out your builds over multiple build agents. For
more info see, Organizing Azure Pipeline into Jobs.
Move a part of the test feedback loop to the release pipeline. This improves the build speed, and
hence the speed of the build feedback loop.
Publish the build artifacts to the package management system solution, and hence publish to a
NuGet feed at the end of a build.

Human intervention
It's important to select different builds for different purpose.
CI builds: Purpose of this build is to ensure it compiles and unit tests run. This build gets triggered at each
commit or set of commits over a period of time. It serves as the heartbeat of the project, provides quality
feedback to the team immediately. For more info see, CI triggers or Batching CI builds.
Nightly build: Purpose of this build is not only to compile but also ensure necessary
integration/regression tests are run. This build can take up some more time, because we need to do some
extra steps to get additional information about the product. For example, metrics about the state of the
software using SonarQube. It may also contain a set of regression tests and integration tests and it may also
deploy the solution to a temporary machine to verify the solution is continuing to work. For more info see,
scheduling builds using cron syntax
Release build: Besides compiling, running test this build additionally compiles the API documentation,
compliance reports, code signing, and other steps which are not required every time the code is built. Finally
this build provide the golden copy that will be pushed to the release pipeline to finally deploy in the
production environment. Generally release build is gets triggered manually instead of a CI trigger.

Next steps
Release Engineering: Deployment
 Deployment considerations for DevOps
12/18/2020 • 4 minutes to read • Edit Online

As you provision and update Azure resources, application code, and configuration settings, a repeatable and
predictable process will help you avoid errors and downtime. We recommend automated processes for
deployment that you can run on demand and rerun if something fails. After your deployment processes are
running smoothly, process documentation can keep them that way.

Automation
To activate resources on demand, deploy solutions rapidly, minimize human error, and produce consistent and
repeatable results, be sure to automate deployments and updates.
Automate as many processes as possible
The most reliable deployment processes are automated and idempotent — that is, repeatable to produce the same
results.
To automate provisioning of Azure resources, you can use Terraform, Ansible, Chef, Puppet, Azure PowerShell,
Azure CLI, or Azure Resource Manager templates.
To configure VMs, you can use cloud-init (for Linux VMs) or Azure Automation State Configuration (DSC).
To automate application deployment, you can use Azure DevOps Services, Jenkins, or other CI/CD solutions.
As a best practice, create a repository of categorized automation scripts for quick access, documented with
explanations of parameters and examples of script use. Keep this documentation in sync with your Azure
deployments, and designate a primary person to manage the repository.
Automation scripts can also activate resources on demand for disaster recovery.
Automate and test deployment and maintenance tasks
Distributed applications consist of multiple parts that must work together. Deployment should take advantage of
proven mechanisms, such as scripts, that can update and validate configuration and automate the deployment
process. Test all processes fully to ensure that errors don't cause additional downtime.
Implement deployment security measures
All deployment tools must incorporate security restrictions to protect the deployed application. Define and enforce
deployment policies carefully, and minimize the need for human intervention.

Release process
One of the challenges with automating deployment is the cut-over itself, taking software from the final stage of
testing to live production. You usually need to do this quickly in order to minimize downtime. The blue-green
deployment approach does this by ensuring you have two production environments, as identical as possible.

Document release process

Without detailed release process documentation, an operator might deploy a bad update or might improperly
configure settings for your application. Clearly define and document your release process, and ensure that it's
available to the entire operations team.

Stage your workloads

Deployment to various stages and running tests/validations at each stage before moving on to the next ensures
friction free production deployment.
With good use of staging and production environments, you can push updates to the production environment in a
highly controlled way and minimize disruption from unanticipated deployment issues.
Blue-green deployment involves deploying an update into a production environment that's separate from the
live application. After you validate the deployment, switch the traffic routing to the updated version. One way to
do this is to use the staging slots available in Azure App Service to stage a deployment before moving it to
production.
Canary releases are similar to blue-green deployments. Instead of switching all traffic to the updated
application, you route only a small portion of the traffic to the new deployment. If there's a problem, revert to
the old deployment. If not, gradually route more traffic to the new version. If you're using Azure App Service,
you can use the Testing in production feature to manage a canary release.

Logging and auditing

To capture as much version-specific information as possible, implement a robust logging strategy. If you use staged
deployment techniques, more than one version of your application will be running in production. If a problem
occurs, determine which version is causing it.

High availability considerations

An application that depends on a single instance of a service creates a single point of failure. To improve resiliency
and scalability, provision multiple instances.
For Azure App Service, select an App Service plan that offers multiple instances.
For Azure Cloud Services, configure each of your roles to use multiple instances.
For Azure Virtual Machines, ensure that your architecture includes more than one VM and that each VM is
included in an availability set.
Consider deploying across multiple regions
We recommend deploying all but the least critical applications and application services across multiple regions. If
your application is deployed to a single region, in the rare event that the entire region becomes unavailable, the
application will also be unavailable. If you choose to deploy to a single region, consider preparing to redeploy to a
secondary region as a response to an unexpected failure.
Redeploy to a secondary region
If you run applications and databases in a single, primary region with no replication, your recovery strategy might
be to redeploy to another region. This solution is affordable but most appropriate for non-critical applications that
can tolerate longer recovery times. If you choose this strategy, automate the redeployment process as much as
possible and include redeployment scenarios in your disaster response testing.
To automate your redeployment process, consider using Azure Site Recovery.

Next steps
Release Engineering: Rollback
 Release Engineering: Rollback
12/18/2020 • 2 minutes to read • Edit Online

In some cases, a new software deployment can harm or degrade the functionality of a software system. When
building your solutions, it is essential to anticipate deployment issues and to architect solutions that provide
mechanisms for fixing problematic deployments. Rolling back a deployment involves reverting the deployment to
a known good state. Rollback can be accomplished in many different ways. Several Azure services support native
mechanisms for rolling back to a previous state. Some of these services are detailed in this article.

Azure App Service

When deploying an application to Azure App Service, consider utilizing deployment slots. Deployment slots are
running instances of the application, each with a separate host name. Slots can be used to stage and test
applications before promoting to a production slot. A deployment slot can be created to hold the last known good
instance of your application. In the event of an issue or problematic deployment, the production slot can be
swapped with the known good slot to bring the application back to a known good state.

Learn more
For more information on using Azure App Service deployment slots, see Set up staging environments in Azure App
Service

Azure Kubernetes Service (AKS)

A Kubernetes deployment defines the desired state for a particular workload running in the cluster. For example, a
deployment may declare that a workload consists of 3 replicas of a specific pod that should be running at all times.
The deployment object creates a ReplicaSet and the associate Pods. When updating a workload, the deployment
itself can be revised, which will generally roll out a new container image to the deployment pods. Assuming
multiple replicas of the pods exist, this rollout can happen in a controlled and staged manner such that no
downtime occurs.
If a deployment introduces breaking changes or unintentional results, it can be reverted to an earlier state.
In this case, a deployment named demorollback contains three replicas of a pod.
 $ kubectl get deployments

NAME READY UP-TO-DATE AVAILABLE AGE

demorollback 3/3 3 3 28s

As the deployment is updated, deployment history is retained.

$ kubectl rollout history deployment demorollback

deployment.apps/demorollback
REVISION CHANGE-CAUSE
1 <none>
2 <none>

If an updated deployment has introduced issues, the kubectl rollout undo command can be used to revert to a
previous deployment revision.

$ kubectl rollout undo deploy demorollback --to-revision=1

deployment.apps/demorollback rolled back

Volumes: <none>

Learn more
For more information, see Kubernetes Deployments

Azure Resource Manager (ARM) deployments

A deployment record is created when deploying Azure infrastructure and solutions with Azure Resource Manager
(ARM) templates. When creating a new deployment, you can provide a previously known good deployment so that
if the current deployment fails, the previous know good deployment is redeployed. There are several
considerations and caveats when using this functionality. See the documentation linked below for these details.

Learn more
For more information, see Rollback on an error to successful deployment

Logic apps
When making changes to an Azure logic application, a new version of the application is created. Azure maintains a
history of versions and can revert or promote any previous version. To do so, in the Azure portal, select your logic
app > Versions . Previous versions can be selected on the versions pane, and the application can be inspected both
in the code view and the visual designer view. Select the version you would like to revert to, and click the Promote
option and then Save .
Learn more
For more information, see Manage logic apps in the Azure portal
 Monitoring for DevOps
11/2/2020 • 7 minutes to read • Edit Online

What you cannot see, you cannot measure. What you cannot measure, you cannot improve. This classic
management axiom is true in the cloud as well. Traditional application and infrastructure monitoring is based on
whether the application is running or not, or what response time it is giving. However, cloud-based monitoring
offer many more opportunities that you should be leveraging in order to give your users the best experience.

Application Monitoring
Application Insights is the Azure Service that allows not only to verify that your application is running correctly, but
it makes application troubleshooting easier, and can be used for custom business telemetry that will tell you
whether your application is being used as intended.
Make sure you leverage all the rich information that Application Insights can provide about your application. This
list is not exhaustive, but here you can find some of the visibility that Application Insights can give you:
Application Insights offers you a default dashboard with an educated guess of the most important metrics you
will be interested in. You can then modify it and customize it to your own needs.
By instrumenting your application correctly, Application Insights will give you performance statistics both from
a client and a server perspective
The Application Map will show you application dependencies in other services such as backend APIs or
databases, allowing to determine visually where performance problems lie
Smart Detection will warn you when anomalies in performance or utilization patterns happen
Usage Analysis can give you telemetry on which features of your application are most frequently used, or
whether all your application functionality is being used. This feature is especially useful after changes to the
application functionality, to verify whether those changes were successful
Release annotations are visual indicators in your Application Insights charts of new builds and other events, so
that you can visually correlate changes in application performance to code releases, being able to quickly
pinpoint performance problems.
Cross-component transaction diagnostics allow you to follow failed transactions to find the point in the
architecture where the fault was originated.
Snapshot Debugger, to automatically collect a snapshot of a live application in case of an exception, to analyze it
at a later stage.
To use Application Insights you have two options: you can use codeless monitoring , where onboarding your app
to Application Insights does not require any code change, or code-based monitoring , where you instrument
your code to send telemetry to Application Insights using the Software Development Kit for your programming
language of choice.
You can certainly use other Application Performance Management tools to monitor your application on Azure, such
as NewRelic or AppDynamics, but Application Insights will give you the most seamless and integrated experience.

Platform Monitoring
Application Insights is actually one of the components of Azure Monitor, which gives you rich metrics and logs to
verify the state of your complete Azure landscape. No matter whether your application is running on Virtual
Machines, App Services, or Kubernetes, Azure Monitor will help you to follow the state of your infrastructure, and
to react promptly if there are any issues.
Make sure not only to monitor your compute elements supporting your application code, but your data platform
as well: databases, storage accounts, or data lakes should be closely monitored, since a low performance of the
data tier of an application could have serious consequences.
Container Insights
Should your application run on Azure Kubernetes Service, Azure Monitor allows you to easily monitor the state of
your cluster, nodes, and pods. Easy to configure for AKS clusters, Container Insights delivers quick, visual, and
actionable information: from the CPU and memory pressure of your nodes to the logs of individual Kubernetes
pods.
Additionally, for operators that prefer using the open-source Kubernetes monitoring tool Prometheus but still like
the ease of use of Azure Monitor Container Insights, both solutions can integrate with each other.

The Sidecar Pattern adds a separate container with responsibilities that are required by the main container. A
common use case is for running logging utilities and monitoring agents.
Network monitoring
Regardless the form factor or programming language your application is based on, the network connecting your
code to your users can make or break the experience that your application provides. As a consequence monitoring
and troubleshooting the network can be decisive for an operations team. The component of Azure Monitor that
manages the network components is called Network Watcher, a collection of network monitoring and
troubleshooting tools. Some of these tools are:
Traffic Analytics will give you an overview of the traffic in your Virtual Networks, as well as the percentage
coming from malicious IP addresses, leveraging Microsoft Threat Intelligence databases. This tool will show you
as well the systems in your virtual networks that generate most traffic, so that you can visually identify
bottlenecks before they degenerate into problems.
Network Performance Manager can generate synthetic traffic to measure the performance of network
connections over multiple links, giving you a perspective on the evolution of WAN and Internet connections
over time, as well as offering valuable monitoring information about Microsoft ExpressRoute circuits.
VPN diagnostics can help troubleshooting site-to-site VPN connections connecting your applications to users
on-premises.
Connection Monitor allows you to measure the network availability between sets of endpoints.
Other information sources
Not only your application components are producing data, but there are many other signals that you need to track
to effectively operate a cloud environment:
Activity Log : this is a trail audit that lets you see every change that has gone through Azure APIs. It can be
critical to understand sudden performance changes or problems, that might have been due to a
misconfiguration of the Azure platform.
Azure Ser vice Health : sometimes outages are provoked not by configuration changes, but by glitches in the
Azure platform itself. You can find information about any Azure-related problem impacting your application in
the Azure Service Health logs.
Azure Advisor : find here recommendations about how to optimize your Azure platform to reduce costs,
improve your security posture, or increase the availability of your environment.
Azure Security Center : not a focus of this pillar, but to be included for completeness: Azure Security Center
can help you to understand whether your Azure resources are configured according to security best practices

Monitoring best practices

Event correlation
One critical advantage of Azure Monitor is that it is the monitoring tool for the whole Azure platform. As the
previous sections have shown, Azure Monitor holds metrics and logs relevant to your application code, the
platform where it is running, the data components, as well as the network connecting the application to its users.
This enables operators to compare metrics of different application components to each other, and find out
dependencies that might have been hidden otherwise.
Dashboards in Azure offer a great way of exposing the rich information contained in Azure Monitor to other users.
Make sure to create shared dashboards in order to expose relevant information to the different groups involved in
operating your application, including Developers and Operators. If more complex visualizations are required, Azure
Monitor data can be exported to Power BI for advanced data analysis.
Notifications
Whether it is for application, network or platform monitoring, you should not expect operators to constantly look at
dashboards. Instead, alerts should be used to send proactive notifications to the relevant individuals that will react
on them. Action groups in Azure Monitor can be used to notify multiple recipients, to trigger automated actions, or
even to automatically open tickets in IT Service Management Tools such as ServiceNow.
Automation around alerts is critical due to the highly collaborative nature of DevOps and the inherent speed
needed for effective incident management. Earlier this year, a report from DevOps.com came out stating that 80%
of IT teams are alerted to critical incidents via email. Email is an effective form of communication, but it shouldn’t
be the most common notification method for a critical issue. Instead, if you can define actions to be executed upon
receiving certain alerts (such as scaling up or down) your system will be self-healing.
Other monitoring tasks
Beyond Azure Monitor, you will want to keep an eye on certain events to make sure that your application is running
smoothly:
Review Azure subscription limits for your resources, and make sure you are not coming too close.
Understand Azure support plans. Refer to Azure support FAQs. Familiarize your team with Azure support.
Make sure that you monitor expiration dates of digital certificates, or even better, configure automatic digital
certificate renewal with Azure Key Vault.

Summary
You can use any monitoring platform to manage your Azure resources. Microsoft's first party offering is Azure
Monitor, a comprehensive solution for metrics and logs from the infrastructure to the application code, including
the ability to trigger alerts and automated actions as well as data visualization.
 DevOps Checklist
11/2/2020 • 14 minutes to read • Edit Online

DevOps is the integration of development, quality assurance, and IT operations into a unified culture and set of
processes for delivering software. Use this checklist as a starting point to assess your DevOps culture and process.

Culture
Ensure business alignment across organizations and teams. Conflicts over resources, purpose, goals, and
priorities within an organization can be a risk to successful operations. Ensure that the business, development, and
operations teams are all aligned.
Ensure the entire team understands the software lifecycle. Your team needs to understand the overall
lifecycle of the application, and which part of the lifecycle the application is currently in. This helps all team
members know what they should be doing now, and what they should be planning and preparing for in the future.
Reduce cycle time. Aim to minimize the time it takes to move from ideas to usable developed software. Limit the
size and scope of individual releases to keep the test burden low. Automate the build, test, configuration, and
deployment processes whenever possible. Clear any obstacles to communication among developers, and between
developers and operations.
Review and improve processes. Your processes and procedures, both automated and manual, are never final.
Set up regular reviews of current workflows, procedures, and documentation, with a goal of continual
improvement.
Do proactive planning. Proactively plan for failure. Have processes in place to quickly identify issues when they
occur, escalate to the correct team members to fix, and confirm resolution.
Learn from failures. Failures are inevitable, but it's important to learn from failures to avoid repeating them. If an
operational failure occurs, triage the issue, document the cause and solution, and share any lessons that were
learned. Whenever possible, update your build processes to automatically detect that kind of failure in the future.
Optimize for speed and collect data. Every planned improvement is a hypothesis. Work in the smallest
increments possible. Treat new ideas as experiments. Instrument the experiments so that you can collect production
data to assess their effectiveness. Be prepared to fail fast if the hypothesis is wrong.
Allow time for learning. Both failures and successes provide good opportunities for learning. Before moving on
to new projects, allow enough time to gather the important lessons, and make sure those lessons are absorbed by
your team. Also give the team the time to build skills, experiment, and learn about new tools and techniques.
Document operations. Document all tools, processes, and automated tasks with the same level of quality as your
product code. Document the current design and architecture of any systems you support, along with recovery
processes and other maintenance procedures. Focus on the steps you actually perform, not theoretically optimal
processes. Regularly review and update the documentation. For code, make sure that meaningful comments are
included, especially in public APIs, and use tools to automatically generate code documentation whenever possible.
Share knowledge. Documentation is only useful if people know that it exists and can find it. Ensure the
documentation is organized and easily discoverable. Be creative: Use brown bags (informal presentations), videos,
or newsletters to share knowledge.

Development
Provide developers with production-like environments. If development and test environments don't match
the production environment, it is hard to test and diagnose problems. Therefore, keep development and test
environments as close to the production environment as possible. Make sure that test data is consistent with the
data used in production, even if it's sample data and not real production data (for privacy or compliance reasons).
Plan to generate and anonymize sample test data.
Ensure that all authorized team members can provision infrastructure and deploy the application.
Setting up production-like resources and deploying the application should not involve complicated manual tasks or
detailed technical knowledge of the system. Anyone with the right permissions should be able to create or deploy
production-like resources without going to the operations team.

This recommendation doesn't imply that anyone can push live updates to the production deployment. It's about
reducing friction for the development and QA teams to create production-like environments.

Instrument the application for insight. To understand the health of your application, you need to know how it's
performing and whether it's experiencing any errors or problems. Always include instrumentation as a design
requirement, and build the instrumentation into the application from the start. Instrumentation must include event
logging for root cause analysis, but also telemetry and metrics to monitor the overall health and usage of the
application.
Track your technical debt. In many projects, release schedules can get prioritized over code quality to one degree
or another. Always keep track when this occurs. Document any shortcuts or other suboptimal implementations, and
schedule time in the future to revisit these issues.
Consider pushing updates directly to production. To reduce the overall release cycle time, consider pushing
properly tested code commits directly to production. Use feature toggles to control which features are enabled. This
allows you to move from development to release quickly, using the toggles to enable or disable features. Toggles
are also useful when performing tests such as canary releases, where a particular feature is deployed to a subset of
the production environment.

Testing
Automate testing. Manually testing software is tedious and susceptible to error. Automate common testing tasks
and integrate the tests into your build processes. Automated testing ensures consistent test coverage and
reproducibility. Integrated UI tests should also be performed by an automated tool. Azure offers development and
test resources that can help you configure and execute testing. For more information, see Development and test.
Test for failures. If a system can't connect to a service, how does it respond? Can it recover once the service is
available again? Make fault injection testing a standard part of review on test and staging environments. When your
test process and practices are mature, consider running these tests in production.
Test in production. The release process doesn't end with deployment to production. Have tests in place to ensure
that deployed code works as expected. For deployments that are infrequently updated, schedule production testing
as a regular part of maintenance.
Automate performance testing to identify performance issues early. The impact of a serious performance
issue can be as severe as a bug in the code. While automated functional tests can prevent application bugs, they
might not detect performance problems. Define acceptable performance goals for metrics like latency, load times,
and resource usage. Include automated performance tests in your release pipeline, to make sure the application
meets those goals.
Perform capacity testing. An application might work fine under test conditions, and then have problems in
production due to scale or resource limitations. Always define the maximum expected capacity and usage limits.
Test to make sure the application can handle those limits, but also test what happens when those limits are
exceeded. Capacity testing should be performed at regular intervals.
After the initial release, you should run performance and capacity tests whenever updates are made to production
code. Use historical data to fine-tune tests and to determine what types of tests need to be performed.
Perform automated security penetration testing. Ensuring your application is secure is as important as
testing any other functionality. Make automated penetration testing a standard part of the build and deployment
process. Schedule regular security tests and vulnerability scanning on deployed applications, monitoring for open
ports, endpoints, and attacks. Automated testing does not remove the need for in-depth security reviews at regular
intervals.
Perform automated business continuity testing. Develop tests for large-scale business continuity, including
backup recovery and failover. Set up automated processes to perform these tests regularly.

Release
Automate deployments. Automate deploying the application to test, staging, and production environments.
Automation enables faster and more reliable deployments, and ensures consistent deployments to any supported
environment. It removes the risk of human error caused by manual deployments. It also makes it easy to schedule
releases for convenient times, to minimize any effects of potential downtime. Have systems in place to detect any
problems during rollout, and have an automated way to roll forward fixes or roll back changes.
Use continuous integration. Continuous integration (CI) is the practice of merging all developer code into a
central codebase on a regular schedule, and then automatically performing standard build and test processes. CI
ensures that an entire team can work on a codebase at the same time without having conflicts. It also ensures that
code defects are found as early as possible. Preferably, the CI process should run every time that code is committed
or checked in. At the very least, it should run once per day.

Consider adopting a trunk based development model. In this model, developers commit to a single branch (the
trunk). There is a requirement that commits never break the build. This model facilitates CI, because all feature
work is done in the trunk, and any merge conflicts are resolved when the commit happens.

Consider using continuous deliver y. Continuous delivery (CD) is the practice of ensuring that code is always
ready to deploy, by automatically building, testing, and deploying code to production-like environments. Adding
continuous delivery to create a full CI/CD pipeline will help you detect code defects as soon as possible, and ensures
that properly tested updates can be released in a very short time.

Continuous deployment is an additional process that automatically takes any updates that have passed through
the CI/CD pipeline and deploys them into production. Continuous deployment requires robust automatic
testing and advanced process planning, and may not be appropriate for all teams.

Make small incremental changes. Large code changes have a greater potential to introduce bugs. Whenever
possible, keep changes small. This limits the potential effects of each change, and makes it easier to understand and
debug any issues.
Control exposure to changes. Make sure you're in control of when updates are visible to your end users.
Consider using feature toggles to control when features are enabled for end users.
Implement release management strategies to reduce deployment risk . Deploying an application update to
production always entails some risk. To minimize this risk, use strategies such as canary releases or blue-green
deployments to deploy updates to a subset of users. Confirm the update works as expected, and then roll the
update out to the rest of the system.
Document all changes. Minor updates and configuration changes can be a source of confusion and versioning
conflict. Always keep a clear record of any changes, no matter how small. Log everything that changes, including
patches applied, policy changes, and configuration changes. (Don't include sensitive data in these logs. For example,
log that a credential was updated, and who made the change, but don't record the updated credentials.) The record
of the changes should be visible to the entire team.
Consider making infrastructure immutable. Immutable infrastructure is the principle that you shouldn't
modify infrastructure after it's deployed to production. Otherwise, you can get into a state where ad hoc changes
have been applied, making it hard to know exactly what changed. Immutable infrastructure works by replacing
entire servers as part of any new deployment. This allows the code and the hosting environment to be tested and
deployed as a block. Once deployed, infrastructure components aren't modified until the next build and deploy
cycle.

Monitoring
Make systems obser vable. The operations team should always have clear visibility into the health and status of a
system or service. Set up external health endpoints to monitor status, and ensure that applications are coded to
instrument the operations metrics. Use a common and consistent schema that helps you correlate events across
systems. Azure Diagnostics and Application Insights are the standard method of tracking the health and status of
Azure resources. Microsoft Operation Management Suite also provides centralized monitoring and management
for cloud or hybrid solutions.
Aggregate and correlate logs and metrics . A properly instrumented telemetry system will provide a large
amount of raw performance data and event logs. Make sure that telemetry and log data is processed and correlated
in a short period of time, so that operations staff always have an up-to-date picture of system health. Organize and
display data in ways that give a cohesive view of any issues, so that whenever possible it's clear when events are
related to one another.

Consult your corporate retention policy for requirements on how data is processed and how long it should be
stored.

Implement automated aler ts and notifications. Set up monitoring tools like Azure Monitor to detect patterns
or conditions that indicate potential or current issues, and send alerts to the team members who can address the
issues. Tune the alerts to avoid false positives.
Monitor assets and resources for expirations. Some resources and assets, such as certificates, expire after a
given amount of time. Make sure to track which assets expire, when they expire, and what services or features
depend on them. Use automated processes to monitor these assets. Notify the operations team before an asset
expires, and escalate if expiration threatens to disrupt the application.

Management
Automate operations tasks. Manually handling repetitive operations processes is error-prone. Automate these
tasks whenever possible to ensure consistent execution and quality. Code that implements the automation should
be versioned in source control. As with any other code, automation tools must be tested.
Take an infrastructure-as-code approach to provisioning. Minimize the amount of manual configuration
needed to provision resources. Instead, use scripts and Azure Resource Manager templates. Keep the scripts and
templates in source control, like any other code you maintain.
Consider using containers. Containers provide a standard package-based interface for deploying applications.
Using containers, an application is deployed using self-contained packages that include any software, dependencies,
and files needed to run the application, which greatly simplifies the deployment process.
Containers also create an abstraction layer between the application and the underlying operating system, which
provides consistency across environments. This abstraction can also isolate a container from other processes or
applications running on a host.
Implement resiliency and self-healing. Resiliency is the ability of an application to recover from failures.
Strategies for resiliency include retrying transient failures, and failing over to a secondary instance or even another
region. For more information, see Designing reliable Azure applications . Instrument your applications so that issues
are reported immediately and you can manage outages or other system failures.
Have an operations manual. An operations manual or runbook documents the procedures and management
information needed for operations staff to maintain a system. Also document any operations scenarios and
mitigation plans that might come into play during a failure or other disruption to your service. Create this
documentation during the development process, and keep it up to date afterwards. This is a living document, and
should be reviewed, tested, and improved regularly.
Shared documentation is critical. Encourage team members to contribute and share knowledge. The entire team
should have access to documents. Make it easy for anyone on the team to help keep documents updated.
Document on-call procedures. Make sure on-call duties, schedules, and procedures are documented and shared
to all team members. Keep this information up-to-date at all times.
Document escalation procedures for third-par ty dependencies. If your application depends on external
third-party services that you don't directly control, you must have a plan to deal with outages. Create
documentation for your planned mitigation processes. Include support contacts and escalation paths.
Use configuration management. Configuration changes should be planned, visible to operations, and recorded.
This could take the form of a configuration management database, or a configuration-as-code approach.
Configuration should be audited regularly to ensure that what's expected is actually in place.
Get an Azure suppor t plan and understand the process. Azure offers a number of support plans. Determine
the right plan for your needs, and make sure the entire team knows how to use it. Team members should
understand the details of the plan, how the support process works, and how to open a support ticket with Azure. If
you are anticipating a high-scale event, Azure support can assist you with increasing your service limits. For more
information, see the Azure Support FAQs.
Follow least-privilege principles when granting access to resources. Carefully manage access to resources.
Access should be denied by default, unless a user is explicitly given access to a resource. Only grant a user access to
what they need to complete their tasks. Track user permissions and perform regular security audits.
Use role-based access control. Assigning user accounts and access to resources should not be a manual
process. Use role-based access control (RBAC) grant access based on Azure Active Directory identities and groups.
Use a bug tracking system to track issues. Without a good way to track issues, it's easy to miss items,
duplicate work, or introduce additional problems. Don't rely on informal person-to-person communication to track
the status of bugs. Use a bug tracking tool to record details about problems, assign resources to address them, and
provide an audit trail of progress and status.
Manage all resources in a change management system. All aspects of your DevOps process should be
included in a management and versioning system, so that changes can be easily tracked and audited. This includes
code, infrastructure, configuration, documentation, and scripts. Treat all these types of resources as code throughout
the test/build/review process.
Use checklists. Create operations checklists to ensure processes are followed. It's common to miss something in a
large manual, and following a checklist can force attention to details that might otherwise be overlooked. Maintain
the checklists, and continually look for ways to automate tasks and streamline processes.
For more about DevOps, see What is DevOps? on the Visual Studio site.
 Overview of the performance efficiency pillar
12/18/2020 • 3 minutes to read • Edit Online

Performance efficiency is the ability of your workload to scale to meet the demands placed on it by users in an
efficient manner. Before the cloud became popular, when it came to planning how a system would handle increases
in load, many organizations intentionally provisioned workloads to be oversized to meet business requirements.
This might make sense in on-premises environments because it ensured capacity during peak usage. Capacity
reflects resource availability (CPU and memory). This was a major consideration for processes that would be in
place for a number of years.
Just as you needed to anticipate increases in load in on-premises environments, you need to anticipate increases in
cloud environments to meet business requirements. One difference is that you may no longer need to make long-
term predictions for anticipated changes to ensure that you will have enough capacity in the future. Another
difference is in the approach used to manage performance.

What is scalability and why is it important?

An important consideration in achieving performance efficiency is to consider how your application scales and to
implement PaaS offerings that have built-in scaling operations. Scalability is the ability of a system to handle
increased load. Services covered by Azure Autoscale can scale automatically to match demand to accommodate
workload. They will scale out to ensure capacity during workload peaks and scaling will return to normal
automatically when the peak drops.
In the cloud, the ability to take advantage of scalability depends on your infrastructure and services. Some
platforms, such as Kubernetes, were built with scaling in mind. Virtual machines, on the other hand, may not scale
as easily although scale operations are possible. With virtual machines, you may want to plan ahead to avoid
scaling infrastructure in the future to meet demand. Another option is to select a different platform such as Azure
virtual machines scale sets.
When using scalability, you need only predict the current average and peak times for your workload. Payment plan
options allow you to manage this. You pay either per minute or per-hour depending on the service for a
designated time period.

Principles
Follow these principles to guide you through improving performance efficiency:
Become Data-driven - Embrace a data-driven culture to deliver timely insights to everyone in your
organization across all your data. To harness this culture, get the best performance from your analytics
solution across all your data, ensure data has the security and privacy needed for your business
environment, and make sure you have tools that enable everyone in your organization to gain insights from
your data.
Avoid antipatterns - A performance antipattern is a common practice that is likely to cause scalability
problems when an application is under pressure. For example, you can have an application that behaves as
expected during performance testing. However, when it is released to production and starts to handle live
workloads, performance decreases. Scalability problems such as rejecting user requests, stalling, or
throwing exceptions may arise. To learn how to identify and fix these antipatterns, see Performance
antipatterns for cloud applications.
Perform load testing to set limits - Load testing helps ensure that your applications can scale and do
 not go down during peak traffic. Load test each application to understand how it performs at various scales.
To learn about Azure service limits, see Managing limits.
Understand billing for metered resources - Your business requirements will determine the tradeoffs
between cost and level of performance efficiency. Azure doesn't directly bill based on the resource cost.
Charges for a resource, such as a virtual machine, are calculated by using one or more meters. Meters are
used to track a resource’s usage over time. These meters are then used to calculate the bill.
Monitor and optimize - Lack of monitoring new services and the health of current workloads are major
inhibitors in workload quality. The overall monitoring strategy should take into consideration not only
scalability, but resiliency (infrastructure, application, and dependent services) and application performance
as well. For purposes of scalability, looking at the metrics would allow you to provision resources
dynamically and scale with demand.

Next steps
Performance efficiency impacts the entire architecture spectrum. Bridge gaps in you knowledge of Azure by
reviewing the 5 pillars in the Microsoft Azure Well-Architected Framework.
To assess your workload using the tenets found in the Microsoft Azure Well-Architected Framework, see the
Microsoft Azure Well-Architected Review.
 Design scalable Azure applications
12/18/2020 • 8 minutes to read • Edit Online

Application design is critical to handling scale as load increases. This article will give you insight on the most
important topics. For more topics related to handling scale, see the Design Azure applications for efficiency article
in the Performance efficiency pillar.

Choose the right data storage

The overall design of the data tier can greatly affect an application's performance and scalability. The Azure storage
platform is designed to be massively scalable to meet the data storage and performance needs of modern
applications.
Data services in the Azure storage platform are:
Azure Blob - A massively scalable object store for text and binary data. Includes support for big data analytics
through Data Lake Storage Gen2.
Azure Files - Managed file shares for cloud or on-premises deployments.
Azure Queue - A messaging store for reliable messaging between application components.
Azure Tables - A NoSQL store for schemaless storage of structured data.
Azure Disks - Block-level storage volumes for Azure VMs.
Most cloud workloads adopt the polyglot persistence approach. Instead of using one data store service, a mix of
technologies is used. Your application will most likely require more than one type of data store depending on your
requirements. For examples of when to use these data storage types, see Example scenarios.
Each service is accessed through a storage account. To get started, see Create a storage account.

Database considerations
The choice of database can affect an application's performance and scalability. Database reads and writes involve a
network call and storage I/O, both of which are expense operations. Choosing the right database service to store
and retrieve data is therefore a critical decision and must be considered to ensure application scalability. Azure has
many database services that will fit most needs. In addition, there are third-party options that can be considered
from Azure Marketplace.
To help you choose a database type, determine if application storage requirements fit a relational design (SQL)
versus a key-value/document/graph design (NO-SQL). Some applications may have both a SQL and a NO-SQL
database for different storage needs. Use the Azure data store decision tree to help you find the appropriate
managed data storage solution.
Why use a relational database?
Use a relational database when strong consistency guarantees are important — where all changes are atomic, and
transactions always leave the data in a consistent state. However, a relational database generally can't scale out
horizontally without sharding the data in some way. Implementing manual sharding can be a time consuming task.
Also, the data in relational database must be normalized, which isn't appropriate for every data set.
If a relational database is considered optimal, Azure offers several PaaS options that fully manage hosting and
operations of the database. Azure SQL Database can host single databases or multiple databases (Azure SQL
Database Managed Instance). The suite of offerings spans requirements that cross performance, scale, size,
resiliency, disaster recovery, and migration compatibility. Azure offers the following PaaS relational database
services:
Azure SQL Database
Azure Database for MySQL
Azure Database for PostgreSQL
Azure Database for MariaDB
Why use a NO -SQL database?
Use a NO-SQL database when application performance and availability are more important than strong
consistency. NO-SQL is ideal for handling large, unrelated, indeterminate, or rapidly changing data. NO-SQL
databases have trade-offs. For specifics, see Some challenges with NoSQL databases.
Azure provides two managed services that optimize for NO-SQL solutions: Azure Cosmos DB and Azure Cache for
Redis. For document and graph databases, Cosmos DB offers extreme scale and performance.
For a detailed description of NO-SQL and relational databases, see Understanding the differences.

Choose the right VM size

Choosing the wrong VM size can result in capacity issues as VMs approach their limits. It can also lead to
unnecessary cost. To choose the right VM size, consider your workloads, number of CPUs, RAM capacity, disk size,
and speed according to business requirements. For a snapshot of Azure VM sizes and their purpose, see Sizes for
virtual machines in Azure.
Azure offers the following categories of VM sizes, each designed to run different workloads. Click a category for
details.
General-purpose - Provide balanced CPU-to-memory ratio. Ideal for testing and development, small to medium
databases, and low to medium traffic web servers.
Memory optimized - Offer a high memory-to-CPU ratio that is great for relational database servers, medium to
large caches, and in-memory analytics.
Compute optimized - Have a high CPU-to-memory ratio. These sizes are good for medium traffic web servers,
network appliances, batch processes, and application servers.
GPU optimized - Available with single, multiple, or fractional GPUs. These sizes are designed for compute-
intensive, graphics-intensive, and visualization workloads.
High performance compute - Designed to deliver leadership-class performance, scalability, and cost efficiency
for a variety of real-world HPC workloads.
Storage optimized - Offer high disk throughput and IO, and are ideal for Big Data, SQL, NO-SQL databases, data
warehousing, and large transactional databases. Examples include Cassandra, MongoDB, Cloudera, and Redis.
You can change the sizing requirements according to your needs and requirements.

Build with microservices

Microservices are a popular architectural style for building applications that are resilient, highly scalable,
independently deployable, and able to evolve quickly. A microservices architecture consists of a collection of small,
autonomous services. Each service is self-contained and should implement a single business capability. Breaking up
larger entities into small discrete pieces alone doesn't ensure sizing and scaling capabilities. Application logic needs
to be written to control this.
One of the many benefits of microservices is that they can be scaled independently. This lets you scale out
subsystems that require more resources, without scaling out the entire application. Another benefit is fault
isolation. If an individual microservice becomes unavailable, it won't disrupt the entire application, as long as any
upstream microservices are designed to handle faults correctly (for example, by implementing circuit breaking).
To learn more about the benefits of microservices, see Benefits.
Building with microservices comes with challenges such as development and testing. Writing a small service that
relies on other dependent services requires a different approach than writing a traditional monolithic or layered
application. Existing tools are not always designed to work with service dependencies. Refactoring across service
boundaries can be difficult. It is also challenging to test service dependencies, especially when the application is
evolving quickly.
See Challenges for a list of possible drawbacks of a microservice architecture.

Use dynamic service discovery for microservices applications

When there are many separate services or instances of services in play, they will need to receive instructions on
who to contact and/or other configuration information. Hard coding this information is flawed, and this is where
service discovery steps in. A service instance can spin up and dynamically discover the configuration information it
needs to become functional without having that information hard coded.
When combined with an orchestration platform designed to execute and manage microservices such as Kubernetes
or Service Fabric, individual services can be right sized, scaled up, scaled down, and dynamically configured to
match user demand. Using an orchestrator such as Kubernetes or Service Fabric, you can pack a higher density of
services onto a single host, which allows for more efficient utilization of resources. Both of these platforms provide
built-in services for executing, scaling, and operating a microservices architecture; and one of those key services is
discovery and finding where a particular service is running.
Kubernetes supports pod autoscaling and cluster autoscaling. To learn more, see Autoscaling. A Service Fabric
architecture takes a different approach to scaling for stateless and stateful services. To learn more, see Scaling
considerations.

TIP
When appropriate, decomposing an application into microservices is a level of decoupling that is an architectural best
practice. A microservices architecture can also bring some challenges. The design patterns in Design patterns for
microservices can help mitigate these challenges.

Establish connection pooling

Establishing connections to databases is typically an expensive operation that involves establishing an
authenticated network connection to the remote database server. This is especially true for applications that open
new connections frequently. Use connection pooling to reduce connection latency by reusing existing connections
and enable higher database throughput (transactions per second) on the server. By doing this, you avoid the
expense of opening a new connection for each request.
Pool size limits
Azure limits the number of network connections a virtual machine or AppService instance can make. Exceeding this
limit would cause connections to be slowed down or terminated. With connection pooling, a fixed set of
connections are established at the startup time and maintained. In many cases, a default pool size might only
consist of a small handful of connections that performs quickly in basic test scenarios, but become a bottleneck
under scale when the pool is exhausted. Establishing a pool size that maps to the number of concurrent
transactions supported on each application instance is a best practice.
Each database and application platform will have slightly different requirements for the right way to set up and
leverage the pool. See SQL Server Connection Pooling for a .NET code example using SQL Server and Azure
Database. In all cases, testing is paramount to ensure a connection pool is properly established and working as
designed under load.
 TIP
Use a pool size that uses the same number of concurrent connections. Choose a size that can handle more than the existing
connections so you can quickly handle a new request coming in.

Integrated security
Integrated security is a singular unified solution to protect every service that a business runs through a set of
common policies and configuration settings. In addition to reducing issues associated with scaling, provisioning,
and managing (including higher costs and complexity), integrated security also increases control and overall
security. However, there may be times when you may not want to use connection pooling for security reasons. For
example, although connection pooling improves the performance of subsequent database requests for a single
user, that user cannot take advantage of connections made by other users. It also results in at least one connection
per user to the database server.
Measure your business' security requirements against the advantages and disadvantages of connection pooling. To
learn more, see Pool fragmentation

Next steps
Application efficiency
 Design Azure applications for efficiency
12/18/2020 • 4 minutes to read • Edit Online

Making choices that effect performance efficiency is critical to application design. For additional related topics, see
the Design scalable Azure applications article in the Performance efficiency pillar.

Reduce response time with asynchronous programming

The time for the caller to receive a response could range from milliseconds to minutes. During that time, the thread
is held by the process until the response comes back, or if an exception happens. This is inefficient because it means
that no other requests can be processed during the time waiting for a response. An example when multiple
requests in flight is inefficient is a bank account. In this situation, only one resource can operate on the request at
the same time. Another example is when connection pools can't be shared, and then all of the requests need
separate connections to complete.
Asynchronous programming is an alternative approach. It enables a remote service to be executed without waiting
and blocking resources on the client. This is a critical pattern for enabling cloud scalable software and is available in
most modern programming languages and platforms.
There are many ways to inject asynchronous programming into an application design. In a simplest form, remote
calls can be asynchronously executed using built-in language constructs like "async/await" in .NET C#. Review a
language construct example. .NET has other built-in platform support for asynchronous programming with task
and event based asynchronous patterns.

Process faster by queuing and batching requests

Similar to asynchronous programming, queuing services has long been used as a scalable mechanism to hand off
processing work to a service. Highly scalable queuing services are natively supported in Azure. The queue is a
storage buffer located between the caller and the processing service. It takes requests, stores them in a buffer, and
queues the requests to provide services around the reliable delivery and management of the queued data.
Using a queue is often the best way to hand off work to a processor service. The processor service receives work by
listening on a queue and dequeuing messages. If items to be processed enter too quickly, the queuing service will
keep them in the queue until the processing service has available resources and asks for a new work item
(message). By leveraging the dynamic nature of Azure Functions, the processor service can easily autoscale on
demand as the queue builds up to meet the intake pressure. Developing processor logic with Azure Functions to
run task logic from a queue is a common, scalable, and cost effective way to using queuing between a client and a
processor.
Azure provides some native first-party queueing services with Azure Storage Queues (simple queuing service
based on Azure Storage) and Azure Service Bus (message broker service supporting transactions and reduced
latency). Many other third-party options are also available through Azure Marketplace.
To learn more about queue-based Load Leveling, see Queue-based Load Leveling pattern. To compare and contrast
queues, see Storage queues and Service Bus queues - compared and contrasted.

Optimize with data compression

A well-known optimization best practice for scaling is to use a compression strategy to compress and bundle web
pages or API responses. The idea is to shrink the data returned from a page or API back to the browser or client
app. Compressing the data returned to clients optimizes network traffic and accelerates the application. .NET has
built-in framework support for this technique with GZip compression. For more information, see Response
compression in ASP.NET Core.

Improve scalability with session affinity

If an application is stateful, meaning that data or state will be stored locally in the instance of the application, it may
increase performance by enabling session affinity. When session affinity is enabled, subsequent requests to the
application will be directed to the same server that processed the first request. If session affinity is not enabled,
subsequent requests would be directed to the next available server depending on the load balancing rules. Session
affinity allows the instance to have some persistent or cached data/context, which can speed subsequent requests.

TIP
Migrate an Azure Cloud Services application to Azure Service Fabric describes best practices about stateless services for an
application that is migrated from old Azure Cloud Services to Azure Service Fabric.

Run background jobs to meet integration needs

Many types of applications require background tasks that run independently of the user interface (UI). Examples
include batch jobs, intensive processing tasks, and long-running processes such as workflows. Background jobs can
be executed without requiring user interaction. The application can start the job and then continue to process
interactive requests from users. To learn more, see Background jobs.
Background tasks must offer sufficient performance to ensure they do not block the application, or cause
inconsistencies due to delayed operation when the system is under load. Typically, performance is improved by
scaling the compute instances that host the background tasks. For a list of considerations, see Scaling and
performance considerations.
Logic Apps is a serverless consumption (pay-per-use) service that enables a vast set of out-of-the-box ready-to-use
connectors and a long-running workflow engine to enable cloud-native integration needs quickly. Logic Apps is
flexible enough for scenarios like running tasks/jobs, advanced scheduling, and triggering. Logic Apps also has
advanced hosting options to allow it to run within enterprise restricted cloud environments. Logic Apps can be
combined with all other Azure services to complement one another, or it can be used independently.
Like all serverless services, Logic Apps doesn't require VM instances to be purchased, enabled, and scaled up and
down. Instead, Logic Apps scale automatically on serverless PaaS provided instances, and a consumer only pays
based on usage.

Next steps
Design for scaling
 Design for scaling
12/18/2020 • 4 minutes to read • Edit Online

Scaling is the ability of a system to handle increased load. Services covered by Azure Autoscale can scale
automatically to match demand to accommodate workload. They will scale out to ensure capacity during workload
peaks and scaling will return to normal automatically when the peak drops.

Plan for growth

Planning for growth starts with understanding your current workloads. This can help to anticipate scale needs
based on predictive usage scenarios. An example of a predictive usage scenario is an e-commerce site that
recognizes that its infrastructure should scale appropriately for an anticipated high volume of holiday traffic.
Perform load tests and stress tests to determine the necessary infrastructure to support the predicted spikes in
workloads. A good plan includes incorporating a buffer to accommodate for random spikes.
For more information on how to determine the upper and maximum limits of an application's capacity, see
Performance testing in the Performance Efficiency pillar.
Another critical component of planning for scale is to make sure the region that hosts your application supports the
necessary capacity required to accommodate load increase. If you are using a multiregion architecture, make sure
the secondary regions can also support the increase. A region can offer the product but may not support the
predicted load increase without the necessary SKUs (Stock Keeping Units) so you need to verify this.
To verify your region and available SKUs, first select the product and regions in Products available by region.

Then, check the SKUs available in the Azure portal.

Add scale units
For each resource, know the upper scaling limits, and use sharding or decomposition to go beyond those limits.
Design the application so that it's easily scaled by adding one or more scale units, such as by using the Deployment
Stamps pattern. Determine the scale units for the system in terms of well-defined sets of resources.
The next step might be to use built-in scaling features or tools to understand which resources need to scale
concurrently with other resources. For example, adding X number of front-end VMs might require Y number of
additional queues and Z number of storage accounts to handle the additional workload. So a scale unit could
consist of X VM instances, Y queues, and Z storage accounts.

Use Autoscaling to manage load increases and decreases

Autoscaling enables you to run the right amount of resources to handle the load of your app. It adds resources
(called scaling out) to handle an increase in load such as seasonal workloads. Autoscaling saves money by
removing idle resources (called scaling in) during a decrease in load such as nights and weekends for some
corporate apps.
You automatically scale between the minimum and maximum number of instances to run and add or remove VMs
automatically based on a set of rules.

For more information, see Autoscaling.

Take advantage of platform autoscaling features

Here's how you can benefit from autoscaling features:
Use built-in autoscaling features when possible rather than custom or third-party mechanisms.
Use scheduled scaling rules where possible to ensure that resources are available.
Add reactive autoscaling to the rules where appropriate to cope with unexpected changes in demand.

NOTE
Providing your application is explicitly designed to handle some of its instances being terminated, remember to use
autoscaling to scale down/in resources that are no longer necessary for the given load in order to reduce operational costs.

For more information, see Autoscaling.

Autoscale CPU or memory-intensive applications

In the case of CPU or memory-intensive applications, scaling up to a larger machine SKU with more CPU or
memory may be required. Once the demand for CPU or memory has been reduced, instances can revert back to
the original instance.
For example, you may have an application that processes images, videos or music. Given the process and
requirements, it may make sense to scale up a server (e.g., add CPU or memory) to quickly process the large media
file. While scaling out allows the system to process more files simultaneously, it does not impact processing speed
of each individual file.

Autoscale with Azure compute services

The way autoscaling works is that metrics are collected for the resource (CPU and memory utilization) and the
application (requests queued and requests per second). Rules can then be created to add and remove instances
depending on how the rule evaluates. An App Services App Plan allows autoscale rules to be set for scale-out/scale-
in and scale-up/scale-down. Scaling also applies to Azure Automation.

The Application Service autoscaling sample shows how to create an Azure App Service plan which includes an
Azure App Service.
Azure Kubernetes Service (AKS) offers two levels of autoscale:
Horizontal autoscale - Can be enabled on service containers to add more or fewer pod instances within the
cluster.
Cluster autoscale - Can be enabled on the agent VM instances running an agent node-pool to add more or
 remove VM instances dynamically.
Other Azure services include the following:
Azure Ser vice Fabric - Virtual machine scale sets offer autoscale capabilities for true IaaS scenarios.
Azure App Gateway and Azure API Management - PaaS offerings for ingress services that enable autoscale.
Azure Functions , Azure Logic Apps , and App Ser vices - Serverless pay-per-use consumption modeling
that inherently provide autoscaling capabilities.
Azure SQL Database - PaaS platform to change performance characteristics of a database on the fly and
assign more resources when needed or release the resources when they are not needed. Allows scaling
up/down, read scale-out, and global scale-out/sharding capabilities.
Each service documents its autoscale capabilities. Review Autoscale overview for a general discussion on Azure
platform autoscale.

NOTE
If your application does not have built-in ability to autoscale, or isn't configured to scale out automatically as load increases,
it's possible that your application's services will fail if they become saturated with user requests. See Azure Automation for
possible solutions.

Next steps
Plan for capacity
 Plan for capacity
12/18/2020 • 4 minutes to read • Edit Online

Azure offers many options to meet capacity requirements as your business grows. These options can also minimize
cost.

Prepare infrastructure for large-scale events

Large-scale application design takes careful planning and possibly involves complex implementation. Work with
your business and marketing teams to prepare for large-scale events. Knowing if there will be sudden spikes in
traffic such as Superbowl, Black Friday, or Marketing pushes, can allow you to prepare your infrastructure ahead of
time.
A fundamental design principle in Azure is to scale out by adding machines or service instances based on increased
demand. Scaling out can be a better alternative to purchasing additional hardware, which may not be in your
budget. Depending on your payment plan, you don't pay for idle VMs or need to reserve capacity in advance. A
pay-as-you-go plan is usually ideal for applications that need to meet planned spikes in traffic.

NOTE
Don't plan for capacity to meet the highest level of expected demand. An inappropriate or misconfigured service can impact
cost. For example, building a multiregion service when the service levels don't require high-availability or geo-redundancy will
increase cost without reasonable business justification.

Choose the right resources

Right sizing your infrastructure to meet the needs of your applications can save you considerably as opposed to a
"one size fits all" solution often employed with on-premises hardware. You can choose various options when you
deploy Azure VMs to support workloads.
Each VM type has specific features and different combinations of CPU, memory, and disks. For example, the B-
series VMs are ideal for workloads that don't need the full performance of the CPU continuously, like web servers,
proof of concepts, small databases, and development build environments. The B-Series offers a cost effective way
to deploy these workloads that don't need the full performance of the CPU continuously and burst in their
performance.
For a list of sizes and a description of the recommended use, see sizes for virtual machines in Azure.
Continually monitor workloads after migration to find out if your VMs aren't optimized or have frequent periods
when they aren't used. If you discover this, it makes sense to either shut down the VMs or downscale them by
using virtual machine scale sets. You can optimize a VM with Azure Automation, virtual machine scale sets, auto-
shutdown, and scripted or third-party solutions. To learn more, see Automate VM optimization.
Along with choosing the right VMs, selecting the right storage type can save your organization significant cost
every month. For a list of storage data types, access tiers, storage account types, and storage redundancy options,
see Select the right storage.

Use metrics to fine-tune scaling

It's often difficult to understand the relationship between metrics and capacity requirements, especially when an
application is initially deployed. Provision a little extra capacity at the beginning, and then monitor and tune the
autoscaling rules to bring the capacity closer to the actual load. Autoscaling enables you to run the right amount of
resources to handle the load of your app. It adds resources (called scaling out) to handle an increase in load such as
seasonal workloads and customer facing applications.
After configuring the autoscaling rules, monitor the performance of your application over time. Use the results of
this monitoring to adjust the way in which the system scales if necessary.
Azure Monitor autoscale provides a common set of autoscaling functionality for virtual machine scale sets, Azure
App Service, and Azure Cloud Service. Scaling can be performed on a schedule, or based on a runtime metric, such
as CPU or memory usage. For example, you can scale out by one instance if average CPU usage is above 70%, and
scale in by one instance if CPU usage falls below 50 percent.
The default autoscaling rules are set to know when it's time to execute an autoscaling action in order to prevent the
system from reacting too quickly. To learn more, see Autoscaling.
For a list of built-in metrics, see Azure Monitor autoscaling common metrics. You can also implement custom
metrics by using Application Insights to monitor the performance of your live applications. Some Azure services
use different scaling methods.

Preemptively scaling based on trends

Preemptively scaling based on historical data can ensure your application has consistent performance, even though
your metrics haven't yet indicated the need to scale. Schedule-based rules allow you to scale when you see time
patterns in your load and want to scale before a possible load increase or decrease occurs. For example, you can set
a trigger attribute to scale out to 10 instances on weekdays, and scale in to four (4) instances on Saturday and
Sunday. If you can predict the load on the application, consider using scheduled autoscaling, which adds and
removes instances to meet anticipated peaks in demand.
To learn more, see Use Azure Monitor autoscale.

Next steps
Performance testing
 Performance testing
12/18/2020 • 4 minutes to read • Edit Online

Performance testing helps to maintain systems properly and fix defects before problems reach system users. It
helps maintain the efficiency, responsiveness, scalability, and speed of applications when compared with business
requirements. When done effectively, performance testing should give you the diagnostic information necessary to
eliminate bottlenecks, which lead to poor performance. A bottleneck occurs when data flow is either interrupted or
stops due to insufficient capacity to handle the workload.
To avoid experiencing poor performance, commit time and resources to testing system performance. Two subsets
of performance testing, load testing and stress testing, can determine the upper (close to capacity limit) and
maximum (point of failure) limit, respectively, of the application's capacity. By performing these tests, you can
determine the necessary infrastructure to support the anticipated workloads.
A best practice is to plan for a load buffer to accommodate random spikes without overloading the infrastructure.
For example, if a normal system load is 100,000 requests per second, the infrastructure should support 100,000
requests at 80% of total capacity (i.e., 125,000 requests per second). If you anticipate that the application will
continue to sustain 100,000 requests per second, and the current SKU (Stock Keeping Unit) introduces latency at
65,000 requests per second, you'll most likely need to upgrade your product to the next higher SKU. If there is a
secondary region, you'll need to ensure that it also supports the higher SKU.

Load testing
Load testing measures system performance as the workload increases. It identifies where and when your
application breaks, so you can fix the issue before shipping to production. It does this by testing system behavior
under typical and heavy loads. The following are key points to consider for load testing:
Know the Azure ser vice limits - Different Azure services have soft and hard limits associated with them.
The terms soft limit and hard limit describe the current, adjustable service limit (soft limit) and the maximum
limit (hard limit). Understand the limits for the services you consume so that you are not blocked if you need
to exceed them. For a list of the most common Azure limits, see Azure subscription and service limits,
quotas, and constraints.

The ResourceLimits sample shows how to query the limits and quotas for commonly used resources.
Measure typical loads - Knowing the typical and maximum loads on your system helps you understand
when something is operating outside of its designed limits. Monitor traffic to understand application
behavior.
Understand application behavior under various scales - Load test your application to understand
how it performs at various scales. First, test to see how the application performs under a typical load. Then,
test to see how it performs under load using different scaling operations. To get additional insight into how
to evaluate your application as the amount of traffic sent to it increases, see Autoscale best practices.

Stress testing
Unlike load testing, which ensures that a system can handle what it’s designed to handle, stress testing focuses on
overloading the system until it breaks. A stress test determines how stable a system is and its ability to withstand
extreme increases in load. It does this by testing the maximum number requests from another service (for
example) that a system can handle at a given time before performance is compromised and fails. Find this
maximum to understand what kind of load the current environment can adequately support.
Determine the maximum demand you want to place on memory, CPU, and disk IOPS. Once a stress test has been
performed, you will know the maximum supported load and an operational margin. It is best to choose an
operational threshold so that scaling can be performed before the threshold has been reached.
Once you determine an acceptable operational margin and response time under typical loads, verify that the
environment is configured adequately. To do this, make sure the SKUs that you selected are based on the desired
margins. Be careful to stay as close as possible to your margins. Allocating too much can increase costs and
maintenance unnecessarily; allocating too few can result in poor user experience.
In addition to stress testing through increased load, you can stress test by reducing resources to identify what
happens when the machine runs out of memory. You can also stress test by increasing latency (e.g., the database
takes 10x time to reply, writes to storage takes 10x longer, etc.).

Multiregion testing
A multiregion architecture can provide higher availability than deploying to a single region. If a regional outage
affects the primary region, you can use Front Door to use the secondary region. This architecture can also help if an
individual subsystem of the application fails.
Test the amount of time it would take for users to be rerouted to the paired region so that the region doesn't fail. To
learn more about routing, see Front Door routing methods. Typically, a planned test failover can help determine
how much time would be required to fully scale to support the redirected load.

Configure the environment based on testing results

Once you have performed testing and found an acceptable operational margin and response under increased
levels of load, configure the environment to sustain performance efficiency. Scale out or scale in to handle
increases and decreases in load. For example, you may know that you will encounter high levels of traffic during
the day and low levels on weekends. You may configure the environment to scale out for increases in load or scale
in for decreases before the load actually changes.
For more information on autoscaling, see Design for scaling in the Performance Efficiency pillar.

NOTE
Ensure that a rule has been configured to scale the environment back down once load reaches below the set thresholds. This
will save you money.

Next steps
Testing tools
 Testing tools
12/18/2020 • 4 minutes to read • Edit Online

There are multiple stages in the development and deployment life cycle in which tests can be performed.
Application code, infrastructure automation, and fault tolerance should all be tested. This can ensure that the
application will perform as expected in every situation. You'll want to test early enough in the application life cycle
to catch and fix errors. Errors are cheaper to repair when caught early and can be expensive or impossible to fix
later.
Testing can be automated or manual. Automating tests is the best way to make sure that they are executed.
Depending on how frequently tests are performed, they are typically limited in duration and scope. Manual testing
is run much less frequently. For a list of tests that you should consider while developing and deploying applications,
see Testing your application and Azure environment.

Identify baselines and goals for performance

Knowing where you are (baseline) and where you want to be (goal) make it easier to plan how to get there.
Established baselines and goals will help you to stay on track and measure progress. Testing may also uncover a
need to perform additional testing on areas that you may not have planned.
Baselines can vary based on connections or platforms that a user may leverage for accessing the application. It may
be important to establish baselines that address the different connections, platforms, and elements such as time of
day, or weekday versus weekend.
There are many types of goals when determining baselines for application performance. Some examples are, the
time it takes to render a page, or a desired number of transactions if your site conducts e-commerce. The following
list shows some examples of questions that may help you to determine goals.
What are your baselines and goals for:
Establishing an initial connection to a service?
An API endpoint complete response?
Server response times?
Latency between systems/microservices?
Database queries?

Caching data
Caching can dramatically improve performance, scalability, and availability. The more data that you have, the
greater the benefits of caching become. Caching typically works well with data that is immutable or that changes
infrequently. Examples include reference information such as product and pricing information in an e-commerce
application, or shared static resources that are costly to construct. Some or all of this data can be loaded into the
cache at application startup to minimize demand on resources and to improve performance.
Use performance testing and usage analysis to determine whether pre-populating or on-demand loading of the
cache, or a combination of both, is appropriate. The decision should be based on the volatility and usage pattern of
the data. Cache utilization and performance analysis are particularly important in applications that encounter heavy
loads and must be highly scalable.
To learn more about how to use caching as a solution in testing, see Caching.
Use Azure Redis to cache data
Azure Cache for Redis is a caching service that can be accessed from any Azure application, whether the application
is implemented as a cloud service, a website, or inside an Azure virtual machine. Caches can be shared by client
applications that have the appropriate access key. It is a high-performance caching solution that provides
availability, scalability, and security.
To learn more about using Azure Cache for Redis, see Considerations for implementing caching in Azure.

Content delivery network

Content delivery networks (CDNs) are typically used to deliver static content such as images, style sheets,
documents, client-side scripts, and HTML pages. The major advantages of using a CDN are lower latency and faster
delivery of content to users, regardless of their geographical location in relation to the datacenter where the
application is hosted. CDNs can help to reduce load on a web application because the application does not have to
service requests for the content that is hosted in the CDN. Using a CDN is a good way to minimize the load on your
application, and maximize availability and performance. Consider adopting this strategy for all of the appropriate
content and resources your application uses.
Decide how you will handle local development and testing when some static content is expected to be served from
a CDN. For example, you could pre-deploy the content to the CDN as part of your build script. Alternatively, use
compile directives or flags to control how the application loads the resources. For example, in debug mode, the
application could load static resources from a local folder. In release mode, the application would use the CDN.
To learn more about CDNs, see Best practices for using content delivery networks (CDNs).

Benchmark testing
Benchmarking is the process of simulating different workloads on your application and measuring application
performance for each workload. It is the best way to figure out what resources you will need to host your
application. Use performance indicators to assess whether your application is performing as expected or not. Take
into consideration VM sizes and disk sizes.
See the Optimize IOPS, throughput, and latency table for guidance.

Metrics
Metrics measure trends over time. They are available for interactive analysis in the Azure portal with Azure Metrics
Explorer. Metrics also can be added to an Azure dashboard for visualization in combination with other data and
used for near-real time alerting.
Performance testing gives you the ability to see specific details on the processing capabilities of applications. You'll
most likely want a monitoring tool that allows you to discover proactively if the issues you find through testing are
appearing in both your infrastructure and applications. Azure Monitor Metrics is a feature of Azure Monitor that
collects metrics from monitored resources into a time series database.
With Azure Monitor, you can collect, analyze, and act on telemetry from your cloud and on-premises environments.
It helps you understand how applications are performing and identifies issues affecting them and the resources
they depend on.
For a list of Azure metrics, see Supported metrics with Azure Monitor.

Next steps
Performance monitoring
 Monitoring for Scalability
11/2/2020 • 2 minutes to read • Edit Online

Monitoring for scalability should be part of your overall monitoring strategy that utilizes Azure Monitor. The overall
monitoring strategy should take into consideration not only scalability, but resiliency (infrastructure, application,
and dependent services) and application performance as well. Most services in Azure offer the ability to turn on
both data and management plane logs as well as metrics. For purposes of scalability, looking at the metrics would
allow you to scale based scale up, scale out, scale in, and scale down. The ability to scale dynamically is one of the
biggest values of moving to the cloud.

What are some of the reasons for setting up auto scaling

Are the systems able to handle the current number of requests?
Are the systems meeting the performance NFRs?
Are the systems out of resources (CPU, Memory, I/O)?
Is there an opportunity to have cost savings by not overprovisioning resources?

What is the goal of auto scaling

Scale up or out or in and down resources to meet the current demand.
Decrease costs by scaling down when resources are not needed.
Improve customer experience be offering uninterrupted service even when the demand peaks of valleys.

How can metrics be used to auto scale

As stated before, most Azure services offer the ability to export logs and metrics to services such
as Log Analytics and external service like Splunk through Azure Event Hubs. Furthermore, application leveraging
technologies such as Application Insights can further enhance the telemetry coming out of the applications.
The metrics coming out of Azure services include metrics such as CPU and memory,
utilization, bandwidth information, current storage utilization information, and much more. You can refer to the
supported metrics for Azure Monitor

How do Azure services auto scale

App Services, ASE, and virtual machine scale sets be configured with autoscaling rules that can be based on
several metrics including CPU, memory, bandwidth, etc. These rules can create new instances (scale out) or
remove instances (scale in) or a running service. This capability can be enhanced by generating custom
events from technologies like Application Insight that could be based on some other custom metrics.
Azure Kubernetes Services offers both the ability to scale pods as well as to auto scale nodes. Scaling rules
can be based on internal metrics or can leverage metrics from systems like Prometheus.
Other services, such as Application Gateway, can be scaled manually. In this case, it is important to leverage
services such as Log Analytics to raise alerts when the service is no longer able to handle the load.
Monitor Metrics and auto scale on performance and schedule for VMs and virtual machine scale sets
For Container workloads, container monitoring solution in Azure Monitor should be utilized.

The AppServiceAutoscalingSample sample shows autoscaling scenarios based on CPU utilization.

Monitoring best practices
Know the minimum number of instances that should run at any given time.
Determine what metrics are best for your solution to base your auto scaling rules.
Configure the auto scaling rules for those services that include it.
Create alert rules for the services that could be scaled manually.
Monitor your environment to make sure that auto scaling is working as expected. For example, watch out for
scaling events from the telemetry coming out of the management plane.
Monitor web applications using Azure Application Insights.
Monitor network performance.
Consider reviewing as applicable, network performance monitor, service connectivity monitor,
ExpressRoute monitor
For long-term storage, consider archiving of the Monitoring Data.
Track activities using Azure Security and Audit Logs.

Related Useful Resources

Azure Monitor Data Platform
Auto scaling best practices
Manage log data and workspaces in Azure Monitor.
Azure Diagnostic Logs and Schemas
 Performance efficiency checklist
11/2/2020 • 14 minutes to read • Edit Online

Performance efficiency is the ability of your workload to scale to meet the demands placed on it by users in an
efficient manner, and is one of the pillars of the Microsoft Azure Well-Architected Framework. Use this checklist to
review your application architecture from a performance efficiency standpoint.

Application design
Par tition the workload . Design parts of the process to be discrete and decomposable. Minimize the size of each
part, while following the usual rules for separation of concerns and the single responsibility principle. This allows
the component parts to be distributed in a way that maximizes use of each compute unit (such as a role or
database server). It also makes it easier to scale the application by adding instances of specific resources. For
complex domains, consider adopting a microservices architecture.
Design for scaling . Scaling allows applications to react to variable load by increasing and decreasing the number
of instances of roles, queues, and other services they use. However, the application must be designed with this in
mind. For example, the application and the services it uses must be stateless, to allow requests to be routed to any
instance. This also prevents the addition or removal of specific instances from adversely affecting current users.
You should also implement configuration or autodetection of instances as they are added and removed, so that
code in the application can perform the necessary routing. For example, a web application might use a set of
queues in a round-robin approach to route requests to background services running in worker roles. The web
application must be able to detect changes in the number of queues, to successfully route requests and balance the
load on the application.
Scale as a unit . Plan for additional resources to accommodate growth. For each resource, know the upper scaling
limits, and use sharding or decomposition to go beyond these limits. Determine the scale units for the system in
terms of well-defined sets of resources. This makes applying scale-out operations easier, and less prone to negative
impact on the application through limitations imposed by lack of resources in some part of the overall system. For
example, adding x number of web and worker roles might require y number of additional queues and z number of
storage accounts to handle the additional workload generated by the roles. So a scale unit could consist of x web
and worker roles, y queues, and z storage accounts. Design the application so that it's easily scaled by adding one
or more scale units.
Avoid client affinity . Where possible, ensure that the application does not require affinity. Requests can thus be
routed to any instance, and the number of instances is irrelevant. This also avoids the overhead of storing,
retrieving, and maintaining state information for each user.
Take advantage of platform autoscaling features . Where the hosting platform supports an autoscaling
capability, such as Azure Autoscale, prefer it to custom or third-party mechanisms unless the built-in mechanism
can't fulfill your requirements. Use scheduled scaling rules where possible to ensure resources are available
without a start-up delay, but add reactive autoscaling to the rules where appropriate to cope with unexpected
changes in demand. You can use the autoscaling operations in the classic deployment model to adjust autoscaling,
and to add custom counters to rules. For more information, see Auto-scaling guidance.
Offload CPU-intensive and I/O-intensive tasks as background tasks . If a request to a service is expected to
take a long time to run or absorb considerable resources, offload the processing for this request to a separate task.
Use worker roles or background jobs (depending on the hosting platform) to execute these tasks. This strategy
enables the service to continue receiving further requests and remain responsive. For more information, see
Background jobs guidance.
Distribute the workload for background tasks . Where there are many background tasks, or the tasks require
considerable time or resources, spread the work across multiple compute units (such as worker roles or
background jobs). For one possible solution, see the Competing Consumers pattern.
Consider moving toward a shared-nothing architecture . A shared-nothing architecture uses independent,
self-sufficient nodes that have no single point of contention (such as shared services or storage). In theory, such a
system can scale almost indefinitely. While a fully shared-nothing approach is generally not practical for most
applications, it may provide opportunities to design for better scalability. For example, avoiding the use of server-
side session state, client affinity, and data partitioning are good examples of moving toward a shared-nothing
architecture.

Data management
Use data par titioning . Divide the data across multiple databases and database servers, or design the application
to use data storage services that can provide this partitioning transparently (examples include Azure SQL Database
Elastic Database, and Azure Table storage). This approach can help to maximize performance and allow easier
scaling. There are different partitioning techniques, such as horizontal, vertical, and functional. You can use a
combination of these to achieve maximum benefit from increased query performance, simpler scalability, more
flexible management, better availability, and to match the type of store to the data it will hold. Also, consider using
different types of data store for different types of data, choosing the types based on how well they are optimized
for the specific type of data. This may include using table storage, a document database, or a column-family data
store, instead of, or as well as, a relational database. For more information, see Data partitioning guidance.
Design for eventual consistency . Eventual consistency improves scalability by reducing or removing the time
needed to synchronize related data partitioned across multiple stores. The cost is that data is not always consistent
when it is read, and some write operations may cause conflicts. Eventual consistency is ideal for situations where
the same data is read frequently but written infrequently. For more information, see the Data Consistency Primer.
Reduce chatty interactions between components and ser vices . Avoid designing interactions in which an
application is required to make multiple calls to a service (each of which returns a small amount of data), rather
than a single call that can return all of the data. Where possible, combine several related operations into a single
request when the call is to a service or component that has noticeable latency. This makes it easier to monitor
performance and optimize complex operations. For example, use stored procedures in databases to encapsulate
complex logic, and reduce the number of round trips and resource locking.
Use queues to level the load for high velocity data writes . Surges in demand for a service can overwhelm
that service and cause escalating failures. To prevent this, consider implementing the Queue-Based Load Leveling
pattern. Use a queue that acts as a buffer between a task and a service that it invokes. This can smooth intermittent
heavy loads that may otherwise cause the service to fail or the task to time out.
Minimize the load on the data store . The data store is commonly a processing bottleneck, a costly resource,
and often not easy to scale out. Where possible, remove logic (such as processing XML documents or JSON
objects) from the data store, and perform processing within the application. For example, instead of passing XML to
the database (other than as an opaque string for storage), serialize or deserialize the XML within the application
layer and pass it in a form that is native to the data store. It's typically much easier to scale out the application than
the data store, so you should attempt to do as much of the compute-intensive processing as possible within the
application.
Minimize the volume of data retrieved . Retrieve only the data you require by specifying columns and using
criteria to select rows. Make use of table value parameters and the appropriate isolation level. Use mechanisms like
entity tags to avoid retrieving data unnecessarily.
Aggressively use caching . Use caching wherever possible to reduce the load on resources and services that
generate or deliver data. Caching is typically suited to data that is relatively static, or that requires considerable
processing to obtain. Caching should occur at all levels where appropriate in each layer of the application,
including data access and user interface generation. For more information, see the Caching Guidance.
Handle data growth and retention . The amount of data stored by an application grows over time. This growth
increases storage costs as well as latency when accessing the data, affecting application throughput and
performance. It may be possible to periodically archive some of the old data that is no longer accessed, or move
data that is rarely accessed into long-term storage that is more cost efficient, even if the access latency is higher.
Optimize Data Transfer Objects (DTOs) using an efficient binar y format . DTOs are passed between the
layers of an application many times. Minimizing the size reduces the load on resources and the network. However,
balance the savings with the overhead of converting the data to the required format in each location where it is
used. Adopt a format that has the maximum interoperability to enable easy reuse of a component.
Set cache control . Design and configure the application to use output caching or fragment caching where
possible, to minimize processing load.
Enable client side caching . Web applications should enable cache settings on the content that can be cached.
This is commonly disabled by default. Configure the server to deliver the appropriate cache control headers to
enable caching of content on proxy servers and clients.
Use Azure blob storage and the Azure Content Deliver y Network to reduce the load on the
application . Consider storing static or relatively static public content, such as images, resources, scripts, and style
sheets, in blob storage. This approach relieves the application of the load caused by dynamically generating this
content for each request. Additionally, consider using the Content Delivery Network to cache this content and
deliver it to clients. Using the Content Delivery Network can improve performance at the client because the content
is delivered from the geographically closest datacenter that contains a Content Delivery Network cache. For more
information, see the Content Delivery Network Guidance.
Optimize and tune SQL queries and indexes . Some T-SQL statements or constructs may have an adverse
effect on performance that can be reduced by optimizing the code in a stored procedure. For example, avoid
converting datetime types to a varchar before comparing with a datetime literal value. Use date/time
comparison functions instead. Lack of appropriate indexes can also slow query execution. If you use an
object/relational mapping framework, understand how it works and how it may affect performance of the data
access layer. For more information, see Query Tuning.
Consider denormalizing data . Data normalization helps to avoid duplication and inconsistency. However,
maintaining multiple indexes, checking for referential integrity, performing multiple accesses to small chunks of
data, and joining tables to reassemble the data imposes an overhead that can affect performance. Consider if some
additional storage volume and duplication is acceptable in order to reduce the load on the data store. Also consider
if the application itself (which is typically easier to scale) can be relied on to take over tasks such as managing
referential integrity in order to reduce the load on the data store. For more information, see Data partitioning
guidance.

Implementation
Review the performance antipatterns . See Performance antipatterns for cloud applications for common
practices that are likely to cause scalability problems when an application is under pressure.
Use asynchronous calls . Use asynchronous code wherever possible when accessing resources or services that
may be limited by I/O or network bandwidth, or that have a noticeable latency, in order to avoid locking the calling
thread.
Avoid locking resources, and use an optimistic approach instead . Never lock access to resources such as
storage or other services that have noticeable latency, because this is a primary cause of poor performance. Always
use optimistic approaches to managing concurrent operations, such as writing to storage. Use features of the
storage layer to manage conflicts. In distributed applications, data may be only eventually consistent.
Compress highly compressible data over high latency, low bandwidth networks . In the majority of cases
in a web application, the largest volume of data generated by the application and passed over the network is HTTP
responses to client requests. HTTP compression can reduce this considerably, especially for static content. This can
reduce cost as well as reducing the load on the network, though compressing dynamic content does apply a
fractionally higher load on the server. In other, more generalized environments, data compression can reduce the
volume of data transmitted and minimize transfer time and costs, but the compression and decompression
processes incur overhead. As such, compression should only be used when there is a demonstrable gain in
performance. Other serialization methods, such as JSON or binary encodings, may reduce the payload size while
having less impact on performance, whereas XML is likely to increase it.
Minimize the time that connections and resources are in use . Maintain connections and resources only for
as long as you need to use them. For example, open connections as late as possible, and allow them to be returned
to the connection pool as soon as possible. Acquire resources as late as possible, and dispose of them as soon as
possible.
Minimize the number of connections required . Service connections absorb resources. Limit the number that
are required and ensure that existing connections are reused whenever possible. For example, after performing
authentication, use impersonation where appropriate to run code as a specific identity. This can help to make best
use of the connection pool by reusing connections.

NOTE
APIs for some services automatically reuse connections, provided service-specific guidelines are followed. It's important that
you understand the conditions that enable connection reuse for each service that your application uses.

Send requests in batches to optimize network use . For example, send and read messages in batches when
accessing a queue, and perform multiple reads or writes as a batch when accessing storage or a cache. This can
help to maximize efficiency of the services and data stores by reducing the number of calls across the network.
Avoid a requirement to store ser ver-side session state where possible. Server-side session state
management typically requires client affinity (that is, routing each request to the same server instance), which
affects the ability of the system to scale. Ideally, you should design clients to be stateless with respect to the servers
that they use. However, if the application must maintain session state, store sensitive data or large volumes of per-
client data in a distributed server-side cache that all instances of the application can access.
Optimize table storage schemas . When using table stores that require the table and column names to be
passed and processed with every query, such as Azure table storage, consider using shorter names to reduce this
overhead. However, do not sacrifice readability or manageability by using overly compact names.
Create resource dependencies during deployment or at application star tup . Avoid repeated calls to
methods that test the existence of a resource and then create the resource if it does not exist. Methods such as
CloudTable.CreateIfNotExists and CloudQueue.CreateIfNotExists in the Azure Storage Client Library follow this
pattern. These methods can impose considerable overhead if they are invoked before each access to a storage
table or storage queue. Instead:
Create the required resources when the application is deployed, or when it first starts (a single call to
CreateIfNotExists for each resource in the startup code for a web or worker role is acceptable). However, be sure
to handle exceptions that may arise if your code attempts to access a resource that doesn't exist. In these
situations, you should log the exception, and possibly alert an operator that a resource is missing.
Under some circumstances, it may be appropriate to create the missing resource as part of the exception
handling code. But you should adopt this approach with caution as the non-existence of the resource might be
indicative of a programming error (a misspelled resource name for example), or some other infrastructure-level
issue.
Use lightweight frameworks . Carefully choose the APIs and frameworks you use to minimize resource usage,
execution time, and overall load on the application. For example, using Web API to handle service requests can
reduce the application footprint and increase execution speed, but it may not be suitable for advanced scenarios
where the additional capabilities of Windows Communication Foundation are required.
Consider minimizing the number of ser vice accounts . For example, use a specific account to access
resources or services that impose a limit on connections, or perform better where fewer connections are
maintained. This approach is common for services such as databases, but it can affect the ability to accurately audit
operations due to the impersonation of the original user.
Carr y out performance profiling and load testing during development, as part of test routines, and before
final release to ensure the application performs and scales as required. This testing should occur on the same type
of hardware as the production platform, and with the same types and quantities of data and user load as it will
encounter in production. For more information, see Testing the performance of a cloud service.
 Overview of the reliability pillar
11/2/2020 • 13 minutes to read • Edit Online

Building a reliable application in the cloud is different from traditional application development. While historically
you may have purchased levels of redundant higher-end hardware to minimize the chance of an entire application
platform failing, in the cloud, we acknowledge up front that failures will happen. Instead of trying to prevent
failures altogether, the goal is to minimize the effects of a single failing component.
To assess your workload using the tenets found in the Microsoft Azure Well-Architected Framework, see the
Microsoft Azure Well-Architected Review.
Reliable applications are:
Resilient and recover gracefully from failures, and they continue to function with minimal downtime and data
loss before full recovery.
Highly available (HA) and run as designed in a healthy state with no significant downtime.
Understanding how these elements work together — and how they affect cost — is essential to building a reliable
application. It can help you determine how much downtime is acceptable, the potential cost to your business, and
which functions are necessary during a recovery.
This article provides a brief overview of building reliability into each step of the Azure application design process.
Each section includes a link to an in-depth article on how to integrate reliability into that specific step in the
process. If you're looking for reliability considerations for individual Azure services, review the Resiliency checklist
for specific Azure services.

Build for reliability

This section describes six steps for building a reliable Azure application. Each step links to a section that further
defines the process and terms.
1. Define requirements. Develop availability and recovery requirements based on decomposed workloads and
business needs.
2. Use architectural best practices. Follow proven practices, identify possible failure points in the architecture,
and determine how the application will respond to failure.
3. Test with simulations and forced failovers. Simulate faults, trigger forced failovers, and test detection and
recovery from these failures.
4. Deploy the application consistently. Release to production using reliable and repeatable processes.
5. Monitor application health. Detect failures, monitor indicators of potential failures, and gauge the health of
your applications.
6. Respond to failures and disasters. Identify when a failure occurs, and determine how to address it based
on established strategies.

Define requirements
Identify your business needs, and build your reliability plan to address them. Consider the following:
Identify workloads and usage. A workload is a distinct capability or task that is logically separated from
other tasks, in terms of business logic and data storage requirements. Each workload has different
requirements for availability, scalability, data consistency, and disaster recovery.
Plan for usage patterns. Usage patterns also play a role in requirements. Identify differences in
 requirements during critical and non-critical periods. For example, a tax-filing application can't fail during a
filing deadline. To ensure uptime, plan redundancy across several regions in case one fails. Conversely, to
minimize costs during non-critical periods, you can run your application in a single region.
Identify critical components and paths. Not all components of your system might be as important as
others. For example, you might have an optional component that adds incremental value, but that the
workload can run without if necessary. Understand where these components are, and conversely, where the
critical parts of your workload are. This will help to scope your availability and reliablity metrics and to plan
your recovery strategies to prioritise the highest-importance components.
Establish availability metrics — mean time to recover y (MTTR) and mean time between
failures (MTBF). MTTR is the average time it takes to restore a component after a failure. MTBF is how
long a component can reasonably expect to last between outages. Use these measures to determine where
to add redundancy and to determine service-level agreements (SLAs) for customers.
Establish recover y metrics — recover y time objective (RTO) and recover y point objective
(RPO). RTO is the maximum acceptable time an application can be unavailable after an incident. RPO is the
maximum duration of data loss that is acceptable during a disaster. To derive these values, conduct a risk
assessment and make sure you understand the cost and risk of downtime or data loss in your organization.

NOTE
If the MTTR of any critical component in a highly available setup exceeds the system RTO, a failure in the system
might cause an unacceptable business disruption. That is, you can't restore the system within the defined RTO.

Determine workload availability targets. To ensure that application architecture meets your business
requirements, define target SLAs for each workload. Account for the cost and complexity of meeting
availability requirements, in addition to application dependencies.
Understand ser vice-level agreements. In Azure, the SLA describes the Microsoft commitments for
uptime and connectivity. If the SLA for a particular service is 99.9 percent, you should expect the service to
be available 99.9 percent of the time.
Define your own target SLAs for each workload in your solution, so you can determine whether the
architecture meets the business requirements. For example, if a workload requires 99.99 percent uptime
but depends on a service with a 99.9 percent SLA, that service can't be a single point of failure in the
system.
For more information about developing requirements for reliable applications, see Application design for
resiliency.

Use architectural best practices

During the architectural phase, focus on implementing practices that meet your business requirements, identify
failure points, and minimize the scope of failures.
Perform a failure mode analysis (FMA). FMA builds resiliency into an application early in the design
stage. It helps you identify the types of failures your application might experience, the potential effects of
each, and possible recovery strategies.
Create a redundancy plan. The level of redundancy required for each workload depends on your
business needs and factors into the overall cost of your application.
Design for scalability. A cloud application must be able to scale to accommodate changes in usage.
Begin with discrete components, and design the application to respond automatically to load changes
whenever possible. Keep scaling limits in mind during design so you can expand easily in the future.
 Plan for subscription and ser vice requirements. You might need additional subscriptions to provision
enough resources to meet your business requirements for storage, connections, throughput, and more.
Use load-balancing to distribute requests. Load-balancing distributes your application's requests to
healthy service instances by removing unhealthy instances from rotation.
Implement resiliency strategies. Resiliency is the ability of a system to recover from failures and
continue to function. Implement resiliency design patterns, such as isolating critical resources, using
compensating transactions, and performing asynchronous operations whenever possible.
Build availability requirements into your design. Availability is the proportion of time your system is
functional and working. Take steps to ensure that application availability conforms to your service-level
agreement. For example, avoid single points of failure, decompose workloads by service-level objective,
and throttle high-volume users.
Manage your data. How you store, back up, and replicate data is critical.
Choose replication methods for your application data. Your application data is stored in various
data stores and might have different availability requirements. Evaluate the replication methods and
locations for each type of data store to ensure that they satisfy your requirements.
Document and test your failover and failback processes. Clearly document instructions to fail
over to a new data store, and test them regularly to make sure they are accurate and easy to follow.
Protect your data. Back up and validate data regularly, and make sure no single user account has
access to both production and backup data.
Plan for data recover y. Make sure that your backup and replication strategy provides for data
recovery times that meet your service-level requirements. Account for all types of data your application
uses, including reference data and databases.

Azure service dependencies

Microsoft Azure services are available globally to drive your cloud operations at an optimal level. You can choose
the best region for your needs based on technical and regulatory considerations: service capabilities, data
residency, compliance requirements, and latency.
Azure services deployed to Azure regions are listed on the Azure global infrastructure products page. To better
understand regions and Availability Zones in Azure, see Regions and Availability Zones in Azure.
Azure services are built for resiliency including high availability and disaster recovery. There are no services that
are dependent on a single logical data center (to avoid single points of failure). Non-regional services listed on
Azure global infrastructure products are services for which there is no dependency on a specific Azure region.
Non-regional services are deployed to two or more regions and if there is a regional failure, the instance of the
service in another region continues servicing customers. Certain non-regional services enable customers to
specify the region where the underlying virtual machine (VM) on which service runs will be deployed. For
example, Windows Virtual Desktop enables customers to specify the region location where the VM resides. All
Azure services that store customer data allow the customer to specify the specific regions in which their data will
be stored. The exception is Azure Active Directory (Azure AD), which has geo placement (such as Europe or North
America). For more information about data storage residency, see the Data residency map.

Test with simulations and forced failovers

Testing for reliability requires measuring how the end-to-end workload performs under failure conditions that
only occur intermittently.
Test for common failure scenarios by triggering actual failures or by simulating them. Use fault
injection testing to test common scenarios (including combinations of failures) and recovery time.
 Identify failures that occur only under load. Test for peak load, using production data or synthetic data
that is as close to production data as possible, to see how the application behaves under real-world conditions.
Run disaster recover y drills. Have a disaster recovery plan in place, and test it periodically to make sure it
works.
Perform failover and failback testing. Ensure that your application's dependent services fail over and fail
back in the correct order.
Run simulation tests. Testing real-life scenarios can highlight issues that need to be addressed. Scenarios
should be controllable and non-disruptive to the business. Inform management of simulation testing plans.
Test health probes. Configure health probes for load balancers and traffic managers to check critical system
components. Test them to make sure that they respond appropriately.
Test monitoring systems. Be sure that monitoring systems are reliably reporting critical information and
accurate data to help identify potential failures.
Include third-par ty ser vices in test scenarios. Test possible points of failure due to third-party service
disruption, in addition to recovery.
Testing is an iterative process. Test the application, measure the outcome, analyze and address any failures, and
repeat the process.
For more information about testing for application reliability, see Testing Azure applications for resiliency and
availability.

Deploy the application consistently

Deployment includes provisioning Azure resources, deploying application code, and applying configuration
settings. An update may involve all three tasks or a subset of them.
After an application is deployed to production, updates are a possible source of errors. Minimize errors with
predictable and repeatable deployment processes.
Automate your application deployment process. Automate as many processes as possible.
Design your release process to maximize availability. If your release process requires services to go
offline during deployment, your application is unavailable until they come back online. Take advantage of
platform staging and production features. Use blue-green or canary releases to deploy updates, so if a failure
occurs, you can quickly roll back the update.
Have a rollback plan for deployment. Design a rollback process to return to a last known good version
and to minimize downtime if a deployment fails.
Log and audit deployments. If you use staged deployment techniques, more than one version of your
application is running in production. Implement a robust logging strategy to capture as much version-specific
information as possible.
Document the application release process. Clearly define and document your release process, and ensure
that it's available to the entire operations team.
For more information about application reliability and deployment, see Deploying Azure applications for resiliency
and availability.

Monitor application health

Implement best practices for monitoring and alerts in your application so you can detect failures and alert an
operator to fix them.
Implement health probes and check functions. Run them regularly from outside the application to
identify degradation of application health and performance.
Check long-running workflows. Catching issues early can minimize the need to roll back the entire
 workflow or to execute multiple compensating transactions.
Maintain application logs.
Log applications in production and at service boundaries.
Use semantic and asynchronous logging.
Separate application logs from audit logs.
Measure remote call statistics, and share the data with the application team. To give your
operations team an instantaneous view into application health, summarize remote call metrics, such as
latency, throughput, and errors in the 99 and 95 percentiles. Perform statistical analysis on the metrics to
uncover errors that occur within each percentile.
Track transient exceptions and retries over an appropriate time frame. A trend of increasing
exceptions over time indicates that the service is having an issue and may fail.
Set up an early warning system. Identify the key performance indicators (KPIs) of an application's
health, such as transient exceptions and remote call latency, and set appropriate threshold values for each
of them. Send an alert to operations when the threshold value is reached.
Operate within Azure subscription limits. Azure subscriptions have limits on certain resource types,
such as the number of resource groups, cores, and storage accounts. Watch your use of resource types.
Monitor third-par ty ser vices. Log your invocations and correlate them with your application's health
and diagnostic logging using a unique identifier.
Train multiple operators to monitor the application and to perform manual recover y steps.
Make sure there is always at least one trained operator active.
For more information about monitoring for application reliability, see Monitoring Azure application health.

Respond to failures and disasters

Create a recovery plan, and make sure that it covers data restoration, network outages, dependent service failures,
and region-wide service disruptions. Consider your VMs, storage, databases, and other Azure platform services in
your recovery strategy.
Plan for Azure suppor t interactions. Before the need arises, establish a process for contacting Azure
support.
Document and test your disaster recover y plan. Write a disaster recovery plan that reflects the business
impact of application failures. Automate the recovery process as much as possible, and document any manual
steps. Regularly test your disaster recovery process to validate and improve the plan.
Fail over manually when required. Some systems can't fail over automatically and require a manual
failover. If an application fails over to a secondary region, perform an operational readiness test. Verify that the
primary region is healthy and ready to receive traffic again before failing back. Determine what the reduced
application functionality is and how the app informs users of temporary problems.
Prepare for application failure. Prepare for a range of failures, including faults that are handled
automatically, those that result in reduced functionality, and those that cause the application to become
unavailable. The application should inform users of temporary issues.
Recover from data corruption. If a failure happens in a data store, check for data inconsistencies when the
store becomes available again, especially if the data was replicated. Restore corrupt data from a backup.
Recover from a network outage. You might be able to use cached data to run locally with reduced
application functionality. If not, consider application downtime or fail over to another region. Store your data in
an alternate location until connectivity is restored.
Recover from a dependent ser vice failure. Determine which functionality is still available and how the
application should respond.
 Recover from a region-wide ser vice disruption. Region-wide service disruptions are uncommon, but you
should have a strategy to address them, especially for critical applications. You might be able to redeploy the
application to another region or redistribute traffic.
For more information about responding to failures and disaster recovery, see Failure and disaster recovery for
Azure applications.
 Designing resilient Azure applications
11/2/2020 • 11 minutes to read • Edit Online

Building resiliency (recovering from failures) and availability (running in a healthy state without significant
downtime) into your apps begins with gathering requirements. For example, how much downtime is acceptable?
How much does potential downtime cost your business? What are your customer's availability requirements? How
much do you invest in making your application highly available? What is the risk versus the cost?

Determine subscription and service requirements

Choose the right subscription and service features for your app by working through these tasks:
Evaluate requirements against Azure subscription and ser vice limits . Azure subscriptions have limits
on certain resource types, such as number of resource groups, cores, and storage accounts. If your application
requirements exceed Azure subscription limits, create another Azure subscription and provision sufficient
resources there. Individual Azure services have consumption limits — for example, limits on storage,
throughput, number of connections, requests per second, and other metrics. Your application will fail if it
attempts to use resources beyond these limits, resulting in service throttling and possible downtime for
affected users. Depending on the specific service and your application requirements, you can often avoid these
limits by scaling up (for example, choosing another pricing tier) or scaling out (such as adding new instances).
Determine how many storage accounts you need. Azure allows a specific number of storage accounts
per subscription. For more information, see Azure subscription and service limits, quotas, and constraints.
Select the right ser vice tier for Azure SQL Database. If your application uses Azure SQL Database, select
the appropriate service tier. If the tier cannot handle your application's database transaction unit (DTU)
requirements, your data use will be throttled. For more information on selecting the correct service plan, see
SQL Database options and performance: Understand what's available in each service tier.
Provision sufficient request units (RUs) in Azure Cosmos DB . With Azure Cosmos DB, you pay for the
throughput you provision and the storage you consume on an hourly basis. The cost of all database operations
is normalized as RUs, which abstracts the system resources such as CPU, IOPS, and memory. For more
information, see Request Units in Azure Cosmos DB.

Resiliency strategies
This section describes some common resiliency strategies. Most of these strategies are not limited to a particular
technology. The descriptions summarize the general idea behind each technique and include links to further
reading.
Implement resiliency patterns for remote operations, where appropriate. If your application depends on
communication between remote services, follow design patterns for dealing with transient failures.
Retr y transient failures. These can be caused by momentary loss of network connectivity, a dropped
database connection, or a timeout when a service is busy. Often, a transient failure can be resolved by
retrying the request.
For many Azure services, the client software development kit (SDK) implements automatic retries in a
way that is transparent to the caller. See Retry guidance for specific services.
Or implement the Retry pattern to help the application handle anticipated, temporary failures
transparently when it tries to connect to a service or network resource.

The RetryPatternSample shows an implementation of the retry pattern.

 Use a circuit breaker to handle faults that might take a variable amount of time to fix. The Circuit Breaker
pattern can prevent an application from repeatedly trying an operation that is likely to fail. The circuit
breaker wraps calls to a service and tracks the number of recent failures. If the failure count exceeds a
threshold, the circuit breaker starts returning an error code without calling the service. This gives the
service time to recover and helps avoid cascading failures.
Isolate critical resources. Failures in one subsystem can sometimes cascade, resulting in failures in other
parts of the application. This can happen if a failure prevents resources such as threads or sockets from
being freed, leading to resource exhaustion. To avoid this, you can partition a system into isolated groups so
that a failure in one partition does not bring down the entire system.
Here are some examples of this technique, which is sometimes called the Bulkhead pattern:
Partition a database (for example, by tenant), and assign a separate pool of web server instances for
each partition.
Use separate thread pools to isolate calls to different services. This helps to prevent cascading
failures if one of the services fails. For an example, see the Netflix Hystrix library.
Use containers to limit the resources available to a particular subsystem.

Apply compensating transactions . A compensating transaction is a transaction that undoes the effects
of another completed transaction. In a distributed system, it can be difficult to achieve strong transactional
consistency. Compensating transactions help to achieve consistency by using a series of smaller, individual
transactions that can be undone at each step. For example, to book a trip, a customer might reserve a car, a
hotel room, and a flight. If one of these steps fails, the entire operation fails. Instead of trying to use a single
distributed transaction for the entire operation, you can define a compensating transaction for each step.
Implement asynchronous operations, whenever possible. Synchronous operations can monopolize
resources and block other operations while the caller waits for the process to complete. Design each part of
your application to allow for asynchronous operations, whenever possible. For more information on how to
implement asynchronous programming in C#, see Asynchronous Programming.

Plan for usage patterns

Identify differences in requirements during critical and non-critical periods. Are there certain critical periods when
the system must be available? For example, a tax-filing application can't fail during a filing deadline and a video
streaming service shouldn't lag during a live event. In these situations, weigh the cost against the risk.
To ensure uptime and meet service-level agreements (SLAs) in critical periods, plan redundancy across several
regions in case one fails, even if it costs more.
Conversely, during non-critical periods, run your application in a single region to minimize costs.
 In some cases, you can mitigate additional expenses by using modern serverless techniques that have
consumption-based billing.

Identify distinct workloads

Cloud solutions typically consist of multiple application workloads. A workload is a distinct capability or task that is
logically separated from other tasks in terms of business logic and data storage requirements. For example, an e-
commerce app might have the following workloads:
Browse and search a product catalog.
Create and track orders.
View recommendations.
Each workload has different requirements for availability, scalability, data consistency, and disaster recovery. Make
your business decisions by balancing cost versus risk for each workload.
Also decompose workloads by service-level objective. If a service is composed of critical and less-critical
workloads, manage them differently and specify the service features and number of instances needed to meet
their availability requirements.

Managing third party services

If your application has dependencies on third-party services, identify how these services can fail and what effect
failures will have on your application.
A third-party service might not include monitoring and diagnostics. Log calls to these services and correlate them
with your application's health and diagnostic logging using a unique identifier. For more information on proven
practices for monitoring and diagnostics, see Monitoring and diagnostics guidance.
See the Health Endpoint Monitoring pattern for a solution to track this with code samples.

Monitoring third-party services

If your application has dependencies on third-party services, identify where and how these services can fail and
what effect those failures will have on your application. Keep in mind the service-level agreement (SLA) for the
third-party service and the effect it might have on your disaster recovery plan.
A third-party service might not provide monitoring and diagnostics capabilities, so it's important to log your
invocations of them and to correlate them with your application's health and diagnostic logging using a unique
identifier. For more information on proven practices for monitoring and diagnostics, see Monitoring and
diagnostics guidance.

Load balancing
Proper load-balancing allows you to meet availability requirements and to minimize costs associated with
availability.
Use load-balancing to distribute requests. Load-balancing distributes your application's requests to
healthy service instances by removing unhealthy instances from rotation. If your service uses Azure App
Service or Azure Cloud Services, it's already load-balanced for you. However, if your application uses Azure
VMs, you need to provision a load-balancer. For more information, see What is Azure Load Balancer?
You can use Azure Load Balancer to:
Load-balance incoming Internet traffic to your VMs. This configuration is known as a public Load
Balancer.
 Load-balance traffic across VMs inside a virtual network. You can also reach a Load Balancer front end
from an on-premises network in a hybrid scenario. Both scenarios use a configuration that is known as
an internal Load Balancer.
Port forward traffic to an itemized port on specific VMs with inbound network address translation (NAT)
rules.
Provide outbound connectivity for VMs inside your virtual network by using a public Load Balancer.
Balance loads across regions with a traffic manager, such as Azure Traffic Manager. To load-
balance traffic across regions requires a traffic management solution, and Azure provides Traffic Manager.
You can also take advantage of third-party services that provide similar traffic-management capabilities.

Failure mode analysis

Failure mode analysis (FMA) builds resiliency into a system by identifying possible failure points and defining how
the application responds to those failures. The FMA should be part of the architecture and design phases, so failure
recovery is built into the system from the beginning. The goals of an FMA are to:
Determine what types of failures an application might experience and how the application detects those
failures.
Capture the potential effects of each type of failure and determine how the app responds.
Plan for logging and monitoring the failure and identify recovery strategies.
Here are some examples of failure modes and detection strategies for a specific failure point — a call to an
external web service:

FA IL URE M O DE DET EC T IO N ST RAT EGY

Service is unavailable HTTP 5xx

Throttling HTTP 429 (Too Many Requests)

Authentication HTTP 401 (Unauthorized)

Slow response Request times out

For more information about the FMA process, with specific recommendations for Azure, see Failure mode analysis.

Related samples are here.

Operating in multiple regions

If your application is deployed to a single region, in the rare event the entire region becomes unavailable, your
application will also be unavailable. This may be unacceptable under the terms of your application's SLA. If so,
consider deploying your application and its services across multiple regions. A multi-region deployment can use
an active-active pattern (distributing requests across multiple active instances) or an active-passive pattern
(keeping a "warm" instance in reserve, in case the primary instance fails)
Many failures are manageable within the same Azure region. However, in the unlikely event of a region-wide
service disruption, the locally redundant copies of your data aren't available. If you've enabled geo-replication,
there are three additional copies of your blobs and tables in a different region. If Microsoft declares the region lost,
Azure remaps all the DNS entries to the secondary region.
 NOTE
This process occurs only for region-wide service disruptions and is not within your control. Consider using Azure Site
Recovery to achieve better RPO and RTO. Using Site Recovery, you decide what is an acceptable outage and when to fail over
to the replicated VMs.

NOTE
The selection of the Resource Group location is important. In the event of a regional outage, you will be unable to control
resources inside that Resource Group, regardless of what region those resources are actually in (i.e., the resources in the
other region(s) will continue to function, but management plane operations will be unavailable.

Your response to a region-wide service disruption depends on your deployment and your disaster recovery plan.
As a cost-control strategy, for non-critical applications that don't require a guaranteed recovery time, it might
make sense to redeploy to a different region.
For applications that are hosted in another region with deployed roles but don't distribute traffic across regions
(active/passive deployment), switch to the secondary hosted service in the alternate region.
For applications that have a full-scale secondary deployment in another region (active/active deployment),
route traffic to that region.
To learn more about recovering from a region-wide service disruption, see Recover from a region-wide service
disruption.
VM recovery
For critical apps, plan for recovering VMs in the event of a region-wide service disruption.
Use Azure Backup or another backup method to create cross-region backups that are application consistent.
(Replication of the Backup vault must be configured at the time of creation.)
Use Site Recovery to replicate across regions for one-click application failover and failover testing.
Use Traffic Manager to automate user traffic failover to another region.
To learn more, see Recover from a region-wide service disruption, Virtual machines.
Storage recovery
To protect your storage in the event of a region-wide service disruption:
Use geo-redundant storage.
Know where your storage is geo-replicated. This affects where you deploy other instances of your data that
require regional affinity with your storage.
Check data for consistency after failover and, if necessary, restore from a backup.
To learn more, see Designing highly available applications using RA-GRS.
SQL Database and SQL Server
Azure SQL Database provides two types of recovery:
Use geo-restore to restore a database from a backup copy in another region. For more information, see
Recover an Azure SQL database using automated database backups.
Use active geo-replication to fail over to a secondary database. For more information, see Creating and using
active geo-replication.
For SQL Server running on VMs, see High availability and disaster recovery for SQL Server in Azure Virtual
Machines.
 Error handling for resilient applications in Azure
11/2/2020 • 2 minutes to read • Edit Online

Ensuring your application can recover from errors is critical when working in a distributed system

Transient fault handling

Track the number of transient exceptions and retries over time to uncover issues or failures in your application's
retry logic. A trend of increasing exceptions over time may indicate that the service is having an issue and may fail.
For more information, see Retry service specific guidance.
Use the Retry pattern, paying particular attention to issues and considerations. Avoid overwhelming dependent
services by implementing the Circuit Breaker pattern. Review and incorporate additional best practices guidance for
Transient fault handling. While calling systems that have Throttling pattern implemented, ensure that your retries
are not counter productive.

A reference implementation is available here. It uses Polly and IHttpClientBuilder to implement the Circuit
Breaker pattern.

Request timeouts
When making a service call or a database call ensure that appropriate request timeouts are set. Database
Connection timeouts are typically set to 30s. Use guidance on troubleshoot, diagnose, and prevent SQL connection
errors and transient errors for SQL Database.
Leverage design patterns that encapsulate robust timeout strategies like Choreography pattern or Compensating
Transaction pattern.

A reference implementation is available on GitHub.

Cascading Failures
The Circuit Breaker pattern provides stability while the system recovers from a failure and minimizes the impact on
performance. It can help to maintain the response time of the system by quickly rejecting a request for an operation
that's likely to fail, rather than waiting for the operation to time out, or never return.
Retry pattern. Describes how an application can handle anticipated temporary failures when it tries to connect to a
service or network resource by transparently retrying an operation that has previously failed.
Health Endpoint Monitoring pattern. A circuit breaker might be able to test the health of a service by sending a
request to an endpoint exposed by the service. The service should return information indicating its status.

Samples related to this pattern are here.

Application Health Probes

Configure and test health probes for your load balancers and traffic managers. Ensure that your health endpoint
checks the critical parts of the system and responds appropriately.
For Azure Traffic Manager, the health probe determines whether to fail over to another region. Your health
endpoint should check any critical dependencies that are deployed within the same region.
 For Azure Load Balancer, the health probe determines whether to remove a VM from rotation. The health
endpoint should report the health of the VM. Don't include other tiers or external services. Otherwise, a failure
that occurs outside the VM will cause the load balancer to remove the VM from rotation.
For guidance on implementing health monitoring in your application, see Health Endpoint Monitoring pattern.

Samples related to heath probes are here.

ARM template that deploys an Azure Load Balancer and health probes that detect the health of the sample
service endpoint.
An ASP.NET Core Web API that shows configuration of health checks at startup.

Command and Query Responsibility Segregation (CQRS)

Achieve levels of scale and performance needed for your solution by segregating read and write interfaces by
implementing the CQRS pattern.
 Failure mode analysis for Azure applications
12/18/2020 • 16 minutes to read • Edit Online

Failure mode analysis (FMA) is a process for building resiliency into a system, by identifying possible failure points
in the system. The FMA should be part of the architecture and design phases, so that you can build failure recovery
into the system from the beginning.
Here is the general process to conduct an FMA:
1. Identify all of the components in the system. Include external dependencies, such as identity providers, third-
party services, and so on.
2. For each component, identify potential failures that could occur. A single component may have more than
one failure mode. For example, you should consider read failures and write failures separately, because the
impact and possible mitigation steps will be different.
3. Rate each failure mode according to its overall risk. Consider these factors:
What is the likelihood of the failure. Is it relatively common? Extremely rare? You don't need exact
numbers; the purpose is to help rank the priority.
What is the impact on the application, in terms of availability, data loss, monetary cost, and business
disruption?
4. For each failure mode, determine how the application will respond and recover. Consider tradeoffs in cost
and application complexity.
As a starting point for your FMA process, this article contains a catalog of potential failure modes and their
mitigation steps. The catalog is organized by technology or Azure service, plus a general category for application-
level design. The catalog is not exhaustive, but covers many of the core Azure services.

App Service
App Service app shuts down.
Detection . Possible causes:
Expected shutdown
An operator shuts down the application; for example, using the Azure portal.
The app was unloaded because it was idle. (Only if the Always On setting is disabled.)
Unexpected shutdown
The app crashes.
An App Service VM instance becomes unavailable.
Application_End logging will catch the app domain shutdown (soft process crash) and is the only way to catch the
application domain shutdowns.
Recover y:
If the shutdown was expected, use the application's shutdown event to shut down gracefully. For example, in
ASP.NET, use the Application_End method.
If the application was unloaded while idle, it is automatically restarted on the next request. However, you will
incur the "cold start" cost.
 To prevent the application from being unloaded while idle, enable the Always On setting in the web app. See
Configure web apps in Azure App Service.
To prevent an operator from shutting down the app, set a resource lock with ReadOnly level. See Lock resources
with Azure Resource Manager.
If the app crashes or an App Service VM becomes unavailable, App Service automatically restarts the app.
Diagnostics . Application logs and web server logs. See Enable diagnostics logging for web apps in Azure App
Service.
A particular user repeatedly makes bad requests or overloads the system.
Detection . Authenticate users and include user ID in application logs.
Recover y:
Use Azure API Management to throttle requests from the user. See Advanced request throttling with Azure API
Management
Block the user.
Diagnostics . Log all authentication requests.
A bad update was deployed.
Detection . Monitor the application health through the Azure portal (see Monitor Azure web app performance) or
implement the health endpoint monitoring pattern.
Recover y:. Use multiple deployment slots and roll back to the last-known-good deployment. For more
information, see Basic web application.

Azure Active Directory

OpenID Connect authentication fails.
Detection . Possible failure modes include:
1. Azure AD is not available, or cannot be reached due to a network problem. Redirection to the authentication
endpoint fails, and the OpenID Connect middleware throws an exception.
2. Azure AD tenant does not exist. Redirection to the authentication endpoint returns an HTTP error code, and the
OpenID Connect middleware throws an exception.
3. User cannot authenticate. No detection strategy is necessary; Azure AD handles login failures.
Recover y:
1. Catch unhandled exceptions from the middleware.
2. Handle AuthenticationFailed events.
3. Redirect the user to an error page.
4. User retries.

Azure Search
Writing data to Azure Search fails.
Detection . Catch Microsoft.Rest.Azure.CloudException errors.
Recover y:
The Search .NET SDK automatically retries after transient failures. Any exceptions thrown by the client SDK should
be treated as nontransient errors.
The default retry policy uses exponential back-off. To use a different retry policy, call SetRetryPolicy on the
SearchIndexClient or SearchServiceClient class. For more information, see Automatic Retries.
Diagnostics . Use Search Traffic Analytics.
Reading data from Azure Search fails.
Detection . Catch Microsoft.Rest.Azure.CloudException errors.
Recover y:
The Search .NET SDK automatically retries after transient failures. Any exceptions thrown by the client SDK should
be treated as nontransient errors.
The default retry policy uses exponential back-off. To use a different retry policy, call SetRetryPolicy on the
SearchIndexClient or SearchServiceClient class. For more information, see Automatic Retries.

Diagnostics . Use Search Traffic Analytics.

Cassandra
Reading or writing to a node fails.
Detection . Catch the exception. For .NET clients, this will typically be System.Web.HttpException . Other client may
have other exception types. For more information, see Cassandra error handling done right.
Recover y:
Each Cassandra client has its own retry policies and capabilities. For more information, see Cassandra error
handling done right.
Use a rack-aware deployment, with data nodes distributed across the fault domains.
Deploy to multiple regions with local quorum consistency. If a nontransient failure occurs, fail over to another
region.
Diagnostics . Application logs

Cloud Service
Web or worker roles are unexpectedly being shut down.
Detection . The RoleEnvironment.Stopping event is fired.
Recover y . Override the RoleEntryPoint.OnStop method to gracefully clean up. For more information, see The Right
Way to Handle Azure OnStop Events (blog).

Cosmos DB
Reading data fails.
Detection . Catch System.Net.Http.HttpRequestException or Microsoft.Azure.Documents.DocumentClientException .
Recover y:
The SDK automatically retries failed attempts. To set the number of retries and the maximum wait time,
configure ConnectionPolicy.RetryOptions . Exceptions that the client raises are either beyond the retry policy or
are not transient errors.
If Cosmos DB throttles the client, it returns an HTTP 429 error. Check the status code in the
DocumentClientException . If you are getting error 429 consistently, consider increasing the throughput value of
the collection.
If you are using the MongoDB API, the service returns error code 16500 when throttling.
Replicate the Cosmos DB database across two or more regions. All replicas are readable. Using the client SDKs,
 specify the PreferredLocations parameter. This is an ordered list of Azure regions. All reads will be sent to the
first available region in the list. If the request fails, the client will try the other regions in the list, in order. For
more information, see How to set up Azure Cosmos DB global distribution using the SQL API.
Diagnostics . Log all errors on the client side.
Writing data fails.
Detection . Catch System.Net.Http.HttpRequestException or Microsoft.Azure.Documents.DocumentClientException .
Recover y:
The SDK automatically retries failed attempts. To set the number of retries and the maximum wait time,
configure ConnectionPolicy.RetryOptions . Exceptions that the client raises are either beyond the retry policy or
are not transient errors.
If Cosmos DB throttles the client, it returns an HTTP 429 error. Check the status code in the
DocumentClientException . If you are getting error 429 consistently, consider increasing the throughput value of
the collection.
Replicate the Cosmos DB database across two or more regions. If the primary region fails, another region will
be promoted to write. You can also trigger a failover manually. The SDK does automatic discovery and routing,
so application code continues to work after a failover. During the failover period (typically minutes), write
operations will have higher latency, as the SDK finds the new write region. For more information, see How to set
up Azure Cosmos DB global distribution using the SQL API.
As a fallback, persist the document to a backup queue, and process the queue later.
Diagnostics . Log all errors on the client side.

Queue storage
Writing a message to Azure Queue storage fails consistently.
Detection . After N retry attempts, the write operation still fails.
Recover y:
Store the data in a local cache, and forward the writes to storage later, when the service becomes available.
Create a secondary queue, and write to that queue if the primary queue is unavailable.
Diagnostics . Use storage metrics.
The application cannot process a particular message from the queue.
Detection . Application specific. For example, the message contains invalid data, or the business logic fails for some
reason.
Recover y:
Move the message to a separate queue. Run a separate process to examine the messages in that queue.
Consider using Azure Service Bus Messaging queues, which provides a dead-letter queue functionality for this
purpose.

NOTE
If you are using Storage queues with WebJobs, the WebJobs SDK provides built-in poison message handling. See How to use
Azure queue storage with the WebJobs SDK.

Diagnostics . Use application logging.

Azure Cache for Redis
Reading from the cache fails.
Detection . Catch StackExchange.Redis.RedisConnectionException .
Recover y:
1. Retry on transient failures. Azure Cache for Redis supports built-in retry. For more information, see Azure Cache
for Redis retry guidelines.
2. Treat nontransient failures as a cache miss, and fall back to the original data source.
Diagnostics . Use Azure Cache for Redis diagnostics.
Writing to the cache fails.
Detection . Catch StackExchange.Redis.RedisConnectionException .
Recover y:
1. Retry on transient failures. Azure Cache for Redis supports built-in retry. For more information, see Azure Cache
for Redis retry guidelines.
2. If the error is nontransient, ignore it and let other transactions write to the cache later.
Diagnostics . Use Azure Cache for Redis diagnostics.

SQL Database
Cannot connect to the database in the primary region.
Detection . Connection fails.
Recover y:
Prerequisite: The database must be configured for active geo-replication. See SQL Database Active Geo-Replication.
For queries, read from a secondary replica.
For inserts and updates, manually fail over to a secondary replica. See Initiate a planned or unplanned failover
for Azure SQL Database.
The replica uses a different connection string, so you will need to update the connection string in your application.
Client runs out of connections in the connection pool.
Detection . Catch System.InvalidOperationException errors.
Recover y:
Retry the operation.
As a mitigation plan, isolate the connection pools for each use case, so that one use case can't dominate all the
connections.
Increase the maximum connection pools.
Diagnostics . Application logs.
Database connection limit is reached.
Detection . Azure SQL Database limits the number of concurrent workers, logins, and sessions. The limits depend
on the service tier. For more information, see Azure SQL Database resource limits.
To detect these errors, catch System.Data.SqlClient.SqlException and check the value of SqlException.Number for
the SQL error code. For a list of relevant error codes, see SQL error codes for SQL Database client applications:
Database connection error and other issues.
Recover y . These errors are considered transient, so retrying may resolve the issue. If you consistently hit these
errors, consider scaling the database.
Diagnostics . - The sys.event_log query returns successful database connections, connection failures, and
deadlocks.
Create an alert rule for failed connections.
Enable SQL Database auditing and check for failed logins.

Service Bus Messaging

Reading a message from a Service Bus queue fails.
Detection . Catch exceptions from the client SDK. The base class for Service Bus exceptions is MessagingException.
If the error is transient, the IsTransient property is true.
For more information, see Service Bus messaging exceptions.
Recover y:
1. Retry on transient failures. See Service Bus retry guidelines.
2. Messages that cannot be delivered to any receiver are placed in a dead-letter queue. Use this queue to see
which messages could not be received. There is no automatic cleanup of the dead-letter queue. Messages
remain there until you explicitly retrieve them. See Overview of Service Bus dead-letter queues.
Writing a message to a Service Bus queue fails.
Detection . Catch exceptions from the client SDK. The base class for Service Bus exceptions is MessagingException.
If the error is transient, the IsTransient property is true.
For more information, see Service Bus messaging exceptions.
Recover y:
1. The Service Bus client automatically retries after transient errors. By default, it uses exponential back-off.
After the maximum retry count or maximum timeout period, the client throws an exception. For more
information, see Service Bus retry guidelines.
2. If the queue quota is exceeded, the client throws QuotaExceededException. The exception message gives
more details. Drain some messages from the queue before retrying, and consider using the Circuit Breaker
pattern to avoid continued retries while the quota is exceeded. Also, make sure the
BrokeredMessage.TimeToLive property is not set too high.
3. Within a region, resiliency can be improved by using partitioned queues or topics. A non-partitioned queue
or topic is assigned to one messaging store. If this messaging store is unavailable, all operations on that
queue or topic will fail. A partitioned queue or topic is partitioned across multiple messaging stores.
4. For additional resiliency, create two Service Bus namespaces in different regions, and replicate the
messages. You can use either active replication or passive replication.
Active replication: The client sends every message to both queues. The receiver listens on both queues.
Tag messages with a unique identifier, so the client can discard duplicate messages.
Passive replication: The client sends the message to one queue. If there is an error, the client falls back to
the other queue. The receiver listens on both queues. This approach reduces the number of duplicate
messages that are sent. However, the receiver must still handle duplicate messages.
For more information, see GeoReplication sample and Best practices for insulating applications against
Service Bus outages and disasters.
Duplicate message.
Detection . Examine the MessageId and DeliveryCount properties of the message.
Recover y:
If possible, design your message processing operations to be idempotent. Otherwise, store message IDs of
messages that are already processed, and check the ID before processing a message.
Enable duplicate detection, by creating the queue with RequiresDuplicateDetection set to true. With this
setting, Service Bus automatically deletes any message that is sent with the same MessageId as a previous
message. Note the following:
This setting prevents duplicate messages from being put into the queue. It doesn't prevent a receiver
from processing the same message more than once.
Duplicate detection has a time window. If a duplicate is sent beyond this window, it won't be detected.
Diagnostics . Log duplicated messages.
The application can't process a particular message from the queue.
Detection . Application specific. For example, the message contains invalid data, or the business logic fails for some
reason.
Recover y:
There are two failure modes to consider.
The receiver detects the failure. In this case, move the message to the dead-letter queue. Later, run a separate
process to examine the messages in the dead-letter queue.
The receiver fails in the middle of processing the message — for example, due to an unhandled exception. To
handle this case, use PeekLock mode. In this mode, if the lock expires, the message becomes available to other
receivers. If the message exceeds the maximum delivery count or the time-to-live, the message is automatically
moved to the dead-letter queue.
For more information, see Overview of Service Bus dead-letter queues.
Diagnostics . Whenever the application moves a message to the dead-letter queue, write an event to the
application logs.

Service Fabric
A request to a service fails.
Detection . The service returns an error.
Recover y:
Locate a proxy again ( ServiceProxy or ActorProxy ) and call the service/actor method again.
Stateful ser vice . Wrap operations on reliable collections in a transaction. If there is an error, the transaction
will be rolled back. The request, if pulled from a queue, will be processed again.
Stateless ser vice . If the service persists data to an external store, all operations need to be idempotent.
Diagnostics . Application log
Service Fabric node is shut down.
Detection . A cancellation token is passed to the service's RunAsync method. Service Fabric cancels the task before
shutting down the node.
Recover y . Use the cancellation token to detect shutdown. When Service Fabric requests cancellation, finish any
work and exit RunAsync as quickly as possible.
Diagnostics . Application logs

Storage
Writing data to Azure Storage fails
Detection . The client receives errors when writing.
Recover y:
1. Retry the operation, to recover from transient failures. The retry policy in the client SDK handles this
automatically.
2. Implement the Circuit Breaker pattern to avoid overwhelming storage.
3. If N retry attempts fail, perform a graceful fallback. For example:
Store the data in a local cache, and forward the writes to storage later, when the service becomes
available.
If the write action was in a transactional scope, compensate the transaction.
Diagnostics . Use storage metrics.
Reading data from Azure Storage fails.
Detection . The client receives errors when reading.
Recover y:
1. Retry the operation, to recover from transient failures. The retry policy in the client SDK handles this
automatically.
2. For RA-GRS storage, if reading from the primary endpoint fails, try reading from the secondary endpoint. The
client SDK can handle this automatically. See Azure Storage replication.
3. If N retry attempts fail, take a fallback action to degrade gracefully. For example, if a product image can't be
retrieved from storage, show a generic placeholder image.
Diagnostics . Use storage metrics.

Virtual machine
Connection to a backend VM fails.
Detection . Network connection errors.
Recover y:
Deploy at least two backend VMs in an availability set, behind a load balancer.
If the connection error is transient, sometimes TCP will successfully retry sending the message.
Implement a retry policy in the application.
For persistent or nontransient errors, implement the Circuit Breaker pattern.
If the calling VM exceeds its network egress limit, the outbound queue will fill up. If the outbound queue is
consistently full, consider scaling out.
Diagnostics . Log events at service boundaries.
VM instance becomes unavailable or unhealthy.
Detection . Configure a Load Balancer health probe that signals whether the VM instance is healthy. The probe
should check whether critical functions are responding correctly.
Recover y . For each application tier, put multiple VM instances into the same availability set, and place a load
balancer in front of the VMs. If the health probe fails, the Load Balancer stops sending new connections to the
unhealthy instance.
Diagnostics . - Use Load Balancer log analytics.
Configure your monitoring system to monitor all of the health monitoring endpoints.
Operator accidentally shuts down a VM.
Detection . N/A
Recover y . Set a resource lock with ReadOnly level. See Lock resources with Azure Resource Manager.
Diagnostics . Use Azure Activity Logs.

WebJobs
Continuous job stops running when the SCM host is idle.
Detection . Pass a cancellation token to the WebJob function. For more information, see Graceful shutdown.
Recover y . Enable the Always On setting in the web app. For more information, see Run Background tasks with
WebJobs.

Application design
Application can't handle a spike in incoming requests.
Detection . Depends on the application. Typical symptoms:
The website starts returning HTTP 5xx error codes.
Dependent services, such as database or storage, start to throttle requests. Look for HTTP errors such as HTTP
429 (Too Many Requests), depending on the service.
HTTP queue length grows.
Recover y:
Scale out to handle increased load.
Mitigate failures to avoid having cascading failures disrupt the entire application. Mitigation strategies
include:
Implement the Throttling pattern to avoid overwhelming backend systems.
Use queue-based load leveling to buffer requests and process them at an appropriate pace.
Prioritize certain clients. For example, if the application has free and paid tiers, throttle customers on the
free tier, but not paid customers. See Priority queue pattern.
Diagnostics . Use App Service diagnostic logging. Use a service such as Azure Log Analytics, Application Insights,
or New Relic to help understand the diagnostic logs.

A sample is available here. It uses Polly for these exceptions:

429 - Throttling
408 - Timeout
401 - Unauthorized
503 or 5xx - Service unavailable
One of the operations in a workflow or distributed transaction fails.
Detection . After N retry attempts, it still fails.
Recover y:
As a mitigation plan, implement the Scheduler Agent Supervisor pattern to manage the entire workflow.
Don't retry on timeouts. There is a low success rate for this error.
Queue work, in order to retry later.
Diagnostics . Log all operations (successful and failed), including compensating actions. Use correlation IDs, so
that you can track all operations within the same transaction.
A call to a remote service fails.
Detection . HTTP error code.
Recover y:
1. Retry on transient failures.
2. If the call fails after N attempts, take a fallback action. (Application specific.)
3. Implement the Circuit Breaker pattern to avoid cascading failures.
Diagnostics . Log all remote call failures.

Next steps
For more information about the FMA process, see Resilience by design for cloud services (PDF download).
 Backup and disaster recover for Azure applications
12/18/2020 • 14 minutes to read • Edit Online

Disaster recovery is the process of restoring application functionality in the wake of a catastrophic loss.
Your tolerance for reduced functionality during a disaster is a business decision that varies from one application to
the next. It might be acceptable for some applications to be unavailable or to be partially available with reduced
functionality or delayed processing for a period of time. For other applications, any reduced functionality is
unacceptable.

Dependent service outage

For each dependent service, you should understand the implications of a service disruption and the way that the
application will respond. Many services include features that support resiliency and availability, so evaluating each
service independently is likely to improve your disaster recovery plan. For example, Azure Event Hubs supports
failing over to the secondary namespace.

Network outage
When parts of the Azure network are inaccessible, you might not be able to access your application or data. In this
situation, we recommend designing the disaster recovery strategy to run most applications with reduced
functionality.
If reducing functionality isn't an option, the remaining options are application downtime or failover to an alternate
region.
In a reduced functionality scenario:
If your application can't access its data because of an Azure network outage, you might be able to run locally
with reduced application functionality by using cached data.
You might be able to store data in an alternate location until connectivity is restored.

Manual responses
Although automation is ideal, some strategies for disaster recovery require manual responses.
Alerts
Monitor your application for warning signs that may require proactive intervention. For example, if Azure SQL
Database or Azure Cosmos DB consistently throttles your application, you might need to increase your database
capacity or optimize your queries. Even though the application might handle the throttling errors transparently,
your telemetry should still raise an alert so that you can follow up.
For service limits and quota thresholds, we recommend configuring alerts on Azure resources metrics and
diagnostics logs. When possible, set up alerts on metrics, which are lower latency than diagnostics logs.
Through Resource Health, Azure provides some built-in health status checks that can help you diagnose Azure
service throttling issues.
Failover
Configure a disaster recovery strategy for each Azure application and its Azure services. Acceptable deployment
strategies to support disaster recovery may vary based on the SLAs required for all components of each
application.
Azure provides different features within many Azure services to allow for manual failover, such as redis cache geo-
replicas, or for automated failover, such as SQL auto-failover groups. For example:
For an application that mainly uses virtual machines, you can use Azure Site Recovery for the web and logic
tiers. For more information, see Azure to Azure disaster recovery architecture. For SQL Server on VMs, use SQL
Server Always On availability groups.
For an application that uses App Service and Azure SQL Database, you can use a smaller tier App Service plan
configured in the secondary region, which autoscales when a failover occurs. Use failover groups for the
database tier.
In either scenario, an Azure Traffic Manager profile provides for the automated failover across regions. Load
balancers or application gateways should be set up in the secondary region to support faster availability on
failover.
Operational readiness testing
Perform an operational readiness test for failover to the secondary region and for failback to the primary region.
Many Azure services support manual failover or test failover for disaster recovery drills. Alternatively, you can
simulate an outage by shutting down or removing Azure services.

Data corruption and restoration

If a data store fails, there might be data inconsistencies when it becomes available again, especially if the data was
replicated. Understanding the recovery time objective (RTO) and recovery point objective (RPO) of replicated data
stores can help you predict the amount of data loss.
To understand whether the cross-regional failover is started manually or by Microsoft, review the Azure service
SLAs. For services with no SLAs for cross-regional failover, Microsoft typically decides when to fail over and
usually prioritizes recovery of data in the primary region. If data in the primary region is deemed unrecoverable,
Microsoft fails over to the secondary region.
Restoring data from backups
Backups protect you from losing a component of the application because of accidental deletion or data corruption.
They preserve a functional version of the component from an earlier time, which you can use to restore it.
Disaster recovery strategies are not a replacement for backups, but regular backups of application data support
some disaster recovery scenarios. Your backup storage choices should be based on your disaster recovery
strategy.
The frequency of running the backup process determines your RPO. For example, if you perform hourly backups
and a disaster occurs two minutes before the backup, you will lose 58 minutes of data. Your disaster recovery plan
should include how you will address lost data.
It's common for data in one data store to reference data in another store. For example, consider a SQL Database
with a column that links to a blob in Azure Storage. If backups don't happen simultaneously, the database might
have a pointer to a blob that wasn't backed up before the failure. The application or the disaster recovery plan
must implement processes to handle this inconsistency after a recovery.

NOTE
In some scenarios, such as that of VMs backed up using Azure Backup, you can restore only from a backup in the same
region. Other Azure services, such as Azure Cache for Redis, provide geo-replicated backups, which you can use to restore
services across regions.

Azure Storage and Azure SQL Database

Azure automatically stores Azure Storage and SQL Database data three times within different fault domains in the
same region. If you use geo-replication, the data is stored three additional times in a different region. However, if
the data is corrupted or deleted in the primary copy (for example, because of user error), the changes replicate to
the other copies.
You have two options for managing potential data corruption or deletion:
Create a custom backup strategy. You can store your backups in Azure or on-premises, depending on your
business requirements and governance regulations.
Use the point-in-time restore option to recover a SQL Database.
Azure Storage recovery
You can develop a custom backup process for Azure Storage or use one of many third-party backup tools.
Azure Storage provides data resiliency through automated replicas, but it doesn't prevent application code or
users from corrupting data. Maintaining data fidelity after application or user error requires more advanced
techniques, such as copying the data to a secondary storage location with an audit log. You have several options:
Block blobs. Create a point-in-time snapshot of each block blob. For each snapshot, you are charged only
for the storage required to store the differences within the blob since the previous snapshot state. The
snapshots are dependent on the original blob, so we recommend copying to another blob or even to
another storage account. This approach ensures that backup data is protected against accidental deletion.
Use AzCopy or Azure PowerShell to copy the blobs to another storage account.
For more information, see Creating a Snapshot of a Blob.
Azure Files. Use share snapshots, AzCopy, or PowerShell to copy your files to another storage account.
Azure Table storage. Use AzCopy to export the table data into another storage account in another region.
SQL Database recovery
To protect your business from data loss, SQL Database automatically performs a combination of full database
backups weekly, differential database backups hourly, and transaction log backups every 5 to 10 minutes. For the
Basic, Standard, and Premium SQL Database tiers, use point-in-time restore to restore a database to an earlier
time. Review the following articles for more information:
Recover an Azure SQL database using automated database backups
Overview of business continuity with Azure SQL Database
Another option is to use active geo-replication for SQL Database, which automatically replicates database changes
to secondary databases in the same or different Azure region. For more information, see Creating and using active
geo-replication.
You can also use a more manual approach for backup and restore:
Use the Database Copy capability to create a backup copy of the database with transactional consistency.
Use the Azure SQL Database Import/Export Service, which supports exporting databases to BACPAC files
(compressed files containing your database schema and associated data) that are stored in Azure Blob storage.
To protect against a region-wide service disruption, copy the BACPAC files to an alternate region.
SQL Server on VMs
For SQL Server running on VMs, you have two options: traditional backups and log shipping.
With traditional backups, you can restore to a specific point in time, but the recovery process is slow. Restoring
traditional backups requires that you start with an initial full backup and then apply any incremental backups.
You can configure a log shipping session to delay the restore of log backups. This provides a window to recover
from errors made on the primary replica.
Azure Database for MySQL and Azure Database for PostgreSQL
In Azure Database for MySQL and Azure Database for PostgreSQL, the database service automatically makes a
backup every five minutes. You can use these automated backups to restore the server and its databases from an
earlier point in time to a new server. For more information, see:
How to back up and restore a server in Azure Database for MySQL by using the Azure portal
How to back up and restore a server in Azure Database for PostgreSQL using the Azure portal
Azure Cosmos DB
Cosmos DB automatically makes a backup at regular intervals. Backups are stored separately in another storage
service and are replicated globally to protect against regional disasters. If you accidentally delete your database or
collection, you can file a support ticket or call Azure support to restore the data from the last automatic backup.
For more information, see Online backup and on-demand restore in Azure Cosmos DB.
Azure Virtual Machines
To protect Azure Virtual Machines from application errors or accidental deletion, use Azure Backup. The created
backups are consistent across multiple VM disks. In addition, the Azure Backup vault can be replicated across
regions to support recovery from a regional loss.

Disaster recovery plan

Start by creating a recovery plan. The plan is considered complete after it has been fully tested. Include the people,
processes, and applications needed to restore functionality within the service-level agreement (SLA) you've
defined for your customers.
Consider the following suggestions when creating and testing your disaster recovery plan:
In your plan, include the process for contacting support and for escalating issues. This information will help to
avoid prolonged downtime as you work out the recovery process for the first time.
Evaluate the business impact of application failures.
Choose a cross-region recovery architecture for mission-critical applications.
Identify a specific owner of the disaster recovery plan, including automation and testing.
Document the process, especially any manual steps.
Automate the process as much as possible.
Establish a backup strategy for all reference and transactional data, and test backup restoration regularly.
Set up alerts for the stack of the Azure services consumed by your application.
Train operations staff to execute the plan.
Perform regular disaster simulations to validate and improve the plan.
If you're using Azure Site Recovery to replicate virtual machines (VMs), create a fully automated recovery plan to
fail over the entire application.

Backup strategy
Many alternative strategies are available for implementing distributed compute across regions. These must be
tailored to the specific business requirements and circumstances of the application. At a high level, the approaches
can be divided into the following categories:
Redeploy on disaster : In this approach, the application is redeployed from scratch at the time of disaster.
This is appropriate for non-critical applications that don’t require a guaranteed recovery time. Redeploy to a
new region
Warm Spare (Active/Passive) : A secondary hosted service is created in an alternate region, and roles are
deployed to guarantee minimal capacity; however, the roles don’t receive production traffic. This approach
is useful for applications that have not been designed to distribute traffic across regions. Basic Web
 Application example, Replicate VM to another region
Hot Spare (Active/Active) : The application is designed to receive production load in multiple regions. The
cloud services in each region might be configured for higher capacity than required for disaster recovery
purposes. Alternatively, the cloud services might scale out as necessary at the time of a disaster and failover.
This approach requires substantial investment in application design, but it has significant benefits. These
include low and guaranteed recovery time, continuous testing of all recovery locations, and efficient usage
of capacity. Multi tier DR example

Resource management
You can distribute compute instances across regions by creating a separate cloud service in each target region,
and then publishing the deployment package to each cloud service. However, distributing traffic across cloud
services in different regions must be implemented by the application developer or with a traffic management
service.
Determining the number of spare role instances to deploy in advance for disaster recovery is an important aspect
of capacity planning. Having a full-scale secondary deployment ensures that capacity is already available when
needed; however, this effectively doubles the cost. A common pattern is to have a small, secondary deployment,
just large enough to run critical services. This small secondary deployment is a good idea, both to reserve capacity,
and for testing the configuration of the secondary environment.

NOTE
The subscription quota is not a capacity guarantee. The quota is simply a credit limit. To guarantee capacity, the required
number of roles must be defined in the service model, and the roles must be deployed.

Failover and failback testing

Test failover and failback to verify that your application's dependent services come back up in a synchronized
manner during disaster recovery. Changes to systems and operations may affect failover and failback functions,
but the impact may not be detected until the main system fails or becomes overloaded. Test failover capabilities
before they are required to compensate for a live problem. Also be sure that dependent services fail over and fail
back in the correct order.
If you are using Azure Site Recovery to replicate VMs, run disaster recovery drills periodically by doing test
failovers to validate your replication strategy. A test failover does not affect the ongoing VM replication or your
production environment. For more information, see Run a disaster recovery drill to Azure.

Validating backups
Regularly verify that your backup data is what you expect by running a script to validate data integrity, schema,
and queries. There's no point having a backup if it's not useful to restore your data sources. Log and report any
inconsistencies so the backup service can be repaired.

Backup Storage
Backups are about protecting data, applications, and systems that are important to the organization. In operations
environments, it's easy to provide backups: pick the workload that needs hyper-availability and back it up.
Operations environments are relatively static – in that, the systems and applications used remain relatively
consistent, with only the data changing daily.

Application archives
It's important to remember, that a DR plan is more than just an ordered restoration from backup and validation
process. Applications may require post-restoration configuration due to site changes, or reinstallation may be
necessary with restored data imported after.

Outage retrospectives
No amount of safeguards or preparation can prevent every possible incident, and sometimes simple human error
can have significant consequences to a development project. You can't avoid it, but you can learn from it and take
steps to minimize the chances of a similar incident in the future. The question is how software organizations can
best go about learning from mistakes with agile postmortems.

Planning for regional failures

Azure is divided physically and logically into units called regions. A region consists of one or more datacenters in
close proximity.
Under rare circumstances, it is possible that facilities in an entire region can become inaccessible, for example due
to network failures. Or facilities can be lost entirely, for example due to a natural disaster. This section explains the
capabilities of Azure for creating applications that are distributed across regions. Such distribution helps to
minimize the possibility that a failure in one region could affect other regions.
Review Recover from loss of an Azure region for guidance on specific Azure services.

Service-specific guidance
The following articles describe disaster recovery for specific Azure services:

SERVIC E A RT IC L E

Azure Database for MySQL Overview of business continuity with Azure Database for
MySQL

Azure Database for PostgreSQL Overview of business continuity with Azure Database for
PostgreSQL

Azure Cloud Services What to do in the event of an Azure service disruption that
impacts Azure Cloud Services

Cosmos DB High availability with Azure Cosmos DB

Azure Key Vault Azure Key Vault availability and redundancy

Azure Storage Disaster recovery and storage account failover (preview) in

Azure Storage

SQL Database Restore an Azure SQL Database or failover to a secondary

region

Virtual Machines What to do in the event of an Azure service disruption

impacts Azure Cloud

Azure Virtual Network Virtual Network – Business Continuity

Next steps
Recover from data corruption or accidental deletion
Recover from a region-wide service disruption
 Using business metrics to design resilient Azure
applications
11/2/2020 • 5 minutes to read • Edit Online

Workload availability targets

Define your own target SLAs for each workload in your solution so you can determine whether the architecture
meets the business requirements.
Consider cost and complexity
Everything else being equal, higher availability is better. But as you strive for more nines, the cost and complexity
grow. An uptime of 99.99% translates to about five minutes of total downtime per month. Is it worth the additional
complexity and cost to reach five nines? The answer depends on the business requirements.
Here are some other considerations when defining an SLA:
To achieve four nines (99.99%), you can't rely on manual intervention to recover from failures. The application
must be self-diagnosing and self-healing.
Beyond four nines, it's challenging to detect outages quickly enough to meet the SLA.
Think about the time window that your SLA is measured against. The smaller the window, the tighter the
tolerances. It doesn't make sense to define your SLA in terms of hourly or daily uptime.
Consider the MTBF and MTTR measurements. The higher your SLA, the less frequently the service can go down
and the quicker the service must recover.
Get agreement from your customers for the availability targets of each piece of your application, and document
it. Otherwise, your design may not meet the customers' expectations.
Identify dependencies
Perform dependency-mapping exercises to identify internal and external dependencies. Examples include
dependencies relating to security or identity, such as Active Directory, or third-party services such as a payment
provider or e-mail messaging service.
Pay particular attention to external dependencies that might be a single point of failure or cause bottlenecks. If a
workload requires 99.99% uptime but depends on a service with a 99.9% SLA, that service can't be a single point of
failure in the system. One remedy is to have a fallback path in case the service fails. Alternatively, take other
measures to recover from a failure in that service.
The following table shows the potential cumulative downtime for various SLA levels.

SL A DO W N T IM E P ER W EEK DO W N T IM E P ER M O N T H DO W N T IM E P ER Y EA R

99% 1.68 hours 7.2 hours 3.65 days

99.9% 10.1 minutes 43.2 minutes 8.76 hours

99.95% 5 minutes 21.6 minutes 4.38 hours

99.99% 1.01 minutes 4.32 minutes 52.56 minutes

99.999% 6 seconds 25.9 seconds 5.26 minutes

Recovery metrics
Derive these values by conducting a risk assessment, and make sure you understand the cost and risk of downtime
and data loss. These are nonfunctional requirements of a system and should be dictated by business requirements.
Recover y time objective (RTO) is the maximum acceptable time an application is unavailable after an
incident.
Recover y point objective (RPO) is the maximum duration of data loss that's acceptable during a disaster.
If the MTTR value of any critical component in a highly available setup exceeds the system RTO, a failure in the
system might cause an unacceptable business disruption. That is, you can't restore the system within the defined
RTO.

Availability metrics
Use these measures to plan for redundancy and determine customer SLAs.
Mean time to recover (MTTR) is the average time it takes to restore a component after a failure.
Mean time between failures (MTBF) is the how long a component can reasonably expect to last between
outages.

Understand service-level agreements

In Azure, the Service Level Agreement describes Microsoft's commitments for uptime and connectivity. If the SLA
for a particular service is 99.9%, you should expect the service to be available 99.9% of the time. Different services
have different SLAs.
The Azure SLA also includes provisions for obtaining a service credit if the SLA is not met, along with specific
definitions of availability for each service. That aspect of the SLA acts as an enforcement policy.

The Service Level Agreement Estimator sample shows how to calculate the SLA of your architecture.
Composite SLAs
Composite SLAs involve multiple services supporting an application, each with differing levels of availability. For
example, consider an App Service web app that writes to Azure SQL Database. At the time of this writing, these
Azure services have the following SLAs:
App Service web apps = 99.95%
SQL Database = 99.99%
What is the maximum downtime you would expect for this application? If either service fails, the whole application
fails. The probability of each service failing is independent, so the composite SLA for this application is 99.95% ×
99.99% = 99.94%. That's lower than the individual SLAs, which isn't surprising because an application that relies on
multiple services has more potential failure points.
You can improve the composite SLA by creating independent fallback paths. For example, if SQL Database is
unavailable, put transactions into a queue to be processed later.
With this design, the application is still available even if it can't connect to the database. However, it fails if the
database and the queue both fail at the same time. The expected percentage of time for a simultaneous failure is
0.0001 × 0.001, so the composite SLA for this combined path is:
Database or queue = 1.0 − (0.0001 × 0.001) = 99.99999%
The total composite SLA is:
Web app and (database or queue) = 99.95% × 99.99999% = ~99.95%
There are tradeoffs to this approach. The application logic is more complex, you are paying for the queue, and you
need to consider data consistency issues.
SLAs for multiregion deployments
SLAs for multiregion deployments involve a high-availability technique to deploy the application in more than one
region and use Azure Traffic Manager to fail over if the application fails in one region.
The composite SLA for a multiregion deployment is calculated as follows:
N is the composite SLA for the application deployed in one region.
R is the number of regions where the application is deployed.
The expected chance that the application fails in all regions at the same time is ((1 − N) ^ R). For example, if the
single-region SLA is 99.95%:
The combined SLA for two regions = (1 − (1 − 0.9995) ^ 2) = 99.999975%
The combined SLA for four regions = (1 − (1 − 0.9995) ^ 4) = 99.999999%
The SLA for Traffic Manager is also a factor. Failing over is not instantaneous in active-passive configurations, which
can result in downtime during a failover. See Traffic Manager endpoint monitoring and failover.
 Chaos engineering
12/18/2020 • 5 minutes to read • Edit Online

Chaos engineering is a methodology that helps developers attain consistent reliability by hardening services
against failures in production. A common way to introduce chaos is to deliberately inject faults that cause system
components to fail. The goal is to observe, monitor, respond to, and improve your system's reliability under adverse
circumstances.

Context
It's difficult to simulate the characteristics of a service's behavior at scale outside a production environment. The
transient nature of cloud platforms can exacerbate this difficulty. Architecting your service to expect failure is a core
approach to creating a modern service. Chaos engineering embraces the uncertainty of the production
environment and strives to anticipate rare, unpredictable, and disruptive outcomes, so that you can minimize any
potential impact on your customers.

Principles
Chaos engineering is aimed at increasing your service’s resiliency and its ability to react to failures. By conducting
experiments in a controlled environment, you can identify issues that are likely to arise during development and
deployment. During this process, be vigilant in adopting the following guidelines:
Be proactive.
Embrace failure.
Break the system.
Identify and address single points of failure early.
Install guardrails and graceful mitigations.
Minimize the blast radius.
Build immunity.
Chaos engineering should be an integral part of development team culture and an ongoing practice, not a short-
term tactical effort in response to a single outage.
Development team members are partners in the process. They must be equipped with the resources to triage
issues, implement the testability that's required for fault injection, and drive the necessary product changes.

When to apply chaos

Ideally, you should apply chaos principles continuously. There's constant change in the environments in which
software and hardware run, so monitoring the changes is key. By constantly applying stress or faults on
components, you can help expose issues early, before small problems are compounded by a number of other
factors.
Apply chaos engineering principles when you're:
Deploying new code.
Adding dependencies.
Observing changes in usage patterns.
Mitigating problems.
Process
Chaos engineering requires specialized expertise, technology, and practices. As with security and performance
teams, the model of a central team supporting the service teams is a common, effective approach.
If you plan to practice the simulated handling of potentially catastrophic scenarios under controlled conditions,
here's a simplified way to organize your teams:

AT TA C K ER DEF EN DER

Inject faults Assess

Provide hints Analyze

Mitigate

Goals
Familiarize team members with monitoring tools.
Recognize outage patterns.
Learn how to assess the impact.
Determine the root cause and mitigate accordingly.
Practice log analysis.
Overall method
1. Start with a hypothesis.
2. Measure baseline behavior.
3. Inject a fault or faults.
4. Monitor the resulting behavior.
5. Document the process and observations.
6. Identify and act on the result.
Periodically validate your process, architecture choices, and code. By conducting fault-injection experiments, you
can confirm that monitoring is in place and alerts are set up, the directly responsible individual (DRI) process is
effective, and your documentation and investigation processes are up to date. Keep in mind a few key
considerations:
Challenge system assumptions.
Validate change (topology, platform, resources).
Use service-level agreement (SLA) buffers.
Use live-site outages as opportunities.

Best practices
Shift left
Shift-left testing means experiment early, experiment often. Incorporate fault-injection configurations and create
resiliency-validation gates during the development stages and in the deployment pipeline.
Shift right
Shift-right testing means that you verify that the service is resilient where it counts in a pre-production or
production environment with actual customer load. Adopt a proactive approach as opposed to reacting to failures.
Be a part of determining and controlling requirements for the blast radius.
Blast radius
Stop the experiment when it goes beyond scope. Unknown results are an expected outcome of chaos experiments.
Strive to achieve balance between collecting substantial result data and affecting as few production users as
possible. For an example of this principle in practice, see the Bulkhead pattern article.
Error budget testing
Establish an error budget as an investment in chaos and fault injection. Your error budget is the difference between
achieving 100% of the service-level objective (SLO) and achieving the agreed-upon SLO.

Considerations
The following sections discuss additional considerations about chaos engineering, based on its application inside
Azure.
Identify faults that are relevant to the development team
Work closely with the development teams to ensure the relevance of the injected failures. Use past incidents or
issues as a guide. Examine dependencies and evaluate the results when those dependencies are removed.
An external team can't hypothesize faults for your team. A study of failures from an artificial source might be
relevant to your team's purposes, but the effort must be justified.
Inject faults in a way that accurately reflects production failures
Simulate production failures. Treat injected faults in the same way that you would treat production-level faults.
Enforcing a tighter limit on the blast radius will enable you to simulate a production environment. Each fault-
injection effort must be accompanied by tooling that's designed to inject the types of faults that are relevant to
your team's scenarios. Here are two basic ways:
Inject faults in a non-production environment, such as Canary or Test In Production (TIP).
Partition the production service or environment.
Halt all faults and roll back the state to its last-known good configuration if the state seems severe.
Build confidence incrementally
Start by hardening the core, and then expand out in layers. At each point, lock in progress with automated
regression tests. Each team should have a long-term strategy based on a progression that makes sense for the
team's circumstances.
By applying the shift left strategy, you can help ensure that any obstacles to developer usage are removed early and
the testing results are actionable.
The process must be very low tax. That is, the process must make it easy for developers to understand what
happened and to fix the issues. The effort must fit easily into their normal workflow, not burden them with one-off
special activities.

Faults
The following table lists faults that you can apply to inject chaos. The list represents commonly injected faults and
isn't intended to be exhaustive.
Resource pressure
CPU
Memory
Physical
Virtual
bad checksum
Hard disk
Capacity
Read
Write
Availability
Data corruption
Read / Write Latency
Network
Layers
Transport (TCP/UDP)
Application layer (HTTP)
Types
Disconnect
Latency
Alter response codes (HTTP)
Packet reorder / loss (TCP/UDP)
# of connections (active / passive)
DOS attack
Filters
Domain / IP / Subnet
URL path
Port / Protocol
DNS Host Name resolution
Process
Stop / Kill
Restart
Stop service
Start
Crash
Hang
Vir tual Machine
Stop
Restart
BSOD
Change date
Re-image
Live Migration
Platform
Quorum loss
Data loss
Move primary node
Remove replica
Functions
Latency
Exceptions
Status codes
Intercept / Denylist calls
Disk capacity
Application specific
Intercept / Re-route calls
No access to service code
Hardware
Machine
Storage
Network devices
Rack
UPS
Datacenter
 Data Management for Reliability
12/18/2020 • 4 minutes to read • Edit Online

Database resiliency
Azure SQL Database
SQL Database automatically performs a combination of full database backups weekly, differential database
backups hourly, and transaction log backups every five - 10 minutes to protect your business from data loss. Use
point-in-time restore to restore a database to an earlier time. For more information, see:
Recover an Azure SQL database using automated database backups
Overview of business continuity with Azure SQL Database
SQL Server on VMs
For SQL Server running on VMs, there are two options: traditional backups and log shipping. Traditional backups
enables you to restore to a specific point in time, but the recovery process is slow. Restoring traditional backups
requires starting with an initial full backup, and then applying any backups taken after that. The second option is to
configure a log shipping session to delay the restore of log backups (for example, by two hours). This provides a
window to recover from errors made on the primary.
Azure Cosmos DB
Azure Cosmos DB automatically takes backups at regular intervals. Backups are stored separately in another
storage service, and those backups are globally replicated for resiliency against regional disasters. If you
accidentally delete your database or collection, you can file a support ticket or call Azure support to restore the
data from the last automatic backup. For more information, see Automatic online backup and restore with Azure
Cosmos DB.
Azure Database for MySQL, Azure Database for PostgreSQL
When using Azure Database for MySQL or Azure Database for PostgreSQL, the database service automatically
makes a backup of the service every five minutes. Using this automatic backup feature you may restore the server
and all its databases into a new server to an earlier point-in-time. For more information, see:
How to back up and restore a server in Azure Database for MySQL by using the Azure portal
How to backup and restore a server in Azure Database for PostgreSQL using the Azure portal

Distribute data geographically

SQL Database
Azure SQL Database provides two types of recovery: geo-restore and active geo-replication.
Geo-restore
Geo-restore is also available with Basic, Standard, and Premium databases. It provides the default recovery option
when the database is unavailable because of an incident in the region where your database is hosted. Similar to
point-in-time restore, geo-restore relies on database backups in geo-redundant Azure storage. It restores from the
geo-replicated backup copy, and therefore is resilient to the storage outages in the primary region. For more
information, see Restore an Azure SQL Database or failover to a secondary.
Active geo-replication
Active geo-replication is available for all database tiers. It’s designed for applications that have more aggressive
recovery requirements than geo-restore can offer. Using active geo-replication, you can create up to four readable
secondaries on servers in different regions. You can initiate failover to any of the secondaries. In addition, active
geo-replication can be used to support the application upgrade or relocation scenarios, as well as load balancing
for read-only workloads. For details, see Configure active geo-replication for Azure SQL Database and initiate
failover. Refer to Designing globally available services using Azure SQL Database and Managing rolling upgrades
of cloud applications by using SQL Database active geo-replication for details on how to design and implement
applications and applications upgrade without downtime.
SQL Server on Azure Virtual Machines
A variety of options are available for recovery and high availability for SQL Server 2012 (and later) running in
Azure Virtual Machines. For more information, see High availability and disaster recovery for SQL Server in Azure
Virtual Machines.
SQL Server Always On Availability Groups across regions
Alternatively, you can use SQL Always On Availability Groups for high availability by creating a single availability
group that includes the SQL Server instances in both regions.
As an example, Multi-region N-tier application reference architecture shows a set of practices for running an N-tier
application in multiple Azure regions to achieve availability and a robust disaster recovery infrastructure. It uses a
SQL Server Always On Availability Group and Azure Traffic Manager.

Storage resiliency
Azure Storage provides data resiliency through automated replicas. However, this does not prevent application
code or users from corrupting data, whether accidentally or maliciously. Maintaining data fidelity in the face of
application or user error requires more advanced techniques, such as copying the data to a secondary storage
location with an audit log.
Block blobs . Create a point-in-time snapshot of each block blob. For more information, see Creating a
Snapshot of a Blob. For each snapshot, you are only charged for the storage required to store the differences
within the blob since the last snapshot state. The snapshots are dependent on the existence of the original
blob they are based on, so a copy operation to another blob or even another storage account is advisable.
This ensures that backup data is properly protected against accidental deletion. You can use AzCopy or Azure
PowerShell to copy the blobs to another storage account.
Files . Use share snapshots, or use AzCopy or PowerShell to copy your files to another storage account.
Tables . Use AzCopy to export the table data into another storage account in another region.

Samples related to storage resiliency are here. The scripts perform these tasks:
ARM template to deploy a storage account and blob container.
Copies file into the blob container.
Creates the blob snapshot.
Creates a share snapshot.
Calls AzCopy for table storage
 Monitoring application health for reliability in Azure
12/18/2020 • 8 minutes to read • Edit Online

Monitoring and diagnostics are crucial for resiliency. If something fails, you need to know that it failed, when it
failed — and why.
Monitoring is not the same as failure detection. For example, your application might detect a transient error and
retry, avoiding downtime. But it should also log the retry operation so that you can monitor the error rate to get an
overall picture of application health.
Think of the monitoring and diagnostics process as a pipeline with four distinct stages: Instrumentation, collection
and storage, analysis and diagnosis, and visualization and alerts.

Early warning system

Monitor your application for warning signs that might require proactive intervention. Tools that assess the overall
health of the application and its dependencies help you to recognize quickly when a system or its components
suddenly become unavailable. Use them to implement an early warning system.
1. Identify the key performance indicators of your application's health, such as transient exceptions and remote call
latency.
2. Set thresholds at levels that identify issues before they become critical and require a recovery response.
3. Send an alert to operations when the threshold value is reached.
Consider Microsoft System Center 2016 or third-party tools to provide monitoring capabilities. Most monitoring
solutions track key performance counters and service availability. Azure resource health provides some built-in
health status checks, which can help diagnose throttling of Azure services.

Remote call statistics

Track and report remote call statistics in real time and provide an easy way to review this information, so the
operations team has an instantaneous view into the health of your application. Summarize remote call metrics,
such as latency, throughput, and errors in the 99 and 95 percentiles.

Long-running workflow failures

Long-running workflows often include multiple steps, each of which should be independent.
Track the progress of long-running processes to minimize the likelihood that the entire workflow will need to be
rolled back or that multiple compensating transactions will need to be executed.

TIP
Monitor and manage the progress of long-running workflows by implementing a pattern such as Scheduler Agent
Supervisor.

Visualization and alerts

Present telemetry data in a format that makes it easy for an operator to notice problems or trends quickly, such as
a dashboard or email alert.
Get a full-stack view of application state by using Azure dashboards to create a consolidated view of monitoring
graphs from Application Insights, Log Analytics, Azure Monitor metrics, and Service Health. And use Azure Monitor
alerts to create notifications on Service Health, resource health, Azure Monitor metrics, logs in Log Analytics, and
Application Insights.
For more information about monitoring and diagnostics, see Monitoring and diagnostics.

Here are some samples about creating and querying alerts:

HealthAlert. A sample about creating resource-level health activity log alerts. The sample uses Azure Resource
Manager to create alerts.
GraphAlertsPsSample. A set of PowerShell commands that queries for alerts generated against your
subscription.

Test Monitoring
Include monitoring systems in your test plan. Automated failover and failback systems depend on the correct
functioning of monitoring and instrumentation. Dashboards to visualize system health and operator alerts also
depend on having accurate monitoring and instrumentation. If these elements fail, miss critical information, or
report inaccurate data, an operator might not realize that the system is unhealthy or failing.

Azure subscription and service limits

Azure subscriptions have limits on certain resource types, such as number of resource groups, cores, and storage
accounts. To ensure that your application doesn't run up against Azure subscription limits, create alerts that poll for
services nearing their limits and quotas.
Address the following subscription limits with alerts.

A reference implementation is available on GitHub.

For information about that implementation, see Event-based cloud automation on Azure.
Individual services
Individual Azure services have consumption limits on storage, throughput, number of connections, requests per
second, and other metrics. Your application will fail if it attempts to use resources beyond these limits, resulting in
service throttling and possible downtime.
Depending on the specific service and your application requirements, you can often stay under these limits by
scaling up (choosing another pricing tier, for example) or scaling out (adding new instances).
Azure storage scalability and performance targets
Azure allows a maximum number of storage accounts per subscription. If your application requires more storage
accounts than are currently available in your subscription, create a new subscription with additional storage
accounts. For more information, see Azure subscription and service limits, quotas, and constraints.
If you exceed Azure storage scalability and performance targets, your application will experience storage throttling.
For more information, see Azure Storage scalability and performance targets.
Scalability targets for virtual machine disks
An Azure infrastructure as a service (IaaS) VM supports attaching a number of data disks, depending on several
factors, including the VM size and the type of storage account. If your application exceeds the scalability targets for
virtual machine disks, provision additional storage accounts and create the virtual machine disks there. For more
information, see Scalability and performance targets for VM disks on Windows.
VM size
If the actual CPU, memory, disk, and I/O of your VMs approach the limits of the VM size, your application may
experience capacity issues. To correct the issues, increase the VM size. VM sizes are described in Sizes for virtual
machines in Azure.
If your workload fluctuates over time, consider using virtual machine scale sets to automatically scale the number
of VM instances. Otherwise, you need to manually increase or decrease the number of VMs. For more information,
see the virtual machine scale sets overview.
Azure SQL Database
If your Azure SQL Database tier isn't adequate to handle your application's Database Transaction Unit (DTU)
requirements, your data use will be throttled. For more information on selecting the correct service plan, see Azure
SQL Database purchasing models.
Sample:
This reference implementation monitors a Cosmos DB database for throttling. Azure Monitor alerts are triggered
when data access requests to Cosmos DB exceed the capacity in Request Units (or RUs). For more context, see the
Reference Architecture (Throttling response scenario)

Instrumentation
It's not practical to monitor your application directly, so instrumentation is key. A large-scale distributed system
might run on dozens of virtual machines (VMs), which are added and removed over time. Likewise, a cloud
application might use a number of data stores and a single user action might span multiple subsystems.
Provide rich instrumentation:
For failures that are likely but have not yet occurred, provide enough data to determine the cause, mitigate the
situation, and ensure that the system remains available.
For failures that have already occurred, the application should return an appropriate error message to the user
but should attempt to continue running, albeit with reduced functionality.
Monitoring systems should capture comprehensive details so that applications can be restored efficiently and, if
necessary, designers and developers can modify the system to prevent the situation from recurring.
The raw data for monitoring can come from a variety of sources, including:
Application logs, such as those produced by the Azure Application Insights service.
Operating system performance metrics collected by Azure monitoring agents.
Azure resources, including metrics collected by Azure Monitor.
Azure Service Health, which offers a dashboard to help you track active events.
Azure AD logs built into the Azure platform.
Most Azure services have metrics and diagnostics that you can configure to analyze and determine the cause of
problems. To learn more, see Monitoring data collected by Azure Monitor.

Collection and storage

Raw instrumentation data can be held in various locations and formats, including:
Application trace logs
IIS logs
Performance counters
These disparate sources are collected, consolidated, and placed in reliable data stores in Azure, such as Application
Insights, Azure Monitor metrics, Service Health, storage accounts, and Azure Log Analytics.
Creating good health probes
The health and performance of an application can degrade over time, and degradation might not be noticeable
until the application fails.
Implement probes or check functions, and run them regularly from outside the application. These checks can be as
simple as measuring response time for the application as a whole, for individual parts of the application, for
specific services that the application uses, or for separate components.
Check functions can run processes to ensure that they produce valid results, measure latency and check availability,
and extract information from the system.

The HealthProbesSample sample shows how to set up health probes. It provides an ARM template to set up
the infrastructure. A load balancer accepts public requests and load balance to a set of virtual machines. The health
probe is set up so that it can check for service's path /Health.

Application logs
Application logs are an important source of diagnostics data. To gain insight when you need it most, follow best
practices for application logging.
Log data in the production environment
Capture robust telemetry data while the application is running in the production environment, so you have
sufficient information to diagnose the cause of issues in the production state.
Log events at service boundaries
Include a correlation ID that flows across service boundaries. If a transaction flows through multiple services and
one of them fails, the correlation ID helps you track requests across your application and pinpoints why the
transaction failed.
Use semantic (structured) logging
With structured logs, it's easier to automate the consumption and analysis of the log data, which is especially
important at cloud scale. Generally, we recommend storing Azure resources metrics and diagnostics data in a Log
Analytics workspace rather than in a storage account. This way, you can use Kusto queries to obtain the data you
want quickly and in a structured format. You can also use Azure Monitor APIs and Azure Log Analytics APIs.
Use asynchronous logging
Synchronous logging operations sometimes block your application code, causing requests to back up as logs are
written. Use asynchronous logging to preserve availability during application logging.
Separate application logging from auditing
Audit records are commonly maintained for compliance or regulatory requirements and must be complete. To
avoid dropped transactions, maintain audit logs separately from diagnostic logs.

Analysis and diagnosis

Analyze data consolidated in these data stores to troubleshoot issues and gain an overall view of application health.
Generally, you can search for and analyze the data in Application Insights and Log Analytics using Kusto queries or
view preconfigured graphs using management solutions. Or use Azure Advisor to view recommendations with a
focus on resiliency and performance.
 Recover from a region-wide service disruption
12/18/2020 • 13 minutes to read • Edit Online

Azure is divided physically and logically into units called regions. A region consists of one or more datacenters in
close proximity.
Under rare circumstances, it is possible that facilities in an entire region can become inaccessible, for example due
to network failures. Or facilities can be lost entirely, for example due to a natural disaster. This section explains the
capabilities of Azure for creating applications that are distributed across regions. Such distribution helps to
minimize the possibility that a failure in one region could affect other regions.

Cloud services
Resource management
You can distribute compute instances across regions by creating a separate cloud service in each target region, and
then publishing the deployment package to each cloud service. However, distributing traffic across cloud services
in different regions must be implemented by the application developer or with a traffic management service.
Determining the number of spare role instances to deploy in advance for disaster recovery is an important aspect
of capacity planning. Having a full-scale secondary deployment ensures that capacity is already available when
needed; however, this effectively doubles the cost. A common pattern is to have a small, secondary deployment,
just large enough to run critical services. This small secondary deployment is a good idea, both to reserve capacity,
and for testing the configuration of the secondary environment.

NOTE
The subscription quota is not a capacity guarantee. The quota is simply a credit limit. To guarantee capacity, the required
number of roles must be defined in the service model, and the roles must be deployed.

Load Balancing
To load balance traffic across regions requires a traffic management solution. Azure provides Azure Traffic
Manager. You can also take advantage of third-party services that provide similar traffic management capabilities.
Strategies
Many alternative strategies are available for implementing distributed compute across regions. These must be
tailored to the specific business requirements and circumstances of the application. At a high level, the approaches
can be divided into the following categories:
Redeploy on disaster : In this approach, the application is redeployed from scratch at the time of disaster.
This is appropriate for non-critical applications that don’t require a guaranteed recovery time.
Warm Spare (Active/Passive) : A secondary hosted service is created in an alternate region, and roles are
deployed to guarantee minimal capacity; however, the roles don’t receive production traffic. This approach
is useful for applications that have not been designed to distribute traffic across regions.
Hot Spare (Active/Active) : The application is designed to receive production load in multiple regions. The
cloud services in each region might be configured for higher capacity than required for disaster recovery
purposes. Alternatively, the cloud services might scale out as necessary at the time of a disaster and fail
over. This approach requires substantial investment in application design, but it has significant benefits.
These include low and guaranteed recovery time, continuous testing of all recovery locations, and efficient
usage of capacity.
A complete discussion of distributed design is outside the scope of this document. For more information, see
Disaster Recovery and High Availability for Azure Applications.

Virtual machines
Recovery of infrastructure as a service (IaaS) virtual machines (VMs) is similar to platform as a service (PaaS)
compute recovery in many respects. There are important differences, however, due to the fact that an IaaS VM
consists of both the VM and the VM disk.
Use Azure Backup to create cross region backups that are application consistent . Azure Backup
enables customers to create application consistent backups across multiple VM disks, and support
replication of backups across regions. You can do this by choosing to geo-replicate the backup vault at the
time of creation. Replication of the backup vault must be configured at the time of creation. It can't be set
later. If a region is lost, Microsoft will make the backups available to customers. Customers will be able to
restore to any of their configured restore points.
Separate the data disk from the operating system disk . An important consideration for IaaS VMs is
that you cannot change the operating system disk without re-creating the VM. This is not a problem if your
recovery strategy is to redeploy after disaster. However, it might be a problem if you are using the Warm
Spare approach to reserve capacity. To implement this properly, you must have the correct operating
system disk deployed to both the primary and secondary locations, and the application data must be stored
on a separate drive. If possible, use a standard operating system configuration that can be provided on both
locations. After a failover, you must then attach the data drive to your existing IaaS VMs in the secondary
DC. Use AzCopy to copy snapshots of the data disk(s) to a remote site.
Be aware of potential consistency issues after a geo-failover of multiple VM Disks . VM Disks are
implemented as Azure Storage blobs, and have the same geo-replication characteristic. Unless Azure
Backup is used, there are no guarantees of consistency across disks, because geo-replication is
asynchronous and replicates independently. Individual VM disks are guaranteed to be in a crash consistent
state after a geo-failover, but not consistent across disks. This could cause problems in some cases (for
example, in the case of disk striping).
Use Azure Site Recover y to replicate applications across regions . With Azure Site Recovery,
customers don't need to worry about separating data disks from operating system disks or about potential
consistency issues. Azure Site Recovery replicates workloads running on physical and virtual machines
from a primary site (either on-premises or in Azure) to a secondary location (in Azure). When an outage
occurs at the customer's primary site, a failover can be triggered to quickly return the customer to an
operational state. After the primary location is restored, customers can then fail back. They can easily
replicate using recovery points with application-consistent snapshots. These snapshots capture disk data, all
in-memory data, and all in-process transactions. Azure Site Recovery allows customers to keep recovery
time objectives (RTO) and recovery point objectives (RPO) within organizational limits. Customers can also
easily run disaster recovery drills without affecting applications in production. Using recovery plans,
customers can sequence the failover and recovery of multitier applications running on multiple VMs. They
can group machines together in a recovery plan, and optionally add scripts and manual actions. Recovery
plans can be integrated with Azure Automation runbooks.

Storage
Recovery by using geo -redundant storage of blob, table, queue, and VM disk storage
In Azure, blobs, tables, queues, and VM disks are all geo-replicated by default. This is referred to as geo-redundant
storage (GRS). GRS replicates storage data to a paired datacenter located hundreds of miles apart within a specific
geographic region. GRS is designed to provide additional durability in case there is a major datacenter disaster.
Microsoft controls when failover occurs, and failover is limited to major disasters in which the original primary
location is deemed unrecoverable in a reasonable amount of time. Under some scenarios, this can be several days.
Data is typically replicated within a few minutes, although synchronization interval is not yet covered by a service
level agreement.
If a geo-failover occurs, there will be no change to how the account is accessed (the URL and account key will not
change). The storage account will, however, be in a different region after failover. This could impact applications
that require regional affinity with their storage account. Even for services and applications that do not require a
storage account in the same datacenter, the cross-datacenter latency and bandwidth charges might be compelling
reasons to move traffic to the failover region temporarily. This could factor into an overall disaster recovery
strategy.
In addition to automatic failover provided by GRS, Azure has introduced a service that gives you read access to the
copy of your data in the secondary storage location. This is called read-access geo-redundant storage (RA-GRS).
For more information about both GRS and RA-GRS storage, see Azure Storage replication.
Geo -replication region mappings
It is important to know where your data is geo-replicated, in order to know where to deploy the other instances of
your data that require regional affinity with your storage. For more information, see Azure Paired Regions.
Geo -replication pricing
Geo-replication is included in current pricing for Azure Storage. This is called geo-redundant storage (GRS). If you
do not want your data geo-replicated, you can disable geo-replication for your account. This is called locally
redundant storage (LRS), and it is charged at a discounted price compared to GRS.
Determining if a geo -failover has occurred
If a geo-failover occurs, this will be posted to the Azure Service Health Dashboard. Applications can implement an
automated means of detecting this, however, by monitoring the geo-region for their storage account. This can be
used to trigger other recovery operations, such as activation of compute resources in the geo-region where their
storage moved to. You can perform a query for this from the service management API, by using Get Storage
Account Properties. The relevant properties are:

<GeoPrimaryRegion>primary-region</GeoPrimaryRegion>
<StatusOfPrimary>[Available|Unavailable]</StatusOfPrimary>
<LastGeoFailoverTime>DateTime</LastGeoFailoverTime>
<GeoSecondaryRegion>secondary-region</GeoSecondaryRegion>
<StatusOfSecondary>[Available|Unavailable]</StatusOfSecondary>

Database
SQL Database
Azure SQL Database provides two types of recovery: geo-restore and active geo-replication.
Geo-restore
Geo-restore is also available with Basic, Standard, and Premium databases. It provides the default recovery option
when the database is unavailable because of an incident in the region where your database is hosted. Similar to
point-in-time restore, geo-restore relies on database backups in geo-redundant Azure storage. It restores from the
geo-replicated backup copy, and therefore is resilient to the storage outages in the primary region. For more
information, see Restore an Azure SQL Database or failover to a secondary.
Active geo-replication
Active geo-replication is available for all database tiers. It’s designed for applications that have more aggressive
recovery requirements than geo-restore can offer. Using active geo-replication, you can create up to four readable
secondaries on servers in different regions. You can initiate failover to any of the secondaries. In addition, active
geo-replication can be used to support the application upgrade or relocation scenarios, as well as load balancing
for read-only workloads. For details, see Configure active geo-replication for Azure SQL Database and initiate
failover. Refer to Designing globally available services using Azure SQL Database and Managing rolling upgrades
of cloud applications by using SQL Database active geo-replication for details on how to design and implement
applications and applications upgrade without downtime.
SQL Server on Azure Virtual Machines
A variety of options are available for recovery and high availability for SQL Server 2012 (and later) running in
Azure Virtual Machines. For more information, see High availability and disaster recovery for SQL Server in Azure
Virtual Machines.

Other Azure platform services

When attempting to run your cloud service in multiple Azure regions, you must consider the implications for each
of your dependencies. In the following sections, the service-specific guidance assumes that you must use the same
Azure service in an alternate Azure datacenter. This involves both configuration and data-replication tasks.

NOTE
In some cases, these steps can help to mitigate a service-specific outage rather than an entire datacenter event. From the
application perspective, a service-specific outage might be just as limiting and would require temporarily migrating the
service to an alternate Azure region.

Service Bus
Azure Service Bus uses a unique namespace that does not span Azure regions. So the first requirement is to set up
the necessary service bus namespaces in the alternate region. However, there are also considerations for the
durability of the queued messages. There are several strategies for replicating messages across Azure regions. For
the details on these replication strategies and other disaster recovery strategies, see Best practices for insulating
applications against Service Bus outages and disasters.
App Service
To migrate an Azure App Service application, such as Web Apps or Mobile Apps, to a secondary Azure region, you
must have a backup of the website available for publishing. If the outage does not involve the entire Azure
datacenter, it might be possible to use FTP to download a recent backup of the site content. Then create a new app
in the alternate region, unless you have previously done this to reserve capacity. Publish the site to the new region,
and make any necessary configuration changes. These changes could include database connection strings or other
region-specific settings. If necessary, add the site’s SSL certificate and change the DNS CNAME record so that the
custom domain name points to the redeployed Azure Web App URL.
HDInsight
The data associated with HDInsight is stored by default in Azure Blob Storage. HDInsight requires that a Hadoop
cluster processing MapReduce jobs must be colocated in the same region as the storage account that contains the
data being analyzed. Provided you use the geo-replication feature available to Azure Storage, you can access your
data in the secondary region where the data was replicated if for some reason the primary region is no longer
available. You can create a new Hadoop cluster in the region where the data has been replicated and continue
processing it.
SQL Reporting
At this time, recovering from the loss of an Azure region requires multiple SQL Reporting instances in different
Azure regions. These SQL Reporting instances should access the same data, and that data should have its own
recovery plan in the event of a disaster. You can also maintain external backup copies of the RDL file for each
report.
Media Services
Azure Media Services has a different recovery approach for encoding and streaming. Typically, streaming is more
critical during a regional outage. To prepare for this, you should have a Media Services account in two different
Azure regions. The encoded content should be located in both regions. During a failure, you can redirect the
streaming traffic to the alternate region. Encoding can be performed in any Azure region. If encoding is time-
sensitive, for example during live event processing, you must be prepared to submit jobs to an alternate
datacenter during failures.
Virtual network
Configuration files provide the quickest way to set up a virtual network in an alternate Azure region. After
configuring the virtual network in the primary Azure region, export the virtual network settings for the current
network to a network configuration file. If an outage occurs in the primary region, restore the virtual network from
the stored configuration file. Then configure other cloud services, virtual machines, or cross-premises settings to
work with the new virtual network.

Checklists for disaster recovery

Cloud Services checklist
1. Review the Cloud Services section of this document.
2. Create a cross-region disaster recovery strategy.
3. Understand trade-offs in reserving capacity in alternate regions.
4. Use traffic routing tools, such as Azure Traffic Manager.
Virtual Machines checklist
1. Review the Virtual Machines section of this document.
2. Use Azure Backup to create application consistent backups across regions.
Storage checklist
1. Review the Storage section of this document.
2. Do not disable geo-replication of storage resources.
3. Understand alternate region for geo-replication if a failover occurs.
4. Create custom backup strategies for user-controlled failover strategies.
SQL Database checklist
1. Review the SQL Database section of this document.
2. Use Geo-restore or geo-replication as appropriate.
SQL Server on Virtual Machines checklist
1. Review the SQL Server on Virtual Machines section of this document.
2. Use cross-region AlwaysOn Availability Groups or database mirroring.
3. Alternately use backup and restore to blob storage.
Service Bus checklist
1. Review the Service Bus section of this document.
2. Configure a Service Bus namespace in an alternate region.
3. Consider custom replication strategies for messages across regions.
App Service checklist
1. Review the App Service section of this document.
2. Maintain site backups outside of the primary region.
3. If outage is partial, attempt to retrieve current site with FTP.
4. Plan to deploy the site to new or existing site in an alternate region.
5. Plan configuration changes for both application and DNS CNAME records.
HDInsight checklist
1. Review the HDInsight section of this document.
2. Create a new Hadoop cluster in the region with replicated data.
SQL Reporting checklist
1. Review the SQL Reporting section of this document.
2. Maintain an alternate SQL Reporting instance in a different region.
3. Maintain a separate plan to replicate the target to that region.
Media Services checklist
1. Review the Media Services section of this document.
2. Create a Media Services account in an alternate region.
3. Encode the same content in both regions to support streaming failover.
4. Submit encoding jobs to an alternate region if a service disruption occurs.
Virtual Network checklist
1. Review the Virtual Network section of this document.
2. Use exported virtual network settings to re-create it in another region.
 Testing Azure applications for resiliency and
availability
11/2/2020 • 2 minutes to read • Edit Online

To test resiliency, you should verify how the end-to-end workload performs under intermittent failure conditions.
Run tests in production using both synthetic and real user data. Test and production are rarely identical, so it's
important to validate your application in production using a blue-green or canary deployment. This way, you're
testing the application under real conditions, so you can be sure that it will function as expected when fully
deployed.
As part of your test plan, include:
Chaos engineering
Automated predeployment testing
Fault injection testing
Peak load testing
Disaster recovery testing
Third-party service testing

Chaos engineering
Harden services against failures in production to attain consistent reliability by adopting chaos engineering.

Simulation testing
Simulation testing involves creating small, real-life situations. Simulations demonstrate the effectiveness of the
solutions in the recovery plan and highlight any issues that weren't adequately addressed.
As you perform simulation testing, follow best practices:
Conduct simulations in a manner that doesn't disrupt actual business but feels like a real situation.
Make sure that simulated scenarios are completely controllable. If the recovery plan seems to be failing, you can
restore the situation back to normal without causing damage.
Inform management about when and how the simulation exercises will be conducted. Your plan should detail
the time frame and the resources affected during the simulation.

Perform fault injection testing

For fault injection testing, check the resiliency of the system during failures, either by triggering actual failures or
by simulating them. Here are some strategies to induce failures:
Shut down virtual machine (VM) instances.
Crash processes.
Expire certificates.
Change access keys.
Shut down the DNS service on domain controllers.
Limit available system resources, such as RAM or number of threads.
Unmount disks.
 Redeploy a VM.
Your test plan should incorporate possible failure points identified during the design phase, in addition to common
failure scenarios:
Test your application in an environment as close to production as possible.
Test failures in combination.
Measure the recovery times, and be sure that your business requirements are met.
Verify that failures don't cascade and are handled in an isolated way.
Best practices for Chaos Engineering, for more information about failure scenarios, see Failure and disaster
recovery for Azure applications.

Test under peak loads

Load testing is crucial for identifying failures that only happen under load, such as the back-end database being
overwhelmed or service throttling. Test for peak load and anticipated increase in peak load, using production data
or synthetic data that is as close to production data as possible. Your goal is to see how the application behaves
under real-world conditions.
 Resiliency checklist for specific Azure services
12/18/2020 • 14 minutes to read • Edit Online

Resiliency is the ability of a system to recover from failures and continue to function. Every technology has its own
particular failure modes, which you must consider when designing and implementing your application. Use this
checklist to review the resiliency considerations for specific Azure services. For more information about designing
resilient applications, see Design reliable Azure applications.

App Service
Use Standard or Premium tier. These tiers support staging slots and automated backups. For more information,
see Azure App Service plans in-depth overview
Avoid scaling up or down. Instead, select a tier and instance size that meet your performance requirements
under typical load, and then scale out the instances to handle changes in traffic volume. Scaling up and down may
trigger an application restart.
Store configuration as app settings. Use app settings to hold configuration settings as app settings. Define the
settings in your Resource Manager templates, or using PowerShell, so that you can apply them as part of an
automated deployment / update process, which is more reliable. For more information, see Configure web apps in
Azure App Service.
Create separate App Ser vice plans for production and test. Don't use slots on your production deployment
for testing. All apps within the same App Service plan share the same VM instances. If you put production and test
deployments in the same plan, it can negatively affect the production deployment. For example, load tests might
degrade the live production site. By putting test deployments into a separate plan, you isolate them from the
production version.
Separate web apps from web APIs. If your solution has both a web front end and a web API, consider
decomposing them into separate App Service apps. This design makes it easier to decompose the solution by
workload. You can run the web app and the API in separate App Service plans, so they can be scaled independently.
If you don't need that level of scalability at first, you can deploy the apps into the same plan, and move them into
separate plans later, if needed.
Avoid using the App Ser vice backup feature to back up Azure SQL databases. Instead, use SQL Database
automated backups. App Service backup exports the database to a SQL BACPAC file, which costs DTUs.
Deploy to a staging slot. Create a deployment slot for staging. Deploy application updates to the staging slot,
and verify the deployment before swapping it into production. This reduces the chance of a bad update in
production. It also ensures that all instances are warmed up before being swapped into production. Many
applications have a significant warmup and cold-start time. For more information, see Set up staging environments
for web apps in Azure App Service.
Create a deployment slot to hold the last-known-good (LKG) deployment. When you deploy an update to
production, move the previous production deployment into the LKG slot. This makes it easier to roll back a bad
deployment. If you discover a problem later, you can quickly revert to the LKG version. For more information, see
Basic web application.
Enable diagnostics logging , including application logging and web server logging. Logging is important for
monitoring and diagnostics. See Enable diagnostics logging for web apps in Azure App Service
Log to blob storage. This makes it easier to collect and analyze the data.
Create a separate storage account for logs. Don't use the same storage account for logs and application data.
This helps to prevent logging from reducing application performance.
Monitor performance. Use a performance monitoring service such as New Relic or Application Insights to
monitor application performance and behavior under load. Performance monitoring gives you real-time insight
into the application. It enables you to diagnose issues and perform root-cause analysis of failures.

Application Gateway
Provision at least two instances. Deploy Application Gateway with at least two instances. A single instance is a
single point of failure. Use two or more instances for redundancy and scalability. In order to qualify for the SLA, you
must provision two or more medium or larger instances.

Cosmos DB
Replicate the database across regions. Cosmos DB allows you to associate any number of Azure regions with
a Cosmos DB database account. A Cosmos DB database can have one write region and multiple read regions. If
there is a failure in the write region, you can read from another replica. The Client SDK handles this automatically.
You can also fail over the write region to another region. For more information, see How to distribute data globally
with Azure Cosmos DB.

Event Hubs
Use checkpoints . An event consumer should write its current position to persistent storage at some predefined
interval. That way, if the consumer experiences a fault (for example, the consumer crashes, or the host fails), then a
new instance can resume reading the stream from the last recorded position. For more information, see Event
consumers.
Handle duplicate messages. If an event consumer fails, message processing is resumed from the last recorded
checkpoint. Any messages that were already processed after the last checkpoint will be processed again. Therefore,
your message processing logic must be idempotent, or the application must be able to deduplicate messages.
Handle exceptions.. An event consumer typically processes a batch of messages in a loop. You should handle
exceptions within this processing loop to avoid losing an entire batch of messages if a single message causes an
exception.
Use a dead-letter queue. If processing a message results in a nontransient failure, put the message onto a dead-
letter queue, so that you can track the status. Depending on the scenario, you might retry the message later, apply a
compensating transaction, or take some other action. Note that Event Hubs does not have any built-in dead-letter
queue functionality. You can use Azure Queue Storage or Service Bus to implement a dead-letter queue, or use
Azure Functions or some other eventing mechanism.
Implement disaster recover y by failing over to a secondar y Event Hubs namespace. For more
information, see Azure Event Hubs Geo-disaster recovery.

Azure Cache for Redis

Configure Geo-replication . Geo-replication provides a mechanism for linking two Premium-tier Azure Cache for
Redis instances. Data written to the primary cache is replicated to a secondary read-only cache. For more
information, see How to configure geo-replication for Azure Cache for Redis
Configure data persistence. Redis persistence allows you to persist data stored in Redis. You can also take
snapshots and back up the data, which you can load in case of a hardware failure. For more information, see How
to configure data persistence for a Premium-tier Azure Cache for Redis
If you are using Azure Cache for Redis as a temporary data cache and not as a persistent store, these
recommendations may not apply.

Search
Provision more than one replica. Use at least two replicas for read high-availability, or three for read-write
high-availability.
Configure indexers for multi-region deployments. If you have a multi-region deployment, consider your
options for continuity in indexing.
If the data source is geo-replicated, you should generally point each indexer of each regional Azure Search
service to its local data source replica. However, that approach is not recommended for large datasets stored
in Azure SQL Database. The reason is that Azure Search cannot perform incremental indexing from
secondary SQL Database replicas, only from primary replicas. Instead, point all indexers to the primary
replica. After a failover, point the Azure Search indexers at the new primary replica.
If the data source is not geo-replicated, point multiple indexers at the same data source, so that Azure Search
services in multiple regions continuously and independently index from the data source. For more
information, see Azure Search performance and optimization considerations.

Service Bus
Use Premium tier for production workloads . Service Bus Premium Messaging provides dedicated and
reserved processing resources, and memory capacity to support predictable performance and throughput.
Premium Messaging tier also gives you access to new features that are available only to premium customers at
first. You can decide the number of messaging units based on expected workloads.
Handle duplicate messages . If a publisher fails immediately after sending a message, or experiences network or
system issues, it may erroneously fail to record that the message was delivered, and may send the same message
to the system twice. Service Bus can handle this issue by enabling duplicate detection. For more information, see
Duplicate detection.
Handle exceptions . Messaging APIs generate exceptions when a user error, configuration error, or other error
occurs. The client code (senders and receivers) should handle these exceptions in their code. This is especially
important in batch processing, where exception handling can be used to avoid losing an entire batch of messages.
For more information, see Service Bus messaging exceptions.
Retr y policy . Service Bus allows you to pick the best retry policy for your applications. The default policy is to
allow 9 maximum retry attempts, and wait for 30 seconds but this can be further adjusted. For more information,
see Retry policy – Service Bus.
Use a dead-letter queue . If a message cannot be processed or delivered to any receiver after multiple retries, it
is moved to a dead letter queue. Implement a process to read messages from the dead letter queue, inspect them,
and remediate the problem. Depending on the scenario, you might retry the message as-is, make changes and
retry, or discard the message. For more information, see Overview of Service Bus dead-letter queues.
Use Geo-Disaster Recover y . Geo-disaster recovery ensures that data processing continues to operate in a
different region or datacenter if an entire Azure region or datacenter becomes unavailable due to a disaster. For
more information, see Azure Service Bus Geo-disaster recovery.

Storage
For application data, use read-access geo-redundant storage (RA-GRS). RA-GRS storage replicates the
data to a secondary region, and provides read-only access from the secondary region. If there is a storage outage
in the primary region, the application can read the data from the secondary region. For more information, see
Azure Storage replication.
For VM disks, use managed disks. Managed disks provide better reliability for VMs in an availability set,
because the disks are sufficiently isolated from each other to avoid single points of failure. Also, managed disks
aren't subject to the IOPS limits of VHDs created in a storage account. For more information, see Manage the
availability of Windows virtual machines in Azure.
For Queue storage, create a backup queue in another region. For Queue storage, a read-only replica has
limited use, because you can't queue or dequeue items. Instead, create a backup queue in a storage account in
another region. If there is a storage outage, the application can use the backup queue, until the primary region
becomes available again. That way, the application can still process new requests.

SQL Database
Use Standard or Premium tier. These tiers provide a longer point-in-time restore period (35 days). For more
information, see SQL Database options and performance.
Enable SQL Database auditing. Auditing can be used to diagnose malicious attacks or human error. For more
information, see Get started with SQL database auditing.
Use Active Geo-Replication Use Active Geo-Replication to create a readable secondary in a different region. If
your primary database fails, or simply needs to be taken offline, perform a manual failover to the secondary
database. Until you fail over, the secondary database remains read-only. For more information, see SQL Database
Active Geo-Replication.
Use sharding. Consider using sharding to partition the database horizontally. Sharding can provide fault isolation.
For more information, see Scaling out with Azure SQL Database.
Use point-in-time restore to recover from human error. Point-in-time restore returns your database to an
earlier point in time. For more information, see Recover an Azure SQL database using automated database
backups.
Use geo-restore to recover from a ser vice outage. Geo-restore restores a database from a geo-redundant
backup. For more information, see Recover an Azure SQL database using automated database backups.

Azure Synapse Analytics

Do not disable geo-backup. By default, Synapse Analytics takes a full backup of your data every 24 hours for
disaster recovery. It is not recommended to turn this feature off. For more information, see Geo-backups.

SQL Server running in a VM

Replicate the database. Use SQL Server Always On availability groups to replicate the database. Provides high
availability if one SQL Server instance fails. For more information, see Run Windows VMs for an N-tier application
Back up the database . If you are already using Azure Backup to back up your VMs, consider using Azure Backup
for SQL Server workloads using DPM. With this approach, there is one backup administrator role for the
organization and a unified recovery procedure for VMs and SQL Server. Otherwise, use SQL Server Managed
Backup to Microsoft Azure.

Traffic Manager
Perform manual failback . After a Traffic Manager failover, perform manual failback, rather than automatically
failing back. Before failing back, verify that all application subsystems are healthy. Otherwise, you can create a
situation where the application flips back and forth between datacenters. For more information, see Run VMs in
multiple regions for high availability.
Create a health probe endpoint. Create a custom endpoint that reports on the overall health of the application.
This enables Traffic Manager to fail over if any critical path fails, not just the front end. The endpoint should return
an HTTP error code if any critical dependency is unhealthy or unreachable. Don't report errors for non-critical
services, however. Otherwise, the health probe might trigger failover when it's not needed, creating false positives.
For more information, see Traffic Manager endpoint monitoring and failover.

Virtual Machines
Avoid running a production workload on a single VM. A single VM deployment is not resilient to planned or
unplanned maintenance. Instead, put multiple VMs in an availability set or virtual machine scale set, with a load
balancer in front.
Specify an availability set when you provision the VM. Currently, there is no way to add a VM to an
availability set after the VM is provisioned. When you add a new VM to an existing availability set, make sure to
create a NIC for the VM, and add the NIC to the back-end address pool on the load balancer. Otherwise, the load
balancer won't route network traffic to that VM.
Put each application tier into a separate Availability Set. In an N-tier application, don't put VMs from
different tiers into the same availability set. VMs in an availability set are placed across fault domains (FDs) and
update domains (UD). However, to get the redundancy benefit of FDs and UDs, every VM in the availability set must
be able to handle the same client requests.
Replicate VMs using Azure Site Recover y. When you replicate Azure VMs using Site Recovery, all the VM
disks are continuously replicated to the target region asynchronously. The recovery points are created every few
minutes. This gives you a Recovery Point Objective (RPO) in the order of minutes. You can conduct disaster
recovery drills as many times as you want, without affecting the production application or the ongoing replication.
For more information, see Run a disaster recovery drill to Azure.
Choose the right VM size based on performance requirements. When moving an existing workload to
Azure, start with the VM size that's the closest match to your on-premises servers. Then measure the performance
of your actual workload with respect to CPU, memory, and disk IOPS, and adjust the size if needed. This helps to
ensure the application behaves as expected in a cloud environment. Also, if you need multiple NICs, be aware of the
NIC limit for each size.
Use managed disks for VHDs. Managed disks provide better reliability for VMs in an availability set, because
the disks are sufficiently isolated from each other to avoid single points of failure. Also, managed disks aren't
subject to the IOPS limits of VHDs created in a storage account. For more information, see Manage the availability
of Windows virtual machines in Azure.
Install applications on a data disk , not the OS disk . Otherwise, you may reach the disk size limit.
Use Azure Backup to back up VMs. Backups protect against accidental data loss. For more information, see
Protect Azure VMs with a Recovery Services vault.
Enable diagnostic logs. Include basic health metrics, infrastructure logs, and boot diagnostics. Boot diagnostics
can help you diagnose a boot failure if your VM gets into a nonbootable state. For more information, see Overview
of Azure Diagnostic Logs.
Configure Azure Monitor. Collect and analyze monitoring data from Azure virtual machines including the guest
operating system and the workloads that run in it, see Azure Monitor and Quickstart: Azure Monitor.

Virtual Network
To allow or block public IP addresses, add a network security group to the subnet. Block access from
malicious users, or allow access only from users who have privilege to access the application.
Create a custom health probe. Load Balancer Health Probes can test either HTTP or TCP. If a VM runs an HTTP
server, the HTTP probe is a better indicator of health status than a TCP probe. For an HTTP probe, use a custom
endpoint that reports the overall health of the application, including all critical dependencies. For more information,
see Azure Load Balancer overview.
Don't block the health probe. The Load Balancer Health probe is sent from a known IP address, 168.63.129.16.
Don't block traffic to or from this IP in any firewall policies or network security group rules. Blocking the health
probe would cause the load balancer to remove the VM from rotation.
Enable Load Balancer logging. The logs show how many VMs on the back-end are not receiving network traffic
due to failed probe responses. For more information, see Log analytics for Azure Load Balancer.
 Overview of the security pillar
11/2/2020 • 3 minutes to read • Edit Online

Information Security has always been a complex subject, and it evolves quickly with the creative ideas and
implementations of attackers and security researchers. The origin of security vulnerabilities started with
identifying and exploiting common programming errors and unexpected edge cases. However over time, the
attack surface that an attacker may explore and exploit has expanded well beyond that. Attackers now freely exploit
vulnerabilities in system configurations, operational practices, and the social habits of the systems’ users. As
system complexity, connectedness, and the variety of users increase, attackers have more opportunities to identify
unprotected edge cases and to “hack” systems into doing things they were not designed to do.
Security is one of the most important aspects of any architecture. It provides confidentiality, integrity, and
availability assurances against deliberate attacks and abuse of your valuable data and systems. Losing these
assurances can negatively impact your business operations and revenue, as well as your organization’s reputation
in the marketplace. In the following series of articles, we’ll discuss key architectural considerations and principles
for security and how they apply to Azure.
To assess your workload using the tenets found in the Microsoft Azure Well-Architected Framework, see the
Microsoft Azure Well-Architected Review.
These are the topics we cover in the security pillar of the Microsoft Azure Well-Architected Framework

SEC URIT Y TO P IC DESC RIP T IO N

Role of security Security is one of the most important aspects of any

architecture. Security provides confidentiality, integrity, and
availability assurances against deliberate attacks and abuse of
your valuable data and systems.

Security design principles These principles support these three key strategies and
describe a securely architected system hosted on cloud or on-
premises datacenters (or a combination of both).

Types of attacks to resist An architecture built on good security practices should be

resilient to attacks. It should both resist attacks and recover
rapidly from disruption to the security assurances of
confidentiality, integrity, and availability.

Regulatory compliance Governments and other organizations frequently publish

standards to help define good security practices (due
diligence) so that organizations can avoid being negligent in
security.

Reduce organizational risk Much like physical safety, success in information security is
defined more as an ongoing task of applying good security
practices and principles and hygiene rather than a static
absolute state.
SEC URIT Y TO P IC DESC RIP T IO N

Administration Administration is the practice of monitoring, maintaining, and

operating Information Technology (IT) systems to meet
service levels that the business requires. Administration
introduces some of the highest impact security risks because
performing these tasks requires privileged access to a very
broad set of these systems and applications.

Applications and services Applications and the data associated with them ultimately act
as the primary store of business value on a cloud platform.

Governance, risk, and compliance How is the organization’s security going to be monitored,
audited, and reported? What types of risks does the
organization face while trying to protect identifiable
information, Intellectual Property (IP), financial information?
Are there specific industry, government, or regulatory
requirements that dictate or provide recommendation on
criteria that your organization’s security controls must meet?

Identity and access management Identity provides the basis of a large percentage of security
assurances.

Info protection and storage Protecting data at rest is required to maintain confidentiality,
integrity, and availability assurances across all workloads.

Network security and containment Network security has been the traditional linchpin of
enterprise security efforts. However, cloud computing has
increased the requirement for network perimeters to be more
porous and many attackers have mastered the art of attacks
on identity system elements (which nearly always bypass
network controls).

Security Operations Security operations maintain and restores the security

assurances of the system as live adversaries attack it. The
tasks of security operations are described well by the NIST
Cybersecurity Framework functions of Detect, Respond, and
Recover.
 Security design principles
11/2/2020 • 5 minutes to read • Edit Online

These principles support these three key strategies and describe a securely architected system hosted on cloud or
on-premises datacenters (or a combination of both). Application of these principles will dramatically increase the
likelihood your security architecture will maintain assurances of confidentiality, integrity, and availability.
Each recommendation in this document includes a description of why it is recommended which maps to one of
more of these principles:
Align Security Priorities to Mission – Security resources are almost always limited, so prioritize efforts
and assurances by aligning security strategy and technical controls to the business using classification of
data and systems. Security resources should be focused first on people and assets (systems, data, accounts,
etc.) with intrinsic business value and those with administrative privileges over business critical assets.
Build a Comprehensive Strategy – A security strategy should consider investments in culture, processes,
and security controls across all system components. The strategy should also consider security for the full
lifecycle of system components including the supply chain of software, hardware, and services.
Drive Simplicity – Complexity in systems leads to increased human confusion, errors, automation failures,
and difficulty of recovering from an issue. Favor simple and consistent architectures and implementations.
Design for Attackers – Your security design and prioritization should be focused on the way attackers see
your environment, which is often not the way IT and application teams see it. Inform your security design
and test it with penetration testing to simulate one time attacks and red teams to simulate long-term
persistent attack groups. Design your enterprise segmentation strategy and other security controls to
contain attacker lateral movement within your environment. Actively measure and reduce the potential
Attack Surface that attackers target for exploitation for resources within the environment.
Leverage Native Controls – Favor native security controls built into cloud services over external controls
from third parties. Native security controls are maintained and supported by the service provider,
eliminating or reducing effort required to integrate external security tooling and update those integrations
over time.
Use Identity as Primar y Access Control – Access to resources in cloud architectures is primarily
governed by identity-based authentication and authorization for access controls. Your account control
strategy should rely on identity systems for controlling access rather than relying on network controls or
direct use of cryptographic keys.
Accountability – Designate clear ownership of assets and security responsibilities and ensure actions are
traceable for nonrepudiation. You should also ensure entities have been granted the least privilege required
(to a manageable level of granularity).
Embrace Automation - Automation of tasks decreases the chance of human error that can create risk, so
both IT operations and security best practices should be automated as much as possible to reduce human
errors (while ensuring skilled humans govern and audit the automation).
Focus on Information Protection – Intellectual property is frequently one of the biggest repositories of
organizational value and this data should be protected anywhere it goes including cloud services, mobile
devices, workstations, or collaboration platforms (without impeding collaboration that allows for business
value creation). Your security strategy should be built around classifying information and assets to enable
security prioritization, leveraging strong access control and encryption technology, and meeting business
needs like productivity, usability, and flexibility.
Design for Resilience – Your security strategy should assume that controls will fail and design
accordingly. Making your security posture more resilient requires several approaches working together
Balanced Investment – across core functions spanning the full NIST Cybersecurity Framework
lifecycle (identify, protect, detect, respond, and recover) to ensure that attackers who successfully
evade preventive controls lose access from detection, response, and recovery capabilities.
Ongoing maintenance – of security controls and assurances to ensure that they don’t decay over time
with changes to the environment or neglect
Ongoing vigilance – to ensure that anomalies and potential threats that could pose risks to the
organizations are addressed in a timely manner.
Defense in depth – approach includes additional controls in the design to mitigate risk to the
organization in the event a primary security control fails. This design should consider how likely the
primary control is to fail, the potential organizational risk if it does, and the effectiveness of the
additional control (especially in the likely cases that would cause the primary control to fail).
Least Privilege – This is a form of defense in depth to limit the damage that can be done by any one
account. Accounts should be granted the least amount of privileged required to accomplish their
assigned tasks by access permissions and by time. This helps mitigate the damage of an external
attacker who gains access to the account and/or an internal employee that inadvertently or
deliberately (for example, insider attack) compromises security assurances.
Baseline and Benchmark – To ensure your organization considers current thinking from outside sources,
evaluate your strategy and configuration against external references (including compliance requirements).
This helps to validate your approaches, minimize risk of inadvertent oversight, and the risk of punitive fines
from noncompliance.
Drive Continuous Improvement – Systems and existing practices should be regularly evaluated and
improved to ensure they are and remain effective against attackers who continuously improve and the
continuous digital transformation of the enterprise. This should include processes that proactively integrate
learnings from real world attacks, realistic penetration testing and red team activities, and other sources as
available.
Assume Zero Trust – When evaluating access requests, all requesting users, devices, and applications
should be considered untrusted until their integrity can be sufficiently validated. Access requests should be
granted conditionally based on the requestors trust level and the target resource’s sensitivity. Reasonable
attempts should be made to offer means to increase trust validation (for example, request multi-factor
authentication) and remediate known risks (change known-leaked password, remediate malware infection)
to support productivity goals.
Educate and incentivize security – The humans that are designing and operating the cloud workloads
are part of the whole system. It is critical to ensure that these people are educated, informed, and
incentivized to support the security assurance goals of the system. This is particularly important for people
with accounts granted broad administrative privileges.
 Enforce governance to reduce risks
12/18/2020 • 2 minutes to read • Edit Online

As part of overall design, prioritize where to invest the available resources; financial, people, and time. Constraints
on those resources also affect the security implementation across the organization. Set organizational policies for
operations, technologies, and configurations based on internal factors (business requirements, risks, asset
evaluation) and external factors (benchmarks, regulatory standards, threat environment).
Before defining the policies, consider:
How is the organization’s security monitored, audited, and reported? Is there mandatory reporting?
Are the existing security practices are working?
Are there new requirements?
Are there any requirements specific to industry, government, or regulatory requirements?
Designate group(s) (or individual roles) for central functions that affect shared services and applications.
After the policies are set, continuously improve those standards incrementally. Make sure that the security posture
doesn’t degrade over time by having auditing and monitoring compliance. For information about managing
security standards of an organization, see governance, risk, and compliance (GRC).

In this section
A RT IC L E DESC RIP T IO N

Reference model: Segmentation Reference model and strategies of how the functions and
teams can be segmented.

Management groups and permissions Strategies using management groups to manage resources
across multiple subscriptions consistently and efficiently.

Regulatory compliance Guidance on standards published by law, authorities, and

regulators.
 Segmentation strategies
12/18/2020 • 3 minutes to read • Edit Online

Segmentation refers to the isolation of resources from other parts of the organization. It's an effective way of
detecting and containing adversary movements.
An approach to segmentation is network isolation. This approach is not recommended because different technical
teams are not aligned with the overall business use cases and application workloads. An outcome is complexity
especially with on-premises network and leads to broad network firewall exceptions. While network control should
be considered as one of the strategies, it must be part of a unified segmentation strategy.
An effective segmentation strategy will guide all technical teams (IT, security, applications) to consistently isolate
access using networking, applications, identity, and any other access controls. The strategy should aim to:
Minimize operation friction by aligning to business practices and applications.
Contain Risk by adding cost attackers. This is done by:
Isolating sensitive workloads from compromise of other assets
Isolating high exposure systems from being used as a pivot to other systems
Monitor operations that can lead to potential violations of the integrity of the segments (account usage,
unexpected traffic.).
Here are some recommendations for creating a unified strategy.
Ensure alignment of technical teams to a single strategy that is based on assessing business risks.
Establish a modern perimeter based on zero-trust principles focused on identity, device, applications, and other
signals. This will overcome the limitation of network controls to protect new resources and attack types.
Reinforce network controls for legacy applications by exploring microsegmentation strategies.

Reference model
Start with this reference model and adapt it to your organization’s needs. This model shows how functions,
resources, and teams can be segmented.

Example segments
Consider isolating shared and individual resources as shown in the preceding image.
Core services segment
This segment hosts shared services utilized across the organization. These shared services typically include Active
Directory Domain Services, DNS/DHCP, System Management Tools hosted on Azure Infrastructure as a Service
(IaaS) virtual machines.
Individual segments
There are other segments that can contain resources based on some criteria. For instance, resources used by a
workload application can be contained in a separate segment. Another way is segment by lifecycle stages:
development, test, and production. Some resources might intersect, such as applications can use virtual networks
used for lifecycle stages.
Functions and teams
These are the main functions for this reference model. Mapping between central functions, responsibilities, and
teams are described in Team roles and responsibilities.

F UN C T IO N SC O P E RESP O N SIB IL IT Y

Policy management (Core and individual Some or all resources. Monitor and enforce compliance with
segments) external (or internal) regulations,
standards, and security policy assign
appropriate permission to those roles.

Central IT operations (Core) Across all resources. Grant permissions to the central IT
department (often the infrastructure
team) to create, modify, and delete
resources like virtual machines and
storage.

Central networking group (Core and All network resources. Ensure consistency and avoid technical
individual segments) conflicts, assign network resource
responsibilities to a single central
networking organization. These
resources should include virtual
networks, subnets, Network Security
Groups (NSG), and the virtual machines
hosting virtual network appliances.

Resource role permissions (Core) - For most core services, administrative

privileges required are granted through
the application (Active Directory,
DNS/DHCP, System Management Tools).
No additional Azure resource
permissions are required. If your
organizational model requires these
teams to manage their own VMs,
storage, or other Azure resources, you
can assign these permissions to those
roles.

Security operations (Core and individual All resources. Assess risk factors, identify potential
segments) mitigations, and advise organizational
stakeholders who accept the risk.
 F UN C T IO N SC O P E RESP O N SIB IL IT Y

IT operations (individual segments) All resources. Grant permission to create, modify, and
delete resources. The purpose of the
segment (and resulting permissions) will
depend on your organization structure.
Segments with resources
managed by a centralized IT
organization can grant the
central IT department (often the
infrastructure team) permission
to modify these resources.
Segments managed by
independent business units or
functions (such as a Human
Resources IT Team) can grant
those teams permission to all
resources in the segment.
Segments with autonomous
DevOps teams don’t need to
grant permissions across all
resources because the resource
role (below) grants permissions
to application teams. For
emergencies, use the service
admin account (break-glass
account).

Service admin (Core and individual Use the service admin role only for
segments) emergencies (and initial setup if
required). Do not use this role for daily
tasks.

See also
Management groups

Next steps
Start with this reference model and manage resources across multiple subscriptions consistently and efficiently
with management groups.
Management groups
 Security management groups
11/2/2020 • 5 minutes to read • Edit Online

Management groups can manage resources across multiple subscriptions consistently and efficiently. However,
due to its flexibility, your design can become complex and compromise security and operations.
The recommendations in this article are based on an example reference model. Start with this model and adapt it
to your organization’s needs. For more information, see Enterprise segmentation strategy.

Support your segmentation strategy with management groups

It's recommended that you align the top level of management groups into a simple enterprise segmentation
strategy and limit the levels to no more than two.
In the example reference, there are enterprise-wide resources used by all segments, a set of core services that
share services, additional segments for each workload.
The recommended approach is to use these management groups:
Root management group for enterprise-wide resources.
Use the root management group to include identities that have the requirement to apply policies across
every resource. For example, regulatory requirements, such as restrictions related to data sovereignty. This
group is effective in by applying policies, permissions, tags, across all subscriptions.
Management group for each workload segment.
Use a separate management group for teams with limited scope of responsibility. This group is typically
required because of organizational boundaries or regulatory requirements.
Root or segment management group for the core set of services.

Azure role-based access control (RBAC) role assignment

Grant roles the appropriate permissions that start with least privilege and add more based your operational needs.
Provide clear guidance to your technical teams that implement permissions. This clarity makes it easier to detect
and correct that reduces human errors such as overpermissioning.
Assign permissions at management group for the segment rather than the individual subscriptions. This
will drive consistency and ensure application to future subscriptions.
Consider the built-in roles before creating custom roles to grant the appropriate permissions to VMs and
other objects.
Security managers group membership may be appropriate for smaller teams/organizations where
security teams have extensive operational responsibilities.
When assigning permissions for a segment, consider consistency while allowing flexibility to accommodate
several organizational models. These models can range from a single centralized IT group to mostly independent
IT and DevOps teams.
The reference model has several groups assigned for core services and workload segments. Here are some other
considerations when assigning permissions for the common teams across those segments.
Security team's visibility
Grant read-only access to security attributes for all technical environments. Assign security teams with the
Security Readers permission that provides access needed to assess risk factors, identify potential mitigations,
without providing access to the data.
You can assign this permission by using:
Root management group – for teams responsible for assessing and reporting risk on all resources.
Segment management group(s) – for teams with limited scope of responsibility. This group is typically required
because of organizational boundaries or regulatory requirements.

IMPORTANT
Treat security teams as critical impact accounts and apply the same protections as administrators.

Policy management across some or all resources

Assign appropriate permission to roles that monitor and enforce compliance with external (or internal)
regulations, standards, and security policy. The roles and permissions you choose will depend on the
organizational culture and expectations of the policy program.
The reference model assigns Read Contributor permissions to roles related to policy management.
Central IT operations across all resources
Grant permissions to the central IT department (often the infrastructure team) to create, modify, and delete
resources like virtual machines and storage. Contributor or Owner roles are appropriate for this function.
Central networking group across network resources
Assign network resource responsibilities to a single central networking organization. The Network Contributor
role is appropriate for this group.
IT Operations across all resources
Grant permission to create, modify, and delete resources that belong to a segment. The purpose of the segment
(and resulting permissions) will depend on your organization structure.
Segments with resources managed by a centralized IT organization can grant the central IT department (often
the infrastructure team) permission to modify these resources.
Segments managed by independent business units or functions (such as a Human Resources IT Team) can
grant those teams permission to all resources in the segment.
Segments with autonomous DevOps teams don’t need to grant permissions across all resources because the
resource role (below) grants permissions to application teams. For emergencies, use the service admin account
(break-glass account).
Resource role permissions
For most core services, administrative privileges required to manage them are granted through the application
(Active Directory, DNS/DHCP, System Management Tools), so no additional Azure resource permissions are
required. If your organizational model requires these teams to manage their own VMs, storage, or other Azure
resources, you can assign these permissions to those roles.
Workload segments with autonomous DevOps teams will manage the resources associated with each application.
The actual roles and their permissions depend on the application size and complexity, the application team size
and complexity, and the culture of the organization and application team.
Break-glass account
Use the Ser vice Administrator role only for emergencies and initial setup. Do not use this role for daily tasks.
Use root management group with caution
Use the root management group to drive consistency across the enterprise by applying policies, permissions, tags,
across all subscriptions. This group can affect every all resources on Azure and potentially cause downtime or
other negative impacts.
Select enterprise-wide identities that have a clear requirement to be applied across every resources. These
requirements could be for regulatory reasons. Also, select identities that have near-zero negative impact on
operations. For example, policy with audit effect, tag assignment, RBAC permissions assignments that have been
carefully reviewed.
Use a dedicated service principal name (SPN) to execute management group management operations,
subscription management operations, and role assignment. SPN reduces the number of users who have elevated
rights and follows least-privilege guidelines. Assign the User Access Administrator at the root management
group scope (/) to grant the SPN just mentioned access at the root level. After the SPN is granted permissions, the
User Access Administrator role can be safely removed. In this way, only the SPN is part of the User Access
Administrator role.
Assign Contributor permission to the SPN, which allows tenant-level operations. This permission level ensures
that the SPN can be used to deploy and manage resources to any subscription within your organization.
Limit the number of Azure Policy assignments made at the root management group scope (/). This limitation
minimizes debugging inherited policies in lower-level management groups.
Don't create any subscriptions under the root management group. This hierarchy ensures that subscriptions don't
only inherit the small set of Azure policies assigned at the root-level management group, which don't represent a
full set necessary for a workload.

IMPORTANT
Test all enterprise-wide changes on the root management group before applying (policy, tags, RBAC model, and so on). You
can use a test lab. This can be representative lab tenant or lab segment in production tenant. Another option is to use a
production pilot. This can be a segment management group or designated subset in subscription(s) management group.
Validate changes to make sure the requirements have the desired effect.
 Administrative account security
12/18/2020 • 9 minutes to read • Edit Online

Administration is the practice of monitoring, maintaining, and operating Information Technology (IT) systems to
meet service levels that the business requires. Administration introduces some of the highest impact security risks
because performing these tasks requires privileged access to a very broad set of these systems and applications.
Attackers know that gaining access to an account with administrative privileges can get them access to most or all
of the data they would target, making the security of administration one of the most critical security areas.
As an example, Microsoft makes significant investments in protection and training of administrators for our cloud
systems and IT systems:

Microsoft's recommended core strategy for administrative privileges is to use the available controls to reduce risk
Reduce risk exposure (scope and time) – The principle of least privilege is best accomplished with modern
controls that provide privileges on demand. This help to limit risk by limiting administrative privileges exposure by:
Scope – Just Enough Access (JEA) provides only the required privileges for the administrative operation
required (vs. having direct and immediate privileges to many or all systems at a time, which is almost never
required).
Time – Just in Time (JIT) approaches provided the required privileged as they are needed.
Mitigate the remaining risks – Use a combination of preventive and detective controls to reduce risks
such as isolating administrator accounts from the most common risks phishing and general web browsing,
simplifying and optimizing their workflow, increasing assurance of authentication decisions, and identifying
anomalies from normal baseline behavior that can be blocked or investigated.
Microsoft has captured and documented best practices for protecting administrative accounts and published
prioritized roadmaps for protecting privileged access that can be used as references for prioritizing mitigations for
accounts with privileged access.
Securing Privileged Access (SPA) roadmap for administrators of on-premises Active Directory
 Guidance for securing administrators of Azure Active Directory

Minimize number of critical impact admins

Grant the fewest number of accounts to privileges that can have a critical business impact
Each admin account represents potential attack surface that an attacker can target, so minimizing the number of
accounts with that privilege helps limit the overall organizational risk. Experience has taught us that membership of
these privileged groups grows naturally over time as people change roles if membership not actively limited and
managed.
We recommend an approach that reduces this attack surface risk while ensuring business continuity in case
something happens to an administrator:
Assign at least two accounts to the privileged group for business continuity
When two or more accounts are required, provide justification for each member including the original two
Regularly review membership & justification for each group member

Managed accounts for admins

Ensure all critical impact admins in are managed by enterprise directory to follow organizational policy
enforcement.
Consumer accounts such as Microsoft accounts like @Hotmail.com, @live.com, @outlook.com, don't offer sufficient
security visibility and control to ensure the organization's policies and any regulatory requirements are being
followed. Because Azure deployments often start small and informally before growing into enterprise-managed
tenants, some consumer accounts remain as administrative accounts long afterward for example, original Azure
project managers, creating blind spots, and potential risks.

Separate accounts for admins

Ensure all critical impact admins have a separate account for administrative tasks (vs the account they use for email,
web browsing, and other productivity tasks).
Phishing and web browser attacks represent the most common attack vectors to compromise accounts, including
administrative accounts.
Create a separate administrative account for all users that have a role requiring critical privileges. For these
administrative accounts, block productivity tools like Office 365 email (remove license). If possible, block arbitrary
web browsing (with proxy and/or application controls) while allowing exceptions for browsing to the Azure portal
and other sites required for administrative tasks.

No standing access / Just in Time privileges

Avoid providing permanent "standing" access for any critical impact accounts
Permanent privileges increase business risk by increasing the time an attacker can use the account to do damage.
Temporary privileges force attackers targeting an account to either work within the limited times the admin is
already using the account or to initiate privilege elevation (which increases their chance of being detected and
removed from the environment).
Grant privileges required only as required using one of these methods:
Just in Time - Enable Azure AD Privileged Identity Management (PIM) or a third party solution to require
following an approval workflow to obtain privileges for critical impact accounts
 Break glass – For rarely used accounts, follow an emergency access process to gain access to the accounts.
This is preferred for privileges that have little need for regular operational usage like members of global
admin accounts.

Emergency access or 'Break Glass' accounts

Ensure you have a mechanism for obtaining administrative access in case of an emergency
While rare, sometimes extreme circumstances arise where all normal means of administrative access are
unavailable.
We recommend following the instructions at Managing emergency access administrative accounts in Azure AD and
ensure that security operations monitor these accounts carefully.

Admin workstation security

Ensure critical impact admins use a workstation with elevated security protections and monitoring
Attack vectors that use browsing and email like phishing are cheap and common. Isolating critical impact admins
from these risks will significantly lower your risk of a major incident where one of these accounts is compromised
and used to materially damage your business or mission.
Choose level of admin workstation security based on the options available at https://aka.ms/securedworkstation
Highly Secure Productivity Device (Enhanced Security Workstation or Specialized Workstation)
You can start this security journey for critical impact admins by providing them with a higher security
workstation that still allows for general browsing and productivity tasks. Using this as an interim step helps
ease the transition to fully isolated workstations for both the critical impact admins as well as the IT staff
supporting these users and their workstations.
Privileged Access Workstation (Specialized Workstation or Secured Workstation)
These configurations represent the ideal security state for critical impact admins as they heavily restrict
access to phishing, browser, and productivity application attack vectors. These workstations don't allow
general internet browsing, only allow browser access to Azure portal and other administrative sites.

Critical impact admin dependencies – Account/Workstation

Carefully choose the on-premises security dependencies for critical impact accounts and their workstations
To contain the risk from a major incident on-premises spilling over to become a major compromise of cloud assets,
you must eliminate or minimize the means of control that on premises resources have to critical impact accounts in
the cloud. As an example, attackers who compromise the on premises Active Directory can access and compromise
cloud-based assets that rely on those accounts like resources in Azure, Amazon Web Services (AWS), ServiceNow,
and so on. Attackers can also use workstations joined to those on premises domains to gain access to accounts and
services managed from them.
Choose the level of isolation from on premises means of control also known as security dependencies for critical
impact accounts
User Accounts – Choose where to host the critical impact accounts
Native Azure AD Accounts -*Create Native Azure AD Accounts that are not synchronized with on-
premises active directory
Synchronize from On Premises Active Directory
Use existing accounts hosted in the on premises active directory.
 Workstations – Choose how you will manage and secure the workstations used by critical admin accounts:
Native Cloud Management and Security (Recommended): Join workstations to Azure AD &
Manage/Patch them with Intune or other cloud services. Protect and Monitor with Windows Microsoft
Defender ATP or another cloud service not managed by on premises based accounts.
Manage with Existing Systems: Join existing AD domain and use existing management/security.

Passwordless or multi-factor authentication for admins

Require all critical impact admins to use passwordless authentication or multi-factor authentication (MFA).
Attack methods have evolved to the point where passwords alone cannot reliably protect an account. This is well
documented in a Microsoft Ignite Session.
Administrative accounts and all critical accounts should use one of the following methods of authentication. These
capabilities are listed in preference order by highest cost/difficulty to attack (strongest/preferred options) to lowest
cost/difficult to attack:
Passwordless (such as Windows Hello)
https://aka.ms/HelloForBusiness
Passwordless (Authenticator App)
/azure/active-directory/authentication/howto-authentication-phone-sign-in
Multifactor Authentication
/azure/active-directory/authentication/howto-mfa-userstates
Note that SMS Text Message based MFA has become very inexpensive for attackers to bypass, so we recommend
you avoid relying on it. This option is still stronger than passwords alone, but is much weaker than other MFA
options

Enforce conditional access for admins - Zero Trust

Authentication for all admins and other critical impact accounts should include measurement and enforcement of
key security attributes to support a Zero Trust strategy.
Attackers compromising Azure Admin accounts can cause significant harm. Conditional Access can significantly
reduce that risk by enforcing security hygiene before allowing access to Azure management.
Configure Conditional Access policy for Azure management that meets your organization's risk appetite and
operational needs.
Require Multifactor Authentication and/or connection from designated work network
Require Device integrity with Microsoft Defender ATP (Strong Assurance)

Avoid granular and custom permissions

Avoid permissions that specifically reference individual resources or users
Specific permissions create unneeded complexity and confusion as they don't carry the intention to new similar
resources. This then accumulates into a complex legacy configuration that is difficult to maintain or change without
fear of "breaking something" – negatively impacting both security and solution agility.
Instead of assigning specific resource-specific permissions, use either
Management Groups for enterprise-wide permissions
 Resource groups for permissions within subscriptions
Instead of granting permissions to specific users, assign access to groups in Azure AD. If there isn't an appropriate
group, work with the identity team to create one. This allows you to add and remove group members externally to
Azure and ensure permissions are current, while also allowing the group to be used for other purposes such as
mailing lists.

Use built-in roles

Use built-in roles for assigning permissions where possible.
Customization leads to complexity that increases confusion and makes automation more complex, challenging, and
fragile. These factors all negatively impact security
We recommend that you evaluate the built-in roles designed to cover most normal scenarios. Custom roles are a
powerful and sometimes useful capability, but they should be reserved for cases when built in roles won't work.

Establish lifecycle management for critical impact accounts

Ensure you have a process for disabling or deleting administrative accounts when admin personnel leave the
organization (or leave administrative positions)
See Regularly review critical access for more details.

Attack simulation for critical impact accounts

Regularly simulate attacks against administrative users with current attack techniques to educate and empower
them.
People are a critical part of your defense, especially your personnel with access to critical impact accounts. Ensuring
these users (and ideally all users) have the knowledge and skills to avoid and resist attacks will reduce your overall
organizational risk.
You can use Office 365 Attack Simulation capabilities or any number of third party offerings.
 Azure identity and access management
considerations
12/18/2020 • 2 minutes to read • Edit Online

Most architectures have shared services that are hosted and accessed across networks. Those services share
common infrastructure and users need to access resources and data from anywhere. For such architectures, a
common way to secure resources is to use network controls. However, that isn't enough.
Provide security assurance through identity management: the process of authenticating and authorizing security
principals. Use identity management services to authenticate and grant permission to users, partners, customers,
applications, services, and other entities.

Checklist
How are you managing the identity for your workload?

Define clear lines of responsibility and separation of duties for each function. Restrict access based on a need-
to-know basis and least privilege security principles.
Assign permissions to users, groups, and applications at a certain scope through Azure RBAC. Use built-in roles
when possible.
Prevent deletion or modification of a resource, resource group, or subscription through management locks.
Use Managed Identities to access resources in Azure.
Support a single enterprise directory. Keep the cloud and on-premises directories synchronized, except for
critical-impact accounts.
Set up Azure AD Conditional Access. Enforce and measure key security attributes when authenticating all users,
especially for critical-impact accounts.
Have a separate identity source for non-employees.
Preferably use passwordless methods or opt for modern password methods.
Block legacy protocols and authentication methods.

In this section
Follow these questions to assess the workload at a deeper level. The recommendations in this section are based on
using Azure AD.

A SSESSM EN T DESC RIP T IO N

Does the application team have a clear view on Designate groups (or individual roles) that will be responsible
responsibilities and individual/group access levels? for central functions, such as network, policy management
and so on.

Is the workload infrastructure protected with RBAC Azure Resource Manager handles all control plane requests
(role-based access control)? and applies restrictions that you specify through Azure role-
based access control (Azure RBAC), Azure Policy, locks.

Has role-based and/or resource-based authorization Use a mix of role-based and resource-based authorization.
been configured within Azure AD? Start with the principle of least privilege and add more actions
based your needs.
 A SSESSM EN T DESC RIP T IO N

How is the workload authenticated when Authenticate using Managed Identities, use passwordless
communicating with Azure platform ser vices? protections, and keep all (except critical accounts) identities at
a central location.

Azure security benchmark

The Azure Security Benchmark includes a collection of high-impact security recommendations you can use to help
secure the services you use in Azure:

The questions in this section are aligned to the Azure Security Benchmarks Identity and Access Control.

Azure services for identity

The considerations and best practices in this section are based on these Azure services:
Azure AD
Azure AD B2B
Azure AD B2C

Reference architecture
Here are some reference architectures related to identity and access management:
Integrate on-premises AD domains with Azure AD
Integrate on-premises AD with Azure

Next steps
We recommend applying as many as of the best practices as early as possible, and then working to retrofit any
gaps over time as you mature your security program.
Monitor identity, network, data risks

Related links
Five steps to securing your identity infrastructure

Go back to the main article: Security

 Team roles and responsibilities
12/18/2020 • 2 minutes to read • Edit Online

In an organization, several teams work together to make sure that the workload and the supporting infrastructure
are secure. To avoid confusion that can create security risks, define clear lines of responsibility and separation of
duties.
Do the teams have a clear view on responsibilities and individual/group access levels?

Designate groups (or individual roles) that will be responsible for central functions, such as network, policy
management as shown in this table. Document responsibilities of each group, share contacts to facilitate
communication.

C EN T RA L F UN C T IO N RESP O N SIB IL IT Y

Network security Configuration and maintenance of Azure Firewall, Network

Virtual Appliances (and associated routing), Web Application
Firewall (WAF), Network Security Groups, Application Security
Groups (ASG), and other cross-network traffic.

Network operations Enterprise-wide virtual network and subnet allocation.

IT operations Server endpoint security includes monitoring and remediating

server security. This includes tasks such as patching,
configuration, endpoint security,and so on.

Security operations Incident monitoring and response to investigate and

remediate security incidents in Security Information and Event
Management (SIEM) or source console such as Azure Security
Center Azure AD Identity Protection.

Policy management Apply governance based on risk analysis and compliance

requirements. Set direction for use of Roles Based Access
Control (RBAC), Azure Security Center, Administrator
protection strategy, and Azure Policy to govern Azure
resources.

Identity Security and Standards Set direction for Azure AD directories, PIM/PAM usage, MFA,
password/synchronization configuration, Application Identity
Standards.

Application roles and responsibilities should cover different access level of each operational function. For
example, publish production release, access customer data, manipulate database records, and so on.
Application teams should include central functions listed in the preceding table.

Next
Restrict access to Azure resources based on a need-to-know basis starting with the principle of least privilege
security and add more based your operational needs.
Azure control plane security
Related links
For considerations about using management groups to reflect the organization's structure within an Azure Active
Directory (Azure AD) tenant, see CAF: Management group and subscription organization.

Back to the main article: Azure identity and access management considerations
 Azure control plane security
12/18/2020 • 3 minutes to read • Edit Online

The term control plane refers to the management of resources in your subscription. These activities include
creating, updating, and deleting Azure resources as required by the technical team.
Azure Resource Manager handles all control plane requests and applies restrictions that you specify through Azure
role-based access control (Azure RBAC), Azure Policy, locks. Apply those restrictions are based on the requirement
of the organization.

Key points
Restrict access based on a need-to-know basis and least privilege security principles.
Assign permissions to users, groups, and applications at a certain scope through Azure RBAC.
Use built-in roles when possible.
Prevent deletion or modification of a resource, resource group, or subscription through management locks.
Use less critical control in your CI/CD pipeline for development and test environments.

Roles and permission assignment

Is the workload infrastructure protected with RBAC (role-based access control)?

Role-based access control (RBAC) provides the separation when accessing the resources that an application uses.
Decide who has access to resources at the granular level and what they can do with those resources. For example:
Developers can't access production infrastructure.
Only the SecOps team can read and manage Key Vault secrets.
If there are multiple teams, Project A team can access and manage Resource Group A and all resources within.

Grant roles the appropriate permissions that start with least privilege and add more based on your
operational needs. Provide clear guidance to your technical teams that implement permissions. This clarity
makes it easier to detect and correct that reduces human errors such as overpermissioning.

Azure RBAC helps you manage that separation. You can assign permissions to users, groups, and applications at a
certain scope. The scope of a role assignment can be a subscription, a resource group, or a single resource. For
details, see Azure role-based access control (Azure RBAC).
Assign permissions at management group instead of individual subscriptions to drive consistency and ensure
application to future subscriptions.
Consider the built-in roles before creating custom roles to grant the appropriate permissions to resources and
other objects.
For example, assign security teams with the Security Readers permission that provides access needed to assess
risk factors, identify potential mitigations, without providing access to the data.

IMPORTANT
Treat security teams as critical accounts and apply the same protections as administrators.
Management locks
Are there resource locks applied on critical par ts of the infrastructure?

Unlike role-based access control, management locks are used to apply a restriction across all users and roles.
For critical infrastructure, use management locks to prevent deletion or modification of a resource, resource group,
or subscription. Lock in use cases where only specific roles and users with permissions should be able to
delete/modify resources.
Set locks in the DevOps process carefully because modification locks can sometimes block automation. For
examples of those blocks and considerations, see Considerations before applying locks.

Application deployment
Is there a direct access to the application infrastructure through Azure por tal, Command-line
Interface (CLI), or REST API?

It's recommended that you deploy application infrastructure through automation and CI/CD. To maximize
application autonomy and agility, balance restrictive access control on less critical development and test
environments.
Are CI/CD pipeline roles clearly defined and permissions set?

Azure DevOps offers built-in roles that can be assigned to individual users of groups. For example, using them
properly can make sure that only users responsible for production releases are able to initiate the process and that
only developers can access the source code. Variable groups often contain sensitive configuration information and
can be protected as well.

Next
Grant or deny access to a system by verifying whether the accessor has the permissions to perform the requested
action.
Authentication

Related links
Back to the main article: Azure identity and access management considerations
 Authentication with Azure AD
12/18/2020 • 6 minutes to read • Edit Online

Authentication is a process that grants or denies access to a system by verifying the accessor's identity. Use a
managed identity service for all resources to simplify overall management (such as password policies) and
minimize the risk of oversights or human errors. Azure Active Directory (Azure AD) is the one-stop-shop for
identity and access management service for Azure.

Key points
Use Managed Identities to access resources in Azure.
Keep the cloud and on-premises directories synchronized, except for high-privilege accounts.
Preferably use passwordless methods or opt for modern password methods.
Enable Azure AD conditional access based on key security attributes when authenticating all users, especially for
high-privilege accounts.

Use identity-based authentication

How is the application authenticated when communicating with Azure platform ser vices?

Authenticate with identity services instead of cryptographic keys. On Azure, Managed Identities eliminate the need
to store credentials that might be leaked inadvertently. When Managed Identity is enabled for an Azure resource,
it's assigned an identity that you can use to obtain Azure AD tokens. For more information, see Azure AD-managed
identities for Azure resources.
For example, an Azure Kubernetes Service (AKS) cluster needs to pull images from Azure Container Registry (ACR).
To access the image, the cluster needs to know the ACR credentials. The recommended way is to enable Managed
Identities during cluster configuration. That configuration assigns an identity to the cluster and allows it to obtain
Azure AD tokens.
This approach is secure because Azure handles the management of the underlying credentials for you.
The identity is tied to the lifecycle of the resource, in the example the AKS cluster. When the resource is deleted,
Azure automatically deletes the identity.
Azure AD manages the timely rotation of secrets for you.

TIP
Here are the resources for the preceding example:

GitHub: Azure Kubernetes Service (AKS) Secure Baseline Reference Implementation.

The design considerations are described in Azure Kubernetes Service (AKS) production baseline.

What kind of authentication is required by application APIs?

Don't assume that API URLs used by a workload are hidden and can't get exposed to attackers. For example,
JavaScript code on a website can be viewed. A mobile application can be decompiled and inspected. Even for
internal APIs used only on the backend, a requirement of authentication can increase the difficulty of lateral
movement if an attacker gets network access. Typical mechanisms include API keys, authorization tokens, IP
restrictions.
Managed Identity can help an API be more secure because it replaces the use of human-managed service
principals and can request authorization tokens.
How is user authentication handled in the application?

Don't use custom implementations to manage user credentials. Instead, use Azure AD or other managed identity
providers such as Microsoft account Azure B2C. Managed identity providers provide additional security features
such as modern password protections, MFA (Multi-factor authentication), and resets. In general, passwordless
protections are preferred. Also, modern protocols like OAuth 2.0 use token-based authentication with limited
timespan.
Are authentication tokens cached securely and encr ypted when sharing across web ser vers?

Application code should first try to get OAuth access tokens silently from a cache before attempting to acquire a
token from the identity provider, to optimize performance and maximize availability. Tokens should be stored
securely and handled as any other credentials. When there's a need to share tokens across application servers
(instead of each server acquiring and caching their own) encryption should be used.
For information, see Acquire and cache tokens.

Choose a system with cross-platform support

Use a single identity provider for authentication on all platforms (operating systems, cloud providers, and third-
party services.
Azure AD can be used to authenticate Windows, Linux, Azure, Office 365, other cloud providers, and third-party
services as service providers.
For example, improve the security of Linux virtual machines (VMs) in Azure with Azure AD integration. For details,
see Log in to a Linux virtual machine in Azure using Azure Active Directory authentication.

Centralize all identity systems

Keep your cloud identity synchronized with the existing identity systems to ensure consistency and reduce human
errors.
Consider using Azure AD Connect for synchronizing Azure AD with your existing on-premises directory. For
migration projects, have a requirement to complete this task before an Azure migration and development projects
begin.

IMPORTANT
Don’t synchronize high-privilege accounts to an on-premises directory. If an attacker gets full control of on-premises assets,
they can compromise a cloud account. This strategy will limit the scope of an incident. For more information, see Critical
impact account dependencies.
Synchronization is blocked by default in the default Azure AD Connect configuration. Make sure that you haven’t customized
this configuration. For information about filtering in Azure AD, see Azure AD Connect sync: Configure filtering.

For more information, see hybrid identity providers.

 TIP
Here are the resources for the preceding example::

GitHub: Integrate on-premises Active Directory domains with Azure Active Directory.
The design considerations are described in Integrate on-premises Active Directory domains with Azure AD.

Use passwordless authentication

Remove the use of passwords, when possible. Also, require the same set of credentials to sign in and access the
resources on-premises or in the cloud. This requirement is crucial for accounts that require passwords, such as
admin accounts.
Here are some methods of authentication. The list is ordered by highest cost/difficulty to attack
(strongest/preferred options) to lowest cost/difficult to attack:
Passwordless authentication. Some examples of this method include Windows Hello or Authenticator App.
Multifactor authentication. Although this method is more effective than passwords, we recommend that you
avoid relying on SMS text message-based MFA. For more information, see Enable per-user Azure Multi-Factor
Authentication to secure sign-in events.
Managed Identities. See Use identity-based authentication.
Those methods apply to all users, but should be applied first and strongest to accounts with administrative
privileges.
An implementation of this strategy is enabling single sign-on (SSO) to devices, apps, and services. By signing in
once using a single user account, you can grant access to all the applications and resources as per the business
needs. Users don't have to manage multiple sets of usernames and passwords and you can provision or de-
provision application access automatically. For more information, see Single sign-on.

Use modern password protection

Require modern protections through methods that reduce the use of passwords.
For Azure, enable protections in Azure AD.
1. Configure Azure AD Connect to synchronize password hashes. For information, see Implement password
hash synchronization with Azure AD Connect sync.
2. Choose whether to automatically or manually remediate issues found in a report. For more information, see
Monitor identity risks.
For more information about supporting modern passwords in Azure AD, see these articles.
What is Identity Protection?
Enforce on-premises Azure AD Password Protection for Active Directory Domain Services
Users at risk security report
Risky sign-ins security report

Enable conditional access

Grant access requests based on the requestors' trust level and the target resources' sensitivity.
Are there any conditional access requirements for the application?
Workloads can be exposed over public internet and location-based network controls are not applicable. To enable
conditional access, understand what restrictions are required for the use case. For example, multi-factor
authentication (MFA) is a necessity for remote access; IP-based filtering can be used to enable adhoc debugging
(VPNs are preferred).
Configure Azure AD Conditional Access by setting up Access policy for Azure management based on your
operational needs. For information, see Manage access to Azure management with Conditional Access.
Conditional Access can be an effective in phasing out legacy authentication and associated protocols. The policies
must be enforced for all admins and other critical impact accounts. Start by using metrics and logs to determine
users who still authenticate with old clients. Next, disable any down-level protocols that aren’t used, and set up
conditional access for all users who aren’t using legacy protocols. Finally, give notice and guidance to users about
upgrading before blocking legacy authentication completely. For more information, see Azure AD Conditional
Access support for blocking legacy auth.

Next
Grant or deny access to a system by verifying the accessor's identity.
Authorization

Related links
Back to the main article: Azure identity and access management considerations
 Authorization with Azure AD
12/18/2020 • 4 minutes to read • Edit Online

Authorization is a process that grants or denies access to a system by verifying whether the accessor has the
permissions to perform the requested action. The accessor in this context is the workload (cloud application) or the
user of the workload. The action might be operational or related to resource management. There are two main
approaches to authorization: role-based and resource-based. Both can be configured with Azure AD.

Key points
Use a mix of role-based and resource-based authorization. Start with the principle of least privilege and add
more actions based on your needs.
Define clear lines of responsibility and separation of duties for application roles and the resources it can
manage. Consider the access levels of each operational function, such as permissions needed to publish
production release, access customer data, manipulate database records.
Do not provide permanent access for any critical accounts. Elevate access permissions that are based on
approval and is time bound using Azure AD Privileged Identity Management (Azure AD PIM).|

Role-based authorization
This approach authorizes an action based on the role assigned to a user. For example, some actions require an
administrator role.
A role is a set of permissions. For example, the administrator role has permissions to perform all read, write, and
delete operations. Also, the role has a scope. The scope specifies the management groups, subscriptions, or
resource groups within which the role is allowed to operate.
When assigning a role to a user consider what actions the role can perform and what is the scope of those
operations. Here are some considerations for role assignment:
Use built-in roles before creating custom roles to grant the appropriate permissions to VMs and other
objects. You can assign built-in roles to users, groups, service principals, and managed identities. For more
information, see Azure built-in roles.
If you need to create custom roles, grant roles with the appropriate action. Actions are categorized into
operational and data actions. To avoid overpermissioning, start with actions that have least privilege and
add more based your operational or data access needs. Provide clear guidance to your technical teams that
implement permissions. For more information, see Azure custom roles.
If you have a segmentation strategy, assign permissions with a scope. For example, if you use management
group to support your strategy, set the scope to the group rather than the individual subscriptions. This will
drive consistency and ensure application to future subscriptions. When assigning permissions for a
segment, consider consistency while allowing flexibility to accommodate several organizational models.
These models can range from a single centralized IT group to mostly independent IT and DevOps teams. For
information about assigning scope, see AssignableScopes.
You can use security groups to assign permissions. However, there are disadvantages. It can get complex
because the workload needs to keep track of which security groups correspond to which application roles,
for each tenant. Also, access tokens can grow significantly and Azure AD includes an "overage" claim to limit
the token size. See Microsoft identity platform access tokens.
For more information, see Azure role-based access control (RBAC).
For information about implementing role-based authorization in an ASP.NET application, see Role-based
authorization.

Resource-based authorization
With role-based authorization, a user gets the same level of control on a resource based on the user's role.
However, there might be situations where you need to define access rights per resource. For example, in a resource
group, you want to allow some users to delete the resource; other users cannot. In such situations, use resource-
based authorization that authorizes an action based on a particular resource. Every resource has an Owner. Owner
can delete the resource. Contributors can read and update but can't delete it.

NOTE
The "owner" and "contributor" roles for a resource are not the same as application roles.

You'll need to implement custom logic for resource-based authorization. That logic might be a mapping of
resources, Azure AD object (like role, group, user), and permissions.
For information and code sample about implementing resource-based authorization in an ASP.NET application, see
Resource-based authorization.

Authorization for critical accounts

There might be cases when you need to do activities that require access to important resources. Those resources
might already be accessible to critical accounts such as an administrator account. Or, you might need to elevate the
access permissions until the activities are complete. Both approaches can pose significant risks.
Are there any processes and tools leveraged to manage privileged activities?

Do not provide permanent access for any critical accounts and lower permissions when access is no longer
required. Some strategies include:
Just-in-time privileged access to Azure AD and Azure resources.
Time-bound access.
Approval-based access.
Break glass for emergency access process to gain access.
Limit write access to production systems to service principals. No user accounts should have have regular write-
access.
You can use native and third-party options to elevate access permissions for at least highly privileged if not all
activities. Azure AD Privileged Identity Management (Azure AD PIM) is the recommended native solution on Azure.
For more information about PIM, see What is Azure AD Privileged Identity Management?.

Related links
Back to the main article: Azure identity and access management considerations
 Network security review
12/18/2020 • 4 minutes to read • Edit Online

Assess your workload in areas such as network boundary security, network security, database security, data storage
security, identity management, and operational security.
How do you implement DDoS protection?

Risk : Potential of smaller-scale attack that doesn't trip the platform-level protection.
Azure Virtual Network resources offers Basic and Standard tiers for DDoS protection. Enable DDoS Protection
Standard for all business-critical web application and services.
The Windows N-tier application on Azure with SQL Server reference architecture uses DDoS Protection Standard
because this option:
Uses adaptive tuning, based on the application's network traffic patterns, to detect threats.
Guarantees 100% SLA.
Can be cost effective. For example, during a DDoS attack, the first set of attacks cause the provisioned resources
to scale out. For a resource such as a virtual machine scale set, 10 machines can grow to 100, increasing overall
costs. With Standard protection, you don't have to worry about the cost of the scaled resources because Azure
will provide the cost credit.
For information about Standard DDoS Protection, see Azure DDoS Protection Service.
How do you configure public IPs for which traffic is passed in, and how and where it's translated?

Risk : Inability to provision VMs with private IP addresses for protection.

Use Azure Firewall for built-in high availability and unrestricted cloud scalability.
Utilize Azure IP address to determine which traffic is passed in, and how and where it's translated on to the virtual
network.
How do you isolate network traffic in Azure?

Risk : Inability to ensure VMs and communication between them remains private within a network boundary.
Use Azure Virtual Network to allow VMs to securely communicate with each other, the internet, and on-premises
networks. Otherwise, communication between VMs may not be fully private within a network boundary.
How do you configure traffic flow between multiple application tiers?

Risk : Inability to define different access policies based on the workload types, and to control traffic flows between
them.
You may want to define different access policies based on the workload types and control flow between them. Use
Azure Virtual Network Subnet to allocate separate address spaces for different elements or tiers within the
workload. Then, define different access policies, and control traffic flows between those tiers.
How do you route network traffic through security appliances for security boundar y policy
enforcement, auditing, and inspection?

Risk : Inability to define communication paths between different tiers within a network boundary.
Use Azure Virtual Network User Defined Routes (UDR) to control next hop for traffic between Azure, on-premises,
and Internet resources through virtual appliance, virtual network gateway, virtual network, or internet.
Do you use firewalls, load balancers, and Network Intrusion Detection/Network Intrusion Prevention
(NIDS/NIPS)?

Risk : Possibility of not being able to select comprehensive solutions for secure network boundaries.
Use Network Appliances from Azure Marketplace to deploy a variety of preconfigured network virtual appliances.
Utilize Application Gateway WAF to detect and protect against common web attacks.
How do you segment the larger address space into subnets?

Risk : Inability to allow or deny inbound network traffic to, or outbound network traffic from, within larger network
space.
Use network security groups (NSGs) to allow or deny traffic to and from single IP address, to and from multiple IP
addresses, or even to and from entire subnets.
How do you control routing behavior between VM connectivity?

Risk : Inability to customize the routing configuration.

Employ Azure Virtual Network User Defined Routes (UDR) to customize the routing configuration for deployments.
How do you implement forced tunneling?

Risk : Potential of outbound connections from any VM increasing attack surface area leveraged by attackers.
Utilize forced tunneling to ensure that connections to the internet go through corporate network security devices.
How do you implement enhanced levels of security - firewall, IPS/IDS, antivirus, vulnerability
management, botnet protection on top of network-level controls?

Risk : Inability to enable security for other OSI model layers other than network and transport layer.
Use Azure Marketplace to provision devices for higher levels of network security than with network-level access
controls.
How do you establish cross premises connectivity?

Risk : Potential of access to company’s information assets on-premises.

Use Azure site-to-site VPN or ExpressRoute to set up cross premises connectivity to on-premises networks.
How do you implement global load balancing?

Risk : Inability to make services available even when datacenters might become unavailable.
Utilize Azure Traffic Manager to load balance connections to services based on the location of the user and/or other
criteria
How do you disable RDP/SSH access to vir tual machines?

Risk : Potential for attackers to use brute force techniques to gain access and launch other attacks.
Disable RDP/SSH access to Azure Virtual Machines and use VPN/ExpressRoute to access these virtual machines for
remote management.
How do you prevent, detect, and respond to threats?
Risk : Inability to have a single pane of visibility to prevent, detect, and respond to threats.
Employ Azure Security Center for increased visibility into, and control over, the security of Azure resources,
integrated security monitoring, and policy management across Azure subscriptions.
How do you monitor and diagnose conditions of the network?

Risk : Inability to understand, diagnose, and gain insights to the network in Azure.
Use Network Watcher to understand, diagnose, and gain insights to the network in Azure.
How do you gain access to real time performance information at the packet level?

Risk : Inability to investigate an issue in detail for better diagnoses.

Utilize packet capture to set alerts and gain access to real time performance information at the packet level.
How do you gather data for compliance, auditing, and monitoring the network security profile?

Risk : Inability to build a deeper understanding of the network traffic pattern.

Use network security group flow logs to gather data for compliance, auditing, and monitoring of your network
security profile.
How do you diagnose VPN connectivity issues? Risk : Inability to identify the issue and use the detailed logs
for further investigation.
Use Network Watcher troubleshooter to diagnose most common VPN gateway and connections issues
 Network security
11/2/2020 • 8 minutes to read • Edit Online

Protect assets by placing controls on network traffic originating in Azure, between on-premises and Azure hosted
resources, and traffic to and from Azure. If security measures aren't in place attackers can gain access, for instance,
by scanning across public IP ranges. Proper network security controls can provide defense-in-depth elements that
help detect, contain, and stop attackers who gain entry into your cloud deployments.
Here are some recommendations for network security and containment:
Align network segmentation with overall strategy
Centralize network management and security
Build a network containment strategy
Define an internet edge strategy

Centralize network management and security

Centralize the management and security of core networking functions such as cross-premises links, virtual
networking, subnetting, and IP address schemes. This also applies to security elements such as virtual network
appliances, encryption of cloud virtual network activity and cross-premises traffic, network-based access controls,
and other traditional network security components.
Centralizing can reduce the potential for inconsistent strategies that a potential attacker can misuse. Also, the entire
organization benefits from the centralized network team’s expertise in network management, security knowledge,
and tooling.
Azure Security Center can be used to help centralize the management of network security.

Build a network segmentation strategy

At an organizational level, align your network segmentation strategy with the enterprise segmentation strategy to
have a unified strategy. This approach helps reduce human errors, and increases reliability through automation.
Here's an example of network architecture for a hybrid cloud infrastructure.
Evolve security beyond network controls
In modern architectures, mobile and other devices outside the network access various services for applications
over the internet or on cloud provider networks. Traditional network controls that are based on a trusted intranet
approach can’t effectively provide security assurances for those applications.
Combine network controls with application, identity, and other technical control types. This approach is effective in
preventing, detecting, and responding to threats outside the networks you control.
Ensure that resource grouping and administrative privileges align to the segmentation model.
Design security controls that identify and allow expected traffic, access requests, and application
communication between segments. Monitor the communication between segment. Use data to identify
anomalies, set alerts, or block traffic to mitigate the risk of attackers crossing segmentation boundaries.
For information, see Jericho Forum and ‘Zero Trust’ approaches.

Build a risk containment strategy

Containment of attack vectors within an environment is critical. Here are some considerations for creating a
containment strategy.
Use native capabilities provided by a cloud service provider, dynamic just-in-time (JIT) methods, and integrated
identity and password controls, such as those recommended by zero trust and continuous validation
approaches.
Make sure that there are network access controls between network constructs. These constructs can represent
virtual networks, or subnets within those virtual networks. This works to protect and contain East-West traffic
within your cloud network infrastructure.
Determine whether to use host-based firewalls. If host-based firewalls have been effective in helping you protect
and discover threats in the past, consider using them for your cloud-based resources. Otherwise, explore native
solutions on your cloud service provider’s platform.
Adopt a zero-trust strategy based on user, device, and application identities. Network access controls are based
on elements such as source and destination IP address, protocols, and port numbers. Adding zero trust enforces
and validates access control during “access time”. This strategy avoids the need for complex combinations of
open ports, network routes, available or serviced protocols for several types of changes. Only the destination
 resource needs to provide the necessary access controls.
These technology options can be used to set up network control:
Azure Network Security Groups. Use for basic layer 3 & 4 access controls between Azure Virtual Networks,
their subnets, and the Internet.
Azure Web Application Firewall and the Azure Firewall. Use for more advanced network access controls that
require application layer support.
Local Admin Password Solution (LAPS) or a third-party Privileged Access Management. Set strong local
admin passwords and just in time access to them.
Third parties offer microsegmentation approaches that may enhance network controls by applying zero-trust
principles to networks you control with legacy assets on them.

Define an internet edge Strategy

Decide whether to use native cloud service provider controls or virtual network appliances for internet edge
security.
Internet edge traffic (also known as North-South traffic) represents network connectivity between cloud resources
and the internet. Legacy workloads require protection from internet endpoints because they were built with the
assumption of an internet firewall. An internet edge strategy is intended to mitigate attacks from the internet and
detect or block threats.
There are two primary choices that can provide internet edge security controls and monitoring:
Cloud service provider native controls (Azure Firewall and Web Application Firewall (WAF)).
Partner virtual network appliances (Firewall and WAF Vendors available in Azure Marketplace. Partner
solutions consistently cost more than native controls.
Native controls offer basic security and are insufficient for common attacks, such as the OWASP Top 10. Partner
capabilities can provide advanced features to protect against sophisticated (but typically uncommon) attacks.
Configuration of partner solutions can be complex and more fragile because they don't integrate with cloud’s fabric
controllers. From a cost perspective, native control is cheaper than partner solutions.
The decision to use native as opposed to partner controls should be based on your organization’s experience and
requirements. If the features of the advanced firewall solutions don’t provide sufficient return on investment,
consider using the native capabilities that are easy to configure and scale.

Evaluate the use of legacy network security technology

Carefully plan your use of signature-based Network Intrusion Detection/Network Intrusion Prevention (NIDS/NIPS)
Systems and Network Data Leakage/Loss Prevention (DLP) as you adopt cloud applications services.
IDS/IPS can generate a high number of false positive alerts. A well-tuned IDS/IPS system can be effective for classic
application architectures but don't work well for SaaS and PaaS architectures.
Network-based DLP is ineffective at identifying both inadvertent and deliberate data loss because most modern
protocols and attackers use network-level encryption for inbound and outbound communications. SSL-bridging
can provide an authorized man-in-the-middle entity that terminates and then reestablishes encrypted network
connections. However, this can also introduce privacy, security, and reliability challenges.
Keep reviewing and updating your network security strategy on these considerations as you migrate existing
workloads to Azure:
Cloud service providers filter malformed packets and common network layer attacks.
 Many traditional NIDS/NIPS solutions use signature-based approaches per packet. This can be easily evaded by
attackers and typically produce a high rate of false positives.
Ensure your IDS/IPS system(s) are providing positive value from alerts.
Measure alert quality by the percentage of real attacks detections versus false alarms in the alerts raised
by the system.
Provide high-quality alerts to security analysts. Ideally, alerts should have a 90% true positive rate for
creating incidents in the primary queue to which triage investigation teams must respond. Lower quality
alerts would go to proactive hunting exercises to reduce analyst fatigue and burnout.
Adopt modern zero-trust identity approaches for protecting modern SaaS and PaaS applications. See Zero-Trust
for more information
For IaaS workloads, focus on network security solutions that provide per network context rather than per
packet/session context. Software-defined networks in the cloud are naturally instrumented and can achieve this
much more easily than on-premises equipment.
Favor solutions use machine learning techniques across these large volumes of traffic. ML technology is far
superior to static/manual human analysis at rapidly identifying anomalies that could be attacker activity out of
normal traffic patterns.

Design virtual network subnet security

Design virtual networks and subnets for growth.
Typically, you'll add more network resources as the design matures. This will require refactoring of IP addressing
and subnetting schemes to accommodate the extra resources. There is limited security value in creating many small
subnets and then trying to map network access controls (such as security groups) to each of them.
Plan your subnets based on roles and functions that use same protocols. That way, you can add resources to the
subnet without making changes to security groups that enforce network level access controls.
Do not use all open rules that allow inbound and outbound traffic to and from 0.0.0.0-255.255.255.255. Use a least
privilege approach and only allow relevant protocols. It will decrease your overall network attack surface on the
subnet. All open rules provide a false sense of security because such a rule enforces no security.
The exception is when you want to use security groups only for network logging purposes. We don't recommend
this option. But it's an option if you have another network access control solution in place.
Azure Virtual Network subnets can be designed in this way.

Mitigate DDoS attacks

Enable Distributed Denial of Service (DDoS) mitigations for all business-critical web application and services. DDoS
attacks are common and are easily conducted by unsophisticated attackers. DDoS attacks can be debilitating. An
attack can completely block access to your services and even take down the services. Responsible cloud service
providers offer DDoS protection of services with varying effectiveness and capacity. Mostly they provide two
options:
DDoS protection at the cloud network fabric level – all customers of the cloud service provider benefit from
these protections. The protection usually focuses on the network (layer 3) level.
DDoS protection at higher levels that profile your services – this option provides a baseline for your
deployments and then uses machine learning techniques to detect anomalous traffic and proactively protects
based on protection level set prior to service degradation. Adopt the advance protection for any services where
downtime will negatively impact the business.
For an example of advanced DDoS protection, see Azure DDoS Protection Service.
Decide on an internet ingress/egress policy
Decide whether to route traffic for the internet through on-premises security devices (also called forced tunneling)
or allow internet connectivity through cloud-based network security devices.
For production enterprise, allow cloud resources to start and respond to internet request directly through cloud
network security devices defined by your internet edge strategy. This approach fits the Nth datacenter paradigm,
that is Azure datacenters are a part of your enterprise. It scales better for an enterprise deployment because it
removes hops that add load, latency, and cost.
Forced tunneling is achieved through a cross-premise WAN link. The goal is to provide the network security teams
with greater security and visibility to internet traffic. Even when your resources in the cloud try to respond to
incoming requests from the internet, the responses are force tunneled. Alternately, forced tunneling fits a
datacenter expansion paradigm and can work well for a quick proof of concept, but scales poorly because of the
increased traffic load, latency, and cost. For those reasons, we recommend that you avoid forced tunneling.
 Storage
11/2/2020 • 2 minutes to read • Edit Online

Protecting data at rest is required to maintain confidentiality, integrity, and availability assurances across all
workloads. Storage in a cloud service like Azure is architected and implemented quite differently than on premises
solutions to enable massive scaling, modern access through REST APIs, and isolation between tenants.
Granting access to Azure storage is possible through Azure Active Directory (Azure AD) as well as key based
authentication mechanisms (Symmetric Shared Key Authentication, or Shared Access Signature (SAS)).
Storage in Azure includes a number of native security design attributes:
All data is encrypted by the service.
Data in the storage system cannot be read by a tenant if it has not been written by that tenant (to mitigate
the risk of cross tenant data leakage).
Data will remain only in the region you choose.
The system maintains three synchronous copies of data in the region you choose.
Detailed activity logging is available on an opt-in basis.
Additional security features can be configured such as a storage firewall to provide an additional layer of access
control as well as storage threat protection to detect anomalous access and activities.
 Applications and services
12/18/2020 • 2 minutes to read • Edit Online

Applications and the data associated with them act as the primary store of business value on a cloud platform.
Applications can play a role in risks to the business because:
Business processes are encapsulated and executed by applications and services need to be available and
provided with high integrity.
Business data is stored and processed by application workloads and requires high assurances of
confidentiality, integrity, and availability.

In this section
A SSESSM EN T DESC RIP T IO N

What aspects of the application do you need to Understanding the hosting models and the security
protect? responsibility.

Does the organization identify the highest severity Identify risks to the application and risks it may pose to your
threats to this workload through threat modeling? enterprise through threat modeling.

Do you have any regulator y or governance Guidance on standards published by law, authorities, and
requirements? regulators.

Are you exposing information through exception Consider the way you store secrets and handle exceptions.
handling or HTTP headers? Here are some considerations.

Are the frameworks and libraries used by the Evaluate frameworks and libraries used by the application and
application secure? the resulting vulnerabilities.

Next steps
See these best practices related to PaaS applications.
Securing PaaS deployments
Secure communication paths between applications and the services. Make sure that there's a distinction between
the endpoints exposed to the public internet and private ones. Also, the public endpoints are protected with web
application firewall.
Network security
 Application classification for security
12/18/2020 • 4 minutes to read • Edit Online

Azure can host both legacy and modern applications through Infrastructure as a Service (IaaS) virtual machines
and Platform as a Service (PaaS). With legacy applications, you have the responsibility of securing all dependencies
including OS, middleware, and other components. For PaaS applications, you don't need to manage and secure the
underlying server OS. You are responsible for the application configuration.
This article describes the considerations for understanding the hosting models and the security responsibility of
each, identifying critical applications.

Understand your responsibility as an owner

Securing an application requires security assurances for three aspects:

Application code . The logic that defines the custom application that you write. Securing that code requires
identifying and mitigating risks from the design and implementation of the application and assessing
supply chain risk of included components.
Application ser vices . The cloud services that the application uses such as databases, identity providers,
event hubs, IoT device management, and so on. Security for cloud services is a shared responsibility. The
cloud provider ensures the security of the underlying service. The application owner is responsible for
security implications of the configuration and operation of the service instance(s) used by the application
including any data stored and processed on the service.
Application hosting platform . The computing environment where the application runs. This could take
many forms with significant variations on who is responsible for security:
Legacy applications . typically require a full operating system (and any middleware) hosted on
physical or virtualized hardware. This operating system and installed middleware/other components
are operated and secured by the application owner or their infrastructure team(s). The security
responsibility for the physical hardware and OS virtualization components (virtualization hosts,
 operating systems, and management services) varies:
On-premises - The application owner is responsible for maintenance and security.
IaaS – The cloud provider is responsible for the underlying infrastructure and the application
owner’s organization is responsible for the VM configuration, operating system, and any
components installed on it.
Modern applications are hosted on PaaS environments such as an Azure application service. The
underlying operating system is secured by the cloud provider. Application owners are responsible for
the security of the application service configurations.
Containers are an application packaging mechanism in which applications are abstracted from the
environment in which they run. The containerized applications can run on a container service by the
cloud provider (modern applications) or on a server managed on premises or in IaaS.

Identify and classify applications

Identify applications that have a high potential impact and,or a high potential exposure to risk.
Business critical data . Applications that process or store information must have assurance of confidentiality,
integrity, and availability.
Regulated data . Applications that handle monetary instruments and sensitive personal information regulated
by standards such as the payment card industry (PCI), General Data Protection Regulation (GDPR), and Health
Information Portability and Accountability Act (HIPAA).
Business critical availability . Applications whose functionality is critical to the business mission, such as
production lines generating revenue, devices or services critical to life and safety, and other critical functions.
Significant Access . Applications that have access to systems with a high impact through technical means
such as
Stored Credentials or keys/certificates that grant access to the data/service.
Permissions granted through access control lists or other methods.
High exposure to attacks . Applications that are easily accessible to attackers such as web applications on the
public internet. Legacy applications can also be higher exposure as attackers (and penetration testers)
frequently target them because they know these legacy applications often have vulnerabilities that are difficult
to fix.

Use Azure services for fundamental components

Use Azure services databases, encryption, identity directory, and authentication instead of developing custom
solutions.
Identity – User directories and other authentication functions are complex to develop and critically
important to security assurances. Avoid custom authentication solutions. Instead choose native capabilities
like Azure Active Directory (Azure AD), Azure AD B2B, Azure AD B2C, or third-party solutions to
authenticate and grant permission to users, partners, customers, applications, services, and other entities.
For more information, see Security with identity and access management (IAM) in Azure.
Data Protection – Use established capabilities from cloud providers such as native encryption in cloud
services to encrypt and protect data. If direct use of cryptography is required, use well-established
cryptographic algorithms and not attempt to invent their own.
Key management – Always authenticate with identity services rather than handling cryptographic key. For
situations where you need to keys, use a managed key store such as Azure Key Vault. This will make sure
keys are handled safely in application code. You can use CredScan to discover potentially exposed keys in
your application code.
 Application Configurations – Inconsistent configurations for applications can create security risks. Azure
App Configuration provides a service to centrally manage application settings and feature flags, which
helps mitigate this risk.

Use native capabilities

Use native security capabilities built into cloud services instead of adding external security components, such as
data encryption, network traffic filtering, threat detection, and other functions.
Azure controls are maintained and supported by Microsoft. You don't have to invest in additional security tooling.
List of Azure Services
Native security capabilities of each service

Next steps
Applications and services
Application classification
Application threat analysis
Regulatory compliance
 Application threat analysis
11/2/2020 • 3 minutes to read • Edit Online

Do a comprehensive analysis to identify threats, attacks, vulnerabilities, and counter measures. Having this
information can protect the application and threats it might pose to the system. Start with simple questions to gain
insight into potential risks. Then, progress to advanced techniques using threat modeling.

1- Gather information about the basic security controls

Start by gathering information about each component of the application. The answers to these questions will
identify gaps in basic protection and understand the attack vectors.

A SK T H IS Q UEST IO N . . . TO DET ERM IN E C O N T RO L S T H AT . . .

Are connections authenticated using Azure AD, TLS (with Prevent unauthorized access to the application component
mutual authentication), or another modern security protocol and data.
approved by the security team?
Between users and the application
Between different application components and
services

Are you limiting access to only those accounts that have the Prevent unauthorized data tampering or alteration.
need to write or modify data in the application

Is the application activity logged and fed into a Security Detect and investigate attacks quickly.
Information and Event Management (SIEM) through Azure
Monitor or a similar solution?

Is critical data protected with encryption that has been Prevent unauthorized copying of data at rest.
approved by the security team?

Is inbound and outbound network traffic encrypted using Prevent unauthorized copying of data in transit.
TLS?

Is the application protected against Distributed Denial of Detect attacks designed to overload the application so it can’t
Service (DDoS) attacks using services such as Azure DDoS be used.
protection?

Does the application store any logon credentials or keys to Identify whether an attack can use your application to attack
access other applications, databases, or services? other systems.

Do the application controls allow you to fulfill regulatory Protect user’s private data and avoid compliance fines.
requirements?

2- Evaluate the application design progressively

Analyze application components and connections and their relationships. Threat modeling is a crucial engineering
exercise that includes defining security requirements, identifying and mitigating threats, and validating those
mitigations. This technique can be used at any stage of application development or production, but it's most
effective during the design stages of a new functionality.
Popular methodologies include:
STRIDE:
Spoofing
Tampering
Repudiation
Information Disclosure
Denial of Service
Elevation of Privilege
Microsoft Security Development Lifecycle uses STRIDE and provides a tool to assist with this process. This tool is
available at no additional cost. For more information, see Microsoft Threat Modeling Tool.
Open Web Application Security Project (OWASP) has documented a threat modeling approach for applications.

Integrate threat modeling through automation using secure operations. Here are some resources:
Toolkit for Secure DevOps on Azure.
Guidance on DevOps pipeline security by OWASP.

3- Mitigate the identified threats

The threat modeling tool produces a report of all the threats identified. After a potential threat is identified,
determine and how it can be detected and the response to that attack.
Use the Defense-in-Depth approach. This can help identify controls needed in the design to mitigate risk if a
primary security control fails. Evaluate how likely it is for the primary control to fail. If it does, what is the extent of
the potential organizational risk? Also, what is the effectiveness of the additional control (especially in cases that
would cause the primary control to fail). Based on the evaluation apply Defense-in-Depth measures to address
potential failures of security controls.
The principle of least privilege is one way of implementing Defense-in-Depth. It limits the damage that can be
done by a single account. Grant least number of privileges to accounts that allows them to accomplish with the
required permissions within a time period. This helps mitigate the damage of an attacker who gains access to the
account to compromise security assurances.
How are threats addressed once found?

Here are some best practices:

Make sure the results are communicated to the interested teams.
Upload the threat modeling report to a tracking tool. Create work items that can be validated and addressed by
the developers. Cyber security teams can also use the report to determine attack vectors during a penetration
test.
As new features are added to the application, update the threat model report and integrate it into the code
management process. Triage security issues into the next release cycle or a faster release, depending on the
severity.
For information about mitigation strategies, see RapidAttack.
How long does it typically take to deploy a security fix into production?

If a security vulnerability is discovered, update the software with the fix as soon as possible. Have processes, tools,
and approvals in place to roll out the fix quickly.
Next steps
Applications and services
Application classification
Regulatory compliance
 Regulatory compliance
12/18/2020 • 2 minutes to read • Edit Online

Cloud architectures can have challenges when strictly applying standards to workloads. This article provides some
security recommendations for adhering to standards.

Gather regulatory requirements

What do the law, authorities, and regulators require?

Governments and regulatory organizations frequently publish standards to help define good security practices so
that organizations can avoid negligence. The purpose and scope of these standards and regulations vary, but the
security requirements can influence the design for data protection and retention, data privacy, and system security.
Noncompliance can lead to fines or other business impact. Work with your regulators and carefully review the
standards to understand both the intent and the literal wording. Start by answering these questions:
What are the regulatory requirements for the solution?
How is compliance measured?
Who approves that solution meets the requirements?
Are there processes for obtaining attestations?

Use the Microsoft Trust Center

Outdated security practices won’t sufficiently protect cloud workloads. Keep checking the Microsoft Trust Center
for the latest information, news, and best practices in security, privacy, and compliance.
Data governance . Focus on protecting information in cloud services, mobile devices, workstations, or
collaboration platforms. Build the security strategy by classifying and labeling information. Use strong
access control and encryption technology.
Compliance offerings . Microsoft offers a comprehensive set of compliance offerings to help your
organization comply with national, regional, and industry-specific requirements governing the collection
and use of data. For information, see Compliance offerings.
Compliance score . Use Microsoft Compliance Score to assess your data protection controls on an
ongoing basis. Act on the recommendations to make progress toward compliance.
Audit repor ts . Use audit reports to stay current on the latest privacy, security, and compliance-related
information for Microsoft’s cloud services. See Audit Reports.
Shared responsibility . The workload can be hosted on Software as a Service (SaaS), Platform as a
Service (PaaS), Infrastructure as a Service (IaaS), or in an on-premises datacenter. Have a clear
understanding about the portions of the architecture for which you are responsible versus Azure. Whatever
the hosting model, the following responsibilities are always retained by you:
Data
Endpoints
Account
Access management
For more information, see Shared responsibility in the cloud.
Next steps
Applications and services
Application classification
Application threat analysis
 Secure application configuration and dependencies
11/2/2020 • 2 minutes to read • Edit Online

Security of an application that is hosted in Azure is a shared responsibility between you as the application owner
and Azure. For IaaS, you are responsible for configurations related to VM, operating system, and components
installed on it. For PaaS, you are responsible for the security of the application service configurations and making
sure that the dependencies used by the application are also secure.

Key points
Do not store secrets in source code or configuration files. Instead keep them in a secure store, such as Azure
Key Vault.
Do not expose detailed error information when handling application exceptions.
Do not expose platform-specific information.
Restrict access to Azure resources that do not meet the security requirements.
Validate the security of any open-source code added to your application.
Update frameworks and libraries as part of the application lifecycle.

Configuration security
During the design phase, consider the way you store secrets and handle exceptions. Here are some considerations.
Are application keys and secrets stored securely?

Configuration within in the application can include secrets like database connection strings, certificate keys, and so
on. Do not store secrets in source code or configuration files. Instead keep them in a secure store, such as Azure
Key Vault. If you need to store secrets in code, identify them with static code scanning tools. Add the scanning
process in your continuous integration (CI) pipeline.
Are errors and exceptions handled properly without exposing that information to users?

When handling application exceptions, make the application fail gracefully and log the error. Do not provide
detailed information related to the failure, such as call stack, SQL queries, or out of range errors. This information
can provide attackers with valuable information about the internals of the application.
Is platform-specific information removed from ser ver-client communication?

Do not reveal information about the application platform. Such information (for example, "X-Powered-By", "X-
ASPNET-VERSION") can get exposed through HTTP banners HTTP headers, error messages, website footers.
Malicious actors can use this information when mapping attack vectors of the application.
Consider using Azure CDN to separate the hosting platform from end users. Azure API Management offers
transformation policies that allow you to modify HTTP headers and remove sensitive information.
Are Azure policies used to control the configuration of the solution resources?

Use Azure Policy to deploy desired settings where applicable. Block resources that do not meet the proper security
requirements defined during service enablement.

Dependencies, frameworks, and libraries

What are the frameworks and libraries used by the application?

It's important to be aware of the implications of using third-party frameworks and libraries in your application
code. These components can result in vulnerabilities. Here are some best practices:
Validate the security of any open-source code added to your application. Tools that can help this assessment:
OWASP Dependency-Check
NPM audit
Maintain a list of frameworks and libraries as part of the application inventory. Also, keep track of versions
in use.
Update frameworks and libraries as part of the application lifecycle. Prioritize critical security patches.

Referenced Azure services

Azure Key Vault
Azure CDN
Azure Policy

Next steps
Applications and services
Application classification
Application threat analysis
Regulatory compliance
 Security health modeling
12/18/2020 • 2 minutes to read • Edit Online

Health modeling refers to the activities that maintain the security posture of a workload through monitoring. These
activities can highlight, if the current security practices are effective or are there new requirements. Health
modeling can be categorized as follows:

Key points
Monitor the workload and the infrastructure in which it runs.
Conduct audits.
Enable, acquire, and store audit logs.
Update and patch security fixes.
Respond to incidents.

In this section
A SSESSM EN T DESC RIP T IO N

How is security monitored in the application context? Use Azure tools and services to monitor your security posture
and also remediate incidents.

Is access to the control plane and data plane of the Monitor network conditions and identity-related risk events
application periodically reviewed? regularly.

Do you implement security practices and tools during Activities related to enabling, acquiring, and storing audit logs
the development lifecycle? for Azure services.

Are operational processes for incident response Guidance for the central SecOps team for monitoring security-
defined and tested? related telemetry data and investigating security breaches.

Next steps
We recommend applying as many as of the best practices as early as possible, and then working to retrofit any
gaps over time as you mature your security program.
Optimize security investments
Assign stakeholders to use Secure Score in Azure Security Center to monitor risk profile and continuously improve
security posture.
Operationalize Azure Secure Score
 Security monitoring tools in Azure
12/18/2020 • 3 minutes to read • Edit Online

Use security monitoring tools to monitor for anomalous behavior and enable investigation of incidents.
Use this article to learn how to strengthen your security posture using the following monitor tools:
Azure Security Center - Prevent, detect, and respond to threats with increased visibility and control over the
security of your resources.
Azure Sentinel - Receive intelligent security analytics and threat intelligence across the enterprise.
Azure DDoS Protection - Defend against DDoS attacks.
Azure Rights Management (RMS) - Protect files and emails across multiple devices.
Azure Information Protection - Secure email, documents, and sensitive data that you share outside your
company.

Prevent, detect, and respond to threats

If your business has hybrid workloads, you may be experiencing weak security posture. For example, if you are
deploying new resources across workloads, they may not be configured according to security best practices. You
may also need to continuously monitor the security status of the network in order to block unwanted connections
that could potentially make it easier for an attacker to creep along your network.
Azure Security Center strengthens the security posture of your data centers, and provides advanced threat
protection across your hybrid workloads in the cloud (whether they're in Azure or not) as well as on-premises.
For information on the Azure Security Center tools, see Strengthen security posture.
For frequently asked questions on Azure Security Center, see FAQ - General Questions.

Detect threats early

Your business may be experiencing increasingly sophisticated attacks, increasing volumes of alerts, and long
resolution timeframes.
To combat these issues, Azure Sentinel uses intelligent security analytics and threat intelligence to provide a single
solution for alert detection, threat visibility, proactive hunting, and threat response.
Azure Sentinel's "birds-eye view" across the enterprise allows you to:
Collect data at cloud scale.
Detect previously undetected threats and minimize false positives.
Investigate threats with artificial intelligence and hunt for suspicious activities at scale.
Respond to incidents rapidly.
For information on the Azure Sentinel tools that will help to meet these requirements, see What is Azure Sentinel?

Protect resources against DDoS attacks

A Distributed Denial of Service (DDoS) attack attempts to exhaust an application's resources, making the
application unavailable to legitimate users. DDoS attacks can be targeted at any endpoint that is publicly reachable
through the internet.
Azure DDoS Protection, combined with application design best practices, provide defense against DDoS attacks.
The service tier that is used (Basic or Standard) determines the available features.
DDoS attack protection features include:
Always-on traffic monitoring which provides near real-time detection of a DDoS attack.
Automatic configuration and tuning of your DDoS Protection settings using intelligent traffic-profiling.
Detailed reports generated in five-minute increments during an attack, and a complete summary after the
attack ends (Standard service tier only).
For types of DDoS attacks that DDoS Protection Standard mitigates as well as more features, see Azure DDoS
Protection Standard overview.

Protect files and emails across multiple devices

Your business may encounter challenges with protecting documents and emails. For example, file protection,
collaboration, and sharing may be issues. You also might be experiencing problems regarding platform support or
infrastructure.
Azure Rights Management (RMS) is a cloud-based protection service that uses encryption, identity, and
authorization policies to help secure files and emails across multiple devices, including phones, tablets, and PCs.
To learn more about how RMS can address these issues, see Business problems solved by Azure Rights
Management.

Classify and protect documents and emails

Organizations that are weak on data classification and file protection might be more susceptible to data leakage or
data misuse.
Azure Information Protection (AIP) is a cloud-based solution that enables organizations to classify and protect
documents and emails by applying labels.
The data classification process categorizes data by sensitivity and business impact in order to identify risks. When
data is classified, you can manage it in ways that protect sensitive or important data from theft or loss.
With proper file protection, you can analyze data flows to gain insight into your business, detect risky behaviors
and take corrective measures, track access to documents, and more. The protection technology in AIP uses
encryption, identity, and authorization policies. Protection stays with the documents and emails, independently of
the location, regardless of whether they are inside or outside your organization, networks, file servers, and
applications.

Next steps
Security health modeling
Security logs and audits
Security operations in Azure
Check for identity, network, data risks
 Security operations in Azure
12/18/2020 • 3 minutes to read • Edit Online

The responsibility of the security operation team (also known as Security Operations Center (SOC), or SecOps) is
to rapidly detect, prioritize, and triage potential attacks. These operations help eliminate false positives and focus
on real attacks, reducing the mean time to remediate real incidents. Central SecOps team monitors security-
related telemetry data and investigates security breaches. It's important that any communication, investigation,
and hunting activities are aligned with the application team.

Security operations principles

Here are some general best practices for conducting security operations:
Follow the NIST Cybersecurity Framework functions as part of operations.
Detect the presence of adversaries in the system.
Respond by quickly investigating whether it's an actual attack or a false alarm.
Recover and restore the confidentiality, integrity, and availability of the workload during and after an
attack.
For information about the framework, see NIST Cybersecurity Framework.
Acknowledge an alert quickly. A detected adversary must not be ignored while defenders are triaging false
positives.
Reduce the time to remediate a detected adversary. Reduce their opportunity time to conduct and attack
and reach sensitive systems.
Prioritize security investments into systems that have high intrinsic value. For example, administrator
accounts.
Proactively hunt for adversaries as your system matures. This effort will reduce the time that a higher
skilled adversary can operate in the environment. For example, skilled enough to evade reactive alerts.
For information about the metrics that the Microsoft’s SOC team uses , see Microsoft SOC.

Tools
Here are some Azure tools that a SOC team can use investigate and remediate incidents.

TO O L P URP O SE

Azure Sentinel Centralized Security Information and Event Management

(SIEM) to get enterprise-wide visibility into logs.

Azure Security Center Alert generation. Use security playbook in response to an

alert.

Azure Monitor Event logs from application and Azure services.

Azure Network Security Group (NSG) Visibility into network activities.

 TO O L P URP O SE

Azure Information Protection Secure email, documents, and sensitive data that you share
outside your company.

Investigation practices should use native tools with deep knowledge of the asset type such as an Endpoint
detection and response (EDR) solution, Identity tools, and Azure Sentinel.
For more information about monitoring tools, see Security monitoring tools in Azure.

Incident response
Make sure that a security contact receives Azure incident notifications from Microsoft.
In most cases, such notifications indicate that your resource is compromised and,or attacking another customer.
This enables your security operations team to rapidly respond to potential security risks and remediate them.
Ensure administrator contact information in the Azure enrollment portal includes contact information that will
notify security operations directly or rapidly through an internal process.
Are operational processes for incident response defined and tested?

Actions executed during an incident and response investigation could impact application availability or
performance. Define these processes and align them with the responsible (and in most cases central) SecOps
team. The impact of such an investigation on the application has to be analyzed.
Are there tools to help incident responders quickly understand the application and components to
do an investigation?

Incident responders are part of a central SecOps team and need to understand security insights of an application.
Security playbook in Azure Sentinel can help to understand the security concepts and cover the typical
investigation activities.

Hybrid enterprise view

Security operations tooling and processes should be designed for attacks on cloud and on-premises assets.
Attackers don’t restrict their actions to a particular environment when targeting an organization. They attack
resources on any platform using any method available. They can pivot between cloud and on-premises resources
using identity or other means. This enterprise-wide view will enable SecOps to rapidly detect, respond, and
recover from attacks, reducing organizational risk.

Leverage native detections and controls

Use Azure security detections and controls instead of creating custom features for viewing and analyzing event
logs. Azure services are updated with new features and have the ability to detect false positive with a higher
accuracy rate.
To get a unified view across the enterprise, feed the logs collected through native detections (such as Azure
Monitor) into a centralized SIEM like Azure Sentinel. Avoid using generalized log analysis tools and queries. Within
Azure Monitor, create Log Analytics Workspace to store logs. You can also review logs and perform queries on log
data. These tools can offer high-quality alerts.

Next steps
Security health modeling
Security tools
Security logs and audits
Check for identity, network, data risks
 Security audits
12/18/2020 • 3 minutes to read • Edit Online

To make sure that the security posture doesn’t degrade over time, have regular auditing that checks compliance
with organizational standards. Enable, acquire, and store audit logs for Azure services.

Key points
Use approved time synchronization sources.
Configure central security log management.
Enable audit logging for Azure resources.
Collect security logs from operating systems.
Configure security log storage retention.
Monitor and review logs.
Enable alerts for anomalous activities.
Centralize anti-malware logging.
Enable DNS query logging.
Enable command-line audit logging.

Review critical access

Is access to the control plane and data plane of the application periodically reviewed?

Regularly review roles that have high privileges. Set up a recurring review pattern to ensure that accounts are
removed from permissions as roles change. Consider auditing at least twice a year.
As people in the organization and on the project change, make sure that only the right people have access to the
application infrastructure. Auditing and reviewing the access control reduces the attack vector to the application.
Azure control plane depends on Azure AD. You can conduct the review manually or through an automated process
by using tools such as Azure AD access reviews. These reviews are often centrally performed often as part of
internal or external audit activities.

Enforce policy compliance

Make sure that the security team is auditing the environment to report on compliance with the security policy of
the organization. Security teams may also enforce compliance with these policies.
Enforce and audit industry, government, and internal corporate security policies. Policy monitoring checks that
initial configurations are correct and that it continues to be compliant over time.
For Azure, use Azure Policy to create and manage policies that enforce compliance. Azure Policies are built on the
Azure Resource Manager capabilities. Azure Policy can also be assigned through Azure Blueprints. For more
information, see Tutorial: Create and manage policies to enforce compliance.

Use native logging through Azure Monitor

Use Azure Monitor to log application activity and feed into a Security Information and Event Management (SIEM).
You can also use Azure Monitor to collect activity logs transmitted by Azure resources. Activity logs provide
detailed diagnostic and auditing information.
For more information, see these articles:
How to get started with Azure Monitor and third-party SIEM integration
How to collect platform logs and metrics with Azure Monitor
You can use Azure Monitor to collect information about the operating system running on Azure compute. If your
are running your own compute, use Azure Security Center. For more information, see Understand Azure Security
Center data collection.
Within Azure Monitor, create Log Analytics Workspace to store logs. You can also review logs and perform queries
on log data. Set the retention period according to your organization's compliance regulations. Use Azure Storage
Accounts for long-term/archival storage.

Prioritize alert and log integration

Ensure that you are integrating critical security alerts and logs into SIEMs without introducing a high volume of
low value data. Doing so can increase SIEM cost, false positives, and lower performance.
Use the data to support these activities:
Alerts. Use existing tools or data required for generating custom alerts.
Investigation of an incident. For example, required for common queries.
Proactive hunting activities.
Integrating more data can allow you to enrich alerts with additional context that enable rapid response and
remediation (filter false positives, and elevate true positives, and so on.), but collection is not detection. If you don’t
have a reasonable expectation that the data will provide value (for example, high volume of firewall denies events),
you may deprioritize integration of these events.

Next steps
Security health modeling
Security operations in Azure
Check for identity, network, data risks
Security tools
 Check for identity, network, data risks
12/18/2020 • 3 minutes to read • Edit Online

Along with identity-based access control, network-based access control is a top priority for protecting assets.
Monitor identity-related risk events using adaptive machine learning algorithms, heuristics quickly before the
attacker can gain deeper access into the system. Also, monitor all communications between segments to detect
potential security threats flowing over the wire.
This article describes some considerations that can help monitor the workload for those risks.

Key points
View identity related risks in Azure Active Directory (Azure AD) reporting and Azure AD Identity Protection.
Obfuscate Personally Identifiable Information (PII) by using PII Detection cognitive skill.
Enable logs (including raw traffic) from your network devices.
Set alerts and gain access to real-time performance information at the packet level.

Review identity risks

Most security incidents take place after an attacker initially gains access using a stolen identity.
Suppose an attacker gains access using a stolen identity. Even though the identity has low privileges, the attacker
can use it to traverse laterally and gain access to more privileged identities. This way the attacker can control
access to the target data or systems.
Does the organization actively monitor identity related risk events related to potentially
compromised identities?

Monitor identity-related risk events on potentially compromised identities and remediate those risks. Review the
reported risk events in these ways:
Azure AD reporting. For information see users at risk security report and the risky sign-ins security report.
Use the reporting capabilities of Azure Active Directory Identity Protection.
Use the Identity Protection risk events API to gain get programmatic access to security detections by using
Microsoft Graph. See riskDetection and riskyUser APIs.
Azure AD uses adaptive machine learning algorithms, heuristics, and known compromised credentials
(username/password pairs) to detect suspicious actions that are related to your user accounts. These
username/password pairs come from monitoring public and dark web and by working with security researchers,
law enforcement, security teams at Microsoft, and others.
Remediate risks by manually addressing each reported account or by setting up a user risk policy to require a
password change for high risk events.

Mask PII information

Is Personally Identifiable Information (PII) detected and removed/obfuscated automatically?

Be cautious when logging sensitive application information. Don't store PII (contact information, payment
information etc.), any application logs. Apply protective measures, such as obfuscation. Machine learning tools like
Cognitive Search PII detection can help with this. For more information, see PII Detection cognitive skill.
Enable network visibility
One way to enable network visibility is by integrating network logs and analyzing the data to identify anomalies.
Based on those insights, you can choose to set alerts or block traffic crossing segmentation boundaries.
How do you monitor and diagnose conditions of the network?

Enable logs (including raw traffic) from your network devices.

Integrate network logs into a security information and event management (SIEM) service, such as Azure Sentinel.
Other popular choices include Splunk, QRadar, or ArcSight ESM.
Use machine learning analytics platforms that support ingestion of large amounts of information and can analyze
large datasets quickly. Also, these solutions can be tuned to significantly reduce the false positive alerts.
Here are some ways to integrate network logs:
Security group logs – flow logs and diagnostic logs
Web application firewall logs
Virtual network taps
Azure Network Watcher

Proactive monitoring
How do you gain access to real-time performance information at the packet level?

Take advantage of packet capture to set alerts and gain access to real-time performance information at the packet
level.
Packet capture tracks traffic in and out of virtual machines. It gives you the capability to run proactive captures
based on defined network anomalies including information about network intrusions.
For an example, see Scenario: Get alerts when VM is sending you more TCP segments than usual.

Next steps
Security health modeling
Security operations in Azure
Security tools
Security logs and audits
 Governance, risk, and compliance
12/18/2020 • 22 minutes to read • Edit Online

Organizations of all sizes are constrained by their available resources; financial, people, and time. To achieve an
effective return on investment (ROI) organizations must prioritize where they will invest. Implementation of security
across the organization is also constrained by this, so to achieve an appropriate ROI on security the organization
needs to first understand and define its security priorities.
Governance: How is the organization’s security going to be monitored, audited, and reported? Design and
implementation of security controls within an organization is only the beginning of the story. How does the
organization know that things are actually working? Are they improving? Are there new requirements? Is there
mandatory reporting? Similar to compliance there may be external industry, government or regulatory standards
that need to be considered.
Risk : What types of risks does the organization face while trying to protect identifiable information, Intellectual
Property (IP), financial information? Who may be interested or could use this information if stolen, including
external and internal threats as well as unintentional or malicious? A commonly forgotten but extremely important
consideration within risk is addressing Disaster Recovery and Business Continuity.
Compliance: Is there a specific industry, government, or regulatory requirements that dictate or provide
recommendation on criteria that your organization’s security controls must meet? Examples of such standards,
organizations, controls, and legislation are ISO27001, NIST, PCI-DSS.
The collective role of organization(s) is to manage the security standards of the organization through their lifecycle:
Define - Set organizational standards and policies for practices, technologies, and configurations based on
internal factors (organizational culture, risk appetite, asset valuation, business initiatives, etc.) and external
factors (benchmarks, regulatory standards, threat environment, and more)
Improve – Continually push these standards incrementally forward towards the ideal state to ensure
continual risk reduction.
Sustain – Ensure the security posture doesn’t degrade naturally over time by instituting auditing and
monitoring compliance with organizational standards.

Prioritize security best practices investments

Security best practices are ideally applied proactively and completely to all systems as you build your cloud
program, but this isn’t reality for most enterprise organizations. Business goals, project constraints, and other
factors often cause organizations to balance security risk against other risks and apply a subset of best practices at
any given point.
We recommend applying as many as of the best practices as early as possible, and then working to retrofit any
gaps over time as you mature your security program. We recommend evaluating the following considerations
when prioritizing which to follow first:
High business impact and highly exposed systems: These include systems with direct intrinsic value as
well as the systems that provide attackers a path to them. For more information, see Identify and classify
business critical applications.
Easiest to implement mitigations: Identify quick wins by prioritizing the best practices, which your
organization can execute quickly because you already have the required skills, tools, and knowledge to do it
(for example, implementing a Web App Firewall (WAF) to protect a legacy application). Be careful not to
 exclusively use (or overuse) this short-term prioritization method. Doing so can increase your risk by
preventing your program from growing and leaving critical risks exposed for extended periods.
Microsoft has provided some prioritized lists of security initiatives to help organizations start with these decisions
based on our experience with threats and mitigation initiatives in our own environments and across our customers.
See Module 4a of the Microsoft CISO Workshop.

Operationalize Azure Secure Score

Assign stakeholders to use Secure Score in Azure Security Center to monitor risk profile and continuously improve
security posture. This will help you continuously improve your security posture by rapidly identifying and
remediating common security hygiene risks, which can significantly reduce your organization's risk. Set up a
regular cadence (typically monthly) to review Azure secure score and plan initiatives with specific improvement
goals. You should also look into gamifying the activity if possible to increase engagement and focus from the
responsible teams (e.g. creating a friendly competition on who increased their score more).
NOTE: Security posture management is an emerging function, this may be a single dedicated team or the
responsibility may be taken on by one or more existing teams described below.
Monitor Secure Score
SC O RE A REA RESP O N SIB L E T EA M

Monitor Secure Score and Follow up with technical teams Security Posture Management Team
Vulnerability Management (or Governance/Risk/Compliance
team)
Architecture Team
Responsible Technical Team (listed below)

Improve Secure Score

C O M P UT E A N D A P P S RESO URC ES RESP O N SIB L E T EC H N IC A L T EA M

App Services Application Development/Security Team(s)

Containers Application Development and/or Infrastructure/IT Operations

VMs/Scale sets/compute IT/Infrastructure Operations

DATA & STO RA GE RESO URC ES RESP O N SIB L E T EC H N IC A L T EA M

SQL/Redis/Data Lake Analytics/Data Lake Store Database Team

Storage Accounts Storage/Infrastructure Team

IDEN T IT Y A N D A C C ESS RESO URC ES RESP O N SIB L E T EC H N IC A L T EA M

Subscriptions Identity Team(s)

Key Vault Information/Data Security Team

N ET W O RK IN G RESO URC ES RESP O N SIB L E T EC H N IC A L T EA M

 N ET W O RK IN G RESO URC ES RESP O N SIB L E T EC H N IC A L T EA M

Networking Resources Networking Team

Network Security Team

IOT SEC URIT Y RESP O N SIB L E T EC H N IC A L T EA M

IoT Resources IoT Operations Team

NOTE: Note that in the DevOps model, some application teams may be responsible for their own application
resources.

The Azure Secure Score sample shows how to get your Azure Secure Score for a subscription by calling the
Azure Security Center REST API. The API methods provide the flexibility to query the data and build your own
reporting mechanism of your secure scores over time.

Manage connected tenants

Ensure your security organization is aware of all enrollments and associated subscriptions connected to your
existing environment (via ExpressRoute or Site-Site VPN) and monitoring as part of the overall enterprise.
These Azure resources are part of your enterprise environment and security organizations require visibility into
them. Security organizations need this access to assess risk and to identify whether organizational policies and
applicable regulatory requirements are being followed.
Ensure all Azure environments that connect to your production environment/network apply your organization’s
policy and IT governance controls for security. You can discover existing connected tenants using a tool provided by
Microsoft. Guidance on permissions you may assign to security is in the Assign privileges for managing the
environment section.

Clear lines of responsibility

Designate the parties responsible for specific functions in Azure
Clearly documenting and sharing the contacts responsible for each of these functions will create consistency and
facilitate communication. Based on our experience with many cloud adoption projects, this will avoid confusion that
can lead to human and automation errors that create security risk.
Designate groups (or individual roles) that will be responsible for these key functions:

GRO UP O R IN DIVIDUA L RO L E RESP O N SIB IL IT Y

Network Security Typically existing network security team. Configuration and

maintenance of Azure Firewall, Network Virtual Appliances
(and associated routing), WAFs, NSGs, ASGs, etc.

Network Management Typically existing network operations team. Enterprise-wide

virtual network and subnet allocation.

Ser ver Endpoint Security Typically IT operations, security, or jointly. Monitor and
remediate server security (patching, configuration, endpoint
security, etc.).
 GRO UP O R IN DIVIDUA L RO L E RESP O N SIB IL IT Y

Incident Monitoring and Response Typically security operations team. Investigate and remediate
security incidents in Security Information and Event
Management (SIEM) or source console.

Policy Management Typically GRC team + Architecture. Set Direction for use of
Role-Based Access Control (RBAC), Azure Security Center,
Administrator protection strategy, and Azure Policy to govern
Azure resources.

Identity Security and Standards Typically Security Team + Identity Team jointly. Set direction for
Azure AD directories, PIM/PAM usage, MFA,
password/synchronization configuration, Application Identity
Standards.

Enterprise segmentation strategy

Identify groups of resources that can be isolated from other parts of the enterprise to contain (and detect)
adversary movement within your enterprise. This unified enterprise segmentation strategy will guide all technical
teams to consistently segment access using networking, applications, identity, and any other access controls.
A clear and simple segmentation strategy helps contain risk while enabling productivity and business operations.
An enterprise segmentation strategy is defined higher than a traditional “network segmentation” security strategy.
Traditional segmentation approaches for on premises environments frequently failed to achieve their goals because
they were developed “bottom-up” by different technical teams and were not aligned well with business use cases
and application workloads. This resulted in overwhelming complexity that generates support issues and often
undermines the original purpose with broad network firewall exceptions.
Creating a unified enterprise segmentation strategy enables to guide all technical teams stakeholders (IT, Security,
Applications, etc.) Business Units that is built around the business risks and needs will increase alignment to and
understand and support sustainability of the security containment promises. This clarity and alignment will also
reduce s the risk of human errors and automation failures that can lead to security vulnerabilities, operational
downtime, or both.
While network micro-segmentation also offers promise to reduce risk (discussed more in Network Security and
Containment section), it doesn’t eliminate the need to align technical teams. Micro segmentation should be
considered after to and plans to ensure the ensuring technical teams are aligned so you can avoid a recurrence of
the internal conflicts that plagued and confusion of the on-premises network generation segmentation strategies.
Here are Microsoft's recommendations for prioritizing initiatives on containment and segmentation (based on Zero
Trust principles). These recommendations are listed in priority order by highest importance.
Ensure alignment of technical teams to a single enterprise segmentation strategy.
Invest in broadening containment by establishing a modern perimeter based on zero trust principles focused
on identity, device, applications, and other signals (to overcome limitation of network controls to protect new
resources and attack types).
Bolster network controls for legacy applications by exploring micro segmentation strategies.
A good enterprise segmentation strategy meets these criteria:
Enables Operations – Minimizes operation friction by aligning to business practices and applications
Contains Risk - Adds cost and friction to attackers by
 Isolating sensitive workloads from compromise of other assets
Isolating high exposure systems from being used as a pivot to other systems
Monitored – Security Operations should monitor for potential violations of the integrity of the segments
(account usage, unexpected traffic, etc.)

Security team visibility

Provide security teams read-only access to the security aspects of all technical resources in their purview
Security organizations require visibility into the technical environment to perform their duties of assessing and
reporting on organizational risk. Without this visibility, security will have to rely on information provided from
groups, operating the environment, who have a potential conflict of interest (and other priorities).
Note that security teams may separately be granted additional privileges if they have operational responsibilities or
a requirement to enforce compliance on Azure resources.
For example in Azure, assign security teams to the Security Readers permission that provides access to measure
security risk (without providing access to the data itself)
For enterprise security groups with broad responsibility for security of Azure, you can assign this permission using:
Root management group – for teams responsible for assessing and reporting risk on all resources
Segment management group(s) – for teams with limited scope of responsibility (typically required because
of organizational boundaries or regulatory requirements)
Because security will have broad access to the environment (and visibility into potentially exploitable
vulnerabilities), you should consider them critical impact accounts and apply the same protections as
administrators. The Administration section details these controls for Azure.

Assign privileges for managing the environment

Grant roles with operational responsibilities in Azure the appropriate permissions based on a clearly documented
strategy built from the principle of least privilege and your operational needs.
Providing clear guidance that follows a reference model will reduce risk because by increasing it provides clarity for
your technical teams implementing these permissions. This clarity makes it easier to detect and correct human
errors like overpermissioning, reducing your overall risk.
Microsoft recommends starting from these Microsoft reference models and adapting to your organization.

Core Services Reference Permissions

This segment hosts shared services utilized across the organization. These shared services typically include Active
Directory Domain Services, DNS/DHCP, System Management Tools hosted on Azure Infrastructure as a Service
(IaaS) virtual machines.
Security Visibility across all resources – For security teams, grant read-only access to security attributes for all
technical environments. This access level is needed to assess risk factors, identify potential mitigations, and advise
organizational stakeholders who accept the risk. See Security Team Visibility for more details.
Policy management across some or all resources – To monitor and enforce compliance with external (or
internal) regulations, standards, and security policy, assign appropriate permission to those roles. The roles and
permissions you choose will depend on the organizational culture and expectations of the policy program. See
Microsoft Cloud Adoption Framework for Azure.
Central IT operations across all resources – Grant permissions to the central IT department (often the
infrastructure team) to create, modify, and delete resources like virtual machines and storage.
Central networking group across network resources – To ensure consistency and avoid technical conflicts,
assign network resource responsibilities to a single central networking organization. These resources should
include virtual networks, subnets, Network Security Groups (NSG), and the virtual machines hosting virtual
network appliances. See Centralize Network Management And Security for more details
Resource Role Permissions – For most core services, administrative privileges required to manage them are
granted via the application itself (Active Directory, DNS/DHCP, System Management Tools, etc.), so no additional
Azure resource permissions are required. If your organizational model requires these teams to manage their own
VMs, storage, or other Azure resources, you can assign these permissions to those roles.
Ser vice admin (Break Glass Account) – Use the service admin role only for emergencies (and initial setup if
required). Do not use this role for daily tasks. See Emergency Access (‘Break Glass’ Accounts) for more details.
Segment reference permissions
This segment permission design provides consistency while allowing flexibility to accommodate the range of
organizational models from a single centralized IT group to mostly independent IT and DevOps teams.
Security visibility across all resources – For security teams, grant read-only access to security attributes for all
technical environments. This access level is needed to assess risk factors, identify potential mitigations, and advise
organizational stakeholders who accept the risk. See Security Team Visibility.
Policy management across some or all resources – To monitor and enforce compliance with external (or
internal) regulations, standards, and security policy assign appropriate permission to those roles. The roles and
permissions you choose will depend on the organizational culture and expectations of the policy program. See
Microsoft Cloud Adoption Framework for Azure.
IT Operations across all resources – Grant permission to create, modify, and delete resources. The purpose of
the segment (and resulting permissions) will depend on your organization structure.
Segments with resources managed by a centralized IT organization can grant the central IT department
(often the infrastructure team) permission to modify these resources.
Segments managed by independent business units or functions (such as a Human Resources IT Team) can
grant those teams permission to all resources in the segment.
Segments with autonomous DevOps teams don’t need to grant permissions across all resources because the
resource role (below) grants permissions to application teams. For emergencies, use the service admin
account (break-glass account).
Central networking group across network resources – To ensure consistency and avoid technical conflicts,
assign network resource responsibilities to a single central networking organization. These resources should
include virtual networks, subnets, Network Security Groups (NSG), and the virtual machines hosting virtual
network appliances. See Centralize Network Management And Security.
Resource Role Permissions – Segments with autonomous DevOps teams will manage the resources associated
with each application. The actual roles and their permissions depend on the application size and complexity, the
application team size and complexity, and the culture of the organization and application team.
Ser vice Admin (Break Glass Account) – Use the service admin role only for emergencies (and initial setup if
required). Do not use this role for daily tasks. See Emergency Access (‘Break Glass’ Accounts) for more details.
Permission Guidance and Tips
To drive consistency and ensure application to future subscriptions, permissions should be assigned at
management group for the segment rather than the individual subscriptions. See Avoid Granular and
Custom Permissions for more details.
You should first review the built-in roles to see if one is applicable before creating a custom role to grant the
appropriate permissions to VMs and other objects. See Use Built in Roles for more details
Security managers group membership may be appropriate for smaller teams/organizations where
security teams have extensive operational responsibilities.

Establish segmentation with management groups

Structure management groups into a simple design that guides the enterprise segmentation model.
Management groups offer the ability to consistently and efficiently manage resources (including multiple
subscriptions as needed). However, because of their flexibility, it's possible to create an overly complex design.
Complexity creates confusion and negatively impacts both operations and security (as illustrated by overly complex
Organizational Unit (OU) and Group Policy Object (GPO) designs for Active Directory).
Microsoft recommends aligning the top level of management groups (MGs) into a simple enterprise segmentation
strategy limited to 1 or 2 levels.

Use root management group carefully

Use the Root Management Group (MG) for enterprise consistency, but test changes carefully to minimize risk of
operational disruption.
The root management group enables you to ensure consistency across the enterprise by applying policies,
permissions, and tags across all subscriptions. Care must be taken when planning and implementing assignments
to the root management group because this can affect every resource on Azure and potentially cause downtime or
other negative impacts on productivity in the event of errors or unanticipated effects.
Root management group guidance:
Plan Carefully - Select enterprise-wide elements to the root management group that have a clear
requirement to be applied across every resource and/or low impact.
Good candidates include:
Regulator y requirements with clear business risk/impact (for example, restrictions related to data
sovereignty).
Near-zero potential negative impact on operations such as policy with audit effect, Tag
assignment, RBAC permissions assignments that have been carefully reviewed.
Test First - Carefully test all enterprise-wide changes on the root management group before applying
(policy, tags, RBAC model, etc.) using a
Test Lab - Representative lab tenant or lab segment in production tenant.
Production Pilot - Segment MG or Designated subset in subscription(s) / MG.
Validate Changes – to ensure they have the desired effect.

Virtual Machine (VM) security updates and strong passwords

Ensure policy and processes enable (and require) rapid application of security updates to virtual machines.
Attackers constantly scan public cloud IP ranges for open management ports and attempt “easy” attacks like
common passwords and unpatched vulnerabilities.
Enable Azure Security Center to identify missing security updates & apply them.
Local Admin Password Solution (LAPS) or a third-party Privileged Access Management can set strong local admin
passwords and just in time access to them.

Remove Virtual Machine (VM) direct internet connectivity

Make sure policies and processes require restricting and monitoring direct internet connectivity by virtual
machines.
For Azure, you can enforce policies by,
Enterprise-wide prevention - Prevent inadvertent exposure by following the permissions and roles
described in the reference model.
Ensures that network traffic is routed through approved egress points by default.
Exceptions (such as adding a public IP address to a resource) must go through a centralized group
that evaluates exception requests and makes sure appropriate controls are applied.
Identify and remediate exposed virtual machines by using the Azure Security Center network visualization
to quickly identify internet exposed resources.
Restrict management por ts (RDP, SSH) using Just in Time access in Azure Security Center.
One way of managing VMs in the virtual network is by using Azure Bastion. This service allows you to log into VMs
in the virtual network through SSH or remote desktop protocol (RDP) without exposing the VMs directly to the
internet. To see a reference architecture that uses Bastion, see Network DMZ between Azure and an on-premises
datacenter.

Assign incident notification contact

Ensure a security contact receives Azure incident notifications from Microsoft typically a notification that your
resource is compromised and/or attacking another customer.
This enables your security operations team to rapidly respond to potential security risks and remediate them.
Ensure administrator contact information in the Azure enrollment portal includes contact information that will
notify security operations (directly or rapidly via an internal process)

Regularly review critical access

Regularly review roles that are assigned privileges with a business-critical impact.
Set up a recurring review pattern to ensure that accounts are removed from permissions as roles change. You can
conduct the review manually or through an automated process by using tools such as Azure AD access reviews.

Discover and remediate common risks

Identity well-known risks for your Azure tenants, remediate those risks, and track your progress using Secure Score.
Identifying and remediating common security hygiene risks significantly reduces overall risk to your organization
by increasing cost to attackers. When you remove cheap and well-established attack vectors, attackers are forced to
acquire and use advanced or untested attack methods.
Azure Secure Score in Azure Security Center monitors the security posture of machines, networks, storage and data
services, and applications to discover potential security issues (internet connected VMs, or missing security updates,
missing endpoint protection or encryption, deviations from baseline security configurations, missing Web
Application Firewall (WAF), and more). You should enable this capability (no additional cost), review the findings,
and follow the included recommendations to plan and execute technical remediations starting with the highest
priority items.
As you address risks, track progress and prioritize ongoing investments in your governance and risk reduction
programs.

Increase automation with Azure Blueprints

Use Azure’s native automation capabilities to increase consistency, compliance, and deployment speed for
workloads.
Automation of deployment and maintenance tasks reduces security and compliance risk by limiting opportunity to
introduce human errors during manual tasks. This will also allow both IT Operations teams and security teams to
shift their focus from repeated manual tasks to higher value tasks like enabling developers and business initiatives,
protecting information, and so on.
Utilize the Azure Blueprint service to rapidly and consistently deploy application environments that are compliant
with your organization’s policies and external regulations. Azure Blueprint Service automates deployment of
environments including RBAC roles, policies, resources (VM/Net/Storage/etc.), and more. Azure Blueprints builds on
Microsoft’s significant investment into the Azure Resource Manager to standardize resource deployment in Azure
and enable resource deployment and governance based on a desired-state approach. You can use built in
configurations in Azure Blueprint, make your own, or just use Resource Manager scripts for smaller scope.
Several Security and Compliance Blueprints samples are available to use as a starting template.

Evaluate security using benchmarks

Use an industry standard benchmark to evaluate your organizations current security posture.
Benchmarking allows you to improve your security program by learning from external organizations.
Benchmarking lets you know how your current security state compares to that of other organizations, providing
both external validation for successful elements of your current system as well as identifying gaps that serve as
opportunities to enrich your team’s overall security strategy. Even if your security program isn’t tied to a specific
benchmark or regulatory standard, you will benefit from understanding the documented ideal states by those
outside and inside of your industry.
As an example, the Center for Internet Security (CIS) has created security benchmarks for Azure that map to the
CIS Control Framework. Another reference example is the MITRE ATT&CK™ framework that defines the various
adversary tactics and techniques based on real-world observations. These external references control mappings
help you to understand any gaps between your current strategy what you have and what other experts in the
industry.

Audit and enforce policy compliance

Ensure that the security team is auditing the environment to report on compliance with the security policy of the
organization. Security teams may also enforce compliance with these policies.
Organizations of all sizes will have security compliance requirements. Industry, government, and internal corporate
security policies all need to be audited and enforced. Policy monitoring is critical to check that initial configurations
are correct and that it continues to be compliant over time.
In Azure, you can take advantage of Azure Policy to create and manage policies that enforce compliance. Like Azure
Blueprints, Azure Policies are built on the underlying Azure Resource Manager capabilities in the Azure platform
(and Azure Policy can also be assigned via Azure Blueprints).
For more information on how to do this in Azure, please review Tutorial: Create and manage policies to
enforce compliance .

Monitor identity Risk

Monitor identity-related risk events for warning on potentially compromised identities and remediate those risks.
Most security incidents take place after an attacker initially gains access using a stolen identity. These identities can
often start with low privileges, but the attackers then use that identity to traverse laterally and gain access to more
privileged identities. This repeats as needed until the attacker controls access to the ultimate target data or systems.
Azure Active Directory uses adaptive machine learning algorithms, heuristics, and known compromised credentials
(username/password pairs) to detect suspicious actions that are related to your user accounts. These
username/password pairs come from monitoring public and dark web sites (where attackers often dump
compromised passwords) and by working with security researchers, law enforcement, Security teams at Microsoft,
and others.
There are two places where you review reported risk events:
Azure AD repor ting - Risk events are part of Azure AD's security reports. For more information, see
the users at risk security report and the risky sign-ins security report.
Azure AD Identity Protection - Risk events are also part of the reporting capabilities of Azure Active
Directory Identity Protection.
In addition, you can use the Identity Protection risk events API to gain programmatic access to security detections
using Microsoft Graph.
Remediate these risks by manually addressing each reported account or by setting up a user risk policy to require a
password change for these high risk events.

Penetration testing
Use Penetration Testing to validate security defenses.
Real world validation of security defenses is critical to validate your defense strategy and implementation. This can
be accomplished by a penetration test (simulates a one time attack) or a red team program (simulates a persistent
threat actor targeting your environment).
Follow the guidance published by Microsoft for planning and executing simulated attacks.

Discover & replace insecure protocols

Discover and disable the use of legacy insecure protocols SMBv1, LM/NTLMv1, wDigest, Unsigned LDAP Binds, and
Weak ciphers in Kerberos.
Authentication protocols are a critical foundation of nearly all security assurances. These older versions can be
exploited by attackers with access to your network and are often used extensively on legacy systems on
Infrastructure as a Service (IaaS).
Here are ways to reduce your risk:
Discover protocol usage by reviewing logs with Azure Sentinel’s Insecure Protocol Dashboard or third-party
tools.
Restrict or Disable use of these protocols by following guidance for SMB, NTLM, WDigest
We recommend implementing changes using pilot or other testing method to mitigate risk of operational
interruption.

Elevated security capabilities

Consider whether to utilize specialized security capabilities in your enterprise architecture.
These measures have the potential to enhance security and meet regulatory requirements, but can introduce
complexity that may negatively impact your operations and efficiency.
We recommend careful consideration and judicious use of these security measures as required:
Dedicated Hardware Security Modules (HSMs)
Dedicated Hardware Security Modules (HSMs) may help meet regulatory or security requirements.
Confidential Computing
Confidential Computing may help meet regulatory or security requirements.
 Governance, risk, and compliance
12/18/2020 • 22 minutes to read • Edit Online

Organizations of all sizes are constrained by their available resources; financial, people, and time. To achieve an
effective return on investment (ROI) organizations must prioritize where they will invest. Implementation of security
across the organization is also constrained by this, so to achieve an appropriate ROI on security the organization
needs to first understand and define its security priorities.
Governance: How is the organization’s security going to be monitored, audited, and reported? Design and
implementation of security controls within an organization is only the beginning of the story. How does the
organization know that things are actually working? Are they improving? Are there new requirements? Is there
mandatory reporting? Similar to compliance there may be external industry, government or regulatory standards
that need to be considered.
Risk : What types of risks does the organization face while trying to protect identifiable information, Intellectual
Property (IP), financial information? Who may be interested or could use this information if stolen, including
external and internal threats as well as unintentional or malicious? A commonly forgotten but extremely important
consideration within risk is addressing Disaster Recovery and Business Continuity.
Compliance: Is there a specific industry, government, or regulatory requirements that dictate or provide
recommendation on criteria that your organization’s security controls must meet? Examples of such standards,
organizations, controls, and legislation are ISO27001, NIST, PCI-DSS.
The collective role of organization(s) is to manage the security standards of the organization through their lifecycle:
Define - Set organizational standards and policies for practices, technologies, and configurations based on
internal factors (organizational culture, risk appetite, asset valuation, business initiatives, etc.) and external
factors (benchmarks, regulatory standards, threat environment, and more)
Improve – Continually push these standards incrementally forward towards the ideal state to ensure
continual risk reduction.
Sustain – Ensure the security posture doesn’t degrade naturally over time by instituting auditing and
monitoring compliance with organizational standards.

Prioritize security best practices investments

Security best practices are ideally applied proactively and completely to all systems as you build your cloud
program, but this isn’t reality for most enterprise organizations. Business goals, project constraints, and other
factors often cause organizations to balance security risk against other risks and apply a subset of best practices at
any given point.
We recommend applying as many as of the best practices as early as possible, and then working to retrofit any
gaps over time as you mature your security program. We recommend evaluating the following considerations
when prioritizing which to follow first:
High business impact and highly exposed systems: These include systems with direct intrinsic value as
well as the systems that provide attackers a path to them. For more information, see Identify and classify
business critical applications.
Easiest to implement mitigations: Identify quick wins by prioritizing the best practices, which your
organization can execute quickly because you already have the required skills, tools, and knowledge to do it
(for example, implementing a Web App Firewall (WAF) to protect a legacy application). Be careful not to
 exclusively use (or overuse) this short-term prioritization method. Doing so can increase your risk by
preventing your program from growing and leaving critical risks exposed for extended periods.
Microsoft has provided some prioritized lists of security initiatives to help organizations start with these decisions
based on our experience with threats and mitigation initiatives in our own environments and across our customers.
See Module 4a of the Microsoft CISO Workshop.

Operationalize Azure Secure Score

Assign stakeholders to use Secure Score in Azure Security Center to monitor risk profile and continuously improve
security posture. This will help you continuously improve your security posture by rapidly identifying and
remediating common security hygiene risks, which can significantly reduce your organization's risk. Set up a
regular cadence (typically monthly) to review Azure secure score and plan initiatives with specific improvement
goals. You should also look into gamifying the activity if possible to increase engagement and focus from the
responsible teams (e.g. creating a friendly competition on who increased their score more).
NOTE: Security posture management is an emerging function, this may be a single dedicated team or the
responsibility may be taken on by one or more existing teams described below.
Monitor Secure Score
SC O RE A REA RESP O N SIB L E T EA M

Monitor Secure Score and Follow up with technical teams Security Posture Management Team
Vulnerability Management (or Governance/Risk/Compliance
team)
Architecture Team
Responsible Technical Team (listed below)

Improve Secure Score

C O M P UT E A N D A P P S RESO URC ES RESP O N SIB L E T EC H N IC A L T EA M

App Services Application Development/Security Team(s)

Containers Application Development and/or Infrastructure/IT Operations

VMs/Scale sets/compute IT/Infrastructure Operations

DATA & STO RA GE RESO URC ES RESP O N SIB L E T EC H N IC A L T EA M

SQL/Redis/Data Lake Analytics/Data Lake Store Database Team

Storage Accounts Storage/Infrastructure Team

IDEN T IT Y A N D A C C ESS RESO URC ES RESP O N SIB L E T EC H N IC A L T EA M

Subscriptions Identity Team(s)

Key Vault Information/Data Security Team

N ET W O RK IN G RESO URC ES RESP O N SIB L E T EC H N IC A L T EA M

 N ET W O RK IN G RESO URC ES RESP O N SIB L E T EC H N IC A L T EA M

Networking Resources Networking Team

Network Security Team

IOT SEC URIT Y RESP O N SIB L E T EC H N IC A L T EA M

IoT Resources IoT Operations Team

NOTE: Note that in the DevOps model, some application teams may be responsible for their own application
resources.

The Azure Secure Score sample shows how to get your Azure Secure Score for a subscription by calling the
Azure Security Center REST API. The API methods provide the flexibility to query the data and build your own
reporting mechanism of your secure scores over time.

Manage connected tenants

Ensure your security organization is aware of all enrollments and associated subscriptions connected to your
existing environment (via ExpressRoute or Site-Site VPN) and monitoring as part of the overall enterprise.
These Azure resources are part of your enterprise environment and security organizations require visibility into
them. Security organizations need this access to assess risk and to identify whether organizational policies and
applicable regulatory requirements are being followed.
Ensure all Azure environments that connect to your production environment/network apply your organization’s
policy and IT governance controls for security. You can discover existing connected tenants using a tool provided by
Microsoft. Guidance on permissions you may assign to security is in the Assign privileges for managing the
environment section.

Clear lines of responsibility

Designate the parties responsible for specific functions in Azure
Clearly documenting and sharing the contacts responsible for each of these functions will create consistency and
facilitate communication. Based on our experience with many cloud adoption projects, this will avoid confusion that
can lead to human and automation errors that create security risk.
Designate groups (or individual roles) that will be responsible for these key functions:

GRO UP O R IN DIVIDUA L RO L E RESP O N SIB IL IT Y

Network Security Typically existing network security team. Configuration and

maintenance of Azure Firewall, Network Virtual Appliances
(and associated routing), WAFs, NSGs, ASGs, etc.

Network Management Typically existing network operations team. Enterprise-wide

virtual network and subnet allocation.

Ser ver Endpoint Security Typically IT operations, security, or jointly. Monitor and
remediate server security (patching, configuration, endpoint
security, etc.).
 GRO UP O R IN DIVIDUA L RO L E RESP O N SIB IL IT Y

Incident Monitoring and Response Typically security operations team. Investigate and remediate
security incidents in Security Information and Event
Management (SIEM) or source console.

Policy Management Typically GRC team + Architecture. Set Direction for use of
Role-Based Access Control (RBAC), Azure Security Center,
Administrator protection strategy, and Azure Policy to govern
Azure resources.

Identity Security and Standards Typically Security Team + Identity Team jointly. Set direction for
Azure AD directories, PIM/PAM usage, MFA,
password/synchronization configuration, Application Identity
Standards.

Enterprise segmentation strategy

Identify groups of resources that can be isolated from other parts of the enterprise to contain (and detect)
adversary movement within your enterprise. This unified enterprise segmentation strategy will guide all technical
teams to consistently segment access using networking, applications, identity, and any other access controls.
A clear and simple segmentation strategy helps contain risk while enabling productivity and business operations.
An enterprise segmentation strategy is defined higher than a traditional “network segmentation” security strategy.
Traditional segmentation approaches for on premises environments frequently failed to achieve their goals because
they were developed “bottom-up” by different technical teams and were not aligned well with business use cases
and application workloads. This resulted in overwhelming complexity that generates support issues and often
undermines the original purpose with broad network firewall exceptions.
Creating a unified enterprise segmentation strategy enables to guide all technical teams stakeholders (IT, Security,
Applications, etc.) Business Units that is built around the business risks and needs will increase alignment to and
understand and support sustainability of the security containment promises. This clarity and alignment will also
reduce s the risk of human errors and automation failures that can lead to security vulnerabilities, operational
downtime, or both.
While network micro-segmentation also offers promise to reduce risk (discussed more in Network Security and
Containment section), it doesn’t eliminate the need to align technical teams. Micro segmentation should be
considered after to and plans to ensure the ensuring technical teams are aligned so you can avoid a recurrence of
the internal conflicts that plagued and confusion of the on-premises network generation segmentation strategies.
Here are Microsoft's recommendations for prioritizing initiatives on containment and segmentation (based on Zero
Trust principles). These recommendations are listed in priority order by highest importance.
Ensure alignment of technical teams to a single enterprise segmentation strategy.
Invest in broadening containment by establishing a modern perimeter based on zero trust principles focused
on identity, device, applications, and other signals (to overcome limitation of network controls to protect new
resources and attack types).
Bolster network controls for legacy applications by exploring micro segmentation strategies.
A good enterprise segmentation strategy meets these criteria:
Enables Operations – Minimizes operation friction by aligning to business practices and applications
Contains Risk - Adds cost and friction to attackers by
 Isolating sensitive workloads from compromise of other assets
Isolating high exposure systems from being used as a pivot to other systems
Monitored – Security Operations should monitor for potential violations of the integrity of the segments
(account usage, unexpected traffic, etc.)

Security team visibility

Provide security teams read-only access to the security aspects of all technical resources in their purview
Security organizations require visibility into the technical environment to perform their duties of assessing and
reporting on organizational risk. Without this visibility, security will have to rely on information provided from
groups, operating the environment, who have a potential conflict of interest (and other priorities).
Note that security teams may separately be granted additional privileges if they have operational responsibilities or
a requirement to enforce compliance on Azure resources.
For example in Azure, assign security teams to the Security Readers permission that provides access to measure
security risk (without providing access to the data itself)
For enterprise security groups with broad responsibility for security of Azure, you can assign this permission using:
Root management group – for teams responsible for assessing and reporting risk on all resources
Segment management group(s) – for teams with limited scope of responsibility (typically required because
of organizational boundaries or regulatory requirements)
Because security will have broad access to the environment (and visibility into potentially exploitable
vulnerabilities), you should consider them critical impact accounts and apply the same protections as
administrators. The Administration section details these controls for Azure.

Assign privileges for managing the environment

Grant roles with operational responsibilities in Azure the appropriate permissions based on a clearly documented
strategy built from the principle of least privilege and your operational needs.
Providing clear guidance that follows a reference model will reduce risk because by increasing it provides clarity for
your technical teams implementing these permissions. This clarity makes it easier to detect and correct human
errors like overpermissioning, reducing your overall risk.
Microsoft recommends starting from these Microsoft reference models and adapting to your organization.

Core Services Reference Permissions

This segment hosts shared services utilized across the organization. These shared services typically include Active
Directory Domain Services, DNS/DHCP, System Management Tools hosted on Azure Infrastructure as a Service
(IaaS) virtual machines.
Security Visibility across all resources – For security teams, grant read-only access to security attributes for all
technical environments. This access level is needed to assess risk factors, identify potential mitigations, and advise
organizational stakeholders who accept the risk. See Security Team Visibility for more details.
Policy management across some or all resources – To monitor and enforce compliance with external (or
internal) regulations, standards, and security policy, assign appropriate permission to those roles. The roles and
permissions you choose will depend on the organizational culture and expectations of the policy program. See
Microsoft Cloud Adoption Framework for Azure.
Central IT operations across all resources – Grant permissions to the central IT department (often the
infrastructure team) to create, modify, and delete resources like virtual machines and storage.
Central networking group across network resources – To ensure consistency and avoid technical conflicts,
assign network resource responsibilities to a single central networking organization. These resources should
include virtual networks, subnets, Network Security Groups (NSG), and the virtual machines hosting virtual
network appliances. See Centralize Network Management And Security for more details
Resource Role Permissions – For most core services, administrative privileges required to manage them are
granted via the application itself (Active Directory, DNS/DHCP, System Management Tools, etc.), so no additional
Azure resource permissions are required. If your organizational model requires these teams to manage their own
VMs, storage, or other Azure resources, you can assign these permissions to those roles.
Ser vice admin (Break Glass Account) – Use the service admin role only for emergencies (and initial setup if
required). Do not use this role for daily tasks. See Emergency Access (‘Break Glass’ Accounts) for more details.
Segment reference permissions
This segment permission design provides consistency while allowing flexibility to accommodate the range of
organizational models from a single centralized IT group to mostly independent IT and DevOps teams.
Security visibility across all resources – For security teams, grant read-only access to security attributes for all
technical environments. This access level is needed to assess risk factors, identify potential mitigations, and advise
organizational stakeholders who accept the risk. See Security Team Visibility.
Policy management across some or all resources – To monitor and enforce compliance with external (or
internal) regulations, standards, and security policy assign appropriate permission to those roles. The roles and
permissions you choose will depend on the organizational culture and expectations of the policy program. See
Microsoft Cloud Adoption Framework for Azure.
IT Operations across all resources – Grant permission to create, modify, and delete resources. The purpose of
the segment (and resulting permissions) will depend on your organization structure.
Segments with resources managed by a centralized IT organization can grant the central IT department
(often the infrastructure team) permission to modify these resources.
Segments managed by independent business units or functions (such as a Human Resources IT Team) can
grant those teams permission to all resources in the segment.
Segments with autonomous DevOps teams don’t need to grant permissions across all resources because the
resource role (below) grants permissions to application teams. For emergencies, use the service admin
account (break-glass account).
Central networking group across network resources – To ensure consistency and avoid technical conflicts,
assign network resource responsibilities to a single central networking organization. These resources should
include virtual networks, subnets, Network Security Groups (NSG), and the virtual machines hosting virtual
network appliances. See Centralize Network Management And Security.
Resource Role Permissions – Segments with autonomous DevOps teams will manage the resources associated
with each application. The actual roles and their permissions depend on the application size and complexity, the
application team size and complexity, and the culture of the organization and application team.
Ser vice Admin (Break Glass Account) – Use the service admin role only for emergencies (and initial setup if
required). Do not use this role for daily tasks. See Emergency Access (‘Break Glass’ Accounts) for more details.
Permission Guidance and Tips
To drive consistency and ensure application to future subscriptions, permissions should be assigned at
management group for the segment rather than the individual subscriptions. See Avoid Granular and
Custom Permissions for more details.
You should first review the built-in roles to see if one is applicable before creating a custom role to grant the
appropriate permissions to VMs and other objects. See Use Built in Roles for more details
Security managers group membership may be appropriate for smaller teams/organizations where
security teams have extensive operational responsibilities.

Establish segmentation with management groups

Structure management groups into a simple design that guides the enterprise segmentation model.
Management groups offer the ability to consistently and efficiently manage resources (including multiple
subscriptions as needed). However, because of their flexibility, it's possible to create an overly complex design.
Complexity creates confusion and negatively impacts both operations and security (as illustrated by overly complex
Organizational Unit (OU) and Group Policy Object (GPO) designs for Active Directory).
Microsoft recommends aligning the top level of management groups (MGs) into a simple enterprise segmentation
strategy limited to 1 or 2 levels.

Use root management group carefully

Use the Root Management Group (MG) for enterprise consistency, but test changes carefully to minimize risk of
operational disruption.
The root management group enables you to ensure consistency across the enterprise by applying policies,
permissions, and tags across all subscriptions. Care must be taken when planning and implementing assignments
to the root management group because this can affect every resource on Azure and potentially cause downtime or
other negative impacts on productivity in the event of errors or unanticipated effects.
Root management group guidance:
Plan Carefully - Select enterprise-wide elements to the root management group that have a clear
requirement to be applied across every resource and/or low impact.
Good candidates include:
Regulator y requirements with clear business risk/impact (for example, restrictions related to data
sovereignty).
Near-zero potential negative impact on operations such as policy with audit effect, Tag
assignment, RBAC permissions assignments that have been carefully reviewed.
Test First - Carefully test all enterprise-wide changes on the root management group before applying
(policy, tags, RBAC model, etc.) using a
Test Lab - Representative lab tenant or lab segment in production tenant.
Production Pilot - Segment MG or Designated subset in subscription(s) / MG.
Validate Changes – to ensure they have the desired effect.

Virtual Machine (VM) security updates and strong passwords

Ensure policy and processes enable (and require) rapid application of security updates to virtual machines.
Attackers constantly scan public cloud IP ranges for open management ports and attempt “easy” attacks like
common passwords and unpatched vulnerabilities.
Enable Azure Security Center to identify missing security updates & apply them.
Local Admin Password Solution (LAPS) or a third-party Privileged Access Management can set strong local admin
passwords and just in time access to them.

Remove Virtual Machine (VM) direct internet connectivity

Make sure policies and processes require restricting and monitoring direct internet connectivity by virtual
machines.
For Azure, you can enforce policies by,
Enterprise-wide prevention - Prevent inadvertent exposure by following the permissions and roles
described in the reference model.
Ensures that network traffic is routed through approved egress points by default.
Exceptions (such as adding a public IP address to a resource) must go through a centralized group
that evaluates exception requests and makes sure appropriate controls are applied.
Identify and remediate exposed virtual machines by using the Azure Security Center network visualization
to quickly identify internet exposed resources.
Restrict management por ts (RDP, SSH) using Just in Time access in Azure Security Center.
One way of managing VMs in the virtual network is by using Azure Bastion. This service allows you to log into VMs
in the virtual network through SSH or remote desktop protocol (RDP) without exposing the VMs directly to the
internet. To see a reference architecture that uses Bastion, see Network DMZ between Azure and an on-premises
datacenter.

Assign incident notification contact

Ensure a security contact receives Azure incident notifications from Microsoft typically a notification that your
resource is compromised and/or attacking another customer.
This enables your security operations team to rapidly respond to potential security risks and remediate them.
Ensure administrator contact information in the Azure enrollment portal includes contact information that will
notify security operations (directly or rapidly via an internal process)

Regularly review critical access

Regularly review roles that are assigned privileges with a business-critical impact.
Set up a recurring review pattern to ensure that accounts are removed from permissions as roles change. You can
conduct the review manually or through an automated process by using tools such as Azure AD access reviews.

Discover and remediate common risks

Identity well-known risks for your Azure tenants, remediate those risks, and track your progress using Secure Score.
Identifying and remediating common security hygiene risks significantly reduces overall risk to your organization
by increasing cost to attackers. When you remove cheap and well-established attack vectors, attackers are forced to
acquire and use advanced or untested attack methods.
Azure Secure Score in Azure Security Center monitors the security posture of machines, networks, storage and data
services, and applications to discover potential security issues (internet connected VMs, or missing security updates,
missing endpoint protection or encryption, deviations from baseline security configurations, missing Web
Application Firewall (WAF), and more). You should enable this capability (no additional cost), review the findings,
and follow the included recommendations to plan and execute technical remediations starting with the highest
priority items.
As you address risks, track progress and prioritize ongoing investments in your governance and risk reduction
programs.

Increase automation with Azure Blueprints

Use Azure’s native automation capabilities to increase consistency, compliance, and deployment speed for
workloads.
Automation of deployment and maintenance tasks reduces security and compliance risk by limiting opportunity to
introduce human errors during manual tasks. This will also allow both IT Operations teams and security teams to
shift their focus from repeated manual tasks to higher value tasks like enabling developers and business initiatives,
protecting information, and so on.
Utilize the Azure Blueprint service to rapidly and consistently deploy application environments that are compliant
with your organization’s policies and external regulations. Azure Blueprint Service automates deployment of
environments including RBAC roles, policies, resources (VM/Net/Storage/etc.), and more. Azure Blueprints builds on
Microsoft’s significant investment into the Azure Resource Manager to standardize resource deployment in Azure
and enable resource deployment and governance based on a desired-state approach. You can use built in
configurations in Azure Blueprint, make your own, or just use Resource Manager scripts for smaller scope.
Several Security and Compliance Blueprints samples are available to use as a starting template.

Evaluate security using benchmarks

Use an industry standard benchmark to evaluate your organizations current security posture.
Benchmarking allows you to improve your security program by learning from external organizations.
Benchmarking lets you know how your current security state compares to that of other organizations, providing
both external validation for successful elements of your current system as well as identifying gaps that serve as
opportunities to enrich your team’s overall security strategy. Even if your security program isn’t tied to a specific
benchmark or regulatory standard, you will benefit from understanding the documented ideal states by those
outside and inside of your industry.
As an example, the Center for Internet Security (CIS) has created security benchmarks for Azure that map to the
CIS Control Framework. Another reference example is the MITRE ATT&CK™ framework that defines the various
adversary tactics and techniques based on real-world observations. These external references control mappings
help you to understand any gaps between your current strategy what you have and what other experts in the
industry.

Audit and enforce policy compliance

Ensure that the security team is auditing the environment to report on compliance with the security policy of the
organization. Security teams may also enforce compliance with these policies.
Organizations of all sizes will have security compliance requirements. Industry, government, and internal corporate
security policies all need to be audited and enforced. Policy monitoring is critical to check that initial configurations
are correct and that it continues to be compliant over time.
In Azure, you can take advantage of Azure Policy to create and manage policies that enforce compliance. Like Azure
Blueprints, Azure Policies are built on the underlying Azure Resource Manager capabilities in the Azure platform
(and Azure Policy can also be assigned via Azure Blueprints).
For more information on how to do this in Azure, please review Tutorial: Create and manage policies to
enforce compliance .

Monitor identity Risk

Monitor identity-related risk events for warning on potentially compromised identities and remediate those risks.
Most security incidents take place after an attacker initially gains access using a stolen identity. These identities can
often start with low privileges, but the attackers then use that identity to traverse laterally and gain access to more
privileged identities. This repeats as needed until the attacker controls access to the ultimate target data or systems.
Azure Active Directory uses adaptive machine learning algorithms, heuristics, and known compromised credentials
(username/password pairs) to detect suspicious actions that are related to your user accounts. These
username/password pairs come from monitoring public and dark web sites (where attackers often dump
compromised passwords) and by working with security researchers, law enforcement, Security teams at Microsoft,
and others.
There are two places where you review reported risk events:
Azure AD repor ting - Risk events are part of Azure AD's security reports. For more information, see
the users at risk security report and the risky sign-ins security report.
Azure AD Identity Protection - Risk events are also part of the reporting capabilities of Azure Active
Directory Identity Protection.
In addition, you can use the Identity Protection risk events API to gain programmatic access to security detections
using Microsoft Graph.
Remediate these risks by manually addressing each reported account or by setting up a user risk policy to require a
password change for these high risk events.

Penetration testing
Use Penetration Testing to validate security defenses.
Real world validation of security defenses is critical to validate your defense strategy and implementation. This can
be accomplished by a penetration test (simulates a one time attack) or a red team program (simulates a persistent
threat actor targeting your environment).
Follow the guidance published by Microsoft for planning and executing simulated attacks.

Discover & replace insecure protocols

Discover and disable the use of legacy insecure protocols SMBv1, LM/NTLMv1, wDigest, Unsigned LDAP Binds, and
Weak ciphers in Kerberos.
Authentication protocols are a critical foundation of nearly all security assurances. These older versions can be
exploited by attackers with access to your network and are often used extensively on legacy systems on
Infrastructure as a Service (IaaS).
Here are ways to reduce your risk:
Discover protocol usage by reviewing logs with Azure Sentinel’s Insecure Protocol Dashboard or third-party
tools.
Restrict or Disable use of these protocols by following guidance for SMB, NTLM, WDigest
We recommend implementing changes using pilot or other testing method to mitigate risk of operational
interruption.

Elevated security capabilities

Consider whether to utilize specialized security capabilities in your enterprise architecture.
These measures have the potential to enhance security and meet regulatory requirements, but can introduce
complexity that may negatively impact your operations and efficiency.
We recommend careful consideration and judicious use of these security measures as required:
Dedicated Hardware Security Modules (HSMs)
Dedicated Hardware Security Modules (HSMs) may help meet regulatory or security requirements.
Confidential Computing
Confidential Computing may help meet regulatory or security requirements.
 Governance, risk, and compliance
12/18/2020 • 22 minutes to read • Edit Online

Organizations of all sizes are constrained by their available resources; financial, people, and time. To achieve an
effective return on investment (ROI) organizations must prioritize where they will invest. Implementation of
security across the organization is also constrained by this, so to achieve an appropriate ROI on security the
organization needs to first understand and define its security priorities.
Governance: How is the organization’s security going to be monitored, audited, and reported? Design and
implementation of security controls within an organization is only the beginning of the story. How does the
organization know that things are actually working? Are they improving? Are there new requirements? Is there
mandatory reporting? Similar to compliance there may be external industry, government or regulatory standards
that need to be considered.
Risk : What types of risks does the organization face while trying to protect identifiable information, Intellectual
Property (IP), financial information? Who may be interested or could use this information if stolen, including
external and internal threats as well as unintentional or malicious? A commonly forgotten but extremely important
consideration within risk is addressing Disaster Recovery and Business Continuity.
Compliance: Is there a specific industry, government, or regulatory requirements that dictate or provide
recommendation on criteria that your organization’s security controls must meet? Examples of such standards,
organizations, controls, and legislation are ISO27001, NIST, PCI-DSS.
The collective role of organization(s) is to manage the security standards of the organization through their
lifecycle:
Define - Set organizational standards and policies for practices, technologies, and configurations based on
internal factors (organizational culture, risk appetite, asset valuation, business initiatives, etc.) and external
factors (benchmarks, regulatory standards, threat environment, and more)
Improve – Continually push these standards incrementally forward towards the ideal state to ensure
continual risk reduction.
Sustain – Ensure the security posture doesn’t degrade naturally over time by instituting auditing and
monitoring compliance with organizational standards.

Prioritize security best practices investments

Security best practices are ideally applied proactively and completely to all systems as you build your cloud
program, but this isn’t reality for most enterprise organizations. Business goals, project constraints, and other
factors often cause organizations to balance security risk against other risks and apply a subset of best practices at
any given point.
We recommend applying as many as of the best practices as early as possible, and then working to retrofit any
gaps over time as you mature your security program. We recommend evaluating the following considerations
when prioritizing which to follow first:
High business impact and highly exposed systems: These include systems with direct intrinsic value
as well as the systems that provide attackers a path to them. For more information, see Identify and classify
business critical applications.
Easiest to implement mitigations: Identify quick wins by prioritizing the best practices, which your
organization can execute quickly because you already have the required skills, tools, and knowledge to do it
 (for example, implementing a Web App Firewall (WAF) to protect a legacy application). Be careful not to
exclusively use (or overuse) this short-term prioritization method. Doing so can increase your risk by
preventing your program from growing and leaving critical risks exposed for extended periods.
Microsoft has provided some prioritized lists of security initiatives to help organizations start with these decisions
based on our experience with threats and mitigation initiatives in our own environments and across our
customers. See Module 4a of the Microsoft CISO Workshop.

Operationalize Azure Secure Score

Assign stakeholders to use Secure Score in Azure Security Center to monitor risk profile and continuously improve
security posture. This will help you continuously improve your security posture by rapidly identifying and
remediating common security hygiene risks, which can significantly reduce your organization's risk. Set up a
regular cadence (typically monthly) to review Azure secure score and plan initiatives with specific improvement
goals. You should also look into gamifying the activity if possible to increase engagement and focus from the
responsible teams (e.g. creating a friendly competition on who increased their score more).
NOTE: Security posture management is an emerging function, this may be a single dedicated team or the
responsibility may be taken on by one or more existing teams described below.
Monitor Secure Score
SC O RE A REA RESP O N SIB L E T EA M

Monitor Secure Score and Follow up with technical teams Security Posture Management Team
Vulnerability Management (or Governance/Risk/Compliance
team)
Architecture Team
Responsible Technical Team (listed below)

Improve Secure Score

C O M P UT E A N D A P P S RESO URC ES RESP O N SIB L E T EC H N IC A L T EA M

App Services Application Development/Security Team(s)

Containers Application Development and/or Infrastructure/IT Operations

VMs/Scale sets/compute IT/Infrastructure Operations

DATA & STO RA GE RESO URC ES RESP O N SIB L E T EC H N IC A L T EA M

SQL/Redis/Data Lake Analytics/Data Lake Store Database Team

Storage Accounts Storage/Infrastructure Team

IDEN T IT Y A N D A C C ESS RESO URC ES RESP O N SIB L E T EC H N IC A L T EA M

Subscriptions Identity Team(s)

Key Vault Information/Data Security Team

 N ET W O RK IN G RESO URC ES RESP O N SIB L E T EC H N IC A L T EA M

Networking Resources Networking Team

Network Security Team

IOT SEC URIT Y RESP O N SIB L E T EC H N IC A L T EA M

IoT Resources IoT Operations Team

NOTE: Note that in the DevOps model, some application teams may be responsible for their own application
resources.

The Azure Secure Score sample shows how to get your Azure Secure Score for a subscription by calling the
Azure Security Center REST API. The API methods provide the flexibility to query the data and build your own
reporting mechanism of your secure scores over time.

Manage connected tenants

Ensure your security organization is aware of all enrollments and associated subscriptions connected to your
existing environment (via ExpressRoute or Site-Site VPN) and monitoring as part of the overall enterprise.
These Azure resources are part of your enterprise environment and security organizations require visibility into
them. Security organizations need this access to assess risk and to identify whether organizational policies and
applicable regulatory requirements are being followed.
Ensure all Azure environments that connect to your production environment/network apply your organization’s
policy and IT governance controls for security. You can discover existing connected tenants using a tool provided
by Microsoft. Guidance on permissions you may assign to security is in the Assign privileges for managing the
environment section.

Clear lines of responsibility

Designate the parties responsible for specific functions in Azure
Clearly documenting and sharing the contacts responsible for each of these functions will create consistency and
facilitate communication. Based on our experience with many cloud adoption projects, this will avoid confusion
that can lead to human and automation errors that create security risk.
Designate groups (or individual roles) that will be responsible for these key functions:

GRO UP O R IN DIVIDUA L RO L E RESP O N SIB IL IT Y

Network Security Typically existing network security team. Configuration and

maintenance of Azure Firewall, Network Virtual Appliances
(and associated routing), WAFs, NSGs, ASGs, etc.

Network Management Typically existing network operations team. Enterprise-wide

virtual network and subnet allocation.

Ser ver Endpoint Security Typically IT operations, security, or jointly. Monitor and
remediate server security (patching, configuration, endpoint
security, etc.).
 GRO UP O R IN DIVIDUA L RO L E RESP O N SIB IL IT Y

Incident Monitoring and Response Typically security operations team. Investigate and remediate
security incidents in Security Information and Event
Management (SIEM) or source console.

Policy Management Typically GRC team + Architecture. Set Direction for use of
Role-Based Access Control (RBAC), Azure Security Center,
Administrator protection strategy, and Azure Policy to govern
Azure resources.

Identity Security and Standards Typically Security Team + Identity Team jointly. Set direction
for Azure AD directories, PIM/PAM usage, MFA,
password/synchronization configuration, Application Identity
Standards.

Enterprise segmentation strategy

Identify groups of resources that can be isolated from other parts of the enterprise to contain (and detect)
adversary movement within your enterprise. This unified enterprise segmentation strategy will guide all technical
teams to consistently segment access using networking, applications, identity, and any other access controls.
A clear and simple segmentation strategy helps contain risk while enabling productivity and business operations.
An enterprise segmentation strategy is defined higher than a traditional “network segmentation” security strategy.
Traditional segmentation approaches for on premises environments frequently failed to achieve their goals
because they were developed “bottom-up” by different technical teams and were not aligned well with business
use cases and application workloads. This resulted in overwhelming complexity that generates support issues and
often undermines the original purpose with broad network firewall exceptions.
Creating a unified enterprise segmentation strategy enables to guide all technical teams stakeholders (IT, Security,
Applications, etc.) Business Units that is built around the business risks and needs will increase alignment to and
understand and support sustainability of the security containment promises. This clarity and alignment will also
reduce s the risk of human errors and automation failures that can lead to security vulnerabilities, operational
downtime, or both.
While network micro-segmentation also offers promise to reduce risk (discussed more in Network Security and
Containment section), it doesn’t eliminate the need to align technical teams. Micro segmentation should be
considered after to and plans to ensure the ensuring technical teams are aligned so you can avoid a recurrence of
the internal conflicts that plagued and confusion of the on-premises network generation segmentation strategies.
Here are Microsoft's recommendations for prioritizing initiatives on containment and segmentation (based on
Zero Trust principles). These recommendations are listed in priority order by highest importance.
Ensure alignment of technical teams to a single enterprise segmentation strategy.
Invest in broadening containment by establishing a modern perimeter based on zero trust principles
focused on identity, device, applications, and other signals (to overcome limitation of network controls to
protect new resources and attack types).
Bolster network controls for legacy applications by exploring micro segmentation strategies.
A good enterprise segmentation strategy meets these criteria:
Enables Operations – Minimizes operation friction by aligning to business practices and applications
Contains Risk - Adds cost and friction to attackers by
 Isolating sensitive workloads from compromise of other assets
Isolating high exposure systems from being used as a pivot to other systems
Monitored – Security Operations should monitor for potential violations of the integrity of the segments
(account usage, unexpected traffic, etc.)

Security team visibility

Provide security teams read-only access to the security aspects of all technical resources in their purview
Security organizations require visibility into the technical environment to perform their duties of assessing and
reporting on organizational risk. Without this visibility, security will have to rely on information provided from
groups, operating the environment, who have a potential conflict of interest (and other priorities).
Note that security teams may separately be granted additional privileges if they have operational responsibilities
or a requirement to enforce compliance on Azure resources.
For example in Azure, assign security teams to the Security Readers permission that provides access to measure
security risk (without providing access to the data itself)
For enterprise security groups with broad responsibility for security of Azure, you can assign this permission
using:
Root management group – for teams responsible for assessing and reporting risk on all resources
Segment management group(s) – for teams with limited scope of responsibility (typically required because
of organizational boundaries or regulatory requirements)
Because security will have broad access to the environment (and visibility into potentially exploitable
vulnerabilities), you should consider them critical impact accounts and apply the same protections as
administrators. The Administration section details these controls for Azure.

Assign privileges for managing the environment

Grant roles with operational responsibilities in Azure the appropriate permissions based on a clearly documented
strategy built from the principle of least privilege and your operational needs.
Providing clear guidance that follows a reference model will reduce risk because by increasing it provides clarity
for your technical teams implementing these permissions. This clarity makes it easier to detect and correct human
errors like overpermissioning, reducing your overall risk.
Microsoft recommends starting from these Microsoft reference models and adapting to your organization.

Core Services Reference Permissions

This segment hosts shared services utilized across the organization. These shared services typically include Active
Directory Domain Services, DNS/DHCP, System Management Tools hosted on Azure Infrastructure as a Service
(IaaS) virtual machines.
Security Visibility across all resources – For security teams, grant read-only access to security attributes for
all technical environments. This access level is needed to assess risk factors, identify potential mitigations, and
advise organizational stakeholders who accept the risk. See Security Team Visibility for more details.
Policy management across some or all resources – To monitor and enforce compliance with external (or
internal) regulations, standards, and security policy, assign appropriate permission to those roles. The roles and
permissions you choose will depend on the organizational culture and expectations of the policy program. See
Microsoft Cloud Adoption Framework for Azure.
Central IT operations across all resources – Grant permissions to the central IT department (often the
infrastructure team) to create, modify, and delete resources like virtual machines and storage.
Central networking group across network resources – To ensure consistency and avoid technical conflicts,
assign network resource responsibilities to a single central networking organization. These resources should
include virtual networks, subnets, Network Security Groups (NSG), and the virtual machines hosting virtual
network appliances. See Centralize Network Management And Security for more details
Resource Role Permissions – For most core services, administrative privileges required to manage them are
granted via the application itself (Active Directory, DNS/DHCP, System Management Tools, etc.), so no additional
Azure resource permissions are required. If your organizational model requires these teams to manage their own
VMs, storage, or other Azure resources, you can assign these permissions to those roles.
Ser vice admin (Break Glass Account) – Use the service admin role only for emergencies (and initial setup if
required). Do not use this role for daily tasks. See Emergency Access (‘Break Glass’ Accounts) for more details.
Segment reference permissions
This segment permission design provides consistency while allowing flexibility to accommodate the range of
organizational models from a single centralized IT group to mostly independent IT and DevOps teams.
Security visibility across all resources – For security teams, grant read-only access to security attributes for
all technical environments. This access level is needed to assess risk factors, identify potential mitigations, and
advise organizational stakeholders who accept the risk. See Security Team Visibility.
Policy management across some or all resources – To monitor and enforce compliance with external (or
internal) regulations, standards, and security policy assign appropriate permission to those roles. The roles and
permissions you choose will depend on the organizational culture and expectations of the policy program. See
Microsoft Cloud Adoption Framework for Azure.
IT Operations across all resources – Grant permission to create, modify, and delete resources. The purpose of
the segment (and resulting permissions) will depend on your organization structure.
Segments with resources managed by a centralized IT organization can grant the central IT department
(often the infrastructure team) permission to modify these resources.
Segments managed by independent business units or functions (such as a Human Resources IT Team) can
grant those teams permission to all resources in the segment.
Segments with autonomous DevOps teams don’t need to grant permissions across all resources because
the resource role (below) grants permissions to application teams. For emergencies, use the service admin
account (break-glass account).
Central networking group across network resources – To ensure consistency and avoid technical conflicts,
assign network resource responsibilities to a single central networking organization. These resources should
include virtual networks, subnets, Network Security Groups (NSG), and the virtual machines hosting virtual
network appliances. See Centralize Network Management And Security.
Resource Role Permissions – Segments with autonomous DevOps teams will manage the resources associated
with each application. The actual roles and their permissions depend on the application size and complexity, the
application team size and complexity, and the culture of the organization and application team.
Ser vice Admin (Break Glass Account) – Use the service admin role only for emergencies (and initial setup if
required). Do not use this role for daily tasks. See Emergency Access (‘Break Glass’ Accounts) for more details.
Permission Guidance and Tips
To drive consistency and ensure application to future subscriptions, permissions should be assigned at
management group for the segment rather than the individual subscriptions. See Avoid Granular and
Custom Permissions for more details.
You should first review the built-in roles to see if one is applicable before creating a custom role to grant the
appropriate permissions to VMs and other objects. See Use Built in Roles for more details
Security managers group membership may be appropriate for smaller teams/organizations where
security teams have extensive operational responsibilities.

Establish segmentation with management groups

Structure management groups into a simple design that guides the enterprise segmentation model.
Management groups offer the ability to consistently and efficiently manage resources (including multiple
subscriptions as needed). However, because of their flexibility, it's possible to create an overly complex design.
Complexity creates confusion and negatively impacts both operations and security (as illustrated by overly
complex Organizational Unit (OU) and Group Policy Object (GPO) designs for Active Directory).
Microsoft recommends aligning the top level of management groups (MGs) into a simple enterprise segmentation
strategy limited to 1 or 2 levels.

Use root management group carefully

Use the Root Management Group (MG) for enterprise consistency, but test changes carefully to minimize risk of
operational disruption.
The root management group enables you to ensure consistency across the enterprise by applying policies,
permissions, and tags across all subscriptions. Care must be taken when planning and implementing assignments
to the root management group because this can affect every resource on Azure and potentially cause downtime or
other negative impacts on productivity in the event of errors or unanticipated effects.
Root management group guidance:
Plan Carefully - Select enterprise-wide elements to the root management group that have a clear
requirement to be applied across every resource and/or low impact.
Good candidates include:
Regulator y requirements with clear business risk/impact (for example, restrictions related to data
sovereignty).
Near-zero potential negative impact on operations such as policy with audit effect, Tag
assignment, RBAC permissions assignments that have been carefully reviewed.
Test First - Carefully test all enterprise-wide changes on the root management group before applying
(policy, tags, RBAC model, etc.) using a
Test Lab - Representative lab tenant or lab segment in production tenant.
Production Pilot - Segment MG or Designated subset in subscription(s) / MG.
Validate Changes – to ensure they have the desired effect.

Virtual Machine (VM) security updates and strong passwords

Ensure policy and processes enable (and require) rapid application of security updates to virtual machines.
Attackers constantly scan public cloud IP ranges for open management ports and attempt “easy” attacks like
common passwords and unpatched vulnerabilities.
Enable Azure Security Center to identify missing security updates & apply them.
Local Admin Password Solution (LAPS) or a third-party Privileged Access Management can set strong local admin
passwords and just in time access to them.

Remove Virtual Machine (VM) direct internet connectivity

Make sure policies and processes require restricting and monitoring direct internet connectivity by virtual
machines.
For Azure, you can enforce policies by,
Enterprise-wide prevention - Prevent inadvertent exposure by following the permissions and roles
described in the reference model.
Ensures that network traffic is routed through approved egress points by default.
Exceptions (such as adding a public IP address to a resource) must go through a centralized group
that evaluates exception requests and makes sure appropriate controls are applied.
Identify and remediate exposed virtual machines by using the Azure Security Center network
visualization to quickly identify internet exposed resources.
Restrict management por ts (RDP, SSH) using Just in Time access in Azure Security Center.
One way of managing VMs in the virtual network is by using Azure Bastion. This service allows you to log into
VMs in the virtual network through SSH or remote desktop protocol (RDP) without exposing the VMs directly to
the internet. To see a reference architecture that uses Bastion, see Network DMZ between Azure and an on-
premises datacenter.

Assign incident notification contact

Ensure a security contact receives Azure incident notifications from Microsoft typically a notification that your
resource is compromised and/or attacking another customer.
This enables your security operations team to rapidly respond to potential security risks and remediate them.
Ensure administrator contact information in the Azure enrollment portal includes contact information that will
notify security operations (directly or rapidly via an internal process)

Regularly review critical access

Regularly review roles that are assigned privileges with a business-critical impact.
Set up a recurring review pattern to ensure that accounts are removed from permissions as roles change. You can
conduct the review manually or through an automated process by using tools such as Azure AD access reviews.

Discover and remediate common risks

Identity well-known risks for your Azure tenants, remediate those risks, and track your progress using Secure
Score.
Identifying and remediating common security hygiene risks significantly reduces overall risk to your organization
by increasing cost to attackers. When you remove cheap and well-established attack vectors, attackers are forced
to acquire and use advanced or untested attack methods.
Azure Secure Score in Azure Security Center monitors the security posture of machines, networks, storage and
data services, and applications to discover potential security issues (internet connected VMs, or missing security
updates, missing endpoint protection or encryption, deviations from baseline security configurations, missing Web
Application Firewall (WAF), and more). You should enable this capability (no additional cost), review the findings,
and follow the included recommendations to plan and execute technical remediations starting with the highest
priority items.
As you address risks, track progress and prioritize ongoing investments in your governance and risk reduction
programs.

Increase automation with Azure Blueprints

Use Azure’s native automation capabilities to increase consistency, compliance, and deployment speed for
workloads.
Automation of deployment and maintenance tasks reduces security and compliance risk by limiting opportunity to
introduce human errors during manual tasks. This will also allow both IT Operations teams and security teams to
shift their focus from repeated manual tasks to higher value tasks like enabling developers and business initiatives,
protecting information, and so on.
Utilize the Azure Blueprint service to rapidly and consistently deploy application environments that are compliant
with your organization’s policies and external regulations. Azure Blueprint Service automates deployment of
environments including RBAC roles, policies, resources (VM/Net/Storage/etc.), and more. Azure Blueprints builds
on Microsoft’s significant investment into the Azure Resource Manager to standardize resource deployment in
Azure and enable resource deployment and governance based on a desired-state approach. You can use built in
configurations in Azure Blueprint, make your own, or just use Resource Manager scripts for smaller scope.
Several Security and Compliance Blueprints samples are available to use as a starting template.

Evaluate security using benchmarks

Use an industry standard benchmark to evaluate your organizations current security posture.
Benchmarking allows you to improve your security program by learning from external organizations.
Benchmarking lets you know how your current security state compares to that of other organizations, providing
both external validation for successful elements of your current system as well as identifying gaps that serve as
opportunities to enrich your team’s overall security strategy. Even if your security program isn’t tied to a specific
benchmark or regulatory standard, you will benefit from understanding the documented ideal states by those
outside and inside of your industry.
As an example, the Center for Internet Security (CIS) has created security benchmarks for Azure that map to the
CIS Control Framework. Another reference example is the MITRE ATT&CK™ framework that defines the
various adversary tactics and techniques based on real-world observations. These external references control
mappings help you to understand any gaps between your current strategy what you have and what other
experts in the industry.

Audit and enforce policy compliance

Ensure that the security team is auditing the environment to report on compliance with the security policy of the
organization. Security teams may also enforce compliance with these policies.
Organizations of all sizes will have security compliance requirements. Industry, government, and internal
corporate security policies all need to be audited and enforced. Policy monitoring is critical to check that initial
configurations are correct and that it continues to be compliant over time.
In Azure, you can take advantage of Azure Policy to create and manage policies that enforce compliance. Like
Azure Blueprints, Azure Policies are built on the underlying Azure Resource Manager capabilities in the Azure
platform (and Azure Policy can also be assigned via Azure Blueprints).
For more information on how to do this in Azure, please review Tutorial: Create and manage policies to
enforce compliance .

Monitor identity Risk

Monitor identity-related risk events for warning on potentially compromised identities and remediate those risks.
Most security incidents take place after an attacker initially gains access using a stolen identity. These identities can
often start with low privileges, but the attackers then use that identity to traverse laterally and gain access to more
privileged identities. This repeats as needed until the attacker controls access to the ultimate target data or
systems.
Azure Active Directory uses adaptive machine learning algorithms, heuristics, and known compromised
credentials (username/password pairs) to detect suspicious actions that are related to your user accounts. These
username/password pairs come from monitoring public and dark web sites (where attackers often dump
compromised passwords) and by working with security researchers, law enforcement, Security teams at Microsoft,
and others.
There are two places where you review reported risk events:
Azure AD repor ting - Risk events are part of Azure AD's security reports. For more information, see
the users at risk security report and the risky sign-ins security report.
Azure AD Identity Protection - Risk events are also part of the reporting capabilities of Azure Active
Directory Identity Protection.
In addition, you can use the Identity Protection risk events API to gain programmatic access to security detections
using Microsoft Graph.
Remediate these risks by manually addressing each reported account or by setting up a user risk policy to require
a password change for these high risk events.

Penetration testing
Use Penetration Testing to validate security defenses.
Real world validation of security defenses is critical to validate your defense strategy and implementation. This can
be accomplished by a penetration test (simulates a one time attack) or a red team program (simulates a persistent
threat actor targeting your environment).
Follow the guidance published by Microsoft for planning and executing simulated attacks.

Discover & replace insecure protocols

Discover and disable the use of legacy insecure protocols SMBv1, LM/NTLMv1, wDigest, Unsigned LDAP Binds,
and Weak ciphers in Kerberos.
Authentication protocols are a critical foundation of nearly all security assurances. These older versions can be
exploited by attackers with access to your network and are often used extensively on legacy systems on
Infrastructure as a Service (IaaS).
Here are ways to reduce your risk:
Discover protocol usage by reviewing logs with Azure Sentinel’s Insecure Protocol Dashboard or third-
party tools.
 Restrict or Disable use of these protocols by following guidance for SMB, NTLM, WDigest
We recommend implementing changes using pilot or other testing method to mitigate risk of operational
interruption.

Elevated security capabilities

Consider whether to utilize specialized security capabilities in your enterprise architecture.
These measures have the potential to enhance security and meet regulatory requirements, but can introduce
complexity that may negatively impact your operations and efficiency.
We recommend careful consideration and judicious use of these security measures as required:
Dedicated Hardware Security Modules (HSMs)
Dedicated Hardware Security Modules (HSMs) may help meet regulatory or security requirements.
Confidential Computing
Confidential Computing may help meet regulatory or security requirements.
 Cloud Design Patterns
12/18/2020 • 7 minutes to read • Edit Online

These design patterns are useful for building reliable, scalable, secure applications in the cloud.
Each pattern describes the problem that the pattern addresses, considerations for applying the pattern, and an
example based on Microsoft Azure. Most of the patterns include code samples or snippets that show how to
implement the pattern on Azure. However, most of the patterns are relevant to any distributed system, whether
hosted on Azure or on other cloud platforms.

Challenges in cloud development

Availability
Availability is the proportion of time that the system is
functional and working, usually measured as a percentage
of uptime. It can be affected by system errors,
infrastructure problems, malicious attacks, and system
load. Cloud applications typically provide users with a
service level agreement (SLA), so applications must be
designed to maximize availability.

Data Management
Data management is the key element of cloud
applications, and influences most of the quality attributes.
Data is typically hosted in different locations and across
multiple servers for reasons such as performance,
scalability or availability, and this can present a range of
challenges. For example, data consistency must be
maintained, and data will typically need to be
synchronized across different locations.

Design and Implementation

Good design encompasses factors such as consistency
and coherence in component design and deployment,
maintainability to simplify administration and
development, and reusability to allow components and
subsystems to be used in other applications and in other
scenarios. Decisions made during the design and
implementation phase have a huge impact on the quality
and the total cost of ownership of cloud hosted
applications and services.
Messaging
The distributed nature of cloud applications requires a
messaging infrastructure that connects the components
and services, ideally in a loosely coupled manner in order
to maximize scalability. Asynchronous messaging is widely
used, and provides many benefits, but also brings
challenges such as the ordering of messages, poison
message management, idempotency, and more.

Management and Monitoring

Cloud applications run in a remote datacenter where you
do not have full control of the infrastructure or, in some
cases, the operating system. This can make management
and monitoring more difficult than an on-premises
deployment. Applications must expose runtime
information that administrators and operators can use to
manage and monitor the system, as well as supporting
changing business requirements and customization
without requiring the application to be stopped or
redeployed.

Performance and Scalability

Performance is an indication of the responsiveness of a
system to execute any action within a given time interval,
while scalability is ability of a system either to handle
increases in load without impact on performance or for
the available resources to be readily increased. Cloud
applications typically encounter variable workloads and
peaks in activity. Predicting these, especially in a
multitenant scenario, is almost impossible. Instead,
applications should be able to scale out within limits to
meet peaks in demand, and scale in when demand
decreases. Scalability concerns not just compute instances,
but other elements such as data storage, messaging
infrastructure, and more.

Resiliency
Resiliency is the ability of a system to gracefully handle
and recover from failures. The nature of cloud hosting,
where applications are often multitenant, use shared
platform services, compete for resources and bandwidth,
communicate over the Internet, and run on commodity
hardware means there is an increased likelihood that both
transient and more permanent faults will arise. Detecting
failures, and recovering quickly and efficiently, is necessary
to maintain resiliency.
 Security
Security provides confidentiality, integrity, and availability
assurances against malicious attacks on information
systems (and safety assurances for attacks on operational
technology systems). Losing these assurances can
negatively impact your business operations and revenue,
as well as your organization’s reputation in the
marketplace. Maintaining security requires following well-
established practices (security hygiene) and being vigilant
to detect and rapidly remediate vulnerabilities and active
attacks.

Catalog of patterns
PAT T ERN SUM M A RY C AT EGO RY

Ambassador Create helper services that send Design and Implementation,

network requests on behalf of a
consumer service or application. Management and Monitoring

Anti-Corruption Layer Implement a façade or adapter layer Design and Implementation,

between a modern application and a
legacy system. Management and Monitoring

Asynchronous Request-Reply Decouple backend processing from a Messaging

frontend host, where backend
processing needs to be asynchronous,
but the frontend still needs a clear
response.

Backends for Frontends Create separate backend services to be Design and Implementation
consumed by specific frontend
applications or interfaces.

Bulkhead Isolate elements of an application into Resiliency

pools so that if one fails, the others will
continue to function.

Cache-Aside Load data on demand into a cache from Data Management,

a data store
Performance and Scalability

Choreography Let each service decide when and how a Messaging,

business operation is processed, instead
of depending on a central orchestrator. Performance and Scalability

Circuit Breaker Handle faults that might take a variable Resiliency

amount of time to fix when connecting
to a remote service or resource.

Claim Check Split a large message into a claim check Messaging

and a payload to avoid overwhelming a
message bus.
PAT T ERN SUM M A RY C AT EGO RY

Compensating Transaction Undo the work performed by a series of Resiliency

steps, which together define an
eventually consistent operation.

Competing Consumers Enable multiple concurrent consumers Messaging

to process messages received on the
same messaging channel.

Compute Resource Consolidation Consolidate multiple tasks or Design and Implementation

operations into a single computational
unit

CQRS Segregate operations that read data Data Management,

from operations that update data by
using separate interfaces. Design and Implementation,
Performance and Scalability

Deployment Stamps Deploy multiple independent copies of Availability,

application components, including data
stores. Performance and Scalability

Event Sourcing Use an append-only store to record the Data Management,

full series of events that describe
actions taken on data in a domain. Performance and Scalability

External Configuration Store Move configuration information out of Design and Implementation,
the application deployment package to
a centralized location. Management and Monitoring

Federated Identity Delegate authentication to an external Security

identity provider.

Gatekeeper Protect applications and services by Security

using a dedicated host instance that
acts as a broker between clients and the
application or service, validates and
sanitizes requests, and passes requests
and data between them.

Gateway Aggregation Use a gateway to aggregate multiple Design and Implementation,

individual requests into a single request.
Management and Monitoring

Gateway Offloading Offload shared or specialized service Design and Implementation,

functionality to a gateway proxy.
Management and Monitoring

Gateway Routing Route requests to multiple services Design and Implementation,

using a single endpoint.
Management and Monitoring

Geodes Deploy backend services into a set of Availability,

geographical nodes, each of which can
service any client request in any region. Performance and Scalability
PAT T ERN SUM M A RY C AT EGO RY

Health Endpoint Monitoring Implement functional checks in an Availability,

application that external tools can
access through exposed endpoints at Management and Monitoring,
regular intervals.
Resiliency

Index Table Create indexes over the fields in data Data Management,
stores that are frequently referenced by
queries. Performance and Scalability

Leader Election Coordinate the actions performed by a Design and Implementation,

collection of collaborating task instances
in a distributed application by electing Resiliency
one instance as the leader that assumes
responsibility for managing the other
instances.

Materialized View Generate prepopulated views over the Data Management,

data in one or more data stores when
the data isn't ideally formatted for Performance and Scalability
required query operations.

Pipes and Filters Break down a task that performs Design and Implementation,
complex processing into a series of
separate elements that can be reused. Messaging

Priority Queue Prioritize requests sent to services so Messaging,

that requests with a higher priority are
received and processed more quickly Performance and Scalability
than those with a lower priority.

Publisher/Subscriber Enable an application to announce Messaging

events to multiple interested consumers
asynchronously, without coupling the
senders to the receivers.

Queue-Based Load Leveling Use a queue that acts as a buffer Availability,

between a task and a service that it
invokes in order to smooth intermittent Messaging,
heavy loads.
Resiliency,
Performance and Scalability

Retry Enable an application to handle Resiliency

anticipated, temporary failures when it
tries to connect to a service or network
resource by transparently retrying an
operation that's previously failed.

Scheduler Agent Supervisor Coordinate a set of actions across a Messaging,

distributed set of services and other
remote resources. Resiliency

Sequential Convoy Process a set of related messages in a Messaging

defined order, without blocking
processing of other groups of
messages.
PAT T ERN SUM M A RY C AT EGO RY

Sharding Divide a data store into a set of Data Management,

horizontal partitions or shards.
Performance and Scalability

Sidecar Deploy components of an application Design and Implementation,

into a separate process or container to
provide isolation and encapsulation. Management and Monitoring

Static Content Hosting Deploy static content to a cloud-based Design and Implementation,
storage service that can deliver them
directly to the client. Data Management,
Performance and Scalability

Strangler Fig Incrementally migrate a legacy system Design and Implementation,

by gradually replacing specific pieces of
functionality with new applications and Management and Monitoring
services.

Throttling Control the consumption of resources Availability,

used by an instance of an application,
an individual tenant, or an entire Performance and Scalability
service.

Valet Key Use a token or key that provides clients Data Management,
with restricted direct access to a specific
resource or service. Security
 Availability patterns
12/18/2020 • 2 minutes to read • Edit Online

Availability is measured as a percentage of uptime, and defines the proportion of time that a system is functional
and working. Availability is affected by system errors, infrastructure problems, malicious attacks, and system load.
Cloud applications typically provide users with a service level agreement (SLA), which means that applications
must be designed and implemented to maximize availability.

PAT T ERN SUM M A RY

Deployment Stamps Deploy multiple independent copies of application

components, including data stores.

Geodes Deploy backend services into a set of geographical nodes,

each of which can service any client request in any region.

Health Endpoint Monitoring Implement functional checks in an application that external

tools can access through exposed endpoints at regular
intervals.

Queue-Based Load Leveling Use a queue that acts as a buffer between a task and a
service that it invokes, to smooth intermittent heavy loads.

Throttling Control the consumption of resources by an instance of an

application, an individual tenant, or an entire service.

To mitigate against availability risks from malicious Distributed Denial of Service (DDoS) attacks, implement the
native Azure DDoS protection standard service or a 3rd party capability.
 Data Management patterns
12/18/2020 • 2 minutes to read • Edit Online

Data management is the key element of cloud applications, and influences most of the quality attributes. Data is
typically hosted in different locations and across multiple servers for reasons such as performance, scalability or
availability, and this can present a range of challenges. For example, data consistency must be maintained, and
data will typically need to be synchronized across different locations.
Additionally data should be protected at rest, in transit, and via authorized access mechanisms to maintain
security assurances of confidentiality, integrity, and availability. Refer to the Azure Security Benchmark Data
Protection Control for more information.

PAT T ERN SUM M A RY

Cache-Aside Load data on demand into a cache from a data store

CQRS Segregate operations that read data from operations that

update data by using separate interfaces.

Event Sourcing Use an append-only store to record the full series of events
that describe actions taken on data in a domain.

Index Table Create indexes over the fields in data stores that are
frequently referenced by queries.

Materialized View Generate prepopulated views over the data in one or more
data stores when the data isn't ideally formatted for required
query operations.

Sharding Divide a data store into a set of horizontal partitions or

shards.

Static Content Hosting Deploy static content to a cloud-based storage service that
can deliver them directly to the client.

Valet Key Use a token or key that provides clients with restricted direct
access to a specific resource or service.
 Design and Implementation patterns
12/18/2020 • 2 minutes to read • Edit Online

Good design encompasses factors such as consistency and coherence in component design and deployment,
maintainability to simplify administration and development, and reusability to allow components and
subsystems to be used in other applications and in other scenarios. Decisions made during the design and
implementation phase have a huge impact on the quality and the total cost of ownership of cloud hosted
applications and services.

PAT T ERN SUM M A RY

Ambassador Create helper services that send network requests on behalf

of a consumer service or application.

Anti-Corruption Layer Implement a façade or adapter layer between a modern

application and a legacy system.

Backends for Frontends Create separate backend services to be consumed by

specific frontend applications or interfaces.

CQRS Segregate operations that read data from operations that

update data by using separate interfaces.

Compute Resource Consolidation Consolidate multiple tasks or operations into a single

computational unit

External Configuration Store Move configuration information out of the application

deployment package to a centralized location.

Gateway Aggregation Use a gateway to aggregate multiple individual requests

into a single request.

Gateway Offloading Offload shared or specialized service functionality to a

gateway proxy.

Gateway Routing Route requests to multiple services using a single endpoint.

Leader Election Coordinate the actions performed by a collection of

collaborating task instances in a distributed application by
electing one instance as the leader that assumes
responsibility for managing the other instances.

Pipes and Filters Break down a task that performs complex processing into a
series of separate elements that can be reused.

Sidecar Deploy components of an application into a separate

process or container to provide isolation and encapsulation.

Static Content Hosting Deploy static content to a cloud-based storage service that
can deliver them directly to the client.
PAT T ERN SUM M A RY

Strangler Fig Incrementally migrate a legacy system by gradually

replacing specific pieces of functionality with new
applications and services.
 Management and Monitoring patterns
12/18/2020 • 2 minutes to read • Edit Online

Cloud applications run in a remote datacenter where you do not have full control of the infrastructure or, in
some cases, the operating system. This can make management and monitoring more difficult than an on-
premises deployment. Applications must expose runtime information that administrators and operators can use
to manage and monitor the system, as well as supporting changing business requirements and customization
without requiring the application to be stopped or redeployed.

PAT T ERN SUM M A RY

Ambassador Create helper services that send network requests on behalf

of a consumer service or application.

Anti-Corruption Layer Implement a façade or adapter layer between a modern

application and a legacy system.

External Configuration Store Move configuration information out of the application

deployment package to a centralized location.

Gateway Aggregation Use a gateway to aggregate multiple individual requests into

a single request.

Gateway Offloading Offload shared or specialized service functionality to a

gateway proxy.

Gateway Routing Route requests to multiple services using a single endpoint.

Health Endpoint Monitoring Implement functional checks in an application that external

tools can access through exposed endpoints at regular
intervals.

Sidecar Deploy components of an application into a separate

process or container to provide isolation and encapsulation.

Strangler Fig Incrementally migrate a legacy system by gradually replacing

specific pieces of functionality with new applications and
services.
 Messaging patterns
12/18/2020 • 2 minutes to read • Edit Online

The distributed nature of cloud applications requires a messaging infrastructure that connects the components
and services, ideally in a loosely coupled manner in order to maximize scalability. Asynchronous messaging is
widely used, and provides many benefits, but also brings challenges such as the ordering of messages, poison
message management, idempotency, and more.

PAT T ERN SUM M A RY

Asynchronous Request-Reply Decouple backend processing from a frontend host, where

backend processing needs to be asynchronous, but the
frontend still needs a clear response.

Claim Check Split a large message into a claim check and a payload to
avoid overwhelming a message bus.

Choreography Have each component of the system participate in the

decision-making process about the workflow of a business
transaction, instead of relying on a central point of control.

Competing Consumers Enable multiple concurrent consumers to process messages

received on the same messaging channel.

Pipes and Filters Break down a task that performs complex processing into a
series of separate elements that can be reused.

Priority Queue Prioritize requests sent to services so that requests with a

higher priority are received and processed more quickly than
those with a lower priority.

Publisher-Subscriber Enable an application to announce events to multiple

interested consumers asynchronously, without coupling the
senders to the receivers.

Queue-Based Load Leveling Use a queue that acts as a buffer between a task and a
service that it invokes in order to smooth intermittent heavy
loads.

Scheduler Agent Supervisor Coordinate a set of actions across a distributed set of

services and other remote resources.

Sequential Convoy Process a set of related messages in a defined order, without

blocking processing of other groups of messages.
 Performance and Scalability patterns
12/18/2020 • 2 minutes to read • Edit Online

Performance is an indication of the responsiveness of a system to execute any action within a given time
interval, while scalability is ability of a system either to handle increases in load without impact on performance
or for the available resources to be readily increased. Cloud applications typically encounter variable workloads
and peaks in activity. Predicting these, especially in a multi-tenant scenario, is almost impossible. Instead,
applications should be able to scale out within limits to meet peaks in demand, and scale in when demand
decreases. Scalability concerns not just compute instances, but other elements such as data storage, messaging
infrastructure, and more.

PAT T ERN SUM M A RY

Cache-Aside Load data on demand into a cache from a data store

Choreography Have each component of the system participate in the

decision-making process about the workflow of a business
transaction, instead of relying on a central point of control.

CQRS Segregate operations that read data from operations that

update data by using separate interfaces.

Event Sourcing Use an append-only store to record the full series of events
that describe actions taken on data in a domain.

Deployment Stamps Deploy multiple independent copies of application

components, including data stores.

Geodes Deploy backend services into a set of geographical nodes,

each of which can service any client request in any region.

Index Table Create indexes over the fields in data stores that are
frequently referenced by queries.

Materialized View Generate prepopulated views over the data in one or more
data stores when the data isn't ideally formatted for
required query operations.

Priority Queue Prioritize requests sent to services so that requests with a

higher priority are received and processed more quickly
than those with a lower priority.

Queue-Based Load Leveling Use a queue that acts as a buffer between a task and a
service that it invokes in order to smooth intermittent
heavy loads.

Sharding Divide a data store into a set of horizontal partitions or

shards.

Static Content Hosting Deploy static content to a cloud-based storage service that
can deliver them directly to the client.
PAT T ERN SUM M A RY

Throttling Control the consumption of resources used by an instance

of an application, an individual tenant, or an entire service.
 Resiliency patterns
12/18/2020 • 2 minutes to read • Edit Online

Resiliency is the ability of a system to gracefully handle and recover from failures, both inadvertent and
malicious.
The nature of cloud hosting, where applications are often multi-tenant, use shared platform services, compete
for resources and bandwidth, communicate over the Internet, and run on commodity hardware means there is
an increased likelihood that both transient and more permanent faults will arise. The connected nature of the
internet and the rise in sophistication and volume of attacks increase the likelihood of a security disruption.
Detecting failures and recovering quickly and efficiently, is necessary to maintain resiliency.

PAT T ERN SUM M A RY

Bulkhead Isolate elements of an application into pools so that if one

fails, the others will continue to function.

Circuit Breaker Handle faults that might take a variable amount of time to
fix when connecting to a remote service or resource.

Compensating Transaction Undo the work performed by a series of steps, which

together define an eventually consistent operation.

Health Endpoint Monitoring Implement functional checks in an application that external

tools can access through exposed endpoints at regular
intervals.

Leader Election Coordinate the actions performed by a collection of

collaborating task instances in a distributed application by
electing one instance as the leader that assumes
responsibility for managing the other instances.

Queue-Based Load Leveling Use a queue that acts as a buffer between a task and a
service that it invokes in order to smooth intermittent heavy
loads.

Retry Enable an application to handle anticipated, temporary

failures when it tries to connect to a service or network
resource by transparently retrying an operation that's
previously failed.

Scheduler Agent Supervisor Coordinate a set of actions across a distributed set of

services and other remote resources.

Security Resliency
Achieving security resilience requires a combination of preventive measures to block attacks, responsive
measures detect and quickly remediate active attacks, and governance to ensure consistent application of best
practices.
Security strategy should include lessons learned described in security strategy guidance
Azure security configurations should align to the best practices and controls in the Azure Security
 Benchmark (ASB). Security configurations for Azure services should align to the Security baselines for Azure
in the ASB
Azure architectures should integrate native security capabilities to protect and monitor workloads
including Azure Defender, Azure DDoS protection, Azure Firewall, and Azure Web Application Firewall (WAF)
For a more detailed dsicussion, see the Cybersecurity Resilience module in the CISO workshop
 Security patterns
12/18/2020 • 2 minutes to read • Edit Online

Security provides confidentiality, integrity, and availability assurances against malicious attacks on information
systems (and safety assurances for attacks on operational technology systems). Losing these assurances can
negatively impact your business operations and revenue, as well as your organization’s reputation in the
marketplace. Maintaining security requires following well-established practices (security hygiene) and being
vigilant to detect and rapidly remediate vulnerabilities and active attacks.

Patterns
PAT T ERN SUM M A RY

Federated Identity Delegate authentication to an external identity provider.

Gatekeeper Protect applications and services by using a dedicated host

instance that acts as a broker between clients and the
application or service, validates and sanitizes requests, and
passes requests and data between them.

Valet Key Use a token or key that provides clients with restricted direct
access to a specific resource or service.

Key Security Resources

RESO URC E SUM M A RY

Azure Security Benchmarks Prescriptive best practices and recommendations to integrate

into architectures for securing workloads, data, services, and
enterprise environments on Azure.

Azure Defender Native security controls to simplify integration of threat

detection and monitoring in Azure architectures

Security Strategy Guidance Building and updating a security strategy for cloud adoption
and modern threat environment

Security Roles and Responsibilities Guidance on security roles and responsibilities including
definitions of mission/outcome for each organizational
function and how each should evolve with the adoption of
cloud.

Getting Started Guide for Security Guidance for planning and implementing security throughout
cloud adoption
 Ambassador pattern
12/18/2020 • 3 minutes to read • Edit Online

Create helper services that send network requests on behalf of a consumer service or application. An ambassador
service can be thought of as an out-of-process proxy that is co-located with the client.
This pattern can be useful for offloading common client connectivity tasks such as monitoring, logging, routing,
security (such as TLS), and resiliency patterns in a language agnostic way. It is often used with legacy applications,
or other applications that are difficult to modify, in order to extend their networking capabilities. It can also enable
a specialized team to implement those features.

Context and problem

Resilient cloud-based applications require features such as circuit breaking, routing, metering and monitoring,
and the ability to make network-related configuration updates. It may be difficult or impossible to update legacy
applications or existing code libraries to add these features, because the code is no longer maintained or can't be
easily modified by the development team.
Network calls may also require substantial configuration for connection, authentication, and authorization. If these
calls are used across multiple applications, built using multiple languages and frameworks, the calls must be
configured for each of these instances. In addition, network and security functionality may need to be managed by
a central team within your organization. With a large code base, it can be risky for that team to update application
code they aren't familiar with.

Solution
Put client frameworks and libraries into an external process that acts as a proxy between your application and
external services. Deploy the proxy on the same host environment as your application to allow control over
routing, resiliency, security features, and to avoid any host-related access restrictions. You can also use the
ambassador pattern to standardize and extend instrumentation. The proxy can monitor performance metrics such
as latency or resource usage, and this monitoring happens in the same host environment as the application.

Features that are offloaded to the ambassador can be managed independently of the application. You can update
and modify the ambassador without disturbing the application's legacy functionality. It also allows for separate,
specialized teams to implement and maintain security, networking, or authentication features that have been
moved to the ambassador.
Ambassador services can be deployed as a sidecar to accompany the lifecycle of a consuming application or
service. Alternatively, if an ambassador is shared by multiple separate processes on a common host, it can be
deployed as a daemon or Windows service. If the consuming service is containerized, the ambassador should be
created as a separate container on the same host, with the appropriate links configured for communication.

Issues and considerations

The proxy adds some latency overhead. Consider whether a client library, invoked directly by the application, is
a better approach.
Consider the possible impact of including generalized features in the proxy. For example, the ambassador
could handle retries, but that might not be safe unless all operations are idempotent.
Consider a mechanism to allow the client to pass some context to the proxy, as well as back to the client. For
example, include HTTP request headers to opt out of retry or specify the maximum number of times to retry.
Consider how you will package and deploy the proxy.
Consider whether to use a single shared instance for all clients or an instance for each client.

When to use this pattern

Use this pattern when you:
Need to build a common set of client connectivity features for multiple languages or frameworks.
Need to offload cross-cutting client connectivity concerns to infrastructure developers or other more
specialized teams.
Need to support cloud or cluster connectivity requirements in a legacy application or an application that is
difficult to modify.
This pattern may not be suitable:
When network request latency is critical. A proxy will introduce some overhead, although minimal, and in
some cases this may affect the application.
When client connectivity features are consumed by a single language. In that case, a better option might be a
client library that is distributed to the development teams as a package.
When connectivity features cannot be generalized and require deeper integration with the client application.

Example
The following diagram shows an application making a request to a remote service via an ambassador proxy. The
ambassador provides routing, circuit breaking, and logging. It calls the remote service and then returns the
response to the client application:
Related guidance
Sidecar pattern
 Anti-Corruption Layer pattern
12/18/2020 • 2 minutes to read • Edit Online

Implement a façade or adapter layer between different subsystems that don't share the same semantics. This layer
translates requests that one subsystem makes to the other subsystem. Use this pattern to ensure that an
application's design is not limited by dependencies on outside subsystems. This pattern was first described by Eric
Evans in Domain-Driven Design.

Context and problem

Most applications rely on other systems for some data or functionality. For example, when a legacy application is
migrated to a modern system, it may still need existing legacy resources. New features must be able to call the
legacy system. This is especially true of gradual migrations, where different features of a larger application are
moved to a modern system over time.
Often these legacy systems suffer from quality issues such as convoluted data schemas or obsolete APIs. The
features and technologies used in legacy systems can vary widely from more modern systems. To interoperate
with the legacy system, the new application may need to support outdated infrastructure, protocols, data models,
APIs, or other features that you wouldn't otherwise put into a modern application.
Maintaining access between new and legacy systems can force the new system to adhere to at least some of the
legacy system's APIs or other semantics. When these legacy features have quality issues, supporting them
"corrupts" what might otherwise be a cleanly designed modern application.
Similar issues can arise with any external system that your development team doesn't control, not just legacy
systems.

Solution
Isolate the different subsystems by placing an anti-corruption layer between them. This layer translates
communications between the two systems, allowing one system to remain unchanged while the other can avoid
compromising its design and technological approach.

The diagram above shows an application with two subsystems. Subsystem A calls to subsystem B through an anti-
corruption layer. Communication between subsystem A and the anti-corruption layer always uses the data model
and architecture of subsystem A. Calls from the anti-corruption layer to subsystem B conform to that subsystem's
data model or methods. The anti-corruption layer contains all of the logic necessary to translate between the two
systems. The layer can be implemented as a component within the application or as an independent service.

Issues and considerations

The anti-corruption layer may add latency to calls made between the two systems.
The anti-corruption layer adds an additional service that must be managed and maintained.
Consider how your anti-corruption layer will scale.
Consider whether you need more than one anti-corruption layer. You may want to decompose functionality
into multiple services using different technologies or languages, or there may be other reasons to partition the
anti-corruption layer.
Consider how the anti-corruption layer will be managed in relation with your other applications or services.
How will it be integrated into your monitoring, release, and configuration processes?
Make sure transaction and data consistency are maintained and can be monitored.
Consider whether the anti-corruption layer needs to handle all communication between different subsystems,
or just a subset of features.
If the anti-corruption layer is part of an application migration strategy, consider whether it will be permanent,
or will be retired after all legacy functionality has been migrated.

When to use this pattern

Use this pattern when:
A migration is planned to happen over multiple stages, but integration between new and legacy systems needs
to be maintained.
Two or more subsystems have different semantics, but still need to communicate.
This pattern may not be suitable if there are no significant semantic differences between new and legacy systems.

Related guidance
Strangler Fig pattern
 Asynchronous Request-Reply pattern
12/18/2020 • 9 minutes to read • Edit Online

Decouple backend processing from a frontend host, where backend processing needs to be asynchronous, but the
frontend still needs a clear response.

Context and problem

In modern application development, it's normal for client applications — often code running in a web-client
(browser) — to depend on remote APIs to provide business logic and compose functionality. These APIs may be
directly related to the application or may be shared services provided by a third party. Commonly these API calls
take place over the HTTP(S) protocol and follow REST semantics.
In most cases, APIs for a client application are designed to respond quickly, on the order of 100 ms or less. Many
factors can affect the response latency, including:
An application's hosting stack.
Security components.
The relative geographic location of the caller and the backend.
Network infrastructure.
Current load.
The size of the request payload.
Processing queue length.
The time for the backend to process the request.
Any of these factors can add latency to the response. Some can be mitigated by scaling out the backend. Others,
such as network infrastructure, are largely out of the control of the application developer. Most APIs can respond
quickly enough for responses to arrive back over the same connection. Application code can make a synchronous
API call in a non-blocking way, giving the appearance of asynchronous processing, which is recommended for I/O-
bound operations.
In some scenarios, however, the work done by backend may be long-running, on the order of seconds, or might be
a background process that is executed in minutes or even hours. In that case, it isn't feasible to wait for the work to
complete before responding to the request. This situation is a potential problem for any synchronous request-
reply pattern.
Some architectures solve this problem by using a message broker to separate the request and response stages.
This separation is often achieved by use of the Queue-Based Load Leveling pattern. This separation can allow the
client process and the backend API to scale independently. But this separation also brings additional complexity
when the client requires success notification, as this step needs to become asynchronous.
Many of the same considerations discussed for client applications also apply for server-to-server REST API calls in
distributed systems — for example, in a microservices architecture.

Solution
One solution to this problem is to use HTTP polling. Polling is useful to client-side code, as it can be hard to provide
call-back endpoints or use long running connections. Even when callbacks are possible, the extra libraries and
services that are required can sometimes add too much extra complexity.
The client application makes a synchronous call to the API, triggering a long-running operation on the
 backend.
The API responds synchronously as quickly as possible. It returns an HTTP 202 (Accepted) status code,
acknowledging that the request has been received for processing.

NOTE
The API should validate both the request and the action to be performed before starting the long running process. If
the request is invalid, reply immediately with an error code such as HTTP 400 (Bad Request).

The response holds a location reference pointing to an endpoint that the client can poll to check for the
result of the long running operation.
The API offloads processing to another component, such as a message queue.
While the work is still pending, the status endpoint returns HTTP 202. Once the work is complete, the status
endpoint can either return a resource that indicates completion, or redirect to another resource URL. For
example, if the asynchronous operation creates a new resource, the status endpoint would redirect to the
URL for that resource.
The following diagram shows a typical flow:

1. The client sends a request and receives an HTTP 202 (Accepted) response.
2. The client sends an HTTP GET request to the status endpoint. The work is still pending, so this call also returns
HTTP 202.
3. At some point, the work is complete and the status endpoint returns 302 (Found) redirecting to the resource.
4. The client fetches the resource at the specified URL.

Issues and considerations

There are a number of possible ways to implement this pattern over HTTP and not all upstream services
have the same semantics. For example, most services won't return an HTTP 202 response back from a GET
method when a remote process hasn't finished. Following pure REST semantics, they should return HTTP
404 (Not Found). This response makes sense when you consider the result of the call isn't present yet.
An HTTP 202 response should indicate the location and frequency that the client should poll for the
response. It should have the following additional headers:
 H EA DER DESC RIP T IO N N OT ES

Location A URL the client should poll for a This URL could be a SAS token with
response status. the Valet Key Pattern being
appropriate if this location needs
access control. The valet key pattern
is also valid when response polling
needs offloading to another backend

Retry-After An estimate of when processing will This header is designed to prevent

complete polling clients from overwhelming
the back-end with retries.

You may need to use a processing proxy or facade to manipulate the response headers or payload
depending on the underlying services used.
If the status endpoint redirects on completion, either HTTP 302 or HTTP 303 are appropriate return codes,
depending on the exact semantics you support.
Upon successful processing, the resource specified by the Location header should return an appropriate
HTTP response code such as 200 (OK), 201 (Created), or 204 (No Content).
If an error occurs during processing, persist the error at the resource URL described in the Location header
and ideally return an appropriate response code to the client from that resource (4xx code).
Not all solutions will implement this pattern in the same way and some services will include additional or
alternate headers. For example, Azure Resource Manager uses a modified variant of this pattern. For more
information, see Azure Resource Manager Async Operations.
Legacy clients might not support this pattern. In that case, you might need to place a facade over the
asynchronous API to hide the asynchronous processing from the original client. For example, Azure Logic
Apps supports this pattern natively can be used as an integration layer between an asynchronous API and a
client that makes synchronous calls. See Perform long-running tasks with the webhook action pattern.
In some scenarios, you might want to provide a way for clients to cancel a long-running request. In that
case, the backend service must support some form of cancellation instruction.

When to use this pattern

Use this pattern for:
Client-side code, such as browser applications, where it's difficult to provide call-back endpoints, or the use
of long-running connections adds too much additional complexity.
Service calls where only the HTTP protocol is available and the return service can't fire callbacks because of
firewall restrictions on the client-side.
Service calls that need to be integrated with legacy architectures that don't support modern callback
technologies such as WebSockets or webhooks.
This pattern might not be suitable when:
You can use a service built for asynchronous notifications instead, such as Azure Event Grid.
Responses must stream in real time to the client.
The client needs to collect many results, and received latency of those results is important. Consider a service
bus pattern instead.
You can use server-side persistent network connections such as WebSockets or SignalR. These services can be
used to notify the caller of the result.
 The network design allows you to open up ports to receive asynchronous callbacks or webhooks.

Example
The following code shows excerpts from an application that uses Azure Functions to implement this pattern. There
are three functions in the solution:
The asynchronous API endpoint.
The status endpoint.
A backend function that takes queued work items and executes them.

This sample is available on GitHub.

AsyncProcessingWorkAcceptor function
The AsyncProcessingWorkAcceptor function implements an endpoint that accepts work from a client application and
puts it on a queue for processing.
The function generates a request ID and adds it as metadata to the queue message.
The HTTP response includes a location header pointing to a status endpoint. The request ID is part of the URL
path.
 public static class AsyncProcessingWorkAcceptor
{
[FunctionName("AsyncProcessingWorkAcceptor")]
public static async Task<IActionResult> Run(
[HttpTrigger(AuthorizationLevel.Anonymous, "post", Route = null)] CustomerPOCO customer,
[ServiceBus("outqueue", Connection = "ServiceBusConnectionAppSetting")] IAsyncCollector<Message>
OutMessage,
ILogger log)
{
if (String.IsNullOrEmpty(customer.id) || String.IsNullOrEmpty(customer.customername))
{
return new BadRequestResult();
}

string reqid = Guid.NewGuid().ToString();

string rqs =
$"http://{Environment.GetEnvironmentVariable("WEBSITE_HOSTNAME")}/api/RequestStatus/{reqid}";

var messagePayload = JsonConvert.SerializeObject(customer);

Message m = new Message(Encoding.UTF8.GetBytes(messagePayload));
m.UserProperties["RequestGUID"] = reqid;
m.UserProperties["RequestSubmittedAt"] = DateTime.Now;
m.UserProperties["RequestStatusURL"] = rqs;

await OutMessage.AddAsync(m);

return (ActionResult) new AcceptedResult(rqs, $"Request Accepted for

Processing{Environment.NewLine}ProxyStatus: {rqs}");
}
}

AsyncProcessingBackgroundWorker function
The AsyncProcessingBackgroundWorker function picks up the operation from the queue, does some work based on
the message payload, and writes the result to the SAS signature location.

public static class AsyncProcessingBackgroundWorker

{
[FunctionName("AsyncProcessingBackgroundWorker")]
public static void Run(
[ServiceBusTrigger("outqueue", Connection = "ServiceBusConnectionAppSetting")]Message myQueueItem,
[Blob("data", FileAccess.ReadWrite, Connection = "StorageConnectionAppSetting")] CloudBlobContainer
inputBlob,
ILogger log)
{
// Perform an actual action against the blob data source for the async readers to be able to check
against.
// This is where your actual service worker processing will be performed.

var id = myQueueItem.UserProperties["RequestGUID"] as string;

CloudBlockBlob cbb = inputBlob.GetBlockBlobReference($"{id}.blobdata");

// Now write the results to blob storage.

cbb.UploadFromByteArrayAsync(myQueueItem.Body, 0, myQueueItem.Body.Length);
}
}

AsyncOperationStatusChecker function
The AsyncOperationStatusChecker function implements the status endpoint. This function first checks whether the
request was completed
If the request was completed, the function either returns a valet-key to the response, or redirects the call
immediately to the valet-key URL.
If the request is still pending, then we should return a 202 accepted with a self-referencing Location header,
putting an ETA for a completed response in the http Retry-After header.

public static class AsyncOperationStatusChecker

{
[FunctionName("AsyncOperationStatusChecker")]
public static async Task<IActionResult> Run(
[HttpTrigger(AuthorizationLevel.Anonymous, "get", Route = "RequestStatus/{thisGUID}")] HttpRequest
req,
[Blob("data/{thisGuid}.blobdata", FileAccess.Read, Connection = "StorageConnectionAppSetting")]
CloudBlockBlob inputBlob, string thisGUID,
ILogger log)
{

OnCompleteEnum OnComplete = Enum.Parse<OnCompleteEnum>(req.Query["OnComplete"].FirstOrDefault() ??

"Redirect");
OnPendingEnum OnPending = Enum.Parse<OnPendingEnum>(req.Query["OnPending"].FirstOrDefault() ??
"Accepted");

log.LogInformation($"C# HTTP trigger function processed a request for status on {thisGUID} -

OnComplete {OnComplete} - OnPending {OnPending}");

// Check to see if the blob is present.

if (await inputBlob.ExistsAsync())
{
// If it's present, depending on the value of the optional "OnComplete" parameter choose what to
do.
return await OnCompleted(OnComplete, inputBlob, thisGUID);
}
else
{
// If it's NOT present, check the optional "OnPending" parameter.
string rqs =
$"http://{Environment.GetEnvironmentVariable("WEBSITE_HOSTNAME")}/api/RequestStatus/{thisGUID}";

switch (OnPending)
{
case OnPendingEnum.Accepted:
{
// Return an HTTP 202 status code.
return (ActionResult)new AcceptedResult() { Location = rqs };
}

case OnPendingEnum.Synchronous:
{
// Back off and retry. Time out if the backoff period hits one minute
int backoff = 250;

while (!await inputBlob.ExistsAsync() && backoff < 64000)

{
log.LogInformation($"Synchronous mode {thisGUID}.blob - retrying in {backoff}
ms");
backoff = backoff * 2;
await Task.Delay(backoff);
}

if (await inputBlob.ExistsAsync())
{
log.LogInformation($"Synchronous Redirect mode {thisGUID}.blob - completed after
{backoff} ms");
return await OnCompleted(OnComplete, inputBlob, thisGUID);
}
else
{
log.LogInformation($"Synchronous mode {thisGUID}.blob - NOT FOUND after timeout
 log.LogInformation($"Synchronous mode {thisGUID}.blob - NOT FOUND after timeout
{backoff} ms");
return (ActionResult)new NotFoundResult();
}
}

default:
{
throw new InvalidOperationException($"Unexpected value: {OnPending}");
}
}
}
}

private static async Task<IActionResult> OnCompleted(OnCompleteEnum OnComplete, CloudBlockBlob inputBlob,

string thisGUID)
{
switch (OnComplete)
{
case OnCompleteEnum.Redirect:
{
// Redirect to the SAS URI to blob storage
return (ActionResult)new RedirectResult(inputBlob.GenerateSASURI());
}

case OnCompleteEnum.Stream:
{
// Download the file and return it directly to the caller.
// For larger files, use a stream to minimize RAM usage.
return (ActionResult)new OkObjectResult(await inputBlob.DownloadTextAsync());
}

default:
{
throw new InvalidOperationException($"Unexpected value: {OnComplete}");
}
}
}
}

public enum OnCompleteEnum {

Redirect,
Stream
}

public enum OnPendingEnum {

Accepted,
Synchronous
}

Related guidance and next steps

The following information may be relevant when implementing this pattern:
Asynchronous operations in REST
Azure Logic Apps - Perform long-running tasks with the polling action pattern.
For general best practices when designing a web API, see Web API design.
 Backends for Frontends pattern
12/18/2020 • 3 minutes to read • Edit Online

Create separate backend services to be consumed by specific frontend applications or interfaces. This pattern is
useful when you want to avoid customizing a single backend for multiple interfaces. This pattern was first
described by Sam Newman.

Context and problem

An application may initially be targeted at a desktop web UI. Typically, a backend service is developed in parallel
that provides the features needed for that UI. As the application's user base grows, a mobile application is
developed that must interact with the same backend. The backend service becomes a general-purpose backend,
serving the requirements of both the desktop and mobile interfaces.
But the capabilities of a mobile device differ significantly from a desktop browser, in terms of screen size,
performance, and display limitations. As a result, the requirements for a mobile application backend differ from
the desktop web UI.
These differences result in competing requirements for the backend. The backend requires regular and significant
changes to serve both the desktop web UI and the mobile application. Often, separate interface teams work on
each frontend, causing the backend to become a bottleneck in the development process. Conflicting update
requirements, and the need to keep the service working for both frontends, can result in spending a lot of effort
on a single deployable resource.

As the development activity focuses on the backend service, a separate team may be created to manage and
maintain the backend. Ultimately, this results in a disconnect between the interface and backend development
teams, placing a burden on the backend team to balance the competing requirements of the different UI teams.
When one interface team requires changes to the backend, those changes must be validated with other interface
teams before they can be integrated into the backend.

Solution
Create one backend per user interface. Fine-tune the behavior and performance of each backend to best match the
needs of the frontend environment, without worrying about affecting other frontend experiences.

Because each backend is specific to one interface, it can be optimized for that interface. As a result, it will be
smaller, less complex, and likely faster than a generic backend that tries to satisfy the requirements for all
interfaces. Each interface team has autonomy to control their own backend and doesn't rely on a centralized
backend development team. This gives the interface team flexibility in language selection, release cadence,
prioritization of workload, and feature integration in their backend.
For more information, see Pattern: Backends For Frontends.

Issues and considerations

Consider how many backends to deploy.
If different interfaces (such as mobile clients) will make the same requests, consider whether it is necessary to
implement a backend for each interface, or if a single backend will suffice.
Code duplication across services is highly likely when implementing this pattern.
Frontend-focused backend services should only contain client-specific logic and behavior. General business
logic and other global features should be managed elsewhere in your application.
Think about how this pattern might be reflected in the responsibilities of a development team.
Consider how long it will take to implement this pattern. Will the effort of building the new backends incur
technical debt, while you continue to support the existing generic backend?

When to use this pattern

Use this pattern when:
A shared or general purpose backend service must be maintained with significant development overhead.
You want to optimize the backend for the requirements of specific client interfaces.
Customizations are made to a general-purpose backend to accommodate multiple interfaces.
An alternative language is better suited for the backend of a different user interface.
This pattern may not be suitable:
 When interfaces make the same or similar requests to the backend.
When only one interface is used to interact with the backend.

Related guidance
Pattern: Backends For Frontends
Gateway Aggregation pattern
Gateway Offloading pattern
Gateway Routing pattern
 Bulkhead pattern
12/18/2020 • 4 minutes to read • Edit Online

The Bulkhead pattern is a type of application design that is tolerant of failure. In a bulkhead architecture, elements
of an application are isolated into pools so that if one fails, the others will continue to function. It's named after the
sectioned partitions (bulkheads) of a ship's hull. If the hull of a ship is compromised, only the damaged section
fills with water, which prevents the ship from sinking.

Context and problem

A cloud-based application may include multiple services, with each service having one or more consumers.
Excessive load or failure in a service will impact all consumers of the service.
Moreover, a consumer may send requests to multiple services simultaneously, using resources for each request.
When the consumer sends a request to a service that is misconfigured or not responding, the resources used by
the client's request may not be freed in a timely manner. As requests to the service continue, those resources may
be exhausted. For example, the client's connection pool may be exhausted. At that point, requests by the consumer
to other services are affected. Eventually the consumer can no longer send requests to other services, not just the
original unresponsive service.
The same issue of resource exhaustion affects services with multiple consumers. A large number of requests
originating from one client may exhaust available resources in the service. Other consumers are no longer able to
consume the service, causing a cascading failure effect.

Solution
Partition service instances into different groups, based on consumer load and availability requirements. This
design helps to isolate failures, and allows you to sustain service functionality for some consumers, even during a
failure.
A consumer can also partition resources, to ensure that resources used to call one service don't affect the
resources used to call another service. For example, a consumer that calls multiple services may be assigned a
connection pool for each service. If a service begins to fail, it only affects the connection pool assigned for that
service, allowing the consumer to continue using the other services.
The benefits of this pattern include:
Isolates consumers and services from cascading failures. An issue affecting a consumer or service can be
isolated within its own bulkhead, preventing the entire solution from failing.
Allows you to preserve some functionality in the event of a service failure. Other services and features of the
application will continue to work.
Allows you to deploy services that offer a different quality of service for consuming applications. A high-
priority consumer pool can be configured to use high-priority services.
The following diagram shows bulkheads structured around connection pools that call individual services. If
Service A fails or causes some other issue, the connection pool is isolated, so only workloads using the thread
pool assigned to Service A are affected. Workloads that use Service B and C are not affected and can continue
working without interruption.
The next diagram shows multiple clients calling a single service. Each client is assigned a separate service
instance. Client 1 has made too many requests and overwhelmed its instance. Because each service instance is
isolated from the others, the other clients can continue making calls.

Issues and considerations

Define partitions around the business and technical requirements of the application.
When partitioning services or consumers into bulkheads, consider the level of isolation offered by the
technology as well as the overhead in terms of cost, performance and manageability.
Consider combining bulkheads with retry, circuit breaker, and throttling patterns to provide more sophisticated
fault handling.
When partitioning consumers into bulkheads, consider using processes, thread pools, and semaphores.
Projects like resilience4j and Polly offer a framework for creating consumer bulkheads.
When partitioning services into bulkheads, consider deploying them into separate virtual machines,
containers, or processes. Containers offer a good balance of resource isolation with fairly low overhead.
Services that communicate using asynchronous messages can be isolated through different sets of queues.
Each queue can have a dedicated set of instances processing messages on the queue, or a single group of
instances using an algorithm to dequeue and dispatch processing.
Determine the level of granularity for the bulkheads. For example, if you want to distribute tenants across
partitions, you could place each tenant into a separate partition, or put several tenants into one partition.
Monitor each partition’s performance and SLA.
When to use this pattern
Use this pattern to:
Isolate resources used to consume a set of backend services, especially if the application can provide some
level of functionality even when one of the services is not responding.
Isolate critical consumers from standard consumers.
Protect the application from cascading failures.
This pattern may not be suitable when:
Less efficient use of resources may not be acceptable in the project.
The added complexity is not necessary

Example
The following Kubernetes configuration file creates an isolated container to run a single service, with its own CPU
and memory resources and limits.

apiVersion: v1
kind: Pod
metadata:
name: drone-management
spec:
containers:
- name: drone-management-container
image: drone-service
resources:
requests:
memory: "64Mi"
cpu: "250m"
limits:
memory: "128Mi"
cpu: "1"

Related guidance
Designing reliable Azure applications
Circuit Breaker pattern
Retry pattern
Throttling pattern
 Cache-Aside pattern
12/18/2020 • 7 minutes to read • Edit Online

Load data on demand into a cache from a data store. This can improve performance and also helps to maintain
consistency between data held in the cache and data in the underlying data store.

Context and problem

Applications use a cache to improve repeated access to information held in a data store. However, it's impractical
to expect that cached data will always be completely consistent with the data in the data store. Applications
should implement a strategy that helps to ensure that the data in the cache is as up-to-date as possible, but can
also detect and handle situations that arise when the data in the cache has become stale.

Solution
Many commercial caching systems provide read-through and write-through/write-behind operations. In these
systems, an application retrieves data by referencing the cache. If the data isn't in the cache, it's retrieved from the
data store and added to the cache. Any modifications to data held in the cache are automatically written back to
the data store as well.
For caches that don't provide this functionality, it's the responsibility of the applications that use the cache to
maintain the data.
An application can emulate the functionality of read-through caching by implementing the cache-aside strategy.
This strategy loads data into the cache on demand. The figure illustrates using the Cache-Aside pattern to store
data in the cache.

If an application updates information, it can follow the write-through strategy by making the modification to the
data store, and by invalidating the corresponding item in the cache.
When the item is next required, using the cache-aside strategy will cause the updated data to be retrieved from
the data store and added back into the cache.
Issues and considerations
Consider the following points when deciding how to implement this pattern:
Lifetime of cached data . Many caches implement an expiration policy that invalidates data and removes it
from the cache if it's not accessed for a specified period. For cache-aside to be effective, ensure that the expiration
policy matches the pattern of access for applications that use the data. Don't make the expiration period too short
because this can cause applications to continually retrieve data from the data store and add it to the cache.
Similarly, don't make the expiration period so long that the cached data is likely to become stale. Remember that
caching is most effective for relatively static data, or data that is read frequently.
Evicting data . Most caches have a limited size compared to the data store where the data originates, and they'll
evict data if necessary. Most caches adopt a least-recently-used policy for selecting items to evict, but this might
be customizable. Configure the global expiration property and other properties of the cache, and the expiration
property of each cached item, to ensure that the cache is cost effective. It isn't always appropriate to apply a
global eviction policy to every item in the cache. For example, if a cached item is very expensive to retrieve from
the data store, it can be beneficial to keep this item in the cache at the expense of more frequently accessed but
less costly items.
Priming the cache . Many solutions prepopulate the cache with the data that an application is likely to need as
part of the startup processing. The Cache-Aside pattern can still be useful if some of this data expires or is
evicted.
Consistency . Implementing the Cache-Aside pattern doesn't guarantee consistency between the data store and
the cache. An item in the data store can be changed at any time by an external process, and this change might not
be reflected in the cache until the next time the item is loaded. In a system that replicates data across data stores,
this problem can become serious if synchronization occurs frequently.
Local (in-memor y) caching . A cache could be local to an application instance and stored in-memory. Cache-
aside can be useful in this environment if an application repeatedly accesses the same data. However, a local
cache is private and so different application instances could each have a copy of the same cached data. This data
could quickly become inconsistent between caches, so it might be necessary to expire data held in a private cache
and refresh it more frequently. In these scenarios, consider investigating the use of a shared or a distributed
caching mechanism.

When to use this pattern

Use this pattern when:
A cache doesn't provide native read-through and write-through operations.
Resource demand is unpredictable. This pattern enables applications to load data on demand. It makes no
assumptions about which data an application will require in advance.
This pattern might not be suitable:
When the cached data set is static. If the data will fit into the available cache space, prime the cache with the
data on startup and apply a policy that prevents the data from expiring.
For caching session state information in a web application hosted in a web farm. In this environment, you
should avoid introducing dependencies based on client-server affinity.

Example
In Microsoft Azure you can use Azure Cache for Redis to create a distributed cache that can be shared by multiple
instances of an application.
This following code examples use the StackExchange.Redis client, which is a Redis client library written for .NET.
To connect to an Azure Cache for Redis instance, call the static ConnectionMultiplexer.Connect method and pass
in the connection string. The method returns a ConnectionMultiplexer that represents the connection. One
approach to sharing a ConnectionMultiplexer instance in your application is to have a static property that returns
a connected instance, similar to the following example. This approach provides a thread-safe way to initialize only
a single connected instance.

private static ConnectionMultiplexer Connection;

// Redis connection string information

private static Lazy<ConnectionMultiplexer> lazyConnection = new Lazy<ConnectionMultiplexer>(() =>
{
string cacheConnection = ConfigurationManager.AppSettings["CacheConnection"].ToString();
return ConnectionMultiplexer.Connect(cacheConnection);
});

public static ConnectionMultiplexer Connection => lazyConnection.Value;

The GetMyEntityAsync method in the following code example shows an implementation of the Cache-Aside
pattern. This method retrieves an object from the cache using the read-through approach.
An object is identified by using an integer ID as the key. The GetMyEntityAsync method tries to retrieve an item
with this key from the cache. If a matching item is found, it's returned. If there's no match in the cache, the
GetMyEntityAsync method retrieves the object from a data store, adds it to the cache, and then returns it. The
code that actually reads the data from the data store is not shown here, because it depends on the data store.
Note that the cached item is configured to expire to prevent it from becoming stale if it's updated elsewhere.

// Set five minute expiration as a default

private const double DefaultExpirationTimeInMinutes = 5.0;

public async Task<MyEntity> GetMyEntityAsync(int id)

{
// Define a unique key for this method and its parameters.
var key = $"MyEntity:{id}";
var cache = Connection.GetDatabase();

// Try to get the entity from the cache.

var json = await cache.StringGetAsync(key).ConfigureAwait(false);
var value = string.IsNullOrWhiteSpace(json)
? default(MyEntity)
: JsonConvert.DeserializeObject<MyEntity>(json);

if (value == null) // Cache miss

{
// If there's a cache miss, get the entity from the original store and cache it.
// Code has been omitted because it is data store dependent.
value = ...;

// Avoid caching a null value.

if (value != null)
{
// Put the item in the cache with a custom expiration time that
// depends on how critical it is to have stale data.
await cache.StringSetAsync(key, JsonConvert.SerializeObject(value)).ConfigureAwait(false);
await cache.KeyExpireAsync(key,
TimeSpan.FromMinutes(DefaultExpirationTimeInMinutes)).ConfigureAwait(false);
}
}

return value;
}
 The examples use Azure Cache for Redis to access the store and retrieve information from the cache. For
more information, see Using Azure Cache for Redis and How to create a Web App with Azure Cache for Redis.

The UpdateEntityAsync method shown below demonstrates how to invalidate an object in the cache when the
value is changed by the application. The code updates the original data store and then removes the cached item
from the cache.

public async Task UpdateEntityAsync(MyEntity entity)

{
// Update the object in the original data store.
await this.store.UpdateEntityAsync(entity).ConfigureAwait(false);

// Invalidate the current cache object.

var cache = Connection.GetDatabase();
var id = entity.Id;
var key = $"MyEntity:{id}"; // The key for the cached object.
await cache.KeyDeleteAsync(key).ConfigureAwait(false); // Delete this key from the cache.
}

NOTE
The order of the steps is important. Update the data store before removing the item from the cache. If you remove the
cached item first, there is a small window of time when a client might fetch the item before the data store is updated. That
will result in a cache miss (because the item was removed from the cache), causing the earlier version of the item to be
fetched from the data store and added back into the cache. The result will be stale cache data.

Related guidance
The following information may be relevant when implementing this pattern:
Caching Guidance. Provides additional information on how you can cache data in a cloud solution, and the
issues that you should consider when you implement a cache.
Data Consistency Primer. Cloud applications typically use data that's spread across data stores. Managing
and maintaining data consistency in this environment is a critical aspect of the system, particularly the
concurrency and availability issues that can arise. This primer describes issues about consistency across
distributed data, and summarizes how an application can implement eventual consistency to maintain the
availability of data.
 Choreography
12/18/2020 • 7 minutes to read • Edit Online

Have each component of the system participate in the decision-making process about the workflow of a business
transaction, instead of relying on a central point of control.

Context and problem

In microservices architecture, it’s often the case that a cloud-based application is divided into several small
services that work together to process a business transaction end-to-end. To lower coupling between services,
each service is responsible for a single business operation. Some benefits include faster development, smaller
code base, and scalability. However, designing an efficient and scalable workflow is a challenge and often requires
complex interservice communication.
The services communicate with each other by using well-defined APIs. Even a single business operation can result
in multiple point-to-point calls among all services. A common pattern for communication is to use a centralized
service that acts as the orchestrator. It acknowledges all incoming requests and delegates operations to the
respective services. In doing so, it also manages the workflow of the entire business transaction. Each service just
completes an operation and is not aware of the overall workflow.
The orchestrator pattern reduces point-to-point communication between services but has some drawbacks
because of the tight coupling between the orchestrator and other services that participate in processing of the
business transaction. To execute tasks in a sequence, the orchestrator needs to have some domain knowledge
about the responsibilities of those services. If you want to add or remove services, existing logic will break, and
you'll need to rewire portions of the communication path. While you can configure the workflow, add or remove
services easily with a well-designed orchestrator, such an implementation is complex and hard to maintain.

Solution
Let each service decide when and how a business operation is processed, instead of depending on a central
orchestrator.
One way to implement choreography is to use the asynchronous messaging pattern to coordinate the business
operations.
A client request publishes messages to a message queue. As messages arrive, they are pushed to subscribers, or
services, interested in that message. Each subscribed service does their operation as indicated by the message
and responds to the message queue with success or failure of the operation. In case of success, the service can
push a message back to the same queue or a different message queue so that another service can continue the
workflow if needed. If an operation fails, the message bus can retry that operation.
This way, the services choreograph the workflow among themselves without depending on an orchestrator or
having direct communication between them.
Because there isn't point-to-point communication, this pattern helps reduce coupling between services. Also, it
can remove the performance bottleneck caused by the orchestrator when it has to deal with all transactions.

When to use this pattern

Use the choreography pattern if you expect to update, remove, or add new services frequently. The entire app can
be modified with lesser effort and minimal disruption to existing services.
Consider this pattern if you experience performance bottlenecks in the central orchestrator.
This pattern is a natural model for the serverless architecture where all services can be short lived, or event
driven. Services can spin up because of an event, do their task, and are removed when the task is finished.

Issues and considerations

Decentralizing the orchestrator can cause issues while managing the workflow.
If a service fails to complete a business operation, it can be difficult to recover from that failure. One way is to
have the service indicate failure by firing an event. Another service subscribes to those failed events takes
necessary actions such as applying compensating transactions to undo successful operations in a request. The
failed service might also fail to fire an event for the failure. In that case, consider using a retry and, or time out
mechanism to recognize that operation as a failure. For an example, see the Example section.
It's simple to implement a workflow when you want to process independent business operations in parallel. You
can use a single message bus. However, the workflow can become complicated when choreography needs to
occur in a sequence. For instance, Service C can start its operation only after Service A and Service B have
completed their operations with success. One approach is to have multiple message buses that get messages in
the required order. For more information, see the Example section.
The choreography pattern becomes a challenge if the number of services grow rapidly. Given the high number of
independent moving parts, the workflow between services tends to get complex. Also, distributed tracing
becomes difficult.
The orchestrator centrally manages the resiliency of the workflow and it can become a single point of failure. On
the other hand, for choreography, the role is distributed between all services and resiliency becomes less robust.
Each service isn't only responsible for the resiliency of its operation but also the workflow. This responsibility can
be burdensome for the service and hard to implement. Each service must retry transient, nontransient, and time-
out failures, so that the request terminates gracefully, if needed. Also, the service must be diligent about
communicating the success or failure of the operation so that other services can act accordingly.

Example
This example shows the choreography pattern with the Drone Delivery app. When a client requests a pickup, the
app assigns a drone and notifies the client.

An example of this pattern is available on GitHub.

A single client business transaction requires three distinct business operations: creating or updating a package,
assigning a drone to deliver the package, and checking the delivery status. Those operations are performed by
three microservices: Package, Drone Scheduler, and Delivery services. Instead of a central orchestrator, the
services use messaging to collaborate and coordinate the request among themselves.
Design
The business transaction is processed in a sequence through multiple hops. Each hop has a message bus and the
respective business service.
When a client sends a delivery request through an HTTP endpoint, the Ingestion service receives it, raises an
operation event, and sends it to a message bus. The bus invokes the subscribed business service and sends the
event in a POST request. On receiving the event, the business service can complete the operation with success,
failure, or the request can time out. If successful, the service responds to the bus with the Ok status code, raises a
new operation event, and sends it to the message bus of the next hop. In case of a failure or time-out, the service
reports failure by sending the BadRequest code to the message bus that sent the original POST request. The
message bus retries the operation based on a retry policy. After that period expires, message bus flags the failed
operation and further processing of the entire request stops.
This workflow continues until the entire request has been processed.
The design uses multiple message buses to process the entire business transaction. Microsoft Azure Event Grid
provides the messaging service. The app is deployed in an Azure Kubernetes Service (AKS) cluster with two
containers in the same pod. One container runs the ambassador that interacts with Event Grid while the other
runs a business service. The approach with two containers in the same pod improves performance and scalability.
The ambassador and the business service share the same network allowing for low latency and high throughput.
To avoid cascading retry operations that might lead to multiple efforts, only Event Grid retries an operation
instead of the business service. It flags a failed request by sending a messaging to a dead letter queue (DLQ).
The business services are idempotent to make sure retry operations don’t result in duplicate resources. For
example, the Package service uses upsert operations to add data to the data store.
The example implements a custom solution to correlate calls across all services and Event Grid hops.
Here’s a code example that shows the choreography pattern between all business services. It shows the workflow
of the Drone Delivery app transactions. Code for exception handling and logging have been removed for brevity.
 [HttpPost]
[Route("/api/[controller]/operation")]
[ProducesResponseType(typeof(void), 200)]
[ProducesResponseType(typeof(void), 400)]
[ProducesResponseType(typeof(void), 500)]

public async Task<IActionResult> Post([FromBody] EventGridEvent[] events)

{

if (events == null)
{
return BadRequest("No Event for Choreography");
}

foreach(var e in events)
{

List<EventGridEvent> listEvents = new List<EventGridEvent>();

e.Topic = eventRepository.GetTopic();
e.EventTime = DateTime.Now;
switch (e.EventType)
{
case Operations.ChoreographyOperation.ScheduleDelivery:
{
var packageGen = await
packageServiceCaller.UpsertPackageAsync(delivery.PackageInfo).ConfigureAwait(false);
if (packageGen is null)
{
//BadRequest allows the event to be reprocessed by Event Grid
return BadRequest("Package creation failed.");
}

//we set the event type to the next choreography step

e.EventType = Operations.ChoreographyOperation.CreatePackage;
listEvents.Add(e);
await eventRepository.SendEventAsync(listEvents);
return Ok("Created Package Completed");
}
case Operations.ChoreographyOperation.CreatePackage:
{
var droneId = await
droneSchedulerServiceCaller.GetDroneIdAsync(delivery).ConfigureAwait(false);
if (droneId is null)
{
//BadRequest allows the event to be reprocessed by Event Grid
return BadRequest("could not get a drone id");
}
e.Subject = droneId;
e.EventType = Operations.ChoreographyOperation.GetDrone;
listEvents.Add(e);
await eventRepository.SendEventAsync(listEvents);
return Ok("Drone Completed");
}
case Operations.ChoreographyOperation.GetDrone:
{
var deliverySchedule = await deliveryServiceCaller.ScheduleDeliveryAsync(delivery,
e.Subject);
return Ok("Delivery Completed");
}
return BadRequest();
}
}

Related patterns and guidance

Consider these patterns in your design for choreography.
Modularize the business service by using the ambassador design pattern.
Implement queue-based load leveling pattern to handle spikes of the workload.
Use asynchronous distributed messaging through the publisher-subscriber pattern.
Use compensating transactions to undo a series of successful operations in case one or more related
operation fails.
For information about using a message broker in a messaging infrastructure, see Asynchronous
messaging options in Azure.
 Circuit Breaker pattern
12/18/2020 • 17 minutes to read • Edit Online

Handle faults that might take a variable amount of time to recover from, when connecting to a remote service
or resource. This can improve the stability and resiliency of an application.

Context and problem

In a distributed environment, calls to remote resources and services can fail due to transient faults, such as
slow network connections, timeouts, or the resources being overcommitted or temporarily unavailable. These
faults typically correct themselves after a short period of time, and a robust cloud application should be
prepared to handle them by using a strategy such as the Retry pattern.
However, there can also be situations where faults are due to unanticipated events, and that might take much
longer to fix. These faults can range in severity from a partial loss of connectivity to the complete failure of a
service. In these situations it might be pointless for an application to continually retry an operation that is
unlikely to succeed, and instead the application should quickly accept that the operation has failed and handle
this failure accordingly.
Additionally, if a service is very busy, failure in one part of the system might lead to cascading failures. For
example, an operation that invokes a service could be configured to implement a timeout, and reply with a
failure message if the service fails to respond within this period. However, this strategy could cause many
concurrent requests to the same operation to be blocked until the timeout period expires. These blocked
requests might hold critical system resources such as memory, threads, database connections, and so on.
Consequently, these resources could become exhausted, causing failure of other possibly unrelated parts of
the system that need to use the same resources. In these situations, it would be preferable for the operation to
fail immediately, and only attempt to invoke the service if it's likely to succeed. Note that setting a shorter
timeout might help to resolve this problem, but the timeout shouldn't be so short that the operation fails most
of the time, even if the request to the service would eventually succeed.

Solution
The Circuit Breaker pattern, popularized by Michael Nygard in his book, Release It!, can prevent an application
from repeatedly trying to execute an operation that's likely to fail. Allowing it to continue without waiting for
the fault to be fixed or wasting CPU cycles while it determines that the fault is long lasting. The Circuit Breaker
pattern also enables an application to detect whether the fault has been resolved. If the problem appears to
have been fixed, the application can try to invoke the operation.

The purpose of the Circuit Breaker pattern is different than the Retry pattern. The Retry pattern enables an
application to retry an operation in the expectation that it'll succeed. The Circuit Breaker pattern prevents
an application from performing an operation that is likely to fail. An application can combine these two
patterns by using the Retry pattern to invoke an operation through a circuit breaker. However, the retry
logic should be sensitive to any exceptions returned by the circuit breaker and abandon retry attempts if
the circuit breaker indicates that a fault is not transient.

A circuit breaker acts as a proxy for operations that might fail. The proxy should monitor the number of recent
failures that have occurred, and use this information to decide whether to allow the operation to proceed, or
simply return an exception immediately.
The proxy can be implemented as a state machine with the following states that mimic the functionality of an
electrical circuit breaker:
Closed : The request from the application is routed to the operation. The proxy maintains a count of the
number of recent failures, and if the call to the operation is unsuccessful the proxy increments this
count. If the number of recent failures exceeds a specified threshold within a given time period, the
proxy is placed into the Open state. At this point the proxy starts a timeout timer, and when this timer
expires the proxy is placed into the Half-Open state.

The purpose of the timeout timer is to give the system time to fix the problem that caused the
failure before allowing the application to try to perform the operation again.

Open : The request from the application fails immediately and an exception is returned to the
application.
Half-Open : A limited number of requests from the application are allowed to pass through and invoke
the operation. If these requests are successful, it's assumed that the fault that was previously causing
the failure has been fixed and the circuit breaker switches to the Closed state (the failure counter is
reset). If any request fails, the circuit breaker assumes that the fault is still present so it reverts back to
the Open state and restarts the timeout timer to give the system a further period of time to recover
from the failure.

The Half-Open state is useful to prevent a recovering service from suddenly being flooded with
requests. As a service recovers, it might be able to support a limited volume of requests until the
recovery is complete, but while recovery is in progress a flood of work can cause the service to
time out or fail again.

In the figure, the failure counter used by the Closed state is time based. It's automatically reset at periodic
intervals. This helps to prevent the circuit breaker from entering the Open state if it experiences occasional
failures. The failure threshold that trips the circuit breaker into the Open state is only reached when a
specified number of failures have occurred during a specified interval. The counter used by the Half-Open
state records the number of successful attempts to invoke the operation. The circuit breaker reverts to the
Closed state after a specified number of consecutive operation invocations have been successful. If any
invocation fails, the circuit breaker enters the Open state immediately and the success counter will be reset
the next time it enters the Half-Open state.

How the system recovers is handled externally, possibly by restoring or restarting a failed component or
repairing a network connection.

The Circuit Breaker pattern provides stability while the system recovers from a failure and minimizes the
impact on performance. It can help to maintain the response time of the system by quickly rejecting a request
for an operation that's likely to fail, rather than waiting for the operation to time out, or never return. If the
circuit breaker raises an event each time it changes state, this information can be used to monitor the health of
the part of the system protected by the circuit breaker, or to alert an administrator when a circuit breaker trips
to the Open state.
The pattern is customizable and can be adapted according to the type of the possible failure. For example, you
can apply an increasing timeout timer to a circuit breaker. You could place the circuit breaker in the Open state
for a few seconds initially, and then if the failure hasn't been resolved increase the timeout to a few minutes,
and so on. In some cases, rather than the Open state returning failure and raising an exception, it could be
useful to return a default value that is meaningful to the application.

Issues and considerations

You should consider the following points when deciding how to implement this pattern:
Exception Handling . An application invoking an operation through a circuit breaker must be prepared to
handle the exceptions raised if the operation is unavailable. The way exceptions are handled will be application
specific. For example, an application could temporarily degrade its functionality, invoke an alternative
operation to try to perform the same task or obtain the same data, or report the exception to the user and ask
them to try again later.
Types of Exceptions . A request might fail for many reasons, some of which might indicate a more severe
type of failure than others. For example, a request might fail because a remote service has crashed and will
take several minutes to recover, or because of a timeout due to the service being temporarily overloaded. A
circuit breaker might be able to examine the types of exceptions that occur and adjust its strategy depending
on the nature of these exceptions. For example, it might require a larger number of timeout exceptions to trip
the circuit breaker to the Open state compared to the number of failures due to the service being completely
unavailable.
Logging . A circuit breaker should log all failed requests (and possibly successful requests) to enable an
administrator to monitor the health of the operation.
Recoverability . You should configure the circuit breaker to match the likely recovery pattern of the operation
it's protecting. For example, if the circuit breaker remains in the Open state for a long period, it could raise
exceptions even if the reason for the failure has been resolved. Similarly, a circuit breaker could fluctuate and
reduce the response times of applications if it switches from the Open state to the Half-Open state too
quickly.
Testing Failed Operations . In the Open state, rather than using a timer to determine when to switch to the
Half-Open state, a circuit breaker can instead periodically ping the remote service or resource to determine
whether it's become available again. This ping could take the form of an attempt to invoke an operation that
had previously failed, or it could use a special operation provided by the remote service specifically for testing
the health of the service, as described by the Health Endpoint Monitoring pattern.
Manual Override . In a system where the recovery time for a failing operation is extremely variable, it's
beneficial to provide a manual reset option that enables an administrator to close a circuit breaker (and reset
the failure counter). Similarly, an administrator could force a circuit breaker into the Open state (and restart
the timeout timer) if the operation protected by the circuit breaker is temporarily unavailable.
Concurrency . The same circuit breaker could be accessed by a large number of concurrent instances of an
application. The implementation shouldn't block concurrent requests or add excessive overhead to each call to
an operation.
Resource Differentiation . Be careful when using a single circuit breaker for one type of resource if there
might be multiple underlying independent providers. For example, in a data store that contains multiple
shards, one shard might be fully accessible while another is experiencing a temporary issue. If the error
responses in these scenarios are merged, an application might try to access some shards even when failure is
highly likely, while access to other shards might be blocked even though it's likely to succeed.
Accelerated Circuit Breaking . Sometimes a failure response can contain enough information for the circuit
breaker to trip immediately and stay tripped for a minimum amount of time. For example, the error response
from a shared resource that's overloaded could indicate that an immediate retry isn't recommended and that
the application should instead try again in a few minutes.

NOTE
A service can return HTTP 429 (Too Many Requests) if it is throttling the client, or HTTP 503 (Service Unavailable) if the
service is not currently available. The response can include additional information, such as the anticipated duration of
the delay.

Replaying Failed Requests . In the Open state, rather than simply failing quickly, a circuit breaker could also
record the details of each request to a journal and arrange for these requests to be replayed when the remote
resource or service becomes available.
Inappropriate Timeouts on External Ser vices . A circuit breaker might not be able to fully protect
applications from operations that fail in external services that are configured with a lengthy timeout period. If
the timeout is too long, a thread running a circuit breaker might be blocked for an extended period before the
circuit breaker indicates that the operation has failed. In this time, many other application instances might also
try to invoke the service through the circuit breaker and tie up a significant number of threads before they all
fail.

When to use this pattern

Use this pattern:
To prevent an application from trying to invoke a remote service or access a shared resource if this
operation is highly likely to fail.
This pattern isn't recommended:
For handling access to local private resources in an application, such as in-memory data structure. In this
environment, using a circuit breaker would add overhead to your system.
As a substitute for handling exceptions in the business logic of your applications.

Example
In a web application, several of the pages are populated with data retrieved from an external service. If the
system implements minimal caching, most hits to these pages will cause a round trip to the service.
Connections from the web application to the service could be configured with a timeout period (typically 60
seconds), and if the service doesn't respond in this time the logic in each web page will assume that the
service is unavailable and throw an exception.
However, if the service fails and the system is very busy, users could be forced to wait for up to 60 seconds
before an exception occurs. Eventually resources such as memory, connections, and threads could be
exhausted, preventing other users from connecting to the system, even if they aren't accessing pages that
retrieve data from the service.
Scaling the system by adding further web servers and implementing load balancing might delay when
resources become exhausted, but it won't resolve the issue because user requests will still be unresponsive
and all web servers could still eventually run out of resources.
Wrapping the logic that connects to the service and retrieves the data in a circuit breaker could help to solve
this problem and handle the service failure more elegantly. User requests will still fail, but they'll fail more
quickly and the resources won't be blocked.
The class maintains state information about a circuit breaker in an object that implements the
CircuitBreaker
ICircuitBreakerStateStore interface shown in the following code.

interface ICircuitBreakerStateStore
{
CircuitBreakerStateEnum State { get; }

Exception LastException { get; }

DateTime LastStateChangedDateUtc { get; }

void Trip(Exception ex);

void Reset();

void HalfOpen();

bool IsClosed { get; }

}

The State property indicates the current state of the circuit breaker, and will be either Open , HalfOpen , or
Closed as defined by the CircuitBreakerStateEnum enumeration. The IsClosed property should be true if the
circuit breaker is closed, but false if it's open or half open. The Trip method switches the state of the circuit
breaker to the open state and records the exception that caused the change in state, together with the date and
time that the exception occurred. The LastException and the LastStateChangedDateUtc properties return this
information. The Reset method closes the circuit breaker, and the HalfOpen method sets the circuit breaker
to half open.
The InMemoryCircuitBreakerStateStore class in the example contains an implementation of the
ICircuitBreakerStateStore interface. The CircuitBreaker class creates an instance of this class to hold the
state of the circuit breaker.
The ExecuteAction method in the CircuitBreaker class wraps an operation, specified as an Action delegate.
If the circuit breaker is closed, ExecuteAction invokes the Action delegate. If the operation fails, an exception
handler calls TrackException , which sets the circuit breaker state to open. The following code example
highlights this flow.
 public class CircuitBreaker
{
private readonly ICircuitBreakerStateStore stateStore =
CircuitBreakerStateStoreFactory.GetCircuitBreakerStateStore();

private readonly object halfOpenSyncObject = new object ();

...
public bool IsClosed { get { return stateStore.IsClosed; } }

public bool IsOpen { get { return !IsClosed; } }

public void ExecuteAction(Action action)

{
...
if (IsOpen)
{
// The circuit breaker is Open.
... (see code sample below for details)
}

// The circuit breaker is Closed, execute the action.

try
{
action();
}
catch (Exception ex)
{
// If an exception still occurs here, simply
// retrip the breaker immediately.
this.TrackException(ex);

// Throw the exception so that the caller can tell

// the type of exception that was thrown.
throw;
}
}

private void TrackException(Exception ex)

{
// For simplicity in this example, open the circuit breaker on the first exception.
// In reality this would be more complex. A certain type of exception, such as one
// that indicates a service is offline, might trip the circuit breaker immediately.
// Alternatively it might count exceptions locally or across multiple instances and
// use this value over time, or the exception/success ratio based on the exception
// types, to open the circuit breaker.
this.stateStore.Trip(ex);
}
}

The following example shows the code (omitted from the previous example) that is executed if the circuit
breaker isn't closed. It first checks if the circuit breaker has been open for a period longer than the time
specified by the local OpenToHalfOpenWaitTime field in the CircuitBreaker class. If this is the case, the
ExecuteAction method sets the circuit breaker to half open, then tries to perform the operation specified by
the Action delegate.
If the operation is successful, the circuit breaker is reset to the closed state. If the operation fails, it is tripped
back to the open state and the time the exception occurred is updated so that the circuit breaker will wait for a
further period before trying to perform the operation again.
If the circuit breaker has only been open for a short time, less than the OpenToHalfOpenWaitTime value, the
ExecuteAction method simply throws a CircuitBreakerOpenException exception and returns the error that
caused the circuit breaker to transition to the open state.
Additionally, it uses a lock to prevent the circuit breaker from trying to perform concurrent calls to the
operation while it's half open. A concurrent attempt to invoke the operation will be handled as if the circuit
breaker was open, and it'll fail with an exception as described later.

...
if (IsOpen)
{
// The circuit breaker is Open. Check if the Open timeout has expired.
// If it has, set the state to HalfOpen. Another approach might be to
// check for the HalfOpen state that had be set by some other operation.
if (stateStore.LastStateChangedDateUtc + OpenToHalfOpenWaitTime < DateTime.UtcNow)
{
// The Open timeout has expired. Allow one operation to execute. Note that, in
// this example, the circuit breaker is set to HalfOpen after being
// in the Open state for some period of time. An alternative would be to set
// this using some other approach such as a timer, test method, manually, and
// so on, and check the state here to determine how to handle execution
// of the action.
// Limit the number of threads to be executed when the breaker is HalfOpen.
// An alternative would be to use a more complex approach to determine which
// threads or how many are allowed to execute, or to execute a simple test
// method instead.
bool lockTaken = false;
try
{
Monitor.TryEnter(halfOpenSyncObject, ref lockTaken);
if (lockTaken)
{
// Set the circuit breaker state to HalfOpen.
stateStore.HalfOpen();

// Attempt the operation.

action();

// If this action succeeds, reset the state and allow other operations.
// In reality, instead of immediately returning to the Closed state, a counter
// here would record the number of successful operations and return the
// circuit breaker to the Closed state only after a specified number succeed.
this.stateStore.Reset();
return;
}
}
catch (Exception ex)
{
// If there's still an exception, trip the breaker again immediately.
this.stateStore.Trip(ex);

// Throw the exception so that the caller knows which exception occurred.
throw;
}
finally
{
if (lockTaken)
{
Monitor.Exit(halfOpenSyncObject);
}
}
}
// The Open timeout hasn't yet expired. Throw a CircuitBreakerOpen exception to
// inform the caller that the call was not actually attempted,
// and return the most recent exception received.
throw new CircuitBreakerOpenException(stateStore.LastException);
}
...

To use a CircuitBreaker object to protect an operation, an application creates an instance of the

CircuitBreaker class and invokes the ExecuteAction method, specifying the operation to be performed as the
parameter. The application should be prepared to catch the CircuitBreakerOpenException exception if the
operation fails because the circuit breaker is open. The following code shows an example:

var breaker = new CircuitBreaker();

try
{
breaker.ExecuteAction(() =>
{
// Operation protected by the circuit breaker.
...
});
}
catch (CircuitBreakerOpenException ex)
{
// Perform some different action when the breaker is open.
// Last exception details are in the inner exception.
...
}
catch (Exception ex)
{
...
}

Related patterns and guidance

The following patterns might also be useful when implementing this pattern:
Retry pattern. Describes how an application can handle anticipated temporary failures when it tries to
connect to a service or network resource by transparently retrying an operation that has previously
failed.
Health Endpoint Monitoring pattern. A circuit breaker might be able to test the health of a service by
sending a request to an endpoint exposed by the service. The service should return information
indicating its status.
 Claim-Check Pattern
12/18/2020 • 6 minutes to read • Edit Online

Split a large message into a claim check and a payload. Send the claim check to the messaging platform and store
the payload to an external service. This pattern allows large messages to be processed, while protecting the
message bus and the client from being overwhelmed or slowed down. This pattern also helps to reduce costs, as
storage is usually cheaper than resource units used by the messaging platform.
This pattern is also known as Reference-Based Messaging, and was originally described in the book Enterprise
Integration Patterns, by Gregor Hohpe and Bobby Woolf.

Context and problem

A messaging-based architecture at some point must be able to send, receive, and manipulate large messages.
Such messages may contain anything, including images (for example, MRI scans), sound files (for example, call-
center calls), text documents, or any kind of binary data of arbitrary size.
Sending such large messages to the message bus directly is not recommended, because they require more
resources and bandwidth to be consumed. Large messages can also slow down the entire solution, because
messaging platforms are usually fine-tuned to handle huge quantities of small messages. Also, most messaging
platforms have limits on message size, so you may need to work around these limits for large messages.

Solution
Store the entire message payload into an external service, such as a database. Get the reference to the stored
payload, and send just that reference to the message bus. The reference acts like a claim check used to retrieve a
piece of luggage, hence the name of the pattern. Clients interested in processing that specific message can use the
obtained reference to retrieve the payload, if needed.

Issues and considerations

Consider the following points when deciding how to implement this pattern:
Consider deleting the message data after consuming it, if you don't need to archive the messages. Although
blob storage is relatively cheap, it costs some money in the long run, especially if there is a lot of data.
Deleting the message can be done synchronously by the application that receives and processes the
message, or asynchronously by a separate dedicated process. The asynchronous approach removes old
data with no impact on the throughput and message processing performance of the receiving application.
 Storing and retrieving the message causes some additional overhead and latency. You may want to
implement logic in the sending application to use this pattern only when the message size exceeds the data
limit of the message bus. The pattern would be skipped for smaller messages. This approach would result in
a conditional claim-check pattern.

When to use this pattern

This pattern should be used whenever a message cannot fit the supported message limit of the chosen message
bus technology. For example, Event Hubs currently has a limit of 256 KB (Basic Tier), while Event Grid supports
only 64-KB messages.
The pattern can also be used if the payload should be accessed only by services that are authorized to see it. By
offloading the payload to an external resource, stricter authentication and authorization rules can be put in place,
to ensure that security is enforced when sensitive data is stored in the payload.

Examples
On Azure, this pattern can be implemented in several ways and with different technologies, but there are two main
categories. In both cases, the receiver has the responsibility to read the claim check and use it to retrieve the
payload.
Automatic claim-check generation . This approach uses Azure Event Grid to automatically generate the
claim check and push it into the message bus.
Manual claim-check generation . In this approach, the sender is responsible for managing the payload.
The sender stores the payload using the appropriate service, gets or generates the claim check, and sends
the claim check to the message bus.
Event Grid is an event routing service and tries to deliver events within a configurable amount of time up to 24
hours. After that, events are either discarded or dead lettered. If you need to archive the event payloads or replay
the event stream, you can add an Event Grid subscription to Event Hubs or Queue Storage, where messages can be
retained for longer periods and archiving messages is supported. For information about fine tuning Event Grid
message delivery and retry, and dead letter configuration, see Dead letter and retry policies.
Automatic claim-check generation with Blob Storage and Event Grid
In this approach, the sender drops the message payload into a designated Azure Blob Storage container. Event Grid
automatically generates a tag/reference and sends it to a supported message bus, such as Azure Storage Queues.
The receiver can poll the queue, get the message, and then use the stored reference data to download the payload
directly from Blob Storage.
The same Event Grid message can be directly consumed by Azure Functions, without needing to go through a
message bus. This approach takes full advantage of the serverless nature of both Event Grid and Functions.
You can find example code for this approach here.
Event Grid with Event Hubs
Similar to the previous example, Event Grid automatically generates a message when a payload is written to an
Azure Blob container. But in this example, the message bus is implemented using Event Hubs. A client can register
itself to receive the stream of messages as they are written to the event hub. The event hub can also be configured
to archive messages, making them available as an Avro file that can be queried using tools like Apache Spark,
Apache Drill, or any of the available Avro libraries.
You can find example code for this approach here.
Claim check generation with Service Bus
This solution takes advantage of a specific Service Bus plugin, ServiceBus.AttachmentPlugin, which makes the
claim-check workflow easy to implement. The plugin converts any message body into an attachment that gets
stored in Azure Blob Storage when the message is sent.

using ServiceBus.AttachmentPlugin;
...

// Getting connection information

var serviceBusConnectionString = Environment.GetEnvironmentVariable("SERVICE_BUS_CONNECTION_STRING");
var queueName = Environment.GetEnvironmentVariable("QUEUE_NAME");
var storageConnectionString = Environment.GetEnvironmentVariable("STORAGE_CONNECTION_STRING");

// Creating config for sending message

var config = new AzureStorageAttachmentConfiguration(storageConnectionString);

// Creating and registering the sender using Service Bus Connection String and Queue Name
var sender = new MessageSender(serviceBusConnectionString, queueName);
sender.RegisterAzureStorageAttachmentPlugin(config);

// Create payload
var payload = new { data = "random data string for testing" };
var serialized = JsonConvert.SerializeObject(payload);
var payloadAsBytes = Encoding.UTF8.GetBytes(serialized);
var message = new Message(payloadAsBytes);

// Send the message

await sender.SendAsync(message);

The Service Bus message acts as a notification queue, which a client can subscribe to. When the consumer receives
the message, the plugin makes it possible to directly read the message data from Blob Storage. You can then
choose how to process the message further. An advantage of this approach is that it abstracts the claim-check
workflow from the sender and receiver.
You can find example code for this approach here.
Manual claim-check generation with Kafka
In this example, a Kafka client writes the payload to Azure Blob Storage. Then it sends a notification message using
Kafka-enabled Event Hubs. The consumer receives the message and can access the payload from Blob Storage.
This example shows how a different messaging protocol can be used to implement the claim-check pattern in
Azure. For example, you might need to support existing Kafka clients.
You can find example code for this approach here.

Related patterns and guidance

The examples described above are available on GitHub.
The Enterprise Integration Patterns site has a description of this pattern.
For another example, see Dealing with large Service Bus messages using claim check pattern (blog post).
An alternative pattern for handling large messages is Split and Aggregate.
 Command and Query Responsibility Segregation
(CQRS) pattern
12/18/2020 • 11 minutes to read • Edit Online

The Command and Query Responsibility Segregation (CQRS) pattern separates read and update operations for a
data store. Implementing CQRS in your application can maximize its performance, scalability, and security. The
flexibility created by migrating to CQRS allows a system to better evolve over time and prevents update
commands from causing merge conflicts at the domain level.

The problem
In traditional architectures, the same data model is used to query and update a database. That's simple and works
well for basic CRUD operations. In more complex applications, however, this approach can become unwieldy. For
example, on the read side, the application may perform many different queries, returning data transfer objects
(DTOs) with different shapes. Object mapping can become complicated. On the write side, the model may
implement complex validation and business logic. As a result, you can end up with an overly complex model that
does too much.
Read and write workloads are often asymmetrical, with very different performance and scale requirements.

There is often a mismatch between the read and write representations of the data, such as additional
columns or properties that must be updated correctly even though they aren't required as part of an
operation.
Data contention can occur when operations are performed in parallel on the same set of data.
The traditional approach can have a negative effect on performance due to load on the data store and data
access layer, and the complexity of queries required to retrieve information.
Managing security and permissions can become complex, because each entity is subject to both read and
write operations, which might expose data in the wrong context.

Solution
CQRS separates reads and writes into different models, using commands to update data, and queries to read
data.
Commands should be task based, rather than data centric. ("Book hotel room", not "set ReservationStatus to
Reserved").
Commands may be placed on a queue for asynchronous processing, rather than being processed
synchronously.
Queries never modify the database. A query returns a DTO that does not encapsulate any domain knowledge.
The models can then be isolated, as shown in the following diagram, although that's not an absolute requirement.

Having separate query and update models simplifies the design and implementation. However, one disadvantage
is that CQRS code can't automatically be generated from a database schema using scaffolding mechanisms such
as O/RM tools.
For greater isolation, you can physically separate the read data from the write data. In that case, the read database
can use its own data schema that is optimized for queries. For example, it can store a materialized view of the
data, in order to avoid complex joins or complex O/RM mappings. It might even use a different type of data store.
For example, the write database might be relational, while the read database is a document database.
If separate read and write databases are used, they must be kept in sync. Typically this is accomplished by having
the write model publish an event whenever it updates the database. Updating the database and publishing the
event must occur in a single transaction.

The read store can be a read-only replica of the write store, or the read and write stores can have a different
structure altogether. Using multiple read-only replicas can increase query performance, especially in distributed
scenarios where read-only replicas are located close to the application instances.
Separation of the read and write stores also allows each to be scaled appropriately to match the load. For
example, read stores typically encounter a much higher load than write stores.
Some implementations of CQRS use the Event Sourcing pattern. With this pattern, application state is stored as a
sequence of events. Each event represents a set of changes to the data. The current state is constructed by
replaying the events. In a CQRS context, one benefit of Event Sourcing is that the same events can be used to
notify other components — in particular, to notify the read model. The read model uses the events to create a
snapshot of the current state, which is more efficient for queries. However, Event Sourcing adds complexity to the
design.
Benefits of CQRS include:
Independent scaling . CQRS allows the read and write workloads to scale independently, and may result in
fewer lock contentions.
Optimized data schemas . The read side can use a schema that is optimized for queries, while the write side
 uses a schema that is optimized for updates.
Security . It's easier to ensure that only the right domain entities are performing writes on the data.
Separation of concerns . Segregating the read and write sides can result in models that are more
maintainable and flexible. Most of the complex business logic goes into the write model. The read model can
be relatively simple.
Simpler queries . By storing a materialized view in the read database, the application can avoid complex joins
when querying.

Implementation issues and considerations

Some challenges of implementing this pattern include:
Complexity . The basic idea of CQRS is simple. But it can lead to a more complex application design,
especially if they include the Event Sourcing pattern.
Messaging . Although CQRS does not require messaging, it's common to use messaging to process
commands and publish update events. In that case, the application must handle message failures or
duplicate messages.
Eventual consistency . If you separate the read and write databases, the read data may be stale. The read
model store must be updated to reflect changes to the write model store, and it can be difficult to detect
when a user has issued a request based on stale read data.

When to use this pattern

Consider CQRS for the following scenarios:
Collaborative domains where many users access the same data in parallel. CQRS allows you to define
commands with enough granularity to minimize merge conflicts at the domain level, and conflicts that do
arise can be merged by the command.
Task-based user interfaces where users are guided through a complex process as a series of steps or with
complex domain models. The write model has a full command-processing stack with business logic, input
validation, and business validation. The write model may treat a set of associated objects as a single unit
for data changes (an aggregate, in DDD terminology) and ensure that these objects are always in a
consistent state. The read model has no business logic or validation stack, and just returns a DTO for use in
a view model. The read model is eventually consistent with the write model.
Scenarios where performance of data reads must be fine tuned separately from performance of data
writes, especially when the number of reads is much greater than the number of writes. In this scenario,
you can scale out the read model, but run the write model on just a few instances. A small number of write
model instances also helps to minimize the occurrence of merge conflicts.
Scenarios where one team of developers can focus on the complex domain model that is part of the write
model, and another team can focus on the read model and the user interfaces.
Scenarios where the system is expected to evolve over time and might contain multiple versions of the
model, or where business rules change regularly.
Integration with other systems, especially in combination with event sourcing, where the temporal failure
of one subsystem shouldn't affect the availability of the others.
This pattern isn't recommended when:
The domain or the business rules are simple.
A simple CRUD-style user interface and data access operations are sufficient.
Consider applying CQRS to limited sections of your system where it will be most valuable.

Event Sourcing and CQRS

The CQRS pattern is often used along with the Event Sourcing pattern. CQRS-based systems use separate read
and write data models, each tailored to relevant tasks and often located in physically separate stores. When used
with the Event Sourcing pattern, the store of events is the write model, and is the official source of information.
The read model of a CQRS-based system provides materialized views of the data, typically as highly
denormalized views. These views are tailored to the interfaces and display requirements of the application, which
helps to maximize both display and query performance.
Using the stream of events as the write store, rather than the actual data at a point in time, avoids update conflicts
on a single aggregate and maximizes performance and scalability. The events can be used to asynchronously
generate materialized views of the data that are used to populate the read store.
Because the event store is the official source of information, it is possible to delete the materialized views and
replay all past events to create a new representation of the current state when the system evolves, or when the
read model must change. The materialized views are in effect a durable read-only cache of the data.
When using CQRS combined with the Event Sourcing pattern, consider the following:
As with any system where the write and read stores are separate, systems based on this pattern are only
eventually consistent. There will be some delay between the event being generated and the data store
being updated.
The pattern adds complexity because code must be created to initiate and handle events, and assemble or
update the appropriate views or objects required by queries or a read model. The complexity of the CQRS
pattern when used with the Event Sourcing pattern can make a successful implementation more difficult,
and requires a different approach to designing systems. However, event sourcing can make it easier to
model the domain, and makes it easier to rebuild views or create new ones because the intent of the
changes in the data is preserved.
Generating materialized views for use in the read model or projections of the data by replaying and
handling the events for specific entities or collections of entities can require significant processing time
and resource usage. This is especially true if it requires summation or analysis of values over long periods,
because all the associated events might need to be examined. Resolve this by implementing snapshots of
the data at scheduled intervals, such as a total count of the number of a specific action that have occurred,
or the current state of an entity.

Example
The following code shows some extracts from an example of a CQRS implementation that uses different
definitions for the read and the write models. The model interfaces don't dictate any features of the underlying
data stores, and they can evolve and be fine-tuned independently because these interfaces are separated.
The following code shows the read model definition.
 // Query interface
namespace ReadModel
{
public interface ProductsDao
{
ProductDisplay FindById(int productId);
ICollection<ProductDisplay> FindByName(string name);
ICollection<ProductInventory> FindOutOfStockProducts();
ICollection<ProductDisplay> FindRelatedProducts(int productId);
}

public class ProductDisplay

{
public int Id { get; set; }
public string Name { get; set; }
public string Description { get; set; }
public decimal UnitPrice { get; set; }
public bool IsOutOfStock { get; set; }
public double UserRating { get; set; }
}

public class ProductInventory

{
public int Id { get; set; }
public string Name { get; set; }
public int CurrentStock { get; set; }
}
}

The system allows users to rate products. The application code does this using the RateProduct command shown
in the following code.

public interface ICommand

{
Guid Id { get; }
}

public class RateProduct : ICommand

{
public RateProduct()
{
this.Id = Guid.NewGuid();
}
public Guid Id { get; set; }
public int ProductId { get; set; }
public int Rating { get; set; }
public int UserId {get; set; }
}

The system uses the ProductsCommandHandler class to handle commands sent by the application. Clients typically
send commands to the domain through a messaging system such as a queue. The command handler accepts
these commands and invokes methods of the domain interface. The granularity of each command is designed to
reduce the chance of conflicting requests. The following code shows an outline of the ProductsCommandHandler
class.
 public class ProductsCommandHandler :
ICommandHandler<AddNewProduct>,
ICommandHandler<RateProduct>,
ICommandHandler<AddToInventory>,
ICommandHandler<ConfirmItemShipped>,
ICommandHandler<UpdateStockFromInventoryRecount>
{
private readonly IRepository<Product> repository;

public ProductsCommandHandler (IRepository<Product> repository)

{
this.repository = repository;
}

void Handle (AddNewProduct command)

{
...
}

void Handle (RateProduct command)

{
var product = repository.Find(command.ProductId);
if (product != null)
{
product.RateProduct(command.UserId, command.Rating);
repository.Save(product);
}
}

void Handle (AddToInventory command)

{
...
}

void Handle (ConfirmItemsShipped command)

{
...
}

void Handle (UpdateStockFromInventoryRecount command)

{
...
}
}

Related patterns and guidance

The following patterns and guidance are useful when implementing this pattern:
Data Consistency Primer. Explains the issues that are typically encountered due to eventual consistency
between the read and write data stores when using the CQRS pattern, and how these issues can be
resolved.
Data Partitioning Guidance. Describes best practices for dividing data into partitions that can be managed
and accessed separately to improve scalability, reduce contention, and optimize performance.
Event Sourcing pattern. Describes in more detail how Event Sourcing can be used with the CQRS pattern to
simplify tasks in complex domains while improving performance, scalability, and responsiveness. As well
as how to provide consistency for transactional data while maintaining full audit trails and history that can
enable compensating actions.
Materialized View pattern. The read model of a CQRS implementation can contain materialized views of
the write model data, or the read model can be used to generate materialized views.
The patterns & practices guide CQRS Journey. In particular, Introducing the Command Query
Responsibility Segregation pattern explores the pattern and when it's useful, and Epilogue: Lessons
Learned helps you understand some of the issues that come up when using this pattern.
The post CQRS by Martin Fowler, which explains the basics of the pattern and links to other useful
resources.
 Compensating Transaction pattern
12/18/2020 • 7 minutes to read • Edit Online

Undo the work performed by a series of steps, which together define an eventually consistent operation, if one
or more of the steps fail. Operations that follow the eventual consistency model are commonly found in cloud-
hosted applications that implement complex business processes and workflows.

Context and problem

Applications running in the cloud frequently modify data. This data might be spread across various data sources
held in different geographic locations. To avoid contention and improve performance in a distributed
environment, an application shouldn't try to provide strong transactional consistency. Rather, the application
should implement eventual consistency. In this model, a typical business operation consists of a series of
separate steps. While these steps are being performed, the overall view of the system state might be
inconsistent, but when the operation has completed and all of the steps have been executed the system should
become consistent again.

The Data Consistency Primer provides information about why distributed transactions don't scale well, and
the principles of the eventual consistency model.

A challenge in the eventual consistency model is how to handle a step that has failed. In this case it might be
necessary to undo all of the work completed by the previous steps in the operation. However, the data can't
simply be rolled back because other concurrent instances of the application might have changed it. Even in
cases where the data hasn't been changed by a concurrent instance, undoing a step might not simply be a
matter of restoring the original state. It might be necessary to apply various business-specific rules (see the
travel website described in the Example section).
If an operation that implements eventual consistency spans several heterogeneous data stores, undoing the
steps in the operation will require visiting each data store in turn. The work performed in every data store must
be undone reliably to prevent the system from remaining inconsistent.
Not all data affected by an operation that implements eventual consistency might be held in a database. In a
service oriented architecture (SOA) environment an operation could invoke an action in a service, and cause a
change in the state held by that service. To undo the operation, this state change must also be undone. This can
involve invoking the service again and performing another action that reverses the effects of the first.

Solution
The solution is to implement a compensating transaction. The steps in a compensating transaction must undo
the effects of the steps in the original operation. A compensating transaction might not be able to simply
replace the current state with the state the system was in at the start of the operation because this approach
could overwrite changes made by other concurrent instances of an application. Instead, it must be an intelligent
process that takes into account any work done by concurrent instances. This process will usually be application
specific, driven by the nature of the work performed by the original operation.
A common approach is to use a workflow to implement an eventually consistent operation that requires
compensation. As the original operation proceeds, the system records information about each step and how the
work performed by that step can be undone. If the operation fails at any point, the workflow rewinds back
through the steps it's completed and performs the work that reverses each step. Note that a compensating
transaction might not have to undo the work in the exact reverse order of the original operation, and it might be
possible to perform some of the undo steps in parallel.

This approach is similar to the Sagas strategy discussed in Clemens Vasters’ blog.

A compensating transaction is also an eventually consistent operation and it could also fail. The system should
be able to resume the compensating transaction at the point of failure and continue. It might be necessary to
repeat a step that's failed, so the steps in a compensating transaction should be defined as idempotent
commands. For more information, see Idempotency Patterns on Jonathan Oliver’s blog.
In some cases it might not be possible to recover from a step that has failed except through manual
intervention. In these situations the system should raise an alert and provide as much information as possible
about the reason for the failure.

Issues and considerations

Consider the following points when deciding how to implement this pattern:
It might not be easy to determine when a step in an operation that implements eventual consistency has failed.
A step might not fail immediately, but instead could block. It might be necessary to implement some form of
time-out mechanism.
-Compensation logic isn't easily generalized. A compensating transaction is application specific. It relies on the
application having sufficient information to be able to undo the effects of each step in a failed operation.
You should define the steps in a compensating transaction as idempotent commands. This enables the steps to
be repeated if the compensating transaction itself fails.
The infrastructure that handles the steps in the original operation, and the compensating transaction, must be
resilient. It must not lose the information required to compensate for a failing step, and it must be able to
reliably monitor the progress of the compensation logic.
A compensating transaction doesn't necessarily return the data in the system to the state it was in at the start of
the original operation. Instead, it compensates for the work performed by the steps that completed successfully
before the operation failed.
The order of the steps in the compensating transaction doesn't necessarily have to be the exact opposite of the
steps in the original operation. For example, one data store might be more sensitive to inconsistencies than
another, and so the steps in the compensating transaction that undo the changes to this store should occur first.
Placing a short-term timeout-based lock on each resource that's required to complete an operation, and
obtaining these resources in advance, can help increase the likelihood that the overall activity will succeed. The
work should be performed only after all the resources have been acquired. All actions must be finalized before
the locks expire.
Consider using retry logic that is more forgiving than usual to minimize failures that trigger a compensating
transaction. If a step in an operation that implements eventual consistency fails, try handling the failure as a
transient exception and repeat the step. Only stop the operation and initiate a compensating transaction if a step
fails repeatedly or irrecoverably.

Many of the challenges of implementing a compensating transaction are the same as those with
implementing eventual consistency. See the section Considerations for Implementing Eventual Consistency
in the Data Consistency Primer for more information.

When to use this pattern

Use this pattern only for operations that must be undone if they fail. If possible, design solutions to avoid the
complexity of requiring compensating transactions.

Example
A travel website lets customers book itineraries. A single itinerary might comprise a series of flights and hotels.
A customer traveling from Seattle to London and then on to Paris could perform the following steps when
creating an itinerary:
1. Book a seat on flight F1 from Seattle to London.
2. Book a seat on flight F2 from London to Paris.
3. Book a seat on flight F3 from Paris to Seattle.
4. Reserve a room at hotel H1 in London.
5. Reserve a room at hotel H2 in Paris.
These steps constitute an eventually consistent operation, although each step is a separate action. Therefore, as
well as performing these steps, the system must also record the counter operations necessary to undo each
step in case the customer decides to cancel the itinerary. The steps necessary to perform the counter operations
can then run as a compensating transaction.
Notice that the steps in the compensating transaction might not be the exact opposite of the original steps, and
the logic in each step in the compensating transaction must take into account any business-specific rules. For
example, unbooking a seat on a flight might not entitle the customer to a complete refund of any money paid.
The figure illustrates generating a compensating transaction to undo a long-running transaction to book a
travel itinerary.

NOTE
It might be possible for the steps in the compensating transaction to be performed in parallel, depending on how you've
designed the compensating logic for each step.

In many business solutions, failure of a single step doesn't always necessitate rolling the system back by using a
compensating transaction. For example, if—after having booked flights F1, F2, and F3 in the travel website
scenario—the customer is unable to reserve a room at hotel H1, it's preferable to offer the customer a room at a
different hotel in the same city rather than canceling the flights. The customer can still decide to cancel (in which
case the compensating transaction runs and undoes the bookings made on flights F1, F2, and F3), but this
decision should be made by the customer rather than by the system.

Related patterns and guidance

The following patterns and guidance might also be relevant when implementing this pattern:
Data Consistency Primer. The Compensating Transaction pattern is often used to undo operations that
implement the eventual consistency model. This primer provides information on the benefits and
tradeoffs of eventual consistency.
Scheduler-Agent-Supervisor pattern. Describes how to implement resilient systems that perform
business operations that use distributed services and resources. Sometimes, it might be necessary to
undo the work performed by an operation by using a compensating transaction.
Retry pattern. Compensating transactions can be expensive to perform, and it might be possible to
minimize their use by implementing an effective policy of retrying failing operations by following the
Retry pattern.
 Competing Consumers pattern
12/18/2020 • 8 minutes to read • Edit Online

Enable multiple concurrent consumers to process messages received on the same messaging channel. This
enables a system to process multiple messages concurrently to optimize throughput, to improve scalability and
availability, and to balance the workload.

Context and problem

An application running in the cloud is expected to handle a large number of requests. Rather than process each
request synchronously, a common technique is for the application to pass them through a messaging system to
another service (a consumer service) that handles them asynchronously. This strategy helps to ensure that the
business logic in the application isn't blocked while the requests are being processed.
The number of requests can vary significantly over time for many reasons. A sudden increase in user activity or
aggregated requests coming from multiple tenants can cause an unpredictable workload. At peak hours a
system might need to process many hundreds of requests per second, while at other times the number could be
very small. Additionally, the nature of the work performed to handle these requests might be highly variable.
Using a single instance of the consumer service can cause that instance to become flooded with requests, or the
messaging system might be overloaded by an influx of messages coming from the application. To handle this
fluctuating workload, the system can run multiple instances of the consumer service. However, these consumers
must be coordinated to ensure that each message is only delivered to a single consumer. The workload also
needs to be load balanced across consumers to prevent an instance from becoming a bottleneck.

Solution
Use a message queue to implement the communication channel between the application and the instances of
the consumer service. The application posts requests in the form of messages to the queue, and the consumer
service instances receive messages from the queue and process them. This approach enables the same pool of
consumer service instances to handle messages from any instance of the application. The figure illustrates using
a message queue to distribute work to instances of a service.

This solution has the following benefits:

It provides a load-leveled system that can handle wide variations in the volume of requests sent by
application instances. The queue acts as a buffer between the application instances and the consumer
service instances. This can help to minimize the impact on availability and responsiveness for both the
 application and the service instances, as described by the Queue-based Load Leveling pattern. Handling a
message that requires some long-running processing doesn't prevent other messages from being
handled concurrently by other instances of the consumer service.
It improves reliability. If a producer communicates directly with a consumer instead of using this pattern,
but doesn't monitor the consumer, there's a high probability that messages could be lost or fail to be
processed if the consumer fails. In this pattern, messages aren't sent to a specific service instance. A failed
service instance won't block a producer, and messages can be processed by any working service instance.
It doesn't require complex coordination between the consumers, or between the producer and the
consumer instances. The message queue ensures that each message is delivered at least once.
It's scalable. The system can dynamically increase or decrease the number of instances of the consumer
service as the volume of messages fluctuates.
It can improve resiliency if the message queue provides transactional read operations. If a consumer
service instance reads and processes the message as part of a transactional operation, and the consumer
service instance fails, this pattern can ensure that the message will be returned to the queue to be picked
up and handled by another instance of the consumer service.

Issues and considerations

Consider the following points when deciding how to implement this pattern:
Message ordering . The order in which consumer service instances receive messages isn't guaranteed,
and doesn't necessarily reflect the order in which the messages were created. Design the system to
ensure that message processing is idempotent because this will help to eliminate any dependency on the
order in which messages are handled. For more information, see Idempotency Patterns on Jonathon
Oliver’s blog.

Microsoft Azure Service Bus Queues can implement guaranteed first-in-first-out ordering of
messages by using message sessions. For more information, see Messaging Patterns Using Sessions.

Designing ser vices for resiliency . If the system is designed to detect and restart failed service
instances, it might be necessary to implement the processing performed by the service instances as
idempotent operations to minimize the effects of a single message being retrieved and processed more
than once.
Detecting poison messages . A malformed message, or a task that requires access to resources that
aren't available, can cause a service instance to fail. The system should prevent such messages being
returned to the queue, and instead capture and store the details of these messages elsewhere so that they
can be analyzed if necessary.
Handling results . The service instance handling a message is fully decoupled from the application logic
that generates the message, and they might not be able to communicate directly. If the service instance
generates results that must be passed back to the application logic, this information must be stored in a
location that's accessible to both. In order to prevent the application logic from retrieving incomplete data
the system must indicate when processing is complete.

If you're using Azure, a worker process can pass results back to the application logic by using a
dedicated message reply queue. The application logic must be able to correlate these results with the
original message. This scenario is described in more detail in the Asynchronous Messaging Primer.

Scaling the messaging system . In a large-scale solution, a single message queue could be
overwhelmed by the number of messages and become a bottleneck in the system. In this situation,
 consider partitioning the messaging system to send messages from specific producers to a particular
queue, or use load balancing to distribute messages across multiple message queues.
Ensuring reliability of the messaging system . A reliable messaging system is needed to guarantee
that after the application enqueues a message it won't be lost. This is essential for ensuring that all
messages are delivered at least once.

When to use this pattern

Use this pattern when:
The workload for an application is divided into tasks that can run asynchronously.
Tasks are independent and can run in parallel.
The volume of work is highly variable, requiring a scalable solution.
The solution must provide high availability, and must be resilient if the processing for a task fails.
This pattern might not be useful when:
It's not easy to separate the application workload into discrete tasks, or there's a high degree of dependence
between tasks.
Tasks must be performed synchronously, and the application logic must wait for a task to complete before
continuing.
Tasks must be performed in a specific sequence.

Some messaging systems support sessions that enable a producer to group messages together and ensure
that they're all handled by the same consumer. This mechanism can be used with prioritized messages (if
they are supported) to implement a form of message ordering that delivers messages in sequence from a
producer to a single consumer.

Example
Azure provides Service Bus Queues and Azure Function queue triggers that, when combined, are a direct
implementation of this cloud design pattern. Azure Functions integrate with Azure Service Bus via triggers and
bindings. Integrating with Service Bus allows you to build functions that consume queue messages sent by
publishers. The publishing application(s) will post messages to a queue, and consumers, implemented as Azure
Functions, can retrieve messages from this queue and handle them.
For resiliency, a Service Bus queue enables a consumer to use PeekLock mode when it retrieves a message from
the queue; this mode doesn't actually remove the message, but simply hides it from other consumers. The Azure
Functions runtime receives a message in PeekLock mode, if the function finishes successfully it calls Complete
on the message, or it may call Abandon if the function fails, and the message will become visible again, allowing
another consumer to retrieve it. If the function runs for a period longer than the PeekLock timeout, the lock is
automatically renewed as long as the function is running.
Azure Functions can scale out/in based on the depth of the queue, all acting as competing consumers of the
queue. If multiple instances of the functions are created they all compete by independently pulling and
processing the messages.
For detailed information on using Azure Service Bus queues, see Service Bus queues, topics, and subscriptions.
For information on Queue triggered Azure Functions, see Azure Service Bus trigger for Azure Functions.
The following code shows how you can create a new message and send it to a Service Bus Queue by using a
QueueClient instance.
 private string serviceBusConnectionString = ...;
...

public async Task SendMessagesAsync(CancellationToken ct)

{
try
{
var msgNumber = 0;

var queueClient = new QueueClient(serviceBusConnectionString, "myqueue");

while (!ct.IsCancellationRequested)
{
// Create a new message to send to the queue
string messageBody = $"Message {msgNumber}";
var message = new Message(Encoding.UTF8.GetBytes(messageBody));

// Write the body of the message to the console

this._logger.LogInformation($"Sending message: {messageBody}");

// Send the message to the queue

await queueClient.SendAsync(message);

this._logger.LogInformation("Message successfully sent.");

msgNumber++;
}
}
catch (Exception exception)
{
this._logger.LogException(exception.Message);
}
}

The following code example shows a consumer, written as a C# Azure Function, that reads message metadata
and logs a Service Bus Queue message. Note how the ServiceBusTrigger attribute is used to bind it to a Service
Bus Queue.

[FunctionName("ProcessQueueMessage")]
public static void Run(
[ServiceBusTrigger("myqueue", Connection = "ServiceBusConnectionString")]
string myQueueItem,
Int32 deliveryCount,
DateTime enqueuedTimeUtc,
string messageId,
ILogger log)
{
log.LogInformation($"C# ServiceBus queue trigger function consumed message: {myQueueItem}");
log.LogInformation($"EnqueuedTimeUtc={enqueuedTimeUtc}");
log.LogInformation($"DeliveryCount={deliveryCount}");
log.LogInformation($"MessageId={messageId}");
}

Related patterns and guidance

The following patterns and guidance might be relevant when implementing this pattern:
Asynchronous Messaging Primer. Message queues are an asynchronous communications mechanism. If
a consumer service needs to send a reply to an application, it might be necessary to implement some
form of response messaging. The Asynchronous Messaging Primer provides information on how to
implement request/reply messaging using message queues.
Autoscaling Guidance. It might be possible to start and stop instances of a consumer service since the
length of the queue applications post messages on varies. Autoscaling can help to maintain throughput
during times of peak processing.
Compute Resource Consolidation pattern. It might be possible to consolidate multiple instances of a
consumer service into a single process to reduce costs and management overhead. The Compute
Resource Consolidation pattern describes the benefits and tradeoffs of following this approach.
Queue-based Load Leveling pattern. Introducing a message queue can add resiliency to the system,
enabling service instances to handle widely varying volumes of requests from application instances. The
message queue acts as a buffer, which levels the load. The Queue-based Load Leveling pattern describes
this scenario in more detail.
 Compute Resource Consolidation pattern
12/18/2020 • 12 minutes to read • Edit Online

Consolidate multiple tasks or operations into a single computational unit. This can increase compute resource
utilization, and reduce the costs and management overhead associated with performing compute processing in
cloud-hosted applications.

Context and problem

A cloud application often implements a variety of operations. In some solutions it makes sense to follow the
design principle of separation of concerns initially, and divide these operations into separate computational units
that are hosted and deployed individually (for example, as separate App Service web apps, separate Virtual
Machines, or separate Cloud Service roles). However, although this strategy can help simplify the logical design of
the solution, deploying a large number of computational units as part of the same application can increase
runtime hosting costs and make management of the system more complex.
As an example, the figure shows the simplified structure of a cloud-hosted solution that is implemented using
more than one computational unit. Each computational unit runs in its own virtual environment. Each function has
been implemented as a separate task (labeled Task A through Task E) running in its own computational unit.

Each computational unit consumes chargeable resources, even when it's idle or lightly used. Therefore, this isn't
always the most cost-effective solution.
In Azure, this concern applies to roles in a Cloud Service, App Services, and Virtual Machines. These items run in
their own virtual environment. Running a collection of separate roles, websites, or virtual machines that are
designed to perform a set of well-defined operations, but that need to communicate and cooperate as part of a
single solution, can be an inefficient use of resources.

Solution
To help reduce costs, increase utilization, improve communication speed, and reduce management it's possible to
consolidate multiple tasks or operations into a single computational unit.
Tasks can be grouped according to criteria based on the features provided by the environment and the costs
associated with these features. A common approach is to look for tasks that have a similar profile concerning their
scalability, lifetime, and processing requirements. Grouping these together allows them to scale as a unit. The
elasticity provided by many cloud environments enables additional instances of a computational unit to be started
and stopped according to the workload. For example, Azure provides autoscaling that you can apply to roles in a
Cloud Service, App Services, and Virtual Machines. For more information, see Autoscaling Guidance.
As a counter example to show how scalability can be used to determine which operations shouldn't be grouped
together, consider the following two tasks:
Task 1 polls for infrequent, time-insensitive messages sent to a queue.
Task 2 handles high-volume bursts of network traffic.
The second task requires elasticity that can involve starting and stopping a large number of instances of the
computational unit. Applying the same scaling to the first task would simply result in more tasks listening for
infrequent messages on the same queue, and is a waste of resources.
In many cloud environments it's possible to specify the resources available to a computational unit in terms of the
number of CPU cores, memory, disk space, and so on. Generally, the more resources specified, the greater the
cost. To save money, it's important to maximize the work an expensive computational unit performs, and not let it
become inactive for an extended period.
If there are tasks that require a great deal of CPU power in short bursts, consider consolidating these into a single
computational unit that provides the necessary power. However, it's important to balance this need to keep
expensive resources busy against the contention that could occur if they are over stressed. Long-running,
compute-intensive tasks shouldn't share the same computational unit, for example.

Issues and considerations

Consider the following points when implementing this pattern:
Scalability and elasticity . Many cloud solutions implement scalability and elasticity at the level of the
computational unit by starting and stopping instances of units. Avoid grouping tasks that have conflicting
scalability requirements in the same computational unit.
Lifetime . The cloud infrastructure periodically recycles the virtual environment that hosts a computational unit.
When there are many long-running tasks inside a computational unit, it might be necessary to configure the unit
to prevent it from being recycled until these tasks have finished. Alternatively, design the tasks by using a check-
pointing approach that enables them to stop cleanly, and continue at the point they were interrupted when the
computational unit is restarted.
Release cadence . If the implementation or configuration of a task changes frequently, it might be necessary to
stop the computational unit hosting the updated code, reconfigure and redeploy the unit, and then restart it. This
process will also require that all other tasks within the same computational unit are stopped, redeployed, and
restarted.
Security . Tasks in the same computational unit might share the same security context and be able to access the
same resources. There must be a high degree of trust between the tasks, and confidence that one task isn't going
to corrupt or adversely affect another. Additionally, increasing the number of tasks running in a computational unit
increases the attack surface of the unit. Each task is only as secure as the one with the most vulnerabilities.
Fault tolerance . If one task in a computational unit fails or behaves abnormally, it can affect the other tasks
running within the same unit. For example, if one task fails to start correctly it can cause the entire startup logic
for the computational unit to fail, and prevent other tasks in the same unit from running.
Contention . Avoid introducing contention between tasks that compete for resources in the same computational
unit. Ideally, tasks that share the same computational unit should exhibit different resource utilization
characteristics. For example, two compute-intensive tasks should probably not reside in the same computational
unit, and neither should two tasks that consume large amounts of memory. However, mixing a compute-intensive
task with a task that requires a large amount of memory is a workable combination.

NOTE
Consider consolidating compute resources only for a system that's been in production for a period of time so that operators
and developers can monitor the system and create a heat map that identifies how each task uses differing resources. This
map can be used to determine which tasks are good candidates for sharing compute resources.

Complexity . Combining multiple tasks into a single computational unit adds complexity to the code in the unit,
possibly making it more difficult to test, debug, and maintain.
Stable logical architecture . Design and implement the code in each task so that it shouldn't need to change,
even if the physical environment the task runs in does change.
Other strategies . Consolidating compute resources is only one way to help reduce costs associated with running
multiple tasks concurrently. It requires careful planning and monitoring to ensure that it remains an effective
approach. Other strategies might be more appropriate, depending on the nature of the work and where the users
these tasks are running are located. For example, functional decomposition of the workload (as described by the
Compute Partitioning Guidance) might be a better option.

When to use this pattern

Use this pattern for tasks that are not cost effective if they run in their own computational units. If a task spends
much of its time idle, running this task in a dedicated unit can be expensive.
This pattern might not be suitable for tasks that perform critical fault-tolerant operations, or tasks that process
highly sensitive or private data and require their own security context. These tasks should run in their own isolated
environment, in a separate computational unit.

Example
When building a cloud service on Azure, it’s possible to consolidate the processing performed by multiple tasks
into a single role. Typically this is a worker role that performs background or asynchronous processing tasks.

In some cases it's possible to include background or asynchronous processing tasks in the web role. This
technique helps to reduce costs and simplify deployment, although it can impact the scalability and
responsiveness of the public-facing interface provided by the web role.

The role is responsible for starting and stopping the tasks. When the Azure fabric controller loads a role, it raises
the Start event for the role. You can override the OnStart method of the WebRole or WorkerRole class to handle
this event, perhaps to initialize the data and other resources the tasks in this method depend on.
When the OnStart method completes, the role can start responding to requests. You can find more information
and guidance about using the OnStart and Run methods in a role in the Application Startup Processes section in
the patterns & practices guide Moving Applications to the Cloud.

Keep the code in the OnStart method as concise as possible. Azure doesn't impose any limit on the time
taken for this method to complete, but the role won't be able to start responding to network requests sent to it
until this method completes.

When the OnStart method has finished, the role executes the Run method. At this point, the fabric controller can
start sending requests to the role.
Place the code that actually creates the tasks in the Run method. Note that the Run method defines the lifetime
of the role instance. When this method completes, the fabric controller will arrange for the role to be shut down.
When a role shuts down or is recycled, the fabric controller prevents any more incoming requests being received
from the load balancer and raises the Stop event. You can capture this event by overriding the OnStop method of
the role and perform any tidying up required before the role terminates.

Any actions performed in the OnStop method must be completed within five minutes (or 30 seconds if you
are using the Azure emulator on a local computer). Otherwise the Azure fabric controller assumes that the
role has stalled and will force it to stop.

The tasks are started by the Run method that waits for the tasks to complete. The tasks implement the business
logic of the cloud service, and can respond to messages posted to the role through the Azure load balancer. The
figure shows the lifecycle of tasks and resources in a role in an Azure cloud service.

The WorkerRole.cs file in the ComputeResourceConsolidation.Worker project shows an example of how you might
implement this pattern in an Azure cloud service.

The ComputeResourceConsolidation.Worker project is part of the ComputeResourceConsolidation solution

available for download from GitHub.

The MyWorkerTask1 and the MyWorkerTask2 methods illustrate how to perform different tasks within the same
worker role. The following code shows MyWorkerTask1 . This is a simple task that sleeps for 30 seconds and then
outputs a trace message. It repeats this process until the task is canceled. The code in MyWorkerTask2 is similar.
 // A sample worker role task.
private static async Task MyWorkerTask1(CancellationToken ct)
{
// Fixed interval to wake up and check for work and/or do work.
var interval = TimeSpan.FromSeconds(30);

try
{
while (!ct.IsCancellationRequested)
{
// Wake up and do some background processing if not canceled.
// TASK PROCESSING CODE HERE
Trace.TraceInformation("Doing Worker Task 1 Work");

// Go back to sleep for a period of time unless asked to cancel.

// Task.Delay will throw an OperationCanceledException when canceled.
await Task.Delay(interval, ct);
}
}
catch (OperationCanceledException)
{
// Expect this exception to be thrown in normal circumstances or check
// the cancellation token. If the role instances are shutting down, a
// cancellation request will be signaled.
Trace.TraceInformation("Stopping service, cancellation requested");

// Rethrow the exception.

throw;
}
}

The sample code shows a common implementation of a background process. In a real world application you
can follow this same structure, except that you should place your own processing logic in the body of the loop
that waits for the cancellation request.

After the worker role has initialized the resources it uses, the Run method starts the two tasks concurrently, as
shown here.
 /// <summary>
/// The cancellation token source use to cooperatively cancel running tasks
/// </summary>
private readonly CancellationTokenSource cts = new CancellationTokenSource();

/// <summary>
/// List of running tasks on the role instance
/// </summary>
private readonly List<Task> tasks = new List<Task>();

// RoleEntry Run() is called after OnStart().

// Returning from Run() will cause a role instance to recycle.
public override void Run()
{
// Start worker tasks and add to the task list
tasks.Add(MyWorkerTask1(cts.Token));
tasks.Add(MyWorkerTask2(cts.Token));

foreach (var worker in this.workerTasks)

{
this.tasks.Add(worker);
}

Trace.TraceInformation("Worker host tasks started");

// The assumption is that all tasks should remain running and not return,
// similar to role entry Run() behavior.
try
{
Task.WaitAll(tasks.ToArray());
}
catch (AggregateException ex)
{
Trace.TraceError(ex.Message);

// If any of the inner exceptions in the aggregate exception

// are not cancellation exceptions then re-throw the exception.
ex.Handle(innerEx => (innerEx is OperationCanceledException));
}

// If there wasn't a cancellation request, stop all tasks and return from Run()
// An alternative to canceling and returning when a task exits would be to
// restart the task.
if (!cts.IsCancellationRequested)
{
Trace.TraceInformation("Task returned without cancellation request");
Stop(TimeSpan.FromMinutes(5));
}
}
...

In this example, the Run method waits for tasks to be completed. If a task is canceled, the Run method assumes
that the role is being shut down and waits for the remaining tasks to be canceled before finishing (it waits for a
maximum of five minutes before terminating). If a task fails due to an expected exception, the Run method
cancels the task.

You could implement more comprehensive monitoring and exception handling strategies in the Run method
such as restarting tasks that have failed, or including code that enables the role to stop and start individual
tasks.

The Stop method shown in the following code is called when the fabric controller shuts down the role instance
(it's invoked from the OnStop method). The code stops each task gracefully by canceling it. If any task takes more
than five minutes to complete, the cancellation processing in the Stop method ceases waiting and the role is
terminated.

// Stop running tasks and wait for tasks to complete before returning
// unless the timeout expires.
private void Stop(TimeSpan timeout)
{
Trace.TraceInformation("Stop called. Canceling tasks.");
// Cancel running tasks.
cts.Cancel();

Trace.TraceInformation("Waiting for canceled tasks to finish and return");

// Wait for all the tasks to complete before returning. Note that the
// emulator currently allows 30 seconds and Azure allows five
// minutes for processing to complete.
try
{
Task.WaitAll(tasks.ToArray(), timeout);
}
catch (AggregateException ex)
{
Trace.TraceError(ex.Message);

// If any of the inner exceptions in the aggregate exception

// are not cancellation exceptions then rethrow the exception.
ex.Handle(innerEx => (innerEx is OperationCanceledException));
}
}

Related patterns and guidance

The following patterns and guidance might also be relevant when implementing this pattern:
Autoscaling Guidance. Autoscaling can be used to start and stop instances of service hosting computational
resources, depending on the anticipated demand for processing.
Compute Partitioning Guidance. Describes how to allocate the services and components in a cloud service
in a way that helps to minimize running costs while maintaining the scalability, performance, availability,
and security of the service.
This pattern includes a downloadable sample application.
 Deployment stamps
12/18/2020 • 12 minutes to read • Edit Online

The deployment stamp pattern involves deploying multiple independent copies of application components,
including data stores. Each individual copy is called a stamp, or sometimes a service unit or scale unit. This
approach can improve the scalability of your solution, allow you to deploy instances across multiple regions, and
separate your customer data.
When hosting an application in the cloud there are certain considerations to be made. One key thing to keep in
mind is the performance and reliability of your application. If you host a single instance of your solution, you
might be subject to the following limitations:
Scale limits. Deploying a single instance of your application may result in natural scaling limits. For example,
you may use services that have limits on the number of inbound connections, host names, TCP sockets, or
other resources.
Non-linear scaling or cost. Some of your solution's components may not scale linearly with the number of
requests or the amount of data. Instead, there can be a sudden decrease in performance or increase in cost
once a threshold has been met. For example, you may use a database and discover that the marginal cost of
adding more capacity (scaling up) becomes prohibitive, and that scaling out is a more cost-effective strategy.
Similarly, Azure Front Door has higher per-domain pricing when a high number of custom domains are
deployed, and it may be better to spread the custom domains across multiple Front Door instances.
Separation of customers. You may need to keep certain customers' data isolated from other customers'
data. Similarly, you may have some customers that require more system resources to service than others, and
consider grouping them on different sets of infrastructure.
Handling single- and multi-tenant instances. You may have some large customers who need their own
independent instances of your solution. You may also have a pool of smaller customers who can share a multi-
tenant deployment.
Complex deployment requirements. You may need to deploy updates to your service in a controlled
manner, and to deploy to different subsets of your customer base at different times.
Update frequency. You may have some customers who are tolerant of having frequent updates to your
system, while others may be risk-averse and want infrequent updates to the system that services their
requests. It may make sense to have these customers deployed to isolated environments.
Geographical or geopolitical restrictions. To architect for low latency, or to comply with data sovereignty
requirements, you may deploy some of your customers into specific regions.

Implementation
To avoid these issues, consider deploying copies of your solution's components multiple times. Stamps operate
independently of each other and can be deployed and updated independently. A single geographical region may
contain a single stamp, or may contain multiple stamps to allow for horizontal scale-out within the region. Stamps
contain a subset of your customers.
Deployment stamps can apply whether your solution uses infrastructure as a service (IaaS) or platform as a
service (PaaS) components, or a mixture of both. Typically IaaS workloads require more intervention to scale, so
the pattern may be useful for IaaS-heavy workloads to allow for scaling out.
Stamps can be used to implement deployment rings. If different customers want to receive service updates at
different frequencies, they can be grouped onto different stamps, and each stamp could have updates deployed at
different cadences.
Because stamps run independently from each other, data is implicitly sharded. Furthermore, a single stamp can
make use of further sharding to internally allow for scalability and elasticity within the stamp.
The deployment stamp pattern is used internally by many Azure services, including App Service, Azure Stack, and
Azure Storage.
Deployment stamps are related to, but distinct from, geodes. In a deployment stamp architecture, multiple
independent instances of your system are deployed and contain a subset of your customers and users. In geodes,
all instances can serve requests from any users, but this architecture is often more complex to design and build.
You may also consider mixing the two patterns within one solution; the traffic routing approach described below
is an example of such a hybrid scenario.
Deployment
Because of the complexity that is involved in deploying identical copies of the same components, good DevOps
practices are critical to ensure success when implementing this pattern. Consider describing your infrastructure as
code, such as using Azure Resource Manager templates and scripts. With this approach, you can ensure that the
deployment of each stamp is predictable. It also reduces the likelihood of human errors such as accidental
mismatches in configuration between stamps.
You can deploy updates automatically to all stamps in parallel, in which case you might consider technologies like
Resource Manager templates or Azure Deployment Manager to coordinate the deployment of your infrastructure
and applications. Alternatively, you may decide to gradually roll out updates to some stamps first, and then
progressively to others. Azure Deployment Manager can also manage this type of staged rollout, or you could
consider using a release management tool like Azure Pipelines to orchestrate deployments to each stamp.
Carefully consider the topology of the Azure subscriptions and resource groups for your deployments:
Typically a subscription will contain all resources for a single solution, so in general consider using a single
subscription for all stamps. However, some Azure services impose subscription-wide quotas, so if you are
using this pattern to allow for a high degree of scale-out, you may need to consider deploying stamps across
different subscriptions.
Resource groups are generally used to deploy components with the same lifecycle. If you plan to deploy
updates to all of your stamps at once, consider using a single resource group to contain all of the components
for all of your stamps, and use resource naming conventions and tags to identify the components that belong
to each stamp. Alternatively, if you plan to deploy updates to each stamp independently, consider deploying
each stamp into its own resource group.
Capacity planning
Use load and performance testing to determine the approximate load that a given stamp can accommodate. Load
metrics may be based on the number of customers/tenants that a single stamp can accommodate, or metrics
from the services that the components within the stamp emit. Ensure that you have sufficient instrumentation to
measure when a given stamp is approaching its capacity, and the ability to deploy new stamps quickly to respond
to demand.
Traffic routing
The Deployment Stamp pattern works well if each stamp is addressed independently. For example, if Contoso
deploys the same API application across multiple stamps, they might consider using DNS to route traffic to the
relevant stamp:
unit1.aus.myapi.contoso.com routes traffic to stamp unit1 within an Australian region.
unit2.aus.myapi.contoso.com routes traffic to stamp unit2 within an Australian region.
unit1.eu.myapi.contoso.com routes traffic to stamp unit1 within a European region.

Clients are then responsible for connecting to the correct stamp.

If a single ingress point for all traffic is required, a traffic routing service may be used to resolve the stamp for a
given request, customer, or tenant. The traffic routing service either directs the client to the relevant URL for the
stamp (for example, using an HTTP 302 response status code), or may act as a reverse proxy and forward the
traffic to the relevant stamp without the client being aware.
A centralized traffic routing service can be a complex component to design, especially when a solution runs across
multiple regions. Consider deploying the traffic routing service into multiple regions (potentially including every
region that stamps are deployed into), and then ensuring the data store (mapping tenants to stamps) is
synchronized. The traffic routing component may itself by an instance of the geode pattern.
For example, Azure API Management could be deployed to act in the traffic routing service role. It can determine
the appropriate stamp for a request by looking up data in a Cosmos DB collection storing the mapping between
tenants and stamps. API Management can then dynamically set the back-end URL to the relevant stamp's API
service.
To enable geo-distribution of requests and geo-redundancy of the traffic routing service, API Management can be
deployed across multiple regions, or Azure Front Door can be used to direct traffic to the closest instance. Front
Door can be configured with a backend pool, enabling requests to be directed to the closest available API
Management instance. The global distribution features of Cosmos DB can be used to keep the mapping
information updated across each region.

If a traffic routing service is included in your solution, consider whether it acts as a gateway and could therefore
perform gateway offloading for services such as token validation, throttling, and authorization.

Issues and considerations

You should consider the following points when deciding how to implement this pattern:
Deployment process. When deploying multiple stamps, it is highly advisable to have automated and fully
repeatable deployment processes. Consider using Resource Manager templates or Terraform templates to
declaratively define the stamp.
Cross-stamp operations. When your solution is deployed independently across multiple stamps, questions
like "how many customers do we have across all of our stamps?" can become more complex to answer.
Queries may need to be executed against each stamp and the results aggregated. Alternatively, consider
having all of the stamps publish data into a centralized data warehouse for consolidated reporting.
 Determining scale-out policies. Stamps have a finite capacity, which might be defined using a proxy metric
such as the number of tenants that can be deployed to the stamp. It is important to monitor the available
capacity and used capacity for each stamp, and to proactively deploy additional stamps to allow for new
tenants to be directed to them.
Cost. The Deployment Stamp pattern involves deploying multiple copies of your infrastructure component,
which will likely involve a substantial increase in the cost of operating your solution.
Moving between stamps. As each stamp is deployed and operated independently, moving tenants between
stamps can be difficult. Your application would need custom logic to transmit the information about a given
customer to a different stamp, and then to remove the tenant's information from the original stamp. This
process may require a backplane for communication between stamps, further increasing the complexity of the
overall solution.
Traffic routing. As described above, routing traffic to the correct stamp for a given request can require an
additional component to resolve tenants to stamps. This component, in turn, may need to be made highly
available.
Shared components. You may have some components that can be shared across stamps. For example, if you
have a shared single-page app for all tenants, consider deploying that into one region and using Azure CDN to
replicate it globally.

When to use deployment stamps

This pattern is useful when you have:
Natural limits on scalability. For example, if some components cannot or should not scale beyond a certain
number of customers or requests, consider scaling out using stamps.
A requirement to separate certain tenants from others. If you have customers that cannot be deployed into a
multi-tenant stamp with other customers due to security concerns, they can be deployed onto their own
isolated stamp.
A need to have some tenants on different versions of your solution at the same time.
Multi-region applications where each tenant's data and traffic should be directed to a specific region.
A desire to achieve resiliency during outages. As stamps are independent of one another, if an outage affects a
single stamp then the tenants deployed to other stamps should not be affected. This isolation helps to contain
the 'blast radius' of an incident or outage.
This pattern is not suitable for:
Simple solutions that do not need to scale to a high degree.
Systems that can be easily scaled out or up within a single instance, such as by increasing the size of the
application layer or by increasing the reserved capacity for databases and the storage tier.
Solutions in which data should be replicated across all deployed instances. Consider the geode pattern for this
scenario.
Solutions in which only some components need to be scaled, but not others. For example, consider whether
your solution could be scaled by sharding the data store rather than deploying a new copy of all of the solution
components.
Solutions comprised solely of static content, such as a front-end JavaScript application. Consider storing such
content in a storage account and using Azure CDN.

Supporting technologies
Infrastructure as code. For example, Resource Manager templates, Azure CLI, Terraform, PowerShell, Bash.
Deployment Manager, which can orchestrate deployments of a solution across multiple stamps.
Azure Front Door, which can route traffic to a specific stamp or to a traffic routing service.
Example
The following example deploys multiple stamps of a simple PaaS solution, with an app service and a SQL
Database in each stamp. While stamps can be configured in any region that support the services deployed in the
template, for illustration purposes this template deploys two stamps within the West US 2 region and a further
stamp in the West Europe region. Within a stamp, the app service receives its own public DNS hostname and it
can receive connections independently of all other stamps.

WARNING
The example below uses a SQL Server administrator account. It's generally not a good practice to use an administrative
account from your application. For a real application, consider using a managed identity to connect from your application to
a SQL database, or use a least-privilege account.

Click the link below to deploy the solution.

NOTE
There are alternative approaches to deploying stamps with a Resource Manager template, including using nested templates
or linked templates to decouple the definition of each stamp from the iteration required to deploy multiple copies.

Example traffic routing approach

The following example deploys an implementation of a traffic routing solution that could be used with a set of
deployment stamps for a hypothetical API application. To allow for geographical distribution of incoming requests,
Front Door is deployed alongside multiple instances of API Management on the consumption tier. Each API
Management instance reads the tenant ID from the request URL and then looks up the relevant stamp for the
request from a geo-distributed Cosmos DB data store. The request is then forwarded to the relevant back-end
stamp.
Click the link below to deploy the solution.

Next steps
Sharding can be used as another, simpler, approach to scale out your data tier. Stamps implicitly shard their
data, but sharding does not require a Deployment Stamp. For more information, see Sharding Pattern.
If a traffic routing service is deployed, the Gateway Routing and Gateway Offloading patterns can be used
together to make the best use of this component.
 Event Sourcing pattern
12/18/2020 • 14 minutes to read • Edit Online

Instead of storing just the current state of the data in a domain, use an append-only store to record the full series
of actions taken on that data. The store acts as the system of record and can be used to materialize the domain
objects. This can simplify tasks in complex domains, by avoiding the need to synchronize the data model and the
business domain, while improving performance, scalability, and responsiveness. It can also provide consistency
for transactional data, and maintain full audit trails and history that can enable compensating actions.

Context and problem

Most applications work with data, and the typical approach is for the application to maintain the current state of
the data by updating it as users work with it. For example, in the traditional create, read, update, and delete
(CRUD) model a typical data process is to read data from the store, make some modifications to it, and update
the current state of the data with the new values—often by using transactions that lock the data.
The CRUD approach has some limitations:
CRUD systems perform update operations directly against a data store, which can slow down performance
and responsiveness, and limit scalability, due to the processing overhead it requires.
In a collaborative domain with many concurrent users, data update conflicts are more likely because the
update operations take place on a single item of data.
Unless there's an additional auditing mechanism that records the details of each operation in a separate
log, history is lost.

For a deeper understanding of the limits of the CRUD approach, see CRUD, Only When You Can Afford It.

Solution
The Event Sourcing pattern defines an approach to handling operations on data that's driven by a sequence of
events, each of which is recorded in an append-only store. Application code sends a series of events that
imperatively describe each action that has occurred on the data to the event store, where they're persisted. Each
event represents a set of changes to the data (such as AddedItemToOrder ).
The events are persisted in an event store that acts as the system of record (the authoritative data source) about
the current state of the data. The event store typically publishes these events so that consumers can be notified
and can handle them if needed. Consumers could, for example, initiate tasks that apply the operations in the
events to other systems, or perform any other associated action that's required to complete the operation. Notice
that the application code that generates the events is decoupled from the systems that subscribe to the events.
Typical uses of the events published by the event store are to maintain materialized views of entities as actions in
the application change them, and for integration with external systems. For example, a system can maintain a
materialized view of all customer orders that's used to populate parts of the UI. As the application adds new
orders, adds or removes items on the order, and adds shipping information, the events that describe these
changes can be handled and used to update the materialized view.
In addition, at any point it's possible for applications to read the history of events, and use it to materialize the
current state of an entity by playing back and consuming all the events related to that entity. This can occur on
demand to materialize a domain object when handling a request, or through a scheduled task so that the state of
the entity can be stored as a materialized view to support the presentation layer.
The figure shows an overview of the pattern, including some of the options for using the event stream such as
creating a materialized view, integrating events with external applications and systems, and replaying events to
create projections of the current state of specific entities.

The Event Sourcing pattern provides the following advantages:

Events are immutable and can be stored using an append-only operation. The user interface, workflow, or
process that initiated an event can continue, and tasks that handle the events can run in the background.
This, combined with the fact that there's no contention during the processing of transactions, can vastly
improve performance and scalability for applications, especially for the presentation level or user interface.
Events are simple objects that describe some action that occurred, together with any associated data
required to describe the action represented by the event. Events don't directly update a data store. They're
simply recorded for handling at the appropriate time. This can simplify implementation and management.
Events typically have meaning for a domain expert, whereas object-relational impedance mismatch can
make complex database tables hard to understand. Tables are artificial constructs that represent the
current state of the system, not the events that occurred.
Event sourcing can help prevent concurrent updates from causing conflicts because it avoids the
requirement to directly update objects in the data store. However, the domain model must still be designed
to protect itself from requests that might result in an inconsistent state.
The append-only storage of events provides an audit trail that can be used to monitor actions taken
against a data store, regenerate the current state as materialized views or projections by replaying the
events at any time, and assist in testing and debugging the system. In addition, the requirement to use
compensating events to cancel changes provides a history of changes that were reversed, which wouldn't
be the case if the model simply stored the current state. The list of events can also be used to analyze
application performance and detect user behavior trends, or to obtain other useful business information.
The event store raises events, and tasks perform operations in response to those events. This decoupling
of the tasks from the events provides flexibility and extensibility. Tasks know about the type of event and
the event data, but not about the operation that triggered the event. In addition, multiple tasks can handle
each event. This enables easy integration with other services and systems that only listen for new events
 raised by the event store. However, the event sourcing events tend to be very low level, and it might be
necessary to generate specific integration events instead.

Event sourcing is commonly combined with the CQRS pattern by performing the data management tasks in
response to the events, and by materializing views from the stored events.

Issues and considerations

Consider the following points when deciding how to implement this pattern:
The system will only be eventually consistent when creating materialized views or generating projections of data
by replaying events. There's some delay between an application adding events to the event store as the result of
handling a request, the events being published, and consumers of the events handling them. During this period,
new events that describe further changes to entities might have arrived at the event store.

NOTE
See the Data Consistency Primer for information about eventual consistency.

The event store is the permanent source of information, and so the event data should never be updated. The only
way to update an entity to undo a change is to add a compensating event to the event store. If the format (rather
than the data) of the persisted events needs to change, perhaps during a migration, it can be difficult to combine
existing events in the store with the new version. It might be necessary to iterate through all the events making
changes so they're compliant with the new format, or add new events that use the new format. Consider using a
version stamp on each version of the event schema to maintain both the old and the new event formats.
Multi-threaded applications and multiple instances of applications might be storing events in the event store. The
consistency of events in the event store is vital, as is the order of events that affect a specific entity (the order that
changes occur to an entity affects its current state). Adding a timestamp to every event can help to avoid issues.
Another common practice is to annotate each event resulting from a request with an incremental identifier. If two
actions attempt to add events for the same entity at the same time, the event store can reject an event that
matches an existing entity identifier and event identifier.
There's no standard approach, or existing mechanisms such as SQL queries, for reading the events to obtain
information. The only data that can be extracted is a stream of events using an event identifier as the criteria. The
event ID typically maps to individual entities. The current state of an entity can be determined only by replaying
all of the events that relate to it against the original state of that entity.
The length of each event stream affects managing and updating the system. If the streams are large, consider
creating snapshots at specific intervals such as a specified number of events. The current state of the entity can be
obtained from the snapshot and by replaying any events that occurred after that point in time. For more
information about creating snapshots of data, see Snapshot on Martin Fowler’s Enterprise Application
Architecture website and Primary-Subordinate Snapshot Replication.
Even though event sourcing minimizes the chance of conflicting updates to the data, the application must still be
able to deal with inconsistencies that result from eventual consistency and the lack of transactions. For example,
an event that indicates a reduction in stock inventory might arrive in the data store while an order for that item is
being placed, resulting in a requirement to reconcile the two operations either by advising the customer or
creating a back order.
Event publication might be "at least once," and so consumers of the events must be idempotent. They must not
reapply the update described in an event if the event is handled more than once. For example, if multiple
instances of a consumer maintain an aggregate an entity's property, such as the total number of orders placed,
only one must succeed in incrementing the aggregate when an order placed event occurs. While this isn't a key
characteristic of event sourcing, it's the usual implementation decision.
When to use this pattern
Use this pattern in the following scenarios:
When you want to capture intent, purpose, or reason in the data. For example, changes to a customer
entity can be captured as a series of specific event types such as Moved home, Closed account, or
Deceased.
When it's vital to minimize or completely avoid the occurrence of conflicting updates to data.
When you want to record events that occur, and be able to replay them to restore the state of a system, roll
back changes, or keep a history and audit log. For example, when a task involves multiple steps you might
need to execute actions to revert updates and then replay some steps to bring the data back into a
consistent state.
When using events is a natural feature of the operation of the application, and requires little additional
development or implementation effort.
When you need to decouple the process of inputting or updating data from the tasks required to apply
these actions. This might be to improve UI performance, or to distribute events to other listeners that take
action when the events occur. For example, integrating a payroll system with an expense submission
website so that events raised by the event store in response to data updates made in the website are
consumed by both the website and the payroll system.
When you want flexibility to be able to change the format of materialized models and entity data if
requirements change, or—when used in conjunction with CQRS—you need to adapt a read model or the
views that expose the data.
When used in conjunction with CQRS, and eventual consistency is acceptable while a read model is
updated, or the performance impact of rehydrating entities and data from an event stream is acceptable.
This pattern might not be useful in the following situations:
Small or simple domains, systems that have little or no business logic, or nondomain systems that
naturally work well with traditional CRUD data management mechanisms.
Systems where consistency and real-time updates to the views of the data are required.
Systems where audit trails, history, and capabilities to roll back and replay actions are not required.
Systems where there's only a very low occurrence of conflicting updates to the underlying data. For
example, systems that predominantly add data rather than updating it.

Example
A conference management system needs to track the number of completed bookings for a conference so that it
can check whether there are seats still available when a potential attendee tries to make a booking. The system
could store the total number of bookings for a conference in at least two ways:
The system could store the information about the total number of bookings as a separate entity in a
database that holds booking information. As bookings are made or canceled, the system could increment
or decrement this number as appropriate. This approach is simple in theory, but can cause scalability
issues if a large number of attendees are attempting to book seats during a short period of time. For
example, in the last day or so prior to the booking period closing.
The system could store information about bookings and cancellations as events held in an event store. It
could then calculate the number of seats available by replaying these events. This approach can be more
scalable due to the immutability of events. The system only needs to be able to read data from the event
 store, or append data to the event store. Event information about bookings and cancellations is never
modified.
The following diagram illustrates how the seat reservation subsystem of the conference management system
might be implemented using event sourcing.

The sequence of actions for reserving two seats is as follows:

1. The user interface issues a command to reserve seats for two attendees. The command is handled by a
separate command handler. A piece of logic that is decoupled from the user interface and is responsible
for handling requests posted as commands.
2. An aggregate containing information about all reservations for the conference is constructed by querying
the events that describe bookings and cancellations. This aggregate is called SeatAvailability , and is
contained within a domain model that exposes methods for querying and modifying the data in the
aggregate.

Some optimizations to consider are using snapshots (so that you don’t need to query and replay the
full list of events to obtain the current state of the aggregate), and maintaining a cached copy of the
aggregate in memory.
3. The command handler invokes a method exposed by the domain model to make the reservations.
4. The SeatAvailability aggregate records an event containing the number of seats that were reserved. The
next time the aggregate applies events, all the reservations will be used to compute how many seats
remain.
5. The system appends the new event to the list of events in the event store.
If a user cancels a seat, the system follows a similar process except the command handler issues a command that
generates a seat cancellation event and appends it to the event store.
As well as providing more scope for scalability, using an event store also provides a complete history, or audit
trail, of the bookings and cancellations for a conference. The events in the event store are the accurate record.
There is no need to persist aggregates in any other way because the system can easily replay the events and
restore the state to any point in time.

You can find more information about this example in Introducing Event Sourcing.

Related patterns and guidance

The following patterns and guidance might also be relevant when implementing this pattern:
Command and Query Responsibility Segregation (CQRS) pattern. The write store that provides the
permanent source of information for a CQRS implementation is often based on an implementation of the
Event Sourcing pattern. Describes how to segregate the operations that read data in an application from
the operations that update data by using separate interfaces.
Materialized View pattern. The data store used in a system based on event sourcing is typically not well
suited to efficient querying. Instead, a common approach is to generate prepopulated views of the data at
regular intervals, or when the data changes. Shows how this can be done.
Compensating Transaction pattern. The existing data in an event sourcing store is not updated, instead new
entries are added that transition the state of entities to the new values. To reverse a change, compensating
entries are used because it isn't possible to simply reverse the previous change. Describes how to undo the
work that was performed by a previous operation.
Data Consistency Primer. When using event sourcing with a separate read store or materialized views, the
read data won't be immediately consistent, instead it'll be only eventually consistent. Summarizes the
issues surrounding maintaining consistency over distributed data.
Data Partitioning Guidance. Data is often partitioned when using event sourcing to improve scalability,
reduce contention, and optimize performance. Describes how to divide data into discrete partitions, and
the issues that can arise.
 External Configuration Store pattern
12/18/2020 • 10 minutes to read • Edit Online

Move configuration information out of the application deployment package to a centralized location. This can
provide opportunities for easier management and control of configuration data, and for sharing configuration
data across applications and application instances.

Context and problem

The majority of application runtime environments include configuration information that's held in files deployed
with the application. In some cases, it's possible to edit these files to change the application behavior after it's been
deployed. However, changes to the configuration require the application be redeployed, often resulting in
unacceptable downtime and other administrative overhead.
Local configuration files also limit the configuration to a single application, but sometimes it would be useful to
share configuration settings across multiple applications. Examples include database connection strings, UI theme
information, or the URLs of queues and storage used by a related set of applications.
It's challenging to manage changes to local configurations across multiple running instances of the application,
especially in a cloud-hosted scenario. It can result in instances using different configuration settings while the
update is being deployed.
In addition, updates to applications and components might require changes to configuration schemas. Many
configuration systems don't support different versions of configuration information.

Solution
Store the configuration information in external storage, and provide an interface that can be used to quickly and
efficiently read and update configuration settings. The type of external store depends on the hosting and runtime
environment of the application. In a cloud-hosted scenario it's typically a cloud-based storage service, but could be
a hosted database or other system.
The backing store you choose for configuration information should have an interface that provides consistent and
easy-to-use access. It should expose the information in a correctly typed and structured format. The
implementation might also need to authorize users’ access in order to protect configuration data, and be flexible
enough to allow storage of multiple versions of the configuration (such as development, staging, or production,
including multiple release versions of each one).

Many built-in configuration systems read the data when the application starts up, and cache the data in
memory to provide fast access and minimize the impact on application performance. Depending on the type
of backing store used, and the latency of this store, it might be helpful to implement a caching mechanism
within the external configuration store. For more information, see the Caching Guidance. The figure illustrates
an overview of the External Configuration Store pattern with optional local cache.
Issues and considerations
Consider the following points when deciding how to implement this pattern:
Choose a backing store that offers acceptable performance, high availability, robustness, and can be backed up as
part of the application maintenance and administration process. In a cloud-hosted application, using a cloud
storage mechanism is usually a good choice to meet these requirements.
Design the schema of the backing store to allow flexibility in the types of information it can hold. Ensure that it
provides for all configuration requirements such as typed data, collections of settings, multiple versions of settings,
and any other features that the applications using it require. The schema should be easy to extend to support
additional settings as requirements change.
Consider the physical capabilities of the backing store, how it relates to the way configuration information is
stored, and the effects on performance. For example, storing an XML document containing configuration
information will require either the configuration interface or the application to parse the document in order to
read individual settings. It'll make updating a setting more complicated, though caching the settings can help to
offset slower read performance.
Consider how the configuration interface will permit control of the scope and inheritance of configuration settings.
For example, it might be a requirement to scope configuration settings at the organization, application, and the
machine level. It might need to support delegation of control over access to different scopes, and to prevent or
allow individual applications to override settings.
Ensure that the configuration interface can expose the configuration data in the required formats such as typed
values, collections, key/value pairs, or property bags.
Consider how the configuration store interface will behave when settings contain errors, or don't exist in the
backing store. It might be appropriate to return default settings and log errors. Also consider aspects such as the
case sensitivity of configuration setting keys or names, the storage and handling of binary data, and the ways that
null or empty values are handled.
Consider how to protect the configuration data to allow access to only the appropriate users and applications. This
is likely a feature of the configuration store interface, but it's also necessary to ensure that the data in the backing
store can't be accessed directly without the appropriate permission. Ensure strict separation between the
permissions required to read and to write configuration data. Also consider whether you need to encrypt some or
all of the configuration settings, and how this'll be implemented in the configuration store interface.
Centrally stored configurations, which change application behavior during runtime, are critically important and
should be deployed, updated, and managed using the same mechanisms as deploying application code. For
example, changes that can affect more than one application must be carried out using a full test and staged
deployment approach to ensure that the change is appropriate for all applications that use this configuration. If an
administrator edits a setting to update one application, it could adversely impact other applications that use the
same setting.
If an application caches configuration information, the application needs to be alerted if the configuration changes.
It might be possible to implement an expiration policy over cached configuration data so that this information is
automatically refreshed periodically and any changes picked up (and acted on).

When to use this pattern

This pattern is useful for:
Configuration settings that are shared between multiple applications and application instances, or where a
standard configuration must be enforced across multiple applications and application instances.
A standard configuration system that doesn't support all of the required configuration settings, such as
storing images or complex data types.
As a complementary store for some of the settings for applications, perhaps allowing applications to
override some or all of the centrally-stored settings.
As a way to simplify administration of multiple applications, and optionally for monitoring use of
configuration settings by logging some or all types of access to the configuration store.

Example
In a Microsoft Azure hosted application, a typical choice for storing configuration information externally is to use
Azure Storage. This is resilient, offers high performance, and is replicated three times with automatic failover to
offer high availability. Azure Table storage provides a key/value store with the ability to use a flexible schema for
the values. Azure Blob storage provides a hierarchical, container-based store that can hold any type of data in
individually named blobs.
The following example shows how a configuration store can be implemented over Blob storage to store and
expose configuration information. The BlobSettingsStore class abstracts Blob storage for holding configuration
information, and implements the ISettingsStore interface shown in the following code.

This code is provided in the ExternalConfigurationStore.Cloud project in the ExternalConfigurationStore

solution, available from GitHub.

public interface ISettingsStore

{
Task<ETag> GetVersionAsync();

Task<Dictionary<string, string>> FindAllAsync();

}

This interface defines methods for retrieving and updating configuration settings held in the configuration store,
and includes a version number that can be used to detect whether any configuration settings have been modified
recently. The BlobSettingsStore class uses the ETag property of the blob to implement versioning. The ETag
property is updated automatically each time the blob is written.

By design, this simple solution exposes all configuration settings as string values rather than typed values.

The ExternalConfigurationManager class provides a wrapper around a BlobSettingsStore object. An application

can use this class to store and retrieve configuration information. This class uses the Microsoft Reactive Extensions
library to expose any changes made to the configuration through an implementation of the IObservable interface.
If a setting is modified by calling the SetAppSetting method, the Changed event is raised and all subscribers to this
event will be notified.
Note that all settings are also cached in a Dictionary object inside the ExternalConfigurationManager class for fast
access. The GetSetting method used to retrieve a configuration setting reads the data from the cache. If the
setting isn't found in the cache, it's retrieved from the BlobSettingsStore object instead.
The GetSettings method invokes the CheckForConfigurationChanges method to detect whether the configuration
information in blob storage has changed. It does this by examining the version number and comparing it with the
current version number held by the ExternalConfigurationManager object. If one or more changes have occurred,
the Changed event is raised and the configuration settings cached in the Dictionary object are refreshed. This is
an application of the Cache-Aside pattern.
The following code sample shows how the Changed event, the GetSettings method, and the
CheckForConfigurationChanges method are implemented:

public class ExternalConfigurationManager : IDisposable

{
// An abstraction of the configuration store.
private readonly ISettingsStore settings;
private readonly ISubject<KeyValuePair<string, string>> changed;
...
private readonly ReaderWriterLockSlim settingsCacheLock = new ReaderWriterLockSlim();
private readonly SemaphoreSlim syncCacheSemaphore = new SemaphoreSlim(1);
...
private Dictionary<string, string> settingsCache;
private ETag currentVersion;
...

public ExternalConfigurationManager(ISettingsStore settings, TimeSpan interval, string environment)

{
this.settings = settings;
this.interval = interval;
this.CheckForConfigurationChangesAsync().Wait();
this.changed = new Subject<KeyValuePair<string, string>>();
this.Environment = environment;
}

...
public IObservable<KeyValuePair<string, string>> Changed => this.changed.AsObservable();
...

public string GetAppSetting(string key)

{
...
// Try to get the value from the settings cache.
// If there's a cache miss, get the setting from the settings store and refresh the settings cache.

string value;
try
{
this.settingsCacheLock.EnterReadLock();

this.settingsCache.TryGetValue(key, out value);

}
finally
{
this.settingsCacheLock.ExitReadLock();
}

return value;
}
...
private void CheckForConfigurationChanges()
{
try
 {
// It is assumed that updates are infrequent.
// To avoid race conditions in refreshing the cache, synchronize access to the in-memory cache.
await this.syncCacheSemaphore.WaitAsync();

var latestVersion = await this.settings.GetVersionAsync();

// If the versions are the same, nothing has changed in the configuration.
if (this.currentVersion == latestVersion) return;

// Get the latest settings from the settings store and publish changes.
var latestSettings = await this.settings.FindAllAsync();

// Refresh the settings cache.

try
{
this.settingsCacheLock.EnterWriteLock();

if (this.settingsCache != null)
{
//Notify settings changed
latestSettings.Except(this.settingsCache).ToList().ForEach(kv => this.changed.OnNext(kv));
}
this.settingsCache = latestSettings;
}
finally
{
this.settingsCacheLock.ExitWriteLock();
}

// Update the current version.

this.currentVersion = latestVersion;
}
catch (Exception ex)
{
this.changed.OnError(ex);
}
finally
{
this.syncCacheSemaphore.Release();
}
}
}

The ExternalConfigurationManager class also provides a property named Environment . This property supports
varying configurations for an application running in different environments, such as staging and production.

An ExternalConfigurationManager object can also query the BlobSettingsStore object periodically for any changes.
In the following code, the StartMonitor method calls CheckForConfigurationChanges at an interval to detect any
changes and raise the Changed event, as described earlier.

public class ExternalConfigurationManager : IDisposable

{
...
private readonly ISubject<KeyValuePair<string, string>> changed;
private Dictionary<string, string> settingsCache;
private readonly CancellationTokenSource cts = new CancellationTokenSource();
private Task monitoringTask;
private readonly TimeSpan interval;

private readonly SemaphoreSlim timerSemaphore = new SemaphoreSlim(1);

...
public ExternalConfigurationManager(string environment) : this(new BlobSettingsStore(environment),
TimeSpan.FromSeconds(15), environment)
{
{
}

public ExternalConfigurationManager(ISettingsStore settings, TimeSpan interval, string environment)

{
this.settings = settings;
this.interval = interval;
this.CheckForConfigurationChangesAsync().Wait();
this.changed = new Subject<KeyValuePair<string, string>>();
this.Environment = environment;
}
...
/// <summary>
/// Check to see if the current instance is monitoring for changes
/// </summary>
public bool IsMonitoring => this.monitoringTask != null && !this.monitoringTask.IsCompleted;

/// <summary>
/// Start the background monitoring for configuration changes in the central store
/// </summary>
public void StartMonitor()
{
if (this.IsMonitoring)
return;

try
{
this.timerSemaphore.Wait();

// Check again to make sure we are not already running.

if (this.IsMonitoring)
return;

// Start running our task loop.

this.monitoringTask = ConfigChangeMonitor();
}
finally
{
this.timerSemaphore.Release();
}
}

/// <summary>
/// Loop that monitors for configuration changes
/// </summary>
/// <returns></returns>
public async Task ConfigChangeMonitor()
{
while (!cts.Token.IsCancellationRequested)
{
await this.CheckForConfigurationChangesAsync();
await Task.Delay(this.interval, cts.Token);
}
}

/// <summary>
/// Stop monitoring for configuration changes
/// </summary>
public void StopMonitor()
{
try
{
this.timerSemaphore.Wait();

// Signal the task to stop.

this.cts.Cancel();

// Wait for the loop to stop.

this.monitoringTask.Wait();
 this.monitoringTask = null;
}
finally
{
this.timerSemaphore.Release();
}
}

public void Dispose()

{
this.cts.Cancel();
}
...
}

The ExternalConfigurationManager class is instantiated as a singleton instance by the ExternalConfiguration class

shown below.

public static class ExternalConfiguration

{
private static readonly Lazy<ExternalConfigurationManager> configuredInstance = new
Lazy<ExternalConfigurationManager>(
() =>
{
var environment = CloudConfigurationManager.GetSetting("environment");
return new ExternalConfigurationManager(environment);
});

public static ExternalConfigurationManager Instance => configuredInstance.Value;

}

The following code is taken from the WorkerRole class in the ExternalConfigurationStore.Cloud project. It shows
how the application uses the ExternalConfiguration class to read a setting.

public override void Run()

{
// Start monitoring configuration changes.
ExternalConfiguration.Instance.StartMonitor();

// Get a setting.
var setting = ExternalConfiguration.Instance.GetAppSetting("setting1");
Trace.TraceInformation("Worker Role: Get setting1, value: " + setting);

this.completeEvent.WaitOne();
}

The following code, also from the WorkerRole class, shows how the application subscribes to configuration events.

public override bool OnStart()

{
...
// Subscribe to the event.
ExternalConfiguration.Instance.Changed.Subscribe(
m => Trace.TraceInformation("Configuration has changed. Key:{0} Value:{1}",
m.Key, m.Value),
ex => Trace.TraceError("Error detected: " + ex.Message));
...
}
Related patterns and guidance
A sample that demonstrates this pattern is available on GitHub.
 Federated Identity pattern
12/18/2020 • 7 minutes to read • Edit Online

Delegate authentication to an external identity provider. This can simplify development, minimize the requirement
for user administration, and improve the user experience of the application.

Context and problem

Users typically need to work with multiple applications provided and hosted by different organizations they have a
business relationship with. These users might be required to use specific (and different) credentials for each one.
This can:
Cause a disjointed user experience . Users often forget sign-in credentials when they have many
different ones.
Expose security vulnerabilities . When a user leaves the company the account must immediately be
deprovisioned. It's easy to overlook this in large organizations.
Complicate user management . Administrators must manage credentials for all of the users, and perform
additional tasks such as providing password reminders.
Users typically prefer to use the same credentials for all these applications.

Solution
Implement an authentication mechanism that can use federated identity. Separate user authentication from the
application code, and delegate authentication to a trusted identity provider. This can simplify development and
allow users to authenticate using a wider range of identity providers (IdP) while minimizing the administrative
overhead. It also allows you to clearly decouple authentication from authorization.
The trusted identity providers include corporate directories, on-premises federation services, other security token
services (STS) provided by business partners, or social identity providers that can authenticate users who have, for
example, a Microsoft, Google, Yahoo!, or Facebook account.
The figure illustrates the Federated Identity pattern when a client application needs to access a service that requires
authentication. The authentication is performed by an IdP that works in concert with an STS. The IdP issues security
tokens that provide information about the authenticated user. This information, referred to as claims, includes the
user’s identity, and might also include other information such as role membership and more granular access
rights.
This model is often called claims-based access control. Applications and services authorize access to features and
functionality based on the claims contained in the token. The service that requires authentication must trust the IdP.
The client application contacts the IdP that performs the authentication. If the authentication is successful, the IdP
returns a token containing the claims that identify the user to the STS (note that the IdP and STS can be the same
service). The STS can transform and augment the claims in the token based on predefined rules, before returning it
to the client. The client application can then pass this token to the service as proof of its identity.

There might be additional security token services in the chain of trust. For example, in the scenario described
later, an on-premises STS trusts another STS that is responsible for accessing an identity provider to
authenticate the user. This approach is common in enterprise scenarios where there's an on-premises STS and
directory.

Federated authentication provides a standards-based solution to the issue of trusting identities across diverse
domains, and can support single sign-on. It's becoming more common across all types of applications, especially
cloud-hosted applications, because it supports single sign-on without requiring a direct network connection to
identity providers. The user doesn't have to enter credentials for every application. This increases security because
it prevents the creation of credentials required to access many different applications, and it also hides the user’s
credentials from all but the original identity provider. Applications see just the authenticated identity information
contained within the token.
Federated identity also has the major advantage that management of the identity and credentials is the
responsibility of the identity provider. The application or service doesn't need to provide identity management
features. In addition, in corporate scenarios, the corporate directory doesn't need to know about the user if it trusts
the identity provider. This removes all the administrative overhead of managing the user identity within the
directory.

Issues and considerations

Consider the following when designing applications that implement federated authentication:
Authentication can be a single point of failure. If you deploy your application to multiple datacenters,
consider deploying your identity management mechanism to the same datacenters to maintain application
reliability and availability.
Authentication tools make it possible to configure access control based on role claims contained in the
authentication token. This is often referred to as role-based access control (RBAC), and it can allow a more
granular level of control over access to features and resources.
 Unlike a corporate directory, claims-based authentication using social identity providers doesn't usually
provide information about the authenticated user other than an email address, and perhaps a name. Some
social identity providers, such as a Microsoft account, provide only a unique identifier. The application
usually needs to maintain some information on registered users, and be able to match this information to
the identifier contained in the claims in the token. Typically this is done through registration when the user
first accesses the application, and information is then injected into the token as additional claims after each
authentication.
If there's more than one identity provider configured for the STS, it must detect which identity provider the
user should be redirected to for authentication. This process is called home realm discovery. The STS might
be able to do this automatically based on an email address or user name that the user provides, a
subdomain of the application that the user is accessing, the user’s IP address scope, or on the contents of a
cookie stored in the user’s browser. For example, if the user entered an email address in the Microsoft
domain, such as user@live.com, the STS will redirect the user to the Microsoft account sign-in page. On later
visits, the STS could use a cookie to indicate that the last sign in was with a Microsoft account. If automatic
discovery can't determine the home realm, the STS will display a home realm discovery page that lists the
trusted identity providers, and the user must select the one they want to use.

When to use this pattern

This pattern is useful for scenarios such as:
Single sign-on in the enterprise . In this scenario you need to authenticate employees for corporate
applications that are hosted in the cloud outside the corporate security boundary, without requiring them to
sign in every time they visit an application. The user experience is the same as when using on-premises
applications where they're authenticated when signing in to a corporate network, and from then on have
access to all relevant applications without needing to sign in again.
Federated identity with multiple par tners . In this scenario you need to authenticate both corporate
employees and business partners who don't have accounts in the corporate directory. This is common in
business-to-business applications, applications that integrate with third-party services, and where
companies with different IT systems have merged or shared resources.
Federated identity in SaaS applications . In this scenario independent software vendors provide a
ready-to-use service for multiple clients or tenants. Each tenant authenticates using a suitable identity
provider. For example, business users will use their corporate credentials, while consumers and clients of the
tenant will use their social identity credentials.
This pattern might not be useful in the following situations:
All users of the application can be authenticated by one identity provider, and there's no requirement to
authenticate using any other identity provider. This is typical in business applications that use a corporate
directory (accessible within the application) for authentication, by using a VPN, or (in a cloud-hosted
scenario) through a virtual network connection between the on-premises directory and the application.
The application was originally built using a different authentication mechanism, perhaps with custom user
stores, or doesn't have the capability to handle the negotiation standards used by claims-based
technologies. Retrofitting claims-based authentication and access control into existing applications can be
complex, and probably not cost effective.

Example
An organization hosts a multi-tenant software as a service (SaaS) application in Microsoft Azure. The application
includes a website that tenants can use to manage the application for their own users. The application allows
tenants to access the website by using a federated identity that is generated by Active Directory Federation
Services (AD FS) when a user is authenticated by that organization’s own Active Directory.

The figure shows how tenants authenticate with their own identity provider (step 1), in this case AD FS. After
successfully authenticating a tenant, AD FS issues a token. The client browser forwards this token to the SaaS
application’s federation provider, which trusts tokens issued by the tenant’s AD FS, in order to get back a token that
is valid for the SaaS federation provider (step 2). If necessary, the SaaS federation provider performs a
transformation on the claims in the token into claims that the application recognizes (step 3) before returning the
new token to the client browser. The application trusts tokens issued by the SaaS federation provider and uses the
claims in the token to apply authorization rules (step 4).
Tenants won't need to remember separate credentials to access the application, and an administrator at the
tenant’s company can configure in its own AD FS the list of users that can access the application.

Related guidance
Microsoft Azure Active Directory
Active Directory Domain Services
Active Directory Federation Services
Identity management for multitenant applications in Microsoft Azure
Multitenant Applications in Azure
 Gatekeeper pattern
12/18/2020 • 4 minutes to read • Edit Online

Protect applications and services by using a dedicated host instance that acts as a broker between clients and the
application or service, validates and sanitizes requests, and passes requests and data between them. This can
provide an additional layer of security, and limit the attack surface of the system.

Context and problem

Applications expose their functionality to clients by accepting and processing requests. In cloud-hosted scenarios,
applications expose endpoints clients connect to, and typically include the code to handle the requests from clients.
This code performs authentication and validation, some or all request processing, and is likely to accesses storage
and other services on behalf of the client.
If a malicious user is able to compromise the system and gain access to the application’s hosting environment, the
security mechanisms it uses such as credentials and storage keys, and the services and data it accesses, are
exposed. As a result, the malicious user can gain unrestrained access to sensitive information and other services.

Solution
To minimize the risk of clients gaining access to sensitive information and services, decouple hosts or tasks that
expose public endpoints from the code that processes requests and accesses storage. You can achieve this by using
a façade or a dedicated task that interacts with clients and then hands off the request—perhaps through a
decoupled interface—to the hosts or tasks that'll handle the request. The figure provides a high-level overview of
this pattern.

The gatekeeper pattern can be used to simply protect storage, or it can be used as a more comprehensive façade
to protect all of the functions of the application. The important factors are:
Controlled validation . The gatekeeper validates all requests, and rejects those that don't meet validation
requirements.
Limited risk and exposure . The gatekeeper doesn't have access to the credentials or keys used by the trusted
host to access storage and services. If the gatekeeper is compromised, the attacker doesn't get access to these
credentials or keys.
Appropriate security . The gatekeeper runs in a limited privilege mode, while the rest of the application runs
in the full trust mode required to access storage and services. If the gatekeeper is compromised, it can't directly
access the application services or data.
This pattern acts like a firewall in a typical network topography. It allows the gatekeeper to examine requests and
make a decision about whether to pass the request on to the trusted host that performs the required tasks. This
decision typically requires the gatekeeper to validate and sanitize the request content before passing it on to the
trusted host.

Issues and considerations

Consider the following points when deciding how to implement this pattern:
Ensure that the trusted hosts the gatekeeper passes requests to expose only internal or protected endpoints,
and connect only to the gatekeeper. The trusted hosts shouldn't expose any external endpoints or interfaces.
The gatekeeper must run in a limited privilege mode. Typically this means running the gatekeeper and the
trusted host in separate hosted services or virtual machines.
The gatekeeper shouldn't perform any processing related to the application or services, or access any data. Its
function is purely to validate and sanitize requests. The trusted hosts might need to perform additional
validation of requests, but the core validation should be performed by the gatekeeper.
Use a secure communication channel (HTTPS, SSL, or TLS) between the gatekeeper and the trusted hosts or
tasks where this is possible. However, some hosting environments don't support HTTPS on internal endpoints.
Adding the extra layer to the application to implement the gatekeeper pattern is likely to have some impact on
performance due to the additional processing and network communication it requires.
The gatekeeper instance could be a single point of failure. To minimize the impact of a failure, consider
deploying additional instances and using an autoscaling mechanism to ensure capacity to maintain availability.

When to use this pattern

This pattern is useful for:
Applications that handle sensitive information, expose services that must have a high degree of protection from
malicious attacks, or perform mission-critical operations that shouldn't be disrupted.
Distributed applications where it's necessary to perform request validation separately from the main tasks, or
to centralize this validation to simplify maintenance and administration.

Example
In a cloud-hosted scenario, this pattern can be implemented by decoupling the gatekeeper role or virtual machine
from the trusted roles and services in an application. Do this by using an internal endpoint, a queue, or storage as
an intermediate communication mechanism. The figure illustrates using an internal endpoint.

Related patterns
The Valet Key pattern might also be relevant when implementing the Gatekeeper pattern. When communicating
between the Gatekeeper and trusted roles it's good practice to enhance security by using keys or tokens that limit
permissions for accessing resources. Describes how to use a token or key that provides clients with restricted
direct access to a specific resource or service.
 Gateway Aggregation pattern
12/18/2020 • 3 minutes to read • Edit Online

Use a gateway to aggregate multiple individual requests into a single request. This pattern is useful when a client
must make multiple calls to different backend systems to perform an operation.

Context and problem

To perform a single task, a client may have to make multiple calls to various backend services. An application that
relies on many services to perform a task must expend resources on each request. When any new feature or
service is added to the application, additional requests are needed, further increasing resource requirements and
network calls. This chattiness between a client and a backend can adversely impact the performance and scale of
the application. Microservice architectures have made this problem more common, as applications built around
many smaller services naturally have a higher amount of cross-service calls.
In the following diagram, the client sends requests to each service (1,2,3). Each service processes the request and
sends the response back to the application (4,5,6). Over a cellular network with typically high latency, using
individual requests in this manner is inefficient and could result in broken connectivity or incomplete requests.
While each request may be done in parallel, the application must send, wait, and process data for each request, all
on separate connections, increasing the chance of failure.

Solution
Use a gateway to reduce chattiness between the client and the services. The gateway receives client requests,
dispatches requests to the various backend systems, and then aggregates the results and sends them back to the
requesting client.
This pattern can reduce the number of requests that the application makes to backend services, and improve
application performance over high-latency networks.
In the following diagram, the application sends a request to the gateway (1). The request contains a package of
additional requests. The gateway decomposes these and processes each request by sending it to the relevant
service (2). Each service returns a response to the gateway (3). The gateway combines the responses from each
service and sends the response to the application (4). The application makes a single request and receives only a
single response from the gateway.
Issues and considerations
The gateway should not introduce service coupling across the backend services.
The gateway should be located near the backend services to reduce latency as much as possible.
The gateway service may introduce a single point of failure. Ensure the gateway is properly designed to meet
your application's availability requirements.
The gateway may introduce a bottleneck. Ensure the gateway has adequate performance to handle load and
can be scaled to meet your anticipated growth.
Perform load testing against the gateway to ensure you don't introduce cascading failures for services.
Implement a resilient design, using techniques such as bulkheads, circuit breaking, retry, and timeouts.
If one or more service calls takes too long, it may be acceptable to timeout and return a partial set of data.
Consider how your application will handle this scenario.
Use asynchronous I/O to ensure that a delay at the backend doesn't cause performance issues in the
application.
Implement distributed tracing using correlation IDs to track each individual call.
Monitor request metrics and response sizes.
Consider returning cached data as a failover strategy to handle failures.
Instead of building aggregation into the gateway, consider placing an aggregation service behind the gateway.
Request aggregation will likely have different resource requirements than other services in the gateway and
may impact the gateway's routing and offloading functionality.

When to use this pattern

Use this pattern when:
A client needs to communicate with multiple backend services to perform an operation.
The client may use networks with significant latency, such as cellular networks.
This pattern may not be suitable when:
You want to reduce the number of calls between a client and a single service across multiple operations. In
that scenario, it may be better to add a batch operation to the service.
The client or application is located near the backend services and latency is not a significant factor.

Example
The following example illustrates how to create a simple a gateway aggregation NGINX service using Lua.

worker_processes 4;

events {
worker_connections 1024;
}

http {
server {
listen 80;

location = /batch {
content_by_lua '
ngx.req.read_body()

-- read json body content

local cjson = require "cjson"
local batch = cjson.decode(ngx.req.get_body_data())["batch"]

-- create capture_multi table

local requests = {}
for i, item in ipairs(batch) do
table.insert(requests, {item.relative_url, { method = ngx.HTTP_GET}})
end

-- execute batch requests in parallel

local results = {}
local resps = { ngx.location.capture_multi(requests) }
for i, res in ipairs(resps) do
table.insert(results, {status = res.status, body = cjson.decode(res.body), header = res.header})
end

ngx.say(cjson.encode({results = results}))
';
}

location = /service1 {
default_type application/json;
echo '{"attr1":"val1"}';
}

location = /service2 {
default_type application/json;
echo '{"attr2":"val2"}';
}
}
}

Related guidance
Backends for Frontends pattern
Gateway Offloading pattern
Gateway Routing pattern
 Gateway Offloading pattern
12/18/2020 • 3 minutes to read • Edit Online

Offload shared or specialized service functionality to a gateway proxy. This pattern can simplify application
development by moving shared service functionality, such as the use of SSL certificates, from other parts of the
application into the gateway.

Context and problem

Some features are commonly used across multiple services, and these features require configuration,
management, and maintenance. A shared or specialized service that is distributed with every application
deployment increases the administrative overhead and increases the likelihood of deployment error. Any updates
to a shared feature must be deployed across all services that share that feature.
Properly handling security issues (token validation, encryption, SSL certificate management) and other complex
tasks can require team members to have highly specialized skills. For example, a certificate needed by an
application must be configured and deployed on all application instances. With each new deployment, the
certificate must be managed to ensure that it does not expire. Any common certificate that is due to expire must
be updated, tested, and verified on every application deployment.
Other common services such as authentication, authorization, logging, monitoring, or throttling can be difficult
to implement and manage across a large number of deployments. It may be better to consolidate this type of
functionality, in order to reduce overhead and the chance of errors.

Solution
Offload some features into a gateway, particularly cross-cutting concerns such as certificate management,
authentication, SSL termination, monitoring, protocol translation, or throttling.
The following diagram shows a gateway that terminates inbound SSL connections. It requests data on behalf of
the original requestor from any HTTP server upstream of the gateway.

Benefits of this pattern include:

Simplify the development of services by removing the need to distribute and maintain supporting
resources, such as web server certificates and configuration for secure websites. Simpler configuration
results in easier management and scalability and makes service upgrades simpler.
Allow dedicated teams to implement features that require specialized expertise, such as security. This
allows your core team to focus on the application functionality, leaving these specialized but cross-cutting
concerns to the relevant experts.
Provide some consistency for request and response logging and monitoring. Even if a service is not
 correctly instrumented, the gateway can be configured to ensure a minimum level of monitoring and
logging.

Issues and considerations

Ensure the gateway is highly available and resilient to failure. Avoid single points of failure by running
multiple instances of your gateway.
Ensure the gateway is designed for the capacity and scaling requirements of your application and endpoints.
Make sure the gateway does not become a bottleneck for the application and is sufficiently scalable.
Only offload features that are used by the entire application, such as security or data transfer.
Business logic should never be offloaded to the gateway.
If you need to track transactions, consider generating correlation IDs for logging purposes.

When to use this pattern

Use this pattern when:
An application deployment has a shared concern such as SSL certificates or encryption.
A feature that is common across application deployments that may have different resource requirements,
such as memory resources, storage capacity or network connections.
You wish to move the responsibility for issues such as network security, throttling, or other network boundary
concerns to a more specialized team.
This pattern may not be suitable if it introduces coupling across services.

Example
Using Nginx as the SSL offload appliance, the following configuration terminates an inbound SSL connection and
distributes the connection to one of three upstream HTTP servers.

upstream iis {
server 10.3.0.10 max_fails=3 fail_timeout=15s;
server 10.3.0.20 max_fails=3 fail_timeout=15s;
server 10.3.0.30 max_fails=3 fail_timeout=15s;
}

server {
listen 443;
ssl on;
ssl_certificate /etc/nginx/ssl/domain.cer;
ssl_certificate_key /etc/nginx/ssl/domain.key;

location / {
set $targ iis;
proxy_pass http://$targ;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Proto https;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header Host $host;
}
}

On Azure, this can be achieved by setting up SSL termination on Application Gateway.

Related guidance
Backends for Frontends pattern
Gateway Aggregation pattern
Gateway Routing pattern
 Gateway Routing pattern
12/18/2020 • 3 minutes to read • Edit Online

Route requests to multiple services using a single endpoint. This pattern is useful when you wish to expose
multiple services on a single endpoint and route to the appropriate service based on the request.

Context and problem

When a client needs to consume multiple services, setting up a separate endpoint for each service and having
the client manage each endpoint can be challenging. For example, an e-commerce application might provide
services such as search, reviews, cart, checkout, and order history. Each service has a different API that the client
must interact with, and the client must know about each endpoint in order to connect to the services. If an API
changes, the client must be updated as well. If you refactor a service into two or more separate services, the code
must change in both the service and the client.

Solution
Place a gateway in front of a set of applications, services, or deployments. Use application Layer 7 routing to
route the request to the appropriate instances.
With this pattern, the client application only needs to know about and communicate with a single endpoint. If a
service is consolidated or decomposed, the client does not necessarily require updating. It can continue making
requests to the gateway, and only the routing changes.
A gateway also lets you abstract backend services from the clients, allowing you to keep client calls simple while
enabling changes in the backend services behind the gateway. Client calls can be routed to whatever service or
services need to handle the expected client behavior, allowing you to add, split, and reorganize services behind
the gateway without changing the client.
This pattern can also help with deployment, by allowing you to manage how updates are rolled out to users.
When a new version of your service is deployed, it can be deployed in parallel with the existing version. Routing
lets you control what version of the service is presented to the clients, giving you the flexibility to use various
release strategies, whether incremental, parallel, or complete rollouts of updates. Any issues discovered after the
new service is deployed can be quickly reverted by making a configuration change at the gateway, without
affecting clients.

Issues and considerations

The gateway service may introduce a single point of failure. Ensure it is properly designed to meet your
availability requirements. Consider resiliency and fault tolerance capabilities when implementing.
The gateway service may introduce a bottleneck. Ensure the gateway has adequate performance to handle
load and can easily scale in line with your growth expectations.
Perform load testing against the gateway to ensure you don't introduce cascading failures for services.
Gateway routing is level 7. It can be based on IP, port, header, or URL.

When to use this pattern

Use this pattern when:
A client needs to consume multiple services that can be accessed behind a gateway.
You wish to simplify client applications by using a single endpoint.
You need to route requests from externally addressable endpoints to internal virtual endpoints, such as
exposing ports on a VM to cluster virtual IP addresses.
This pattern may not be suitable when you have a simple application that uses only one or two services.

Example
Using Nginx as the router, the following is a simple example configuration file for a server that routes requests
for applications residing on different virtual directories to different machines at the back end.

server {
listen 80;
server_name domain.com;

location /app1 {
proxy_pass http://10.0.3.10:80;
}

location /app2 {
proxy_pass http://10.0.3.20:80;
}

location /app3 {
proxy_pass http://10.0.3.30:80;
}
}

On Azure, multiple services can be set up behind an Application Gateway instance, which provides layer-7
routing.

Related guidance
Backends for Frontends pattern
Gateway Aggregation pattern
Gateway Offloading pattern
 Geodes
12/18/2020 • 5 minutes to read • Edit Online

The geode pattern involves deploying a collection of backend services into a set of ge ographical node s, each of
which can service any request for any client in any region. This pattern allows serving requests in an active-active
style, improving latency and increasing availability by distributing request processing around the globe.

Context and problem

Many large-scale services have specific challenges around geo-availability and scale. Classic designs often bring
the data to the compute by storing data in a remote SQL server that serves as the compute tier for that data,
relying on scale-up for growth.
The classic approach may present a number of challenges:
Network latency issues for users coming from the other side of the globe to connect to the hosting endpoint
Traffic management for demand bursts that can overwhelm the services in a single region
Cost-prohibitive complexity of deploying copies of app infrastructure into multiple regions for a 24x7 service
Modern cloud infrastructure has evolved to enable geographic load balancing of front-end services, while
allowing for geographic replication of backend services. For availability and performance, getting data closer to
the user is good. When data is geo-distributed across a far-flung user base, the geo-distributed datastores should
also be colocated with the compute resources that process the data. The geode pattern brings the compute to the
data.

Solution
Deploy the service into a number of satellite deployments spread around the globe, each of which is called a
geode. The geode pattern harnesses key features of Azure to route traffic via the shortest path to a nearby geode,
which improves latency and performance. Each geode is behind a global load balancer, and uses a geo-replicated
read-write service like Azure Cosmos DB to host the data plane, ensuring cross-geode data consistency. Data
replication services ensure that data stores are identical across geodes, so all requests can be served from all
geodes.
The key difference between a deployment stamp and a geode is that geodes never exist in isolation. There should
always be more than one geode in a production platform.
Geodes have the following characteristics:
Consist of a collection of disparate types of resources, often defined in a template.
Have no dependencies outside of the geode footprint and are self-contained. No geode is dependent on
another to operate, and if one dies, the others continue to operate.
Are loosely coupled via an edge network and replication backplane. For example, you can use Azure Traffic
Manager or Azure Front Door for fronting the geodes, while Azure Cosmos DB can act as the replication
backplane. Geodes are not the same as clusters because they share a replication backplane, so the platform
takes care of quorum issues.
The geode pattern occurs in big data architectures that use commodity hardware to process data colocated on the
same machine, and MapReduce to consolidate results across machines. Another usage is near-edge compute,
which brings compute closer to the intelligent edge of the network to reduce response time.
Services can use this pattern over dozens or hundreds of geodes. Furthermore, the resiliency of the whole
solution increases with each added geode, since any geodes can take over if a regional outage takes one or more
geodes offline.
It's also possible to augment local availability techniques, such as availability zones or paired regions, with the
geode pattern for global availability. This increases complexity, but is useful if your architecture is underpinned by
a storage engine such as blob storage that can only replicate to a paired region. You can deploy geodes into an
intra-zone, zonal, or regional footprint, with a mind to regulatory or latency constraints on location.

Issues and considerations

Use the following techniques and technologies to implement this pattern:
Modern DevOps practices and tools to produce and rapidly deploy identical geodes across a large number of
regions or instances.
Autoscaling to scale out compute and database throughput instances within a geode. Each geode individually
scales out, within the common backplane constraints.
A front-end service like Azure Front Door that does dynamic content acceleration, split TCP, and Anycast
routing.
A replicating data store like Azure Cosmos DB to control data consistency.
Serverless technologies where possible, to reduce always-on deployment cost, especially when load is
frequently rebalanced around the globe. This strategy allows for many geodes to be deployed with minimal
additional investment. Serverless and consumption-based billing technologies reduce waste and cost from
 duplicate geo-distributed deployments.
Consider the following points when deciding how to implement this pattern:
Choose whether to process data locally in each region, or to distribute aggregations in a single geode and
replicate the result across the globe. The Azure Cosmos DB change feed processor offers this granular control
using its lease container concept, and the leasecollectionprefix in the corresponding Azure Functions binding.
Each approach has distinct advantages and drawbacks.
Geodes can work in tandem, using the Azure Cosmos DB change feed and a real-time communication
platform like SignalR. Geodes can communicate with remote users via other geodes in a mesh pattern, without
knowing or caring where the remote user is located.
This design pattern implicitly decouples everything, resulting in an ultra-highly distributed and decoupled
architecture. Decoupling is a good thing, but consider how to track different components of the same request
as they might execute asynchronously on different instances. Get a good monitoring strategy in place.

When to use this pattern

Use this pattern:
To implement a high-scale platform that has users distributed over a wide area.
For any service that requires extreme availability and resilience characteristics, because services based on the
geode pattern can survive the loss of multiple service regions at the same time.
This pattern might not be suitable for
Architectures that have constraints so that all geodes can't be equal for data storage. For example, there may
be data residency requirements, an application that needs to maintain temporary state for a particular session,
or a heavy weighting of requests towards a single region. In this case, consider using deployment stamps in
combination with a global routing plane that is aware of where a user’s data sits, such as the traffic routing
component described within the deployment stamps pattern.
Situations where there's no geographical distribution required. Instead, consider availability zones and paired
regions for clustering.
Situations where a legacy platform needs to be retrofitted. This pattern works for cloud-native development
only, and can be difficult to retrofit.
Simple architectures and requirements, where geo-redundancy and geo-distribution aren't required or
advantageous.

Examples
Windows Active Directory implements an early variant of this pattern. Multi-primary replication means all
updates and requests can in theory be served from all serviceable nodes, but FSMO roles mean that all geodes
aren't equal.
A QnA sample application on GitHub showcases this design pattern in practice.
 Health Endpoint Monitoring pattern
12/18/2020 • 13 minutes to read • Edit Online

Implement functional checks in an application that external tools can access through exposed endpoints at
regular intervals. This can help to verify that applications and services are performing correctly.

Context and problem

It's a good practice, and often a business requirement, to monitor web applications and back-end services, to
ensure they're available and performing correctly. However, it's more difficult to monitor services running in the
cloud than it is to monitor on-premises services. For example, you don't have full control of the hosting
environment, and the services typically depend on other services provided by platform vendors and others.
There are many factors that affect cloud-hosted applications such as network latency, the performance and
availability of the underlying compute and storage systems, and the network bandwidth between them. The
service can fail entirely or partially due to any of these factors. Therefore, you must verify at regular intervals
that the service is performing correctly to ensure the required level of availability, which might be part of your
service level agreement (SLA).

Solution
Implement health monitoring by sending requests to an endpoint on the application. The application should
perform the necessary checks, and return an indication of its status.
A health monitoring check typically combines two factors:
The checks (if any) performed by the application or service in response to the request to the health
verification endpoint.
Analysis of the results by the tool or framework that performs the health verification check.
The response code indicates the status of the application and, optionally, any components or services it uses.
The latency or response time check is performed by the monitoring tool or framework. The figure provides an
overview of the pattern.
Other checks that might be carried out by the health monitoring code in the application include:
Checking cloud storage or a database for availability and response time.
Checking other resources or services located in the application, or located elsewhere but used by the
application.
Services and tools are available that monitor web applications by submitting a request to a configurable set of
endpoints, and evaluating the results against a set of configurable rules. It's relatively easy to create a service
endpoint whose sole purpose is to perform some functional tests on the system.
Typical checks that can be performed by the monitoring tools include:
Validating the response code. For example, an HTTP response of 200 (OK) indicates that the application
responded without error. The monitoring system might also check for other response codes to give more
comprehensive results.
Checking the content of the response to detect errors, even when a 200 (OK) status code is returned. This
can detect errors that affect only a section of the returned web page or service response. For example,
checking the title of a page or looking for a specific phrase that indicates the correct page was returned.
Measuring the response time, which indicates a combination of the network latency and the time that the
application took to execute the request. An increasing value can indicate an emerging problem with the
application or network.
Checking resources or services located outside the application, such as a content delivery network used by
the application to deliver content from global caches.
Checking for expiration of SSL certificates.
Measuring the response time of a DNS lookup for the URL of the application to measure DNS latency and
DNS failures.
Validating the URL returned by the DNS lookup to ensure correct entries. This can help to avoid malicious
request redirection through a successful attack on the DNS server.
It's also useful, where possible, to run these checks from different on-premises or hosted locations to measure
and compare response times. Ideally you should monitor applications from locations that are close to
customers to get an accurate view of the performance from each location. In addition to providing a more
robust checking mechanism, the results can help you decide on the deployment location for the application—
and whether to deploy it in more than one datacenter.
Tests should also be run against all the service instances that customers use to ensure the application is working
correctly for all customers. For example, if customer storage is spread across more than one storage account,
the monitoring process should check all of these.

Issues and considerations

Consider the following points when deciding how to implement this pattern:
How to validate the response. For example, is just a single 200 (OK) status code sufficient to verify the
application is working correctly? While this provides the most basic measure of application availability, and is
the minimum implementation of this pattern, it provides little information about the operations, trends, and
possible upcoming issues in the application.

Make sure that the application correctly returns a 200 (OK) only when the target resource is found and
processed. In some scenarios, such as when using a master page to host the target web page, the server
sends back a 200 (OK) status code instead of a 404 (Not Found) code, even when the target content page
was not found.

The number of endpoints to expose for an application. One approach is to expose at least one endpoint for the
core services that the application uses and another for lower priority services, allowing different levels of
importance to be assigned to each monitoring result. Also consider exposing more endpoints, such as one for
each core service, for additional monitoring granularity. For example, a health verification check might check the
database, storage, and an external geocoding service that an application uses, with each requiring a different
level of uptime and response time. The application could still be healthy if the geocoding service, or some other
background task, is unavailable for a few minutes.
Whether to use the same endpoint for monitoring as is used for general access, but to a specific path designed
for health verification checks, for example, /HealthCheck/{GUID}/ on the general access endpoint. This allows
some functional tests in the application to be run by the monitoring tools, such as adding a new user
registration, signing in, and placing a test order, while also verifying that the general access endpoint is
available.
The type of information to collect in the service in response to monitoring requests, and how to return this
information. Most existing tools and frameworks look only at the HTTP status code that the endpoint returns. To
return and validate additional information, you might have to create a custom monitoring utility or service.
How much information to collect. Performing excessive processing during the check can overload the
application and impact other users. The time it takes might exceed the timeout of the monitoring system so it
marks the application as unavailable. Most applications include instrumentation such as error handlers and
performance counters that log performance and detailed error information, this might be sufficient instead of
returning additional information from a health verification check.
Caching the endpoint status. It could be expensive to run the health check too frequently. If the health status is
reported through a dashboard, for example, you don't want every request from the dashboard to trigger a
health check. Instead, periodically check the system health and cache the status. Expose an endpoint that returns
the cached status.
How to configure security for the monitoring endpoints to protect them from public access, which might expose
the application to malicious attacks, risk the exposure of sensitive information, or attract denial of service (DoS)
attacks. Typically this should be done in the application configuration so that it can be updated easily without
restarting the application. Consider using one or more of the following techniques:
Secure the endpoint by requiring authentication. You can do this by using an authentication security key
in the request header or by passing credentials with the request, provided that the monitoring service or
tool supports authentication.
Use an obscure or hidden endpoint. For example, expose the endpoint on a different IP address to
 that used by the default application URL, configure the endpoint on a nonstandard HTTP port,
and/or use a complex path to the test page. You can usually specify additional endpoint addresses
and ports in the application configuration, and add entries for these endpoints to the DNS server
if required to avoid having to specify the IP address directly.
Expose a method on an endpoint that accepts a parameter such as a key value or an operation
mode value. Depending on the value supplied for this parameter, when a request is received the
code can perform a specific test or set of tests, or return a 404 (Not Found) error if the parameter
value isn't recognized. The recognized parameter values could be set in the application
configuration.

DoS attacks are likely to have less impact on a separate endpoint that performs basic
functional tests without compromising the operation of the application. Ideally, avoid using a
test that might expose sensitive information. If you must return information that might be
useful to an attacker, consider how you'll protect the endpoint and the data from unauthorized
access. In this case just relying on obscurity isn't enough. You should also consider using an
HTTPS connection and encrypting any sensitive data, although this will increase the load on
the server.

How to access an endpoint that's secured using authentication. Not all tools and frameworks can be
configured to include credentials with the health verification request. For example, Microsoft Azure built-
in health verification features can't provide authentication credentials. Some third-party alternatives are
Pingdom, Panopta, NewRelic, and Statuscake.
How to ensure that the monitoring agent is performing correctly. One approach is to expose an endpoint
that simply returns a value from the application configuration or a random value that can be used to test
the agent.

Also ensure that the monitoring system performs checks on itself, such as a self-test and built-in test,
to avoid it issuing false positive results.

When to use this pattern

This pattern is useful for:
Monitoring websites and web applications to verify availability.
Monitoring websites and web applications to check for correct operation.
Monitoring middle-tier or shared services to detect and isolate a failure that could disrupt other applications.
Complementing existing instrumentation in the application, such as performance counters and error
handlers. Health verification checking doesn't replace the requirement for logging and auditing in the
application. Instrumentation can provide valuable information for an existing framework that monitors
counters and error logs to detect failures or other issues. However, it can't provide information if the
application is unavailable.

Example
The following code examples, taken from the HealthCheckController class (a sample that demonstrates this
pattern is available on GitHub), demonstrates exposing an endpoint for performing a range of health checks.
The CoreServices method, shown below in C#, performs a series of checks on services used in the application.
If all of the tests run without error, the method returns a 200 (OK) status code. If any of the tests raises an
exception, the method returns a 500 (Internal Error) status code. The method could optionally return additional
information when an error occurs, if the monitoring tool or framework is able to use it.
 public ActionResult CoreServices()
{
try
{
// Run a simple check to ensure the database is available.
DataStore.Instance.CoreHealthCheck();

// Run a simple check on our external service.

MyExternalService.Instance.CoreHealthCheck();
}
catch (Exception ex)
{
Trace.TraceError("Exception in basic health check: {0}", ex.Message);

// This can optionally return different status codes based on the exception.
// Optionally it could return more details about the exception.
// The additional information could be used by administrators who access the
// endpoint with a browser, or using a ping utility that can display the
// additional information.
return new HttpStatusCodeResult((int)HttpStatusCode.InternalServerError);
}
return new HttpStatusCodeResult((int)HttpStatusCode.OK);
}

The ObscurePath method shows how you can read a path from the application configuration and use it as the
endpoint for tests. This example, in C#, also shows how you can accept an ID as a parameter and use it to check
for valid requests.

public ActionResult ObscurePath(string id)

{
// The id could be used as a simple way to obscure or hide the endpoint.
// The id to match could be retrieved from configuration and, if matched,
// perform a specific set of tests and return the result. If not matched it
// could return a 404 (Not Found) status.

// The obscure path can be set through configuration to hide the endpoint.
var hiddenPathKey = CloudConfigurationManager.GetSetting("Test.ObscurePath");

// If the value passed does not match that in configuration, return 404 (Not Found).
if (!string.Equals(id, hiddenPathKey))
{
return new HttpStatusCodeResult((int)HttpStatusCode.NotFound);
}

// Else continue and run the tests...

// Return results from the core services test.
return this.CoreServices();
}

The TestResponseFromConfig method shows how you can expose an endpoint that performs a check for a
specified configuration setting value.
 public ActionResult TestResponseFromConfig()
{
// Health check that returns a response code set in configuration for testing.
var returnStatusCodeSetting = CloudConfigurationManager.GetSetting(
"Test.ReturnStatusCode");

int returnStatusCode;

if (!int.TryParse(returnStatusCodeSetting, out returnStatusCode))

{
returnStatusCode = (int)HttpStatusCode.OK;
}

return new HttpStatusCodeResult(returnStatusCode);

}

Monitoring endpoints in Azure hosted applications

Some options for monitoring endpoints in Azure applications are:
Use the built-in monitoring features of Azure.
Use a third-party service or a framework such as Microsoft System Center Operations Manager.
Create a custom utility or a service that runs on your own or on a hosted server.

Even though Azure provides a reasonably comprehensive set of monitoring options, you can use
additional services and tools to provide extra information. Azure Management Services provides a
built-in monitoring mechanism for alert rules. The alerts section of the management services page in
the Azure portal allows you to configure up to ten alert rules per subscription for your services.
These rules specify a condition and a threshold value for a service such as CPU load, or the number
of requests or errors per second, and the service can automatically send email notifications to
addresses you define in each rule.

The conditions you can monitor vary depending on the hosting mechanism you choose for your application
(such as Web Sites, Cloud Services, Virtual Machines, or Mobile Services), but all of these include the ability to
create an alert rule that uses a web endpoint you specify in the settings for your service. This endpoint should
respond in a timely way so that the alert system can detect that the application is operating correctly.

Read more information about creating alert notifications.

If you host your application in Azure Cloud Services web and worker roles or Virtual Machines, you can take
advantage of one of the built-in services in Azure called Traffic Manager. Traffic Manager is a routing and load-
balancing service that can distribute requests to specific instances of your Cloud Services hosted application
based on a range of rules and settings.
In addition to routing requests, Traffic Manager pings a URL, port, and relative path that you specify on a regular
basis to determine which instances of the application defined in its rules are active and are responding to
requests. If it detects a status code 200 (OK), it marks the application as available. Any other status code causes
Traffic Manager to mark the application as offline. You can view the status in the Traffic Manager console, and
configure the rule to reroute requests to other instances of the application that are responding.
However, Traffic Manager will only wait ten seconds to receive a response from the monitoring URL. Therefore,
you should ensure that your health verification code executes in this time, allowing for network latency for the
round trip from Traffic Manager to your application and back again.
 Read more information about using Traffic Manager to monitor your applications. Traffic Manager is also
discussed in Multiple Datacenter Deployment Guidance.

Related guidance
The following guidance can be useful when implementing this pattern:
Instrumentation and Telemetry Guidance. Checking the health of services and components is typically done
by probing, but it's also useful to have information in place to monitor application performance and detect
events that occur at runtime. This data can be transmitted back to monitoring tools as additional information
for health monitoring. Instrumentation and Telemetry Guidance explores gathering remote diagnostics
information that's collected by instrumentation in applications.
Receiving alert notifications.
This pattern includes a downloadable sample application.
 Index Table pattern
12/18/2020 • 9 minutes to read • Edit Online

Create indexes over the fields in data stores that are frequently referenced by queries. This pattern can improve
query performance by allowing applications to more quickly locate the data to retrieve from a data store.

Context and problem

Many data stores organize the data for a collection of entities using the primary key. An application can use this
key to locate and retrieve data. The figure shows an example of a data store holding customer information. The
primary key is the Customer ID. The figure shows customer information organized by the primary key (Customer
ID).

While the primary key is valuable for queries that fetch data based on the value of this key, an application might
not be able to use the primary key if it needs to retrieve data based on some other field. In the customers
example, an application can't use the Customer ID primary key to retrieve customers if it queries data solely by
referencing the value of some other attribute, such as the town in which the customer is located. To perform a
query such as this, the application might have to fetch and examine every customer record, which could be a slow
process.
Many relational database management systems support secondary indexes. A secondary index is a separate data
structure that's organized by one or more nonprimary (secondary) key fields, and it indicates where the data for
each indexed value is stored. The items in a secondary index are typically sorted by the value of the secondary
keys to enable fast lookup of data. These indexes are usually maintained automatically by the database
management system.
You can create as many secondary indexes as you need to support the different queries that your application
performs. For example, in a Customers table in a relational database where the Customer ID is the primary key,
it's beneficial to add a secondary index over the town field if the application frequently looks up customers by the
town where they reside.
However, although secondary indexes are common in relational systems, most NoSQL data stores used by cloud
applications don't provide an equivalent feature.

Solution
If the data store doesn't support secondary indexes, you can emulate them manually by creating your own index
tables. An index table organizes the data by a specified key. Three strategies are commonly used for structuring
an index table, depending on the number of secondary indexes that are required and the nature of the queries
that an application performs.
The first strategy is to duplicate the data in each index table but organize it by different keys (complete
denormalization). The next figure shows index tables that organize the same customer information by Town and
LastName.

This strategy is appropriate if the data is relatively static compared to the number of times it's queried using each
key. If the data is more dynamic, the processing overhead of maintaining each index table becomes too large for
this approach to be useful. Also, if the volume of data is very large, the amount of space required to store the
duplicate data is significant.
The second strategy is to create normalized index tables organized by different keys and reference the original
data by using the primary key rather than duplicating it, as shown in the following figure. The original data is
called a fact table.

This technique saves space and reduces the overhead of maintaining duplicate data. The disadvantage is that an
application has to perform two lookup operations to find data using a secondary key. It has to find the primary
key for the data in the index table, and then use the primary key to look up the data in the fact table.
The third strategy is to create partially normalized index tables organized by different keys that duplicate
frequently retrieved fields. Reference the fact table to access less frequently accessed fields. The next figure shows
how commonly accessed data is duplicated in each index table.
With this strategy, you can strike a balance between the first two approaches. The data for common queries can
be retrieved quickly by using a single lookup, while the space and maintenance overhead isn't as significant as
duplicating the entire data set.
If an application frequently queries data by specifying a combination of values (for example, “Find all customers
that live in Redmond and that have a last name of Smith”), you could implement the keys to the items in the index
table as a concatenation of the Town attribute and the LastName attribute. The next figure shows an index table
based on composite keys. The keys are sorted by Town, and then by LastName for records that have the same
value for Town.

Index tables can speed up query operations over sharded data, and are especially useful where the shard key is
hashed. The next figure shows an example where the shard key is a hash of the Customer ID. The index table can
organize data by the nonhashed value (Town and LastName), and provide the hashed shard key as the lookup
data. This can save the application from repeatedly calculating hash keys (an expensive operation) if it needs to
retrieve data that falls within a range, or it needs to fetch data in order of the nonhashed key. For example, a
query such as “Find all customers that live in Redmond” can be quickly resolved by locating the matching items in
the index table, where they're all stored in a contiguous block. Then, follow the references to the customer data
using the shard keys stored in the index table.
Issues and considerations
Consider the following points when deciding how to implement this pattern:
The overhead of maintaining secondary indexes can be significant. You must analyze and understand the
queries that your application uses. Only create index tables when they're likely to be used regularly. Don't
create speculative index tables to support queries that an application doesn't perform, or performs only
occasionally.
Duplicating data in an index table can add significant overhead in storage costs and the effort required to
maintain multiple copies of data.
Implementing an index table as a normalized structure that references the original data requires an
application to perform two lookup operations to find data. The first operation searches the index table to
retrieve the primary key, and the second uses the primary key to fetch the data.
If a system incorporates a number of index tables over very large data sets, it can be difficult to maintain
consistency between index tables and the original data. It might be possible to design the application
around the eventual consistency model. For example, to insert, update, or delete data, an application could
post a message to a queue and let a separate task perform the operation and maintain the index tables
that reference this data asynchronously. For more information about implementing eventual consistency,
see the Data Consistency Primer.

Microsoft Azure storage tables support transactional updates for changes made to data held in the
same partition (referred to as entity group transactions). If you can store the data for a fact table and
one or more index tables in the same partition, you can use this feature to help ensure consistency.

Index tables might themselves be partitioned or sharded.

When to use this pattern

Use this pattern to improve query performance when an application frequently needs to retrieve data by using a
key other than the primary (or shard) key.
This pattern might not be useful when:
Data is volatile. An index table can become out of date very quickly, making it ineffective or making the
overhead of maintaining the index table greater than any savings made by using it.
 A field selected as the secondary key for an index table is nondiscriminating and can only have a small set of
values (for example, gender).
The balance of the data values for a field selected as the secondary key for an index table are highly skewed.
For example, if 90% of the records contain the same value in a field, then creating and maintaining an index
table to look up data based on this field might create more overhead than scanning sequentially through the
data. However, if queries very frequently target values that lie in the remaining 10%, this index can be useful.
You should understand the queries that your application is performing, and how frequently they're performed.

Example
Azure storage tables provide a highly scalable key/value data store for applications running in the cloud.
Applications store and retrieve data values by specifying a key. The data values can contain multiple fields, but the
structure of a data item is opaque to table storage, which simply handles a data item as an array of bytes.
Azure storage tables also support sharding. The sharding key includes two elements, a partition key and a row
key. Items that have the same partition key are stored in the same partition (shard), and the items are stored in
row key order within a shard. Table storage is optimized for performing queries that fetch data falling within a
contiguous range of row key values within a partition. If you're building cloud applications that store information
in Azure tables, you should structure your data with this feature in mind.
For example, consider an application that stores information about movies. The application frequently queries
movies by genre (action, documentary, historical, comedy, drama, and so on). You could create an Azure table
with partitions for each genre by using the genre as the partition key, and specifying the movie name as the row
key, as shown in the next figure.

This approach is less effective if the application also needs to query movies by starring actor. In this case, you can
create a separate Azure table that acts as an index table. The partition key is the actor and the row key is the
movie name. The data for each actor will be stored in separate partitions. If a movie stars more than one actor, the
same movie will occur in multiple partitions.
You can duplicate the movie data in the values held by each partition by adopting the first approach described in
the Solution section above. However, it's likely that each movie will be replicated several times (once for each
actor), so it might be more efficient to partially denormalize the data to support the most common queries (such
as the names of the other actors) and enable an application to retrieve any remaining details by including the
partition key necessary to find the complete information in the genre partitions. This approach is described by
the third option in the Solution section. The next figure shows this approach.
Related patterns and guidance
The following patterns and guidance might also be relevant when implementing this pattern:
Data Consistency Primer. An index table must be maintained as the data that it indexes changes. In the cloud, it
might not be possible or appropriate to perform operations that update an index as part of the same
transaction that modifies the data. In that case, an eventually consistent approach is more suitable. Provides
information on the issues surrounding eventual consistency.
Sharding pattern. The Index Table pattern is frequently used in conjunction with data partitioned by using
shards. The Sharding pattern provides more information on how to divide a data store into a set of shards.
Materialized View pattern. Instead of indexing data to support queries that summarize data, it might be more
appropriate to create a materialized view of the data. Describes how to support efficient summary queries by
generating prepopulated views over data.
 Leader Election pattern
12/18/2020 • 11 minutes to read • Edit Online

Coordinate the actions performed by a collection of collaborating instances in a distributed application by

electing one instance as the leader that assumes responsibility for managing the others. This can help to ensure
that instances don't conflict with each other, cause contention for shared resources, or inadvertently interfere with
the work that other instances are performing.

Context and problem

A typical cloud application has many tasks acting in a coordinated manner. These tasks could all be instances
running the same code and requiring access to the same resources, or they might be working together in parallel
to perform the individual parts of a complex calculation.
The task instances might run separately for much of the time, but it might also be necessary to coordinate the
actions of each instance to ensure that they don’t conflict, cause contention for shared resources, or accidentally
interfere with the work that other task instances are performing.
For example:
In a cloud-based system that implements horizontal scaling, multiple instances of the same task could be
running at the same time with each instance serving a different user. If these instances write to a shared
resource, it's necessary to coordinate their actions to prevent each instance from overwriting the changes
made by the others.
If the tasks are performing individual elements of a complex calculation in parallel, the results need to be
aggregated when they all complete.
The task instances are all peers, so there isn't a natural leader that can act as the coordinator or aggregator.

Solution
A single task instance should be elected to act as the leader, and this instance should coordinate the actions of the
other subordinate task instances. If all of the task instances are running the same code, they are each capable of
acting as the leader. Therefore, the election process must be managed carefully to prevent two or more instances
taking over the leader role at the same time.
The system must provide a robust mechanism for selecting the leader. This method has to cope with events such
as network outages or process failures. In many solutions, the subordinate task instances monitor the leader
through some type of heartbeat method, or by polling. If the designated leader terminates unexpectedly, or a
network failure makes the leader unavailable to the subordinate task instances, it's necessary for them to elect a
new leader.
There are several strategies for electing a leader among a set of tasks in a distributed environment, including:
Selecting the task instance with the lowest-ranked instance or process ID.
Racing to acquire a shared, distributed mutex. The first task instance that acquires the mutex is the leader.
However, the system must ensure that, if the leader terminates or becomes disconnected from the rest of the
system, the mutex is released to allow another task instance to become the leader.
Implementing one of the common leader election algorithms such as the Bully Algorithm or the Ring
Algorithm. These algorithms assume that each candidate in the election has a unique ID, and that it can
communicate with the other candidates reliably.
Issues and considerations
Consider the following points when deciding how to implement this pattern:
The process of electing a leader should be resilient to transient and persistent failures.
It must be possible to detect when the leader has failed or has become otherwise unavailable (such as due to a
communications failure). How quickly detection is needed is system dependent. Some systems might be able
to function for a short time without a leader, during which a transient fault might be fixed. In other cases, it
might be necessary to detect leader failure immediately and trigger a new election.
In a system that implements horizontal autoscaling, the leader could be terminated if the system scales back
and shuts down some of the computing resources.
Using a shared, distributed mutex introduces a dependency on the external service that provides the mutex.
The service constitutes a single point of failure. If it becomes unavailable for any reason, the system won't be
able to elect a leader.
Using a single dedicated process as the leader is a straightforward approach. However, if the process fails
there could be a significant delay while it's restarted. The resulting latency can affect the performance and
response times of other processes if they're waiting for the leader to coordinate an operation.
Implementing one of the leader election algorithms manually provides the greatest flexibility for tuning and
optimizing the code.

When to use this pattern

Use this pattern when the tasks in a distributed application, such as a cloud-hosted solution, need careful
coordination and there's no natural leader.

Avoid making the leader a bottleneck in the system. The purpose of the leader is to coordinate the work of the
subordinate tasks, and it doesn't necessarily have to participate in this work itself—although it should be able
to do so if the task isn't elected as the leader.

This pattern might not be useful if:

There's a natural leader or dedicated process that can always act as the leader. For example, it might be
possible to implement a singleton process that coordinates the task instances. If this process fails or becomes
unhealthy, the system can shut it down and restart it.
The coordination between tasks can be achieved using a more lightweight method. For example, if several task
instances simply need coordinated access to a shared resource, a better solution is to use optimistic or
pessimistic locking to control access.
A third-party solution is more appropriate. For example, the Microsoft Azure HDInsight service (based on
Apache Hadoop) uses the services provided by Apache Zookeeper to coordinate the map and reduce tasks
that collect and summarize data.

Example
The DistributedMutex project in the LeaderElection solution (a sample that demonstrates this pattern is available
on GitHub) shows how to use a lease on an Azure Storage blob to provide a mechanism for implementing a
shared, distributed mutex. This mutex can be used to elect a leader among a group of role instances in an Azure
cloud service. The first role instance to acquire the lease is elected the leader, and remains the leader until it
releases the lease or isn't able to renew the lease. Other role instances can continue to monitor the blob lease in
case the leader is no longer available.

A blob lease is an exclusive write lock over a blob. A single blob can be the subject of only one lease at any
point in time. A role instance can request a lease over a specified blob, and it'll be granted the lease if no other
role instance holds a lease over the same blob. Otherwise the request will throw an exception.
 To avoid a faulted role instance retaining the lease indefinitely, specify a lifetime for the lease. When this
expires, the lease becomes available. However, while a role instance holds the lease it can request that the
lease is renewed, and it'll be granted the lease for a further period of time. The role instance can continually
repeat this process if it wants to retain the lease. For more information on how to lease a blob, see Lease Blob
(REST API).

The BlobDistributedMutex class in the C# example below contains the RunTaskWhenMutexAcquired method that
enables a role instance to attempt to acquire a lease over a specified blob. The details of the blob (the name,
container, and storage account) are passed to the constructor in a BlobSettings object when the
BlobDistributedMutex object is created (this object is a simple struct that is included in the sample code). The
constructor also accepts a Task that references the code that the role instance should run if it successfully
acquires the lease over the blob and is elected the leader. Note that the code that handles the low-level details of
acquiring the lease is implemented in a separate helper class named BlobLeaseManager .

public class BlobDistributedMutex

{
...
private readonly BlobSettings blobSettings;
private readonly Func<CancellationToken, Task> taskToRunWhenLeaseAcquired;
...

public BlobDistributedMutex(BlobSettings blobSettings,

Func<CancellationToken, Task> taskToRunWhenLeaseAcquired)
{
this.blobSettings = blobSettings;
this.taskToRunWhenLeaseAcquired = taskToRunWhenLeaseAcquired;
}

public async Task RunTaskWhenMutexAcquired(CancellationToken token)

{
var leaseManager = new BlobLeaseManager(blobSettings);
await this.RunTaskWhenBlobLeaseAcquired(leaseManager, token);
}
...

The RunTaskWhenMutexAcquired method in the code sample above invokes the RunTaskWhenBlobLeaseAcquired
method shown in the following code sample to actually acquire the lease. The RunTaskWhenBlobLeaseAcquired
method runs asynchronously. If the lease is successfully acquired, the role instance has been elected the leader.
The purpose of the taskToRunWhenLeaseAcquired delegate is to perform the work that coordinates the other role
instances. If the lease isn't acquired, another role instance has been elected as the leader and the current role
instance remains a subordinate. Note that the TryAcquireLeaseOrWait method is a helper method that uses the
BlobLeaseManager object to acquire the lease.
 private async Task RunTaskWhenBlobLeaseAcquired(
BlobLeaseManager leaseManager, CancellationToken token)
{
while (!token.IsCancellationRequested)
{
// Try to acquire the blob lease.
// Otherwise wait for a short time before trying again.
string leaseId = await this.TryAcquireLeaseOrWait(leaseManager, token);

if (!string.IsNullOrEmpty(leaseId))
{
// Create a new linked cancellation token source so that if either the
// original token is canceled or the lease can't be renewed, the
// leader task can be canceled.
using (var leaseCts =
CancellationTokenSource.CreateLinkedTokenSource(new[] { token }))
{
// Run the leader task.
var leaderTask = this.taskToRunWhenLeaseAcquired.Invoke(leaseCts.Token);
...
}
}
}
...
}

The task started by the leader also runs asynchronously. While this task is running, the
RunTaskWhenBlobLeaseAcquired method shown in the following code sample periodically attempts to renew the
lease. This helps to ensure that the role instance remains the leader. In the sample solution, the delay between
renewal requests is less than the time specified for the duration of the lease in order to prevent another role
instance from being elected the leader. If the renewal fails for any reason, the task is canceled.
If the lease fails to be renewed or the task is canceled (possibly as a result of the role instance shutting down), the
lease is released. At this point, this or another role instance might be elected as the leader. The code extract below
shows this part of the process.

private async Task RunTaskWhenBlobLeaseAcquired(

BlobLeaseManager leaseManager, CancellationToken token)
{
while (...)
{
...
if (...)
{
...
using (var leaseCts = ...)
{
...
// Keep renewing the lease in regular intervals.
// If the lease can't be renewed, then the task completes.
var renewLeaseTask =
this.KeepRenewingLease(leaseManager, leaseId, leaseCts.Token);

// When any task completes (either the leader task itself or when it
// couldn't renew the lease) then cancel the other task.
await CancelAllWhenAnyCompletes(leaderTask, renewLeaseTask, leaseCts);
}
}
}
}
...
}
The KeepRenewingLease method is another helper method that uses the BlobLeaseManager object to renew the
lease. The CancelAllWhenAnyCompletes method cancels the tasks specified as the first two parameters. The
following diagram illustrates using the BlobDistributedMutex class to elect a leader and run a task that
coordinates operations.

The following code example shows how to use the BlobDistributedMutex class in a worker role. This code
acquires a lease over a blob named MyLeaderCoordinatorTask in the lease's container in development storage, and
specifies that the code defined in the MyLeaderCoordinatorTask method should run if the role instance is elected
the leader.
 var settings = new BlobSettings(CloudStorageAccount.DevelopmentStorageAccount,
"leases", "MyLeaderCoordinatorTask");
var cts = new CancellationTokenSource();
var mutex = new BlobDistributedMutex(settings, MyLeaderCoordinatorTask);
mutex.RunTaskWhenMutexAcquired(this.cts.Token);
...

// Method that runs if the role instance is elected the leader

private static async Task MyLeaderCoordinatorTask(CancellationToken token)
{
...
}

Note the following points about the sample solution:

The blob is a potential single point of failure. If the blob service becomes unavailable, or is inaccessible, the
leader won't be able to renew the lease and no other role instance will be able to acquire the lease. In this case,
no role instance will be able to act as the leader. However, the blob service is designed to be resilient, so
complete failure of the blob service is considered to be extremely unlikely.
If the task being performed by the leader stalls, the leader might continue to renew the lease, preventing any
other role instance from acquiring the lease and taking over the leader role in order to coordinate tasks. In the
real world, the health of the leader should be checked at frequent intervals.
The election process is nondeterministic. You can't make any assumptions about which role instance will
acquire the blob lease and become the leader.
The blob used as the target of the blob lease shouldn't be used for any other purpose. If a role instance
attempts to store data in this blob, this data won't be accessible unless the role instance is the leader and holds
the blob lease.

Related patterns and guidance

The following guidance might also be relevant when implementing this pattern:
This pattern has a downloadable sample application.
Autoscaling Guidance. It's possible to start and stop instances of the task hosts as the load on the application
varies. Autoscaling can help to maintain throughput and performance during times of peak processing.
Compute Partitioning Guidance. This guidance describes how to allocate tasks to hosts in a cloud service in a
way that helps to minimize running costs while maintaining the scalability, performance, availability, and
security of the service.
The Task-based Asynchronous Pattern.
An example illustrating the Bully Algorithm.
An example illustrating the Ring Algorithm.
Apache Curator a client library for Apache ZooKeeper.
The article Lease Blob (REST API) on MSDN.
 Materialized View pattern
12/18/2020 • 7 minutes to read • Edit Online

Generate prepopulated views over the data in one or more data stores when the data isn't ideally formatted for
required query operations. This can help support efficient querying and data extraction, and improve application
performance.

Context and problem

When storing data, the priority for developers and data administrators is often focused on how the data is stored,
as opposed to how it's read. The chosen storage format is usually closely related to the format of the data,
requirements for managing data size and data integrity, and the kind of store in use. For example, when using
NoSQL document store, the data is often represented as a series of aggregates, each containing all of the
information for that entity.
However, this can have a negative effect on queries. When a query only needs a subset of the data from some
entities, such as a summary of orders for several customers without all of the order details, it must extract all of
the data for the relevant entities in order to obtain the required information.

Solution
To support efficient querying, a common solution is to generate, in advance, a view that materializes the data in a
format suited to the required results set. The Materialized View pattern describes generating prepopulated views
of data in environments where the source data isn't in a suitable format for querying, where generating a
suitable query is difficult, or where query performance is poor due to the nature of the data or the data store.
These materialized views, which only contain data required by a query, allow applications to quickly obtain the
information they need. In addition to joining tables or combining data entities, materialized views can include the
current values of calculated columns or data items, the results of combining values or executing transformations
on the data items, and values specified as part of the query. A materialized view can even be optimized for just a
single query.
A key point is that a materialized view and the data it contains is completely disposable because it can be entirely
rebuilt from the source data stores. A materialized view is never updated directly by an application, and so it's a
specialized cache.
When the source data for the view changes, the view must be updated to include the new information. You can
schedule this to happen automatically, or when the system detects a change to the original data. In some cases it
might be necessary to regenerate the view manually. The figure shows an example of how the Materialized View
pattern might be used.
Issues and considerations
Consider the following points when deciding how to implement this pattern:
How and when the view will be updated. Ideally it'll regenerate in response to an event indicating a change to the
source data, although this can lead to excessive overhead if the source data changes rapidly. Alternatively,
consider using a scheduled task, an external trigger, or a manual action to regenerate the view.
In some systems, such as when using the Event Sourcing pattern to maintain a store of only the events that
modified the data, materialized views are necessary. Prepopulating views by examining all events to determine
the current state might be the only way to obtain information from the event store. If you're not using Event
Sourcing, you need to consider whether a materialized view is helpful or not. Materialized views tend to be
specifically tailored to one, or a small number of queries. If many queries are used, materialized views can result
in unacceptable storage capacity requirements and storage cost.
Consider the impact on data consistency when generating the view, and when updating the view if this occurs on
a schedule. If the source data is changing at the point when the view is generated, the copy of the data in the
view won't be fully consistent with the original data.
Consider where you'll store the view. The view doesn't have to be located in the same store or partition as the
original data. It can be a subset from a few different partitions combined.
A view can be rebuilt if lost. Because of that, if the view is transient and is only used to improve query
performance by reflecting the current state of the data, or to improve scalability, it can be stored in a cache or in a
less reliable location.
When defining a materialized view, maximize its value by adding data items or columns to it based on
computation or transformation of existing data items, on values passed in the query, or on combinations of these
values when appropriate.
Where the storage mechanism supports it, consider indexing the materialized view to further increase
performance. Most relational databases support indexing for views, as do big data solutions based on Apache
Hadoop.

When to use this pattern

This pattern is useful when:
Creating materialized views over data that's difficult to query directly, or where queries must be very complex
to extract data that's stored in a normalized, semi-structured, or unstructured way.
Creating temporary views that can dramatically improve query performance, or can act directly as source
 views or data transfer objects for the UI, for reporting, or for display.
Supporting occasionally connected or disconnected scenarios where connection to the data store isn't always
available. The view can be cached locally in this case.
Simplifying queries and exposing data for experimentation in a way that doesn't require knowledge of the
source data format. For example, by joining different tables in one or more databases, or one or more
domains in NoSQL stores, and then formatting the data to fit its eventual use.
Providing access to specific subsets of the source data that, for security or privacy reasons, shouldn't be
generally accessible, open to modification, or fully exposed to users.
Bridging different data stores, to take advantage of their individual capabilities. For example, using a cloud
store that's efficient for writing as the reference data store, and a relational database that offers good query
and read performance to hold the materialized views.
This pattern isn't useful in the following situations:
The source data is simple and easy to query.
The source data changes very quickly, or can be accessed without using a view. In these cases, you should
avoid the processing overhead of creating views.
Consistency is a high priority. The views might not always be fully consistent with the original data.
When using microservices. Microservices typically have well defined boundries aligning to domain driven
design (DDD).

Example
The following figure shows an example of using the Materialized View pattern to generate a summary of sales.
Data in the Order, OrderItem, and Customer tables in separate partitions in an Azure storage account are
combined to generate a view containing the total sales value for each product in the Electronics category, along
with a count of the number of customers who made purchases of each item.

Creating this materialized view requires complex queries. However, by exposing the query result as a
materialized view, users can easily obtain the results and use them directly or incorporate them in another query.
The view is likely to be used in a reporting system or dashboard, and can be updated on a scheduled basis such
as weekly.

Although this example uses Azure table storage, many relational database management systems also
provide native support for materialized views.

Related patterns and guidance

The following patterns and guidance might also be relevant when implementing this pattern:
Data Consistency Primer. The summary information in a materialized view has to be maintained so that it
reflects the underlying data values. As the data values change, it might not be practical to update the
summary data in real time, and instead you'll have to adopt an eventually consistent approach. Summarizes
the issues surrounding maintaining consistency over distributed data, and describes the benefits and
tradeoffs of different consistency models.
Command and Query Responsibility Segregation (CQRS) pattern. Use to update the information in a
materialized view by responding to events that occur when the underlying data values change.
Event Sourcing pattern. Use in conjunction with the CQRS pattern to maintain the information in a
materialized view. When the data values a materialized view is based on are changed, the system can raise
events that describe these changes and save them in an event store.
Index Table pattern. The data in a materialized view is typically organized by a primary key, but queries might
need to retrieve information from this view by examining data in other fields. Use to create secondary indexes
over data sets for data stores that don't support native secondary indexes.
 Pipes and Filters pattern
12/18/2020 • 12 minutes to read • Edit Online

Decompose a task that performs complex processing into a series of separate elements that can be reused. This
can improve performance, scalability, and reusability by allowing task elements that perform the processing to be
deployed and scaled independently.

Context and problem

An application is required to perform a variety of tasks of varying complexity on the information that it processes.
A straightforward but inflexible approach to implementing an application is to perform this processing as a
monolithic module. However, this approach is likely to reduce the opportunities for refactoring the code,
optimizing it, or reusing it if parts of the same processing are required elsewhere within the application.
The figure illustrates the issues with processing data using the monolithic approach. An application receives and
processes data from two sources. The data from each source is processed by a separate module that performs a
series of tasks to transform this data, before passing the result to the business logic of the application.

Some of the tasks that the monolithic modules perform are functionally very similar, but the modules have been
designed separately. The code that implements the tasks is closely coupled in a module, and has been developed
with little or no thought given to reuse or scalability.
However, the processing tasks performed by each module, or the deployment requirements for each task, could
change as business requirements are updated. Some tasks might be compute intensive and could benefit from
running on powerful hardware, while others might not require such expensive resources. Also, additional
processing might be required in the future, or the order in which the tasks performed by the processing could
change. A solution is required that addresses these issues, and increases the possibilities for code reuse.

Solution
Break down the processing required for each stream into a set of separate components (or filters), each
performing a single task. By standardizing the format of the data that each component receives and sends, these
filters can be combined together into a pipeline. This helps to avoid duplicating code, and makes it easy to
remove, replace, or integrate additional components if the processing requirements change. The next figure
shows a solution implemented using pipes and filters.

The time it takes to process a single request depends on the speed of the slowest filter in the pipeline. One or
more filters could be a bottleneck, especially if a large number of requests appear in a stream from a particular
data source. A key advantage of the pipeline structure is that it provides opportunities for running parallel
instances of slow filters, enabling the system to spread the load and improve throughput.
The filters that make up a pipeline can run on different machines, enabling them to be scaled independently and
take advantage of the elasticity that many cloud environments provide. A filter that is computationally intensive
can run on high-performance hardware, while other less demanding filters can be hosted on less expensive
commodity hardware. The filters don't even have to be in the same datacenter or geographic location, which
allows each element in a pipeline to run in an environment close to the resources it requires. The next figure
shows an example applied to the pipeline for the data from Source 1.

If the input and output of a filter are structured as a stream, it's possible to perform the processing for each filter
in parallel. The first filter in the pipeline can start its work and output its results, which are passed directly on to
the next filter in the sequence before the first filter has completed its work.
Another benefit is the resiliency that this model can provide. If a filter fails or the machine it's running on is no
longer available, the pipeline can reschedule the work that the filter was performing and direct this work to
another instance of the component. Failure of a single filter doesn't necessarily result in failure of the entire
pipeline.
Using the Pipes and Filters pattern in conjunction with the Compensating Transaction pattern is an alternative
approach to implementing distributed transactions. A distributed transaction can be broken down into separate,
compensable tasks, each of which can be implemented by using a filter that also implements the Compensating
Transaction pattern. The filters in a pipeline can be implemented as separate hosted tasks running close to the
data that they maintain.

Issues and considerations

You should consider the following points when deciding how to implement this pattern:
Complexity . The increased flexibility that this pattern provides can also introduce complexity, especially if
the filters in a pipeline are distributed across different servers.
Reliability . Use an infrastructure that ensures that data flowing between filters in a pipeline won't be lost.
Idempotency . If a filter in a pipeline fails after receiving a message and the work is rescheduled to another
instance of the filter, part of the work might have already been completed. If this work updates some aspect
of the global state (such as information stored in a database), the same update could be repeated. A similar
issue might occur if a filter fails after posting its results to the next filter in the pipeline, but before
indicating that it's completed its work successfully. In these cases, the same work could be repeated by
another instance of the filter, causing the same results to be posted twice. This could result in subsequent
filters in the pipeline processing the same data twice. Therefore filters in a pipeline should be designed to
be idempotent. For more information, see Idempotency Patterns on Jonathan Oliver’s blog.
Repeated messages . If a filter in a pipeline fails after posting a message to the next stage of the pipeline,
another instance of the filter might be run, and it'll post a copy of the same message to the pipeline. This
could cause two instances of the same message to be passed to the next filter. To avoid this, the pipeline
should detect and eliminate duplicate messages.

If you're implementing the pipeline by using message queues (such as Microsoft Azure Service Bus
queues), the message queuing infrastructure might provide automatic duplicate message detection
and removal.

Context and state . In a pipeline, each filter essentially runs in isolation and shouldn't make any
assumptions about how it was invoked. This means that each filter should be provided with sufficient
context to perform its work. This context could include a large amount of state information.

When to use this pattern

Use this pattern when:
The processing required by an application can easily be broken down into a set of independent steps.
The processing steps performed by an application have different scalability requirements.

It's possible to group filters that should scale together in the same process. For more information, see
the Compute Resource Consolidation pattern.

Flexibility is required to allow reordering of the processing steps performed by an application, or the
capability to add and remove steps.
The system can benefit from distributing the processing for steps across different servers.
 A reliable solution is required that minimizes the effects of failure in a step while data is being processed.
This pattern might not be useful when:
The processing steps performed by an application aren't independent, or they have to be performed
together as part of the same transaction.
The amount of context or state information required by a step makes this approach inefficient. It might be
possible to persist state information to a database instead, but don't use this strategy if the additional load
on the database causes excessive contention.

Example
You can use a sequence of message queues to provide the infrastructure required to implement a pipeline. An
initial message queue receives unprocessed messages. A component implemented as a filter task listens for a
message on this queue, performs its work, and then posts the transformed message to the next queue in the
sequence. Another filter task can listen for messages on this queue, process them, post the results to another
queue, and so on until the fully transformed data appears in the final message in the queue. The next figure
illustrates implementing a pipeline using message queues.

If you're building a solution on Azure you can use Service Bus queues to provide a reliable and scalable queuing
mechanism. The ServiceBusPipeFilter class shown below in C# demonstrates how you can implement a filter
that receives input messages from a queue, processes these messages, and posts the results to another queue.

The ServiceBusPipeFilter class is defined in the PipesAndFilters.Shared project available from GitHub.

public class ServiceBusPipeFilter

{
...
private readonly string inQueuePath;
private readonly string outQueuePath;
...
private QueueClient inQueue;
private QueueClient outQueue;
...

public ServiceBusPipeFilter(..., string inQueuePath, string outQueuePath = null)

{
...
this.inQueuePath = inQueuePath;
this.outQueuePath = outQueuePath;
}

public void Start()

{
...
// Create the outbound filter queue if it doesn't exist.
...
this.outQueue = QueueClient.CreateFromConnectionString(...);

...
// Create the inbound and outbound queue clients.
this.inQueue = QueueClient.CreateFromConnectionString(...);
}
 public void OnPipeFilterMessageAsync(
Func<BrokeredMessage, Task<BrokeredMessage>> asyncFilterTask, ...)
{
...

this.inQueue.OnMessageAsync(
async (msg) =>
{
...
// Process the filter and send the output to the
// next queue in the pipeline.
var outMessage = await asyncFilterTask(msg);

// Send the message from the filter processor

// to the next queue in the pipeline.
if (outQueue != null)
{
await outQueue.SendAsync(outMessage);
}

// Note: There's a chance that the same message could be sent twice
// or that a message gets processed by an upstream or downstream
// filter at the same time.
// This would happen in a situation where processing of a message was
// completed, it was sent to the next pipe/queue, and then failed
// to complete when using the PeekLock method.
// Idempotent message processing and concurrency should be considered
// in a real-world implementation.
},
options);
}

public async Task Close(TimeSpan timespan)

{
// Pause the processing threads.
this.pauseProcessingEvent.Reset();

// There's no clean approach for waiting for the threads to complete

// the processing. This example simply stops any new processing, waits
// for the existing thread to complete, then closes the message pump
// and finally returns.
Thread.Sleep(timespan);

this.inQueue.Close();
...
}

...
}

The method in the ServiceBusPipeFilter class connects to a pair of input and output queues, and the
Start
Close method disconnects from the input queue. The OnPipeFilterMessageAsync method performs the actual
processing of messages, the asyncFilterTask parameter to this method specifies the processing to be performed.
The OnPipeFilterMessageAsync method waits for incoming messages on the input queue, runs the code specified
by the asyncFilterTask parameter over each message as it arrives, and posts the results to the output queue. The
queues themselves are specified by the constructor.
The sample solution implements filters in a set of worker roles. Each worker role can be scaled independently,
depending on the complexity of the business processing that it performs or the resources required for processing.
Additionally, multiple instances of each worker role can be run in parallel to improve throughput.
The following code shows an Azure worker role named PipeFilterARoleEntry , defined in the PipeFilterA project in
the sample solution.
 public class PipeFilterARoleEntry : RoleEntryPoint
{
...
private ServiceBusPipeFilter pipeFilterA;

public override bool OnStart()

{
...
this.pipeFilterA = new ServiceBusPipeFilter(
...,
Constants.QueueAPath,
Constants.QueueBPath);

this.pipeFilterA.Start();
...
}

public override void Run()

{
this.pipeFilterA.OnPipeFilterMessageAsync(async (msg) =>
{
// Clone the message and update it.
// Properties set by the broker (Deliver count, enqueue time, ...)
// aren't cloned and must be copied over if required.
var newMsg = msg.Clone();

await Task.Delay(500); // DOING WORK

Trace.TraceInformation("Filter A processed message:{0} at {1}",

msg.MessageId, DateTime.UtcNow);

newMsg.Properties.Add(Constants.FilterAMessageKey, "Complete");

return newMsg;
});

...
}

...
}

This role contains a ServiceBusPipeFilter object. The OnStart method in the role connects to the queues for
receiving input messages and posting output messages (the names of the queues are defined in the Constants
class). The Run method invokes the OnPipeFilterMessagesAsync method to perform some processing on each
message that's received (in this example, the processing is simulated by waiting for a short period of time). When
processing is complete, a new message is constructed containing the results (in this case, the input message has a
custom property added), and this message is posted to the output queue.
The sample code contains another worker role named PipeFilterBRoleEntry in the PipeFilterB project. This role is
similar to PipeFilterARoleEntry except that it performs different processing in the Run method. In the example
solution, these two roles are combined to construct a pipeline, the output queue for the PipeFilterARoleEntry
role is the input queue for the PipeFilterBRoleEntry role.
The sample solution also provides two additional roles named InitialSenderRoleEntry (in the InitialSender
project) and FinalReceiverRoleEntry (in the FinalReceiver project). The InitialSenderRoleEntry role provides the
initial message in the pipeline. The OnStart method connects to a single queue and the Run method posts a
method to this queue. This queue is the input queue used by the PipeFilterARoleEntry role, so posting a message
to it causes the message to be received and processed by the PipeFilterARoleEntry role. The processed message
then passes through the PipeFilterBRoleEntry role.
The input queue for the FinalReceiveRoleEntry role is the output queue for the PipeFilterBRoleEntry role. The
Run method in the FinalReceiveRoleEntry role, shown below, receives the message and performs some final
processing. Then it writes the values of the custom properties added by the filters in the pipeline to the trace
output.

public class FinalReceiverRoleEntry : RoleEntryPoint

{
...
// Final queue/pipe in the pipeline to process data from.
private ServiceBusPipeFilter queueFinal;

public override bool OnStart()

{
...
// Set up the queue.
this.queueFinal = new ServiceBusPipeFilter(...,Constants.QueueFinalPath);
this.queueFinal.Start();
...
}

public override void Run()

{
this.queueFinal.OnPipeFilterMessageAsync(
async (msg) =>
{
await Task.Delay(500); // DOING WORK

// The pipeline message was received.

Trace.TraceInformation(
"Pipeline Message Complete - FilterA:{0} FilterB:{1}",
msg.Properties[Constants.FilterAMessageKey],
msg.Properties[Constants.FilterBMessageKey]);

return null;
});
...
}

...
}

Related patterns and guidance

The following patterns and guidance might also be relevant when implementing this pattern:
A sample that demonstrates this pattern is available on GitHub.
Competing Consumers pattern. A pipeline can contain multiple instances of one or more filters. This approach
is useful for running parallel instances of slow filters, enabling the system to spread the load and improve
throughput. Each instance of a filter will compete for input with the other instances, two instances of a filter
shouldn't be able to process the same data. Provides an explanation of this approach.
Compute Resource Consolidation pattern. It might be possible to group filters that should scale together into
the same process. Provides more information about the benefits and tradeoffs of this strategy.
Compensating Transaction pattern. A filter can be implemented as an operation that can be reversed, or that
has a compensating operation that restores the state to a previous version in the event of a failure. Explains
how this can be implemented to maintain or achieve eventual consistency.
Idempotency Patterns on Jonathan Oliver’s blog.
 Priority Queue pattern
12/18/2020 • 9 minutes to read • Edit Online

Prioritize requests sent to services so that requests with a higher priority are received and processed more
quickly than those with a lower priority. This pattern is useful in applications that offer different service level
guarantees to individual clients.

Context and Problem

Applications can delegate specific tasks to other services, for example, to perform background processing or to
integrate with other applications or services. In the cloud, a message queue is typically used to delegate tasks to
background processing. In many cases the order requests are received in by a service isn't important. In some
cases, though, it's necessary to prioritize specific requests. These requests should be processed earlier than lower
priority requests that were sent previously by the application.

Solution
A queue is usually a first-in, first-out (FIFO) structure, and consumers typically receive messages in the same
order that they were posted to the queue. However, some message queues support priority messaging. The
application posting a message can assign a priority and the messages in the queue are automatically reordered
so that those with a higher priority will be received before those with a lower priority. The figure illustrates a
queue with priority messaging.

Most message queue implementations support multiple consumers (following the Competing Consumers
pattern), and the number of consumer processes can be scaled up or down depending on demand.

In systems that don't support priority-based message queues, an alternative solution is to maintain a separate
queue for each priority. The application is responsible for posting messages to the appropriate queue. Each
queue can have a separate pool of consumers. Higher priority queues can have a larger pool of consumers
running on faster hardware than lower priority queues. The next figure illustrates using separate message
queues for each priority.
A variation on this strategy is to have a single pool of consumers that check for messages on high priority
queues first, and only then start to fetch messages from lower priority queues. There are some semantic
differences between a solution that uses a single pool of consumer processes (either with a single queue that
supports messages with different priorities or with multiple queues that each handle messages of a single
priority), and a solution that uses multiple queues with a separate pool for each queue.
In the single pool approach, higher priority messages are always received and processed before lower priority
messages. In theory, messages that have a very low priority could be continually superseded and might never be
processed. In the multiple pool approach, lower priority messages will always be processed, just not as quickly as
those of a higher priority (depending on the relative size of the pools and the resources that they have available).
Using a priority queuing mechanism can provide the following advantages:
It allows applications to meet business requirements that require prioritization of availability or
performance, such as offering different levels of service to specific groups of customers.
It can help to minimize operational costs. In the single queue approach, you can scale back the number of
consumers if necessary. High priority messages will still be processed first (although possibly more
slowly), and lower priority messages might be delayed for longer. If you've implemented the multiple
message queue approach with separate pools of consumers for each queue, you can reduce the pool of
consumers for lower priority queues, or even suspend processing for some very low priority queues by
stopping all the consumers that listen for messages on those queues.
The multiple message queue approach can help maximize application performance and scalability by
partitioning messages based on processing requirements. For example, vital tasks can be prioritized to be
handled by receivers that run immediately while less important background tasks can be handled by
receivers that are scheduled to run at less busy periods.

Issues and Considerations

Consider the following points when deciding how to implement this pattern:
Define the priorities in the context of the solution. For example, high priority could mean that messages should
be processed within ten seconds. Identify the requirements for handling high priority items, and the other
resources that should be allocated to meet these criteria.
Decide if all high priority items must be processed before any lower priority items. If the messages are being
processed by a single pool of consumers, you have to provide a mechanism that can preempt and suspend a task
that's handling a low priority message if a higher priority message becomes available.
In the multiple queue approach, when using a single pool of consumer processes that listen on all queues rather
than a dedicated consumer pool for each queue, the consumer must apply an algorithm that ensures it always
services messages from higher priority queues before those from lower priority queues.
Monitor the processing speed on high and low priority queues to ensure that messages in these queues are
processed at the expected rates.
If you need to guarantee that low priority messages will be processed, it's necessary to implement the multiple
message queue approach with multiple pools of consumers. Alternatively, in a queue that supports message
prioritization, it's possible to dynamically increase the priority of a queued message as it ages. However, this
approach depends on the message queue providing this feature.
Using a separate queue for each message priority works best for systems that have a small number of well-
defined priorities.
Message priorities can be determined logically by the system. For example, rather than having explicit high and
low priority messages, they could be designated as “fee paying customer,” or “non-fee paying customer.”
Depending on your business model, your system can allocate more resources to processing messages from fee
paying customers than non-fee paying ones.
There might be a financial and processing cost associated with checking a queue for a message (some
commercial messaging systems charge a small fee each time a message is posted or retrieved, and each time a
queue is queried for messages). This cost increases when checking multiple queues.
It's possible to dynamically adjust the size of a pool of consumers based on the length of the queue that the pool
is servicing. For more information, see the Autoscaling Guidance.

When to use this pattern

This pattern is useful in scenarios where:
The system must handle multiple tasks that have different priorities.
Different users or tenants should be served with different priority.

Example
Microsoft Azure doesn't provide a queuing mechanism that natively supports automatic prioritization of
messages through sorting. However, it does provide Azure Service Bus topics and subscriptions that support a
queuing mechanism that provides message filtering, together with a wide range of flexible capabilities that make
it ideal for use in most priority queue implementations.
An Azure solution can implement a Service Bus topic an application can post messages to, in the same way as a
queue. Messages can contain metadata in the form of application-defined custom properties. Service Bus
subscriptions can be associated with the topic, and these subscriptions can filter messages based on their
properties. When an application sends a message to a topic, the message is directed to the appropriate
subscription where it can be read by a consumer. Consumer processes can retrieve messages from a subscription
using the same semantics as a message queue (a subscription is a logical queue). The following figure illustrates
implementing a priority queue with Azure Service Bus topics and subscriptions.
In the figure above, the application creates several messages and assigns a custom property called Priority in
each message with a value, either High or Low . The application posts these messages to a topic. The topic has
two associated subscriptions that both filter messages by examining the Priority property. One subscription
accepts messages where the Priority property is set to High , and the other accepts messages where the
Priority property is set to Low . A pool of consumers reads messages from each subscription. The high priority
subscription has a larger pool, and these consumers might be running on more powerful computers with more
resources available than the consumers in the low priority pool.
Note that there's nothing special about the designation of high and low priority messages in this example.
They're simply labels specified as properties in each message, and are used to direct messages to a specific
subscription. If additional priorities are required, it's relatively easy to create further subscriptions and pools of
consumer processes to handle these priorities.
The PriorityQueue solution available on GitHub contains an implementation of this approach. This solution
contains two worker role projects named PriorityQueue.High and PriorityQueue.Low . These worker roles inherit
from the PriorityWorkerRole class that contains the functionality for connecting to a specified subscription in the
OnStart method.

The PriorityQueue.High and PriorityQueue.Low worker roles connect to different subscriptions, defined by their
configuration settings. An administrator can configure different numbers of each role to be run. Typically there'll
be more instances of the PriorityQueue.High worker role than the PriorityQueue.Low worker role.
The Run method in the PriorityWorkerRole class arranges for the virtual ProcessMessage method (also defined
in the PriorityWorkerRole class) to be run for each message received on the queue. The following code shows
the Run and ProcessMessage methods. The QueueManager class, defined in the PriorityQueue.Shared project,
provides helper methods for using Azure Service Bus queues.
 public class PriorityWorkerRole : RoleEntryPoint
{
private QueueManager queueManager;
...

public override void Run()

{
// Start listening for messages on the subscription.
var subscriptionName = CloudConfigurationManager.GetSetting("SubscriptionName");
this.queueManager.ReceiveMessages(subscriptionName, this.ProcessMessage);
...;
}
...

protected virtual async Task ProcessMessage(BrokeredMessage message)

{
// Simulating processing.
await Task.Delay(TimeSpan.FromSeconds(2));
}
}

The PriorityQueue.Highand PriorityQueue.Low worker roles both override the default functionality of the
ProcessMessage method. The code below shows the ProcessMessage method for the PriorityQueue.High worker
role.

protected override async Task ProcessMessage(BrokeredMessage message)

{
// Simulate message processing for High priority messages.
await base.ProcessMessage(message);
Trace.TraceInformation("High priority message processed by " +
RoleEnvironment.CurrentRoleInstance.Id + " MessageId: " + message.MessageId);
}

When an application posts messages to the topic associated with the subscriptions used by the
PriorityQueue.High and PriorityQueue.Low worker roles, it specifies the priority by using the Priority custom
property, as shown in the following code example. This code (implemented in the WorkerRole class in the
PriorityQueue.Sender project), uses the SendBatchAsync helper method of the QueueManager class to post
messages to a topic in batches.
 // Send a low priority batch.
var lowMessages = new List<BrokeredMessage>();

for (int i = 0; i < 10; i++)

{
var message = new BrokeredMessage() { MessageId = Guid.NewGuid().ToString() };
message.Properties["Priority"] = Priority.Low;
lowMessages.Add(message);
}

this.queueManager.SendBatchAsync(lowMessages).Wait();
...

// Send a high priority batch.

var highMessages = new List<BrokeredMessage>();

for (int i = 0; i < 10; i++)

{
var message = new BrokeredMessage() { MessageId = Guid.NewGuid().ToString() };
message.Properties["Priority"] = Priority.High;
highMessages.Add(message);
}

this.queueManager.SendBatchAsync(highMessages).Wait();

Related patterns and guidance

The following patterns and guidance might also be relevant when implementing this pattern:
A sample that demonstrates this pattern is available on GitHub.
Asynchronous Messaging Primer. A consumer service that processes a request might need to send a reply
to the instance of the application that posted the request. Provides information on the strategies that you
can use to implement request/response messaging.
Competing Consumers pattern. To increase the throughput of the queues, it’s possible to have multiple
consumers that listen on the same queue, and process the tasks in parallel. These consumers will compete
for messages, but only one should be able to process each message. Provides more information on the
benefits and tradeoffs of implementing this approach.
Throttling pattern. You can implement throttling by using queues. Priority messaging can be used to
ensure that requests from critical applications, or applications being run by high-value customers, are
given priority over requests from less important applications.
Autoscaling Guidance. It might be possible to scale the size of the pool of consumer processes handling a
queue depending on the length of the queue. This strategy can help to improve performance, especially
for pools handling high priority messages.
Enterprise Integration Patterns with Service Bus on Abhishek Lal’s blog.
 Publisher-Subscriber pattern
12/18/2020 • 6 minutes to read • Edit Online

Enable an application to announce events to multiple interested consumers asynchronously, without coupling the
senders to the receivers.
Also called : Pub/sub messaging

Context and problem

In cloud-based and distributed applications, components of the system often need to provide information to other
components as events happen.
Asynchronous messaging is an effective way to decouple senders from consumers, and avoid blocking the sender
to wait for a response. However, using a dedicated message queue for each consumer does not effectively scale to
many consumers. Also, some of the consumers might be interested in only a subset of the information. How can
the sender announce events to all interested consumers without knowing their identities?

Solution
Introduce an asynchronous messaging subsystem that includes the following:
An input messaging channel used by the sender. The sender packages events into messages, using a known
message format, and sends these messages via the input channel. The sender in this pattern is also called
the publisher.

NOTE
A message is a packet of data. An event is a message that notifies other components about a change or an action
that has taken place.

One output messaging channel per consumer. The consumers are known as subscribers.
A mechanism for copying each message from the input channel to the output channels for all subscribers
interested in that message. This operation is typically handled by an intermediary such as a message
broker or event bus.
The following diagram shows the logical components of this pattern:
Pub/sub messaging has the following benefits:
It decouples subsystems that still need to communicate. Subsystems can be managed independently, and
messages can be properly managed even if one or more receivers are offline.
It increases scalability and improves responsiveness of the sender. The sender can quickly send a single
message to the input channel, then return to its core processing responsibilities. The messaging
infrastructure is responsible for ensuring messages are delivered to interested subscribers.
It improves reliability. Asynchronous messaging helps applications continue to run smoothly under
increased loads and handle intermittent failures more effectively.
It allows for deferred or scheduled processing. Subscribers can wait to pick up messages until off-peak
hours, or messages can be routed or processed according to a specific schedule.
It enables simpler integration between systems using different platforms, programming languages, or
communication protocols, as well as between on-premises systems and applications running in the cloud.
It facilitates asynchronous workflows across an enterprise.
It improves testability. Channels can be monitored and messages can be inspected or logged as part of an
overall integration test strategy.
It provides separation of concerns for your applications. Each application can focus on its core capabilities,
while the messaging infrastructure handles everything required to reliably route messages to multiple
consumers.

Issues and considerations

Consider the following points when deciding how to implement this pattern:
Existing technologies. It is strongly recommended to use available messaging products and services that
support a publish-subscribe model, rather than building your own. In Azure, consider using Service Bus or
Event Grid. Other technologies that can be used for pub/sub messaging include Redis, RabbitMQ, and
Apache Kafka.
Subscription handling. The messaging infrastructure must provide mechanisms that consumers can use
to subscribe to or unsubscribe from available channels.
Security. Connecting to any message channel must be restricted by security policy to prevent
eavesdropping by unauthorized users or applications.
Subsets of messages. Subscribers are usually only interested in subset of the messages distributed by a
publisher. Messaging services often allow subscribers to narrow the set of messages received by:
Topics. Each topic has a dedicated output channel, and each consumer can subscribe to all relevant
topics.
Content filtering. Messages are inspected and distributed based on the content of each message. Each
subscriber can specify the content it is interested in.
Wildcard subscribers. Consider allowing subscribers to subscribe to multiple topics via wildcards.
Bi-directional communication. The channels in a publish-subscribe system are treated as unidirectional.
If a specific subscriber needs to send acknowledgment or communicate status back to the publisher,
consider using the Request/Reply Pattern. This pattern uses one channel to send a message to the
subscriber, and a separate reply channel for communicating back to the publisher.
Message ordering. The order in which consumer instances receive messages isn't guaranteed, and
doesn't necessarily reflect the order in which the messages were created. Design the system to ensure that
 message processing is idempotent to help eliminate any dependency on the order of message handling.
Message priority. Some solutions may require that messages are processed in a specific order. The
Priority Queue pattern provides a mechanism for ensuring specific messages are delivered before others.
Poison messages. A malformed message, or a task that requires access to resources that aren't available,
can cause a service instance to fail. The system should prevent such messages being returned to the queue.
Instead, capture and store the details of these messages elsewhere so that they can be analyzed if
necessary.
Repeated messages. The same message might be sent more than once. For example, the sender might
fail after posting a message. Then a new instance of the sender might start up and repeat the message. The
messaging infrastructure should implement duplicate message detection and removal (also known as de-
duping) based on message IDs in order to provide at-most-once delivery of messages.
Message expiration. A message might have a limited lifetime. If it isn't processed within this period, it
might no longer be relevant and should be discarded. A sender can specify an expiration time as part of the
data in the message. A receiver can examine this information before deciding whether to perform the
business logic associated with the message.
Message scheduling. A message might be temporarily embargoed and should not be processed until a
specific date and time. The message should not be available to a receiver until this time.

When to use this pattern

Use this pattern when:
An application needs to broadcast information to a significant number of consumers.
An application needs to communicate with one or more independently-developed applications or services,
which may use different platforms, programming languages, and communication protocols.
An application can send information to consumers without requiring real-time responses from the
consumers.
The systems being integrated are designed to support an eventual consistency model for their data.
An application needs to communicate information to multiple consumers, which may have different
availability requirements or uptime schedules than the sender.
This pattern might not be useful when:
An application has only a few consumers who need significantly different information from the producing
application.
An application requires near real-time interaction with consumers.

Example
The following diagram shows an enterprise integration architecture that uses Service Bus to coordinate
workflows, and Event Grid to notify subsystems of events that occur. For more information, see Enterprise
integration on Azure using message queues and events.
Related patterns and guidance
The following patterns and guidance might be relevant when implementing this pattern:
Choose between Azure services that deliver messages.
The Event-driven architecture style is an architecture style that uses pub/sub messaging.
Asynchronous Messaging Primer. Message queues are an asynchronous communications mechanism. If a
consumer service needs to send a reply to an application, it might be necessary to implement some form
of response messaging. The Asynchronous Messaging Primer provides information on how to implement
request/reply messaging using message queues.
Observer Pattern. The Publish-Subscribe pattern builds on the Observer pattern by decoupling subjects
from observers via asynchronous messaging.
Message Broker Pattern. Many messaging subsystems that support a publish-subscribe model are
implemented via a message broker.
 Queue-Based Load Leveling pattern
12/18/2020 • 5 minutes to read • Edit Online

Use a queue that acts as a buffer between a task and a service it invokes in order to smooth intermittent heavy
loads that can cause the service to fail or the task to time out. This can help to minimize the impact of peaks in
demand on availability and responsiveness for both the task and the service.

Context and problem

Many solutions in the cloud involve running tasks that invoke services. In this environment, if a service is
subjected to intermittent heavy loads, it can cause performance or reliability issues.
A service could be part of the same solution as the tasks that use it, or it could be a third-party service
providing access to frequently used resources such as a cache or a storage service. If the same service is used
by a number of tasks running concurrently, it can be difficult to predict the volume of requests to the service at
any time.
A service might experience peaks in demand that cause it to overload and be unable to respond to requests in
a timely manner. Flooding a service with a large number of concurrent requests can also result in the service
failing if it's unable to handle the contention these requests cause.

Solution
Refactor the solution and introduce a queue between the task and the service. The task and the service run
asynchronously. The task posts a message containing the data required by the service to a queue. The queue
acts as a buffer, storing the message until it's retrieved by the service. The service retrieves the messages from
the queue and processes them. Requests from a number of tasks, which can be generated at a highly variable
rate, can be passed to the service through the same message queue. This figure shows using a queue to level
the load on a service.

The queue decouples the tasks from the service, and the service can handle the messages at its own pace
regardless of the volume of requests from concurrent tasks. Additionally, there's no delay to a task if the
service isn't available at the time it posts a message to the queue.
This pattern provides the following benefits:
It can help to maximize availability because delays arising in services won't have an immediate and
direct impact on the application, which can continue to post messages to the queue even when the
service isn't available or isn't currently processing messages.
 It can help to maximize scalability because both the number of queues and the number of services can
be varied to meet demand.
It can help to control costs because the number of service instances deployed only have to be adequate
to meet average load rather than the peak load.

Some services implement throttling when demand reaches a threshold beyond which the system
could fail. Throttling can reduce the functionality available. You can implement load leveling with
these services to ensure that this threshold isn't reached.

Issues and considerations

Consider the following points when deciding how to implement this pattern:
It's necessary to implement application logic that controls the rate at which services handle messages to
avoid overwhelming the target resource. Avoid passing spikes in demand to the next stage of the system.
Test the system under load to ensure that it provides the required leveling, and adjust the number of
queues and the number of service instances that handle messages to achieve this.
Message queues are a one-way communication mechanism. If a task expects a reply from a service, it might
be necessary to implement a mechanism that the service can use to send a response. For more information,
see the Asynchronous Messaging Primer.
Be careful if you apply autoscaling to services that are listening for requests on the queue. This can result in
increased contention for any resources that these services share and diminish the effectiveness of using the
queue to level the load.

When to use this pattern

This pattern is useful to any application that uses services that are subject to overloading.
This pattern isn't useful if the application expects a response from the service with minimal latency.

Example
A web app writes data to an external data store. If a large number of instances of the web app run concurrently,
the data store might be unable to respond to requests quickly enough, causing requests to time out, be
throttled, or otherwise fail. The following diagram shows a data store being overwhelmed by a large number
of concurrent requests from instances of an application.
To resolve this, you can use a queue to level the load between the application instances and the data store. An
Azure Functions app reads the messages from the queue and performs the read/write requests to the data
store. The application logic in the function app can control the rate at which it passes requests to the data store,
to prevent the store from being overwhelmed. (Otherwise the function app will just re-introduce the same
problem at the back end.)

Related patterns and guidance

The following patterns and guidance might also be relevant when implementing this pattern:
Asynchronous Messaging Primer. Message queues are inherently asynchronous. It might be necessary
to redesign the application logic in a task if it's adapted from communicating directly with a service to
using a message queue. Similarly, it might be necessary to refactor a service to accept requests from a
message queue. Alternatively, it might be possible to implement a proxy service, as described in the
example.
Competing Consumers pattern. It might be possible to run multiple instances of a service, each acting
as a message consumer from the load-leveling queue. You can use this approach to adjust the rate at
which messages are received and passed to a service.
Throttling pattern. A simple way to implement throttling with a service is to use queue-based load
leveling and route all requests to a service through a message queue. The service can process requests
at a rate that ensures that resources required by the service aren't exhausted, and to reduce the amount
of contention that could occur.
Choose between Azure messaging services. Information about choosing a messaging and queuing
mechanism in Azure applications.
Improve scalability in an Azure web application. This reference architecture includes queue-based load
leveling as part of the architecture.
 Retry pattern
12/18/2020 • 10 minutes to read • Edit Online

Enable an application to handle transient failures when it tries to connect to a service or network resource, by
transparently retrying a failed operation. This can improve the stability of the application.

Context and problem

An application that communicates with elements running in the cloud has to be sensitive to the transient faults
that can occur in this environment. Faults include the momentary loss of network connectivity to components
and services, the temporary unavailability of a service, or timeouts that occur when a service is busy.
These faults are typically self-correcting, and if the action that triggered a fault is repeated after a suitable delay
it's likely to be successful. For example, a database service that's processing a large number of concurrent
requests can implement a throttling strategy that temporarily rejects any further requests until its workload has
eased. An application trying to access the database might fail to connect, but if it tries again after a delay it
might succeed.

Solution
In the cloud, transient faults aren't uncommon and an application should be designed to handle them elegantly
and transparently. This minimizes the effects faults can have on the business tasks the application is performing.
If an application detects a failure when it tries to send a request to a remote service, it can handle the failure
using the following strategies:
Cancel . If the fault indicates that the failure isn't transient or is unlikely to be successful if repeated, the
application should cancel the operation and report an exception. For example, an authentication failure
caused by providing invalid credentials is not likely to succeed no matter how many times it's attempted.
Retr y . If the specific fault reported is unusual or rare, it might have been caused by unusual
circumstances such as a network packet becoming corrupted while it was being transmitted. In this case,
the application could retry the failing request again immediately because the same failure is unlikely to
be repeated and the request will probably be successful.
Retr y after delay . If the fault is caused by one of the more commonplace connectivity or busy failures,
the network or service might need a short period while the connectivity issues are corrected or the
backlog of work is cleared. The application should wait for a suitable time before retrying the request.
For the more common transient failures, the period between retries should be chosen to spread requests from
multiple instances of the application as evenly as possible. This reduces the chance of a busy service continuing
to be overloaded. If many instances of an application are continually overwhelming a service with retry
requests, it'll take the service longer to recover.
If the request still fails, the application can wait and make another attempt. If necessary, this process can be
repeated with increasing delays between retry attempts, until some maximum number of requests have been
attempted. The delay can be increased incrementally or exponentially, depending on the type of failure and the
probability that it'll be corrected during this time.
The following diagram illustrates invoking an operation in a hosted service using this pattern. If the request is
unsuccessful after a predefined number of attempts, the application should treat the fault as an exception and
handle it accordingly.
The application should wrap all attempts to access a remote service in code that implements a retry policy
matching one of the strategies listed above. Requests sent to different services can be subject to different
policies. Some vendors provide libraries that implement retry policies, where the application can specify the
maximum number of retries, the time between retry attempts, and other parameters.
An application should log the details of faults and failing operations. This information is useful to operators. If a
service is frequently unavailable or busy, it's often because the service has exhausted its resources. You can
reduce the frequency of these faults by scaling out the service. For example, if a database service is continually
overloaded, it might be beneficial to partition the database and spread the load across multiple servers.

Microsoft Entity Framework provides facilities for retrying database operations. Also, most Azure services
and client SDKs include a retry mechanism. For more information, see Retry guidance for specific services.

Issues and considerations

You should consider the following points when deciding how to implement this pattern.
The retry policy should be tuned to match the business requirements of the application and the nature of the
failure. For some noncritical operations, it's better to fail fast rather than retry several times and impact the
throughput of the application. For example, in an interactive web application accessing a remote service, it's
better to fail after a smaller number of retries with only a short delay between retry attempts, and display a
suitable message to the user (for example, “please try again later”). For a batch application, it might be more
appropriate to increase the number of retry attempts with an exponentially increasing delay between attempts.
An aggressive retry policy with minimal delay between attempts, and a large number of retries, could further
degrade a busy service that's running close to or at capacity. This retry policy could also affect the
responsiveness of the application if it's continually trying to perform a failing operation.
If a request still fails after a significant number of retries, it's better for the application to prevent further
requests going to the same resource and simply report a failure immediately. When the period expires, the
application can tentatively allow one or more requests through to see whether they're successful. For more
details of this strategy, see the Circuit Breaker pattern.
Consider whether the operation is idempotent. If so, it's inherently safe to retry. Otherwise, retries could cause
the operation to be executed more than once, with unintended side effects. For example, a service might receive
the request, process the request successfully, but fail to send a response. At that point, the retry logic might re-
send the request, assuming that the first request wasn't received.
A request to a service can fail for a variety of reasons raising different exceptions depending on the nature of
the failure. Some exceptions indicate a failure that can be resolved quickly, while others indicate that the failure
is longer lasting. It's useful for the retry policy to adjust the time between retry attempts based on the type of
the exception.
Consider how retrying an operation that's part of a transaction will affect the overall transaction consistency.
Fine tune the retry policy for transactional operations to maximize the chance of success and reduce the need to
undo all the transaction steps.
Ensure that all retry code is fully tested against a variety of failure conditions. Check that it doesn't severely
impact the performance or reliability of the application, cause excessive load on services and resources, or
generate race conditions or bottlenecks.
Implement retry logic only where the full context of a failing operation is understood. For example, if a task that
contains a retry policy invokes another task that also contains a retry policy, this extra layer of retries can add
long delays to the processing. It might be better to configure the lower-level task to fail fast and report the
reason for the failure back to the task that invoked it. This higher-level task can then handle the failure based on
its own policy.
It's important to log all connectivity failures that cause a retry so that underlying problems with the application,
services, or resources can be identified.
Investigate the faults that are most likely to occur for a service or a resource to discover if they're likely to be
long lasting or terminal. If they are, it's better to handle the fault as an exception. The application can report or
log the exception, and then try to continue either by invoking an alternative service (if one is available), or by
offering degraded functionality. For more information on how to detect and handle long-lasting faults, see the
Circuit Breaker pattern.

When to use this pattern

Use this pattern when an application could experience transient faults as it interacts with a remote service or
accesses a remote resource. These faults are expected to be short lived, and repeating a request that has
previously failed could succeed on a subsequent attempt.
This pattern might not be useful:
When a fault is likely to be long lasting, because this can affect the responsiveness of an application. The
application might be wasting time and resources trying to repeat a request that's likely to fail.
For handling failures that aren't due to transient faults, such as internal exceptions caused by errors in the
business logic of an application.
As an alternative to addressing scalability issues in a system. If an application experiences frequent busy
faults, it's often a sign that the service or resource being accessed should be scaled up.

Example
This example in C# illustrates an implementation of the Retry pattern. The OperationWithBasicRetryAsync
method, shown below, invokes an external service asynchronously through the TransientOperationAsync
method. The details of the TransientOperationAsync method will be specific to the service and are omitted from
the sample code.
 private int retryCount = 3;
private readonly TimeSpan delay = TimeSpan.FromSeconds(5);

public async Task OperationWithBasicRetryAsync()

{
int currentRetry = 0;

for (;;)
{
try
{
// Call external service.
await TransientOperationAsync();

// Return or break.
break;
}
catch (Exception ex)
{
Trace.TraceError("Operation Exception");

currentRetry++;

// Check if the exception thrown was a transient exception

// based on the logic in the error detection strategy.
// Determine whether to retry the operation, as well as how
// long to wait, based on the retry strategy.
if (currentRetry > this.retryCount || !IsTransient(ex))
{
// If this isn't a transient error or we shouldn't retry,
// rethrow the exception.
throw;
}
}

// Wait to retry the operation.

// Consider calculating an exponential delay here and
// using a strategy best suited for the operation and fault.
await Task.Delay(delay);
}
}

// Async method that wraps a call to a remote service (details not shown).
private async Task TransientOperationAsync()
{
...
}

The statement that invokes this method is contained in a try/catch block wrapped in a for loop. The for loop
exits if the call to the TransientOperationAsync method succeeds without throwing an exception. If the
TransientOperationAsync method fails, the catch block examines the reason for the failure. If it's believed to be a
transient error the code waits for a short delay before retrying the operation.
The for loop also tracks the number of times that the operation has been attempted, and if the code fails three
times the exception is assumed to be more long lasting. If the exception isn't transient or it's long lasting, the
catch handler throws an exception. This exception exits the for loop and should be caught by the code that
invokes the OperationWithBasicRetryAsync method.
The IsTransient method, shown below, checks for a specific set of exceptions that are relevant to the
environment the code is run in. The definition of a transient exception will vary according to the resources being
accessed and the environment the operation is being performed in.
 private bool IsTransient(Exception ex)
{
// Determine if the exception is transient.
// In some cases this is as simple as checking the exception type, in other
// cases it might be necessary to inspect other properties of the exception.
if (ex is OperationTransientException)
return true;

var webException = ex as WebException;

if (webException != null)
{
// If the web exception contains one of the following status values
// it might be transient.
return new[] {WebExceptionStatus.ConnectionClosed,
WebExceptionStatus.Timeout,
WebExceptionStatus.RequestCanceled }.
Contains(webException.Status);
}

// Additional exception checking logic goes here.

return false;
}

Related patterns and guidance

Circuit Breaker pattern. If a failure is expected to be more long lasting, it might be more appropriate to
implement the Circuit Breaker pattern. Combining the Retry and Circuit Breaker patterns provides a
comprehensive approach to handling faults.
For most Azure services, the client SDKs include built-in retry logic. For more information, see Retry
guidance for Azure services.
Before writing custom retry logic, consider using a general framework such as Polly for .NET or
Resilience4j for Java.
When processing commands that change business data, be aware that retries can result in the action
being performed twice, which could be problematic if that action is something like charging a customer's
credit card. Using the Idempotence pattern described in this blog post can help deal with these situations.
 Scheduler Agent Supervisor pattern
12/18/2020 • 16 minutes to read • Edit Online

Coordinate a set of distributed actions as a single operation. If any of the actions fail, try to handle the failures
transparently, or else undo the work that was performed, so the entire operation succeeds or fails as a whole.
This can add resiliency to a distributed system, by enabling it to recover and retry actions that fail due to
transient exceptions, long-lasting faults, and process failures.

Context and problem

An application performs tasks that include a number of steps, some of which might invoke remote services or
access remote resources. The individual steps might be independent of each other, but they are orchestrated by
the application logic that implements the task.
Whenever possible, the application should ensure that the task runs to completion and resolve any failures that
might occur when accessing remote services or resources. Failures can occur for many reasons. For example, the
network might be down, communications could be interrupted, a remote service might be unresponsive or in an
unstable state, or a remote resource might be temporarily inaccessible, perhaps due to resource constraints. In
many cases the failures will be transient and can be handled by using the Retry pattern.
If the application detects a more permanent fault it can't easily recover from, it must be able to restore the
system to a consistent state and ensure integrity of the entire operation.

Solution
The Scheduler Agent Supervisor pattern defines the following actors. These actors orchestrate the steps to be
performed as part of the overall task.
The Scheduler arranges for the steps that make up the task to be executed and orchestrates their
operation. These steps can be combined into a pipeline or workflow. The Scheduler is responsible for
ensuring that the steps in this workflow are performed in the right order. As each step is performed, the
Scheduler records the state of the workflow, such as "step not yet started," "step running," or "step
completed." The state information should also include an upper limit of the time allowed for the step to
finish, called the complete-by time. If a step requires access to a remote service or resource, the Scheduler
invokes the appropriate Agent, passing it the details of the work to be performed. The Scheduler typically
communicates with an Agent using asynchronous request/response messaging. This can be implemented
using queues, although other distributed messaging technologies could be used instead.

The Scheduler performs a similar function to the Process Manager in the Process Manager pattern.
The actual workflow is typically defined and implemented by a workflow engine that's controlled by
the Scheduler. This approach decouples the business logic in the workflow from the Scheduler.

The Agent contains logic that encapsulates a call to a remote service, or access to a remote resource
referenced by a step in a task. Each Agent typically wraps calls to a single service or resource,
implementing the appropriate error handling and retry logic (subject to a timeout constraint, described
later). If the steps in the workflow being run by the Scheduler use several services and resources across
different steps, each step might reference a different Agent (this is an implementation detail of the
pattern).
The Super visor monitors the status of the steps in the task being performed by the Scheduler. It runs
periodically (the frequency will be system-specific), and examines the status of steps maintained by the
 Scheduler. If it detects any that have timed out or failed, it arranges for the appropriate Agent to recover
the step or execute the appropriate remedial action (this might involve modifying the status of a step).
Note that the recovery or remedial actions are implemented by the Scheduler and Agents. The Supervisor
should simply request that these actions be performed.
The Scheduler, Agent, and Supervisor are logical components and their physical implementation depends on the
technology being used. For example, several logical agents might be implemented as part of a single web
service.
The Scheduler maintains information about the progress of the task and the state of each step in a durable data
store, called the state store. The Supervisor can use this information to help determine whether a step has failed.
The figure illustrates the relationship between the Scheduler, the Agents, the Supervisor, and the state store.

NOTE
This diagram shows a simplified version of the pattern. In a real implementation, there might be many instances of the
Scheduler running concurrently, each a subset of tasks. Similarly, the system could run multiple instances of each Agent, or
even multiple Supervisors. In this case, Supervisors must coordinate their work with each other carefully to ensure that
they don’t compete to recover the same failed steps and tasks. The Leader Election pattern provides one possible solution
to this problem.

When the application is ready to run a task, it submits a request to the Scheduler. The Scheduler records initial
state information about the task and its steps (for example, step not yet started) in the state store and then starts
performing the operations defined by the workflow. As the Scheduler starts each step, it updates the information
about the state of that step in the state store (for example, step running).
If a step references a remote service or resource, the Scheduler sends a message to the appropriate Agent. The
message contains the information that the Agent needs to pass to the service or access the resource, in addition
to the complete-by time for the operation. If the Agent completes its operation successfully, it returns a response
to the Scheduler. The Scheduler can then update the state information in the state store (for example, step
completed) and perform the next step. This process continues until the entire task is complete.
An Agent can implement any retry logic that's necessary to perform its work. However, if the Agent doesn't
complete its work before the complete-by period expires, the Scheduler will assume that the operation has
failed. In this case, the Agent should stop its work and not try to return anything to the Scheduler (not even an
error message), or try any form of recovery. The reason for this restriction is that, after a step has timed out or
failed, another instance of the Agent might be scheduled to run the failing step (this process is described later).
If the Agent fails, the Scheduler won't receive a response. The pattern doesn't make a distinction between a step
that has timed out and one that has genuinely failed.
If a step times out or fails, the state store will contain a record that indicates that the step is running, but the
complete-by time will have passed. The Supervisor looks for steps like this and tries to recover them. One
possible strategy is for the Supervisor to update the complete-by value to extend the time available to complete
the step, and then send a message to the Scheduler identifying the step that has timed out. The Scheduler can
then try to repeat this step. However, this design requires the tasks to be idempotent.
The Supervisor might need to prevent the same step from being retried if it continually fails or times out. To do
this, the Supervisor could maintain a retry count for each step, along with the state information, in the state
store. If this count exceeds a predefined threshold the Supervisor can adopt a strategy of waiting for an extended
period before notifying the Scheduler that it should retry the step, in the expectation that the fault will be
resolved during this period. Alternatively, the Supervisor can send a message to the Scheduler to request the
entire task be undone by implementing a Compensating Transaction pattern. This approach will depend on the
Scheduler and Agents providing the information necessary to implement the compensating operations for each
step that completed successfully.

It isn't the purpose of the Supervisor to monitor the Scheduler and Agents, and restart them if they fail. This
aspect of the system should be handled by the infrastructure these components are running in. Similarly, the
Supervisor shouldn't have knowledge of the actual business operations that the tasks being performed by
the Scheduler are running (including how to compensate should these tasks fail). This is the purpose of the
workflow logic implemented by the Scheduler. The sole responsibility of the Supervisor is to determine
whether a step has failed and arrange either for it to be repeated or for the entire task containing the failed
step to be undone.

If the Scheduler is restarted after a failure, or the workflow being performed by the Scheduler terminates
unexpectedly, the Scheduler should be able to determine the status of any inflight task that it was handling when
it failed, and be prepared to resume this task from that point. The implementation details of this process are
likely to be system-specific. If the task can't be recovered, it might be necessary to undo the work already
performed by the task. This might also require implementing a compensating transaction.
The key advantage of this pattern is that the system is resilient in the event of unexpected temporary or
unrecoverable failures. The system can be constructed to be self-healing. For example, if an Agent or the
Scheduler fails, a new one can be started and the Supervisor can arrange for a task to be resumed. If the
Supervisor fails, another instance can be started and can take over from where the failure occurred. If the
Supervisor is scheduled to run periodically, a new instance can be automatically started after a predefined
interval. The state store can be replicated to reach an even greater degree of resiliency.

Issues and considerations

You should consider the following points when deciding how to implement this pattern:
This pattern can be difficult to implement and requires thorough testing of each possible failure mode of
the system.
 The recovery/retry logic implemented by the Scheduler is complex and dependent on state information
held in the state store. It might also be necessary to record the information required to implement a
compensating transaction in a durable data store.
How often the Supervisor runs will be important. It should run often enough to prevent any failed steps
from blocking an application for an extended period, but it shouldn't run so often that it becomes an
overhead.
The steps performed by an Agent could be run more than once. The logic that implements these steps
should be idempotent.

When to use this pattern

Use this pattern when a process that runs in a distributed environment, such as the cloud, must be resilient to
communications failure and/or operational failure.
This pattern might not be suitable for tasks that don't invoke remote services or access remote resources.

Example
A web application that implements an ecommerce system has been deployed on Microsoft Azure. Users can run
this application to browse the available products and to place orders. The user interface runs as a web role, and
the order processing elements of the application are implemented as a set of worker roles. Part of the order
processing logic involves accessing a remote service, and this aspect of the system could be prone to transient
or more long-lasting faults. For this reason, the designers used the Scheduler Agent Supervisor pattern to
implement the order processing elements of the system.
When a customer places an order, the application constructs a message that describes the order and posts this
message to a queue. A separate submission process, running in a worker role, retrieves the message, inserts the
order details into the orders database, and creates a record for the order process in the state store. Note that the
inserts into the orders database and the state store are performed as part of the same operation. The submission
process is designed to ensure that both inserts complete together.
The state information that the submission process creates for the order includes:
OrderID . The ID of the order in the orders database.
LockedBy . The instance ID of the worker role handling the order. There might be multiple current
instances of the worker role running the Scheduler, but each order should only be handled by a single
instance.
CompleteBy . The time the order should be processed by.
ProcessState . The current state of the task handling the order. The possible states are:
Pending . The order has been created but processing hasn't yet been started.
Processing . The order is currently being processed.
Processed . The order has been processed successfully.
Error . The order processing has failed.
FailureCount . The number of times that processing has been tried for the order.
In this state information, the OrderID field is copied from the order ID of the new order. The LockedBy and
CompleteBy fields are set to null , the ProcessState field is set to Pending , and the FailureCount field is set to
0.
 NOTE
In this example, the order handling logic is relatively simple and only has a single step that invokes a remote service. In a
more complex multistep scenario, the submission process would likely involve several steps, and so several records would
be created in the state store — each one describing the state of an individual step.

The Scheduler also runs as part of a worker role and implements the business logic that handles the order. An
instance of the Scheduler polling for new orders examines the state store for records where the LockedBy field is
null and the ProcessState field is pending. When the Scheduler finds a new order, it immediately populates the
LockedBy field with its own instance ID, sets the CompleteBy field to an appropriate time, and sets the
ProcessState field to processing. The code is designed to be exclusive and atomic to ensure that two concurrent
instances of the Scheduler can't try to handle the same order simultaneously.
The Scheduler then runs the business workflow to process the order asynchronously, passing it the value in the
OrderID field from the state store. The workflow handling the order retrieves the details of the order from the
orders database and performs its work. When a step in the order processing workflow needs to invoke the
remote service, it uses an Agent. The workflow step communicates with the Agent using a pair of Azure Service
Bus message queues acting as a request/response channel. The figure shows a high-level view of the solution.
The message sent to the Agent from a workflow step describes the order and includes the complete-by time. If
the Agent receives a response from the remote service before the complete-by time expires, it posts a reply
message on the Service Bus queue on which the workflow is listening. When the workflow step receives the
valid reply message, it completes its processing and the Scheduler sets the ProcessState field of the order state
to processed. At this point, the order processing has completed successfully.
If the complete-by time expires before the Agent receives a response from the remote service, the Agent simply
halts its processing and terminates handling the order. Similarly, if the workflow handling the order exceeds the
complete-by time, it also terminates. In both cases, the state of the order in the state store remains set to
processing, but the complete-by time indicates that the time for processing the order has passed and the process
is deemed to have failed. Note that if the Agent that's accessing the remote service, or the workflow that's
handling the order (or both) terminate unexpectedly, the information in the state store will again remain set to
processing and eventually will have an expired complete-by value.
If the Agent detects an unrecoverable, nontransient fault while it's trying to contact the remote service, it can
send an error response back to the workflow. The Scheduler can set the status of the order to error and raise an
event that alerts an operator. The operator can then try to resolve the reason for the failure manually and
resubmit the failed processing step.
The Supervisor periodically examines the state store looking for orders with an expired complete-by value. If the
Supervisor finds a record, it increments the FailureCount field. If the failure count value is below a specified
threshold value, the Supervisor resets the LockedBy field to null, updates the CompleteBy field with a new
expiration time, and sets the ProcessState field to pending. An instance of the Scheduler can pick up this order
and perform its processing as before. If the failure count value exceeds a specified threshold, the reason for the
failure is assumed to be nontransient. The Supervisor sets the status of the order to error and raises an event
that alerts an operator.

In this example, the Supervisor is implemented in a separate worker role. You can use a variety of strategies
to arrange for the Supervisor task to be run, including using the Azure Scheduler service (not to be confused
with the Scheduler component in this pattern). For more information about the Azure Scheduler service, visit
the Scheduler page.

Although it isn't shown in this example, the Scheduler might need to keep the application that submitted the
order informed about the progress and status of the order. The application and the Scheduler are isolated from
each other to eliminate any dependencies between them. The application has no knowledge of which instance of
the Scheduler is handling the order, and the Scheduler is unaware of which specific application instance posted
the order.
To allow the order status to be reported, the application could use its own private response queue. The details of
this response queue would be included as part of the request sent to the submission process, which would
include this information in the state store. The Scheduler would then post messages to this queue indicating the
status of the order (request received, order completed, order failed, and so on). It should include the order ID in
these messages so they can be correlated with the original request by the application.

Related patterns and guidance

The following patterns and guidance might also be relevant when implementing this pattern:
Retry pattern. An Agent can use this pattern to transparently retry an operation that accesses a remote
service or resource that has previously failed. Use when the expectation is that the cause of the failure is
transient and can be corrected.
Circuit Breaker pattern. An Agent can use this pattern to handle faults that take a variable amount of time to
correct when connecting to a remote service or resource.
Compensating Transaction pattern. If the workflow being performed by a Scheduler can't be completed
successfully, it might be necessary to undo any work it's previously performed. The Compensating
Transaction pattern describes how this can be achieved for operations that follow the eventual consistency
model. These types of operations are commonly implemented by a Scheduler that performs complex
business processes and workflows.
Asynchronous Messaging Primer. The components in the Scheduler Agent Supervisor pattern typically run
decoupled from each other and communicate asynchronously. Describes some of the approaches that can be
used to implement asynchronous communication based on message queues.
Leader Election pattern. It might be necessary to coordinate the actions of multiple instances of a Supervisor
to prevent them from attempting to recover the same failed process. The Leader Election pattern describes
how to do this.
Cloud Architecture: The Scheduler-Agent-Supervisor Pattern on Clemens Vasters' blog
Process Manager pattern
Reference 6: A Saga on Sagas. An example showing how the CQRS pattern uses a process manager (part of
the CQRS Journey guidance).
Microsoft Azure Scheduler
 Sequential Convoy pattern
11/2/2020 • 3 minutes to read • Edit Online

Process a set of related messages in a defined order, without blocking processing of other groups of messages.

Context and problem

Applications often need to process a sequence of messages in the order they arrive, while still being able to scale
out to handle increased load. In a distributed architecture, processing these messages in order is not
straightforward, because the workers can scale independently and often pull messages independently, using a
Competing Consumers pattern.
For example, an order tracking system receives a ledger containing orders and the relevant operations on those
orders. These operations could be to create an order, add a transaction to the order, modify a past transaction, or
delete an order. In this system, operations must be performed in a first-in-first-out (FIFO) manner, but only at the
order level. However, the initial queue receives a ledger containing transactions for many orders, which may be
interleaved.

Solution
Push related messages into categories within the queuing system, and have the queue listeners lock and pull only
from one category, one message at a time.
Here's what the general Sequential Convoy pattern looks like:

In the queue, messages for different categories may be interleaved, as shown in the following diagram:

Issues and considerations

Consider the following points when deciding how to implement this pattern:
 Category/scale unit. What property of your incoming messages can you scale out on? In the order tracking
scenario, this property is the order ID.
Throughput. What is your target message throughput? If it is very high, you may need to reconsider your FIFO
requirements. For example, can you enforce a start/end message, sort by time, then send a batch for
processing?
Service capabilities. Does your choice of message bus allow for one-at-a-time processing of messages within a
queue or category of a queue?
Evolvability. How will you add a new category of message to the system? For example, suppose the ledger
system described above is specific one customer. If you needed to onboard a new customer, could you have a
set of ledger processors that distribute work per customer ID?
It's possible that consumers might receive a message out of order, due to variable network latency when
sending messages. Consider using sequence numbers to verify ordering. You might also include a special "end
of sequence" flag in the last message of a transaction. Stream processing technologies such as Spark or Azure
Stream Analytics can process messages in order within a time window.

When to use this pattern

Use this pattern when:
You have messages that arrive in order and must be processed in the same order.
Arriving messages are or can be "categorized" in such a way that the category becomes a unit of scale for the
system.
This pattern might not be suitable for:
Extremely high throughput scenarios (millions of messages/minute or second), as the FIFO requirement limits
the scaling that can be done by the system.

Example
On Azure, this pattern can be implemented using Azure Service Bus message sessions. For the consumers, you can
use either Logic Apps with the Service Bus peek-lock connector or Azure Functions with the Service Bus trigger.
For the previous order-tracking example, process each ledger message in the order it's received, and send each
transaction to another queue where the category is set to the order ID. A transaction will never span multiple
orders in this scenario, so consumers process each category in parallel but FIFO within the category.
The ledge processor fan outs the messages by de-batching the content of each message in the first queue:

The ledger processor takes care of:

1. Walking the ledger one transaction at a time.
2. Setting the session ID of the message to match the order ID.
3. Sending each ledger transaction to a secondary queue with the session ID set to the order ID.
The consumers listen to the secondary queue where they process all messages with matching order IDs in order
from the queue. Consumers use peek-lock mode.
When considering scalability, the ledger queue is a primary bottleneck. Different transactions posted to the ledger
could reference the same order ID. However, messages can fan out after the ledger to the number of orders in a
serverless environment.

Next steps
The following information may be relevant when implementing this pattern:
Message sessions: first in, first out (FIFO)
Peek-Lock Message (Non-Destructive Read)
In order delivery of correlated messages in Logic Apps by using Service Bus sessions (MSDN blog)
 Sharding pattern
12/18/2020 • 19 minutes to read • Edit Online

Divide a data store into a set of horizontal partitions or shards. This can improve scalability when storing and
accessing large volumes of data.

Context and problem

A data store hosted by a single server might be subject to the following limitations:
Storage space . A data store for a large-scale cloud application is expected to contain a huge volume of
data that could increase significantly over time. A server typically provides only a finite amount of disk
storage, but you can replace existing disks with larger ones, or add further disks to a machine as data
volumes grow. However, the system will eventually reach a limit where it isn't possible to easily increase
the storage capacity on a given server.
Computing resources . A cloud application is required to support a large number of concurrent users,
each of which run queries that retrieve information from the data store. A single server hosting the data
store might not be able to provide the necessary computing power to support this load, resulting in
extended response times for users and frequent failures as applications attempting to store and retrieve
data time out. It might be possible to add memory or upgrade processors, but the system will reach a
limit when it isn't possible to increase the compute resources any further.
Network bandwidth . Ultimately, the performance of a data store running on a single server is governed
by the rate the server can receive requests and send replies. It's possible that the volume of network
traffic might exceed the capacity of the network used to connect to the server, resulting in failed requests.
Geography . It might be necessary to store data generated by specific users in the same region as those
users for legal, compliance, or performance reasons, or to reduce latency of data access. If the users are
dispersed across different countries or regions, it might not be possible to store the entire data for the
application in a single data store.
Scaling vertically by adding more disk capacity, processing power, memory, and network connections can
postpone the effects of some of these limitations, but it's likely to only be a temporary solution. A commercial
cloud application capable of supporting large numbers of users and high volumes of data must be able to scale
almost indefinitely, so vertical scaling isn't necessarily the best solution.

Solution
Divide the data store into horizontal partitions or shards. Each shard has the same schema, but holds its own
distinct subset of the data. A shard is a data store in its own right (it can contain the data for many entities of
different types), running on a server acting as a storage node.
This pattern has the following benefits:
You can scale the system out by adding further shards running on additional storage nodes.
A system can use off-the-shelf hardware rather than specialized and expensive computers for each
storage node.
You can reduce contention and improve performance by balancing the workload across shards.
In the cloud, shards can be located physically close to the users that'll access the data.
When dividing a data store up into shards, decide which data should be placed in each shard. A shard typically
contains items that fall within a specified range determined by one or more attributes of the data. These
attributes form the shard key (sometimes referred to as the partition key). The shard key should be static. It
shouldn't be based on data that might change.
Sharding physically organizes the data. When an application stores and retrieves data, the sharding logic directs
the application to the appropriate shard. This sharding logic can be implemented as part of the data access code
in the application, or it could be implemented by the data storage system if it transparently supports sharding.
Abstracting the physical location of the data in the sharding logic provides a high level of control over which
shards contain which data. It also enables data to migrate between shards without reworking the business logic
of an application if the data in the shards need to be redistributed later (for example, if the shards become
unbalanced). The tradeoff is the additional data access overhead required in determining the location of each
data item as it's retrieved.
To ensure optimal performance and scalability, it's important to split the data in a way that's appropriate for the
types of queries that the application performs. In many cases, it's unlikely that the sharding scheme will exactly
match the requirements of every query. For example, in a multi-tenant system an application might need to
retrieve tenant data using the tenant ID, but it might also need to look up this data based on some other
attribute such as the tenant’s name or location. To handle these situations, implement a sharding strategy with a
shard key that supports the most commonly performed queries.
If queries regularly retrieve data using a combination of attribute values, you can likely define a composite shard
key by linking attributes together. Alternatively, use a pattern such as Index Table to provide fast lookup to data
based on attributes that aren't covered by the shard key.

Sharding strategies
Three strategies are commonly used when selecting the shard key and deciding how to distribute data across
shards. Note that there doesn't have to be a one-to-one correspondence between shards and the servers that
host them—a single server can host multiple shards. The strategies are:
The Lookup strategy . In this strategy the sharding logic implements a map that routes a request for data to
the shard that contains that data using the shard key. In a multi-tenant application all the data for a tenant might
be stored together in a shard using the tenant ID as the shard key. Multiple tenants might share the same shard,
but the data for a single tenant won't be spread across multiple shards. The figure illustrates sharding tenant
data based on tenant IDs.
The mapping between the shard key and the physical storage can be based on physical shards where each shard
key maps to a physical partition. Alternatively, a more flexible technique for rebalancing shards is virtual
partitioning, where shard keys map to the same number of virtual shards, which in turn map to fewer physical
partitions. In this approach, an application locates data using a shard key that refers to a virtual shard, and the
system transparently maps virtual shards to physical partitions. The mapping between a virtual shard and a
physical partition can change without requiring the application code be modified to use a different set of shard
keys.
The Range strategy . This strategy groups related items together in the same shard, and orders them by shard
key—the shard keys are sequential. It's useful for applications that frequently retrieve sets of items using range
queries (queries that return a set of data items for a shard key that falls within a given range). For example, if an
application regularly needs to find all orders placed in a given month, this data can be retrieved more quickly if
all orders for a month are stored in date and time order in the same shard. If each order was stored in a different
shard, they'd have to be fetched individually by performing a large number of point queries (queries that return
a single data item). The next figure illustrates storing sequential sets (ranges) of data in shard.
In this example, the shard key is a composite key containing the order month as the most significant element,
followed by the order day and the time. The data for orders is naturally sorted when new orders are created and
added to a shard. Some data stores support two-part shard keys containing a partition key element that
identifies the shard and a row key that uniquely identifies an item in the shard. Data is usually held in row key
order in the shard. Items that are subject to range queries and need to be grouped together can use a shard key
that has the same value for the partition key but a unique value for the row key.
The Hash strategy . The purpose of this strategy is to reduce the chance of hotspots (shards that receive a
disproportionate amount of load). It distributes the data across the shards in a way that achieves a balance
between the size of each shard and the average load that each shard will encounter. The sharding logic
computes the shard to store an item in based on a hash of one or more attributes of the data. The chosen
hashing function should distribute data evenly across the shards, possibly by introducing some random element
into the computation. The next figure illustrates sharding tenant data based on a hash of tenant IDs.
To understand the advantage of the Hash strategy over other sharding strategies, consider how a multi-tenant
application that enrolls new tenants sequentially might assign the tenants to shards in the data store. When
using the Range strategy, the data for tenants 1 to n will all be stored in shard A, the data for tenants n+1 to m
will all be stored in shard B, and so on. If the most recently registered tenants are also the most active, most data
activity will occur in a small number of shards, which could cause hotspots. In contrast, the Hash strategy
allocates tenants to shards based on a hash of their tenant ID. This means that sequential tenants are most likely
to be allocated to different shards, which will distribute the load across them. The previous figure shows this for
tenants 55 and 56.
The three sharding strategies have the following advantages and considerations:
Lookup . This offers more control over the way that shards are configured and used. Using virtual shards
reduces the impact when rebalancing data because new physical partitions can be added to even out the
workload. The mapping between a virtual shard and the physical partitions that implement the shard can
be modified without affecting application code that uses a shard key to store and retrieve data. Looking
up shard locations can impose an additional overhead.
Range . This is easy to implement and works well with range queries because they can often fetch
multiple data items from a single shard in a single operation. This strategy offers easier data
management. For example, if users in the same region are in the same shard, updates can be scheduled
in each time zone based on the local load and demand pattern. However, this strategy doesn't provide
optimal balancing between shards. Rebalancing shards is difficult and might not resolve the problem of
uneven load if the majority of activity is for adjacent shard keys.
Hash . This strategy offers a better chance of more even data and load distribution. Request routing can
be accomplished directly by using the hash function. There's no need to maintain a map. Note that
computing the hash might impose an additional overhead. Also, rebalancing shards is difficult.
Most common sharding systems implement one of the approaches described above, but you should also
consider the business requirements of your applications and their patterns of data usage. For example, in a
multi-tenant application:
You can shard data based on workload. You could segregate the data for highly volatile tenants in
separate shards. The speed of data access for other tenants might be improved as a result.
You can shard data based on the location of tenants. You can take the data for tenants in a specific
geographic region offline for backup and maintenance during off-peak hours in that region, while the
data for tenants in other regions remains online and accessible during their business hours.
High-value tenants could be assigned their own private, high performing, lightly loaded shards, whereas
lower-value tenants might be expected to share more densely-packed, busy shards.
The data for tenants that need a high degree of data isolation and privacy can be stored on a completely
separate server.

Scaling and data movement operations

Each of the sharding strategies implies different capabilities and levels of complexity for managing scale in, scale
out, data movement, and maintaining state.
The Lookup strategy permits scaling and data movement operations to be carried out at the user level, either
online or offline. The technique is to suspend some or all user activity (perhaps during off-peak periods), move
the data to the new virtual partition or physical shard, change the mappings, invalidate or refresh any caches
that hold this data, and then allow user activity to resume. Often this type of operation can be centrally
managed. The Lookup strategy requires state to be highly cacheable and replica friendly.
The Range strategy imposes some limitations on scaling and data movement operations, which must typically
be carried out when a part or all of the data store is offline because the data must be split and merged across
the shards. Moving the data to rebalance shards might not resolve the problem of uneven load if the majority of
activity is for adjacent shard keys or data identifiers that are within the same range. The Range strategy might
also require some state to be maintained in order to map ranges to the physical partitions.
The Hash strategy makes scaling and data movement operations more complex because the partition keys are
hashes of the shard keys or data identifiers. The new location of each shard must be determined from the hash
function, or the function modified to provide the correct mappings. However, the Hash strategy doesn't require
maintenance of state.

Issues and considerations

Consider the following points when deciding how to implement this pattern:
Sharding is complementary to other forms of partitioning, such as vertical partitioning and functional
partitioning. For example, a single shard can contain entities that have been partitioned vertically, and a
functional partition can be implemented as multiple shards. For more information about partitioning, see
the Data Partitioning Guidance.
Keep shards balanced so they all handle a similar volume of I/O. As data is inserted and deleted, it's
necessary to periodically rebalance the shards to guarantee an even distribution and to reduce the
chance of hotspots. Rebalancing can be an expensive operation. To reduce the necessity of rebalancing,
plan for growth by ensuring that each shard contains sufficient free space to handle the expected volume
of changes. You should also develop strategies and scripts you can use to quickly rebalance shards if this
becomes necessary.
Use stable data for the shard key. If the shard key changes, the corresponding data item might have to
move between shards, increasing the amount of work performed by update operations. For this reason,
avoid basing the shard key on potentially volatile information. Instead, look for attributes that are
invariant or that naturally form a key.
Ensure that shard keys are unique. For example, avoid using autoincrementing fields as the shard key. Is
some systems, autoincremented fields can't be coordinated across shards, possibly resulting in items in
different shards having the same shard key.

Autoincremented values in other fields that are not shard keys can also cause problems. For example,
if you use autoincremented fields to generate unique IDs, then two different items located in different
shards might be assigned the same ID.

It might not be possible to design a shard key that matches the requirements of every possible query
against the data. Shard the data to support the most frequently performed queries, and if necessary
create secondary index tables to support queries that retrieve data using criteria based on attributes that
aren't part of the shard key. For more information, see the Index Table pattern.
Queries that access only a single shard are more efficient than those that retrieve data from multiple
shards, so avoid implementing a sharding system that results in applications performing large numbers
of queries that join data held in different shards. Remember that a single shard can contain the data for
multiple types of entities. Consider denormalizing your data to keep related entities that are commonly
queried together (such as the details of customers and the orders that they have placed) in the same
shard to reduce the number of separate reads that an application performs.

If an entity in one shard references an entity stored in another shard, include the shard key for the
second entity as part of the schema for the first entity. This can help to improve the performance of
queries that reference related data across shards.
 If an application must perform queries that retrieve data from multiple shards, it might be possible to
fetch this data by using parallel tasks. Examples include fan-out queries, where data from multiple shards
is retrieved in parallel and then aggregated into a single result. However, this approach inevitably adds
some complexity to the data access logic of a solution.
For many applications, creating a larger number of small shards can be more efficient than having a
small number of large shards because they can offer increased opportunities for load balancing. This can
also be useful if you anticipate the need to migrate shards from one physical location to another. Moving
a small shard is quicker than moving a large one.
Make sure the resources available to each shard storage node are sufficient to handle the scalability
requirements in terms of data size and throughput. For more information, see the section “Designing
Partitions for Scalability” in the Data Partitioning Guidance.
Consider replicating reference data to all shards. If an operation that retrieves data from a shard also
references static or slow-moving data as part of the same query, add this data to the shard. The
application can then fetch all of the data for the query easily, without having to make an additional round
trip to a separate data store.

If reference data held in multiple shards changes, the system must synchronize these changes across
all shards. The system can experience a degree of inconsistency while this synchronization occurs. If
you do this, you should design your applications to be able to handle it.

It can be difficult to maintain referential integrity and consistency between shards, so you should
minimize operations that affect data in multiple shards. If an application must modify data across shards,
evaluate whether complete data consistency is actually required. Instead, a common approach in the
cloud is to implement eventual consistency. The data in each partition is updated separately, and the
application logic must take responsibility for ensuring that the updates all complete successfully, as well
as handling the inconsistencies that can arise from querying data while an eventually consistent
operation is running. For more information about implementing eventual consistency, see the Data
Consistency Primer.
Configuring and managing a large number of shards can be a challenge. Tasks such as monitoring,
backing up, checking for consistency, and logging or auditing must be accomplished on multiple shards
and servers, possibly held in multiple locations. These tasks are likely to be implemented using scripts or
other automation solutions, but that might not completely eliminate the additional administrative
requirements.
Shards can be geolocated so that the data that they contain is close to the instances of an application that
use it. This approach can considerably improve performance, but requires additional consideration for
tasks that must access multiple shards in different locations.

When to use this pattern

Use this pattern when a data store is likely to need to scale beyond the resources available to a single storage
node, or to improve performance by reducing contention in a data store.

NOTE
The primary focus of sharding is to improve the performance and scalability of a system, but as a by-product it can also
improve availability due to how the data is divided into separate partitions. A failure in one partition doesn't necessarily
prevent an application from accessing data held in other partitions, and an operator can perform maintenance or
recovery of one or more partitions without making the entire data for an application inaccessible. For more information,
see the Data Partitioning Guidance.
Example
The following example in C# uses a set of SQL Server databases acting as shards. Each database holds a subset
of the data used by an application. The application retrieves data that's distributed across the shards using its
own sharding logic (this is an example of a fan-out query). The details of the data that's located in each shard is
returned by a method called GetShards . This method returns an enumerable list of ShardInformation objects,
where the ShardInformation type contains an identifier for each shard and the SQL Server connection string
that an application should use to connect to the shard (the connection strings aren't shown in the code example).

private IEnumerable<ShardInformation> GetShards()

{
// This retrieves the connection information from a shard store
// (commonly a root database).
return new[]
{
new ShardInformation
{
Id = 1,
ConnectionString = ...
},
new ShardInformation
{
Id = 2,
ConnectionString = ...
}
};
}

The code below shows how the application uses the list of ShardInformation objects to perform a query that
fetches data from each shard in parallel. The details of the query aren't shown, but in this example the data that's
retrieved contains a string that could hold information such as the name of a customer if the shards contain the
details of customers. The results are aggregated into a ConcurrentBag collection for processing by the
application.
 // Retrieve the shards as a ShardInformation[] instance.
var shards = GetShards();

var results = new ConcurrentBag<string>();

// Execute the query against each shard in the shard list.

// This list would typically be retrieved from configuration
// or from a root/primary shard store.
Parallel.ForEach(shards, shard =>
{
// NOTE: Transient fault handling isn't included,
// but should be incorporated when used in a real world application.
using (var con = new SqlConnection(shard.ConnectionString))
{
con.Open();
var cmd = new SqlCommand("SELECT ... FROM ...", con);

Trace.TraceInformation("Executing command against shard: {0}", shard.Id);

var reader = cmd.ExecuteReader();

// Read the results in to a thread-safe data structure.
while (reader.Read())
{
results.Add(reader.GetString(0));
}
}
});

Trace.TraceInformation("Fanout query complete - Record Count: {0}",

results.Count);

Related patterns and guidance

The following patterns and guidance might also be relevant when implementing this pattern:
Data Consistency Primer. It might be necessary to maintain consistency for data distributed across different
shards. Summarizes the issues surrounding maintaining consistency over distributed data, and describes the
benefits and tradeoffs of different consistency models.
Data Partitioning Guidance. Sharding a data store can introduce a range of additional issues. Describes these
issues in relation to partitioning data stores in the cloud to improve scalability, reduce contention, and
optimize performance.
Index Table pattern. Sometimes it isn't possible to completely support queries just through the design of the
shard key. Enables an application to quickly retrieve data from a large data store by specifying a key other
than the shard key.
Materialized View pattern. To maintain the performance of some query operations, it's useful to create
materialized views that aggregate and summarize data, especially if this summary data is based on
information that's distributed across shards. Describes how to generate and populate these views.
 Sidecar pattern
12/18/2020 • 5 minutes to read • Edit Online

Deploy components of an application into a separate process or container to provide isolation and encapsulation.
This pattern can also enable applications to be composed of heterogeneous components and technologies.
This pattern is named Sidecar because it resembles a sidecar attached to a motorcycle. In the pattern, the sidecar
is attached to a parent application and provides supporting features for the application. The sidecar also shares
the same lifecycle as the parent application, being created and retired alongside the parent. The sidecar pattern is
sometimes referred to as the sidekick pattern and is a decomposition pattern.

Context and Problem

Applications and services often require related functionality, such as monitoring, logging, configuration, and
networking services. These peripheral tasks can be implemented as separate components or services.
If they are tightly integrated into the application, they can run in the same process as the application, making
efficient use of shared resources. However, this also means they are not well isolated, and an outage in one of
these components can affect other components or the entire application. Also, they usually need to be
implemented using the same language as the parent application. As a result, the component and the application
have close interdependence on each other.
If the application is decomposed into services, then each service can be built using different languages and
technologies. While this gives more flexibility, it means that each component has its own dependencies and
requires language-specific libraries to access the underlying platform and any resources shared with the parent
application. In addition, deploying these features as separate services can add latency to the application.
Managing the code and dependencies for these language-specific interfaces can also add considerable
complexity, especially for hosting, deployment, and management.

Solution
Co-locate a cohesive set of tasks with the primary application, but place them inside their own process or
container, providing a homogeneous interface for platform services across languages.

A sidecar service is not necessarily part of the application, but is connected to it. It goes wherever the parent
application goes. Sidecars are supporting processes or services that are deployed with the primary application.
On a motorcycle, the sidecar is attached to one motorcycle, and each motorcycle can have its own sidecar. In the
same way, a sidecar service shares the fate of its parent application. For each instance of the application, an
instance of the sidecar is deployed and hosted alongside it.
Advantages of using a sidecar pattern include:
A sidecar is independent from its primary application in terms of runtime environment and programming
language, so you don't need to develop one sidecar per language.
The sidecar can access the same resources as the primary application. For example, a sidecar can monitor
system resources used by both the sidecar and the primary application.
Because of its proximity to the primary application, there’s no significant latency when communicating
between them.
Even for applications that don’t provide an extensibility mechanism, you can use a sidecar to extend
functionality by attaching it as its own process in the same host or sub-container as the primary
application.
The sidecar pattern is often used with containers and referred to as a sidecar container or sidekick container.

Issues and Considerations

Consider the deployment and packaging format you will use to deploy services, processes, or containers.
Containers are particularly well suited to the sidecar pattern.
When designing a sidecar service, carefully decide on the interprocess communication mechanism. Try to use
language- or framework-agnostic technologies unless performance requirements make that impractical.
Before putting functionality into a sidecar, consider whether it would work better as a separate service or a
more traditional daemon.
Also consider whether the functionality could be implemented as a library or using a traditional extension
mechanism. Language-specific libraries may have a deeper level of integration and less network overhead.

When to Use this Pattern

Use this pattern when:
Your primary application uses a heterogeneous set of languages and frameworks. A component located in a
sidecar service can be consumed by applications written in different languages using different frameworks.
A component is owned by a remote team or a different organization.
A component or feature must be co-located on the same host as the application
You need a service that shares the overall lifecycle of your main application, but can be independently updated.
You need fine-grained control over resource limits for a particular resource or component. For example, you
may want to restrict the amount of memory a specific component uses. You can deploy the component as a
sidecar and manage memory usage independently of the main application.
This pattern may not be suitable:
When interprocess communication needs to be optimized. Communication between a parent application and
sidecar services includes some overhead, notably latency in the calls. This may not be an acceptable trade-off
for chatty interfaces.
For small applications where the resource cost of deploying a sidecar service for each instance is not worth
the advantage of isolation.
When the service needs to scale differently than or independently from the main applications. If so, it may be
better to deploy the feature as a separate service.
Example
The sidecar pattern is applicable to many scenarios. Some common examples:
Infrastructure API. The infrastructure development team creates a service that's deployed alongside each
application, instead of a language-specific client library to access the infrastructure. The service is loaded as a
sidecar and provides a common layer for infrastructure services, including logging, environment data,
configuration store, discovery, health checks, and watchdog services. The sidecar also monitors the parent
application's host environment and process (or container) and logs the information to a centralized service.
Manage NGINX/HAProxy. Deploy NGINX with a sidecar service that monitors environment state, then updates
the NGINX configuration file and recycles the process when a change in state is needed.
Ambassador sidecar. Deploy an ambassador service as a sidecar. The application calls through the ambassador,
which handles request logging, routing, circuit breaking, and other connectivity related features.
Offload proxy. Place an NGINX proxy in front of a node.js service instance, to handle serving static file content
for the service.

Related guidance
Ambassador pattern
 Static Content Hosting pattern
12/18/2020 • 5 minutes to read • Edit Online

Deploy static content to a cloud-based storage service that can deliver them directly to the client. This can reduce
the need for potentially expensive compute instances.

Context and problem

Web applications typically include some elements of static content. This static content might include HTML pages
and other resources such as images and documents that are available to the client, either as part of an HTML page
(such as inline images, style sheets, and client-side JavaScript files) or as separate downloads (such as PDF
documents).
Although web servers are optimized for dynamic rendering and output caching, they still have to handle requests
to download static content. This consumes processing cycles that could often be put to better use.

Solution
In most cloud hosting environments, you can put some of an application's resources and static pages in a storage
service. The storage service can serve requests for these resources, reducing load on the compute resources that
handle other web requests. The cost for cloud-hosted storage is typically much less than for compute instances.
When hosting some parts of an application in a storage service, the main considerations are related to
deployment of the application and to securing resources that aren't intended to be available to anonymous users.

Issues and considerations

Consider the following points when deciding how to implement this pattern:
The hosted storage service must expose an HTTP endpoint that users can access to download the static
resources. Some storage services also support HTTPS, so it's possible to host resources in storage services
that require SSL.
For maximum performance and availability, consider using a content delivery network (CDN) to cache the
contents of the storage container in multiple datacenters around the world. However, you'll likely have to
pay for using the CDN.
Storage accounts are often geo-replicated by default to provide resiliency against events that might affect a
datacenter. This means that the IP address might change, but the URL will remain the same.
When some content is located in a storage account and other content is in a hosted compute instance, it
becomes more challenging to deploy and update the application. You might have to perform separate
deployments, and version the application and content to manage it more easily—especially when the static
content includes script files or UI components. However, if only static resources have to be updated, they
can simply be uploaded to the storage account without needing to redeploy the application package.
Storage services might not support the use of custom domain names. In this case it's necessary to specify
the full URL of the resources in links because they'll be in a different domain from the dynamically-
generated content containing the links.
The storage containers must be configured for public read access, but it's vital to ensure that they aren't
configured for public write access to prevent users being able to upload content.
 Consider using a valet key or token to control access to resources that shouldn't be available anonymously.
See the Valet Key pattern for more information.

When to use this pattern

This pattern is useful for:
Minimizing the hosting cost for websites and applications that contain some static resources.
Minimizing the hosting cost for websites that consist of only static content and resources. Depending on
the capabilities of the hosting provider's storage system, it might be possible to entirely host a fully static
website in a storage account.
Exposing static resources and content for applications running in other hosting environments or on-
premises servers.
Locating content in more than one geographical area using a content delivery network that caches the
contents of the storage account in multiple datacenters around the world.
Monitoring costs and bandwidth usage. Using a separate storage account for some or all of the static
content allows the costs to be more easily separated from hosting and runtime costs.
This pattern might not be useful in the following situations:
The application needs to perform some processing on the static content before delivering it to the client.
For example, it might be necessary to add a timestamp to a document.
The volume of static content is very small. The overhead of retrieving this content from separate storage
can outweigh the cost benefit of separating it out from the compute resource.

Example
Azure Storage supports serving static content directly from a storage container. Files are served through
anonymous access requests. By default, files have a URL in a subdomain of core.windows.net , such as
https://contoso.z4.web.core.windows.net/image.png . You can configure a custom domain name, and use Azure
CDN to access the files over HTTPS. For more information, see Static website hosting in Azure Storage.

Static website hosting makes the files available for anonymous access. If you need to control who can access the
files, you can store files in Azure blob storage and then generate shared access signatures to limit access.
The links in the pages delivered to the client must specify the full URL of the resource. If the resource is protected
with a valet key, such as a shared access signature, this signature must be included in the URL.
A sample application that demonstrates using external storage for static resources is available on GitHub. This
sample uses configuration files to specify the storage account and container that holds the static content.

<Setting name="StaticContent.StorageConnectionString"
value="UseDevelopmentStorage=true" />
<Setting name="StaticContent.Container" value="static-content" />

The Settings class in the file Settings.cs of the StaticContentHosting.Web project contains methods to extract
these values and build a string value containing the cloud storage account container URL.

public class Settings

{
public static string StaticContentStorageConnectionString {
get
{
return RoleEnvironment.GetConfigurationSettingValue(
"StaticContent.StorageConnectionString");
}
}

public static string StaticContentContainer

{
get
{
return RoleEnvironment.GetConfigurationSettingValue("StaticContent.Container");
}
}

public static string StaticContentBaseUrl

{
get
{
var blobServiceClient = new BlobServiceClient(StaticContentStorageConnectionString);

return string.Format("{0}/{1}", blobServiceClient.Uri.ToString().TrimEnd('/'),

StaticContentContainer.TrimStart('/'));
}
}
}

The StaticContentUrlHtmlHelperclass in the file StaticContentUrlHtmlHelper.cs exposes a method named

StaticContentUrl that generates a URL containing the path to the cloud storage account if the URL passed to it
starts with the ASP.NET root path character (~).

public static class StaticContentUrlHtmlHelper

{
public static string StaticContentUrl(this HtmlHelper helper, string contentPath)
{
if (contentPath.StartsWith("~"))
{
contentPath = contentPath.Substring(1);
}

contentPath = string.Format("{0}/{1}", Settings.StaticContentBaseUrl.TrimEnd('/'),

contentPath.TrimStart('/'));

var url = new UrlHelper(helper.ViewContext.RequestContext);

return url.Content(contentPath);
}
}

The file Index.cshtml in the Views\Home folder contains an image element that uses the StaticContentUrl method
to create the URL for its src attribute.

<img src="@Html.StaticContentUrl("~/media/orderedList1.png")" alt="Test Image" />

Related patterns and guidance

Static Content Hosting sample. A sample application that demonstrates this pattern.
Valet Key pattern. If the target resources aren't supposed to be available to anonymous users, use this pattern
to restrict direct access.
Serverless web application on Azure. A reference architecture that uses static website hosting with Azure
Functions to implement a serverless web app.
 Strangler Fig pattern
12/18/2020 • 2 minutes to read • Edit Online

Incrementally migrate a legacy system by gradually replacing specific pieces of functionality with new applications
and services. As features from the legacy system are replaced, the new system eventually replaces all of the old
system's features, strangling the old system and allowing you to decommission it.

Context and problem

As systems age, the development tools, hosting technology, and even system architectures they were built on can
become increasingly obsolete. As new features and functionality are added, the complexity of these applications
can increase dramatically, making them harder to maintain or add new features to.
Completely replacing a complex system can be a huge undertaking. Often, you will need a gradual migration to a
new system, while keeping the old system to handle features that haven't been migrated yet. However, running
two separate versions of an application means that clients have to know where particular features are located.
Every time a feature or service is migrated, clients need to be updated to point to the new location.

Solution
Incrementally replace specific pieces of functionality with new applications and services. Create a façade that
intercepts requests going to the backend legacy system. The façade routes these requests either to the legacy
application or the new services. Existing features can be migrated to the new system gradually, and consumers
can continue using the same interface, unaware that any migration has taken place.

This pattern helps to minimize risk from the migration, and spread the development effort over time. With the
façade safely routing users to the correct application, you can add functionality to the new system at whatever
pace you like, while ensuring the legacy application continues to function. Over time, as features are migrated to
the new system, the legacy system is eventually "strangled" and is no longer necessary. Once this process is
complete, the legacy system can safely be retired.

Issues and considerations

Consider how to handle services and data stores that are potentially used by both new and legacy systems.
Make sure both can access these resources side-by-side.
Structure new applications and services in a way that they can easily be intercepted and replaced in future
strangler fig migrations.
At some point, when the migration is complete, the strangler fig façade will either go away or evolve into an
adaptor for legacy clients.
 Make sure the façade keeps up with the migration.
Make sure the façade doesn't become a single point of failure or a performance bottleneck.

When to use this pattern

Use this pattern when gradually migrating a back-end application to a new architecture.
This pattern may not be suitable:
When requests to the back-end system cannot be intercepted.
For smaller systems where the complexity of wholesale replacement is low.

Related guidance
Martin Fowler's blog post on StranglerApplication
 Throttling pattern
12/18/2020 • 7 minutes to read • Edit Online

Control the consumption of resources used by an instance of an application, an individual tenant, or an entire
service. This can allow the system to continue to function and meet service level agreements, even when an
increase in demand places an extreme load on resources.

Context and problem

The load on a cloud application typically varies over time based on the number of active users or the types of
activities they are performing. For example, more users are likely to be active during business hours, or the
system might be required to perform computationally expensive analytics at the end of each month. There might
also be sudden and unanticipated bursts in activity. If the processing requirements of the system exceed the
capacity of the resources that are available, it'll suffer from poor performance and can even fail. If the system has
to meet an agreed level of service, such failure could be unacceptable.
There're many strategies available for handling varying load in the cloud, depending on the business goals for
the application. One strategy is to use autoscaling to match the provisioned resources to the user needs at any
given time. This has the potential to consistently meet user demand, while optimizing running costs. However,
while autoscaling can trigger the provisioning of additional resources, this provisioning isn't immediate. If
demand grows quickly, there can be a window of time where there's a resource deficit.

Solution
An alternative strategy to autoscaling is to allow applications to use resources only up to a limit, and then
throttle them when this limit is reached. The system should monitor how it's using resources so that, when
usage exceeds the threshold, it can throttle requests from one or more users. This will enable the system to
continue functioning and meet any service level agreements (SLAs) that are in place. For more information on
monitoring resource usage, see the Instrumentation and Telemetry Guidance.
The system could implement several throttling strategies, including:
Rejecting requests from an individual user who's already accessed system APIs more than n times per
second over a given period of time. This requires the system to meter the use of resources for each tenant
or user running an application. For more information, see the Service Metering Guidance.
Disabling or degrading the functionality of selected nonessential services so that essential services can
run unimpeded with sufficient resources. For example, if the application is streaming video output, it
could switch to a lower resolution.
Using load leveling to smooth the volume of activity (this approach is covered in more detail by the
Queue-based Load Leveling pattern). In a multi-tenant environment, this approach will reduce the
performance for every tenant. If the system must support a mix of tenants with different SLAs, the work
for high-value tenants might be performed immediately. Requests for other tenants can be held back, and
handled when the backlog has eased. The Priority Queue pattern could be used to help implement this
approach.
Deferring operations being performed on behalf of lower priority applications or tenants. These
operations can be suspended or limited, with an exception generated to inform the tenant that the system
is busy and that the operation should be retried later.
The figure shows an area graph for resource use (a combination of memory, CPU, bandwidth, and other factors)
against time for applications that are making use of three features. A feature is an area of functionality, such as a
component that performs a specific set of tasks, a piece of code that performs a complex calculation, or an
element that provides a service such as an in-memory cache. These features are labeled A, B, and C.

The area immediately below the line for a feature indicates the resources that are used by applications when
they invoke this feature. For example, the area below the line for Feature A shows the resources used by
applications that are making use of Feature A, and the area between the lines for Feature A and Feature B
indicates the resources used by applications invoking Feature B. Aggregating the areas for each feature
shows the total resource use of the system.

The previous figure illustrates the effects of deferring operations. Just prior to time T1, the total resources
allocated to all applications using these features reach a threshold (the limit of resource use). At this point, the
applications are in danger of exhausting the resources available. In this system, Feature B is less critical than
Feature A or Feature C, so it's temporarily disabled and the resources that it was using are released. Between
times T1 and T2, the applications using Feature A and Feature C continue running as normal. Eventually, the
resource use of these two features diminishes to the point when, at time T2, there is sufficient capacity to enable
Feature B again.
The autoscaling and throttling approaches can also be combined to help keep the applications responsive and
within SLAs. If the demand is expected to remain high, throttling provides a temporary solution while the system
scales out. At this point, the full functionality of the system can be restored.
The next figure shows an area graph of the overall resource use by all applications running in a system against
time, and illustrates how throttling can be combined with autoscaling.
At time T1, the threshold specifying the soft limit of resource use is reached. At this point, the system can start to
scale out. However, if the new resources don't become available quickly enough, then the existing resources
might be exhausted and the system could fail. To prevent this from occurring, the system is temporarily throttled,
as described earlier. When autoscaling has completed and the additional resources are available, throttling can
be relaxed.

Issues and considerations

You should consider the following points when deciding how to implement this pattern:
Throttling an application, and the strategy to use, is an architectural decision that impacts the entire
design of a system. Throttling should be considered early in the application design process because it isn't
easy to add once a system has been implemented.
Throttling must be performed quickly. The system must be capable of detecting an increase in activity and
react accordingly. The system must also be able to revert to its original state quickly after the load has
eased. This requires that the appropriate performance data is continually captured and monitored.
If a service needs to temporarily deny a user request, it should return a specific error code so the client
application understands that the reason for the refusal to perform an operation is due to throttling. The
client application can wait for a period before retrying the request.
Throttling can be used as a temporary measure while a system autoscales. In some cases it's better to
simply throttle, rather than to scale, if a burst in activity is sudden and isn't expected to be long lived
because scaling can add considerably to running costs.
If throttling is being used as a temporary measure while a system autoscales, and if resource demands
grow very quickly, the system might not be able to continue functioning—even when operating in a
throttled mode. If this isn't acceptable, consider maintaining larger capacity reserves and configuring
more aggressive autoscaling.

When to use this pattern

Use this pattern:
 To ensure that a system continues to meet service level agreements.
To prevent a single tenant from monopolizing the resources provided by an application.
To handle bursts in activity.
To help cost-optimize a system by limiting the maximum resource levels needed to keep it functioning.

Example
The final figure illustrates how throttling can be implemented in a multi-tenant system. Users from each of the
tenant organizations access a cloud-hosted application where they fill out and submit surveys. The application
contains instrumentation that monitors the rate at which these users are submitting requests to the application.
In order to prevent the users from one tenant affecting the responsiveness and availability of the application for
all other users, a limit is applied to the number of requests per second the users from any one tenant can submit.
The application blocks requests that exceed this limit.

Related patterns and guidance

The following patterns and guidance may also be relevant when implementing this pattern:
Instrumentation and Telemetry Guidance. Throttling depends on gathering information about how heavily a
service is being used. Describes how to generate and capture custom monitoring information.
Service Metering Guidance. Describes how to meter the use of services in order to gain an understanding of
how they are used. This information can be useful in determining how to throttle a service.
Autoscaling Guidance. Throttling can be used as an interim measure while a system autoscales, or to remove
the need for a system to autoscale. Contains information on autoscaling strategies.
Queue-based Load Leveling pattern. Queue-based load leveling is a commonly used mechanism for
implementing throttling. A queue can act as a buffer that helps to even out the rate at which requests sent by
an application are delivered to a service.
Priority Queue pattern. A system can use priority queuing as part of its throttling strategy to maintain
performance for critical or higher value applications, while reducing the performance of less important
applications.
 Valet Key pattern
12/18/2020 • 12 minutes to read • Edit Online

Use a token that provides clients with restricted direct access to a specific resource, in order to offload data
transfer from the application. This is particularly useful in applications that use cloud-hosted storage systems or
queues, and can minimize cost and maximize scalability and performance.

Context and problem

Client programs and web browsers often need to read and write files or data streams to and from an application’s
storage. Typically, the application will handle the movement of the data — either by fetching it from storage and
streaming it to the client, or by reading the uploaded stream from the client and storing it in the data store.
However, this approach absorbs valuable resources such as compute, memory, and bandwidth.
Data stores have the ability to handle upload and download of data directly, without requiring that the application
perform any processing to move this data. But, this typically requires the client to have access to the security
credentials for the store. This can be a useful technique to minimize data transfer costs and the requirement to
scale out the application, and to maximize performance. It means, though, that the application is no longer able to
manage the security of the data. After the client has a connection to the data store for direct access, the
application can't act as the gatekeeper. It's no longer in control of the process and can't prevent subsequent
uploads or downloads from the data store.
This isn't a realistic approach in distributed systems that need to serve untrusted clients. Instead, applications
must be able to securely control access to data in a granular way, but still reduce the load on the server by setting
up this connection and then allowing the client to communicate directly with the data store to perform the
required read or write operations.

Solution
You need to resolve the problem of controlling access to a data store where the store can't manage authentication
and authorization of clients. One typical solution is to restrict access to the data store’s public connection and
provide the client with a key or token that the data store can validate.
This key or token is usually referred to as a valet key. It provides time-limited access to specific resources and
allows only predefined operations such as reading and writing to storage or queues, or uploading and
downloading in a web browser. Applications can create and issue valet keys to client devices and web browsers
quickly and easily, allowing clients to perform the required operations without requiring the application to
directly handle the data transfer. This removes the processing overhead, and the impact on performance and
scalability, from the application and the server.
The client uses this token to access a specific resource in the data store for only a specific period, and with specific
restrictions on access permissions, as shown in the figure. After the specified period, the key becomes invalid and
won't allow access to the resource.
It's also possible to configure a key that has other dependencies, such as the scope of the data. For example,
depending on the data store capabilities, the key can specify a complete table in a data store, or only specific rows
in a table. In cloud storage systems the key can specify a container, or just a specific item within a container.
The key can also be invalidated by the application. This is a useful approach if the client notifies the server that the
data transfer operation is complete. The server can then invalidate that key to prevent further access.
Using this pattern can simplify managing access to resources because there's no requirement to create and
authenticate a user, grant permissions, and then remove the user again. It also makes it easy to limit the location,
the permission, and the validity period—all by simply generating a key at runtime. The important factors are to
limit the validity period, and especially the location of the resource, as tightly as possible so that the recipient can
only use it for the intended purpose.

Issues and considerations

Consider the following points when deciding how to implement this pattern:
Manage the validity status and period of the key . If leaked or compromised, the key effectively unlocks the
target item and makes it available for malicious use during the validity period. A key can usually be revoked or
disabled, depending on how it was issued. Server-side policies can be changed or, the server key it was signed
with can be invalidated. Specify a short validity period to minimize the risk of allowing unauthorized operations
to take place against the data store. However, if the validity period is too short, the client might not be able to
complete the operation before the key expires. Allow authorized users to renew the key before the validity period
expires if multiple accesses to the protected resource are required.
Control the level of access the key will provide . Typically, the key should allow the user to only perform the
actions necessary to complete the operation, such as read-only access if the client shouldn't be able to upload
data to the data store. For file uploads, it's common to specify a key that provides write-only permission, as well
as the location and the validity period. It's critical to accurately specify the resource or the set of resources to
which the key applies.
Consider how to control users’ behavior . Implementing this pattern means some loss of control over the
resources users are granted access to. The level of control that can be exerted is limited by the capabilities of the
policies and permissions available for the service or the target data store. For example, it's usually not possible to
create a key that limits the size of the data to be written to storage, or the number of times the key can be used to
access a file. This can result in huge unexpected costs for data transfer, even when used by the intended client, and
might be caused by an error in the code that causes repeated upload or download. To limit the number of times a
file can be uploaded, where possible, force the client to notify the application when one operation has completed.
For example, some data stores raise events the application code can use to monitor operations and control user
behavior. However, it's hard to enforce quotas for individual users in a multi-tenant scenario where the same key
is used by all the users from one tenant.
Validate, and optionally sanitize, all uploaded data . A malicious user that gains access to the key could
upload data designed to compromise the system. Alternatively, authorized users might upload data that's invalid
and, when processed, could result in an error or system failure. To protect against this, ensure that all uploaded
data is validated and checked for malicious content before use.
Audit all operations . Many key-based mechanisms can log operations such as uploads, downloads, and
failures. These logs can usually be incorporated into an audit process, and also used for billing if the user is
charged based on file size or data volume. Use the logs to detect authentication failures that might be caused by
issues with the key provider, or accidental removal of a stored access policy.
Deliver the key securely . It can be embedded in a URL that the user activates in a web page, or it can be used
in a server redirection operation so that the download occurs automatically. Always use HTTPS to deliver the key
over a secure channel.
Protect sensitive data in transit . Sensitive data delivered through the application will usually take place using
SSL or TLS, and this should be enforced for clients accessing the data store directly.
Other issues to be aware of when implementing this pattern are:
If the client doesn't, or can't, notify the server of completion of the operation, and the only limit is the
expiration period of the key, the application won't be able to perform auditing operations such as counting
the number of uploads or downloads, or preventing multiple uploads or downloads.
The flexibility of key policies that can be generated might be limited. For example, some mechanisms only
allow the use of a timed expiration period. Others aren't able to specify a sufficient granularity of
read/write permissions.
If the start time for the key or token validity period is specified, ensure that it's a little earlier than the
current server time to allow for client clocks that might be slightly out of synchronization. The default, if
not specified, is usually the current server time.
The URL containing the key will be recorded in server log files. While the key will typically have expired
before the log files are used for analysis, ensure that you limit access to them. If log data is transmitted to a
monitoring system or stored in another location, consider implementing a delay to prevent leakage of keys
until after their validity period has expired.
If the client code runs in a web browser, the browser might need to support cross-origin resource sharing
(CORS) to enable code that executes within the web browser to access data in a different domain from the
one that served the page. Some older browsers and some data stores don't support CORS, and code that
runs in these browsers might not be able to use a valet key to provide access to data in a different domain,
such as a cloud storage account.

When to use this pattern

This pattern is useful for the following situations:
To minimize resource loading and maximize performance and scalability. Using a valet key doesn't require
the resource to be locked, no remote server call is required, there's no limit on the number of valet keys
that can be issued, and it avoids a single point of failure resulting from performing the data transfer
through the application code. Creating a valet key is typically a simple cryptographic operation of signing a
string with a key.
To minimize operational cost. Enabling direct access to stores and queues is resource and cost efficient, can
result in fewer network round trips, and might allow for a reduction in the number of compute resources
required.
When clients regularly upload or download data, particularly where there's a large volume or when each
 operation involves large files.
When the application has limited compute resources available, either due to hosting limitations or cost
considerations. In this scenario, the pattern is even more helpful if there are many concurrent data uploads
or downloads because it relieves the application from handling the data transfer.
When the data is stored in a remote data store or a different datacenter. If the application was required to
act as a gatekeeper, there might be a charge for the additional bandwidth of transferring the data between
datacenters, or across public or private networks between the client and the application, and then between
the application and the data store.
This pattern might not be useful in the following situations:
If the application must perform some task on the data before it's stored or before it's sent to the client. For
example, if the application needs to perform validation, log access success, or execute a transformation on
the data. However, some data stores and clients are able to negotiate and carry out simple transformations
such as compression and decompression (for example, a web browser can usually handle GZip formats).
If the design of an existing application makes it difficult to incorporate the pattern. Using this pattern
typically requires a different architectural approach for delivering and receiving data.
If it's necessary to maintain audit trails or control the number of times a data transfer operation is
executed, and the valet key mechanism in use doesn't support notifications that the server can use to
manage these operations.
If it's necessary to limit the size of the data, especially during upload operations. The only solution to this is
for the application to check the data size after the operation is complete, or check the size of uploads after
a specified period or on a scheduled basis.

Example
Azure supports shared access signatures on Azure Storage for granular access control to data in blobs, tables, and
queues, and for Service Bus queues and topics. A shared access signature token can be configured to provide
specific access rights such as read, write, update, and delete to a specific table; a key range within a table; a queue;
a blob; or a blob container. The validity can be a specified time period or with no time limit.
Azure shared access signatures also support server-stored access policies that can be associated with a specific
resource such as a table or blob. This feature provides additional control and flexibility compared to application-
generated shared access signature tokens, and should be used whenever possible. Settings defined in a server-
stored policy can be changed and are reflected in the token without requiring a new token to be issued, but
settings defined in the token can't be changed without issuing a new token. This approach also makes it possible
to revoke a valid shared access signature token before it's expired.

For more information, see Introducing Table SAS (Shared Access Signature), Queue SAS and update to Blob
SAS and Using Shared Access Signatures on MSDN.

The following code shows how to create a shared access signature token that's valid for five minutes. The
GetSharedAccessReferenceForUpload method returns a shared access signatures token that can be used to upload a
file to Azure Blob Storage.
 public class ValuesController : ApiController
{
private readonly BlobServiceClient blobServiceClient;
private readonly string blobContainer;
...
/// <summary>
/// Return a limited access key that allows the caller to upload a file
/// to this specific destination for a defined period of time.
/// </summary>
private StorageEntitySas GetSharedAccessReferenceForUpload(string blobName)
{
var blob = blobServiceClient.GetBlobContainerClient(this.blobContainer).GetBlobClient(blobName);

var storageSharedKeyCredential = new StorageSharedKeyCredential(blobServiceClient.AccountName,

ConfigurationManager.AppSettings["AzureStorageEmulatorAccountKey"]);

var blobSasBuilder = new BlobSasBuilder

{
BlobContainerName = this.blobContainer,
BlobName = blobName,
Resource = "b",
StartsOn = DateTimeOffset.UtcNow.AddMinutes(-5),
ExpiresOn = DateTimeOffset.UtcNow.AddMinutes(5)
};
policy.SetPermissions(BlobSasPermissions.Write);
var sas = policy.ToSasQueryParameters(storageSharedKeyCredential).ToString();

return new StorageEntitySas

{
BlobUri = blob.Uri,
Credentials = sas
};
}
public struct StorageEntitySas
{
public string Credentials;
public Uri BlobUri;
}
}

The complete sample is available in the ValetKey solution available for download from GitHub. The
ValetKey.Web project in this solution contains a web application that includes the ValuesController class
shown above. A sample client application that uses this web application to retrieve a shared access signatures
key and upload a file to blob storage is available in the ValetKey.Client project.

Next steps
The following patterns and guidance might also be relevant when implementing this pattern:
A sample that demonstrates this pattern is available on GitHub.
Gatekeeper pattern. This pattern can be used in conjunction with the Valet Key pattern to protect applications
and services by using a dedicated host instance that acts as a broker between clients and the application or
service. The gatekeeper validates and sanitizes requests, and passes requests and data between the client and
the application. Can provide an additional layer of security, and reduce the attack surface of the system.
Static Content Hosting pattern. Describes how to deploy static resources to a cloud-based storage service that
can deliver these resources directly to the client to reduce the requirement for expensive compute instances.
Where the resources aren't intended to be publicly available, the Valet Key pattern can be used to secure them.
Introducing Table SAS (Shared Access Signature), Queue SAS and update to Blob SAS
Using Shared Access Signatures
Shared Access Signature Authentication with Service Bus
 Artificial intelligence (AI)
12/18/2020 • 15 minutes to read • Edit Online

Artificial intelligence (AI) is the capability of a computer to imitate intelligent human behavior. Through AI, machines
can analyze images, comprehend speech, interact in natural ways, and make predictions using data.

AI concepts
Algorithm
An algorithm is a sequence of calculations and rules used to solve a problem or analyze a set of data. It is like a flow
chart, with step-by-step instructions for questions to ask, but written in math and programming code. An algorithm
may describe how to determine whether a pet is a cat, dog, fish, bird, or lizard. Another far more complicated
algorithm may describe how to identify a written or spoken language, analyze its words, translate them into a
different language, and then check the translation for accuracy.
Machine learning
Machine learning (ML) is an AI technique that uses mathematical algorithms to create predictive models. An
algorithm is used to parse data fields and to "learn" from that data by using patterns found within it to generate
models. Those models are then used to make informed predictions or decisions about new data.
The predictive models are validated against known data, measured by performance metrics selected for specific
business scenarios, and then adjusted as needed. This process of learning and validation is called training. Through
periodic retraining, ML models are improved over time.
Machine learning at scale
What are the machine learning products at Microsoft?
Deep learning
Deep learning is a type of ML that can determine for itself whether its predictions are accurate. It also uses
algorithms to analyze data, but it does so on a larger scale than ML.
Deep learning uses artificial neural networks, which consist of multiple layers of algorithms. Each layer looks at the
incoming data, performs its own specialized analysis, and produces an output that other layers can understand. This
output is then passed to the next layer, where a different algorithm does its own analysis, and so on.
With many layers in each neural network-and sometimes using multiple neural networks-a machine can learn
through its own data processing. This requires much more data and much more computing power than ML.
Deep learning versus machine learning
Distributed training of deep learning models on Azure
Batch scoring of deep learning models on Azure
Training of Python scikit-learn and deep learning models on Azure
Real-time scoring of Python scikit-learn and deep learning models on Azure
Bots
A bot is an automated software program designed to perform a particular task. Think of it as a robot without a
body. Early bots were comparatively simple, handling repetitive and voluminous tasks with relatively
straightforward algorithmic logic. An example would be web crawlers used by search engines to automatically
explore and catalog web content.
Bots have become much more sophisticated, using AI and other technologies to mimic human activity and
decision-making, often while interacting directly with humans through text or even speech. Examples include bots
that can take a dinner reservation, chatbots (or conversational AI) that help with customer service interactions, and
social bots that post breaking news or scientific data to social media sites.
Microsoft offers the Azure Bot Service, a managed service purpose-built for enterprise-grade bot development.
About Azure Bot Service
Ten guidelines for responsible bots
Azure reference architecture: Enterprise-grade conversational bot
Example workload: Conversational chatbot for hotel reservations on Azure
Autonomous systems
Autonomous systems are part of an evolving new class that goes beyond basic automation. Instead of performing
a specific task repeatedly with little or no variation (like bots do), autonomous systems bring intelligence to
machines so they can adapt to changing environments to accomplish a desired goal.
Smart buildings use autonomous systems to automatically control operations like lighting, ventilation, air
conditioning, and security. A more sophisticated example would be a self-directed robot exploring a collapsed mine
shaft to thoroughly map its interior, determine which portions are structurally sound, analyze the air for
breathability, and detect signs of trapped miners in need of rescue-all without a human monitoring in real time on
the remote end.
Autonomous systems and solutions from Microsoft AI
General info on Microsoft AI
Learn more about Microsoft AI, and keep up-to-date with related news:
Microsoft AI School
Azure AI platform page
Microsoft AI platform page
Microsoft AI Blog
 Microsoft AI on GitHub: Samples, reference architectures, and best practices
Azure Architecture Center

High-level architectural types

Prebuilt AI
Prebuilt AI is exactly what it sounds like-off-the-shelf AI models, services, and APIs that are ready to use. These help
you add intelligence to apps, websites, and flows without having to gather data and then build, train, and publish
your own models.
One example of prebuilt AI might be a pretrained model that can be incorporated as is or used to provide a
baseline for further custom training. Another example would be a cloud-based API service that can be called at will
to process natural language in a desired fashion.
Azure Cognitive Services
Cognitive Services provide developers the opportunity to use prebuilt APIs and integration toolkits to create
applications that can see, hear, speak, understand, and even begin to reason. The catalog of services within
Cognitive Services can be categorized into five main pillars: Vision, Speech, Language, Web Search, and
Decision/Recommendation.
Azure Cognitive Services documentation
Try Azure Cognitive Services for free
Choosing a Microsoft cognitive services technology
Choosing a natural language processing technology in Azure
Prebuilt AI models in AI Builder
AI Builder is a new capability in Microsoft Power Platform that provides a point-and-click interface for adding AI to
your apps, even if you have no coding or data science skills. (Some features in AI Builder have not yet released for
general availability and remain in preview status. For more information, refer to the Feature availability by region
page.)
You can build and train your own models, but AI Builder also provides select prebuilt AI models that are ready for
use right away. For example, you can add a component in Microsoft Power Apps based on a prebuilt model that
recognizes contact information from business cards.
Power Apps on Azure
AI Builder documentation
AI model types in AI Builder
Overview of prebuilt AI models in AI Builder
Custom AI
Although prebuilt AI is useful (and increasingly flexible), the best way to get what you need from AI is probably to
build a system yourself. This is obviously a very deep and complex subject, but let's look at some basic concepts
beyond what we've just covered.
Code languages
The core concept of AI is the use of algorithms to analyze data and generate models to describe (or score) it in ways
that are useful. Algorithms are written by developers and data scientists (and sometimes by other algorithms) using
programming code. Two of the most popular programming languages for AI development are currently Python
and R.
Python is a general-purpose, high-level programming language. It has a simple, easy-to-learn syntax that
emphasizes readability. There is no compiling step. Python has a large standard library, but it also supports the
ability to add modules and packages. This encourages modularity and lets you expand capabilities when needed.
There is a large and growing ecosystem of AI and ML libraries for Python, including many that are readily available
in Azure.
Python on Azure product home page
Azure for Python developers
Azure Machine Learning SDK for Python
Introduction to machine learning with Python and Azure Notebooks
scikit-learn. An open-source ML library for Python
PyTorch. An open-source Python library with a rich ecosystem that can be used for deep learning, computer
vision, natural language processing, and more
TensorFlow. An open-source symbolic math library also used for ML applications and neural networks
Tutorial: Apply machine learning models in Azure Functions with Python and TensorFlow
R is a language and environment for statistical computing and graphics. It can be used for everything from
mapping broad social and marketing trends online to developing financial and climate models.
Microsoft has fully embraced the R programming language and provides many different options for R developers
to run their code in Azure.
R developer's guide to Azure
Microsoft R Open. An enhanced distribution of R from Microsoft, fully compatible with R-3.5.3, with
additional capabilities for improved performance and reproducibility, in addition to support for Windows-
and Linux-based platforms
Using R in Azure Machine Learning Studio. A video illustrating how to incorporate your R code in ML studio.
Tutorial: Create a logistic regression model in R with Azure Machine Learning
Training
Training is core to machine learning. It is the iterative process of "teaching" an algorithm to create models, which
are used to analyze data and then make accurate predictions from it. In practice, this process has three general
phases: training, validation, and testing.
During the training phase, a quality set of known data is tagged so that individual fields are identifiable. The tagged
data is fed to an algorithm configured to make a particular prediction. When finished, the algorithm outputs a
model that describes the patterns it found as a set of parameters. During validation, fresh data is tagged and used
to test the model. The algorithm is adjusted as needed and possibly put through more training. Finally, the testing
phase uses real-world data without any tags or preselected targets. Assuming the model's results are accurate, it is
considered ready for use and can be deployed.
Train models with Azure Machine Learning
Hyper par am et er t u n i n g

Hyperparameters are data variables that govern the training process itself. They are configuration variables that
control how the algorithm operates. Hyperparameters are thus typically set before model training begins and are
not modified within the training process in the way that parameters are. Hyperparameter tuning involves running
trials within the training task, assessing how well they are getting the job done, and then adjusting as needed. This
process generates multiple models, each trained using different families of hyperparameters.
Tune hyperparameters for your model with Azure Machine Learning
 Azure Machine Learning Studio (classic): Tune Model Hyperparameters
M o d e l se l e c t i o n

The process of training and hyperparameter tuning produces numerous candidate models. These can have many
different variances, including the effort needed to prepare the data, the flexibility of the model, the amount of
processing time, and of course the degree of accuracy of its results. Choosing the best trained model for your needs
and constraints is called model selection, but this is as much about preplanning before training as it is about
choosing the one that works best.
A u t o m a t e d m a c h i n e l e a r n i n g (A u t o M L )

Automated machine learning, also known as AutoML, is the process of automating the time-consuming, iterative
tasks of machine learning model development. It can significantly reduce the time it takes to get production-ready
ML models. Automated ML can assist with model selection, hyperparameter tuning, model training, and other tasks,
without requiring extensive programming or domain knowledge.
What is automated machine learning?
Scoring
Scoring is also called prediction and is the process of generating values based on a trained machine learning
model, given some new input data. The values, or scores, that are created can represent predictions of future values,
but they might also represent a likely category or outcome. The scoring process can generate many different types
of values:
A list of recommended items and a similarity score
Numeric values, for time series models and regression models
A probability value, indicating the likelihood that a new input belongs to some existing category
The name of a category or cluster to which a new item is most similar
A predicted class or outcome, for classification models
Batch scoring is when data is collected during some fixed period of time and then processed in a batch. This
might include generating business reports or analyzing customer loyalty.
Real-time scoring is exactly that-scoring that is ongoing and performed as quickly as possible. The classic
example is credit card fraud detection, but real-time scoring can also be used in speech recognition, medical
diagnoses, market analyses, and many other applications.
General info on custom AI on Azure
Microsoft AI on GitHub: Samples, reference architectures, and best practices
Custom AI on Azure GitHub repo. A collection of scripts and tutorials to help developers effectively use Azure
for their AI workloads
Azure Machine Learning SDK for Python
Azure Machine Learning service example notebooks (Python). A GitHub repo of example notebooks
demonstrating the Azure Machine Learning Python SDK
Azure Machine Learning SDK for R

Azure AI platform offerings

Following is a breakdown of Azure technologies, platforms, and services you can use to develop AI solutions for
your needs.
Azure Machine Learning
This is an enterprise-grade machine learning service to build and deploy models faster. Azure Machine Learning
offers web interfaces and SDKs so you can quickly train and deploy your machine learning models and pipelines at
scale. Use these capabilities with open-source Python frameworks, such as PyTorch, TensorFlow, and scikit-learn.
What are the machine learning products at Microsoft?
Azure Machine Learning product home page
Azure Machine Learning Data Architecture Guide overview
Azure Machine Learning documentation overview
What is Azure Machine Learning? General orientation with links to many learning resources, SDKs,
documentation, and more
Azure Machine Learning versus Azure Machine Learning Studio (classic)
Machine learning reference architectures for Azure
Training of Python scikit-learn and deep learning models on Azure
Distributed training of deep learning models on Azure
Batch scoring of Python machine learning models on Azure
Batch scoring of deep learning models on Azure
Real-time scoring of Python scikit-learn and deep learning models on Azure
Machine learning operationalization (MLOps) for Python models using Azure Machine Learning
Batch scoring of R machine learning models on Azure
Real-time scoring of R machine learning models on Azure
Batch scoring of Spark machine learning models on Azure Databricks
Enterprise-grade conversational bot
Build a real-time recommendation API on Azure
Azure automated machine learning
Azure provides extensive support for automated ML. Developers can build models using a no-code UI or through a
code-first notebooks experience.
Azure automated machine learning product home page
Azure automated ML infographic (PDF)
Tutorial: Create a classification model with automated ML in Azure Machine Learning
Tutorial: Use automated machine learning to predict taxi fares
Configure automated ML experiments in Python
Use the CLI extension for Azure Machine Learning
Automate machine learning activities with the Azure Machine Learning CLI
Azure Cognitive Services
This is a comprehensive family of AI services and cognitive APIs to help you build intelligent apps. These domain-
specific, pretrained AI models can be customized with your data.
Cognitive Services product home page
Azure Cognitive Services documentation
Azure Cognitive Search
This is an AI-powered cloud search service for mobile and web app development. The service can search over
private heterogenous content, with options for AI enrichment if your content is unstructured or unsearchable in raw
form.
Azure Cognitive Search product home page
Getting started with AI enrichment
Azure Cognitive Search documentation overview
Choosing a natural language processing technology in Azure
Quickstart: Create an Azure Cognitive Search cognitive skill set in the Azure portal
Azure Bot Service
This is a purpose-built bot development environment with out-of-the-box templates to get started quickly.
Azure Bot Service product home page
Azure Bot Service documentation overview
Azure reference architecture: Enterprise-grade conversational bot
Example workload: Conversational chatbot for hotel reservations on Azure
Microsoft Bot Framework
GitHub Bot Builder repo
Apache Spark on Azure
Apache Spark is a parallel processing framework that supports in-memory processing to boost the performance of
big data analytic applications. Spark provides primitives for in-memory cluster computing. A Spark job can load
and cache data into memory and query it repeatedly, which is much faster than disk-based applications, such as
Hadoop.
Apache Spark in Azure HDInsight is the Microsoft implementation of Apache Spark in the cloud. Spark clusters in
HDInsight are compatible with Azure Storage and Azure Data Lake Storage, so you can use HDInsight Spark clusters
to process your data stored in Azure.
The Microsoft Machine Learning library for Apache Spark is MMLSpark (Microsoft ML for Apache Spark). It is an
open-source library that adds many deep learning and data science tools, networking capabilities, and production-
grade performance to the Spark ecosystem. Learn more about MMLSpark features and capabilities.
Azure HDInsight overview. Basic information about features, cluster architecture, and use cases, with
pointers to quickstarts and tutorials.
Tutorial: Build an Apache Spark machine learning application in Azure HDInsight
Apache Spark best practices on HDInsight
Configure HDInsight Apache Spark Cluster settings
Machine learning on HDInsight
GitHub repo for MMLSpark: Microsoft Machine Learning library for Apache Spark
Create an Apache Spark machine learning pipeline on HDInsight
Azure Databricks Runtime for Machine Learning
Azure Databricks is an Apache Spark–based analytics platform with one-click setup, streamlined workflows, and an
interactive workspace for collaboration between data scientists, engineers, and business analysts.
Databricks Runtime for Machine Learning (Databricks Runtime ML) lets you start a Databricks cluster with all of the
libraries required for distributed training. It provides a ready-to-go environment for machine learning and data
science. Plus, it contains multiple popular libraries, including TensorFlow, PyTorch, Keras, and XGBoost. It also
supports distributed training using Horovod.
Azure Databricks product home page
Azure Databricks documentation
Machine learning capabilities in Azure Databricks
How-to guide: Databricks Runtime for Machine Learning
Batch scoring of Spark machine learning models on Azure Databricks
Deep learning overview for Azure Databricks

Customer stories
Different industries are applying AI in innovative and inspiring ways. Following are a number of customer case
studies and success stories:
ASOS: Online retailer solves challenges with Azure Machine Learning service
KPMG helps financial institutions save millions in compliance costs with Azure Cognitive Services
Volkswagen: Machine translation speaks Volkswagen – in 40 languages
Buncee: NYC school empowers readers of all ages and abilities with Azure AI
InterSystems: Data platform company boosts healthcare IT by generating critical information at
unprecedented speed
Zencity: Data-driven startup uses funding to help local governments support better quality of life for
residents
Bosch uses IoT innovation to drive traffic safety improvements by helping drivers avoid serious accidents
Automation Anywhere: Robotic process automation platform developer enriches its software with Azure
Cognitive Services
Wix deploys smart, scalable search across 150 million websites with Azure Cognitive Search
Asklepios Klinik Altona: Precision surgeries with Microsoft HoloLens 2 and 3D visualization
AXA Global P&C: Global insurance firm models complex natural disasters with cloud-based HPC
Browse more AI customer stories

Next steps
To learn about the artificial intelligence development products available from Microsoft, refer to
the Microsoft AI platform page.
For training in how to develop AI solutions, refer to Microsoft AI School.
Microsoft AI on GitHub: Samples, reference architectures, and best practices organizes the Microsoft open
source AI-based repositories, providing tutorials and learning materials.
 Choosing a Microsoft cognitive services technology
12/18/2020 • 3 minutes to read • Edit Online

Microsoft cognitive services are cloud-based APIs that you can use in artificial intelligence (AI) applications and
data flows. They provide you with pretrained models that are ready to use in your application, requiring no data
and no model training on your part. The cognitive services are developed by Microsoft's AI and Research team and
leverage the latest deep learning algorithms. They are consumed over HTTP REST interfaces. In addition, SDKs are
available for many common application development frameworks.
The cognitive services include:
Text analysis
Computer vision
Video analytics
Speech recognition and generation
Natural language understanding
Intelligent search
Key benefits:
Minimal development effort for state-of-the-art AI services.
Easy integration into apps via HTTP REST interfaces.
Built-in support for consuming cognitive services in Azure Data Lake Analytics.
Considerations:
Only available over the web. Internet connectivity is generally required. An exception is the Custom Vision
Service, whose trained model you can export for prediction on devices and at the IoT edge.
Although considerable customization is supported, the available services may not suit all predictive
analytics requirements.

What are your options when choosing amongst the cognitive services?
In Azure, there are dozens of Cognitive Services available. The current listing of these is available in a directory
categorized by the functional area they support:
Vision
Speech
Decision
Search
Language

Key Selection Criteria

To narrow the choices, start by answering these questions:
What type of data are you dealing with? Narrow your options based on the type of input data you are
working with. For example, if your input is text, select from the services that have an input type of text.
Do you have the data to train a model? If yes, consider the custom services that enable you to train their
underlying models with data that you provide, for improved accuracy and performance.
Capability matrix
The following tables summarize the key differences in capabilities.
Uses prebuilt models
C A PA B IL IT Y IN P UT T Y P E K EY B EN EF IT

Text Analytics API Text Evaluate sentiment and topics to

understand what users want.

Entity Linking API Text Power your app's data links with named
entity recognition and disambiguation.

Language Understanding Intelligent Text Teach your apps to understand

Service (LUIS) commands from your users.

QnA Maker Service Text Distill FAQ formatted information into

conversational, easy-to-navigate
answers.

Linguistic Analysis API Text Simplify complex language concepts

and parse text.

Knowledge Exploration Service Text Enable interactive search experiences

over structured data via natural
language inputs.

Web Language Model API Text Use predictive language models trained
on web-scale data.

Academic Knowledge API Text Tap into the wealth of academic content
in the Microsoft Academic Graph
populated by Bing.

Bing Autosuggest API Text Give your app intelligent autosuggest

options for searches.

Bing Spell Check API Text Detect and correct spelling mistakes in
your app.

Translator Text API Text Machine translation.

Recommendations API Text Predict and recommend items your

customers want.

Bing Entity Search API Text (web search query) Identify and augment entity
information from the web.

Bing Image Search API Text (web search query) Search for images.

Bing News Search API Text (web search query) Search for news.

Bing Video Search API Text (web search query) Search for videos.
 C A PA B IL IT Y IN P UT T Y P E K EY B EN EF IT

Bing Web Search API Text (web search query) Get enhanced search details from
billions of web documents.

Bing Speech API Text or Speech Convert speech to text and back again.

Speaker Recognition API Speech Use speech to identify and authenticate

individual speakers.

Translator Speech API Speech Perform real-time speech translation.

Computer Vision API Images (or frames from video) Distill actionable information from
images, automatically create description
of photos, derive tags, recognize
celebrities, extract text, and create
accurate thumbnails.

Content Moderator Text, Images or Video Automated image, text, and video
moderation.

Emotion API Images (photos with human subjects) Identify the range emotions of human
subjects.

Face API Images (photos with human subjects) Detect, identify, analyze, organize, and
tag faces in photos.

Video Indexer Video Video insights such as sentiment,

transcript speech, translate speech,
recognize faces and emotions, and
extract keywords.

Trained with custom data you provide

C A PA B IL IT Y IN P UT T Y P E K EY B EN EF IT

Custom Vision Service Images (or frames from video) Customize your own computer vision
models.

Custom Speech Service Speech Overcome speech recognition barriers

like speaking style, background noise,
and vocabulary.

Custom Decision Service Web content (for example, RSS feed) Use machine learning to automatically
select the appropriate content for your
home page

Bing Custom Search API Text (web search query) Commercial-grade search tool.
 Compare the machine learning products and
technologies from Microsoft
12/18/2020 • 8 minutes to read • Edit Online

Learn about the machine learning products and technologies from Microsoft. Compare options to help you choose
how to most effectively build, deploy, and manage your machine learning solutions.

Cloud-based machine learning products

The following options are available for machine learning in the Azure cloud.

C LO UD O P T IO N S W H AT IT IS W H AT Y O U C A N DO W IT H IT

Azure Machine Learning Managed platform for machine learning Use a pretrained model. Or, train,
deploy, and manage models on Azure
using Python and CLI

Azure Cognitive Services Pre-built AI capabilities implemented Build intelligent applications quickly
through REST APIs and SDKs using standard programming
languages. Doesn't require machine
learning and data science expertise

Azure SQL Managed Instance Machine In-database machine learning for SQL Train and deploy models inside Azure
Learning Services SQL Managed Instance

Machine learning in Azure Synapse Analytics service with machine learning Train and deploy models inside Azure
Analytics SQL Managed Instance

Machine learning and AI with ONNX in Machine learning in SQL on IoT Train and deploy models inside Azure
Azure SQL Edge SQL Edge

Azure Databricks Apache Spark-based analytics platform Build and deploy models and data
workflows using integrations with
open-source machine learning libraries
and the MLFlow platform.

On-premises machine learning products

The following options are available for machine learning on-premises. On-premises servers can also run in a
virtual machine in the cloud.

O N - P REM ISES O P T IO N S W H AT IT IS W H AT Y O U C A N DO W IT H IT

SQL Server Machine Learning Services In-database machine learning for SQL Train and deploy models inside SQL
Server

Machine Learning Services on SQL Machine learning in Big Data Clusters Train and deploy models on SQL Server
Server Big Data Clusters Big Data Clusters

Development platforms and tools

The following development platforms and tools are available for machine learning.

P L AT F O RM S/ TO O L S W H AT IT IS W H AT Y O U C A N DO W IT H IT

Azure Data Science Virtual Machine Virtual machine with pre-installed data Develop machine learning solutions in a
science tools pre-configured environment

ML.NET Open-source, cross-platform machine Develop machine learning solutions for

learning SDK .NET applications

Windows ML Windows 10 machine learning platform Evaluate trained models on a Windows

10 device

MMLSpark Open-source, distributed, machine Create and deploy scalable machine

learning and microservices framework learning applications for Scala and
for Apache Spark Python.

Machine Learning extension for Azure Open-source and cross-platform Manage packages, import machine
Data Studio machine learning extension for Azure learning models, make predictions, and
Data Studio create notebooks to run experiments
for your SQL databases

Azure Machine Learning

Azure Machine Learning is a fully managed cloud service used to train, deploy, and manage machine learning
models at scale. It fully supports open-source technologies, so you can use tens of thousands of open-source
Python packages such as TensorFlow, PyTorch, and scikit-learn. Rich tools are also available, such as Compute
instances, Jupyter notebooks, or the Azure Machine Learning for Visual Studio Code extension, a free extension
that allows you to manage your resources, model training workflows and deployments in Visual Studio Code.
Azure Machine Learning includes features that automate model generation and tuning with ease, efficiency, and
accuracy.
Use Python SDK, Jupyter notebooks, R, and the CLI for machine learning at cloud scale. For a low-code or no-code
option, use Azure Machine Learning's interactive designer in the studio to easily and quickly build, test, and deploy
models using pre-built machine learning algorithms.
Try Azure Machine Learning for free.

Type Cloud-based machine learning solution

Suppor ted languages Python, R

Machine learning phases Model training

Deployment
MLOps/Management

Key benefits Code first (SDK) and studio & drag-and-drop designer web
interface authoring options.

Central management of scripts and run history, making it

easy to compare model versions.

Easy deployment and management of models to the cloud or

edge devices.
 Considerations Requires some familiarity with the model management model.

Azure Cognitive Services

Azure Cognitive Services is a set of pre-built APIs that enable you to build apps that use natural methods of
communication. The term pre-built suggests that you do not need to bring datasets or data science expertise to
train models to use in your applications. That's all done for you and packaged as APIs and SDKs that allow your
apps to see, hear, speak, understand, and interpret user needs with just a few lines of code. You can easily add
intelligent features to your apps, such as:
Vision - Object detection, face recognition, OCR, etc. See Computer Vision, Face, Form Recognizer.
Speech - Speech-to-text, text-to-speech, speaker recognition, etc. See Speech Service.
Language - Translation, Sentiment analysis, key phrase extraction, language understanding, etc. See Translator,
Text Analytics, Language Understanding, QnA Maker
Decision - Anomaly detection, content moderation, reinforcement learning. See Anomaly Detector, Content
Moderator, Personalizer.
Use Cognitive Services to develop apps across devices and platforms. The APIs keep improving, and are easy to set
up.

Type APIs for building intelligent applications

Suppor ted languages Various options depending on the service. Standard ones are
C#, Java, JavaScript, and Python.

Machine learning phases Deployment

Key benefits Build intelligent applications using pre-trained models

available through REST API and SDK.
Variety of models for natural communication methods with
vision, speech, language, and decision.
No machine learning or data science expertise required.

SQL machine learning

SQL machine learning adds statistical analysis, data visualization, and predictive analytics in Python and R for
relational data, both on-premises and in the cloud. Current platforms and tools include:
SQL Server Machine Learning Services
Machine Learning Services on SQL Server Big Data Clusters
Azure SQL Managed Instance Machine Learning Services
Machine learning in Azure Synapse Analytics
Machine learning and AI with ONNX in Azure SQL Edge
Machine Learning extension for Azure Data Studio
Use SQL machine learning when you need built-in AI and predictive analytics on relational data in SQL.

Type On-premises predictive analytics for relational data

 Suppor ted languages Python, R, SQL

Machine learning phases Data preparation

Model training
Deployment

Key benefits Encapsulate predictive logic in a database function, making it

easy to include in data-tier logic.

Considerations Assumes a SQL database as the data tier for your application.

Azure Data Science Virtual Machine

The Azure Data Science Virtual Machine is a customized virtual machine environment on the Microsoft Azure
cloud. It is available in versions for both Windows and Linux Ubuntu. The environment is built specifically for doing
data science and developing ML solutions. It has many popular data science, ML frameworks, and other tools pre-
installed and pre-configured to jump-start building intelligent applications for advanced analytics.
Use the Data Science VM when you need to run or host your jobs on a single node. Or if you need to remotely
scale up your processing on a single machine.

Type Customized virtual machine environment for data science

Key benefits Reduced time to install, manage, and troubleshoot data

science tools and frameworks.

The latest versions of all commonly used tools and

frameworks are included.

Virtual machine options include highly scalable images with

GPU capabilities for intensive data modeling.

Considerations The virtual machine cannot be accessed when offline.

Running a virtual machine incurs Azure charges, so you must

be careful to have it running only when required.

Azure Databricks
Azure Databricks is an Apache Spark-based analytics platform optimized for the Microsoft Azure cloud services
platform. Databricks is integrated with Azure to provide one-click setup, streamlined workflows, and an interactive
workspace that enables collaboration between data scientists, data engineers, and business analysts. Use Python,
R, Scala, and SQL code in web-based notebooks to query, visualize, and model data.
Use Databricks when you want to collaborate on building machine learning solutions on Apache Spark.

Type Apache Spark-based analytics platform

Suppor ted languages Python, R, Scala, SQL

 Machine learning phases Data preparation
Data preprocessing
Model training
Model tuning
Model inference
Management
Deployment

ML.NET
ML.NET is an open-source, and cross-platform machine learning framework. With ML.NET, you can build custom
machine learning solutions and integrate them into your .NET applications. ML.NET offers varying levels of
interoperability with popular frameworks like TensorFlow and ONNX for training and scoring machine learning
and deep learning models. For resource-intensive tasks like training image classification models, you can take
advantage of Azure to train your models in the cloud.
Use ML.NET when you want to integrate machine learning solutions into your .NET applications. Choose between
the API for a code-first experience and Model Builder or the CLI for a low-code experience.

Type Open-source cross-platform framework for developing

custom machine learning applications with .NET

Languages suppor ted C#, F#

Machine learning phases Data preparation

Training
Deployment

Key benefits Data science & ML experience not required

Use familiar tools (Visual Studio, VS Code) and languages
Deploy where .NET runs
Extensible
Scalable
Local-first experience

Windows ML
Windows ML inference engine allows you to use trained machine learning models in your applications, evaluating
trained models locally on Windows 10 devices.
Use Windows ML when you want to use trained machine learning models within your Windows applications.

Type Inference engine for trained models in Windows devices

Languages suppor ted C#/C++, JavaScript

MMLSpark
Microsoft ML for Apache Spark (MMLSpark) is an open-source library that expands the distributed computing
framework Apache Spark. MMLSpark adds many deep learning and data science tools to the Spark ecosystem,
including seamless integration of Spark Machine Learning pipelines with Microsoft Cognitive Toolkit (CNTK),
LightGBM, LIME (Model Interpretability), and OpenCV. You can use these tools to create powerful predictive
models on any Spark cluster, such as Azure Databricks or Cosmic Spark.
MMLSpark also brings new networking capabilities to the Spark ecosystem. With the HTTP on Spark project, users
can embed any web service into their SparkML models. Additionally, MMLSpark provides easy-to-use tools for
orchestrating Azure Cognitive Services at scale. For production-grade deployment, the Spark Serving project
enables high throughput, submillisecond latency web services, backed by your Spark cluster.

Type Open-source, distributed machine learning and microservices

framework for Apache Spark

Languages suppor ted Scala 2.11, Java, Python 3.5+, R (beta)

Machine learning phases Data preparation

Model training
Deployment

Key benefits Scalability

Streaming + Serving compatible
Fault-tolerance

Considerations Requires Apache Spark

Next steps
To learn about all the Artificial Intelligence (AI) development products available from Microsoft, see Microsoft AI
platform
For training in developing AI and Machine Learning solutions with Microsoft, see Microsoft Learn
 Machine learning at scale
12/18/2020 • 3 minutes to read • Edit Online

Machine learning (ML) is a technique used to train predictive models based on mathematical algorithms. Machine
learning analyzes the relationships between data fields to predict unknown values.
Creating and deploying a machine learning model is an iterative process:
Data scientists explore the source data to determine relationships between features and predicted labels.
The data scientists train and validate models based on appropriate algorithms to find the optimal model for
prediction.
The optimal model is deployed into production, as a web service or some other encapsulated function.
As new data is collected, the model is periodically retrained to improve its effectiveness.
Machine learning at scale addresses two different scalability concerns. The first is training a model against large
data sets that require the scale-out capabilities of a cluster to train. The second centers on operationalizing the
learned model so it can scale to meet the demands of the applications that consume it. Typically this is
accomplished by deploying the predictive capabilities as a web service that can then be scaled out.
Machine learning at scale has the benefit that it can produce powerful, predictive capabilities because better
models typically result from more data. Once a model is trained, it can be deployed as a stateless, highly
performant scale-out web service.

Model preparation and training

During the model preparation and training phase, data scientists explore the data interactively using languages like
Python and R to:
Extract samples from high volume data stores.
Find and treat outliers, duplicates, and missing values to clean the data.
Determine correlations and relationships in the data through statistical analysis and visualization.
Generate new calculated features that improve the predictiveness of statistical relationships.
Train ML models based on predictive algorithms.
Validate trained models using data that was withheld during training.
To support this interactive analysis and modeling phase, the data platform must enable data scientists to explore
data using a variety of tools. Additionally, the training of a complex machine learning model can require a lot of
intensive processing of high volumes of data, so sufficient resources for scaling out the model training is essential.

Model deployment and consumption

When a model is ready to be deployed, it can be encapsulated as a web service and deployed in the cloud, to an
edge device, or within an enterprise ML execution environment. This deployment process is referred to as
operationalization.

Challenges
Machine learning at scale produces a few challenges:
You typically need a lot of data to train a model, especially for deep learning models.
You need to prepare these big data sets before you can even begin training your model.
 The model training phase must access the big data stores. It's common to perform the model training using the
same big data cluster, such as Spark, that is used for data preparation.
For scenarios such as deep learning, not only will you need a cluster that can provide you scale-out on CPUs,
but your cluster will need to consist of GPU-enabled nodes.

Machine learning at scale in Azure

Before deciding which ML services to use in training and operationalization, consider whether you need to train a
model at all, or if a prebuilt model can meet your requirements. In many cases, using a prebuilt model is just a
matter of calling a web service or using an ML library to load an existing model. Some options include:
Use the web services provided by Azure Cognitive Services.
Use the pretrained neural network models provided by the Cognitive Toolkit.
Embed the serialized models provided by Core ML for an iOS app.
If a prebuilt model does not fit your data or your scenario, options in Azure include Azure Machine Learning,
HDInsight with Spark MLlib and MMLSpark, Azure Databricks, Cognitive Toolkit, and SQL Machine Learning
Services. If you decide to use a custom model, you must design a pipeline that includes model training and
operationalization.
For a list of technology choices for ML in Azure, see:
Choosing a cognitive services technology
Choosing a machine learning technology
Choosing a natural language processing technology

Next steps
The following reference architectures show machine learning scenarios in Azure:
Batch scoring on Azure for deep learning models
Real-time scoring of Python Scikit-Learn and Deep Learning Models on Azure
 Choosing a natural language processing technology
in Azure
12/18/2020 • 3 minutes to read • Edit Online

Natural language processing (NLP) is used for tasks such as sentiment analysis, topic detection, language
detection, key phrase extraction, and document categorization.

NLP can be use to classify documents, such as labeling documents as sensitive or spam. The output of NLP can be
used for subsequent processing or search. Another use for NLP is to summarize text by identifying the entities
present in the document. These entities can also be used to tag documents with keywords, which enables search
and retrieval based on content. Entities might be combined into topics, with summaries that describe the
important topics present in each document. The detected topics may be used to categorize the documents for
navigation, or to enumerate related documents given a selected topic. Another use for NLP is to score text for
sentiment, to assess the positive or negative tone of a document. These approaches use many techniques from
natural language processing, such as:
Tokenizer . Splitting the text into words or phrases.
Stemming and lemmatization . Normalizing words so that different forms map to the canonical word with
the same meaning. For example, "running" and "ran" map to "run."
Entity extraction . Identifying subjects in the text.
Par t of speech detection . Identifying text as a verb, noun, participle, verb phrase, and so on.
Sentence boundar y detection . Detecting complete sentences within paragraphs of text.
When using NLP to extract information and insight from free-form text, the starting point is typically the raw
documents stored in object storage such as Azure Storage or Azure Data Lake Store.

Challenges
Processing a collection of free-form text documents is typically computationally resource intensive, as well as
being time intensive.
Without a standardized document format, it can be difficult to achieve consistently accurate results using free-
form text processing to extract specific facts from a document. For example, think of a text representation of an
invoice—it can be difficult to build a process that correctly extracts the invoice number and invoice date for
invoices across any number of vendors.
What are your options when choosing an NLP service?
In Azure, the following services provide natural language processing (NLP) capabilities:
Azure HDInsight with Spark and Spark MLlib
Azure Databricks
Microsoft Cognitive Services

Key selection criteria

To narrow the choices, start by answering these questions:
Do you want to use prebuilt models? If yes, consider using the APIs offered by Microsoft Cognitive Services.
Do you need to train custom models against a large corpus of text data? If yes, consider using Azure
HDInsight with Spark MLlib and Spark NLP.
Do you need low-level NLP capabilities like tokenization, stemming, lemmatization, and term
frequency/inverse document frequency (TF/IDF)? If yes, consider using Azure HDInsight with Spark MLlib
and Spark NLP.
Do you need simple, high-level NLP capabilities like entity and intent identification, topic detection, spell
check, or sentiment analysis? If yes, consider using the APIs offered by Microsoft Cognitive Services.

Capability matrix
The following tables summarize the key differences in capabilities.
General capabilities
C A PA B IL IT Y A Z URE H DIN SIGH T M IC RO SO F T C O GN IT IVE SERVIC ES

Provides pretrained models as a service No Yes

REST API Yes Yes

Programmability Python, Scala, Java C#, Java, Node.js, Python, PHP, Ruby

Support processing of big data sets and Yes No

large documents

Low-level natural language processing capabilities

C A PA B IL IT Y A Z URE H DIN SIGH T M IC RO SO F T C O GN IT IVE SERVIC ES

Tokenizer Yes (Spark NLP) Yes (Linguistic Analysis API)

Stemmer Yes (Spark NLP) No

Lemmatizer Yes (Spark NLP) No

Part of speech tagging Yes (Spark NLP) Yes (Linguistic Analysis API)

Term frequency/inverse-document Yes (Spark MLlib) No

frequency (TF/IDF)
 C A PA B IL IT Y A Z URE H DIN SIGH T M IC RO SO F T C O GN IT IVE SERVIC ES

String similarity—edit distance Yes (Spark MLlib) No

calculation

N-gram calculation Yes (Spark MLlib) No

Stop word removal Yes (Spark MLlib) No

High-level natural language processing capabilities

C A PA B IL IT Y A Z URE H DIN SIGH T M IC RO SO F T C O GN IT IVE SERVIC ES

Entity/intent identification and No Yes (Language Understanding

extraction Intelligent Service (LUIS) API)

Topic detection Yes (Spark NLP) Yes (Text Analytics API)

Spell checking Yes (Spark NLP) Yes (Bing Spell Check API)

Sentiment analysis Yes (Spark NLP) Yes (Text Analytics API)

Language detection No Yes (Text Analytics API)

Supports multiple languages besides No Yes (varies by API)

English

See also
Natural language processing
 R developer's guide to Azure
12/18/2020 • 8 minutes to read • Edit Online

Many data scientists dealing with ever-increasing volumes of data are looking for ways to
harness the power of cloud computing for their analyses. This article provides an
overview of the various ways that data scientists can use their existing skills with the R
programming language in Azure.
Microsoft has fully embraced the R programming language as a first-class tool for data
scientists. By providing many different options for R developers to run their code in
Azure, the company is enabling data scientists to extend their data science workloads into the cloud when tackling
large-scale projects.
Let's examine the various options and the most compelling scenarios for each one.

Azure services with R language support

This article covers the following Azure services that support the R language:

SERVIC E DESC RIP T IO N

Azure Machine Learning Server enterprise software for data science, providing R and Python
interpreters

Data Science Virtual Machine a customized VM to use as a data science workstation or as a

custom compute target

ML Services on HDInsight cluster-based system for running R analyses on large datasets

across many nodes

Azure Databricks collaborative Spark environment that supports R and other

languages

Azure Machine Learning cloud service that you use to train, deploy, automate, and
manage machine learning models

Azure Machine Learning Studio (classic) run custom R scripts in Azure's machine learning experiments

Azure Batch offers a variety options for economically running R code across
many nodes in a cluster

Azure Notebooks a no-cost cloud-based version of Jupyter notebooks

Azure SQL Database run R scripts inside of the SQL Server database engine

Data Science Virtual Machine

The Data Science Virtual Machine (DSVM) is a customized VM image on Microsoft’s Azure cloud platform built
specifically for doing data science. It has many popular data science tools, including:
Microsoft R Open
 Microsoft Machine Learning Server
RStudio Desktop
RStudio Server
The DSVM can be provisioned with either Windows or Linux as the operating system. You can use the DSVM in two
different ways: as an interactive workstation or as a compute platform for a custom cluster.
As a workstation
If you want to get started with R in the cloud quickly and easily, this is your best bet. The environment will be
familiar to anyone who has worked with R on a local workstation. However, instead of using local resources, the R
environment runs on a VM in the cloud. If your data is already stored in Azure, this has the added benefit of
allowing your R scripts to run "closer to the data." Instead of transferring the data across the Internet, the data can
be accessed over Azure's internal network, which provides much faster access times.
The DSVM can be particularly useful to small teams of R developers. Instead of investing in powerful workstations
for each developer and requiring team members to synchronize on which versions of the various software
packages they will use, each developer can spin up an instance of the DSVM whenever needed.
As a compute platform
In addition to being used as a workstation, the DSVM is also used as an elastically scalable compute platform for R
projects. Using the AzureDSVM R package, you can programmatically control the creation and deletion of DSVM
instances. You can form the instances into a cluster and deploy a distributed analysis to be performed in the cloud.
This entire process can be controlled by R code running on your local workstation.
To learn more about the DSVM, see Introduction to Azure Data Science Virtual Machine for Linux and Windows.

ML Services on HDInsight
Microsoft ML Services provide data scientists, statisticians, and R programmers with on-demand access to scalable,
distributed methods of analytics on HDInsight. This solution provides the latest capabilities for R-based analytics on
datasets of virtually any size, loaded to either Azure Blob or Data Lake storage.
This is an enterprise-grade solution that allows you to scale your R code across a cluster. By using functions in
Microsoft's RevoScaleR package, your R scripts on HDInsight can run data processing functions in parallel across
many nodes in a cluster. This allows R to crunch data on a much larger scale than is possible with single-threaded R
running on a workstation.
This ability to scale makes ML Services on HDInsight a great option for R developers with massive data sets. It
provides a flexible and scalable platform for running your R scripts in the cloud.
For a walk-through on creating an ML Services cluster, see Get started with ML Services on Azure HDInsight.

Azure Databricks
Azure Databricks is an Apache Spark-based analytics platform optimized for the Microsoft Azure cloud services
platform. Designed with the founders of Apache Spark, Databricks is integrated with Azure to provide one-click
setup, streamlined workflows, and an interactive workspace that enables collaboration between data scientists, data
engineers, and business analysts.
The collaboration in Databricks is enabled by the platform's notebook system. Users can create, share, and edit
notebooks with other users of the systems. These notebooks allow users to write code that executes against Spark
clusters managed in the Databricks environment. These notebooks fully support R and give users access to Spark
through both the SparkR and sparklyr packages.
Since Databricks is built on Spark and has a strong focus on collaboration, the platform is often used by teams of
data scientists that work together on complex analyses of large data sets. Because the notebooks in Databricks
support other languages in addition to R, it is especially useful for teams where analysts use different languages
for their primary work.
The article What is Azure Databricks? can provide more details about the platform and help you get started.

Azure Machine Learning

Azure Machine Learning can be used for any kind of machine learning, from classical machine learning to deep
learning, supervised and unsupervised learning. Whether you prefer to write Python or R code or zero-code/low-
code options such as the designer, you can build, train and track highly accurate machine learning and deep-
learning models in an Azure Machine Learning Workspace.
Start training on your local machine and then scale out to the cloud. Train your first model in R with Azure Machine
Learning today.

Azure Machine Learning Studio (classic)

Azure Machine Learning Studio (classic) is a collaborative, drag-and-drop tool you can use to build, test, and deploy
predictive analytics solutions in the cloud. It enables emerging data scientists to create and deploy machine
learning models without the need to write much code.
Azure Machine Learning Studio (classic) supports both R and Python.
Customers currently using or evaluating Azure Machine Learning Studio (classic) are encouraged to try the
designer in Azure Machine Learning, which provides drag-n-drop ML modules plus scalability, version control, and
enterprise security.

Azure Batch
For large-scale R jobs, you can use Azure Batch. This service provides cloud-scale job scheduling and compute
management so you can scale your R workload across tens, hundreds, or thousands of virtual machines. Since it is
a generalized computing platform, there a few options for running R jobs on Azure Batch.
One option is to use Microsoft's doAzureParallel package. This R package is a parallel backend for the foreach
package. It allows each iteration of the foreach loop to run in parallel on a node within the Azure Batch cluster. For
an introduction to the package, see the blog post doAzureParallel: Take advantage of Azure’s flexible compute
directly from your R session.
Another option for running an R script in Azure Batch is to bundle your code with "RScript.exe" as a Batch App in
the Azure portal. For a detailed walkthrough, see R Workloads on Azure Batch.
A third option is to use the Azure Distributed Data Engineering Toolkit (AZTK), which allows you to provision on-
demand Spark clusters using Docker containers in Azure Batch. This provides an economical way to run Spark jobs
in Azure. By using SparklyR with AZTK, your R scripts can be scaled out in the cloud easily and economically.

Azure Notebooks
Azure Notebooks is a low-cost, low-friction method for R developers who prefer working with notebooks to bring
their code to Azure. It is a free service for anyone to develop and run code in their browser using Jupyter, which is
an open-source project that enables combing markdown prose, executable code, and graphics onto a single canvas.
The free service tier of Azure Notebooks is a viable option for small-scale projects, as it limits each notebook's
process to 4 GB of memory and 1 GB data sets. If you need compute and data power beyond these limitations,
however, you can run notebooks in a Data Science Virtual Machine instance. For more information, see Manage
and configure Azure Notebooks projects - Compute tier.

Azure SQL Database

Azure SQL Database
Azure SQL Database is Microsoft's intelligent, fully managed relational cloud database service. It allows you to use
the full power of SQL Server without any hassle of setting up the infrastructure. This includes Machine Learning
Services in SQL Server, which is one of the more recent additions to SQL.
This feature offers an embedded, predictive analytics and data science engine that can execute R code within a SQL
Server database as stored procedures, as T-SQL scripts containing R statements, or as R code containing T-SQL.
Instead of extracting data from the database and loading it into the R environment, you load your R code directly
into the database and let it run right alongside the data.
While Machine Learning Services has been part of on-premises SQL Server since 2016, it is relatively new to Azure
SQL Database. It is currently in limited preview but will continue to evolve.
Next steps
Running your R code on Azure with mrsdeploy
Machine Learning Server in the Cloud
Additional Resources for Machine Learning Server and Microsoft R
R on Azure: an overview of packages, tools, and case studies for using R with Azure

The R logo is © 2016 The R Foundation and is used under the terms of the Creative Commons Attribution-ShareAlike 4.0 International license.
 Baseball decision analysis with ML.NET and Blazor
12/18/2020 • 5 minutes to read • Edit Online

This scenario describes a web application, called the Baseball Machine Learning Workbench, which provides an
interface for non-technical users to use artificial intelligence (AI) and machine learning (ML) to perform decision
analysis techniques in order to rapidly gain insights and make informed predictions.
This solution uses historical baseball data to generate National Baseball Hall of Fame insights. Machine intelligence
powers the what-if analysis, decision thresholding, and improvements over traditional rule-based systems. User-
friendly interface controls set adjustable parameters and surface the results in real time, with clear visual cues to
highlight positive or negative outcomes.
The architecture provides rapid results by using in-memory models and rapid two-way communication between
the user and server. Delta rendering pushes down only modified content to the web browser, updating the display
without having to reload the entire page. This solution can scale to tens of in-memory models serving hundreds of
concurrent sessions in real time.
The following article explains the architecture of the Baseball Machine Learning Workbench, where to get the
source code for it, and how to deploy it. You can also view a live demo of this solution.

Relevant use cases

Consider this scenario for the following use cases:
Decision analysis system replacement with AI and ML
AI-assisted decision support systems or decision management systems
Fantasy baseball
Financial forecasting
Business goal and objective modeling (budgeting and project planning, for example)

Architecture
The processing sequence in this solution flows as follows:
1. The user accesses the workbench application with any browser. They choose which analysis method to
employ and then which player to analyze.
2. SignalR brokers two-way communication with the server in real time.
3. Azure App Service hosts the application, including AI logic with the machine learning models.
4. One of three different decision analysis mechanisms is utilized, depending on which mode the user has
selected.
5. Historical data is analyzed using the designated set of rules or ML models in ML.NET, operating in-memory
for very quick inference.
6. Blazor Server surfaces the results to the end user’s browser, updating only the portions of the interface that
have changed, and transmits back to the user via SignalR.
7. Azure Application Insights is optionally used to monitor performance and instrumentation resources as
needed.
Components
The following assets and technologies were used to craft the Baseball Machine Learning Workbench:
Azure App Service enables you to build and host web applications in the programming language of your
choice without managing infrastructure.
.NET Core 3.1 is an open-source, cross-platform, general-purpose development framework that runs on
Windows, Linux, and macOS platforms.
ML.NET is a cross-platform, open-source framework for creating machine learning and artificial intelligence
models using .NET. It is used in the inference sections of this application.
Blazor Server lets you build interactive web UIs using C# instead of JavaScript, as in this solution. In this
architecture, Blazor renders the UI on the server and pushes HTML and other changes out to the browser.
SignalR provides asynchronous communication between the browser and the server (including Blazor). It
 handles event updates, UI updates, and any processing done on the server (such as model inference).
Visual Studio 2019 is the software programming environment used for this project. This architecture uses
cross-platform components, so either the Windows or Mac version can be used.
Azure Application Insights (a feature of Azure Monitor) can be used for performance monitoring and
analytics and to drive autoscaling.

Considerations
Scalability
This solution uses the prediction engine functionality in ML.NET to scale the model response times. Object pooling
allows the ML.NET models to be accessed by multiple requests in a thread-safe manner. Learn more about ML.NET
object pooling in Deploy a model in an ASP.NET Core Web API.
Azure App Service is used for hosting the workbench in the cloud. With App Service you can automatically scale the
number of instances that run your app, letting you keep up with customer demand. For more information on
autoscale, refer to Autoscaling best practices in the Azure Architecture Center.
In Blazor Server, the state of many components might be maintained concurrently by the server. Because of this,
memory exhaustion is a concern that must be addressed. For guidance on how to author a Blazor Server app to
help ensure the best use of server memory, consult Threat mitigation guidance for ASP.NET Core Blazor Server.
Applying these best practices allows a server-side Blazor application to scale to thousands of concurrent users—
even on relatively small server hosts.
General guidance on designing scalable solutions is provided in the Azure Architecture Center’s Performance
efficiency checklist.
Resiliency and support
Use .NET Core 3.1.x because it is a Long Term Support (LTS) release. Although Blazor Server is also available in .NET
Core 3.0, that is not an LTS release and thus continuing compatibility with future component updates is not assured.
Learn more about the .NET Core Support Policy.

Deploy this solution

All of the source code for this scenario is available in the Baseball Machine Learning Workbench repository. This
solution is open source and provided with an MIT License.
Prerequisites
For online deployments, you must have an existing Azure account. If you need one, create a free Azure account
before you begin.
For deployment as an Azure application instance, you need the Visual Studio 2019 IDE and you must have Git
installed locally.
The historical baseball data used for the analysis and machine learning models comes from Sean Lahman’s
Baseball Database, which is licensed under a Creative Commons Attribution-ShareAlike 3.0 Unported License. The
Major League Baseball data itself is public domain.
Deployment to Azure
1. Make sure you have your Azure subscription information handy.
2. Start by cloning the workbench GitHub repository:
 git clone
https://github.com/bartczernicki/MachineLearning-BaseballPrediction-BlazorApp.git

3. Follow the instructions provided in the GETSTARTED.md file.

Alternative: Docker container
This application is also available as a complete, ready-to-run Docker container downloadable from Docker Hub.
The container can be run locally (offline) in your own environment. It can also be deployed online in an Azure
Container Instance. Instructions for getting started with either use case are provided in the main GitHub repo’s Get
Started documentation:
Run the Docker Container locally in your own environment
Publish Docker Container to the Azure Cloud using Azure Container Instances

Related resources
View a live demo of this solution
Baseball HOF prediction using R mlr and DALEX packages is a GitHub repo using R and cutting edge “black
box” model techniques to explain ML.NET models related to this workload
Blazor documentation
ML.NET documentation
ASP.NET Core Blazor hosting models
MLOps (DevOps for Machine Learning) helps data science teams deliver innovation faster, increasing the
pace of ML model development
Learn about the National Baseball Hall of Fame voting process and rules
XAI Stories: Case Studies for Explainable Artificial Intelligence (Warsaw University of Technology and
University of Warsaw, 2020)
 High Performance Computing (HPC) on Azure
12/18/2020 • 7 minutes to read • Edit Online

Introduction to HPC
High Performance Computing (HPC), also called "Big Compute", uses a large number of CPU or GPU-based
computers to solve complex mathematical tasks.
Many industries use HPC to solve some of their most difficult problems. These include workloads such as:
Genomics
Oil and gas simulations
Finance
Semiconductor design
Engineering
Weather modeling
How is HPC different on the cloud?
One of the primary differences between an on-premises HPC system and one in the cloud is the ability for
resources to dynamically be added and removed as they're needed. Dynamic scaling removes compute capacity as
a bottleneck and instead allow customers to right size their infrastructure for the requirements of their jobs.
The following articles provide more detail about this dynamic scaling capability.
Big Compute Architecture Style
Autoscaling best practices

Implementation checklist
As you're looking to implement your own HPC solution on Azure, ensure you're reviewed the following topics:
Choose the appropriate architecture based on your requirements
Know which compute options is right for your workload
Identify the right storage solution that meets your needs
Decide how you're going to manage all your resources
Optimize your application for the cloud
Secure your Infrastructure

Infrastructure
There are a number of infrastructure components necessary to build an HPC system. Compute, Storage, and
Networking provide the underlying components, no matter how you choose to manage your HPC workloads.
Example HPC architectures
There are a number of different ways to design and implement your HPC architecture on Azure. HPC applications
can scale to thousands of compute cores, extend on-premises clusters, or run as a 100% cloud-native solution.
The following scenarios outline a few of the common ways HPC solutions are built.
Computer-aided engineering services on Azure
Provide a software-as-a-service (SaaS) platform for computer-aided engineering (CAE) on Azure.

Computational fluid dynamics (CFD) simulations on Azure

Execute computational fluid dynamics (CFD) simulations on Azure.
 3D video rendering on Azure
Run native HPC workloads in Azure using the Azure Batch service
Compute
Azure offers a range of sizes that are optimized for both CPU & GPU intensive workloads.
CPU-based virtual machines
Linux VMs
Windows VMs VMs
GPU-enabled virtual machines
N-series VMs feature NVIDIA GPUs designed for compute-intensive or graphics-intensive applications including
artificial intelligence (AI) learning and visualization.
Linux VMs
Windows VMs
Storage
Large-scale Batch and HPC workloads have demands for data storage and access that exceed the capabilities of
traditional cloud file systems. There are a number of solutions to manage both the speed and capacity needs of
HPC applications on Azure
Avere vFXT for faster, more accessible data storage for high-performance computing at the edge
Azure NetApp Files
BeeGFS
Storage Optimized Virtual Machines
Blob, table, and queue storage
Azure SMB File storage
For more information comparing Lustre, GlusterFS, and BeeGFS on Azure, review the Parallel Files Systems on
Azure e-book and the Lustre on Azure blog.
Networking
H16r, H16mr, A8, and A9 VMs can connect to a high throughput back-end RDMA network. This network can
improve the performance of tightly coupled parallel applications running under Microsoft MPI or Intel MPI.
 RDMA Capable Instances
Virtual Network
ExpressRoute

Management
Do -it-yourself
Building an HPC system from scratch on Azure offers a significant amount of flexibility, but is often very
maintenance intensive.
1. Set up your own cluster environment in Azure virtual machines or virtual machine scale sets.
2. Use Azure Resource Manager templates to deploy leading workload managers, infrastructure, and applications.
3. Choose HPC and GPU VM sizes that include specialized hardware and network connections for MPI or GPU
workloads.
4. Add high performance storage for I/O-intensive workloads.
Hybrid and cloud Bursting
If you have an existing on-premises HPC system that you'd like to connect to Azure, there are a number of
resources to help get you started.
First, review the Options for connecting an on-premises network to Azure article in the documentation. From there,
you may want information on these connectivity options:

Connect an on-premises network to Azure using a VPN gateway

This reference architecture shows how to extend an on-premises network to Azure, using a site-to-site
virtual private network (VPN).

Connect an on-premises network to Azure using ExpressRoute

ExpressRoute connections use a private, dedicated connection through a third-party connectivity provider.
The private connection extends your on-premises network into Azure.

Connect an on-premises network to Azure using ExpressRoute with VPN failover

Implement a highly available and secure site-to-site network architecture that spans an Azure virtual
network and an on-premises network connected using ExpressRoute with VPN gateway failover.
Once network connectivity is securely established, you can start using cloud compute resources on-demand with
the bursting capabilities of your existing workload manager.
Marketplace solutions
There are a number of workload managers offered in the Azure Marketplace.
RogueWave CentOS-based HPC
SUSE Linux Enterprise Server for HPC
TIBCO Grid Server Engine
 Azure Data Science VM for Windows and Linux
D3View
UberCloud
Azure Batch
Azure Batch is a platform service for running large-scale parallel and high-performance computing (HPC)
applications efficiently in the cloud. Azure Batch schedules compute-intensive work to run on a managed pool of
virtual machines, and can automatically scale compute resources to meet the needs of your jobs.
SaaS providers or developers can use the Batch SDKs and tools to integrate HPC applications or container
workloads with Azure, stage data to Azure, and build job execution pipelines.
Azure CycleCloud
Azure CycleCloud Provides the simplest way to manage HPC workloads using any scheduler (like Slurm, Grid
Engine, HPC Pack, HTCondor, LSF, PBS Pro, or Symphony), on Azure
CycleCloud allows you to:
Deploy full clusters and other resources, including scheduler, compute VMs, storage, networking, and cache
Orchestrate job, data, and cloud workflows
Give admins full control over which users can run jobs, as well as where and at what cost
Customize and optimize clusters through advanced policy and governance features, including cost controls,
Active Directory integration, monitoring, and reporting
Use your current job scheduler and applications without modification
Take advantage of built-in autoscaling and battle-tested reference architectures for a wide range of HPC
workloads and industries
Workload managers
The following are examples of cluster and workload managers that can run in Azure infrastructure. Create stand-
alone clusters in Azure VMs or burst to Azure VMs from an on-premises cluster.
Alces Flight Compute
TIBCO DataSynapse GridServer
Bright Cluster Manager
IBM Spectrum Symphony and Symphony LSF
Altair PBS Works
Rescale
Univa Grid Engine
Microsoft HPC Pack
HPC Pack for Windows
HPC Pack for Linux
Containers
Containers can also be used to manage some HPC workloads. Services like the Azure Kubernetes Service (AKS)
makes it simple to deploy a managed Kubernetes cluster in Azure.
Azure Kubernetes Service (AKS)
Container Registry

Cost management
Managing your HPC cost on Azure can be done through a few different ways. Ensure you've reviewed the Azure
purchasing options to find the method that works best for your organization.
Security
For an overview of security best practices on Azure, review the Azure Security Documentation.
In addition to the network configurations available in the Cloud Bursting section, you may want to implement a
hub/spoke configuration to isolate your compute resources:

Implement a hub-spoke network topology in Azure

The hub is a virtual network (VNet) in Azure that acts as a central point of connectivity to your on-premises
network. The spokes are VNets that peer with the hub, and can be used to isolate workloads.

Implement a hub-spoke network topology with shared services in Azure

This reference architecture builds on the hub-spoke reference architecture to include shared services in the
hub that can be consumed by all spokes.

HPC applications
Run custom or commercial HPC applications in Azure. Several examples in this section are benchmarked to scale
efficiently with additional VMs or compute cores. Visit the Azure Marketplace for ready-to-deploy solutions.

NOTE
Check with the vendor of any commercial application for licensing or other restrictions for running in the cloud. Not all
vendors offer pay-as-you-go licensing. You might need a licensing server in the cloud for your solution, or connect to an on-
premises license server.

Engineering applications
Altair RADIOSS
ANSYS CFD
MATLAB Distributed Computing Server
StarCCM+
OpenFOAM
Graphics and rendering
Autodesk Maya, 3ds Max, and Arnold on Azure Batch
AI and deep learning
Microsoft Cognitive Toolkit
Batch Shipyard recipes for deep learning
MPI Providers
Microsoft MPI

Remote visualization
 Linux virtual desktops with Citrix
Build a VDI environment for Linux Desktops using Citrix on Azure.

Performance Benchmarks
Compute benchmarks

Customer stories
There are a number of customers who have seen great success by using Azure for their HPC workloads. You can
find a few of these customer case studies below:
AXA Global P&C
Axioma
d3View
EFS
Hymans Robertson
MetLife
Microsoft Research
Milliman
Mitsubishi UFJ Securities International
NeuroInitiative
Schlumberger
Towers Watson

Other important information

Ensure your vCPU quota has been increased before attempting to run large-scale workloads.

Next steps
For the latest announcements, see:
Microsoft HPC and Batch team blog
Visit the Azure blog.
Microsoft Batch Examples
These tutorials will provide you with details on running applications on Microsoft Batch
Get started developing with Batch
Use Azure Batch code samples
Use low-priority VMs with Batch
Run containerized HPC workloads with Batch Shipyard
Run parallel R workloads on Batch
Run on-demand Spark jobs on Batch
Use compute-intensive VMs in Batch pools
 Azure Kubernetes Service solution journey
12/18/2020 • 3 minutes to read • Edit Online

Kubernetes is an open-source system for automating deployment, scaling, and management of containerized
applications. Azure Kubernetes Service (AKS) makes it simple to deploy a managed Kubernetes cluster in Azure.
Organizations are at various points in their understanding, rationalizing, and adoption of Kubernetes on Azure.
Your organization's journey will likely follow a similar path to many other technologies you've adopted; learning,
aligning your organization around roles & responsibilities, and deploying production-ready workloads. From there,
you'll iterate; growing your product as your customer and business demands change.

Best
Practices
AKS Baseline Cluster Your Workload Operate

Align

Learn

Iterate

Learn about Azure Kubernetes Service (AKS)

If you're new to Kubernetes or AKS, the best place to learn about the service is with Microsoft Learn. Microsoft
Learn is a free, online training platform that provides interactive learning for Microsoft products and more. The
Introduction to Kubernetes on Azure learning path will provide you with foundational knowledge that will take
you through core concepts of containers, AKS cluster management, and workload deployment.
Introduction to Kubernetes on Azure

Organizational readiness
As organizations such as yours have adopted Azure, the Cloud Adoption Framework provides them prescriptive
guidance as they move between the phases of the cloud adoption lifecycle. The Cloud Adoption Framework
includes tools, programs, and content to simplify adoption of Kubernetes and related cloud-native practices at
scale.
Kubernetes in the Cloud Adoption Framework

Path to production
You understand the benefits and trade-offs of Kubernetes, and have decided that AKS is the best Azure compute
platform for your workload. Your organizational controls have been put into place; you're ready to learn how to
deploy production-ready clusters for your workload.
Microsoft's AKS Baseline Cluster is the starting point to help you build production-ready AKS clusters. We
recommend you start from this baseline implementation and modify it to align to your workload's specific needs
and Well-Architected Framework priorities.
Microsoft's AKS Baseline Cluster
Best practices
As part of on going operations, you may wish to spot check your cluster against current recommended best
practices. The best place to start is to ensure your cluster is aligned with Microsoft's AKS Baseline Cluster.
See Best Practices for Cluster Operations and Best Practices for AKS Workloads.

You may also consider evaluating a community-driven utility like The AKS Checklist as a way of organizing and
tracking your alignment to these best practices.

Operations Guide
Getting your workload deployed on AKS is a great milestone and this is when day-2 operations are going to be
top-of-mind. Microsoft's AKS Day 2 Operations Guide was built for your ease of reference. This will help
ensure you are ready to meet the demands of your customers and ensure you are prepared for break-fix situations
via optimized triage processes.
Microsoft's AKS Day 2 Operations Guide

Stay current with AKS

Kubernetes and AKS are both moving fast. The platform is evolving and just knowing what's on the roadmap might
help you make architectural decisions and understand planned deprecations; consider bookmarking it.
AKS product roadmap

Additional resources
The typical AKS solution journey depicted above ranges from learning about AKS to growing your existing clusters
to meet new product and customer demands. However, you might also just be looking for additional reference and
supporting material to help along the way for your specific situation.
Example solutions
If you're seeking additional reference material that use AKS as their foundation, here are a few to consider.
Microservices architecture on AKS
Secure DevOps for AKS
Building a telehealth system
CI/CD pipeline for container-based workloads
Azure Arc
Azure Kubernetes Service offers you a managed Kubernetes experience on Azure, however there are workloads or
situations that might be best suited for placing your own Kubernetes clusters under Azure Arc management. This
includes your clusters such as RedHat OpenShift, RedHat RKE, and Canonical Charmed Kubernetes. Azure Arc
management should also be used for AKS Engine clusters running in your datacenter, in another cloud, or on Azure
Stack Hub.
Azure Arc enabled Kubernetes
Managed service provider
If you're a managed service provider, you already use Azure Lighthouse to manage resources for multiple
customers. Azure Kubernetes Service supports Azure Lighthouse so that you can manage hosted Kubernetes
environments and deploy containerized applications within your customers' tenants.
AKS with Azure Lighthouse
AWS or GCP professionals
These articles provide service mapping and comparison between Azure and other cloud services. This reference
can help you ramp up quickly on Azure.
Containers and container orchestrators for AWS Professionals
Containers and container orchestrators for GCP Professionals
 Azure Kubernetes Services (AKS) Day 2 Operations
Guide
12/18/2020 • 2 minutes to read • Edit Online

After releasing your Azure Kubernetes Service-hosted application, prepare for day-2 operations. The term refers to
the ongoing maintainence of the deployed assets and rollout of upgrades. The operations can help you:
Keep up to date with your service-level agreement (SLA) or Service-level objective (SLO) requirements.
Troubleshoot customer support requests.
Stay current with the latest platform features and security updates.
Plan for future growth.

Prerequisites
The best practices for day-2 operations assume that you've deployed the Azure Kubernetes Service (AKS) Baseline
architecture as an example of a production cluster.

Triage practices
It's often challenging to do root-cause analysis given the different aspects of an AKS cluster. When triaging issues,
consider a top-down approach on the cluster hierarchy. Start at the cluster level and drill down if necessary.
Cluster Node pools Nodes Pods Containers

In this series, we'll walk you through the thought process of this approach. The articles show examples using a set
of tools and dashboards, and how they can highlight some symptoms.
Common causes addressed in this series include:
Network and connectivity problems caused by improper configuration.
Control plane to node communication is broken.
Kubelet pressures caused by insufficient compute, memory, or storage resources.
DNS resolution issues.
Node health is running out of disk IOPS.
Admission control pipeline is blocking a large number of requests to the API server.
The cluster doesn't have permissions to pull from the appropriate container registry.
It’s not intended to resolve specific issues. For information about troubleshooting specific issues, AKS Common
Issues.

In this series
ST EP DESC RIP T IO N

1- Check the AKS cluster health Start by checking the cluster the health of the overall cluster
and networking.

2- Examine the node and pod health Check the health of the AKS worker nodes.
 ST EP DESC RIP T IO N

3- Check the workload deployments Check to see that all deployments and daemonSets are
running.

4- Validate the admission controllers Check whether the admission controllers are working as
expected.

5- Verify the connection to the container registry Verify the connection to the container registry.

Related links
Day-2 operations
AKS periscope
AKS roadmap
AKS documentation
 Check the AKS cluster health
12/18/2020 • 2 minutes to read • Edit Online

Start by checking the health of the overall cluster and networking.

This article is part of a series. Read the introduction here.
Tools
AKS Diagnostics . In Azure portal, navigate to the AKS cluster resource. Select Diagnose and solve problems .

Diagnostics shows a list of results from various test runs. If there are any issues found, More info can show you
information about the underlying issue.
This image indicates that network and connectivity issues are caused by Azure CNI subnet configuration.
To learn more about this feature, see Azure Kubernetes Service Diagnostics overview.

Next steps
Examine the node and pod health
 Examine the node and pod health
11/2/2020 • 5 minutes to read • Edit Online

If the cluster checks are clear, check the health of the AKS worker nodes. Determine the reason for the unhealthy
node and resolve the issue.
This article is part of a series. Read the introduction here.

1- Check the overall health of the worker nodes

A node can be unhealthy because of various reasons. A common reason is when the control plane to node
communication is broken as a result of misconfiguration of routing and firewall rules. As a fix, you can allow the
necessary ports and fully qualified domain names through the firewall according to the AKS egress traffic
guidance. Another reason can be kubelet pressures. In that case, add more compute, memory, or storage
resources.
Tools
You can check node health in one of these ways:
Azure Monitor - Containers health view . In Azure portal, open Azure Monitor. Select Containers . On
the right pane, select Monitored clusters . Select the cluster to view the health of the nodes, user pods, and
system pods.

AKS - Nodes view - In Azure portal, open navigate to the cluster. Select Insights under Monitoring . View
Nodes on the right pane.
 Prometheus and Grafana Dashboard . Open the Node Conditions dashboard.

2- Verify the control plane and worker node connectivity

If worker nodes are healthy, examine the connectivity between the managed AKS control plane and the worker
nodes. Depending on the age and type of cluster configuration, the connectivity pods are either tunnelfront or
aks-link , and located in the kube-system namespace.
Tools
kubectl
Azure Monitor for Containers

If tunnelfront or aks-link connectivity is not working, establish connectivity after checking that the appropriate
AKS egress traffic rules have been allowed. Here are the steps:
1. Restart tunnelfront or aks-link .

kubectl rollout restart deploy aks-link

If restarting the pods doesn't fix the connection, continue to the next step.
2. Check the logs and look for abnormalities. This output shows logs for a working connection.

kubectl logs -l app=aks-link -c openvpn-client --tail=50

You can also retrieve those logs by searching the container logs in the logging and monitoring service. This
example searches Azure Monitor for Containers to check for aks-link connectivity errors.

let ContainerIDs = KubePodInventory

| where ClusterId =~
"/subscriptions/YOUR_SUBSCRIPTION_ID/resourceGroups/RESOURCE_GROUP/providers/Microsoft.ContainerService/manage
dClusters/YOUR_CLUSTER_ID"
| where Name has "aks-link"
| distinct ContainerID;
ContainerLog
| where ContainerID in (ContainerIDs)
| project LogEntrySource, LogEntry, TimeGenerated, Computer, Image, Name, ContainerID
| order by TimeGenerated desc
| limit 200

Here's another example query to check for tunnelfront connectivity errors.

let ContainerIDs = KubePodInventory

| where ClusterId =~
"/subscriptions/YOUR_SUBSCRIPTION_ID/resourceGroups/RESOURCE_GROUP/providers/Microsoft.ContainerService/manage
dClusters/YOUR_CLUSTER_ID"
| where Name has "tunnelfront"
| distinct ContainerID;
ContainerLog
| where ContainerID in (ContainerIDs)
| where LogEntry has "ssh to tunnelend is not connected"
| project LogEntrySource, LogEntry, TimeGenerated, Computer, Image, Name, ContainerID
| order by TimeGenerated desc
| limit 200

If you can't get the logs through the kubectl or queries, use SSH into the node. This example finds the tunnelfront
pod after connecting to the node through SSH.

kubectl pods -n kube-system -o wide | grep tunnelfront

ssh azureuser@<agent node pod is on, output from step above>
docker ps | grep tunnelfront
docker logs …
nslookup <ssh-server_fqdn>
ssh -vv azureuser@<ssh-server_fqdn> -p 9000
docker exec -it <tunnelfront_container_id> /bin/bash -c "ping bing.com"
kubectl get pods -n kube-system -o wide | grep <agent_node_where_tunnelfront_is_running>
kubectl delete po <kube_proxy_pod> -n kube-system
3- Validate DNS resolution when restricting egress
DNS resolution is a critical component of your cluster. If DNS resolution isn't working, then control plane errors or
container image pull failures may occur.
Tools
nslookup
dig

Follow these steps to make sure that DNS resolution is working.

1. Exec into the pod to examine and use nslookup or dig if those tools are installed on the pod.
2. If the pod doesn't have those tools, start a utility pod in the same namespace and retry with the tools.
3. If those steps don't show insights, SSH to one of the nodes and try resolution from there. This step will help
determine if the issue is related to AKS related or networking configuration.
4. If DNS resolves from the node, then the issue is related to Kubernetes DNS and not a networking issue. Restart
Kubernetes DNS and check whether the issue is resolved. If not, open a Microsoft support ticket.
5. If DNS doesn't resolve from the node, then check the networking setup again to make sure that the appropriate
routing paths and ports are open.

4- Check for kubelet errors

Check the kubelet process running on each worker node and make sure it's not experiencing any pressures. Those
pressures can be related to CPU, memory, or storage.
Tools
AKS - Kubelet Workbook
 Prometheus and Grafana Dashboard - Kubelet Dashboard

The pressure increases when kubelet restarts and causes some sporadic, unpredictable behavior. Make sure that
the error count isn't continuously growing. An occasional error is acceptable but a constant growth indicates an
underlying issue that needs to be investigated and resolved.

5- Check disk IOPS for throttling

Check to see that file operations (IOPS) are not getting throttled and impacting services in the cluster.
Tools
Azure Monitor for Containers Disk IO Workbook
 Prometheus and Grafana Dashboard - Node Disk Dashboard

Physical storage devices have limitations, bandwidth, and total number of file operations. Azure Disks are used to
store the OS running on the AKS nodes. They are subject to the same physical storage limitations.
Another way is to look at the throughput. IOPSs is measured with the average IO size * IOPS as the throughput in
MB/s. Large IO sizes will lead to lower IOPS because the throughput of a disk doesn't change.
When a workload exceeds Azure Disks' service limits on Max IOPS, the cluster becomes unresponsive and blocked
in IO Wait. Everything on Linux is a file. This includes network sockets, CNI, Docker, and other services that use
network I/O. All of those files will fail if they're unable to read the disk.
The events that can trigger IOPS throttle include:
High volume of containers running on the nodes. Docker IO is shared on the OS disk.
Custom or third-party tools used for security, monitoring, logging that are running on the OS disk.
Node failover events, and periodic jobs. As the load increases or the pods are scaled, this throttling occurs more
frequently until all nodes go to NotReady state while the IO completes.
Related links
Virtual machine disk limits
Relationship between Virtual Machine and Disk Performance

Next steps
Check the workload deployments
 Check the workload deployments
11/2/2020 • 2 minutes to read • Edit Online

Check to see that all deployments and daemonSets are running. The Ready and the Available matches the
expected count.
This article is part of a series. Read the introduction here.
Tools
AKS - Workloads . In Azure portal, navigate to the AKS cluster resource. Select Workloads .

Prometheus and Grafana Dashboard . Deployment Status Dashboard. This image is from Grafana

Community Chart 10856.

Next steps
Validate the admission controllers
 Validate the admission controllers are working as
expected
11/2/2020 • 2 minutes to read • Edit Online

Check whether the admission controllers are working as expected.

This article is part of a series. Read the introduction here.
Issues because of admission controllers are an edge case but they should be considered. Here are some examples:
Mutating and validating webhooks. Be careful when you add mutating and validating webhooks in your
cluster. Make sure that they're highly available so that an unhealthy node doesn't cause API server requests
to be blocked. AKS Policy, also known as OPA Gatekeeper, is an example of this type of webhooks. If there
are problems in the admission control pipeline, it can block a large number of requests to the API server.
Service meshes. They use admission controllers to automatically inject sidecars for example.
Tools kubectl

These commands check if AKS Policy is running in your cluster and how to validate that all of the admission
controllers are functioning as expected.

# Check AKS Policy pods are running.

kubectl get po -n gatekeeper-system

# Sample Output
...
NAME READY STATUS RESTARTS AGE
gatekeeper-audit-65844778cb-rkflg 1/1 Running 0 163m
gatekeeper-controller-78797d4687-4pf6w 1/1 Running 0 163m
gatekeeper-controller-78797d4687-splzh 1/1 Running 0 163m
...

If this command doesn't run as expected, it could indicate that an admission controller, API service, or CRD isn't
functioning correctly.

# Check that all API Resources are working correctly.

kubectl api-resources

# Sample Output
...
NAME SHORTNAMES APIGROUP NAMESPACED KIND
bindings true Binding
componentstatuses cs false
ComponentStatus
configmaps cm true ConfigMap
...

Next steps
Verify the connection to the container registry
 Verify the connection to the container registry
11/2/2020 • 2 minutes to read • Edit Online

Make sure that the worker nodes have the correct permission to pull the necessary container images from the
container registry.
This article is part of a series. Read the introduction here.
A common symptom of this issue is receiving ImagePullBackoff errors when getting or describing a pod. Be sure
that the registry and image name are correct. Also, the cluster has permissions to pull from the appropriate
container registry.
If you are using Azure Container Registry (ACR), the cluster service principal or managed identity should be
granted AcrPull permissions against the registry.
One way is to run this command using the managed identity of the AKS cluster node pool. This command gets a
list of its permissions.

# Get Kubelet Identity (Nodepool MSI)

ASSIGNEE=$(az aks show -g $RESOURCE_GROUP -n $NAME --query identityProfile.kubeletidentity.clientId -o tsv)
az role assignment list --assignee $ASSIGNEE --all -o table

# Expected Output
...
e5615a90-1767-4a4f-83b6-cecfa0675970 AcrPull
/subscriptions/.../providers/Microsoft.ContainerRegistry/registries/akskhacr
...

If you're using another container registry, check the appropriate ImagePullSecret credentials for the registry.

Related links
Import container images to a container registry
AKS Roadmap
 Azure Data Architecture Guide
11/2/2020 • 2 minutes to read • Edit Online

This guide presents a structured approach for designing data-centric solutions on Microsoft Azure. It is based on
proven practices derived from customer engagements.

Introduction
The cloud is changing the way applications are designed, including how data is processed and stored. Instead of a
single general-purpose database that handles all of a solution's data, polyglot persistence solutions use multiple,
specialized data stores, each optimized to provide specific capabilities. The perspective on data in the solution
changes as a result. There are no longer multiple layers of business logic that read and write to a single data layer.
Instead, solutions are designed around a data pipeline that describes how data flows through a solution, where it is
processed, where it is stored, and how it is consumed by the next component in the pipeline.

How this guide is structured

This guide is structured around two general categories of data solution, traditional RDBMS workloads and big data
solutions.
Traditional RDBMS workloads . These workloads include online transaction processing (OLTP) and online
analytical processing (OLAP). Data in OLTP systems is typically relational data with a predefined schema and a set
of constraints to maintain referential integrity. Often, data from multiple sources in the organization may be
consolidated into a data warehouse, using an ETL process to move and transform the source data.
OLTP Data Warehouse Analysis Reporting

ETL

Big data solutions . A big data architecture is designed to handle the ingestion, processing, and analysis of data
that is too large or complex for traditional database systems. The data may be processed in batch or in real time.
Big data solutions typically involve a large amount of non-relational data, such as key-value data, JSON documents,
or time series data. Often traditional RDBMS systems are not well-suited to store this type of data. The term NoSQL
refers to a family of databases designed to hold non-relational data. The term isn't quite accurate, because many
non-relational data stores support SQL compatible queries. The term NoSQL stands for "Not only SQL".
 Ingestion Processing Storage
ML

Reporting

These two categories are not mutually exclusive, and there is overlap between them, but we feel that it's a useful
way to frame the discussion. Within each category, the guide discusses common scenarios , including relevant
Azure services and the appropriate architecture for the scenario. In addition, the guide compares technology
choices for data solutions in Azure, including open source options. Within each category, we describe the key
selection criteria and a capability matrix, to help you choose the right technology for your scenario.
This guide is not intended to teach you data science or database theory — you can find entire books on those
subjects. Instead, the goal is to help you select the right data architecture or data pipeline for your scenario, and
then select the Azure services and technologies that best fit your requirements. If you already have an architecture
in mind, you can skip directly to the technology choices.
 Extract, transform, and load (ETL)
12/18/2020 • 5 minutes to read • Edit Online

A common problem that organizations face is how to gather data from multiple sources, in multiple formats, and
move it to one or more data stores. The destination may not be the same type of data store as the source, and often
the format is different, or the data needs to be shaped or cleaned before loading it into its final destination.
Various tools, services, and processes have been developed over the years to help address these challenges. No
matter the process used, there is a common need to coordinate the work and apply some level of data
transformation within the data pipeline. The following sections highlight the common methods used to perform
these tasks.

Extract, transform, and load (ETL) process

Extract, transform, and load (ETL) is a data pipeline used to collect data from various sources, transform the data
according to business rules, and load it into a destination data store. The transformation work in ETL takes place in
a specialized engine, and often involves using staging tables to temporarily hold data as it is being transformed and
ultimately loaded to its destination.
The data transformation that takes place usually involves various operations, such as filtering, sorting, aggregating,
joining data, cleaning data, deduplicating, and validating data.

Often, the three ETL phases are run in parallel to save time. For example, while data is being extracted, a
transformation process could be working on data already received and prepare it for loading, and a loading process
can begin working on the prepared data, rather than waiting for the entire extraction process to complete.
Relevant Azure service:
Azure Data Factory v2
Other tools:
SQL Server Integration Services (SSIS)

Extract, load, and transform (ELT)

Extract, load, and transform (ELT) differs from ETL solely in where the transformation takes place. In the ELT pipeline,
the transformation occurs in the target data store. Instead of using a separate transformation engine, the
processing capabilities of the target data store are used to transform data. This simplifies the architecture by
removing the transformation engine from the pipeline. Another benefit to this approach is that scaling the target
data store also scales the ELT pipeline performance. However, ELT only works well when the target system is
powerful enough to transform the data efficiently.

Typical use cases for ELT fall within the big data realm. For example, you might start by extracting all of the source
data to flat files in scalable storage such as Hadoop distributed file system (HDFS) or Azure Data Lake Store.
Technologies such as Spark, Hive, or PolyBase can then be used to query the source data. The key point with ELT is
that the data store used to perform the transformation is the same data store where the data is ultimately
consumed. This data store reads directly from the scalable storage, instead of loading the data into its own
proprietary storage. This approach skips the data copy step present in ETL, which can be a time consuming
operation for large data sets.
In practice, the target data store is a data warehouse using either a Hadoop cluster (using Hive or Spark) or a Azure
Synapse Analytics. In general, a schema is overlaid on the flat file data at query time and stored as a table, enabling
the data to be queried like any other table in the data store. These are referred to as external tables because the
data does not reside in storage managed by the data store itself, but on some external scalable storage.
The data store only manages the schema of the data and applies the schema on read. For example, a Hadoop
cluster using Hive would describe a Hive table where the data source is effectively a path to a set of files in HDFS. In
Azure Synapse, PolyBase can achieve the same result — creating a table against data stored externally to the
database itself. Once the source data is loaded, the data present in the external tables can be processed using the
capabilities of the data store. In big data scenarios, this means the data store must be capable of massively parallel
processing (MPP), which breaks the data into smaller chunks and distributes processing of the chunks across
multiple machines in parallel.
The final phase of the ELT pipeline is typically to transform the source data into a final format that is more efficient
for the types of queries that need to be supported. For example, the data may be partitioned. Also, ELT might use
optimized storage formats like Parquet, which stores row-oriented data in a columnar fashion and provides
optimized indexing.
Relevant Azure service:
Azure Synapse
HDInsight with Hive
Azure Data Factory v2
Oozie on HDInsight
Other tools:
SQL Server Integration Services (SSIS)

Data flow and control flow

In the context of data pipelines, the control flow ensures orderly processing of a set of tasks. To enforce the correct
processing order of these tasks, precedence constraints are used. You can think of these constraints as connectors
in a workflow diagram, as shown in the image below. Each task has an outcome, such as success, failure, or
completion. Any subsequent task does not initiate processing until its predecessor has completed with one of these
outcomes.
Control flows execute data flows as a task. In a data flow task, data is extracted from a source, transformed, or
loaded into a data store. The output of one data flow task can be the input to the next data flow task, and data flows
can run in parallel. Unlike control flows, you cannot add constraints between tasks in a data flow. You can, however,
add a data viewer to observe the data as it is processed by each task.

In the diagram above, there are several tasks within the control flow, one of which is a data flow task. One of the
tasks is nested within a container. Containers can be used to provide structure to tasks, providing a unit of work.
One such example is for repeating elements within a collection, such as files in a folder or database statements.
Relevant Azure service:
Azure Data Factory v2
Other tools:
SQL Server Integration Services (SSIS)

Technology choices
Online Transaction Processing (OLTP) data stores
Online Analytical Processing (OLAP) data stores
Data warehouses
Pipeline orchestration

Next steps
The following reference architectures show end-to-end ELT pipelines on Azure:
Enterprise BI in Azure with Azure Synapse
Automated enterprise BI with Azure Synapse and Azure Data Factory
 Online analytical processing (OLAP)
12/18/2020 • 8 minutes to read • Edit Online

Online analytical processing (OLAP) is a technology that organizes large business databases and supports complex
analysis. It can be used to perform complex analytical queries without negatively affecting transactional systems.
The databases that a business uses to store all its transactions and records are called online transaction processing
(OLTP) databases. These databases usually have records that are entered one at a time. Often they contain a great
deal of information that is valuable to the organization. The databases that are used for OLTP, however, were not
designed for analysis. Therefore, retrieving answers from these databases is costly in terms of time and effort.
OLAP systems were designed to help extract this business intelligence information from the data in a highly
performant way. This is because OLAP databases are optimized for heavy read, low write workloads.

Semantic modeling
A semantic data model is a conceptual model that describes the meaning of the data elements it contains.
Organizations often have their own terms for things, sometimes with synonyms, or even different meanings for the
same term. For example, an inventory database might track a piece of equipment with an asset ID and a serial
number, but a sales database might refer to the serial number as the asset ID. There is no simple way to relate these
values without a model that describes the relationship.
Semantic modeling provides a level of abstraction over the database schema, so that users don't need to know the
underlying data structures. This makes it easier for end users to query data without performing aggregates and
joins over the underlying schema. Also, usually columns are renamed to more user-friendly names, so that the
context and meaning of the data are more obvious.
Semantic modeling is predominately used for read-heavy scenarios, such as analytics and business intelligence
(OLAP), as opposed to more write-heavy transactional data processing (OLTP). This is mostly due to the nature of a
typical semantic layer:
Aggregation behaviors are set so that reporting tools display them properly.
Business logic and calculations are defined.
Time-oriented calculations are included.
Data is often integrated from multiple sources.
Traditionally, the semantic layer is placed over a data warehouse for these reasons.
There are two primary types of semantic models:
Tabular . Uses relational modeling constructs (model, tables, columns). Internally, metadata is inherited from
OLAP modeling constructs (cubes, dimensions, measures). Code and script use OLAP metadata.
Multidimensional . Uses traditional OLAP modeling constructs (cubes, dimensions, measures).
Relevant Azure service:
Azure Analysis Services

Example use case

An organization has data stored in a large database. It wants to make this data available to business users and
customers to create their own reports and do some analysis. One option is just to give those users direct access to
the database. However, there are several drawbacks to doing this, including managing security and controlling
access. Also, the design of the database, including the names of tables and columns, may be hard for a user to
understand. Users would need to know which tables to query, how those tables should be joined, and other
business logic that must be applied to get the correct results. Users would also need to know a query language like
SQL even to get started. Typically this leads to multiple users reporting the same metrics but with different results.
Another option is to encapsulate all of the information that users need into a semantic model. The semantic model
can be more easily queried by users with a reporting tool of their choice. The data provided by the semantic model
is pulled from a data warehouse, ensuring that all users see a single version of the truth. The semantic model also
provides friendly table and column names, relationships between tables, descriptions, calculations, and row-level
security.

Typical traits of semantic modeling

Semantic modeling and analytical processing tends to have the following traits:

REQ UIREM EN T DESC RIP T IO N

Schema Schema on write, strongly enforced

Uses Transactions No

Locking Strategy None

Updateable No (typically requires recomputing cube)

Appendable No (typically requires recomputing cube)

Workload Heavy reads, read-only

 REQ UIREM EN T DESC RIP T IO N

Indexing Multidimensional indexing

Datum size Small to medium sized

Model Multidimensional

Data shape: Cube or star/snowflake schema

Query flexibility Highly flexible

Scale: Large (10s-100s GBs)

When to use this solution

Consider OLAP in the following scenarios:
You need to execute complex analytical and ad hoc queries rapidly, without negatively affecting your OLTP
systems.
You want to provide business users with a simple way to generate reports from your data
You want to provide a number of aggregations that will allow users to get fast, consistent results.
OLAP is especially useful for applying aggregate calculations over large amounts of data. OLAP systems are
optimized for read-heavy scenarios, such as analytics and business intelligence. OLAP allows users to segment
multi-dimensional data into slices that can be viewed in two dimensions (such as a pivot table) or filter the data by
specific values. This process is sometimes called "slicing and dicing" the data, and can be done regardless of
whether the data is partitioned across several data sources. This helps users to find trends, spot patterns, and
explore the data without having to know the details of traditional data analysis.
Semantic models can help business users abstract relationship complexities and make it easier to analyze data
quickly.

Challenges
For all the benefits OLAP systems provide, they do produce a few challenges:
Whereas data in OLTP systems is constantly updated through transactions flowing in from various sources,
OLAP data stores are typically refreshed at a much slower intervals, depending on business needs. This means
OLAP systems are better suited for strategic business decisions, rather than immediate responses to changes.
Also, some level of data cleansing and orchestration needs to be planned to keep the OLAP data stores up-to-
date.
Unlike traditional, normalized, relational tables found in OLTP systems, OLAP data models tend to be
multidimensional. This makes it difficult or impossible to directly map to entity-relationship or object-oriented
models, where each attribute is mapped to one column. Instead, OLAP systems typically use a star or snowflake
schema in place of traditional normalization.

OLAP in Azure
In Azure, data held in OLTP systems such as Azure SQL Database is copied into the OLAP system, such as Azure
Analysis Services. Data exploration and visualization tools like Power BI, Excel, and third-party options connect to
Analysis Services servers and provide users with highly interactive and visually rich insights into the modeled data.
The flow of data from OLTP data to OLAP is typically orchestrated using SQL Server Integration Services, which can
be executed using Azure Data Factory.
In Azure, all of the following data stores will meet the core requirements for OLAP:
SQL Server with Columnstore indexes
Azure Analysis Services
SQL Server Analysis Services (SSAS)
SQL Server Analysis Services (SSAS) offers OLAP and data mining functionality for business intelligence
applications. You can either install SSAS on local servers, or host within a virtual machine in Azure. Azure Analysis
Services is a fully managed service that provides the same major features as SSAS. Azure Analysis Services
supports connecting to various data sources in the cloud and on-premises in your organization.
Clustered Columnstore indexes are available in SQL Server 2014 and above, as well as Azure SQL Database, and
are ideal for OLAP workloads. However, beginning with SQL Server 2016 (including Azure SQL Database), you can
take advantage of hybrid transactional/analytics processing (HTAP) through the use of updateable nonclustered
columnstore indexes. HTAP enables you to perform OLTP and OLAP processing on the same platform, which
removes the need to store multiple copies of your data, and eliminates the need for distinct OLTP and OLAP
systems. For more information, see Get started with Columnstore for real-time operational analytics.

Key selection criteria

To narrow the choices, start by answering these questions:
Do you want a managed service rather than managing your own servers?
Do you require secure authentication using Azure Active Directory (Azure AD)?
Do you want to conduct real-time analytics? If so, narrow your options to those that support real-time
analytics.
Real-time analytics in this context applies to a single data source, such as an enterprise resource planning
(ERP) application, that will run both an operational and an analytics workload. If you need to integrate data
from multiple sources, or require extreme analytics performance by using pre-aggregated data such as
cubes, you might still require a separate data warehouse.
Do you need to use pre-aggregated data, for example to provide semantic models that make analytics more
business user friendly? If yes, choose an option that supports multidimensional cubes or tabular semantic
models.
Providing aggregates can help users consistently calculate data aggregates. Pre-aggregated data can also
provide a large performance boost when dealing with several columns across many rows. Data can be pre-
aggregated in multidimensional cubes or tabular semantic models.
Do you need to integrate data from several sources, beyond your OLTP data store? If so, consider options
that easily integrate multiple data sources.

Capability matrix
The following tables summarize the key differences in capabilities.
General capabilities
 A Z URE SQ L
SQ L SERVER W IT H DATA B A SE W IT H
A Z URE A N A LY SIS SQ L SERVER C O L UM N STO RE C O L UM N STO RE
C A PA B IL IT Y SERVIC ES A N A LY SIS SERVIC ES IN DEXES IN DEXES

Is managed service Yes No No Yes

Supports No Yes No No
multidimensional
cubes

Supports tabular Yes Yes No No

semantic models

Easily integrate Yes Yes No 1 No 1

multiple data sources

Supports real-time No No Yes Yes

analytics

Requires process to Yes Yes No No

copy data from
source(s)

Azure AD integration Yes No No 2 Yes

[1] Although SQL Server and Azure SQL Database cannot be used to query from and integrate multiple external
data sources, you can still build a pipeline that does this for you using SSIS or Azure Data Factory. SQL Server
hosted in an Azure VM has additional options, such as linked servers and PolyBase. For more information, see
Pipeline orchestration, control flow, and data movement.
[2] Connecting to SQL Server running on an Azure Virtual Machine is not supported using an Azure AD account.
Use a domain Active Directory account instead.
Scalability Capabilities
A Z URE SQ L
SQ L SERVER W IT H DATA B A SE W IT H
A Z URE A N A LY SIS SQ L SERVER C O L UM N STO RE C O L UM N STO RE
C A PA B IL IT Y SERVIC ES A N A LY SIS SERVIC ES IN DEXES IN DEXES

Redundant regional Yes No Yes Yes

servers for high
availability

Supports query scale Yes No Yes Yes

out

Dynamic scalability Yes No Yes Yes

(scale up)
 Online transaction processing (OLTP)
12/18/2020 • 6 minutes to read • Edit Online

The management of transactional data using computer systems is referred to as online transaction processing
(OLTP). OLTP systems record business interactions as they occur in the day-to-day operation of the organization,
and support querying of this data to make inferences.

Transactional data
Transactional data is information that tracks the interactions related to an organization's activities. These
interactions are typically business transactions, such as payments received from customers, payments made to
suppliers, products moving through inventory, orders taken, or services delivered. Transactional events, which
represent the transactions themselves, typically contain a time dimension, some numerical values, and references
to other data.
Transactions typically need to be atomic and consistent. Atomicity means that an entire transaction always
succeeds or fails as one unit of work, and is never left in a half-completed state. If a transaction cannot be
completed, the database system must roll back any steps that were already done as part of that transaction. In a
traditional RDBMS, this rollback happens automatically if a transaction cannot be completed. Consistency means
that transactions always leave the data in a valid state. (These are very informal descriptions of atomicity and
consistency. There are more formal definitions of these properties, such as ACID.)
Transactional databases can support strong consistency for transactions using various locking strategies, such as
pessimistic locking, to ensure that all data is strongly consistent within the context of the enterprise, for all users
and processes.
The most common deployment architecture that uses transactional data is the data store tier in a 3-tier
architecture. A 3-tier architecture typically consists of a presentation tier, business logic tier, and data store tier. A
related deployment architecture is the N-tier architecture, which may have multiple middle-tiers handling business
logic.

Typical traits of transactional data

Transactional data tends to have the following traits:

REQ UIREM EN T DESC RIP T IO N

Normalization Highly normalized

Schema Schema on write, strongly enforced

Consistency Strong consistency, ACID guarantees

Integrity High integrity

Uses transactions Yes

Locking strategy Pessimistic or optimistic

Updateable Yes
 REQ UIREM EN T DESC RIP T IO N

Appendable Yes

Workload Heavy writes, moderate reads

Indexing Primary and secondary indexes

Datum size Small to medium sized

Model Relational

Data shape Tabular

Query flexibility Highly flexible

Scale Small (MBs) to Large (a few TBs)

When to use this solution

Choose OLTP when you need to efficiently process and store business transactions and immediately make them
available to client applications in a consistent way. Use this architecture when any tangible delay in processing
would have a negative impact on the day-to-day operations of the business.
OLTP systems are designed to efficiently process and store transactions, as well as query transactional data. The
goal of efficiently processing and storing individual transactions by an OLTP system is partly accomplished by data
normalization — that is, breaking the data up into smaller chunks that are less redundant. This supports efficiency
because it enables the OLTP system to process large numbers of transactions independently, and avoids extra
processing needed to maintain data integrity in the presence of redundant data.

Challenges
Implementing and using an OLTP system can create a few challenges:
OLTP systems are not always good for handling aggregates over large amounts of data, although there are
exceptions, such as a well-planned SQL Server-based solution. Analytics against the data, that rely on aggregate
calculations over millions of individual transactions, are very resource intensive for an OLTP system. They can
be slow to execute and can cause a slow-down by blocking other transactions in the database.
When conducting analytics and reporting on data that is highly normalized, the queries tend to be complex,
because most queries need to de-normalize the data by using joins. Also, naming conventions for database
objects in OLTP systems tend to be terse and succinct. The increased normalization coupled with terse naming
conventions makes OLTP systems difficult for business users to query, without the help of a DBA or data
developer.
Storing the history of transactions indefinitely and storing too much data in any one table can lead to slow
query performance, depending on the number of transactions stored. The common solution is to maintain a
relevant window of time (such as the current fiscal year) in the OLTP system and offload historical data to other
systems, such as a data mart or data warehouse.

OLTP in Azure
Applications such as websites hosted in App Service Web Apps, REST APIs running in App Service, or mobile or
desktop applications communicate with the OLTP system, typically via a REST API intermediary.
In practice, most workloads are not purely OLTP. There tends to be an analytical component as well. In addition,
there is an increasing demand for real-time reporting, such as running reports against the operational system. This
is also referred to as HTAP (Hybrid Transactional and Analytical Processing). For more information, see Online
Analytical Processing (OLAP).
In Azure, all of the following data stores will meet the core requirements for OLTP and the management of
transaction data:
Azure SQL Database
SQL Server in an Azure virtual machine
Azure Database for MySQL
Azure Database for PostgreSQL

Key selection criteria

To narrow the choices, start by answering these questions:
Do you want a managed service rather than managing your own servers?
Does your solution have specific dependencies for Microsoft SQL Server, MySQL or PostgreSQL
compatibility? Your application may limit the data stores you can choose based on the drivers it supports for
communicating with the data store, or the assumptions it makes about which database is used.
Are your write throughput requirements particularly high? If yes, choose an option that provides in-
memory tables.
Is your solution multitenant? If so, consider options that support capacity pools, where multiple database
instances draw from an elastic pool of resources, instead of fixed resources per database. This can help you
better distribute capacity across all database instances, and can make your solution more cost effective.
Does your data need to be readable with low latency in multiple regions? If yes, choose an option that
supports readable secondary replicas.
Does your database need to be highly available across geo-graphic regions? If yes, choose an option that
supports geographic replication. Also consider the options that support automatic failover from the primary
replica to a secondary replica.
Does your database have specific security needs? If yes, examine the options that provide capabilities like
row level security, data masking, and transparent data encryption.

Capability matrix
The following tables summarize the key differences in capabilities.
General capabilities
SQ L SERVER IN A N
A Z URE SQ L A Z URE VIRT UA L A Z URE DATA B A SE A Z URE DATA B A SE
C A PA B IL IT Y DATA B A SE M A C H IN E F O R M Y SQ L F O R P O STGRESQ L

Is Managed Service Yes No Yes Yes

Runs on Platform N/A Windows, Linux, N/A N/A

Docker

Programmability 1 T-SQL, .NET, R T-SQL, .NET, R, SQL SQL, PL/pgSQL

Python
[1] Not including client driver support, which allows many programming languages to connect to and use the
OLTP data store.
Scalability capabilities
SQ L SERVER IN A N
A Z URE SQ L A Z URE VIRT UA L A Z URE DATA B A SE A Z URE DATA B A SE
C A PA B IL IT Y DATA B A SE M A C H IN E F O R M Y SQ L F O R P O STGRESQ L

Maximum database 4 TB 256 TB 16 TB 16 TB

instance size

Supports capacity Yes Yes No No

pools

Supports clusters No Yes No No

scale out

Dynamic scalability Yes No Yes Yes

(scale up)

Analytic workload capabilities

SQ L SERVER IN A N
A Z URE SQ L A Z URE VIRT UA L A Z URE DATA B A SE A Z URE DATA B A SE
C A PA B IL IT Y DATA B A SE M A C H IN E F O R M Y SQ L F O R P O STGRESQ L

Temporal tables Yes Yes No No

In-memory Yes Yes No No

(memory-optimized)
tables

Columnstore support Yes Yes No No

Adaptive query Yes Yes No No

processing

Availability capabilities
SQ L SERVER IN A N
A Z URE SQ L A Z URE VIRT UA L A Z URE DATA B A SE A Z URE DATA B A SE
C A PA B IL IT Y DATA B A SE M A C H IN E F O R M Y SQ L F O R P O STGRESQ L

Readable secondaries Yes Yes Yes Yes

Geographic Yes Yes Yes Yes

replication

Automatic failover to Yes No No No

secondary

Point-in-time restore Yes Yes Yes Yes

Security capabilities
 SQ L SERVER IN A N
A Z URE SQ L A Z URE VIRT UA L A Z URE DATA B A SE A Z URE DATA B A SE
C A PA B IL IT Y DATA B A SE M A C H IN E F O R M Y SQ L F O R P O STGRESQ L

Row level security Yes Yes Yes Yes

Data masking Yes Yes No No

Transparent data Yes Yes Yes Yes

encryption

Restrict access to Yes Yes Yes Yes

specific IP addresses

Restrict access to Yes Yes Yes Yes

allow VNet access
only

Azure Active Yes No Yes Yes

Directory
authentication

Active Directory No Yes No No

authentication

Multi-factor Yes No Yes Yes

authentication

Supports Always Yes Yes No No

Encrypted

Private IP No Yes No No
 Data warehousing
12/18/2020 • 11 minutes to read • Edit Online

A data warehouse is a centralized repository of integrated data from one or more disparate sources. Data
warehouses store current and historical data and are used for reporting and analysis of the data.

To move data into a data warehouse, data is periodically extracted from various sources that contain important
business information. As the data is moved, it can be formatted, cleaned, validated, summarized, and reorganized.
Alternatively, the data can be stored in the lowest level of detail, with aggregated views provided in the warehouse
for reporting. In either case, the data warehouse becomes a permanent data store for reporting, analysis, and
business intelligence (BI).

Data warehouse architectures

The following reference architectures show end-to-end data warehouse architectures on Azure:
Enterprise BI in Azure with Azure Synapse Analytics. This reference architecture implements an extract, load,
and transform (ELT) pipeline that moves data from an on-premises SQL Server database into Azure Synapse.
Automated enterprise BI with Azure Synapse and Azure Data Factory. This reference architecture shows an ELT
pipeline with incremental loading, automated using Azure Data Factory.

When to use this solution

Choose a data warehouse when you need to turn massive amounts of data from operational systems into a format
that is easy to understand. Data warehouses don't need to follow the same terse data structure you may be using
in your OLTP databases. You can use column names that make sense to business users and analysts, restructure
the schema to simplify relationships, and consolidate several tables into one. These steps help guide users who
need to create reports and analyze the data in BI systems, without the help of a database administrator (DBA) or
data developer.
Consider using a data warehouse when you need to keep historical data separate from the source transaction
systems for performance reasons. Data warehouses make it easy to access historical data from multiple locations,
by providing a centralized location using common formats, keys, and data models.
Because data warehouses are optimized for read access, generating reports is faster than using the source
transaction system for reporting.
Other benefits include:
 The data warehouse can store historical data from multiple sources, representing a single source of truth.
You can improve data quality by cleaning up data as it is imported into the data warehouse.
Reporting tools don't compete with the transactional systems for query processing cycles. A data warehouse
allows the transactional system to focus on handling writes, while the data warehouse satisfies the majority of
read requests.
A data warehouse can consolidate data from different software.
Data mining tools can find hidden patterns in the data using automatic methodologies.
Data warehouses make it easier to provide secure access to authorized users, while restricting access to others.
Business users don't need access to the source data, removing a potential attack vector.
Data warehouses make it easier to create business intelligence solutions, such as OLAP cubes.

Challenges
Properly configuring a data warehouse to fit the needs of your business can bring some of the following
challenges:
Committing the time required to properly model your business concepts. Data warehouses are information
driven. You must standardize business-related terms and common formats, such as currency and dates. You
also need to restructure the schema in a way that makes sense to business users but still ensures accuracy
of data aggregates and relationships.
Planning and setting up your data orchestration. Consider how to copy data from the source transactional
system to the data warehouse, and when to move historical data from operational data stores into the
warehouse.
Maintaining or improving data quality by cleaning the data as it is imported into the warehouse.

Data warehousing in Azure

You may have one or more sources of data, whether from customer transactions or business applications. This
data is traditionally stored in one or more OLTP databases. The data could be persisted in other storage mediums
such as network shares, Azure Storage Blobs, or a data lake. The data could also be stored by the data warehouse
itself or in a relational database such as Azure SQL Database. The purpose of the analytical data store layer is to
satisfy queries issued by analytics and reporting tools against the data warehouse. In Azure, this analytical store
capability can be met with Azure Synapse, or with Azure HDInsight using Hive or Interactive Query. In addition,
you will need some level of orchestration to move or copy data from data storage to the data warehouse, which
can be done using Azure Data Factory or Oozie on Azure HDInsight.
There are several options for implementing a data warehouse in Azure, depending on your needs. The following
lists are broken into two categories, symmetric multiprocessing (SMP) and massively parallel processing (MPP).
SMP:
Azure SQL Database
SQL Server in a virtual machine
MPP:
Azure Synapse Analytics (formerly Azure Data Warehouse)
Apache Hive on HDInsight
Interactive Query (Hive LLAP) on HDInsight
As a general rule, SMP-based warehouses are best suited for small to medium data sets (up to 4-100 TB), while
MPP is often used for big data. The delineation between small/medium and big data partly has to do with your
organization's definition and supporting infrastructure. (See Choosing an OLTP data store.)
Beyond data sizes, the type of workload pattern is likely to be a greater determining factor. For example, complex
queries may be too slow for an SMP solution, and require an MPP solution instead. MPP-based systems usually
have a performance penalty with small data sizes, because of how jobs are distributed and consolidated across
nodes. If your data sizes already exceed 1 TB and are expected to continually grow, consider selecting an MPP
solution. However, if your data sizes are smaller, but your workloads are exceeding the available resources of your
SMP solution, then MPP may be your best option as well.
The data accessed or stored by your data warehouse could come from a number of data sources, including a data
lake, such as Azure Data Lake Storage. For a video session that compares the different strengths of MPP services
that can use Azure Data Lake, see Azure Data Lake and Azure Data Warehouse: Applying Modern Practices to Your
App.
SMP systems are characterized by a single instance of a relational database management system sharing all
resources (CPU/Memory/Disk). You can scale up an SMP system. For SQL Server running on a VM, you can scale
up the VM size. For Azure SQL Database, you can scale up by selecting a different service tier.
MPP systems can be scaled out by adding more compute nodes (which have their own CPU, memory, and I/O
subsystems). There are physical limitations to scaling up a server, at which point scaling out is more desirable,
depending on the workload. However, the differences in querying, modeling, and data partitioning mean that MPP
solutions require a different skill set.
When deciding which SMP solution to use, see A closer look at Azure SQL Database and SQL Server on Azure
VMs.
Azure Synapse (formerly Azure SQL Data Warehouse) can also be used for small and medium datasets, where the
workload is compute and memory intensive. Read more about Azure Synapse patterns and common scenarios:
Azure SQL Data Warehouse Workload Patterns and Anti-Patterns
Azure SQL Data Warehouse loading patterns and strategies
Migrating data to Azure SQL Data Warehouse in practice
Common ISV application patterns using Azure SQL Data Warehouse

Key selection criteria

To narrow the choices, start by answering these questions:
Do you want a managed service rather than managing your own servers?
Are you working with extremely large data sets or highly complex, long-running queries? If yes, consider an
MPP option.
For a large data set, is the data source structured or unstructured? Unstructured data may need to be
processed in a big data environment such as Spark on HDInsight, Azure Databricks, Hive LLAP on
HDInsight, or Azure Data Lake Analytics. All of these can serve as ELT (Extract, Load, Transform) and ETL
(Extract, Transform, Load) engines. They can output the processed data into structured data, making it easier
to load into Azure Synapse or one of the other options. For structured data, Azure Synapse has a
performance tier called Optimized for Compute, for compute-intensive workloads requiring ultra-high
performance.
Do you want to separate your historical data from your current, operational data? If so, select one of the
options where orchestration is required. These are standalone warehouses optimized for heavy read access,
and are best suited as a separate historical data store.
Do you need to integrate data from several sources, beyond your OLTP data store? If so, consider options
that easily integrate multiple data sources.
 Do you have a multitenancy requirement? If so, Azure Synapse is not ideal for this requirement. For more
information, see Azure Synapse Patterns and Anti-Patterns.
Do you prefer a relational data store? If so, choose an option with a relational data store, but also note that
you can use a tool like PolyBase to query non-relational data stores if needed. If you decide to use PolyBase,
however, run performance tests against your unstructured data sets for your workload.
Do you have real-time reporting requirements? If you require rapid query response times on high volumes
of singleton inserts, choose an option that supports real-time reporting.
Do you need to support a large number of concurrent users and connections? The ability to support a
number of concurrent users/connections depends on several factors.
For Azure SQL Database, refer to the documented resource limits based on your service tier.
SQL Server allows a maximum of 32,767 user connections. When running on a VM, performance will
depend on the VM size and other factors.
Azure Synapse has limits on concurrent queries and concurrent connections. For more information,
see Concurrency and workload management in Azure Synapse. Consider using complementary
services, such as Azure Analysis Services, to overcome limits in Azure Synapse.
What sort of workload do you have? In general, MPP-based warehouse solutions are best suited for
analytical, batch-oriented workloads. If your workloads are transactional by nature, with many small
read/write operations or multiple row-by-row operations, consider using one of the SMP options. One
exception to this guideline is when using stream processing on an HDInsight cluster, such as Spark
Streaming, and storing the data within a Hive table.

Capability Matrix
The following tables summarize the key differences in capabilities.
General capabilities
A Z URE SQ L A PA C H E H IVE O N H IVE L L A P O N
C A PA B IL IT Y DATA B A SE SQ L SERVER ( VM ) A Z URE SY N A P SE H DIN SIGH T H DIN SIGH T

Is managed Yes No Yes Yes 1 Yes 1

service

Requires data No No Yes Yes Yes

orchestration
(holds copy of
data/historical
data)

Easily integrate No No Yes Yes Yes

multiple data
sources

Supports pausing No No Yes No 2 No 2

compute

Relational data Yes Yes Yes No No

store

Real-time Yes Yes No No Yes

reporting
 A Z URE SQ L A PA C H E H IVE O N H IVE L L A P O N
C A PA B IL IT Y DATA B A SE SQ L SERVER ( VM ) A Z URE SY N A P SE H DIN SIGH T H DIN SIGH T

Flexible backup Yes Yes No 3 Yes 4 Yes 4

restore points

SMP/MPP SMP SMP MPP MPP MPP

[1] Manual configuration and scaling.

[2] HDInsight clusters can be deleted when not needed, and then re-created. Attach an external data store to your
cluster so your data is retained when you delete your cluster. You can use Azure Data Factory to automate your
cluster's lifecycle by creating an on-demand HDInsight cluster to process your workload, then delete it once the
processing is complete.
[3] With Azure Synapse, you can restore a database to any available restore point within the last seven days.
Snapshots start every four to eight hours and are available for seven days. When a snapshot is older than seven
days, it expires and its restore point is no longer available.
[4] Consider using an external Hive metastore that can be backed up and restored as needed. Standard backup and
restore options that apply to Blob Storage or Data Lake Storage can be used for the data, or third-party HDInsight
backup and restore solutions, such as Imanis Data can be used for greater flexibility and ease of use.
Scalability capabilities
A Z URE SQ L A PA C H E H IVE O N H IVE L L A P O N
C A PA B IL IT Y DATA B A SE SQ L SERVER ( VM ) A Z URE SY N A P SE H DIN SIGH T H DIN SIGH T

Redundant Yes Yes Yes No No

regional servers
for high
availability

Supports query No No Yes Yes Yes

scale out
(distributed
queries)

Dynamic Yes No Yes 1 No No

scalability

Supports in- Yes Yes Yes Yes Yes

memory caching
of data

[1] Azure Synapse allows you to scale up or down by adjusting the number of data warehouse units (DWUs). See
Manage compute power in Azure Synapse.
Security capabilities
SQ L SERVER IN A
A Z URE SQ L VIRT UA L A PA C H E H IVE O N H IVE L L A P O N
C A PA B IL IT Y DATA B A SE M A C H IN E A Z URE SY N A P SE H DIN SIGH T H DIN SIGH T

Authentication SQL / Azure SQL / Azure AD / SQL / Azure AD local / Azure AD 1 local / Azure AD 1
Active Directory Active Directory
(Azure AD)
 SQ L SERVER IN A
A Z URE SQ L VIRT UA L A PA C H E H IVE O N H IVE L L A P O N
C A PA B IL IT Y DATA B A SE M A C H IN E A Z URE SY N A P SE H DIN SIGH T H DIN SIGH T

Authorization Yes Yes Yes Yes Yes 1

Auditing Yes Yes Yes Yes Yes 1

Data encryption Yes 2 Yes 2 Yes 2 Yes 2 Yes 1

at rest

Row-level Yes Yes Yes No Yes 1

security

Supports firewalls Yes Yes Yes Yes Yes 3

Dynamic data Yes Yes Yes No Yes 1

masking

[1] Requires using a domain-joined HDInsight cluster.

[2] Requires using Transparent Data Encryption (TDE) to encrypt and decrypt your data at rest.
[3] Supported when used within an Azure Virtual Network.
Read more about securing your data warehouse:
Securing your SQL Database
Secure a database in Azure Synapse
Extend Azure HDInsight using an Azure Virtual Network
Enterprise-level Hadoop security with domain-joined HDInsight clusters
 Non-relational data and NoSQL
12/18/2020 • 12 minutes to read • Edit Online

A non-relational database is a database that does not use the tabular schema of rows and columns found in most
traditional database systems. Instead, non-relational databases use a storage model that is optimized for the
specific requirements of the type of data being stored. For example, data may be stored as simple key/value pairs,
as JSON documents, or as a graph consisting of edges and vertices.
What all of these data stores have in common is that they don't use a relational model. Also, they tend to be more
specific in the type of data they support and how data can be queried. For example, time series data stores are
optimized for queries over time-based sequences of data, while graph data stores are optimized for exploring
weighted relationships between entities. Neither format would generalize well to the task of managing
transactional data.
The term NoSQL refers to data stores that do not use SQL for queries, and instead use other programming
languages and constructs to query the data. In practice, "NoSQL" means "non-relational database," even though
many of these databases do support SQL-compatible queries. However, the underlying query execution strategy is
usually very different from the way a traditional RDBMS would execute the same SQL query.
The following sections describe the major categories of non-relational or NoSQL database.

Document data stores

A document data store manages a set of named string fields and object data values in an entity referred to as a
document. These data stores typically store data in the form of JSON documents. Each field value could be a scalar
item, such as a number, or a compound element, such as a list or a parent-child collection. The data in the fields of a
document can be encoded in a variety of ways, including XML, YAML, JSON, BSON, or even stored as plain text. The
fields within documents are exposed to the storage management system, enabling an application to query and
filter data by using the values in these fields.
Typically, a document contains the entire data for an entity. What items constitute an entity are application-specific.
For example, an entity could contain the details of a customer, an order, or a combination of both. A single
document might contain information that would be spread across several relational tables in a relational database
management system (RDBMS). A document store does not require that all documents have the same structure. This
free-form approach provides a great deal of flexibility. For example, applications can store different data in
documents in response to a change in business requirements.
The application can retrieve documents by using the document key. This is a unique identifier for the document,
which is often hashed, to help distribute data evenly. Some document databases create the document key
automatically. Others enable you to specify an attribute of the document to use as the key. The application can also
query documents based on the value of one or more fields. Some document databases support indexing to
facilitate fast lookup of documents based on one or more indexed fields.
Many document databases support in-place updates, enabling an application to modify the values of specific fields
in a document without rewriting the entire document. Read and write operations over multiple fields in a single
document are typically atomic.
Relevant Azure service:
Azure Cosmos DB

Columnar data stores

A columnar or column-family data store organizes data into columns and rows. In its simplest form, a column-
family data store can appear very similar to a relational database, at least conceptually. The real power of a column-
family database lies in its denormalized approach to structuring sparse data, which stems from the column-
oriented approach to storing data.
You can think of a column-family data store as holding tabular data with rows and columns, but the columns are
divided into groups known as column families. Each column family holds a set of columns that are logically related
and are typically retrieved or manipulated as a unit. Other data that is accessed separately can be stored in separate
column families. Within a column family, new columns can be added dynamically, and rows can be sparse (that is, a
row doesn't need to have a value for every column).
The following diagram shows an example with two column families, Identity and Contact Info . The data for a
single entity has the same row key in each column family. This structure, where the rows for any given object in a
column family can vary dynamically, is an important benefit of the column-family approach, making this form of
data store highly suited for storing data with varying schemas.

Unlike a key/value store or a document database, most column-family databases physically store data in key order,
rather than by computing a hash. The row key is considered the primary index and enables key-based access via a
specific key or a range of keys. Some implementations allow you to create secondary indexes over specific columns
in a column family. Secondary indexes let you retrieve data by columns value, rather than row key.
On disk, all of the columns within a column family are stored together in the same file, with a certain number of
rows in each file. With large data sets, this approach creates a performance benefit by reducing the amount of data
that needs to be read from disk when only a few columns are queried together at a time.
Read and write operations for a row are typically atomic within a single column family, although some
implementations provide atomicity across the entire row, spanning multiple column families.
Relevant Azure service:
 Cosmos DB Cassandra API
HBase in HDInsight

Key/value data stores

A key/value store is essentially a large hash table. You associate each data value with a unique key, and the
key/value store uses this key to store the data by using an appropriate hashing function. The hashing function is
selected to provide an even distribution of hashed keys across the data storage.
Most key/value stores only support simple query, insert, and delete operations. To modify a value (either partially
or completely), an application must overwrite the existing data for the entire value. In most implementations,
reading or writing a single value is an atomic operation. If the value is large, writing may take some time.
An application can store arbitrary data as a set of values, although some key/value stores impose limits on the
maximum size of values. The stored values are opaque to the storage system software. Any schema information
must be provided and interpreted by the application. Essentially, values are blobs and the key/value store simply
retrieves or stores the value by key.

Key/value stores are highly optimized for applications performing simple lookups using the value of the key, or by
a range of keys, but are less suitable for systems that need to query data across different tables of keys/values, such
as joining data across multiple tables.
Key/value stores are also not optimized for scenarios where querying or filtering by non-key values is important,
rather than performing lookups based only on keys. For example, with a relational database, you can find a record
by using a WHERE clause to filter the non-key columns, but key/values stores usually do not have this type of
lookup capability for values, or if they do, it requires a slow scan of all values.
A single key/value store can be extremely scalable, as the data store can easily distribute data across multiple nodes
on separate machines.
Relevant Azure services:
Azure Cosmos DB Table API
Azure Cache for Redis
Azure Table Storage

Graph data stores

A graph data store manages two types of information, nodes and edges. Nodes represent entities, and edges
specify the relationships between these entities. Both nodes and edges can have properties that provide
information about that node or edge, similar to columns in a table. Edges can also have a direction indicating the
nature of the relationship.
The purpose of a graph data store is to allow an application to efficiently perform queries that traverse the network
of nodes and edges, and to analyze the relationships between entities. The following diagram shows an
organization's personnel data structured as a graph. The entities are employees and departments, and the edges
indicate reporting relationships and the department in which employees work. In this graph, the arrows on the
edges show the direction of the relationships.

This structure makes it straightforward to perform queries such as "Find all employees who report directly or
indirectly to Sarah" or "Who works in the same department as John?" For large graphs with lots of entities and
relationships, you can perform complex analyses quickly. Many graph databases provide a query language that you
can use to traverse a network of relationships efficiently.
Relevant Azure service:
Azure Cosmos DB Graph API

Time series data stores

Time series data is a set of values organized by time, and a time series data store is optimized for this type of data.
Time series data stores must support a very high number of writes, as they typically collect large amounts of data
in real time from a large number of sources. Time series data stores are optimized for storing telemetry data.
Scenarios include IoT sensors or application/system counters. Updates are rare, and deletes are often done as bulk
operations.

Although the records written to a time series database are generally small, there are often a large number of
records, and total data size can grow rapidly. Time series data stores also handle out-of-order and late-arriving data,
automatic indexing of data points, and optimizations for queries described in terms of windows of time. This last
feature enables queries to run across millions of data points and multiple data streams quickly, in order to support
time series visualizations, which is a common way that time series data is consumed.
For more information, see Time series solutions
Relevant Azure services:
 Azure Time Series Insights
OpenTSDB with HBase on HDInsight

Object data stores

Object data stores are optimized for storing and retrieving large binary objects or blobs such as images, text files,
video and audio streams, large application data objects and documents, and virtual machine disk images. An object
consists of the stored data, some metadata, and a unique ID for accessing the object. Object stores are designed to
support files that are individually very large, as well provide large amounts of total storage to manage all files.

Some object data stores replicate a given blob across multiple server nodes, which enables fast parallel reads. This
in turn enables the scale-out querying of data contained in large files, because multiple processes, typically running
on different servers, can each query the large data file simultaneously.
One special case of object data stores is the network file share. Using file shares enables files to be accessed across
a network using standard networking protocols like server message block (SMB). Given appropriate security and
concurrent access control mechanisms, sharing data in this way can enable distributed services to provide highly
scalable data access for basic, low-level operations such as simple read and write requests.
Relevant Azure services:
Azure Blob Storage
Azure Data Lake Store
Azure File Storage

External index data stores

External index data stores provide the ability to search for information held in other data stores and services. An
external index acts as a secondary index for any data store, and can be used to index massive volumes of data and
provide near real-time access to these indexes.
For example, you might have text files stored in a file system. Finding a file by its file path is quick, but searching
based on the contents of the file would require a scan of all of the files, which is slow. An external index lets you
create secondary search indexes and then quickly find the path to the files that match your criteria. Another
example application of an external index is with key/value stores that only index by the key. You can build a
secondary index based on the values in the data, and quickly look up the key that uniquely identifies each matched
item.
The indexes are created by running an indexing process. This can be performed using a pull model, triggered by the
data store, or using a push model, initiated by application code. Indexes can be multidimensional and may support
free-text searches across large volumes of text data.
External index data stores are often used to support full text and web-based search. In these cases, searching can be
exact or fuzzy. A fuzzy search finds documents that match a set of terms and calculates how closely they match.
Some external indexes also support linguistic analysis that can return matches based on synonyms, genre
expansions (for example, matching "dogs" to "pets"), and stemming (for example, searching for "run" also matches
"ran" and "running").
Relevant Azure service:
Azure Search

Typical requirements
Non-relational data stores often use a different storage architecture from that used by relational databases.
Specifically, they tend toward having no fixed schema. Also, they tend not to support transactions, or else restrict
the scope of transactions, and they generally don't include secondary indexes for scalability reasons.
The following compares the requirements for each of the non-relational data stores:

C O L UM N - FA M ILY
REQ UIREM EN T DO C UM EN T DATA DATA K EY / VA L UE DATA GRA P H DATA

Normalization Denormalized Denormalized Denormalized Normalized

Schema Schema on read Column families Schema on read Schema on read

defined on write,
column schema on
read

Consistency (across Tunable consistency, Column-family–level Key-level guarantees Graph-level

concurrent document-level guarantees guarantees
transactions) guarantees

Atomicity (transaction Collection Table Table Graph

scope)

Locking Strategy Optimistic (lock free) Pessimistic (row locks) Optimistic (ETag)

Access pattern Random access Aggregates on Random access Random access

tall/wide data
 C O L UM N - FA M ILY
REQ UIREM EN T DO C UM EN T DATA DATA K EY / VA L UE DATA GRA P H DATA

Indexing Primary and Primary and Primary index only Primary and
secondary indexes secondary indexes secondary indexes

Data shape Document Tabular with column Key and value Graph containing
families containing edges and vertices
columns

Sparse Yes Yes Yes No

Wide (lots of Yes Yes No No

columns/attributes)

Datum size Small (KBs) to Medium (MBs) to Small (KBs) Small (KBs)
medium (low MBs) Large (low GBs)

Overall Maximum Very Large (PBs) Very Large (PBs) Very Large (PBs) Large (TBs)
Scale

REQ UIREM EN T T IM E SERIES DATA O B JEC T DATA EXT ERN A L IN DEX DATA

Normalization Normalized Denormalized Denormalized

Schema Schema on read Schema on read Schema on write

Consistency (across N/A N/A N/A

concurrent transactions)

Atomicity (transaction scope) N/A Object N/A

Locking Strategy N/A Pessimistic (blob locks) N/A

Access pattern Random access and Sequential access Random access

aggregation

Indexing Primary and secondary Primary index only N/A

indexes

Data shape Tabular Blob and metadata Document

Sparse No N/A No

Wide (lots of No Yes Yes

columns/attributes)

Datum size Small (KBs) Large (GBs) to Very Large Small (KBs)
(TBs)

Overall Maximum Scale Large (low TBs) Very Large (PBs) Large (low TBs)
 Processing free-form text for search
12/18/2020 • 2 minutes to read • Edit Online

To support search, free-form text processing can be performed against documents containing paragraphs of text.
Text search works by constructing a specialized index that is precomputed against a collection of documents. A
client application submits a query that contains the search terms. The query returns a result set, consisting of a list
of documents sorted by how well each document matches the search criteria. The result set may also include the
context in which the document matches the criteria, which enables the application to highlight the matching phrase
in the document.

Free-form text processing can produce useful, actionable data from large amounts of noisy text data. The results
can give unstructured documents a well-defined and queryable structure.

Challenges
Processing a collection of free-form text documents is typically computationally intensive, as well as time
intensive.
In order to search free-form text effectively, the search index should support fuzzy search based on terms that
have a similar construction. For example, search indexes are built with lemmatization and linguistic stemming,
so that queries for "run" will match documents that contain "ran" and "running."

Architecture
In most scenarios, the source text documents are loaded into object storage such as Azure Storage or Azure Data
Lake Store, and then indexed using an external search service. In this scenario, source text documents are physically
distinct from a resulting search index that's hosted on a search service. An exception is using full text search within
SQL Server or Azure SQL Database. In this case, the document data exists internally in tables managed by the
database. Once stored, the documents are processed in a batch to create the index.

Technology choices
Options for creating a search index include Azure Cognitive Search, Elasticsearch, and HDInsight with Solr. Each of
these technologies can populate a search index from a collection of documents. Cognitive Search provides indexers
that can automatically populate the index for documents ranging from plain text to Excel and PDF formats. You can
also attach machine learning models to an indexer to analyze images and unstructured text for searchable content.
On HDInsight, Apache Solr can index binary files of many types, including plain text, Word, and PDF. Once the index
is constructed, clients can access the search interface by means of a REST API.
If your text data is stored in SQL Server or Azure SQL Database, you can use the full-text search that is built into the
database. The database populates the index from text, binary, or XML data stored within the same database. Clients
search by using T-SQL queries.
For more information, see Search data stores.
 Time series solutions
12/18/2020 • 4 minutes to read • Edit Online

Time series data is a set of values organized by time. Examples of time series data include sensor data, stock prices,
click stream data, and application telemetry. Time series data can be analyzed for historical trends, real-time alerts,
or predictive modeling.

Time series data represents how an asset or process changes over time. The data has a timestamp, but more
importantly, time is the most meaningful axis for viewing or analyzing the data. Time series data typically arrives in
order of time and is usually treated as an insert rather than an update to your database. Because of this, change is
measured over time, enabling you to look backward and to predict future change. As such, time series data is best
visualized with scatter or line charts.

Some examples of time series data are:

Stock prices captured over time to detect trends.
Server performance, such as CPU usage, I/O load, memory usage, and network bandwidth consumption.
Telemetry from sensors on industrial equipment, which can be used to detect pending equipment failure and
trigger alert notifications.
Real-time car telemetry data including speed, braking, and acceleration over a time window to produce an
aggregate risk score for the driver.
In each of these cases, you can see how time is most meaningful as an axis. Displaying the events in the order in
which they arrived is a key characteristic of time series data, as there is a natural temporal ordering. This differs
from data captured for standard OLTP data pipelines where data can be entered in any order, and updated at any
time.

When to use this solution

Choose a time series solution when you need to ingest data whose strategic value is centered around changes over
a period of time, and you are primarily inserting new data and rarely updating, if at all. You can use this information
to detect anomalies, visualize trends, and compare current data to historical data, among other things. This type of
architecture is also best suited for predictive modeling and forecasting results, because you have the historical
record of changes over time, which can be applied to any number of forecasting models.
Using time series offers the following benefits:
Clearly represents how an asset or process changes over time.
Helps you quickly detect changes to a number of related sources, making anomalies and emerging trends
clearly stand out.
Best suited for predictive modeling and forecasting.
Internet of Things (IoT )
Data collected by IoT devices is a natural fit for time series storage and analysis. The incoming data is inserted and
rarely, if ever, updated. The data is timestamped and inserted in the order it was received, and this data is typically
displayed in chronological order, enabling users to discover trends, spot anomalies, and use the information for
predictive analysis.
For more information, see Internet of Things.
Real-time analytics
Time series data is often time sensitive — that is, it must be acted on quickly, to spot trends in real time or generate
alerts. In these scenarios, any delay in insights can cause downtime and business impact. In addition, there is often
a need to correlate data from a variety of different sources, such as sensors.
Ideally, you would have a stream processing layer that can handle the incoming data in real time and process all of
it with high precision and high granularity. This isn't always possible, depending on your streaming architecture
and the components of your stream buffering and stream processing layers. You may need to sacrifice some
precision of the time series data by reducing it. This is done by processing sliding time windows (several seconds,
for example), allowing the processing layer to perform calculations in a timely manner. You may also need to
downsample and aggregate your data when displaying longer periods of time, such as zooming to display data
captured over several months.

Challenges
Time series data is often very high volume, especially in IoT scenarios. Storing, indexing, querying, analyzing,
and visualizing time series data can be challenging.
It can be challenging to find the right combination of high-speed storage and powerful compute operations
for handling real-time analytics, while minimizing time to market and overall cost investment.

Architecture
In many scenarios that involve time series data, such as IoT, the data is captured in real time. As such, a real-time
processing architecture is appropriate.
Data from one or more data sources is ingested into the stream buffering layer by IoT Hub, Event Hubs, or Kafka on
HDInsight. Next, the data is processed in the stream processing layer that can optionally hand off the processed
data to a machine learning service for predictive analytics. The processed data is stored in an analytical data store,
such as HBase, Azure Cosmos DB, Azure Data Lake, or Blob Storage. An analytics and reporting application or
service, like Power BI, Azure Data Explorer, or OpenTSDB (if stored in HBase) can be used to display the time series
data for analysis.
Another option is to use Azure Time Series Insights. Time Series Insights is a fully managed service for time series
data. In this architecture, Time Series Insights performs the roles of stream processing, data store, and analytics and
reporting. It accepts streaming data from either IoT Hub or Event Hubs and stores, processes, analyzes, and
displays the data in near real time. It does not pre-aggregate the data, but stores the raw events.
Time Series Insights is schema adaptive, which means that you do not have to do any data preparation to start
deriving insights. This enables you to explore, compare, and correlate a variety of data sources seamlessly. It also
provides SQL-like filters and aggregates, ability to construct, visualize, compare, and overlay various time series
patterns, heat maps, and the ability to save and share queries.

Technology choices
Data Storage
Analysis, visualizations, and reporting
Analytical Data Stores
Stream processing
 Working with CSV and JSON files for data solutions
12/18/2020 • 4 minutes to read • Edit Online

CSV and JSON are likely the most common formats used for ingesting, exchanging, and storing unstructured or
semi-structured data.

About CSV format

CSV (comma-separated values) files are commonly used to exchange tabular data between systems in plain text.
They typically contain a header row that provides column names for the data, but are otherwise considered semi-
structured. This is due to the fact that CSVs cannot naturally represent hierarchical or relational data. Data
relationships are typically handled with multiple CSV files, where foreign keys are stored in columns of one or more
files, but the relationships between those files are not expressed by the format itself. Files in CSV format may use
other delimiters besides commas, such as tabs or spaces.
Despite their limitations, CSV files are a popular choice for data exchange, because they are supported by a wide
range of business, consumer, and scientific applications. For example, database and spreadsheet programs can
import and export CSV files. Similarly, most batch and stream data processing engines, such as Spark and Hadoop,
natively support serializing and deserializing CSV-formatted files and offer ways to apply a schema on read. This
makes it easier to work with the data, by offering options to query against it and store the information in a more
efficient data format for faster processing.

About JSON format

JSON (JavaScript Object Notation) data is represented as key-value pairs in a semi-structured format. JSON is often
compared to XML, as both are capable of storing data in hierarchical format, with child data represented inline with
its parent. Both are self-describing and human readable, but JSON documents tend to be much smaller, leading to
their popular use in online data exchange, especially with the advent of REST-based web services.
JSON-formatted files have several benefits over CSV:
JSON maintains hierarchical structures, making it easier to hold related data in a single document and represent
complex relationships.
Most programming languages provide native support for deserializing JSON into objects, or provide
lightweight JSON serialization libraries.
JSON supports lists of objects, helping to avoid messy translations of lists into a relational data model.
JSON is a commonly used file format for NoSQL databases, such as MongoDB, Couchbase, and Azure Cosmos
DB.
Since a lot of data coming across the wire is already in JSON format, most web-based programming languages
support working with JSON natively, or through the use of external libraries to serialize and deserialize JSON data.
This universal support for JSON has led to its use in logical formats through data structure representation,
exchange formats for hot data, and data storage for cold data.
Many batch and stream data processing engines natively support JSON serialization and deserialization. Though
the data contained within JSON documents may ultimately be stored in a more performance-optimized formats,
such as Parquet or Avro, it serves as the raw data for source of truth, which is critical for reprocessing the data as
needed.

When to use CSV or JSON formats

CSVs are more commonly used for exporting and importing data, or processing it for analytics and machine
learning. JSON-formatted files have the same benefits, but are more common in hot data exchange solutions. JSON
documents are often sent by web and mobile devices performing online transactions, by IoT (internet of things)
devices for one-way or bidirectional communication, or by client applications communicating with SaaS and PaaS
services or serverless architectures.
CSV and JSON file formats both make it easy to exchange data between dissimilar systems or devices. Their semi-
structured formats allow flexibility in transferring almost any type of data, and universal support for these formats
make them simple to work with. Both can be used as the raw source of truth in cases where the processed data is
stored in binary formats for more efficient querying.

Working with CSV and JSON data in Azure

Azure provides several solutions for working with CSV and JSON files, depending on your needs. The primary
landing place for these files is either Azure Storage or Azure Data Lake Store. Most Azure services that work with
these and other text-based files integrate with either object storage service. In some situations, however, you may
opt to directly import the data into Azure SQL or some other data store. SQL Server has native support for storing
and working with JSON documents, which makes it easy to import and process those types of files. You can use a
utility like SQL Bulk Import to easily import CSV files.
You can also query JSON files directly from Azure Blob Storage without importing them into Azure SQL. For a
complete example of this approach, see Work with JSON files with Azure SQL. Currently this option isn't available
for CSV files.
Depending on the scenario, you may perform batch processing or real-time processing of the data.

Challenges
There are some challenges to consider when working with these formats:
Without any restraints on the data model, CSV and JSON files are prone to data corruption ("garbage in,
garbage out"). For instance, there's no notion of a date/time object in either file, so the file format does not
prevent you from inserting "ABC123" in a date field, for example.
Using CSV and JSON files as your cold storage solution does not scale well when working with big data. In
most cases, they cannot be split into partitions for parallel processing, and cannot be compressed as well as
binary formats. This often leads to processing and storing this data into read-optimized formats such as
Parquet and ORC (optimized row columnar), which also provide indexes and inline statistics about the data
contained.
You may need to apply a schema on the semi-structured data to make it easier to query and analyze.
Typically, this requires storing the data in another form that complies with your environment's data storage
needs, such as within a database.
 Big data architectures
12/18/2020 • 10 minutes to read • Edit Online

A big data architecture is designed to handle the ingestion, processing, and analysis of data that is too large or
complex for traditional database systems. The threshold at which organizations enter into the big data realm
differs, depending on the capabilities of the users and their tools. For some, it can mean hundreds of gigabytes of
data, while for others it means hundreds of terabytes. As tools for working with big data sets advance, so does the
meaning of big data. More and more, this term relates to the value you can extract from your data sets through
advanced analytics, rather than strictly the size of the data, although in these cases they tend to be quite large.
Over the years, the data landscape has changed. What you can do, or are expected to do, with data has changed.
The cost of storage has fallen dramatically, while the means by which data is collected keeps growing. Some data
arrives at a rapid pace, constantly demanding to be collected and observed. Other data arrives more slowly, but in
very large chunks, often in the form of decades of historical data. You might be facing an advanced analytics
problem, or one that requires machine learning. These are challenges that big data architectures seek to solve.
Big data solutions typically involve one or more of the following types of workload:
Batch processing of big data sources at rest.
Real-time processing of big data in motion.
Interactive exploration of big data.
Predictive analytics and machine learning.
Consider big data architectures when you need to:
Store and process data in volumes too large for a traditional database.
Transform unstructured data for analysis and reporting.
Capture, process, and analyze unbounded streams of data in real time, or with low latency.

Components of a big data architecture

The following diagram shows the logical components that fit into a big data architecture. Individual solutions may
not contain every item in this diagram.

Most big data architectures include some or all of the following components:
 Data sources . All big data solutions start with one or more data sources. Examples include:
Application data stores, such as relational databases.
Static files produced by applications, such as web server log files.
Real-time data sources, such as IoT devices.
Data storage . Data for batch processing operations is typically stored in a distributed file store that can
hold high volumes of large files in various formats. This kind of store is often called a data lake. Options for
implementing this storage include Azure Data Lake Store or blob containers in Azure Storage.
Batch processing . Because the data sets are so large, often a big data solution must process data files
using long-running batch jobs to filter, aggregate, and otherwise prepare the data for analysis. Usually these
jobs involve reading source files, processing them, and writing the output to new files. Options include
running U-SQL jobs in Azure Data Lake Analytics, using Hive, Pig, or custom Map/Reduce jobs in an
HDInsight Hadoop cluster, or using Java, Scala, or Python programs in an HDInsight Spark cluster.
Real-time message ingestion . If the solution includes real-time sources, the architecture must include a
way to capture and store real-time messages for stream processing. This might be a simple data store,
where incoming messages are dropped into a folder for processing. However, many solutions need a
message ingestion store to act as a buffer for messages, and to support scale-out processing, reliable
delivery, and other message queuing semantics. This portion of a streaming architecture is often referred to
as stream buffering. Options include Azure Event Hubs, Azure IoT Hub, and Kafka.
Stream processing . After capturing real-time messages, the solution must process them by filtering,
aggregating, and otherwise preparing the data for analysis. The processed stream data is then written to an
output sink. Azure Stream Analytics provides a managed stream processing service based on perpetually
running SQL queries that operate on unbounded streams. You can also use open source Apache streaming
technologies like Storm and Spark Streaming in an HDInsight cluster.
Analytical data store . Many big data solutions prepare data for analysis and then serve the processed
data in a structured format that can be queried using analytical tools. The analytical data store used to serve
these queries can be a Kimball-style relational data warehouse, as seen in most traditional business
intelligence (BI) solutions. Alternatively, the data could be presented through a low-latency NoSQL
technology such as HBase, or an interactive Hive database that provides a metadata abstraction over data
files in the distributed data store. Azure Synapse Analytics provides a managed service for large-scale,
cloud-based data warehousing. HDInsight supports Interactive Hive, HBase, and Spark SQL, which can also
be used to serve data for analysis.
Analysis and repor ting . The goal of most big data solutions is to provide insights into the data through
analysis and reporting. To empower users to analyze the data, the architecture may include a data modeling
layer, such as a multidimensional OLAP cube or tabular data model in Azure Analysis Services. It might also
support self-service BI, using the modeling and visualization technologies in Microsoft Power BI or Microsoft
Excel. Analysis and reporting can also take the form of interactive data exploration by data scientists or data
analysts. For these scenarios, many Azure services support analytical notebooks, such as Jupyter, enabling
these users to leverage their existing skills with Python or R. For large-scale data exploration, you can use
Microsoft R Server, either standalone or with Spark.
Orchestration . Most big data solutions consist of repeated data processing operations, encapsulated in
workflows, that transform source data, move data between multiple sources and sinks, load the processed
data into an analytical data store, or push the results straight to a report or dashboard. To automate these
workflows, you can use an orchestration technology such Azure Data Factory or Apache Oozie and Sqoop.

Lambda architecture
When working with very large data sets, it can take a long time to run the sort of queries that clients need. These
queries can't be performed in real time, and often require algorithms such as MapReduce that operate in parallel
across the entire data set. The results are then stored separately from the raw data and used for querying.
One drawback to this approach is that it introduces latency — if processing takes a few hours, a query may return
results that are several hours old. Ideally, you would like to get some results in real time (perhaps with some loss of
accuracy), and combine these results with the results from the batch analytics.
The lambda architecture , first proposed by Nathan Marz, addresses this problem by creating two paths for data
flow. All data coming into the system goes through these two paths:
A batch layer (cold path) stores all of the incoming data in its raw form and performs batch processing on
the data. The result of this processing is stored as a batch view .
A speed layer (hot path) analyzes data in real time. This layer is designed for low latency, at the expense of
accuracy.
The batch layer feeds into a ser ving layer that indexes the batch view for efficient querying. The speed layer
updates the serving layer with incremental updates based on the most recent data.

Data that flows into the hot path is constrained by latency requirements imposed by the speed layer, so that it can
be processed as quickly as possible. Often, this requires a tradeoff of some level of accuracy in favor of data that is
ready as quickly as possible. For example, consider an IoT scenario where a large number of temperature sensors
are sending telemetry data. The speed layer may be used to process a sliding time window of the incoming data.
Data flowing into the cold path, on the other hand, is not subject to the same low latency requirements. This allows
for high accuracy computation across large data sets, which can be very time intensive.
Eventually, the hot and cold paths converge at the analytics client application. If the client needs to display timely,
yet potentially less accurate data in real time, it will acquire its result from the hot path. Otherwise, it will select
results from the cold path to display less timely but more accurate data. In other words, the hot path has data for a
relatively small window of time, after which the results can be updated with more accurate data from the cold path.
The raw data stored at the batch layer is immutable. Incoming data is always appended to the existing data, and the
previous data is never overwritten. Any changes to the value of a particular datum are stored as a new
timestamped event record. This allows for recomputation at any point in time across the history of the data
collected. The ability to recompute the batch view from the original raw data is important, because it allows for
new views to be created as the system evolves.

Kappa architecture
A drawback to the lambda architecture is its complexity. Processing logic appears in two different places — the cold
and hot paths — using different frameworks. This leads to duplicate computation logic and the complexity of
managing the architecture for both paths.
The kappa architecture was proposed by Jay Kreps as an alternative to the lambda architecture. It has the same
basic goals as the lambda architecture, but with an important distinction: All data flows through a single path,
using a stream processing system.

There are some similarities to the lambda architecture's batch layer, in that the event data is immutable and all of it
is collected, instead of a subset. The data is ingested as a stream of events into a distributed and fault tolerant
unified log. These events are ordered, and the current state of an event is changed only by a new event being
appended. Similar to a lambda architecture's speed layer, all event processing is performed on the input stream and
persisted as a real-time view.
If you need to recompute the entire data set (equivalent to what the batch layer does in lambda), you simply replay
the stream, typically using parallelism to complete the computation in a timely fashion.

Internet of Things (IoT)

From a practical viewpoint, Internet of Things (IoT) represents any device that is connected to the Internet. This
includes your PC, mobile phone, smart watch, smart thermostat, smart refrigerator, connected automobile, heart
monitoring implants, and anything else that connects to the Internet and sends or receives data. The number of
connected devices grows every day, as does the amount of data collected from them. Often this data is being
collected in highly constrained, sometimes high-latency environments. In other cases, data is sent from low-latency
environments by thousands or millions of devices, requiring the ability to rapidly ingest the data and process
accordingly. Therefore, proper planning is required to handle these constraints and unique requirements.
Event-driven architectures are central to IoT solutions. The following diagram shows a possible logical architecture
for IoT. The diagram emphasizes the event-streaming components of the architecture.

The cloud gateway ingests device events at the cloud boundary, using a reliable, low latency messaging system.
Devices might send events directly to the cloud gateway, or through a field gateway . A field gateway is a
specialized device or software, usually collocated with the devices, that receives events and forwards them to the
cloud gateway. The field gateway might also preprocess the raw device events, performing functions such as
filtering, aggregation, or protocol transformation.
After ingestion, events go through one or more stream processors that can route the data (for example, to
storage) or perform analytics and other processing.
The following are some common types of processing. (This list is certainly not exhaustive.)
Writing event data to cold storage, for archiving or batch analytics.
Hot path analytics, analyzing the event stream in (near) real time, to detect anomalies, recognize patterns
over rolling time windows, or trigger alerts when a specific condition occurs in the stream.
Handling special types of nontelemetry messages from devices, such as notifications and alarms.
Machine learning.
The boxes that are shaded gray show components of an IoT system that are not directly related to event streaming,
but are included here for completeness.
The device registr y is a database of the provisioned devices, including the device IDs and usually device
metadata, such as location.
The provisioning API is a common external interface for provisioning and registering new devices.
Some IoT solutions allow command and control messages to be sent to devices.
Relevant Azure services:
Azure IoT Hub
Azure Event Hubs
Azure Stream Analytics
Azure Data Explorer
Learn more about IoT on Azure by reading the Azure IoT reference architecture.
 Batch processing
12/18/2020 • 6 minutes to read • Edit Online

A common big data scenario is batch processing of data at rest. In this scenario, the source data is loaded into data
storage, either by the source application itself or by an orchestration workflow. The data is then processed in-place
by a parallelized job, which can also be initiated by the orchestration workflow. The processing may include
multiple iterative steps before the transformed results are loaded into an analytical data store, which can be
queried by analytics and reporting components.
For example, the logs from a web server might be copied to a folder and then processed overnight to generate
daily reports of web activity.

When to use this solution

Batch processing is used in a variety of scenarios, from simple data transformations to a more complete ETL
(extract-transform-load) pipeline. In a big data context, batch processing may operate over very large data sets,
where the computation takes significant time. (For example, see Lambda architecture.) Batch processing typically
leads to further interactive exploration, provides the modeling-ready data for machine learning, or writes the data
to a data store that is optimized for analytics and visualization.
One example of batch processing is transforming a large set of flat, semi-structured CSV or JSON files into a
schematized and structured format that is ready for further querying. Typically the data is converted from the raw
formats used for ingestion (such as CSV) into binary formats that are more performant for querying because they
store data in a columnar format, and often provide indexes and inline statistics about the data.

Challenges
Data format and encoding . Some of the most difficult issues to debug happen when files use an
unexpected format or encoding. For example, source files might use a mix of UTF-16 and UTF-8 encoding,
or contain unexpected delimiters (space versus tab), or include unexpected characters. Another common
example is text fields that contain tabs, spaces, or commas that are interpreted as delimiters. Data loading
and parsing logic must be flexible enough to detect and handle these issues.
Orchestrating time slices . Often source data is placed in a folder hierarchy that reflects processing
windows, organized by year, month, day, hour, and so on. In some cases, data may arrive late. For example,
suppose that a web server fails, and the logs for March 7th don't end up in the folder for processing until
March 9th. Are they just ignored because they're too late? Can the downstream processing logic handle out-
 of-order records?

Architecture
A batch processing architecture has the following logical components, shown in the diagram above.
Data storage . Typically a distributed file store that can serve as a repository for high volumes of large files
in various formats. Generically, this kind of store is often referred to as a data lake.
Batch processing . The high-volume nature of big data often means that solutions must process data files
using long-running batch jobs to filter, aggregate, and otherwise prepare the data for analysis. Usually these
jobs involve reading source files, processing them, and writing the output to new files.
Analytical data store . Many big data solutions are designed to prepare data for analysis and then serve
the processed data in a structured format that can be queried using analytical tools.
Analysis and repor ting . The goal of most big data solutions is to provide insights into the data through
analysis and reporting.
Orchestration . With batch processing, typically some orchestration is required to migrate or copy the data
into your data storage, batch processing, analytical data store, and reporting layers.

Technology choices
The following technologies are recommended choices for batch processing solutions in Azure.
Data storage
Azure Storage Blob Containers . Many existing Azure business processes already use Azure blob storage,
making this a good choice for a big data store.
Azure Data Lake Store . Azure Data Lake Store offers virtually unlimited storage for any size of file, and
extensive security options, making it a good choice for extremely large-scale big data solutions that require a
centralized store for data in heterogeneous formats.
For more information, see Data storage.
Batch processing
U-SQL . U-SQL is the query processing language used by Azure Data Lake Analytics. It combines the declarative
nature of SQL with the procedural extensibility of C#, and takes advantage of parallelism to enable efficient
processing of data at massive scale.
Hive . Hive is a SQL-like language that is supported in most Hadoop distributions, including HDInsight. It can be
used to process data from any HDFS-compatible store, including Azure blob storage and Azure Data Lake Store.
Pig . Pig is a declarative big data processing language used in many Hadoop distributions, including HDInsight.
It is particularly useful for processing data that is unstructured or semi-structured.
Spark . The Spark engine supports batch processing programs written in a range of languages, including Java,
Scala, and Python. Spark uses a distributed architecture to process data in parallel across multiple worker
nodes.
For more information, see Batch processing.
Analytical data store
Azure Synapse Analytics . Azure Synapse is a managed service based on SQL Server database technologies
and optimized to support large-scale data warehousing workloads.
Spark SQL . Spark SQL is an API built on Spark that supports the creation of dataframes and tables that can be
queried using SQL syntax.
HBase . HBase is a low-latency NoSQL store that offers a high-performance, flexible option for querying
 structured and semi-structured data.
Hive . In addition to being useful for batch processing, Hive offers a database architecture that is conceptually
similar to that of a typical relational database management system. Improvements in Hive query performance
through innovations like the Tez engine and Stinger initiative mean that Hive tables can be used effectively as
sources for analytical queries in some scenarios.
For more information, see Analytical data stores.
Analytics and reporting
Azure Analysis Ser vices . Many big data solutions emulate traditional enterprise business intelligence
architectures by including a centralized online analytical processing (OLAP) data model (often referred to as a
cube) on which reports, dashboards, and interactive "slice and dice" analysis can be based. Azure Analysis
Services supports the creation of tabular models to meet this need.
Power BI . Power BI enables data analysts to create interactive data visualizations based on data models in an
OLAP model or directly from an analytical data store.
Microsoft Excel . Microsoft Excel is one of the most widely used software applications in the world, and offers
a wealth of data analysis and visualization capabilities. Data analysts can use Excel to build document data
models from analytical data stores, or to retrieve data from OLAP data models into interactive PivotTables and
charts.
For more information, see Analytics and reporting.
Orchestration
Azure Data Factor y . Azure Data Factory pipelines can be used to define a sequence of activities, scheduled for
recurring temporal windows. These activities can initiate data copy operations as well as Hive, Pig, MapReduce,
or Spark jobs in on-demand HDInsight clusters; U-SQL jobs in Azure Date Lake Analytics; and stored
procedures in Azure Synapse or Azure SQL Database.
Oozie and Sqoop . Oozie is a job automation engine for the Apache Hadoop ecosystem and can be used to
initiate data copy operations as well as Hive, Pig, and MapReduce jobs to process data and Sqoop jobs to copy
data between HDFS and SQL databases.
For more information, see Pipeline orchestration
 Real time processing
12/18/2020 • 5 minutes to read • Edit Online

Real time processing deals with streams of data that are captured in real-time and processed with minimal latency
to generate real-time (or near-real-time) reports or automated responses. For example, a real-time traffic
monitoring solution might use sensor data to detect high traffic volumes. This data could be used to dynamically
update a map to show congestion, or automatically initiate high-occupancy lanes or other traffic management
systems.

Real-time processing is defined as the processing of unbounded stream of input data, with very short latency
requirements for processing — measured in milliseconds or seconds. This incoming data typically arrives in an
unstructured or semi-structured format, such as JSON, and has the same processing requirements as batch
processing, but with shorter turnaround times to support real-time consumption.
Processed data is often written to an analytical data store, which is optimized for analytics and visualization. The
processed data can also be ingested directly into the analytics and reporting layer for analysis, business
intelligence, and real-time dashboard visualization.

Challenges
One of the big challenges of real-time processing solutions is to ingest, process, and store messages in real time,
especially at high volumes. Processing must be done in such a way that it does not block the ingestion pipeline.
The data store must support high-volume writes. Another challenge is being able to act on the data quickly, such
as generating alerts in real time or presenting the data in a real-time (or near-real-time) dashboard.

Architecture
A real-time processing architecture has the following logical components.
Real-time message ingestion. The architecture must include a way to capture and store real-time
messages to be consumed by a stream processing consumer. In simple cases, this service could be
implemented as a simple data store in which new messages are deposited in a folder. But often the solution
requires a message broker, such as Azure Event Hubs, that acts as a buffer for the messages. The message
broker should support scale-out processing and reliable delivery.
Stream processing. After capturing real-time messages, the solution must process them by filtering,
aggregating, and otherwise preparing the data for analysis.
 Analytical data store. Many big data solutions are designed to prepare data for analysis and then serve
the processed data in a structured format that can be queried using analytical tools.
Analysis and repor ting. The goal of most big data solutions is to provide insights into the data through
analysis and reporting.

Technology choices
The following technologies are recommended choices for real-time processing solutions in Azure.
Real-time message ingestion
Azure Event Hubs . Azure Event Hubs is a messaging solution for ingesting millions of event messages per
second. The captured event data can be processed by multiple consumers in parallel. While Event Hubs natively
supports AMQP (Advanced Message Queuing Protocol 1.0), it also provides a binary compatibility layer that
allows applications using the Kafka protocol (Kafka 1.0 and above) to process events using Event Hubs with no
application changes.
Azure IoT Hub . Azure IoT Hub provides bi-directional communication between Internet-connected devices,
and a scalable message queue that can handle millions of simultaneously connected devices.
Apache Kafka . Kafka is an open source message queuing and stream processing application that can scale to
handle millions of messages per second from multiple message producers, and route them to multiple
consumers. Kafka is available in Azure as an HDInsight cluster type.
For more information, see Real-time message ingestion.
Data storage
Azure Storage Blob Containers or Azure Data Lake Store . Incoming real-time data is usually captured in
a message broker (see above), but in some scenarios, it can make sense to monitor a folder for new files and
process them as they are created or updated. Additionally, many real-time processing solutions combine
streaming data with static reference data, which can be stored in a file store. Finally, file storage may be used as
an output destination for captured real-time data for archiving, or for further batch processing in a lambda
architecture.
For more information, see Data storage.
Stream processing
Azure Stream Analytics . Azure Stream Analytics can run perpetual queries against an unbounded stream of
data. These queries consume streams of data from storage or message brokers, filter and aggregate the data
based on temporal windows, and write the results to sinks such as storage, databases, or directly to reports in
Power BI. Stream Analytics uses a SQL-based query language that supports temporal and geospatial constructs,
and can be extended using JavaScript.
Storm . Apache Storm is an open source framework for stream processing that uses a topology of spouts and
bolts to consume, process, and output the results from real-time streaming data sources. You can provision
Storm in an Azure HDInsight cluster, and implement a topology in Java or C#.
Spark Streaming . Apache Spark is an open source distributed platform for general data processing. Spark
provides the Spark Streaming API, in which you can write code in any supported Spark language, including
Java, Scala, and Python. Spark 2.0 introduced the Spark Structured Streaming API, which provides a simpler and
more consistent programming model. Spark 2.0 is available in an Azure HDInsight cluster.
For more information, see Stream processing.
Analytical data store
Azure Synapse Analytics , Azure Data Explorer , HBase , Spark , or Hive . Processed real-time data can be
stored in a relational database such Synapse Analytics, Azure Data Explorer, a NoSQL store such as HBase, or as
files in distributed storage over which Spark or Hive tables can be defined and queried.
For more information, see Analytical data stores.
Analytics and reporting
Azure Analysis Ser vices , Power BI , and Microsoft Excel . Processed real-time data that is stored in an
analytical data store can be used for historical reporting and analysis in the same way as batch processed data.
Additionally, Power BI can be used to publish real-time (or near-real-time) reports and visualizations from
analytical data sources where latency is sufficiently low, or in some cases directly from the stream processing
output.
For more information, see Analytics and reporting.
In a purely real-time solution, most of the processing orchestration is managed by the message ingestion and
stream processing components. However, in a lambda architecture that combines batch processing and real-time
processing, you may need to use an orchestration framework such as Azure Data Factory or Apache Oozie and
Sqoop to manage batch workflows for captured real-time data.

Next steps
The following reference architecture shows an end-to-end stream processing pipeline:
Stream processing with Azure Stream Analytics
 Choosing an analytical data store in Azure
12/18/2020 • 5 minutes to read • Edit Online

In a big data architecture, there is often a need for an analytical data store that serves processed data in a
structured format that can be queried using analytical tools. Analytical data stores that support querying of both
hot-path and cold-path data are collectively referred to as the serving layer, or data serving storage.
The serving layer deals with processed data from both the hot path and cold path. In the lambda architecture, the
serving layer is subdivided into a speed serving layer, which stores data that has been processed incrementally,
and a batch serving layer, which contains the batch-processed output. The serving layer requires strong support
for random reads with low latency. Data storage for the speed layer should also support random writes, because
batch loading data into this store would introduce undesired delays. On the other hand, data storage for the batch
layer does not need to support random writes, but batch writes instead.
There is no single best data management choice for all data storage tasks. Different data management solutions
are optimized for different tasks. Most real-world cloud apps and big data processes have a variety of data storage
requirements and often use a combination of data storage solutions.

What are your options when choosing an analytical data store?

There are several options for data serving storage in Azure, depending on your needs:
Azure Synapse Analytics
Azure Data Explorer
Azure SQL Database
SQL Server in Azure VM
HBase/Phoenix on HDInsight
Hive LLAP on HDInsight
Azure Analysis Services
Azure Cosmos DB
These options provide various database models that are optimized for different types of tasks:
Key/value databases hold a single serialized object for each key value. They're good for storing large volumes
of data where you want to get one item for a given key value and you don't have to query based on other
properties of the item.
Document databases are key/value databases in which the values are documents. A "document" in this context
is a collection of named fields and values. The database typically stores the data in a format such as XML, YAML,
JSON, or BSON, but may use plain text. Document databases can query on non-key fields and define secondary
indexes to make querying more efficient. This makes a document database more suitable for applications that
need to retrieve data based on criteria more complex than the value of the document key. For example, you
could query on fields such as product ID, customer ID, or customer name.
Column-family databases are key/value data stores that structure data storage into collections of related
columns called column families. For example, a census database might have one group of columns for a
person's name (first, middle, last), one group for the person's address, and one group for the person's profile
information (data of birth, gender). The database can store each column family in a separate partition, while
keeping all of the data for one person related to the same key. An application can read a single column family
without reading through all of the data for an entity.
Graph databases store information as a collection of objects and relationships. A graph database can efficiently
perform queries that traverse the network of objects and the relationships between them. For example, the
 objects might be employees in a human resources database, and you might want to facilitate queries such as
"find all employees who directly or indirectly work for Scott."
Telemetry and time series databases are an append-only collection of objects. Telemetry databases efficiently
index data in a variety of column stores and in-memory structures, making them the optimal choice for storing
and analyzing vast quantities of telemetry and time series data.

Key selection criteria

To narrow the choices, start by answering these questions:
Do you need serving storage that can serve as a hot path for your data? If yes, narrow your options to those
that are optimized for a speed serving layer.
Do you need massively parallel processing (MPP) support, where queries are automatically distributed
across several processes or nodes? If yes, select an option that supports query scale out.
Do you prefer to use a relational data store? If so, narrow your options to those with a relational database
model. However, note that some non-relational stores support SQL syntax for querying, and tools such as
PolyBase can be used to query non-relational data stores.
Do you collect time series data? Do you use append-only data?

Capability matrix
The following tables summarize the key differences in capabilities.
General capabilities
A Z URE H B A SE/ P H H IVE L L A P A Z URE
SQ L A Z URE DATA O EN IX O N ON A N A LY SIS C O SM O S
C A PA B IL IT Y DATA B A SE SY N A P SE EXP LO RER H DIN SIGH T H DIN SIGH T SERVIC ES DB

Is managed Yes Yes Yes Yes 1 Yes 1 Yes Yes

service

Primary Relational Relational Relational Wide Hive/In- Tabular/MO Document

database (columnar tables with (column column Memory LAP store,
model format columnar store), store semantic graph, key-
when using storage telemetry, models value store,
columnstor and time wide
e indexes) series store column
store

SQL Yes Yes Yes Yes (using Yes No Yes

language Phoenix
support JDBC
driver)

Optimized Yes 2 No Yes Yes Yes No Yes

for speed
serving
layer

[1] With manual configuration and scaling.

[2] Using memory-optimized tables and hash or nonclustered indexes.
Scalability capabilities
 A Z URE H B A SE/ P H H IVE L L A P A Z URE
SQ L A Z URE DATA O EN IX O N ON A N A LY SIS C O SM O S
C A PA B IL IT Y DATA B A SE SY N A P SE EXP LO RER H DIN SIGH T H DIN SIGH T SERVIC ES DB

Redundant Yes Yes Yes Yes No No Yes

regional
servers for
high
availability

Supports No Yes Yes Yes Yes Yes Yes

query scale
out

Dynamic Yes Yes Yes No No Yes Yes

scalability
(scale up)

Supports Yes Yes Yes No Yes Yes No

in-memory
caching of
data

Security capabilities
A Z URE H B A SE/ P H H IVE L L A P A Z URE
SQ L A Z URE DATA O EN IX O N ON A N A LY SIS C O SM O S
C A PA B IL IT Y DATA B A SE SY N A P SE EXP LO RER H DIN SIGH T H DIN SIGH T SERVIC ES DB

Authenticati SQL / Azure SQL / Azure Azure AD local / local / Azure AD database
on Active AD Azure AD 1 Azure AD 1 users /
Directory Azure AD
(Azure AD) via access
control
(IAM)

Data Yes 2 Yes 2 Yes Yes 1 Yes 1 Yes Yes

encryption
at rest

Row-level Yes Yes 3 Yes Yes 1 Yes 1 Yes No

security (through
object-level
security in
model)

Supports Yes Yes Yes Yes 4 Yes 4 Yes Yes

firewalls

Dynamic Yes Yes Yes Yes 1 Yes No No

data
masking

[1] Requires using a domain-joined HDInsight cluster.

[2] Requires using transparent data encryption (TDE) to encrypt and decrypt your data at rest.
[3] Filter predicates only. See Row-Level Security
[4] When used within an Azure Virtual Network. See Extend Azure HDInsight using an Azure Virtual Network.
 Choosing a data analytics technology in Azure
12/18/2020 • 4 minutes to read • Edit Online

The goal of most big data solutions is to provide insights into the data through analysis and reporting. This can
include preconfigured reports and visualizations, or interactive data exploration.

What are your options when choosing a data analytics technology?

There are several options for analysis, visualizations, and reporting in Azure, depending on your needs:
Power BI
Jupyter Notebooks
Zeppelin Notebooks
Microsoft Azure Notebooks
Power BI
Power BI is a suite of business analytics tools. It can connect to hundreds of data sources, and can be used for ad
hoc analysis. See this list of the currently available data sources. Use Power BI Embedded to integrate Power BI
within your own applications without requiring any additional licensing.
Organizations can use Power BI to produce reports and publish them to the organization. Everyone can create
personalized dashboards, with governance and security built in. Power BI uses Azure Active Directory (Azure AD)
to authenticate users who log in to the Power BI service, and uses the Power BI login credentials whenever a user
attempts to access resources that require authentication.
Jupyter Notebooks
Jupyter Notebooks provide a browser-based shell that lets data scientists create notebook files that contain
Python, Scala, or R code and markdown text, making it an effective way to collaborate by sharing and
documenting code and results in a single document.
Most varieties of HDInsight clusters, such as Spark or Hadoop, come preconfigured with Jupyter notebooks for
interacting with data and submitting jobs for processing. Depending on the type of HDInsight cluster you are
using, one or more kernels will be provided for interpreting and running your code. For example, Spark clusters on
HDInsight provide Spark-related kernels that you can select from to execute Python or Scala code using the Spark
engine.
Jupyter notebooks provide a great environment for analyzing, visualizing, and processing your data prior to
building more advanced visualizations with a BI/reporting tool like Power BI.
Zeppelin Notebooks
Zeppelin Notebooks are another option for a browser-based shell, similar to Jupyter in functionality. Some
HDInsight clusters come preconfigured with Zeppelin notebooks. However, if you are using an HDInsight
Interactive Query (Hive LLAP) cluster, Zeppelin is currently your only choice of notebook that you can use to run
interactive Hive queries. Also, if you are using a domain-joined HDInsight cluster, Zeppelin notebooks are the only
type that enables you to assign different user logins to control access to notebooks and the underlying Hive tables.
Microsoft Azure Notebooks
Azure Notebooks is an online Jupyter Notebooks-based service that enables data scientists to create, run, and
share Jupyter Notebooks in cloud-based libraries. Azure Notebooks provides execution environments for Python
2, Python 3, F#, and R, and provides several charting libraries for visualizing your data, such as ggplot, matplotlib,
bokeh, and seaborn.
Unlike Jupyter notebooks running on an HDInsight cluster, which are connected to the cluster's default storage
account, Azure Notebooks does not provide any data. You must load data in a variety of ways, such downloading
data from an online source, interacting with Azure Blobs or Table Storage, connecting to a SQL database, or
loading data with the Copy Wizard for Azure Data Factory.
Key benefits:
Free service—no Azure subscription required.
No need to install Jupyter and the supporting R or Python distributions locally—just use a browser.
Manage your own online libraries and access them from any device.
Share your notebooks with collaborators.
Considerations:
You will be unable to access your notebooks when offline.
Limited processing capabilities of the free notebook service may not be enough to train large or complex
models.

Key selection criteria

To narrow the choices, start by answering these questions:
Do you need to connect to numerous data sources, providing a centralized place to create reports for data
spread throughout your domain? If so, choose an option that allows you to connect to 100s of data sources.
Do you want to embed dynamic visualizations in an external website or application? If so, choose an option
that provides embedding capabilities.
Do you want to design your visualizations and reports while offline? If yes, choose an option with offline
capabilities.
Do you need heavy processing power to train large or complex AI models or work with very large data
sets? If yes, choose an option that can connect to a big data cluster.

Capability matrix
The following tables summarize the key differences in capabilities.
General capabilities
Z EP P EL IN M IC RO SO F T A Z URE
C A PA B IL IT Y P O W ER B I JUP Y T ER N OT EB O O K S N OT EB O O K S N OT EB O O K S

Connect to big data Yes Yes Yes No

cluster for advanced
processing

Managed service Yes Yes 1 Yes 1 Yes

Connect to 100s of Yes No No No

data sources

Offline capabilities Yes 2 No No No

Embedding Yes No No No
capabilities
 Z EP P EL IN M IC RO SO F T A Z URE
C A PA B IL IT Y P O W ER B I JUP Y T ER N OT EB O O K S N OT EB O O K S N OT EB O O K S

Automatic data Yes No No No

refresh

Access to numerous No Yes 3 Yes 3 Yes 4

open source packages

Data Power Query, R 40 languages, 20+ interpreters, Python, F#, R

transformation/cleans including Python, R, including Python,
ing options Julia, and Scala JDBC, and R

Pricing Free for Power BI Free Free Free

Desktop (authoring),
see pricing for
hosting options

Multiuser Yes Yes (through sharing Yes Yes (through sharing)

collaboration or with a multiuser
server like
JupyterHub)

[1] When used as part of a managed HDInsight cluster.

[2] With the use of Power BI Desktop.
[2] You can search the Maven repository for community-contributed packages.
[3] Python packages can be installed using either pip or conda. R packages can be installed from CRAN or GitHub.
Packages in F# can be installed via nuget.org using the Paket dependency manager.
 Choosing a batch processing technology in Azure
12/18/2020 • 3 minutes to read • Edit Online

Big data solutions often use long-running batch jobs to filter, aggregate, and otherwise prepare the data for
analysis. Usually these jobs involve reading source files from scalable storage (like HDFS, Azure Data Lake Store,
and Azure Storage), processing them, and writing the output to new files in scalable storage.
The key requirement of such batch processing engines is the ability to scale out computations, in order to handle a
large volume of data. Unlike real-time processing, however, batch processing is expected to have latencies (the time
between data ingestion and computing a result) that measure in minutes to hours.

Technology choices for batch processing

Azure Synapse Analytics
Azure Synapse is a distributed system designed to perform analytics on large data. It supports massive parallel
processing (MPP), which makes it suitable for running high-performance analytics. Consider Azure Synapse when
you have large amounts of data (more than 1 TB) and are running an analytics workload that will benefit from
parallelism.
Azure Data Lake Analytics
Data Lake Analytics is an on-demand analytics job service. It is optimized for distributed processing of very large
data sets stored in Azure Data Lake Store.
Languages: U-SQL (including Python, R, and C# extensions).
Integrates with Azure Data Lake Store, Azure Storage blobs, Azure SQL Database, and Azure Synapse.
Pricing model is per-job.
HDInsight
HDInsight is a managed Hadoop service. Use it deploy and manage Hadoop clusters in Azure. For batch processing,
you can use Spark, Hive, Hive LLAP, MapReduce.
Languages: R, Python, Java, Scala, SQL
Kerberos authentication with Active Directory, Apache Ranger based access control
Gives you full control of the Hadoop cluster
Azure Databricks
Azure Databricks is an Apache Spark-based analytics platform. You can think of it as "Spark as a service." It's the
easiest way to use Spark on the Azure platform.
Languages: R, Python, Java, Scala, Spark SQL
Fast cluster start times, autotermination, autoscaling.
Manages the Spark cluster for you.
Built-in integration with Azure Blob Storage, Azure Data Lake Storage (ADLS), Azure Synapse, and other services.
See Data Sources.
User authentication with Azure Active Directory.
Web-based notebooks for collaboration and data exploration.
Supports GPU-enabled clusters
Azure Distributed Data Engineering Toolkit
The Distributed Data Engineering Toolkit (AZTK) is a tool for provisioning on-demand Spark on Docker clusters in
Azure.
AZTK is not an Azure service. Rather, it's a client-side tool with a CLI and Python SDK interface, that's built on Azure
Batch. This option gives you the most control over the infrastructure when deploying a Spark cluster.
Bring your own Docker image.
Use low-priority VMs for an 80% discount.
Mixed mode clusters that use both low-priority and dedicated VMs.
Built in support for Azure Blob Storage and Azure Data Lake connection.

Key selection criteria

To narrow the choices, start by answering these questions:
Do you want a managed service rather than managing your own servers?
Do you want to author batch processing logic declaratively or imperatively?
Will you perform batch processing in bursts? If yes, consider options that let you auto-terminate the cluster
or whose pricing model is per batch job.
Do you need to query relational data stores along with your batch processing, for example to look up
reference data? If yes, consider the options that enable querying of external relational stores.

Capability matrix
The following tables summarize the key differences in capabilities.
General capabilities
A Z URE DATA L A K E
C A PA B IL IT Y A N A LY T IC S A Z URE SY N A P SE H DIN SIGH T A Z URE DATA B RIC K S

Is managed service Yes Yes Yes 1 Yes

Relational data store Yes Yes No No

Pricing model Per batch job By cluster hour By cluster hour Databricks Unit2 +
cluster hour

[1] With manual configuration and scaling.

[2] A Databricks Unit (DBU) is a unit of processing capability per hour.
Capabilities
A Z URE DATA H DIN SIGH T
LAKE A Z URE H DIN SIGH T H DIN SIGH T W IT H H IVE A Z URE
C A PA B IL IT Y A N A LY T IC S SY N A P SE W IT H SPA RK W IT H H IVE LLAP DATA B RIC K S

Autoscaling No No No No No Yes

Scale-out Per job Per cluster Per cluster Per cluster Per cluster Per cluster
granularity

In-memory No Yes Yes No Yes Yes

caching of
data
 A Z URE DATA H DIN SIGH T
LAKE A Z URE H DIN SIGH T H DIN SIGH T W IT H H IVE A Z URE
C A PA B IL IT Y A N A LY T IC S SY N A P SE W IT H SPA RK W IT H H IVE LLAP DATA B RIC K S

Query from Yes No Yes No No Yes

external
relational
stores

Authenticatio Azure AD SQL / Azure No Azure AD1 Azure AD1 Azure AD

n AD

Auditing Yes Yes No Yes 1 Yes 1 Yes

Row-level No Yes2 No Yes 1 Yes 1 No

security

Supports Yes Yes Yes Yes 3 Yes 3 No

firewalls

Dynamic data No Yes No Yes 1 Yes 1 No

masking

[1] Requires using a domain-joined HDInsight cluster.

[2] Filter predicates only. See Row-Level Security
[3] Supported when used within an Azure Virtual Network.
 Data lakes
12/18/2020 • 2 minutes to read • Edit Online

A data lake is a storage repository that holds a large amount of data in its native, raw format. Data lake stores are
optimized for scaling to terabytes and petabytes of data. The data typically comes from multiple heterogeneous
sources, and may be structured, semi-structured, or unstructured. The idea with a data lake is to store everything in
its original, untransformed state. This approach differs from a traditional data warehouse, which transforms and
processes the data at the time of ingestion.
Advantages of a data lake:
Data is never thrown away, because the data is stored in its raw format. This is especially useful in a big data
environment, when you may not know in advance what insights are available from the data.
Users can explore the data and create their own queries.
May be faster than traditional ETL tools.
More flexible than a data warehouse, because it can store unstructured and semi-structured data.
A complete data lake solution consists of both storage and processing. Data lake storage is designed for fault-
tolerance, infinite scalability, and high-throughput ingestion of data with varying shapes and sizes. Data lake
processing involves one or more processing engines built with these goals in mind, and can operate on data stored
in a data lake at scale.

When to use a data lake

Typical uses for a data lake include data exploration, data analytics, and machine learning.
A data lake can also act as the data source for a data warehouse. With this approach, the raw data is ingested into
the data lake and then transformed into a structured queryable format. Typically this transformation uses an ELT
(extract-load-transform) pipeline, where the data is ingested and transformed in place. Source data that is already
relational may go directly into the data warehouse, using an ETL process, skipping the data lake.
Data lake stores are often used in event streaming or IoT scenarios, because they can persist large amounts of
relational and nonrelational data without transformation or schema definition. They are built to handle high
volumes of small writes at low latency, and are optimized for massive throughput.

Challenges
Lack of a schema or descriptive metadata can make the data hard to consume or query.
Lack of semantic consistency across the data can make it challenging to perform analysis on the data, unless
users are highly skilled at data analytics.
It can be hard to guarantee the quality of the data going into the data lake.
Without proper governance, access control and privacy issues can be problems. What information is going into
the data lake, who can access that data, and for what uses?
A data lake may not be the best way to integrate data that is already relational.
By itself, a data lake does not provide integrated or holistic views across the organization.
A data lake may become a dumping ground for data that is never actually analyzed or mined for insights.

Relevant Azure services

Data Lake Store is a hyperscale, Hadoop-compatible repository.
Data Lake Analytics is an on-demand analytics job service to simplify big data analytics.
 Choosing a big data storage technology in Azure
12/18/2020 • 8 minutes to read • Edit Online

This topic compares options for data storage for big data solutions — specifically, data storage for bulk data
ingestion and batch processing, as opposed to analytical data stores or real-time streaming ingestion.

What are your options when choosing data storage in Azure?

There are several options for ingesting data into Azure, depending on your needs.
File storage:
Azure Storage blobs
Azure Data Lake Store
NoSQL databases:
Azure Cosmos DB
HBase on HDInsight
Analytical databases:
Azure Data Explorer

Azure Storage blobs

Azure Storage is a managed storage service that is highly available, secure, durable, scalable, and redundant.
Microsoft takes care of maintenance and handles critical problems for you. Azure Storage is the most ubiquitous
storage solution Azure provides, due to the number of services and tools that can be used with it.
There are various Azure Storage services you can use to store data. The most flexible option for storing blobs from
a number of data sources is Blob storage. Blobs are basically files. They store pictures, documents, HTML files,
virtual hard disks (VHDs), big data such as logs, database backups — pretty much anything. Blobs are stored in
containers, which are similar to folders. A container provides a grouping of a set of blobs. A storage account can
contain an unlimited number of containers, and a container can store an unlimited number of blobs.
Azure Storage is a good choice for big data and analytics solutions, because of its flexibility, high availability, and
low cost. It provides hot, cool, and archive storage tiers for different use cases. For more information, see Azure
Blob Storage: Hot, cool, and archive storage tiers.
Azure Blob storage can be accessed from Hadoop (available through HDInsight). HDInsight can use a blob
container in Azure Storage as the default file system for the cluster. Through a Hadoop distributed file system
(HDFS) interface provided by a WASB driver, the full set of components in HDInsight can operate directly on
structured or unstructured data stored as blobs. Azure Blob storage can also be accessed via Azure Synapse
Analytics using its PolyBase feature.
Other features that make Azure Storage a good choice are:
Multiple concurrency strategies.
Disaster recovery and high availability options.
Encryption at rest.
Role-based access control (RBAC) to control access using Azure Active Directory users and groups.
Azure Data Lake Store
Azure Data Lake Store is an enterprise-wide hyperscale repository for big data analytic workloads. Data Lake
enables you to capture data of any size, type, and ingestion speed in one single secure location for operational and
exploratory analytics.
Data Lake Store does not impose any limits on account sizes, file sizes, or the amount of data that can be stored in
a data lake. Data is stored durably by making multiple copies and there is no limit on the duration of time that the
data can be stored in the Data Lake. In addition to making multiple copies of files to guard against any unexpected
failures, Data lake spreads parts of a file over a number of individual storage servers. This improves the read
throughput when reading the file in parallel for performing data analytics.
Data Lake Store can be accessed from Hadoop (available through HDInsight) using the WebHDFS-compatible
REST APIs. You may consider using this as an alternative to Azure Storage when your individual or combined file
sizes exceed that which is supported by Azure Storage. However, there are performance tuning guidelines you
should follow when using Data Lake Store as your primary storage for an HDInsight cluster, with specific
guidelines for Spark, Hive, MapReduce, and Storm. Also, be sure to check Data Lake Store's regional availability,
because it is not available in as many regions as Azure Storage, and it needs to be located in the same region as
your HDInsight cluster.
Coupled with Azure Data Lake Analytics, Data Lake Store is specifically designed to enable analytics on the stored
data and is tuned for performance for data analytics scenarios. Data Lake Store can also be accessed via Azure
Synapse using its PolyBase feature.

Azure Cosmos DB
Azure Cosmos DB is Microsoft's globally distributed multi-model database. Cosmos DB guarantees single-digit-
millisecond latencies at the 99th percentile anywhere in the world, offers multiple well-defined consistency models
to fine-tune performance, and guarantees high availability with multi-homing capabilities.
Azure Cosmos DB is schema-agnostic. It automatically indexes all the data without requiring you to deal with
schema and index management. It's also multi-model, natively supporting document, key-value, graph, and
column-family data models.
Azure Cosmos DB features:
Geo-replication
Elastic scaling of throughput and storage worldwide
Five well-defined consistency levels

HBase on HDInsight
Apache HBase is an open-source, NoSQL database that is built on Hadoop and modeled after Google BigTable.
HBase provides random access and strong consistency for large amounts of unstructured and semi-structured
data in a schemaless database organized by column families.
Data is stored in the rows of a table, and data within a row is grouped by column family. HBase is schemaless in
the sense that neither the columns nor the type of data stored in them need to be defined before using them. The
open-source code scales linearly to handle petabytes of data on thousands of nodes. It can rely on data
redundancy, batch processing, and other features that are provided by distributed applications in the Hadoop
ecosystem.
The HDInsight implementation leverages the scale-out architecture of HBase to provide automatic sharding of
tables, strong consistency for reads and writes, and automatic failover. Performance is enhanced by in-memory
caching for reads and high-throughput streaming for writes. In most cases, you'll want to create the HBase cluster
inside a virtual network so other HDInsight clusters and applications can directly access the tables.
Azure Data Explorer
Azure Data Explorer is a fast and highly scalable data exploration service for log and telemetry data. It helps you
handle the many data streams emitted by modern software so you can collect, store, and analyze data. Azure Data
Explorer is ideal for analyzing large volumes of diverse data from any data source, such as websites, applications,
IoT devices, and more. This data is used for diagnostics, monitoring, reporting, machine learning, and additional
analytics capabilities. Azure Data Explorer makes it simple to ingest this data and enables you to do complex ad
hoc queries on the data in seconds.
Azure Data Explorer can be linearly scaled out for increasing ingestion and query processing throughput. An Azure
Data Explorer cluster can be deployed to a Virtual Network for enabling private networks.

Key selection criteria

To narrow the choices, start by answering these questions:
Do you need managed, high-speed, cloud-based storage for any type of text or binary data? If yes, then
select one of the file storage or analytics options.
Do you need file storage that is optimized for parallel analytics workloads and high throughput/IOPS? If
yes, then choose an option that is tuned to analytics workload performance.
Do you need to store unstructured or semi-structured data in a schemaless database? If so, select one of the
non-relational or analytics options. Compare options for indexing and database models. Depending on the
type of data you need to store, the primary database models may be the largest factor.
Can you use the service in your region? Check the regional availability for each Azure service. See Products
available by region.

Capability matrix
The following tables summarize the key differences in capabilities.
File storage capabilities
C A PA B IL IT Y A Z URE DATA L A K E STO RE A Z URE B LO B STO RA GE C O N TA IN ERS

Purpose Optimized storage for big data analytics General purpose object store for a wide
workloads variety of storage scenarios

Use cases Batch, streaming analytics, and machine Any type of text or binary data, such as
learning data such as log files, IoT data, application back end, backup data,
click streams, large datasets media storage for streaming, and
general purpose data

Structure Hierarchical file system Object store with flat namespace

Authentication Based on Azure Active Directory Based on shared secrets Account

Identities Access Keys and Shared Access
Signature Keys, and role-based access
control (RBAC)

Authentication protocol OAuth 2.0. Calls must contain a valid Hash-based message authentication
JWT (JSON web token) issued by Azure code (HMAC). Calls must contain a
Active Directory Base64-encoded SHA-256 hash over a
part of the HTTP request.
 C A PA B IL IT Y A Z URE DATA L A K E STO RE A Z URE B LO B STO RA GE C O N TA IN ERS

Authorization POSIX access control lists (ACLs). ACLs For account-level authorization use
based on Azure Active Directory Account Access Keys. For account,
identities can be set file and folder level. container, or blob authorization use
Shared Access Signature Keys.

Auditing Available. Available

Encryption at rest Transparent, server side Transparent, server side; Client-side

encryption

Developer SDKs .NET, Java, Python, Node.js .NET, Java, Python, Node.js, C++, Ruby

Analytics workload performance Optimized performance for parallel Not optimized for analytics workloads
analytics workloads, High Throughput
and IOPS

Size limits No limits on account sizes, file sizes or Specific limits documented here
number of files

Geo-redundancy Locally-redundant (LRS), globally Locally redundant (LRS), globally

redundant (GRS), read-access globally redundant (GRS), read-access globally
redundant (RA-GRS), zone-redundant redundant (RA-GRS), zone-redundant
(ZRS). (ZRS). See here for more information

NoSQL database capabilities

C A PA B IL IT Y A Z URE C O SM O S DB H B A SE O N H DIN SIGH T

Primary database model Document store, graph, key-value Wide column store
store, wide column store

Secondary indexes Yes No

SQL language support Yes Yes (using the Phoenix JDBC driver)

Consistency Strong, bounded-staleness, session, Strong

consistent prefix, eventual

Native Azure Functions integration Yes No

Automatic global distribution Yes No HBase cluster replication can be

configured across regions with eventual
consistency

Pricing model Elastically scalable request units (RUs) Per-minute pricing for HDInsight cluster
charged per-second as needed, (horizontal scaling of nodes), storage
elastically scalable storage

Analytical database capabilities

C A PA B IL IT Y A Z URE DATA EXP LO RER
C A PA B IL IT Y A Z URE DATA EXP LO RER

Primary database model Relational (column store), telemetry,

and time series store

SQL language support Yes

Pricing model Elastically scalable cluster instances

Authentication Based onAzure Active Directory

identities

Encryption at rest Supported, customer managed keys

Analytics workload performance Optimized performance for parallel

analytics workloads

Size limits Linearly scalable

 Understand data store models
12/18/2020 • 12 minutes to read • Edit Online

Modern business systems manage increasingly large volumes of heterogeneous data. This heterogeneity means
that a single data store is usually not the best approach. Instead, it's often better to store different types of data in
different data stores, each focused toward a specific workload or usage pattern. The term polyglot persistence is
used to describe solutions that use a mix of data store technologies. Therefore, it's important to understand the
main storage models and their tradeoffs.
Selecting the right data store for your requirements is a key design decision. There are literally hundreds of
implementations to choose from among SQL and NoSQL databases. Data stores are often categorized by how
they structure data and the types of operations they support. This article describes several of the most common
storage models. Note that a particular data store technology may support multiple storage models. For example,
a relational database management systems (RDBMS) may also support key/value or graph storage. In fact, there
is a general trend for so-called multi-model support, where a single database system supports several models.
But it's still useful to understand the different models at a high level.
Not all data stores in a given category provide the same feature-set. Most data stores provide server-side
functionality to query and process data. Sometimes this functionality is built into the data storage engine. In
other cases, the data storage and processing capabilities are separated, and there may be several options for
processing and analysis. Data stores also support different programmatic and management interfaces.
Generally, you should start by considering which storage model is best suited for your requirements. Then
consider a particular data store within that category, based on factors such as feature set, cost, and ease of
management.

Relational database management systems

Relational databases organize data as a series of two-dimensional tables with rows and columns. Most vendors
provide a dialect of the Structured Query Language (SQL) for retrieving and managing data. An RDBMS typically
implements a transactionally consistent mechanism that conforms to the ACID (Atomic, Consistent, Isolated,
Durable) model for updating information.
An RDBMS typically supports a schema-on-write model, where the data structure is defined ahead of time, and
all read or write operations must use the schema.
This model is very useful when strong consistency guarantees are important — where all changes are atomic,
and transactions always leave the data in a consistent state. However, an RDBMS generally can't scale out
horizontally without sharding the data in some way. Also, the data in an RDBMS must normalized, which isn't
appropriate for every data set.
Azure services
Azure SQL Database | (Security Baseline)
Azure Database for MySQL | (Security Baseline)
Azure Database for PostgreSQL | (Security Baseline)
Azure Database for MariaDB | (Security Baseline)
Workload
Records are frequently created and updated.
Multiple operations have to be completed in a single transaction.
Relationships are enforced using database constraints.
 Indexes are used to optimize query performance.
Data type
Data is highly normalized.
Database schemas are required and enforced.
Many-to-many relationships between data entities in the database.
Constraints are defined in the schema and imposed on any data in the database.
Data requires high integrity. Indexes and relationships need to be maintained accurately.
Data requires strong consistency. Transactions operate in a way that ensures all data are 100% consistent for
all users and processes.
Size of individual data entries is small to medium-sized.
Examples
Inventory management
Order management
Reporting database
Accounting

Key/value stores
A key/value store associates each data value with a unique key. Most key/value stores only support simple query,
insert, and delete operations. To modify a value (either partially or completely), an application must overwrite the
existing data for the entire value. In most implementations, reading or writing a single value is an atomic
operation.
An application can store arbitrary data as a set of values. Any schema information must be provided by the
application. The key/value store simply retrieves or stores the value by key.

Key/value stores are highly optimized for applications performing simple lookups, but are less suitable if you
need to query data across different key/value stores. Key/value stores are also not optimized for querying by
value.
A single key/value store can be extremely scalable, as the data store can easily distribute data across multiple
nodes on separate machines.
Azure services
Azure Cosmos DB Table API, etcd API (preview), and SQL API | (Cosmos DB Security Baseline)
Azure Cache for Redis | (Security Baseline)
Azure Table Storage | (Security Baseline)
Workload
Data is accessed using a single key, like a dictionary.
No joins, lock, or unions are required.
No aggregation mechanisms are used.
Secondary indexes are generally not used.
Data type
Each key is associated with a single value.
There is no schema enforcement.
No relationships between entities.
Examples
Data caching
Session management
User preference and profile management
Product recommendation and ad serving

Document databases
A document database stores a collection of documents, where each document consists of named fields and data.
The data can be simple values or complex elements such as lists and child collections. Documents are retrieved
by unique keys.
Typically, a document contains the data for single entity, such as a customer or an order. A document may contain
information that would be spread across several relational tables in an RDBMS. Documents don't need to have
the same structure. Applications can store different data in documents as business requirements change.

Azure service
Azure Cosmos DB SQL API | (Cosmos DB Security Baseline)
Workload
Insert and update operations are common.
No object-relational impedance mismatch. Documents can better match the object structures used in
application code.
Individual documents are retrieved and written as a single block.
Data requires index on multiple fields.
Data type
Data can be managed in de-normalized way.
Size of individual document data is relatively small.
Each document type can use its own schema.
 Documents can include optional fields.
Document data is semi-structured, meaning that data types of each field are not strictly defined.
Examples
Product catalog
Content management
Inventory management

Graph databases
A graph database stores two types of information, nodes and edges. Edges specify relationships between nodes.
Nodes and edges can have properties that provide information about that node or edge, similar to columns in a
table. Edges can also have a direction indicating the nature of the relationship.
Graph databases can efficiently perform queries across the network of nodes and edges and analyze the
relationships between entities. The following diagram shows an organization's personnel database structured as
a graph. The entities are employees and departments, and the edges indicate reporting relationships and the
departments in which employees work.

This structure makes it straightforward to perform queries such as "Find all employees who report directly or
indirectly to Sarah" or "Who works in the same department as John?" For large graphs with lots of entities and
relationships, you can perform very complex analyses very quickly. Many graph databases provide a query
language that you can use to traverse a network of relationships efficiently.
Azure services
Azure Cosmos DB Gremlin API | (Security Baseline)
SQL Server | (Security Baseline)
Workload
Complex relationships between data items involving many hops between related data items.
The relationship between data items are dynamic and change over time.
Relationships between objects are first-class citizens, without requiring foreign-keys and joins to traverse.
Data type
Nodes and relationships.
 Nodes are similar to table rows or JSON documents.
Relationships are just as important as nodes, and are exposed directly in the query language.
Composite objects, such as a person with multiple phone numbers, tend to be broken into separate, smaller
nodes, combined with traversable relationships
Examples
Organization charts
Social graphs
Fraud detection
Recommendation engines

Data analytics
Data analytics stores provide massively parallel solutions for ingesting, storing, and analyzing data. The data is
distributed across multiple servers to maximize scalability. Large data file formats such as delimiter files (CSV),
parquet, and ORC are widely used in data analytics. Historical data is typically stored in data stores such as blob
storage or Azure Data Lake Storage Gen2, which are then accessed by Azure Synapse, Databricks, or HDInsight as
external tables. A typical scenario using data stored as parquet files for performance, is described in the article
Use external tables with Synapse SQL.
Azure services
Azure Synapse Analytics | (Security Baseline)
Azure Data Lake | (Security Baseline)
Azure Data Explorer | (Security Baseline)
Azure Analysis Services
HDInsight | (Security Baseline)
Azure Databricks | (Security Baseline)
Workload
Data analytics
Enterprise BI
Data type
Historical data from multiple sources.
Usually denormalized in a "star" or "snowflake" schema, consisting of fact and dimension tables.
Usually loaded with new data on a scheduled basis.
Dimension tables often include multiple historic versions of an entity, referred to as a slowly changing
dimension.
Examples
Enterprise data warehouse

Column-family databases
A column-family database organizes data into rows and columns. In its simplest form, a column-family database
can appear very similar to a relational database, at least conceptually. The real power of a column-family
database lies in its denormalized approach to structuring sparse data.
You can think of a column-family database as holding tabular data with rows and columns, but the columns are
divided into groups known as column families. Each column family holds a set of columns that are logically
related together and are typically retrieved or manipulated as a unit. Other data that is accessed separately can be
stored in separate column families. Within a column family, new columns can be added dynamically, and rows
can be sparse (that is, a row doesn't need to have a value for every column).
The following diagram shows an example with two column families, Identity and Contact Info . The data for a
single entity has the same row key in each column-family. This structure, where the rows for any given object in a
column family can vary dynamically, is an important benefit of the column-family approach, making this form of
data store highly suited for storing structured, volatile data.

Unlike a key/value store or a document database, most column-family databases store data in key order, rather
than by computing a hash. Many implementations allow you to create indexes over specific columns in a column-
family. Indexes let you retrieve data by columns value, rather than row key.
Read and write operations for a row are usually atomic with a single column-family, although some
implementations provide atomicity across the entire row, spanning multiple column-families.
Azure services
Azure Cosmos DB Cassandra API | (Security Baseline)
HBase in HDInsight | (Security Baseline)
Workload
Most column-family databases perform write operations extremely quickly.
Update and delete operations are rare.
Designed to provide high throughput and low-latency access.
Supports easy query access to a particular set of fields within a much larger record.
Massively scalable.
Data type
Data is stored in tables consisting of a key column and one or more column families.
Specific columns can vary by individual rows.
Individual cells are accessed via get and put commands
Multiple rows are returned using a scan command.
Examples
Recommendations
Personalization
Sensor data
Telemetry
Messaging
Social media analytics
Web analytics
Activity monitoring
Weather and other time-series data
Search Engine Databases
A search engine database allows applications to search for information held in external data stores. A search
engine database can index massive volumes of data and provide near real-time access to these indexes.
Indexes can be multi-dimensional and may support free-text searches across large volumes of text data. Indexing
can be performed using a pull model, triggered by the search engine database, or using a push model, initiated
by external application code.
Searching can be exact or fuzzy. A fuzzy search finds documents that match a set of terms and calculates how
closely they match. Some search engines also support linguistic analysis that can return matches based on
synonyms, genre expansions (for example, matching dogs to pets ), and stemming (matching words with the
same root).
Azure service
Azure Search | (Security Baseline)
Workload
Data indexes from multiple sources and services.
Queries are ad-hoc and can be complex.
Full text search is required.
Ad hoc self-service query is required.
Data type
Semi-structured or unstructured text
Text with reference to structured data
Examples
Product catalogs
Site search
Logging

Time series databases

Time series data is a set of values organized by time. Time series databases typically collect large amounts of data
in real time from a large number of sources. Updates are rare, and deletes are often done as bulk operations.
Although the records written to a time-series database are generally small, there are often a large number of
records, and total data size can grow rapidly.
Azure service
Azure Time Series Insights
Workload
Records are generally appended sequentially in time order.
An overwhelming proportion of operations (95-99%) are writes.
Updates are rare.
Deletes occur in bulk, and are made to contiguous blocks or records.
Data is read sequentially in either ascending or descending time order, often in parallel.
Data type
A timestamp is used as the primary key and sorting mechanism.
Tags may define additional information about the type, origin, and other information about the entry.
Examples
 Monitoring and event telemetry.
Sensor or other IoT data.

Object storage
Object storage is optimized for storing and retrieving large binary objects (images, files, video and audio streams,
large application data objects and documents, virtual machine disk images). Large data files are also popularly
used in this model, for example, delimiter file (CSV), parquet, and ORC. Object stores can manage extremely large
amounts of unstructured data.
Azure service
Azure Blob Storage | (Security Baseline)
Azure Data Lake Storage Gen2 | (Security Baseline)
Workload
Identified by key.
Content is typically an asset such as a delimiter, image, or video file.
Content must be durable and external to any application tier.
Data type
Data size is large.
Value is opaque.
Examples
Images, videos, office documents, PDFs
Static HTML, JSON, CSS
Log and audit files
Database backups

Shared files
Sometimes, using simple flat files can be the most effective means of storing and retrieving information. Using
file shares enables files to be accessed across a network. Given appropriate security and concurrent access
control mechanisms, sharing data in this way can enable distributed services to provide highly scalable data
access for performing basic, low-level operations such as simple read and write requests.
Azure service
Azure Files | (Security Baseline)
Workload
Migration from existing apps that interact with the file system.
Requires SMB interface.
Data type
Files in a hierarchical set of folders.
Accessible with standard I/O libraries.
Examples
Legacy files
Shared content accessible among a number of VMs or app instances
Aided with this understanding of different data storage models, the next step is to evaluate your workload and
application, and decide which data store will meet your specific needs. Use the data storage decision tree to help
with this process.
 Choosing a data pipeline orchestration technology in
Azure
12/18/2020 • 2 minutes to read • Edit Online

Most big data solutions consist of repeated data processing operations, encapsulated in workflows. A pipeline
orchestrator is a tool that helps to automate these workflows. An orchestrator can schedule jobs, execute
workflows, and coordinate dependencies among tasks.

What are your options for data pipeline orchestration?

In Azure, the following services and tools will meet the core requirements for pipeline orchestration, control flow,
and data movement:
Azure Data Factory
Oozie on HDInsight
SQL Server Integration Services (SSIS)
These services and tools can be used independently from one another, or used together to create a hybrid
solution. For example, the Integration Runtime (IR) in Azure Data Factory V2 can natively execute SSIS packages in
a managed Azure compute environment. While there is some overlap in functionality between these services,
there are a few key differences.

Key Selection Criteria

To narrow the choices, start by answering these questions:
Do you need big data capabilities for moving and transforming your data? Usually this means multi-
gigabytes to terabytes of data. If yes, then narrow your options to those that best suited for big data.
Do you require a managed service that can operate at scale? If yes, select one of the cloud-based services
that aren't limited by your local processing power.
Are some of your data sources located on-premises? If yes, look for options that can work with both cloud
and on-premises data sources or destinations.
Is your source data stored in Blob storage on an HDFS filesystem? If so, choose an option that supports
Hive queries.

Capability matrix
The following tables summarize the key differences in capabilities.
General capabilities
SQ L SERVER IN T EGRAT IO N
C A PA B IL IT Y A Z URE DATA FA C TO RY SERVIC ES ( SSIS) O O Z IE O N H DIN SIGH T

Managed Yes No Yes

Cloud-based Yes No (local) Yes

 SQ L SERVER IN T EGRAT IO N
C A PA B IL IT Y A Z URE DATA FA C TO RY SERVIC ES ( SSIS) O O Z IE O N H DIN SIGH T

Prerequisite Azure Subscription SQL Server Azure Subscription,

HDInsight cluster

Management tools Azure Portal, PowerShell, SSMS, PowerShell Bash shell, Oozie REST API,
CLI, .NET SDK Oozie web UI

Pricing Pay per usage Licensing / pay for features No additional charge on top
of running the HDInsight
cluster

Pipeline capabilities
SQ L SERVER IN T EGRAT IO N
C A PA B IL IT Y A Z URE DATA FA C TO RY SERVIC ES ( SSIS) O O Z IE O N H DIN SIGH T

Copy data Yes Yes Yes

Custom transformations Yes Yes Yes (MapReduce, Pig, and

Hive jobs)

Azure Machine Learning Yes Yes (with scripting) No

scoring

HDInsight On-Demand Yes No No

Azure Batch Yes No No

Pig, Hive, MapReduce Yes No Yes

Spark Yes No No

Execute SSIS Package Yes Yes No

Control flow Yes Yes Yes

Access on-premises data Yes Yes No

Scalability capabilities
SQ L SERVER IN T EGRAT IO N
C A PA B IL IT Y A Z URE DATA FA C TO RY SERVIC ES ( SSIS) O O Z IE O N H DIN SIGH T

Scale up Yes No No

Scale out Yes No Yes (by adding worker nodes

to cluster)

Optimized for big data Yes No Yes

 Choosing a real-time message ingestion technology
in Azure
12/18/2020 • 2 minutes to read • Edit Online

Real time processing deals with streams of data that are captured in real-time and processed with minimal latency.
Many real-time processing solutions need a message ingestion store to act as a buffer for messages, and to
support scale-out processing, reliable delivery, and other message queuing semantics.

What are your options for real-time message ingestion?

Azure Event Hubs
Azure IoT Hub
Kafka on HDInsight

Azure Event Hubs

Azure Event Hubs is a highly scalable data streaming platform and event ingestion service, capable of receiving
and processing millions of events per second. Event Hubs can process and store events, data, or telemetry
produced by distributed software and devices. Data sent to an event hub can be transformed and stored using any
real-time analytics provider or batching/storage adapters. Event Hubs provides publish-subscribe capabilities with
low latency at massive scale, which makes it appropriate for big data scenarios.

Azure IoT Hub

Azure IoT Hub is a managed service that enables reliable and secure bidirectional communications between
millions of IoT devices and a cloud-based back end.
Feature of IoT Hub include:
Multiple options for device-to-cloud and cloud-to-device communication. These options include one-way
messaging, file transfer, and request-reply methods.
Message routing to other Azure services.
Queryable store for device metadata and synchronized state information.
Secure communications and access control using per-device security keys or X.509 certificates.
Monitoring of device connectivity and device identity management events.
In terms of message ingestion, IoT Hub is similar to Event Hubs. However, it was specifically designed for
managing IoT device connectivity, not just message ingestion. For more information, see Comparison of Azure IoT
Hub and Azure Event Hubs.

Kafka on HDInsight
Apache Kafka is an open-source distributed streaming platform that can be used to build real-time data pipelines
and streaming applications. Kafka also provides message broker functionality similar to a message queue, where
you can publish and subscribe to named data streams. It is horizontally scalable, fault-tolerant, and extremely fast.
Kafka on HDInsight provides a Kafka as a managed, highly scalable, and highly available service in Azure.
Some common use cases for Kafka are:
Messaging . Because it supports the publish-subscribe message pattern, Kafka is often used as a message
 broker.
Activity tracking . Because Kafka provides in-order logging of records, it can be used to track and re-create
activities, such as user actions on a web site.
Aggregation . Using stream processing, you can aggregate information from different streams to combine and
centralize the information into operational data.
Transformation . Using stream processing, you can combine and enrich data from multiple input topics into
one or more output topics.

Key selection criteria

To narrow the choices, start by answering these questions:
Do you need two-way communication between your IoT devices and Azure? If so, choose IoT Hub.
Do you need to manage access for individual devices and be able to revoke access to a specific device? If
yes, choose IoT Hub.

Capability matrix
The following tables summarize the key differences in capabilities.

C A PA B IL IT Y IOT H UB EVEN T H UB S K A F K A O N H DIN SIGH T

Cloud-to-device Yes No No
communications

Device-initiated file upload Yes No No

Device state information Device twins No No

Protocol support MQTT, AMQP, HTTPS 1 AMQP, HTTPS, Kafka Kafka Protocol
Protocol

Security Per-device identity; Shared access policies; Authentication using SASL;

revocable access control. limited revocation through pluggable authorization;
publisher policies. integration with external
authentication services
supported.

[1] You can also use Azure IoT protocol gateway as a custom gateway to enable protocol adaptation for IoT Hub.
For more information, see Comparison of Azure IoT Hub and Azure Event Hubs.
 Choosing a search data store in Azure
12/18/2020 • 2 minutes to read • Edit Online

This article compares technology choices for search data stores in Azure. A search data store is used to create and
store specialized indexes for performing searches on free-form text. The text that is indexed may reside in a
separate data store, such as blob storage. An application submits a query to the search data store, and the result is
a list of matching documents. For more information about this scenario, see Processing free-form text for search.

What are your options when choosing a search data store?

In Azure, all of the following data stores will meet the core requirements for search against free-form text data by
providing a search index:
Azure Cognitive Search
Elasticsearch
HDInsight with Solr
Azure SQL Database with full text search

Key selection criteria

For search scenarios, begin choosing the appropriate search data store for your needs by answering these
questions:
Do you want a managed service rather than managing your own servers?
Can you specify your index schema at design time? If not, choose an option that supports updateable
schemas.
Do you need an index only for full-text search, or do you also need rapid aggregation of numeric data and
other analytics? If you need functionality beyond full-text search, consider options that support additional
analytics.
Do you need a search index for log analytics, with support for log collection, aggregation, and visualizations
on indexed data? If so, consider Elasticsearch, which is part of a log analytics stack.
Do you need to index data in common document formats such as PDF, Word, PowerPoint, and Excel? If yes,
choose an option that provides document indexers.
Does your database have specific security needs? If yes, consider the security features listed below.

Capability matrix
The following tables summarize the key differences in capabilities.
General capabilities
H DIN SIGH T W IT H
C A PA B IL IT Y C O GN IT IVE SEA RC H EL A ST IC SEA RC H SO L R SQ L DATA B A SE

Is managed service Yes No Yes Yes

REST API Yes Yes Yes No

 H DIN SIGH T W IT H
C A PA B IL IT Y C O GN IT IVE SEA RC H EL A ST IC SEA RC H SO L R SQ L DATA B A SE

Programmability .NET, Java, Python, Java Java T-SQL

JavaScript

Document indexers Yes No Yes No

for common file types
(PDF, DOCX, TXT, and
so on)

Manageability capabilities
H DIN SIGH T W IT H
C A PA B IL IT Y C O GN IT IVE SEA RC H EL A ST IC SEA RC H SO L R SQ L DATA B A SE

Updateable schema Yes Yes Yes Yes

Supports scale out Yes Yes Yes No

Analytic workload capabilities

H DIN SIGH T W IT H
C A PA B IL IT Y C O GN IT IVE SEA RC H EL A ST IC SEA RC H SO L R SQ L DATA B A SE

Supports analytics No Yes Yes Yes

beyond full text
search

Part of a log analytics No Yes (ELK) No No

stack

Supports semantic Yes (find similar Yes Yes Yes

search documents only)

Security capabilities
H DIN SIGH T W IT H
C A PA B IL IT Y C O GN IT IVE SEA RC H EL A ST IC SEA RC H SO L R SQ L DATA B A SE

Row-level security Partial (requires Partial (requires Yes Yes

application query to application query to
filter by group id) filter by group id)

Transparent data No No No Yes

encryption

Restrict access to Yes Yes Yes Yes

specific IP addresses

Restrict access to Yes Yes Yes Yes

allow virtual network
access only
 H DIN SIGH T W IT H
C A PA B IL IT Y C O GN IT IVE SEA RC H EL A ST IC SEA RC H SO L R SQ L DATA B A SE

Active Directory No No No Yes

authentication
(integrated
authentication)

See also
Processing free-form text for search
 Choosing a stream processing technology in Azure
12/18/2020 • 2 minutes to read • Edit Online

This article compares technology choices for real-time stream processing in Azure.
Real-time stream processing consumes messages from either queue or file-based storage, process the messages,
and forward the result to another message queue, file store, or database. Processing may include querying,
filtering, and aggregating messages. Stream processing engines must be able to consume an endless streams of
data and produce results with minimal latency. For more information, see Real time processing.

What are your options when choosing a technology for real-time

processing?
In Azure, all of the following data stores will meet the core requirements supporting real-time processing:
Azure Stream Analytics
HDInsight with Spark Streaming
Apache Spark in Azure Databricks
HDInsight with Storm
Azure Functions
Azure App Service WebJobs
Apache Kafka streams API

Key Selection Criteria

For real-time processing scenarios, begin choosing the appropriate service for your needs by answering these
questions:
Do you prefer a declarative or imperative approach to authoring stream processing logic?
Do you need built-in support for temporal processing or windowing?
Does your data arrive in formats besides Avro, JSON, or CSV? If yes, consider options support any format
using custom code.
Do you need to scale your processing beyond 1 GB/s? If yes, consider the options that scale with the cluster
size.

Capability matrix
The following tables summarize the key differences in capabilities.
General capabilities
A PA C H E
A Z URE H DIN SIGH T SPA RK IN A Z URE A P P
ST REA M W IT H SPA RK A Z URE H DIN SIGH T A Z URE SERVIC E
C A PA B IL IT Y A N A LY T IC S ST REA M IN G DATA B RIC K S W IT H STO RM F UN C T IO N S W EB JO B S
 A PA C H E
A Z URE H DIN SIGH T SPA RK IN A Z URE A P P
ST REA M W IT H SPA RK A Z URE H DIN SIGH T A Z URE SERVIC E
C A PA B IL IT Y A N A LY T IC S ST REA M IN G DATA B RIC K S W IT H STO RM F UN C T IO N S W EB JO B S

Programmabil Stream C#/F#, Java, C#/F#, Java, C#, Java C#, F#, Java, C#, Java,
ity analytics Python, Scala Python, R, Node.js, Node.js, PHP,
query Scala Python Python
language,
JavaScript

Programming Declarative Mixture of Mixture of Imperative Imperative Imperative

paradigm declarative declarative
and and
imperative imperative

Pricing model Streaming Per cluster Databricks Per cluster Per function Per app
units hour units hour execution and service plan
resource hour
consumption

Integration capabilities
A PA C H E
A Z URE H DIN SIGH T SPA RK IN A Z URE A P P
ST REA M W IT H SPA RK A Z URE H DIN SIGH T A Z URE SERVIC E
C A PA B IL IT Y A N A LY T IC S ST REA M IN G DATA B RIC K S W IT H STO RM F UN C T IO N S W EB JO B S

Inputs Azure Event Event Hubs, Event Hubs, Event Hubs, Supported Service Bus,
Hubs, Azure IoT Hub, IoT Hub, IoT Hub, bindings Storage
IoT Hub, Kafka, HDFS, Kafka, HDFS, Storage Blobs, Queues,
Azure Blob Storage Blobs, Storage Blobs, Azure Data Storage Blobs,
storage Azure Data Azure Data Lake Store Event Hubs,
Lake Store Lake Store WebHooks,
Cosmos DB,
Files

Sinks Azure Data HDFS, Kafka, HDFS, Kafka, Event Hubs, Supported Service Bus,
Lake Store, Storage Blobs, Storage Blobs, Service Bus, bindings Storage
Azure SQL Azure Data Azure Data Kafka Queues,
Database, Lake Store, Lake Store, Storage Blobs,
Storage Blobs, Cosmos DB Cosmos DB Event Hubs,
Event Hubs, WebHooks,
Power BI, Cosmos DB,
Table Storage, Files
Service Bus
Queues,
Service Bus
Topics,
Cosmos DB,
Azure
Functions

Processing capabilities
 A PA C H E
A Z URE H DIN SIGH T SPA RK IN A Z URE A P P
ST REA M W IT H SPA RK A Z URE H DIN SIGH T A Z URE SERVIC E
C A PA B IL IT Y A N A LY T IC S ST REA M IN G DATA B RIC K S W IT H STO RM F UN C T IO N S W EB JO B S

Built-in Yes Yes Yes Yes No No

temporal/win
dowing
support

Input data Avro, JSON or Any format Any format Any format Any format Any format
formats CSV, UTF-8 using custom using custom using custom using custom using custom
encoded code code code code code

Scalability Query Bounded by Bounded by Bounded by Up to 200 Bounded by

partitions cluster size Databricks cluster size function app app service
cluster scale instances plan capacity
configuration processing in
parallel

Late arrival Yes Yes Yes Yes No No

and out of
order event
handling
support

See also:
Choosing a real-time message ingestion technology
Real time processing
 Tenancy models for SaaS applications
11/2/2020 • 3 minutes to read • Edit Online

In a Software as a Service (SaaS) model, each of your customers are a tenant of your application. Each tenant pays
for access to the SaaS application by paying a subscription fee. This article describes the application tenancy
models available to SaaS application builders.
When designing a SaaS application, you must choose the application tenancy model that best fits the needs of your
customers and your business. In general, the application tenancy model doesn't impact the functionality of an
application. But it likely impacts other aspects of the overall solution including scale, tenant isolation, cost per
tenant and operation complexity.

What are application tenancy models

The sections below explore the concepts of single, mixed, and multi-tenant application models.
Single tenant

In the single tenant model, a single dedicated instance of an application is deployed for each customer. For example,
with a N-tier architecture style application, all customers get a new dedicated instance of the web, middle, and data
tiers. These tiers are not shared between these customers.
Mixed tenant
In this model, one or more parts of an application are deployed as dedicated for each customer, and the rest is
shared between all customers. For example, with a N-tier architecture style application the web and middle tiers are
shared between all customers. However, a dedicated data tier and database is provisioned for each customer.
Multi-tenant

In this model, a single instance of the application is deployed for all customers and shared amongst them. For
example, with a N-tier architecture style application, the web, middle and data tiers are shared between all
customers.
A combination of these models can be provided for customers with different needs. For example, your basic tier of
service would run on a shared multi-tenant instance of your application. As a baseline, your customers can access
your app with lower performance or limited functionality for a lower cost. On top of this baseline, a dedicate service
tier could run on a single tenant model. For customers that need higher performance or additional functionality,
you can provide an isolated instance of your application for a higher cost.

Compare application tenancy models

In general, the tenancy model doesn't impact the functionality of an application, but it likely impacts other aspects
of the overall solution. The following table summarizes the differences between the application tenancy models:

M EA SUREM EN T SIN GL E T EN A N T M IXED M ULT I- T EN A N T

Scale Medium High Very High

Tenant isolation Very High High Low

Cost per tenant High Medium Low

Operational complexity Low-High - Individually Low-Medium - Need to Low-High - Individual tenant

simple, complex at scale. address complexity at scale. management is complex.

The measurement terms are as detailed below:

Scale: The number of concurrent customers (or tenants) your application can service.
Tenant isolation: The degree in which customers' data and performance is separated from other customers.
Cost per tenant: The relative amount it costs to run a single customer's tenant.
Operational complexity: How complex it is to achieve operational tasks like automation, security, monitoring,
maintenance, metering, application deployments, disaster recovery/BCP.
In general, the single tenancy application model and a small number of tenants, are simpler to manage. With a
large number of tenants, it can become highly complex without strong operational excellence practices. Conversely,
in the multi-tenancy application model, it is easier to manage the system as a whole, but highly complex when you
want to manage an individual customer tenant by itself, without strong operational excellence practices.

Additional considerations
Once you know the customer's needs and business goals, start asking these questions:
Do I operate in a highly regulated industry that requires customers data to be isolated from other customers?
Am I looking to rapidly scale my application to many thousands of clients?
Am I concerned about how much it costs to run each tenant/customer instance?
The answers to these questions will help you refine your tenancy requirement.

Next steps
Read more about application tenancy patterns in Multi-tenant SaaS database tenancy patterns.
 Monitoring Azure Databricks
12/18/2020 • 2 minutes to read • Edit Online

Azure Databricks is a fast, powerful Apache Spark–based analytics service that makes it easy to rapidly develop and
deploy big data analytics and artificial intelligence (AI) solutions. Many users take advantage of the simplicity of
notebooks in their Azure Databricks solutions. For users that require more robust computing options, Azure
Databricks supports the distributed execution of custom application code.
Monitoring is a critical part of any production-level solution, and Azure Databricks offers robust functionality for
monitoring custom application metrics, streaming query events, and application log messages. Azure Databricks
can send this monitoring data to different logging services.
The following articles show how to send monitoring data from Azure Databricks to Azure Monitor, the monitoring
data platform for Azure.
Send Azure Databricks application logs to Azure Monitor
Use dashboards to visualize Azure Databricks metrics
Troubleshoot performance bottlenecks
The code library that accompanies these articles extends the core monitoring functionality of Azure Databricks to
send Spark metrics, events, and logging information to Azure Monitor.
The audience for these articles and the accompanying code library are Apache Spark and Azure Databricks solution
developers. The code must be built into Java Archive (JAR) files and then deployed to an Azure Databricks cluster.
The code is a combination of Scala and Java, with a corresponding set of Maven project object model (POM) files to
build the output JAR files. Understanding of Java, Scala, and Maven are recommended as prerequisites.

Next steps
Start by building the code library and deploying it to your Azure Databricks cluster.
Send Azure Databricks application logs to Azure Monitor
 Send Azure Databricks application logs to Azure
Monitor
12/18/2020 • 2 minutes to read • Edit Online

This article shows how to send application logs and metrics from Azure Databricks to a Log Analytics workspace. It
uses the Azure Databricks Monitoring Library, which is available on GitHub.

Prerequisites
Configure your Azure Databricks cluster to use the monitoring library, as described in the GitHub readme.

NOTE
The monitoring library streams Apache Spark level events and Spark Structured Streaming metrics from your jobs to Azure
Monitor. You don't need to make any changes to your application code for these events and metrics.

Send application metrics using Dropwizard

Spark uses a configurable metrics system based on the Dropwizard Metrics Library. For more information, see
Metrics in the Spark documentation.
To send application metrics from Azure Databricks application code to Azure Monitor, follow these steps:
1. Build the spark-listeners-loganalytics-1.0-SNAPSHOT.jar JAR file as described in the GitHub readme.
2. Create Dropwizard gauges or counters in your application code. You can use the UserMetricsSystem class
defined in the monitoring library. The following example creates a counter named counter1 .

import org.apache.spark.metrics.UserMetricsSystems
import org.apache.spark.sql.SparkSession

object StreamingQueryListenerSampleJob {

private final val METRICS_NAMESPACE = "samplejob"

private final val COUNTER_NAME = "counter1"

def main(args: Array[String]): Unit = {

val spark = SparkSession

.builder
.getOrCreate

val driverMetricsSystem = UserMetricsSystems

.getMetricSystem(METRICS_NAMESPACE, builder => {
builder.registerCounter(COUNTER_NAME)
})

driverMetricsSystem.counter(COUNTER_NAME).inc(5)
}
}

The monitoring library includes a sample application that demonstrates how to use the UserMetricsSystem
class.
Send application logs using Log4j
To send your Azure Databricks application logs to Azure Log Analytics using the Log4j appender in the library,
follow these steps:
1. Build the spark-listeners-1.0-SNAPSHOT.jar and the spark-listeners-loganalytics-1.0-
SNAPSHOT.jar JAR file as described in the GitHub readme.
2. Create a log4j.proper ties configuration file for your application. Include the following configuration
properties. Substitute your application package name and log level where indicated:

log4j.appender.A1=com.microsoft.pnp.logging.loganalytics.LogAnalyticsAppender
log4j.appender.A1.layout=com.microsoft.pnp.logging.JSONLayout
log4j.appender.A1.layout.LocationInfo=false
log4j.additivity.<your application package name>=false
log4j.logger.<your application package name>=<log level>, A1

You can find a sample configuration file here.

3. In your application code, include the spark-listeners-loganalytics project, and import
com.microsoft.pnp.logging.Log4jconfiguration to your application code.

import com.microsoft.pnp.logging.Log4jConfiguration

4. Configure Log4j using the log4j.proper ties file you created in step 3:

getClass.getResourceAsStream("<path to file in your JAR file>/log4j.properties") {

stream => {
Log4jConfiguration.configure(stream)
}
}

5. Add Apache Spark log messages at the appropriate level in your code as required. For example, use the
logDebug method to send a debug log message. For more information, see Logging in the Spark
documentation.

logTrace("Trace message")
logDebug("Debug message")
logInfo("Info message")
logWarning("Warning message")
logError("Error message")

Run the sample application

The monitoring library includes a sample application that demonstrates how to send both application metrics and
application logs to Azure Monitor. To run the sample:
1. Build the spark-jobs project in the monitoring library, as described in the GitHub readme.
2. Navigate to your Databricks workspace and create a new job, as described here.
3. In the job detail page, select Set JAR .
4. Upload the JAR file from /src/spark-jobs/target/spark-jobs-1.0-SNAPSHOT.jar .
5. For Main class , enter com.microsoft.pnp.samplejob.StreamingQueryListenerSampleJob .
6. Select a cluster that is already configured to use the monitoring library. See Configure Azure Databricks to
send metrics to Azure Monitor.
When the job runs, you can view the application logs and metrics in your Log Analytics workspace.
Application logs appear under SparkLoggingEvent_CL:

SparkLoggingEvent_CL | where logger_name_s contains "com.microsoft.pnp"

Application metrics appear under SparkMetric_CL:

SparkMetric_CL | where name_s contains "rowcounter" | limit 50

IMPORTANT
After you verify the metrics appear, stop the sample application job.

Next steps
Deploy the performance monitoring dashboard that accompanies this code library to troubleshoot performance
issues in your production Azure Databricks workloads.
Use dashboards to visualize Azure Databricks metrics
 Use dashboards to visualize Azure Databricks metrics
12/18/2020 • 7 minutes to read • Edit Online

This article shows how to set up a Grafana dashboard to monitor Azure Databricks jobs for performance issues.
Azure Databricks is a fast, powerful, and collaborative Apache Spark–based analytics service that makes it easy to
rapidly develop and deploy big data analytics and artificial intelligence (AI) solutions. Monitoring is a critical
component of operating Azure Databricks workloads in production. The first step is to gather metrics into a
workspace for analysis. In Azure, the best solution for managing log data is Azure Monitor. Azure Databricks does
not natively support sending log data to Azure monitor, but a library for this functionality is available in GitHub.
This library enables logging of Azure Databricks service metrics as well as Apache Spark structure streaming
query event metrics. Once you've successfully deployed this library to an Azure Databricks cluster, you can further
deploy a set of Grafana dashboards that you can deploy as part of your production environment.

Prerequisites
Configure your Azure Databricks cluster to use the monitoring library, as described in the GitHub readme.

Deploy the Azure Log Analytics workspace

To deploy the Azure Log Analytics workspace, follow these steps:
1. Navigate to the /perftools/deployment/loganalytics directory.
2. Deploy the logAnalyticsDeploy.json Azure Resource Manager template. For more information about
deploying Resource Manager templates, see Deploy resources with Resource Manager templates and Azure
CLI. The template has the following parameters:
location : The region where the Log Analytics workspace and dashboards are deployed.
 ser viceTier : Rhe workspace pricing tier. See here for a list of valid values.
dataRetention (optional): The number of days the log data is retained in the Log Analytics workspace.
The default value is 30 days. If the pricing tier is Free , the data retention must be seven days.
workspaceName (optional): A name for the workspace. If not specified, the template generates a name.

az group deployment create --resource-group <resource-group-name> --template-file

logAnalyticsDeploy.json --parameters location='East US' serviceTier='Standalone'

This template creates the workspace and also creates a set of predefined queries that are used by dashboard.

Deploy Grafana in a virtual machine

Grafana is an open source project you can deploy to visualize the time series metrics stored in your Azure Log
Analytics workspace using the Grafana plugin for Azure Monitor. Grafana executes on a virtual machine (VM) and
requires a storage account, virtual network, and other resources. To deploy a virtual machine with the bitnami-
certified Grafana image and associated resources, follow these steps:
1. Use the Azure CLI to accept the Azure Marketplace image terms for Grafana.

az vm image accept-terms --publisher bitnami --offer grafana --plan default

2. Navigate to the /spark-monitoring/perftools/deployment/grafana directory in your local copy of the GitHub

repo.
3. Deploy the grafanaDeploy.json Resource Manager template as follows:

export DATA_SOURCE="https://raw.githubusercontent.com/mspnp/spark-
monitoring/master/perftools/deployment/grafana/AzureDataSource.sh"
az group deployment create \
--resource-group <resource-group-name> \
--template-file grafanaDeploy.json \
--parameters adminPass='<vm password>' dataSource=$DATA_SOURCE

Once the deployment is complete, the bitnami image of Grafana is installed on the virtual machine.

Update the Grafana password

As part of the setup process, the Grafana installation script outputs a temporary password for the admin user. You
need this temporary password to sign in. To obtain the temporary password, follow these steps:
1. Log in to the Azure portal.
2. Select the resource group where the resources were deployed.
3. Select the VM where Grafana was installed. If you used the default parameter name in the deployment
template, the VM name is prefaced with sparkmonitoring-vm-grafana .
4. In the Suppor t + troubleshooting section, click Boot diagnostics to open the boot diagnostics page.
5. Click Serial log on the boot diagnostics page.
6. Search for the following string: "Setting Bitnami application password to".
7. Copy the password to a safe location.
Next, change the Grafana administrator password by following these steps:
1. In the Azure portal, select the VM and click Over view .
2. Copy the public IP address.
3. Open a web browser and navigate to the following URL: http://<IP address>:3000 .
4. At the Grafana login screen, enter admin for the user name, and use the Grafana password from the previous
steps.
5. Once logged in, select Configuration (the gear icon).
6. Select Ser ver Admin .
7. On the Users tab, select the admin login.
8. Update the password.

Create an Azure Monitor data source

1. Create a service principal that allows Grafana to manage access to your Log Analytics workspace. For more
information, see Create an Azure service principal with Azure CLI

az ad sp create-for-rbac --name http://<service principal name> --role "Log Analytics Reader"

2. Note the values for appId, password, and tenant in the output from this command:

{
"appId": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
"displayName": "azure-cli-2019-03-27-00-33-39",
"name": "http://<service principal name>",
"password": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
"tenant": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx"
}

3. Log into Grafana as described earlier. Select Configuration (the gear icon) and then Data Sources .
4. In the Data Sources tab, click Add data source .
5. Select Azure Monitor as the data source type.
6. In the Settings section, enter a name for the data source in the Name textbox.
7. In the Azure Monitor API Details section, enter the following information:
Subscription Id: Your Azure subscription ID.
Tenant Id: The tenant ID from earlier.
Client Id: The value of "appId" from earlier.
Client Secret: The value of "password" from earlier.
8. In the Azure Log Analytics API Details section, check the Same Details as Azure Monitor API
checkbox.
9. Click Save & Test . If the Log Analytics data source is correctly configured, a success message is displayed.

Create the dashboard

Create the dashboards in Grafana by following these steps:
1. Navigate to the /perftools/dashboards/grafana directory in your local copy of the GitHub repo.
2. Run the following script:
 export WORKSPACE=<your Azure Log Analytics workspace ID>
export LOGTYPE=SparkListenerEvent_CL

sh DashGen.sh

The output from the script is a file named SparkMonitoringDash.json .

3. Return to the Grafana dashboard and select Create (the plus icon).
4. Select Impor t .
5. Click Upload .json File .
6. Select the SparkMonitoringDash.json file created in step 2.
7. In the Options section, under AL A , select the Azure Monitor data source created earlier.
8. Click Impor t .

Visualizations in the dashboards

Both the Azure Log Analytics and Grafana dashboards include a set of time-series visualizations. Each graph is
time-series plot of metric data related to an Apache Spark job, stages of the job, and tasks that make up each stage.
The visualizations are:
Job latency
This visualization shows execution latency for a job, which is a coarse view on the overall performance of a job.
Displays the job execution duration from start to completion. Note that the job start time is not the same as the job
submission time. Latency is represented as percentiles (10%, 30%, 50%, 90%) of job execution indexed by cluster
ID and application ID.
Stage latency
The visualization shows the latency of each stage per cluster, per application, and per individual stage. This
visualization is useful for identifying a particular stage that is running slowly.
Task latency
This visualization shows task execution latency. Latency is represented as a percentile of task execution per cluster,
stage name, and application.
Sum Task Execution per host
This visualization shows the sum of task execution latency per host running on a cluster. Viewing task execution
latency per host identifies hosts that have much higher overall task latency than other hosts. This may mean that
tasks have been inefficiently or unevenly distributed to hosts.
Task metrics
This visualization shows a set of the execution metrics for a given task's execution. These metrics include the size
and duration of a data shuffle, duration of serialization and deserialization operations, and others. For the full set of
metrics, view the Log Analytics query for the panel. This visualization is useful for understanding the operations
that make up a task and identifying resource consumption of each operation. Spikes in the graph represent costly
operations that should be investigated.
Cluster throughput
This visualization is a high-level view of work items indexed by cluster and application to represent the amount of
work done per cluster and application. It shows the number of jobs, tasks, and stages completed per cluster,
application, and stage in one minute increments.
Streaming Throughput/Latency
This visualization is related to the metrics associated with a structured streaming query. The graph shows the
number of input rows per second and the number of rows processed per second. The streaming metrics are also
represented per application. These metrics are sent when the OnQueryProgress event is generated as the
structured streaming query is processed and the visualization represents streaming latency as the amount of time,
in milliseconds, taken to execute a query batch.
Resource consumption per executor
Next is a set of visualizations for the dashboard show the particular type of resource and how it is consumed per
executor on each cluster. These visualizations help identify outliers in resource consumption per executor. For
example, if the work allocation for a particular executor is skewed, resource consumption will be elevated in
relation to other executors running on the cluster. This can be identified by spikes in the resource consumption for
an executor.
Executor compute time metrics
Next is a set of visualizations for the dashboard that show the ratio of executor serialize time, deserialize time, CPU
time, and Java virtual machine time to overall executor compute time. This demonstrates visually how much each
of these four metrics is contributing to overall executor processing.
Shuffle metrics
The final set of visualizations shows the data shuffle metrics associated with a structured streaming query across
all executors. These include shuffle bytes read, shuffle bytes written, shuffle memory, and disk usage in queries
where the file system is used.

Next steps
Troubleshoot performance bottlenecks
 Troubleshoot performance bottlenecks in Azure
Databricks
12/18/2020 • 6 minutes to read • Edit Online

This article describes how to use monitoring dashboards to find performance bottlenecks in Spark jobs on Azure
Databricks.
Azure Databricks is an Apache Spark–based analytics service that makes it easy to rapidly develop and deploy big
data analytics. Monitoring and troubleshooting performance issues is a critical when operating production Azure
Databricks workloads. To identify common performance issues, it's helpful to use monitoring visualizations based
on telemetry data.

Prerequisites
To set up the Grafana dashboards shown in this article:
Configure your Databricks cluster to send telemetry to a Log Analytics workspace, using the Azure
Databricks Monitoring Library. For details, see the GitHub readme.
Deploy Grafana in a virtual machine. See Use dashboards to visualize Azure Databricks metrics.
The Grafana dashboard that is deployed includes a set of time-series visualizations. Each graph is time-series plot
of metrics related to an Apache Spark job, the stages of the job, and tasks that make up each stage.

Azure Databricks performance overview

Azure Databricks is based on Apache Spark, a general-purpose distributed computing system. Application code,
known as a job , executes on an Apache Spark cluster, coordinated by the cluster manager. In general, a job is the
highest-level unit of computation. A job represents the complete operation performed by the Spark application. A
typical operation includes reading data from a source, applying data transformations, and writing the results to
storage or another destination.
Jobs are broken down into stages . The job advances through the stages sequentially, which means that later
stages must wait for earlier stages to complete. Stages contain groups of identical tasks that can be executed in
parallel on multiple nodes of the Spark cluster. Tasks are the most granular unit of execution taking place on a
subset of the data.
The next sections describe some dashboard visualizations that are useful for performance troubleshooting.

Job and stage latency

Job latency is the duration of a job execution from when it starts until it completes. It is shown as percentiles of a
job execution per cluster and application ID, to allow the visualization of outliers. The following graph shows a job
history where the 90th percentile reached 50 seconds, even though the 50th percentile was consistently around 10
seconds.
Investigate job execution by cluster and application, looking for spikes in latency. Once clusters and applications
with high latency are identified, move on to investigate stage latency.
Stage latency is also shown as percentiles to allow the visualization of outliers. Stage latency is broken out by
cluster, application, and stage name. Identify spikes in task latency in the graph to determine which tasks are
holding back completion of the stage.

The cluster throughput graph shows the number of jobs, stages, and tasks completed per minute. This helps you to
understand the workload in terms of the relative number of stages and tasks per job. Here you can see that the
number of jobs per minute ranges between 2 and 6, while the number of stages is about 12 – 24 per minute.
Sum of task execution latency
This visualization shows the sum of task execution latency per host running on a cluster. Use this graph to detect
tasks that run slowly due to the host slowing down on a cluster, or a misallocation of tasks per executor. In the
following graph, most of the hosts have a sum of about 30 seconds. However, two of the hosts have sums that
hover around 10 minutes. Either the hosts are running slow or the number of tasks per executor is misallocated.

The number of tasks per executor shows that two executors are assigned a disproportionate number of tasks,
causing a bottleneck.
Task metrics per stage
The task metrics visualization gives the cost breakdown for a task execution. You can use it see the relative time
spent on tasks such as serialization and deserialization. This data might show opportunities to optimize — for
example, by using broadcast variables to avoid shipping data. The task metrics also show the shuffle data size for a
task, and the shuffle read and write times. If these values are high, it means that a lot of data is moving across the
network.
Another task metric is the scheduler delay, which measures how long it takes to schedule a task. Ideally, this value
should be low compared to the executor compute time, which is the time spent actually executing the task.
The following graph shows a scheduler delay time (3.7 s) that exceeds the executor compute time (1.1 s). That
means more time is spent waiting for tasks to be scheduled than doing the actual work.

In this case, the problem was caused by having too many partitions, which caused a lot of overhead. Reducing the
number of partitions lowered the scheduler delay time. The next graph shows that most of the time is spent
executing the task.
Streaming throughput and latency
Streaming throughput is directly related to structured streaming. There are two important metrics associated with
streaming throughput: Input rows per second and processed rows per second. If input rows per second outpaces
processed rows per second, it means the stream processing system is falling behind. Also, if the input data comes
from Event Hubs or Kafka, then input rows per second should keep up with the data ingestion rate at the front end.
Two jobs can have similar cluster throughput but very different streaming metrics. The following screenshot shows
two different workloads. They are similar in terms of cluster throughput ( jobs, stages, and tasks per minute). But
the second run processes 12,000 rows/sec versus 4,000 rows/sec.

Streaming throughput is often a better business metric than cluster throughput, because it measures the number
of data records that are processed.

Resource consumption per executor

These metrics help to understand the work that each executor performs.
Percentage metrics measure how much time an executor spends on various things, expressed as a ratio of time
spent versus the overall executor compute time. The metrics are:
% Serialize time
% Deserialize time
% CPU executor time
% JVM time
These visualizations show how much each of these metrics contributes to overall executor processing.

Shuffle metrics are metrics related to data shuffling across the executors.
Shuffle I/O
Shuffle memory
File system usage
 Disk usage

Common performance bottlenecks

Two common performance bottlenecks in Spark are task stragglers and a non-optimal shuffle partition count.
Task stragglers
The stages in a job are executed sequentially, with earlier stages blocking later stages. If one task executes a shuffle
partition more slowly than other tasks, all tasks in the cluster must wait for the slow task to catch up before the
stage can end. This can happen for the following reasons:
1. A host or group of hosts are running slow. Symptoms: High task, stage, or job latency and low cluster
throughput. The summation of tasks latencies per host won't be evenly distributed. However, resource
consumption will be evenly distributed across executors.
2. Tasks have an expensive aggregation to execute (data skewing). Symptoms: High task latency, high stage
latency, high job latency, or low cluster throughput, but the summation of latencies per host is evenly
distributed. Resource consumption will be evenly distributed across executors.
3. If partitions are of unequal size, a larger partition may cause unbalanced task execution (partition skewing).
Symptoms: Executor resource consumption is high compared to other executors running on the cluster. All
tasks running on that executor will run slow and hold the stage execution in the pipeline. Those stages are
said to be stage barriers.
Non-optimal shuffle partition count
During a structured streaming query, the assignment of a task to an executor is a resource-intensive operation for
the cluster. If the shuffle data isn't the optimal size, the amount of delay for a task will negatively impact throughput
and latency. If there are too few partitions, the cores in the cluster will be underutilized which can result in
processing inefficiency. Conversely, if there are too many partitions, there's a great deal of management overhead
for a small number of tasks.
Use the resource consumption metrics to troubleshoot partition skewing and misallocation of executors on the
cluster. If a partition is skewed, executor resources will be elevated in comparison to other executors running on
the cluster.
For example, the following graph shows that the memory used by shuffling on the first two executors is 90X
bigger than the other executors:
 Transferring data to and from Azure
12/18/2020 • 7 minutes to read • Edit Online

There are several options for transferring data to and from Azure, depending on your needs.

Physical transfer
Using physical hardware to transfer data to Azure is a good option when:
Your network is slow or unreliable.
Getting additional network bandwidth is cost-prohibitive.
Security or organizational policies do not allow outbound connections when dealing with sensitive data.
If your primary concern is how long it will take to transfer your data, you may want to run a test to verify whether
network transfer is actually slower than physical transport.
There are two main options for physically transporting data to Azure:
Azure Impor t/Expor t . The Azure Import/Export service lets you securely transfer large amounts of data to
Azure Blob Storage or Azure Files by shipping internal SATA HDDs or SDDs to an Azure datacenter. You can
also use this service to transfer data from Azure Storage to hard disk drives and have these shipped to you
for loading on-premises.
Azure Data Box . Azure Data Box is a Microsoft-provided appliance that works much like the Azure
Import/Export service. Microsoft ships you a proprietary, secure, and tamper-resistant transfer appliance and
handles the end-to-end logistics, which you can track through the portal. One benefit of the Azure Data Box
service is ease of use. You don't need to purchase several hard drives, prepare them, and transfer files to
each one. Azure Data Box is supported by a number of industry-leading Azure partners to make it easier to
seamlessly use offline transport to the cloud from their products.

Command line tools and APIs

Consider these options when you want scripted and programmatic data transfer.
Azure CLI . The Azure CLI is a cross-platform tool that allows you to manage Azure services and upload data
to Azure Storage.
AzCopy . Use AzCopy from a Windows or Linux command-line to easily copy data to and from Azure Blob,
File, and Table storage with optimal performance. AzCopy supports concurrency and parallelism, and the
ability to resume copy operations when interrupted. You can also use AzCopy to copy data from AWS to
Azure. For programmatic access, the Microsoft Azure Storage Data Movement Library is the core framework
that powers AzCopy. It is provided as a .NET Core library.
PowerShell . The Start-AzureStorageBlobCopy PowerShell cmdlet is an option for Windows administrators
who are used to PowerShell.
AdlCopy . AdlCopy enables you to copy data from Azure Storage Blobs into Data Lake Store. It can also be
used to copy data between two Azure Data Lake Store accounts. However, it cannot be used to copy data
from Data Lake Store to Storage Blobs.
Distcp . If you have an HDInsight cluster with access to Data Lake Store, you can use Hadoop ecosystem
tools like Distcp to copy data to and from an HDInsight cluster storage (WASB) into a Data Lake Store
account.
 Sqoop . Sqoop is an Apache project and part of the Hadoop ecosystem. It comes preinstalled on all
HDInsight clusters. It allows data transfer between an HDInsight cluster and relational databases such as
SQL, Oracle, MySQL, and so on. Sqoop is a collection of related tools, including import and export. Sqoop
works with HDInsight clusters using either Azure Storage blobs or Data Lake Store attached storage.
PolyBase . PolyBase is a technology that accesses data outside of the database through the T-SQL language.
In SQL Server 2016, it allows you to run queries on external data in Hadoop or to import/export data from
Azure Blob Storage. In Azure Synapse Analytics, you can import/export data from Azure Blob Storage and
Azure Data Lake Store. Currently, PolyBase is the fastest method of importing data into Azure Synapse.
Hadoop command line . When you have data that resides on an HDInsight cluster head node, you can use
the hadoop -copyFromLocal command to copy that data to your cluster's attached storage, such as Azure
Storage blob or Azure Data Lake Store. In order to use the Hadoop command, you must first connect to the
head node. Once connected, you can upload a file to storage.

Graphical interface
Consider the following options if you are only transferring a few files or data objects and don't need to automate
the process.
Azure Storage Explorer . Azure Storage Explorer is a cross-platform tool that lets you manage the contents
of your Azure storage accounts. It allows you to upload, download, and manage blobs, files, queues, tables,
and Azure Cosmos DB entities. Use it with Blob storage to manage blobs and folders, as well as upload and
download blobs between your local file system and Blob storage, or between storage accounts.
Azure por tal . Both Blob storage and Data Lake Store provide a web-based interface for exploring files and
uploading new files one at a time. This is a good option if you do not want to install any tools or issue
commands to quickly explore your files, or to simply upload a handful of new ones.

Data pipeline
Azure Data Factor y . Azure Data Factory is a managed service best suited for regularly transferring files between
a number of Azure services, on-premises, or a combination of the two. Using Azure Data Factory, you can create
and schedule data-driven workflows (called pipelines) that ingest data from disparate data stores. It can process
and transform the data by using compute services such as Azure HDInsight Hadoop, Spark, Azure Data Lake
Analytics, and Azure Machine Learning. Create data-driven workflows for orchestrating and automating data
movement and data transformation.

Key Selection Criteria

For data transfer scenarios, choose the appropriate system for your needs by answering these questions:
Do you need to transfer very large amounts of data, where doing so over an Internet connection would take
too long, be unreliable, or too expensive? If yes, consider physical transfer.
Do you prefer to script your data transfer tasks, so they are reusable? If so, select one of the command line
options or Azure Data Factory.
Do you need to transfer a very large amount of data over a network connection? If so, select an option that is
optimized for big data.
Do you need to transfer data to or from a relational database? If yes, choose an option that supports one or
more relational databases. Note that some of these options also require a Hadoop cluster.
Do you need an automated data pipeline or workflow orchestration? If yes, consider Azure Data Factory.
Capability matrix
The following tables summarize the key differences in capabilities.
Physical transfer
C A PA B IL IT Y A Z URE IM P O RT / EXP O RT SERVIC E A Z URE DATA B O X

Form factor Internal SATA HDDs or SDDs Secure, tamper-proof, single hardware
appliance

Microsoft manages shipping logistics No Yes

Integrates with partner products No Yes

Custom appliance No Yes

Command line tools

Hadoop/HDInsight:

C A PA B IL IT Y DISTC P SQ O O P H A DO O P C L I

Optimized for big data Yes Yes Yes

Copy to relational database No Yes No

Copy from relational No Yes No

database

Copy to Blob storage Yes Yes Yes

Copy from Blob storage Yes Yes No

Copy to Data Lake Store Yes Yes Yes

Copy from Data Lake Store Yes Yes No

Other :

C A PA B IL IT Y A Z URE C L I AZC OPY P O W ERSH EL L A DL C O P Y P O LY B A SE

Compatible Linux, OS X, Linux, Windows Windows Linux, OS X, SQL Server, Azure

platforms Windows Windows Synapse

Optimized for big No Yes No Yes 1 Yes 2

data

Copy to relational No No No No Yes

database

Copy from No No No No Yes

relational
database
 C A PA B IL IT Y A Z URE C L I AZC OPY P O W ERSH EL L A DL C O P Y P O LY B A SE

Copy to Blob Yes Yes Yes No Yes

storage

Copy from Blob Yes Yes Yes Yes Yes

storage

Copy to Data No Yes Yes Yes Yes

Lake Store

Copy from Data No No Yes Yes Yes

Lake Store

[1] AdlCopy is optimized for transferring big data when used with a Data Lake Analytics account.
[2] PolyBase performance can be increased by pushing computation to Hadoop and using PolyBase scale-out
groups to enable parallel data transfer between SQL Server instances and Hadoop nodes.
Graphical interface and Azure Data Factory
C A PA B IL IT Y A Z URE STO RA GE EXP LO RER A Z URE P O RTA L * A Z URE DATA FA C TO RY

Optimized for big data No No Yes

Copy to relational database No No Yes

Copy from relational No No Yes

database

Copy to Blob storage Yes No Yes

Copy from Blob storage Yes No Yes

Copy to Data Lake Store No No Yes

Copy from Data Lake Store No No Yes

Upload to Blob storage Yes Yes Yes

Upload to Data Lake Store Yes Yes Yes

Orchestrate data transfers No No Yes

Custom data No No Yes

transformations

Pricing model Free Free Pay per usage

* Azure portal in this case means using the web-based exploration tools for Blob storage and Data Lake Store.
 Extending on-premises data solutions to the cloud
12/18/2020 • 7 minutes to read • Edit Online

When organizations move workloads and data to the cloud, their on-premises datacenters often continue to play
an important role. The term hybrid cloud refers to a combination of public cloud and on-premises datacenters, to
create an integrated IT environment that spans both. Some organizations use hybrid cloud as a path to migrate
their entire datacenter to the cloud over time. Other organizations use cloud services to extend their existing on-
premises infrastructure.
This article describes some considerations and best practices for managing data in a hybrid cloud solution,

When to use a hybrid solution

Consider using a hybrid solution in the following scenarios:
As a transition strategy during a longer-term migration to a fully cloud-native solution.
When regulations or policies do not permit moving specific data or workloads to the cloud.
For disaster recovery and fault tolerance, by replicating data and services between on-premises and cloud
environments.
To reduce latency between your on-premises datacenter and remote locations, by hosting part of your
architecture in Azure.

Challenges
Creating a consistent environment in terms of security, management, and development, and avoiding
duplication of work.
Creating a reliable, low latency and secure data connection between your on-premises and cloud
environments.
Replicating your data and modifying applications and tools to use the correct data stores within each
environment.
Securing and encrypting data that is hosted in the cloud but accessed from on-premises, or vice versa.

On-premises data stores

On-premises data stores include databases and files. There may be several reasons to keep these data stores local.
There may be regulations or policies that do not permit moving specific data or workloads to the cloud. Data
sovereignty, privacy, or security concerns may favor on-premises placement. During a migration, you may want to
keep some data local to an application that hasn't been migrated yet.
Considerations in placing application data in a public cloud include:
Cost . The cost of storage in Azure can be significantly lower than the cost of maintaining storage with
similar characteristics in an on-premises datacenter. Of course, many companies have existing investments
in high-end SANs, so these cost advantages may not reach full fruition until existing hardware ages out.
Elastic scale . Planning and managing data capacity growth in an on-premises environment can be
challenging, particularly when data growth is difficult to predict. These applications can take advantage of
the capacity-on-demand and virtually unlimited storage available in the cloud. This consideration is less
relevant for applications that consist of relatively static-sized datasets.
 Disaster recover y . Data stored in Azure can be automatically replicated within an Azure region and across
geographic regions. In hybrid environments, these same technologies can be used to replicate between on-
premises and cloud-based data stores.

Extending data stores to the cloud

There are several options for extending on-premises data stores to the cloud. One option is to have on-premises
and cloud replicas. This can help achieve a high level of fault tolerance, but may require making changes to
applications to connect to the appropriate data store in the event of a failover.
Another option is to move a portion of the data to cloud storage, while keeping the more current or more highly
accessed data on-premises. This method can provide a more cost-effective option for long-term storage, as well as
improve data access response times by reducing your operational data set.
A third option is to keep all data on-premises, but use cloud computing to host applications. To do this, you would
host your application in the cloud and connect it to your on-premises data store over a secure connection.

Azure Stack
For a complete hybrid cloud solution, consider using Microsoft Azure Stack. Azure Stack is a hybrid cloud platform
that lets you provide Azure services from your datacenter. This helps maintain consistency between on-premises
and Azure, by using identical tools and requiring no code changes.
The following are some use cases for Azure and Azure Stack:
Edge and disconnected solutions . Address latency and connectivity requirements by processing data
locally in Azure Stack and then aggregating in Azure for further analytics, with common application logic
across both.
Cloud applications that meet varied regulations . Develop and deploy applications in Azure, with the
flexibility to deploy the same applications on-premises on Azure Stack to meet regulatory or policy
requirements.
Cloud application model on-premises . Use Azure to update and extend existing applications or build
new ones. Use consistent DevOps processes across Azure in the cloud and Azure Stack on-premises.

SQL Server data stores

If you are running SQL Server on-premises, you can use Microsoft Azure Blob Storage service for backup and
restore. For more information, see SQL Server Backup and Restore with Microsoft Azure Blob Storage Service. This
capability gives you limitless offsite storage, and the ability to share the same backups between SQL Server
running on-premises and SQL Server running in a virtual machine in Azure.
Azure SQL Database is a managed relational database-as-a service. Because Azure SQL Database uses the
Microsoft SQL Server Engine, applications can access data in the same way with both technologies. Azure SQL
Database can also be combined with SQL Server in useful ways. For example, the SQL Server Stretch Database
feature lets an application access what looks like a single table in a SQL Server database while some or all rows of
that table might be stored in Azure SQL Database. This technology automatically moves data that's not accessed for
a defined period of time to the cloud. Applications reading this data are unaware that any data has been moved to
the cloud.
Maintaining data stores on-premises and in the cloud can be challenging when you desire to keep the data
synchronized. You can address this with SQL Data Sync, a service built on Azure SQL Database that lets you
synchronize the data you select, bi-directionally across multiple Azure SQL databases and SQL Server instances.
While Data Sync makes it easy to keep your data up-to-date across these various data stores, it should not be used
for disaster recovery or for migrating from on-premises SQL Server to Azure SQL Database.
For disaster recovery and business continuity, you can use AlwaysOn Availability Groups to replicate data across
two or more instances of SQL Server, some of which can be running on Azure virtual machines in another
geographic region.

Network shares and file-based data stores

In a hybrid cloud architecture, it is common for an organization to keep newer files on-premises while archiving
older files to the cloud. This is sometimes called file tiering, where there is seamless access to both sets of files, on-
premises and cloud-hosted. This approach helps to minimize network bandwidth usage and access times for newer
files, which are likely to be accessed the most often. At the same time, you get the benefits of cloud-based storage
for archived data.
Organizations may also wish to move their network shares entirely to the cloud. This would be desirable, for
example, if the applications that access them are also located in the cloud. This procedure can be done using data
orchestration tools.
Azure StorSimple offers the most complete integrated storage solution for managing storage tasks between your
on-premises devices and Azure cloud storage. StorSimple is an efficient, cost-effective, and easily manageable
storage area network (SAN) solution that eliminates many of the issues and expenses associated with enterprise
storage and data protection. It uses the proprietary StorSimple 8000 series device, integrates with cloud services,
and provides a set of integrated management tools.
Another way to use on-premises network shares alongside cloud-based file storage is with Azure Files. Azure Files
offers fully managed file shares that you can access with the standard Server Message Block (SMB) protocol
(sometimes referred to as CIFS). You can mount Azure Files as a file share on your local computer, or use them with
existing applications that access local or network share files.
To synchronize file shares in Azure Files with your on-premises Windows Servers, use Azure File Sync. One major
benefit of Azure File Sync is the ability to tier files between your on-premises file server and Azure Files. This lets
you keep only the newest and most recently accessed files locally.
For more information, see Deciding when to use Azure Blob storage, Azure Files, or Azure Disks.

Hybrid networking
This article focused on hybrid data solutions, but another consideration is how to extend your on-premises
network to Azure. For more information about this aspect of hybrid solutions, see:
Choose a solution for connecting an on-premises network to Azure
Hybrid network reference architectures
 Securing data solutions
12/18/2020 • 5 minutes to read • Edit Online

For many, making data accessible in the cloud, particularly when transitioning from working exclusively in on-
premises data stores, can cause some concern around increased accessibility to that data and new ways in which to
secure it.

Challenges
Centralizing the monitoring and analysis of security events stored in numerous logs.
Implementing encryption and authorization management across your applications and services.
Ensuring that centralized identity management works across all of your solution components, whether on-
premises or in the cloud.

Data Protection
The first step to protecting information is identifying what to protect. Develop clear, simple, and well-communicated
guidelines to identify, protect, and monitor the most important data assets anywhere they reside. Establish the
strongest protection for assets that have a disproportionate impact on the organization's mission or profitability.
These are known as high value assets, or HVAs. Perform stringent analysis of HVA lifecycle and security
dependencies, and establish appropriate security controls and conditions. Similarly, identify and classify sensitive
assets, and define the technologies and processes to automatically apply security controls.
Once the data you need to protect has been identified, consider how you will protect the data at rest and data in
transit.
Data at rest : Data that exists statically on physical media, whether magnetic or optical disk, on premises or in
the cloud.
Data in transit : Data while it is being transferred between components, locations or programs, such as over
the network, across a service bus (from on-premises to cloud and vice-versa), or during an input/output
process.
To learn more about protecting your data at rest or in transit, see Azure Data Security and Encryption Best Practices.

Access Control
Central to protecting your data in the cloud is a combination of identity management and access control. Given the
variety and type of cloud services, as well as the rising popularity of hybrid cloud, there are several key practices
you should follow when it comes to identity and access control:
Centralize your identity management.
Enable Single Sign-On (SSO).
Deploy password management.
Enforce multi-factor authentication for users.
Use role based access control (RBAC).
Conditional Access Policies should be configured, which enhances the classic concept of user identity with
additional properties related to user location, device type, patch level, and so on.
Control locations where resources are created using resource manager.
Actively monitor for suspicious activities
For more information, see Azure Identity Management and access control security best practices.

Auditing
Beyond the identity and access monitoring previously mentioned, the services and applications that you use in the
cloud should be generating security-related events that you can monitor. The primary challenge to monitoring
these events is handling the quantities of logs , in order to avoid potential problems or troubleshoot past ones.
Cloud-based applications tend to contain many moving parts, most of which generate some level of logging and
telemetry. Use centralized monitoring and analysis to help you manage and make sense of the large amount of
information.
For more information, see Azure Logging and Auditing.

Securing data solutions in Azure

Encryption
Vir tual machines . Use Azure Disk Encryption to encrypt the attached disks on Windows or Linux VMs. This
solution integrates with Azure Key Vault to control and manage the disk-encryption keys and secrets.
Azure Storage . Use Azure Storage Service Encryption to automatically encrypt data at rest in Azure Storage.
Encryption, decryption, and key management are totally transparent to users. Data can also be secured in transit by
using client-side encryption with Azure Key Vault. For more information, see Client-Side Encryption and Azure Key
Vault for Microsoft Azure Storage.
SQL Database and Azure Synapse Analytics . Use Transparent Data Encryption (TDE) to perform real-time
encryption and decryption of your databases, associated backups, and transaction log files without requiring any
changes to your applications. SQL Database can also use Always Encrypted to help protect sensitive data at rest on
the server, during movement between client and server, and while the data is in use. You can use Azure Key Vault to
store your Always Encrypted encryption keys.
Rights management
Azure Rights Management is a cloud-based service that uses encryption, identity, and authorization policies to
secure files and email. It works across multiple devices — phones, tablets, and PCs. Information can be protected
both within your organization and outside your organization because that protection remains with the data, even
when it leaves your organization's boundaries.
Access control
Use role-based access control (RBAC) to restrict access to Azure resources based on user roles. If you are using
Active Directory on-premises, you can synchronize with Azure AD to provide users with a cloud identity based on
their on-premises identity.
Use Conditional access in Azure Active Directory to enforce controls on the access to applications in your
environment based on specific conditions. For example, your policy statement could take the form of: When
contractors are trying to access our cloud apps from networks that are not trusted, then block access.
Azure AD Privileged Identity Management can help you manage, control, and monitor your users and what sorts of
tasks they are performing with their admin privileges. This is an important step to limiting who in your
organization can carry out privileged operations in Azure AD, Azure, Microsoft 365, or SaaS apps, as well as
monitor their activities.
Network
To protect data in transit, always use SSL/TLS when exchanging data across different locations. Sometimes you
need to isolate your entire communication channel between your on-premises and cloud infrastructure by using
either a virtual private network (VPN) or ExpressRoute. For more information, see Extending on-premises data
solutions to the cloud.
Use network security groups to reduce the number of potential attack vectors. A network security group contains a
list of security rules that allow or deny inbound or outbound network traffic based on source or destination IP
address, port, and protocol.
Use Virtual Network service endpoints to secure Azure SQL or Azure Storage resources, so that only traffic from
your virtual network can access these resources.
VMs within an Azure Virtual Network (VNet) can securely communicate with other VNets using virtual network
peering. Network traffic between peered virtual networks is private. Traffic between the virtual networks is kept on
the Microsoft backbone network.
For more information, see Azure network security
Monitoring
Azure Security Center automatically collects, analyzes, and integrates log data from your Azure resources, the
network, and connected partner solutions, such as firewall solutions, to detect real threats and reduce false
positives.
Log Analytics provides centralized access to your logs and helps you analyze that data and create custom alerts.
Azure SQL Database Threat Detection detects anomalous activities indicating unusual and potentially harmful
attempts to access or exploit databases. Security officers or other designated administrators can receive an
immediate notification about suspicious database activities as they occur. Each notification provides details of the
suspicious activity and recommends how to further investigate and mitigate the threat.
 Run Apache Cassandra on Azure VMs
12/18/2020 • 8 minutes to read • Edit Online

This article describes performance considerations for running Apache Cassandra on Azure virtual machines.
These recommendations are based on the results of performance tests, which you can find on GitHub. You should
use these recommendations as a baseline and then test against your own workload.

Azure VM sizes and disk types

Cassandra workloads on Azure commonly use either Standard_DS14_v2 or Standard_DS13_v2 virtual machines.
Cassandra workloads benefit from having more memory in the VM, so consider memory optimized virtual
machine sizes, such as Standard_DS14_v2, or local-storage optimized sizes such as Standard_L16s_v2.
For durability, data and commit logs are commonly stored on a stripe set of two to four 1-TB premium managed
disks (P30).
Cassandra nodes should not be too data-dense. We recommend having at most 1 – 2 TB of data per VM and
enough free space for compaction. To achieve the highest possible combined throughput and IOPS using premium
managed disks, we recommend creating a stripe set from a few 1-TB disks, instead of using a single 2-TB or 4-TB
disk. For example, on a DS14_v2 VM, four 1-TB disks have a maximum IOPS of 4 × 5000 = 20 K, versus 7.5 K for a
single 4-TB disk.
As Azure Ultra Disks become more widely available across regions, evaluate them for Cassandra workloads that
need smaller disk capacity. They can provide higher IOPS/throughput and lower latency on VM sizes like
Standard_D32s_v3 and Standard_D16s_v3.
For Cassandra workloads that don't need durable storage — that is, where data can be easily reconstructed from
another storage medium — consider using Standard_L32s_v2 or Standard_L16s_v2 VMs. These VMs sizes have
large and fast local temporary NVMe disks.
For more information, see Comparing performance of Azure local/ephemeral vs attached/persistent disks (GitHub).

Accelerated Networking
Cassandra nodes make heavy use of the network to send and receive data from the client VM and to communicate
between nodes for replication. For optimal performance, Cassandra VMs benefit from high-throughput and low-
latency network.
We recommended enabling Accelerated Networking on the NIC of the Cassandra node and on VMs running client
applications accessing Cassandra.
Accelerated networking requires a modern Linux distribution with the latest drivers, such as Cent OS 7.5+ or
Ubuntu 16.x/18.x. For more information, see Create a Linux virtual machine with Accelerated Networking.

Azure VM data disk caching

Cassandra read workloads perform best when random-access disk latency is low. We recommend using Azure
managed disks with ReadOnly caching enabled. ReadOnly caching provides lower average latency, because the data
is read from the cache on the host instead of going to the backend storage.
Read-heavy, random-read workloads like Cassandra benefit from the lower read latency even though cached mode
has lower throughput limits than uncached mode. (For example, DS14_v2 virtual machines have a maximum
cached throughput of 512 MB/s versus uncached of 768 MB/s.)
ReadOnly caching is particularly helpful for Cassandra time-series and other workloads where the working dataset
fits in the host cache and data is not constantly overwritten. For example, DS14_v2 provides a cache size of 512 GB,
which could store up to 50% of the data from a Cassandra node with 1-2 TB data density.
There is no significant performance penalty from cache-misses when ReadOnly caching is enabled, so cached mode
is recommended for all but the most write-heavy workloads.
For more information, see Comparing Azure VM data disk caching configurations (GitHub).

Linux read-ahead
In most Linux distributions in the Azure Marketplace, the default block device read-ahead setting is 4096 KB.
Cassandra's read IOs are usually random and relatively small. Therefore, having a large read-ahead wastes
throughput by reading parts of files that aren't needed.
To minimize unnecessary lookahead, set the Linux block device read-ahead setting to 8 KB. (See Recommended
production settings in the DataStax documentation.)
Configure 8 KB read-ahead for all block devices in the stripe set and on the array device itself (for example,
/dev/md0 ).

For more information, see Comparing impact of disk read-ahead settings (GitHub).

Disk array mdadm chunk size

When running Cassandra on Azure, it's common to create an mdadm stripe set (that is, RAID 0) of multiple data
disks to increase the overall disk throughput and IOPS closer to the VM limits. Optimal disk stripe size is an
application-specific setting. For example, for SQL Server OLTP workloads, the recommendation is 64 KB. For data
warehousing workloads, the recommendation is 256 KB.
Our tests found no significant difference between chunk sizes of 64k, 128k, and 256k for Cassandra read
workloads. There seems to be a small, slightly noticeable, advantage to the 128k chunk size. Therefore, we
recommend the following:
If you're already using a chunk size of 64 K or 256 K, it doesn't make sense rebuild the disk array to use 128
K size.
For a new configuration, it makes senses to use 128 K from the beginning.
For more information, see Measuring impact of mdadm chunk sizes on Cassandra performance (GitHub).

Commit log filesystem

Cassandra writes perform best when commit logs are on disks with high throughput and low latency. In the default
configuration, Cassandra 3.x flushes data from memory to the commit log file every ~10 seconds and doesn't
touch the disk for every write. In this configuration, write performance is almost identical whether the commit log is
on premium attached disks versus local/ephemeral disks.
Commit logs must be durable, so that a restarted node can reconstruct any data not yet in data files from the
flushed commit logs. For better durability, store commit logs on premium managed disks and not on local storage,
which can be lost if the VM is migrated to another host.
Based on our tests, Cassandra on CentOS 7.x may have lower write performance when commit logs are on the xfs
versus ext4 filesystem. Turning on commit log compression brings xfs performance in line with ext4. Compressed
xfs performs as well as compressed and non-compressed ext4 in our tests.
For more information, see Observations on ext4 and xfs file systems and compressed commit logs (GitHub).

Measuring baseline VM performance

After deploying the VMs for the Cassandra ring, run a few synthetic tests to establish baseline network and disk
performance. Use these tests to confirm that performance is in line with expectations, based on the VM size.
Later, when you run the actual workload, knowing the performance baseline makes it easier to investigate potential
bottlenecks. For example, knowing the baseline performance for network egress on the VM can help to rule out
network as a bottleneck.
For more information about running performance tests, see Validating baseline Azure VM Performance (GitHub).
Document size
Cassandra read and write performance depends on the document size. You can expect to see higher latency and
lower operations/second when reading or writing with larger documents.
For more information, see Comparing relative performance of various Cassandra document sizes (GitHub).
Replication factor
Most Cassandra workloads use a replication factor (RF) of 3 when using attached premium disks and even 5 when
using temporary/ephemeral local disks. The number of nodes in the Cassandra ring should be a multiple of the
replication factor. For example, RF 3 implies a ring of 3, 6, 9, or 12 nodes, while RF 5 would have 5, 10, 15, or 20
nodes. When using RF greater than 1 and a consistency level of LOCAL_QUORUM, it's normal for read and write
performance to be lower than the same workload running with RF 1.
For more information, see Comparing relative performance of various replication factors (GitHub).
Linux page caching
When reading data files, Cassandra's Java code uses regular file I/O and benefits from Linux page caching. After
parts of the file are read one time, the read content is stored in the OS page cache. Subsequent read access to the
same data is much faster.
For this reason, when executing read performance tests against the same data, the second and subsequent reads
will appear to be much faster than the original read, which needed to access data on the remote data disk or from
the host cache when ReadOnly is enabled. To get similar performance measurements on subsequent runs, clear the
Linux page cache and restart the Cassandra service to clear its internal memory. When ReadOnly caching is
enabled, the data might be in the host cache, and subsequent reads will be faster even after clearing the OS page
cache and restarting the Cassandra service.
For more information, see Observations on Cassandra usage of Linux page caching (GitHub).

Multi-datacenter replication
Cassandra natively supports the concept of multiple data centers, making it easy to configure one Cassandra ring
across multiple Azure regions or across availability zones within one region.
For a multiregion deployment, use Azure Global VNet-peering to connect the virtual networks in the different
regions. When VMs are deployed in the same region but in separate availability zones, the VMs can be in the same
virtual network.
It's important to measure the baseline roundtrip latency between regions. Network latency between regions can be
10-100 times higher than latency within a region. Expect a lag between data appearing in the second region when
using LOCAL_QUORUM write consistency, or significantly decreased performance of writes when using
EACH_QUORUM.
When running Apache Cassandra at scale, and specifically in a multi-DC environment, node repair becomes
challenging. Tools such as Reaper can help to coordinate repairs at scale (for example, across all the nodes in a data
center, one data center at a time, to limit the load on the whole cluster). However, node repair for large clusters is
not yet a fully solved problem and applies in all environments, whether on-premises or in the cloud.
When nodes are added to a secondary region, performance will not scale linearly, because some bandwidth and
CPU/disk resources are spent on receiving and sending replication traffic across regions.
For more information, see Measuring impact of multi-dc cross-region replication (GitHub).
Hinted-handoff configuration
In a multiregion Cassandra ring, write workloads with consistency level of LOCAL_QUORUM may lose data in the
secondary region. By default, Cassandra hinted handoff is throttled to a relatively low maximum throughput and
three-hour hint lifetime. For workloads with heavy writes, we recommended increasing the hinted handoff throttle
and hint window time to ensure hints are not dropped before they are replicated.
For more information, see Observations on hinted handoff in cross-region replication (GitHub).

Next steps
For more information about these performance results, see Cassandra on Azure VMs Performance Experiments.
For information on general Cassandra settings, not specific to Azure, see:
DataStax Recommended Production Settings
Apache Cassandra Hardware Choices
Apache Cassandra Configuration File
The following reference architecture deploys Cassandra as part of an n-tier configuration:
Linux N-tier application in Azure with Apache Cassandra
 Building microservices on Azure
12/18/2020 • 5 minutes to read • Edit Online

Microservices are a popular architectural style for building applications that are resilient, highly scalable,
independently deployable, and able to evolve quickly. But a successful microservices architecture requires a
different approach to designing and building applications.
A microservices architecture consists of a collection of small, autonomous services. Each service is self-contained
and should implement a single business capability.

What are microservices?

Microservices are small, independent, and loosely coupled. A single small team of developers can write and
maintain a service.
Each service is a separate codebase, which can be managed by a small development team.
Services can be deployed independently. A team can update an existing service without rebuilding and
redeploying the entire application.
Services are responsible for persisting their own data or external state. This differs from the traditional
model, where a separate data layer handles data persistence.
Services communicate with each other by using well-defined APIs. Internal implementation details of each
service are hidden from other services.
Services don't need to share the same technology stack, libraries, or frameworks.
Besides for the services themselves, some other components appear in a typical microservices architecture:
Management/orchestration . This component is responsible for placing services on nodes, identifying failures,
rebalancing services across nodes, and so forth. Typically this component is an off-the-shelf technology such as
Kubernetes, rather than something custom built.
API Gateway . The API gateway is the entry point for clients. Instead of calling services directly, clients call the API
gateway, which forwards the call to the appropriate services on the back end.
Advantages of using an API gateway include:
It decouples clients from services. Services can be versioned or refactored without needing to update all of
 the clients.
Services can use messaging protocols that are not web friendly, such as AMQP.
The API Gateway can perform other cross-cutting functions such as authentication, logging, SSL termination,
and load balancing.

Benefits
Agility. Because microservices are deployed independently, it's easier to manage bug fixes and feature
releases. You can update a service without redeploying the entire application, and roll back an update if
something goes wrong. In many traditional applications, if a bug is found in one part of the application, it
can block the entire release process. New features may be held up waiting for a bug fix to be integrated,
tested, and published.
Small, focused teams . A microservice should be small enough that a single feature team can build, test,
and deploy it. Small team sizes promote greater agility. Large teams tend be less productive, because
communication is slower, management overhead goes up, and agility diminishes.
Small code base . In a monolithic application, there is a tendency over time for code dependencies to
become tangled Adding a new feature requires touching code in a lot of places. By not sharing code or data
stores, a microservices architecture minimizes dependencies, and that makes it easier to add new features.
Mix of technologies . Teams can pick the technology that best fits their service, using a mix of technology
stacks as appropriate.
Fault isolation . If an individual microservice becomes unavailable, it won't disrupt the entire application, as
long as any upstream microservices are designed to handle faults correctly (for example, by implementing
circuit breaking).
Scalability . Services can be scaled independently, letting you scale out subsystems that require more
resources, without scaling out the entire application. Using an orchestrator such as Kubernetes or Service
Fabric, you can pack a higher density of services onto a single host, which allows for more efficient
utilization of resources.
Data isolation . It is much easier to perform schema updates, because only a single microservice is affected.
In a monolithic application, schema updates can become very challenging, because different parts of the
application may all touch the same data, making any alterations to the schema risky.

Challenges
The benefits of microservices don't come for free. Here are some of the challenges to consider before embarking
on a microservices architecture.
Complexity . A microservices application has more moving parts than the equivalent monolithic
application. Each service is simpler, but the entire system as a whole is more complex.
Development and testing . Writing a small service that relies on other dependent services requires a
different approach than a writing a traditional monolithic or layered application. Existing tools are not
always designed to work with service dependencies. Refactoring across service boundaries can be difficult. It
is also challenging to test service dependencies, especially when the application is evolving quickly.
Lack of governance . The decentralized approach to building microservices has advantages, but it can also
lead to problems. You may end up with so many different languages and frameworks that the application
becomes hard to maintain. It may be useful to put some project-wide standards in place, without overly
restricting teams' flexibility. This especially applies to cross-cutting functionality such as logging.
Network congestion and latency . The use of many small, granular services can result in more
 interservice communication. Also, if the chain of service dependencies gets too long (service A calls B, which
calls C...), the additional latency can become a problem. You will need to design APIs carefully. Avoid overly
chatty APIs, think about serialization formats, and look for places to use asynchronous communication
patterns.
Data integrity . With each microservice responsible for its own data persistence. As a result, data
consistency can be a challenge. Embrace eventual consistency where possible.
Management . To be successful with microservices requires a mature DevOps culture. Correlated logging
across services can be challenging. Typically, logging must correlate multiple service calls for a single user
operation.
Versioning . Updates to a service must not break services that depend on it. Multiple services could be
updated at any given time, so without careful design, you might have problems with backward or forward
compatibility.
Skillset . Microservices are highly distributed systems. Carefully evaluate whether the team has the skills
and experience to be successful.

Process for building a microservices architecture

The articles listed here present a structured approach for designing, building, and operating a microservices
architecture.
Domain analysis. To avoid some common pitfalls when designing microservices, use domain analysis to define
your microservice boundaries. Follow these steps:
1. Use domain analysis to model microservices.
2. Use tactical DDD to design microservices.
3. Identify microservice boundaries.
Design the ser vices . Microservices require a different approach to designing and building applications. For more
information, see Designing a microservices architecture.
Operate in production . Because microservices architectures are distributed, you must have robust operations
for deployment and monitoring.
CI/CD for microservices architectures
Build a CI/CD pipeline for microservices on Kubernetes
Monitor microservices running on Azure Kubernetes Service (AKS)

Microservices reference architectures for Azure

Microservices architecture on Azure Kubernetes Service (AKS)
Microservices architecture on Azure Service Fabric
 Using domain analysis to model microservices
12/18/2020 • 8 minutes to read • Edit Online

One of the biggest challenges of microservices is to define the boundaries of individual services. The general rule
is that a service should do "one thing" — but putting that rule into practice requires careful thought. There is no
mechanical process that will produce the "right" design. You have to think deeply about your business domain,
requirements, and goals. Otherwise, you can end up with a haphazard design that exhibits some undesirable
characteristics, such as hidden dependencies between services, tight coupling, or poorly designed interfaces. This
article shows a domain-driven approach to designing microservices.
This article uses a drone delivery service as a running example. You can read more about the scenario and the
corresponding reference implementation here.

Introduction
Microservices should be designed around business capabilities, not horizontal layers such as data access or
messaging. In addition, they should have loose coupling and high functional cohesion. Microservices are loosely
coupled if you can change one service without requiring other services to be updated at the same time. A
microservice is cohesive if it has a single, well-defined purpose, such as managing user accounts or tracking
delivery history. A service should encapsulate domain knowledge and abstract that knowledge from clients. For
example, a client should be able to schedule a drone without knowing the details of the scheduling algorithm or
how the drone fleet is managed.
Domain-driven design (DDD) provides a framework that can get you most of the way to a set of well-designed
microservices. DDD has two distinct phases, strategic and tactical. In strategic DDD, you are defining the large-scale
structure of the system. Strategic DDD helps to ensure that your architecture remains focused on business
capabilities. Tactical DDD provides a set of design patterns that you can use to create the domain model. These
patterns include entities, aggregates, and domain services. These tactical patterns will help you to design
microservices that are both loosely coupled and cohesive.

In this article and the next, we'll walk through the following steps, applying them to the Drone Delivery application:
1. Start by analyzing the business domain to understand the application's functional requirements. The output
of this step is an informal description of the domain, which can be refined into a more formal set of domain
models.
2. Next, define the bounded contexts of the domain. Each bounded context contains a domain model that
represents a particular subdomain of the larger application.
3. Within a bounded context, apply tactical DDD patterns to define entities, aggregates, and domain services.
4. Use the results from the previous step to identify the microservices in your application.
In this article, we cover the first three steps, which are primarily concerned with DDD. In the next article, we'll
identify the microservices. However, it's important to remember that DDD is an iterative, ongoing process. Service
boundaries aren't fixed in stone. As an application evolves, you may decide to break apart a service into several
smaller services.
 NOTE
This article doesn't show a complete and comprehensive domain analysis. We deliberately kept the example brief, to illustrate
the main points. For more background on DDD, we recommend Eric Evans' Domain-Driven Design, the book that first
introduced the term. Another good reference is Implementing Domain-Driven Design by Vaughn Vernon.

Scenario: Drone delivery

Fabrikam, Inc. is starting a drone delivery service. The company manages a fleet of drone aircraft. Businesses
register with the service, and users can request a drone to pick up goods for delivery. When a customer schedules
a pickup, a backend system assigns a drone and notifies the user with an estimated delivery time. While the
delivery is in progress, the customer can track the location of the drone, with a continuously updated ETA.
This scenario involves a fairly complicated domain. Some of the business concerns include scheduling drones,
tracking packages, managing user accounts, and storing and analyzing historical data. Moreover, Fabrikam wants
to get to market quickly and then iterate quickly, adding new functionality and capabilities. The application needs
to operate at cloud scale, with a high service level objective (SLO). Fabrikam also expects that different parts of the
system will have very different requirements for data storage and querying. All of these considerations lead
Fabrikam to choose a microservices architecture for the Drone Delivery application.

Analyze the domain

Using a DDD approach will help you to design microservices so that every service forms a natural fit to a
functional business requirement. It can help you to avoid the trap of letting organizational boundaries or
technology choices dictate your design.
Before writing any code, you need a bird's eye view of the system that you are creating. DDD starts by modeling
the business domain and creating a domain model. The domain model is an abstract model of the business
domain. It distills and organizes domain knowledge, and provides a common language for developers and domain
experts.
Start by mapping all of the business functions and their connections. This will likely be a collaborative effort that
involves domain experts, software architects, and other stakeholders. You don't need to use any particular
formalism. Sketch a diagram or draw on whiteboard.
As you fill in the diagram, you may start to identify discrete subdomains. Which functions are closely related?
Which functions are core to the business, and which provide ancillary services? What is the dependency graph?
During this initial phase, you aren't concerned with technologies or implementation details. That said, you should
note the place where the application will need to integrate with external systems, such as CRM, payment
processing, or billing systems.

Example: Drone delivery application

After some initial domain analysis, the Fabrikam team came up with a rough sketch that depicts the Drone Delivery
domain.

Shipping is placed in the center of the diagram, because it's core to the business. Everything else in the
diagram exists to enable this functionality.
Drone management is also core to the business. Functionality that is closely related to drone management
includes drone repair and using predictive analysis to predict when drones need servicing and
maintenance.
ETA analysis provides time estimates for pickup and delivery.
 Third-par ty transpor tation will enable the application to schedule alternative transportation methods if a
package cannot be shipped entirely by drone.
Drone sharing is a possible extension of the core business. The company may have excess drone capacity
during certain hours, and could rent out drones that would otherwise be idle. This feature will not be in the
initial release.
Video sur veillance is another area that the company might expand into later.
User accounts , Invoicing , and Call center are subdomains that support the core business.
Notice that at this point in the process, we haven't made any decisions about implementation or technologies.
Some of the subsystems may involve external software systems or third-party services. Even so, the application
needs to interact with these systems and services, so it's important to include them in the domain model.

NOTE
When an application depends on an external system, there is a risk that the external system's data schema or API will leak
into your application, ultimately compromising the architectural design. This is particularly true with legacy systems that may
not follow modern best practices, and may use convoluted data schemas or obsolete APIs. In that case, it's important to
have a well-defined boundary between these external systems and the application. Consider using the Strangler Fig pattern
or the Anti-Corruption Layer pattern for this purpose.

Define bounded contexts

The domain model will include representations of real things in the world — users, drones, packages, and so forth.
But that doesn't mean that every part of the system needs to use the same representations for the same things.
For example, subsystems that handle drone repair and predictive analysis will need to represent many physical
characteristics of drones, such as their maintenance history, mileage, age, model number, performance
characteristics, and so on. But when it's time to schedule a delivery, we don't care about those things. The
scheduling subsystem only needs to know whether a drone is available, and the ETA for pickup and delivery.
If we tried to create a single model for both of these subsystems, it would be unnecessarily complex. It would also
become harder for the model to evolve over time, because any changes will need to satisfy multiple teams
working on separate subsystems. Therefore, it's often better to design separate models that represent the same
real-world entity (in this case, a drone) in two different contexts. Each model contains only the features and
attributes that are relevant within its particular context.
This is where the DDD concept of bounded contexts comes into play. A bounded context is simply the boundary
within a domain where a particular domain model applies. Looking at the previous diagram, we can group
functionality according to whether various functions will share a single domain model.

Bounded contexts are not necessarily isolated from one another. In this diagram, the solid lines connecting the
bounded contexts represent places where two bounded contexts interact. For example, Shipping depends on User
Accounts to get information about customers, and on Drone Management to schedule drones from the fleet.
In the book Domain Driven Design, Eric Evans describes several patterns for maintaining the integrity of a domain
model when it interacts with another bounded context. One of the main principles of microservices is that services
communicate through well-defined APIs. This approach corresponds to two patterns that Evans calls Open Host
Service and Published Language. The idea of Open Host Service is that a subsystem defines a formal protocol (API)
for other subsystems to communicate with it. Published Language extends this idea by publishing the API in a
form that other teams can use to write clients. In the article Designing APIs for microservices, we discuss using
OpenAPI Specification (formerly known as Swagger) to define language-agnostic interface descriptions for REST
APIs, expressed in JSON or YAML format.
For the rest of this journey, we will focus on the Shipping bounded context.

Next steps
After completing a domain analysis, the next step is to apply tactical DDD, to define your domain models with
more precision.
Tactical DDD
 Using tactical DDD to design microservices
12/18/2020 • 5 minutes to read • Edit Online

During the strategic phase of DDD, you are mapping out the business domain and defining bounded contexts for
your domain models. Tactical DDD is when you define your domain models with more precision. The tactical
patterns are applied within a single bounded context. In a microservices architecture, we are particularly interested
in the entity and aggregate patterns. Applying these patterns will help us to identify natural boundaries for the
services in our application (see the next article in this series). As a general principle, a microservice should be no
smaller than an aggregate, and no larger than a bounded context. First, we'll review the tactical patterns. Then we'll
apply them to the Shipping bounded context in the Drone Delivery application.

Overview of the tactical patterns

This section provides a brief summary of the tactical DDD patterns, so if you are already familiar with DDD, you
can probably skip this section. The patterns are described in more detail in chapters 5 – 6 of Eric Evans' book, and
in Implementing Domain-Driven Design by Vaughn Vernon.

Entities . An entity is an object with a unique identity that persists over time. For example, in a banking application,
customers and accounts would be entities.
An entity has a unique identifier in the system, which can be used to look up or retrieve the entity. That doesn't
mean the identifier is always exposed directly to users. It could be a GUID or a primary key in a database.
An identity may span multiple bounded contexts, and may endure beyond the lifetime of the application. For
example, bank account numbers or government-issued IDs are not tied to the lifetime of a particular
application.
The attributes of an entity may change over time. For example, a person's name or address might change, but
they are still the same person.
An entity can hold references to other entities.
Value objects . A value object has no identity. It is defined only by the values of its attributes. Value objects are
also immutable. To update a value object, you always create a new instance to replace the old one. Value objects
can have methods that encapsulate domain logic, but those methods should have no side-effects on the object's
state. Typical examples of value objects include colors, dates and times, and currency values.
Aggregates . An aggregate defines a consistency boundary around one or more entities. Exactly one entity in an
aggregate is the root. Lookup is done using the root entity's identifier. Any other entities in the aggregate are
children of the root, and are referenced by following pointers from the root.
The purpose of an aggregate is to model transactional invariants. Things in the real world have complex webs of
relationships. Customers create orders, orders contain products, products have suppliers, and so on. If the
application modifies several related objects, how does it guarantee consistency? How do we keep track of
invariants and enforce them?
Traditional applications have often used database transactions to enforce consistency. In a distributed application,
however, that's often not feasible. A single business transaction may span multiple data stores, or may be long
running, or may involve third-party services. Ultimately it's up to the application, not the data layer, to enforce the
invariants required for the domain. That's what aggregates are meant to model.

NOTE
An aggregate might consist of a single entity, without child entities. What makes it an aggregate is the transactional
boundary.

Domain and application ser vices . In DDD terminology, a service is an object that implements some logic
without holding any state. Evans distinguishes between domain services, which encapsulate domain logic, and
application services, which provide technical functionality, such as user authentication or sending an SMS
message. Domain services are often used to model behavior that spans multiple entities.

NOTE
The term service is overloaded in software development. The definition here is not directly related to microservices.

Domain events . Domain events can be used to notify other parts of the system when something happens. As the
name suggests, domain events should mean something within the domain. For example, "a record was inserted
into a table" is not a domain event. "A delivery was cancelled" is a domain event. Domain events are especially
relevant in a microservices architecture. Because microservices are distributed and don't share data stores, domain
events provide a way for microservices to coordinate with each other. The article Interservice communication
discusses asynchronous messaging in more detail.
There are a few other DDD patterns not listed here, including factories, repositories, and modules. These can be
useful patterns for when you are implementing a microservice, but they are less relevant when designing the
boundaries between microservice.

Drone delivery: Applying the patterns

We start with the scenarios that the Shipping bounded context must handle.
A customer can request a drone to pick up goods from a business that is registered with the drone delivery
service.
The sender generates a tag (barcode or RFID) to put on the package.
A drone will pick up and deliver a package from the source location to the destination location.
When a customer schedules a delivery, the system provides an ETA based on route information, weather
conditions, and historical data.
When the drone is in flight, a user can track the current location and the latest ETA.
Until a drone has picked up the package, the customer can cancel a delivery.
The customer is notified when the delivery is completed.
The sender can request delivery confirmation from the customer, in the form of a signature or finger print.
Users can look up the history of a completed delivery.
From these scenarios, the development team identified the following entities .
Delivery
Package
 Drone
Account
Confirmation
Notification
Tag
The first four, Delivery, Package, Drone, and Account, are all aggregates that represent transactional consistency
boundaries. Confirmations and Notifications are child entities of Deliveries, and Tags are child entities of Packages.
The value objects in this design include Location, ETA, PackageWeight, and PackageSize.
To illustrate, here is a UML diagram of the Delivery aggregate. Notice that it holds references to other aggregates,
including Account, Package, and Drone.

There are two domain events:

While a drone is in flight, the Drone entity sends DroneStatus events that describe the drone's location and
status (in-flight, landed).
The Delivery entity sends DeliveryTracking events whenever the stage of a delivery changes. These include
DeliveryCreated, DeliveryRescheduled, DeliveryHeadedToDropoff, and DeliveryCompleted.
Notice that these events describe things that are meaningful within the domain model. They describe something
about the domain, and aren't tied to a particular programming language construct.
The development team identified one more area of functionality, which doesn't fit neatly into any of the entities
described so far. Some part of the system must coordinate all of the steps involved in scheduling or updating a
delivery. Therefore, the development team added two domain ser vices to the design: a Scheduler that
coordinates the steps, and a Supervisor that monitors the status of each step, in order to detect whether any steps
have failed or timed out. This is a variation of the Scheduler Agent Supervisor pattern.

Next steps
The next step is to define the boundaries for each microservice.
Identify microservice boundaries
 Identifying microservice boundaries
12/18/2020 • 5 minutes to read • Edit Online

What is the right size for a microservice? You often hear something to the effect of, "not too big and not too small"
— and while that's certainly correct, it's not very helpful in practice. But if you start from a carefully designed
domain model, it's much easier to reason about microservices.

This article uses a drone delivery service as a running example. You can read more about the scenario and the
corresponding reference implementation here.

From domain model to microservices

In the previous article, we defined a set of bounded contexts for a Drone Delivery application. Then we looked
more closely at one of these bounded contexts, the Shipping bounded context, and identified a set of entities,
aggregates, and domain services for that bounded context.
Now we're ready to go from domain model to application design. Here's an approach that you can use to derive
microservices from the domain model.
1. Start with a bounded context. In general, the functionality in a microservice should not span more than one
bounded context. By definition, a bounded context marks the boundary of a particular domain model. If you
find that a microservice mixes different domain models together, that's a sign that you may need to go back
and refine your domain analysis.
2. Next, look at the aggregates in your domain model. Aggregates are often good candidates for
microservices. A well-designed aggregate exhibits many of the characteristics of a well-designed
microservice, such as:
An aggregate is derived from business requirements, rather than technical concerns such as data access
or messaging.
An aggregate should have high functional cohesion.
An aggregate is a boundary of persistence.
Aggregates should be loosely coupled.
3. Domain services are also good candidates for microservices. Domain services are stateless operations
 across multiple aggregates. A typical example is a workflow that involves several microservices. We'll see an
example of this in the Drone Delivery application.
4. Finally, consider non-functional requirements. Look at factors such as team size, data types, technologies,
scalability requirements, availability requirements, and security requirements. These factors may lead you to
further decompose a microservice into two or more smaller services, or do the opposite and combine
several microservices into one.
After you identify the microservices in your application, validate your design against the following criteria:
Each service has a single responsibility.
There are no chatty calls between services. If splitting functionality into two services causes them to be overly
chatty, it may be a symptom that these functions belong in the same service.
Each service is small enough that it can be built by a small team working independently.
There are no inter-dependencies that will require two or more services to be deployed in lock-step. It should
always be possible to deploy a service without redeploying any other services.
Services are not tightly coupled, and can evolve independently.
Your service boundaries will not create problems with data consistency or integrity. Sometimes it's important to
maintain data consistency by putting functionality into a single microservice. That said, consider whether you
really need strong consistency. There are strategies for addressing eventual consistency in a distributed system,
and the benefits of decomposing services often outweigh the challenges of managing eventual consistency.
Above all, it's important to be pragmatic, and remember that domain-driven design is an iterative process. When
in doubt, start with more coarse-grained microservices. Splitting a microservice into two smaller services is easier
than refactoring functionality across several existing microservices.

Example: Defining microservices for the Drone Delivery application

Recall that the development team had identified the four aggregates — Delivery, Package, Drone, and Account —
and two domain services, Scheduler and Supervisor.
Delivery and Package are obvious candidates for microservices. The Scheduler and Supervisor coordinate the
activities performed by other microservices, so it makes sense to implement these domain services as
microservices.
Drone and Account are interesting because they belong to other bounded contexts. One option is for the
Scheduler to call the Drone and Account bounded contexts directly. Another option is to create Drone and Account
microservices inside the Shipping bounded context. These microservices would mediate between the bounded
contexts, by exposing APIs or data schemas that are more suited to the Shipping context.
The details of the Drone and Account bounded contexts are beyond the scope of this guidance, so we created
mock services for them in our reference implementation. But here are some factors to consider in this situation:
What is the network overhead of calling directly into the other bounded context?
Is the data schema for the other bounded context suitable for this context, or is it better to have a schema
that's tailored to this bounded context?
Is the other bounded context a legacy system? If so, you might create a service that acts as an anti-
corruption layer to translate between the legacy system and the modern application.
What is the team structure? Is it easy to communicate with the team that's responsible for the other
bounded context? If not, creating a service that mediates between the two contexts can help to mitigate the
cost of cross-team communication.
So far, we haven't considered any non-functional requirements. Thinking about the application's throughput
requirements, the development team decided to create a separate Ingestion microservice that is responsible for
ingesting client requests. This microservice will implement load leveling by putting incoming requests into a buffer
for processing. The Scheduler will read the requests from the buffer and execute the workflow.
Non-functional requirements led the team to create one additional service. All of the services so far have been
about the process of scheduling and delivering packages in real time. But the system also needs to store the
history of every delivery in long-term storage for data analysis. The team considered making this the
responsibility of the Delivery service. However, the data storage requirements are quite different for historical
analysis versus in-flight operations (see Data considerations). Therefore, the team decided to create a separate
Delivery History service, which will listen for DeliveryTracking events from the Delivery service and write the
events into long-term storage.
The following diagram shows the design at this point:

Next steps
At this point, you should have a clear understanding of the purpose and functionality of each microservice in your
design. Now you can architect the system.
Design a microservices architecture
 Designing a microservices architecture
12/18/2020 • 2 minutes to read • Edit Online

Microservices have become a popular architectural style for building cloud applications that are resilient, highly
scalable, independently deployable, and able to evolve quickly. To be more than just a buzzword, however,
microservices require a different approach to designing and building applications.
In this set of articles, we explore how to build a microservices architecture on Azure. Topics include:
Compute options for microservices
Interservice communication
API design
API gateways
Data considerations
Design patterns

Prerequisites
Before reading these articles, you might start with the following:
Introduction to microservices architectures. Understand the benefits and challenges of microservices, and
when to use this style of architecture.
Using domain analysis to model microservices. Learn a domain-driven approach to modeling microservices.

Reference implementation
To illustrate best practices for a microservices architecture, we created a reference implementation that we call the
Drone Delivery application. This implementation runs on Kubernetes using Azure Kubernetes Service (AKS). You
can find the reference implementation on GitHub.

Scenario
Fabrikam, Inc. is starting a drone delivery service. The company manages a fleet of drone aircraft. Businesses
register with the service, and users can request a drone to pick up goods for delivery. When a customer schedules
a pickup, a backend system assigns a drone and notifies the user with an estimated delivery time. While the
delivery is in progress, the customer can track the location of the drone, with a continuously updated ETA.
This scenario involves a fairly complicated domain. Some of the business concerns include scheduling drones,
tracking packages, managing user accounts, and storing and analyzing historical data. Moreover, Fabrikam wants
to get to market quickly and then iterate quickly, adding new functionality and capabilities. The application needs
to operate at cloud scale, with a high service level objective (SLO). Fabrikam also expects that different parts of the
system will have very different requirements for data storage and querying. All of these considerations lead
Fabrikam to choose a microservices architecture for the Drone Delivery application.

NOTE
For help in choosing between a microservices architecture and other architectural styles, see the Azure Application
Architecture Guide.

Our reference implementation uses Kubernetes with Azure Kubernetes Service (AKS). However, many of the high-
level architectural decisions and challenges will apply to any container orchestrator, including Azure Service
Fabric.

Next steps
Choose a compute option
 Choosing an Azure compute option for microservices
12/18/2020 • 4 minutes to read • Edit Online

The term compute refers to the hosting model for the computing resources that your application runs on. For a
microservices architecture, two approaches are especially popular:
A service orchestrator that manages services running on dedicated nodes (VMs).
A serverless architecture using functions as a service (FaaS).
While these aren't the only options, they are both proven approaches to building microservices. An application
might include both approaches.

Service orchestrators
An orchestrator handles tasks related to deploying and managing a set of services. These tasks include placing
services on nodes, monitoring the health of services, restarting unhealthy services, load balancing network traffic
across service instances, service discovery, scaling the number of instances of a service, and applying
configuration updates. Popular orchestrators include Kubernetes, Service Fabric, DC/OS, and Docker Swarm.
On the Azure platform, consider the following options:
Azure Kubernetes Service (AKS) is a managed Kubernetes service. AKS provisions Kubernetes and exposes
the Kubernetes API endpoints, but hosts and manages the Kubernetes control plane, performing automated
upgrades, automated patching, autoscaling, and other management tasks. You can think of AKS as being
"Kubernetes APIs as a service."
Service Fabric is a distributed systems platform for packaging, deploying, and managing microservices.
Microservices can be deployed to Service Fabric as containers, as binary executables, or as Reliable Services.
Using the Reliable Services programming model, services can directly use Service Fabric programming APIs
to query the system, report health, receive notifications about configuration and code changes, and discover
other services. A key differentiation with Service Fabric is its strong focus on building stateful services using
Reliable Collections.
Other options such as Docker Enterprise Edition and Mesosphere DC/OS can run in an IaaS environment on
Azure. You can find deployment templates on Azure Marketplace.

Containers
Sometimes people talk about containers and microservices as if they were the same thing. While that's not true —
you don't need containers to build microservices — containers do have some benefits that are particularly relevant
to microservices, such as:
Por tability . A container image is a standalone package that runs without needing to install libraries or
other dependencies. That makes them easy to deploy. Containers can be started and stopped quickly, so you
can spin up new instances to handle more load or to recover from node failures.
Density . Containers are lightweight compared with running a virtual machine, because they share OS
resources. That makes it possible to pack multiple containers onto a single node, which is especially useful
when the application consists of many small services.
Resource isolation . You can limit the amount of memory and CPU that is available to a container, which
can help to ensure that a runaway process doesn't exhaust the host resources. See the Bulkhead pattern for
more information.
Serverless (Functions as a Service)
With a serverless architecture, you don't manage the VMs or the virtual network infrastructure. Instead, you deploy
code and the hosting service handles putting that code onto a VM and executing it. This approach tends to favor
small granular functions that are coordinated using event-based triggers. For example, a message being placed
onto a queue might trigger a function that reads from the queue and processes the message.
Azure Functions is a serverless compute service that supports various function triggers, including HTTP requests,
Service Bus queues, and Event Hubs events. For a complete list, see Azure Functions triggers and bindings
concepts. Also consider Azure Event Grid, which is a managed event routing service in Azure.

Orchestrator or serverless?
Here are some factors to consider when choosing between an orchestrator approach and a serverless approach.
Manageability A serverless application is easy to manage, because the platform manages all the of compute
resources for you. While an orchestrator abstracts some aspects of managing and configuring a cluster, it does not
completely hide the underlying VMs. With an orchestrator, you will need to think about issues such as load
balancing, CPU and memory usage, and networking.
Flexibility and control . An orchestrator gives you a great deal of control over configuring and managing your
services and the cluster. The tradeoff is additional complexity. With a serverless architecture, you give up some
degree of control because these details are abstracted.
Por tability . All of the orchestrators listed here (Kubernetes, DC/OS, Docker Swarm, and Service Fabric) can run
on-premises or in multiple public clouds.
Application integration . It can be challenging to build a complex application using a serverless architecture, due
to the need to coordinate, deploy, and manage many small independent functions. One option in Azure is to use
Azure Logic Apps to coordinate a set of Azure Functions. For an example of this approach, see Create a function
that integrates with Azure Logic Apps.
Cost . With an orchestrator, you pay for the VMs that are running in the cluster. With a serverless application, you
pay only for the actual compute resources consumed. In both cases, you need to factor in the cost of any additional
services, such as storage, databases, and messaging services.
Scalability . Azure Functions scales automatically to meet demand, based on the number of incoming events. With
an orchestrator, you can scale out by increasing the number of service instances running in the cluster. You can also
scale by adding additional VMs to the cluster.
Our reference implementation primarily uses Kubernetes, but we did use Azure Functions for one service, namely
the Delivery History service. Azure Functions was a good fit for this particular service, because it's is an event-
driven workload. By using an Event Hubs trigger to invoke the function, the service needed a minimal amount of
code. Also, the Delivery History service is not part of the main workflow, so running it outside of the Kubernetes
cluster doesn't affect the end-to-end latency of user-initiated operations.

Next steps
Interservice communication
 Designing interservice communication for
microservices
12/18/2020 • 12 minutes to read • Edit Online

Communication between microservices must be efficient and robust. With lots of small services interacting to
complete a single transaction, this can be a challenge. In this article, we look at the tradeoffs between
asynchronous messaging versus synchronous APIs. Then we look at some of the challenges in designing resilient
interservice communication.

Challenges
Here are some of the main challenges arising from service-to-service communication. Service meshes, described
later in this article, are designed to handle many of these challenges.
Resiliency . There may be dozens or even hundreds of instances of any given microservice. An instance can fail for
any number of reasons. There can be a node-level failure, such as a hardware failure or a VM reboot. An instance
might crash, or be overwhelmed with requests and unable to process any new requests. Any of these events can
cause a network call to fail. There are two design patterns that can help make service-to-service network calls
more resilient:
Retr y . A network call may fail because of a transient fault that goes away by itself. Rather than fail outright,
the caller should typically retry the operation a certain number of times, or until a configured time-out
period elapses. However, if an operation is not idempotent, retries can cause unintended side effects. The
original call might succeed, but the caller never gets a response. If the caller retries, the operation may be
invoked twice. Generally, it's not safe to retry POST or PATCH methods, because these are not guaranteed to
be idempotent.
Circuit Breaker . Too many failed requests can cause a bottleneck, as pending requests accumulate in the
queue. These blocked requests might hold critical system resources such as memory, threads, database
connections, and so on, which can cause cascading failures. The Circuit Breaker pattern can prevent a
service from repeatedly trying an operation that is likely to fail.
Load balancing . When service "A" calls service "B", the request must reach a running instance of service "B". In
Kubernetes, the Service resource type provides a stable IP address for a group of pods. Network traffic to the
service's IP address gets forwarded to a pod by means of iptable rules. By default, a random pod is chosen. A
service mesh (see below) can provide more intelligent load balancing algorithms based on observed latency or
other metrics.
Distributed tracing . A single transaction may span multiple services. That can make it hard to monitor the
overall performance and health of the system. Even if every service generates logs and metrics, without some way
to tie them together, they are of limited use. The article Logging and monitoring talks more about distributed
tracing, but we mention it here as a challenge.
Ser vice versioning . When a team deploys a new version of a service, they must avoid breaking any other
services or external clients that depend on it. In addition, you might want to run multiple versions of a service
side-by-side, and route requests to a particular version. See API Versioning for more discussion of this issue.
TLS encr yption and mutual TLS authentication . For security reasons, you may want to encrypt traffic
between services with TLS, and use mutual TLS authentication to authenticate callers.
Synchronous versus asynchronous messaging
There are two basic messaging patterns that microservices can use to communicate with other microservices.
1. Synchronous communication. In this pattern, a service calls an API that another service exposes, using a
protocol such as HTTP or gRPC. This option is a synchronous messaging pattern because the caller waits for
a response from the receiver.
2. Asynchronous message passing. In this pattern, a service sends message without waiting for a response,
and one or more services process the message asynchronously.
It's important to distinguish between asynchronous I/O and an asynchronous protocol. Asynchronous I/O means
the calling thread is not blocked while the I/O completes. That's important for performance, but is an
implementation detail in terms of the architecture. An asynchronous protocol means the sender doesn't wait for a
response. HTTP is a synchronous protocol, even though an HTTP client may use asynchronous I/O when it sends a
request.
There are tradeoffs to each pattern. Request/response is a well-understood paradigm, so designing an API may
feel more natural than designing a messaging system. However, asynchronous messaging has some advantages
that can be useful in a microservices architecture:
Reduced coupling . The message sender does not need to know about the consumer.
Multiple subscribers . Using a pub/sub model, multiple consumers can subscribe to receive events. See
Event-driven architecture style.
Failure isolation . If the consumer fails, the sender can still send messages. The messages will be picked up
when the consumer recovers. This ability is especially useful in a microservices architecture, because each
service has its own lifecycle. A service could become unavailable or be replaced with a newer version at any
given time. Asynchronous messaging can handle intermittent downtime. Synchronous APIs, on the other
hand, require the downstream service to be available or the operation fails.
Responsiveness . An upstream service can reply faster if it does not wait on downstream services. This is
especially useful in a microservices architecture. If there is a chain of service dependencies (service A calls
B, which calls C, and so on), waiting on synchronous calls can add unacceptable amounts of latency.
Load leveling . A queue can act as a buffer to level the workload, so that receivers can process messages at
their own rate.
Workflows . Queues can be used to manage a workflow, by check-pointing the message after each step in
the workflow.
However, there are also some challenges to using asynchronous messaging effectively.
Coupling with the messaging infrastructure . Using a particular messaging infrastructure may cause
tight coupling with that infrastructure. It will be difficult to switch to another messaging infrastructure later.
Latency . End-to-end latency for an operation may become high if the message queues fill up.
Cost . At high throughputs, the monetary cost of the messaging infrastructure could be significant.
Complexity . Handling asynchronous messaging is not a trivial task. For example, you must handle
duplicated messages, either by de-duplicating or by making operations idempotent. It's also hard to
implement request-response semantics using asynchronous messaging. To send a response, you need
another queue, plus a way to correlate request and response messages.
Throughput . If messages require queue semantics, the queue can become a bottleneck in the system. Each
message requires at least one queue operation and one dequeue operation. Moreover, queue semantics
generally require some kind of locking inside the messaging infrastructure. If the queue is a managed
 service, there may be additional latency, because the queue is external to the cluster's virtual network. You
can mitigate these issues by batching messages, but that complicates the code. If the messages don't
require queue semantics, you might be able to use an event stream instead of a queue. For more
information, see Event-driven architectural style.

Drone Delivery: Choosing the messaging patterns

With these considerations in mind, the development team made the following design choices for the Drone
Delivery application
The Ingestion service exposes a public REST API that client applications use to schedule, update, or cancel
deliveries.
The Ingestion service uses Event Hubs to send asynchronous messages to the Scheduler service.
Asynchronous messages are necessary to implement the load-leveling that is required for ingestion.
The Account, Delivery, Package, Drone, and Third-party Transport services all expose internal REST APIs. The
Scheduler service calls these APIs to carry out a user request. One reason to use synchronous APIs is that
the Scheduler needs to get a response from each of the downstream services. A failure in any of the
downstream services means the entire operation failed. However, a potential issue is the amount of latency
that is introduced by calling the backend services.
If any downstream service has a nontransient failure, the entire transaction should be marked as failed. To
handle this case, the Scheduler service sends an asynchronous message to the Supervisor, so that the
Supervisor can schedule compensating transactions.
The Delivery service exposes a public API that clients can use to get the status of a delivery. In the article API
gateway, we discuss how an API gateway can hide the underlying services from the client, so the client
doesn't need to know which services expose which APIs.
While a drone is in flight, the Drone service sends events that contain the drone's current location and
status. The Delivery service listens to these events in order to track the status of a delivery.
When the status of a delivery changes, the Delivery service sends a delivery status event, such as
DeliveryCreated or DeliveryCompleted . Any service can subscribe to these events. In the current design, the
Delivery service is the only subscriber, but there might be other subscribers later. For example, the events
might go to a real-time analytics service. And because the Scheduler doesn't have to wait for a response,
adding more subscribers doesn't affect the main workflow path.
Notice that delivery status events are derived from drone location events. For example, when a drone reaches a
delivery location and drops off a package, the Delivery service translates this into a DeliveryCompleted event. This
is an example of thinking in terms of domain models. As described earlier, Drone Management belongs in a
separate bounded context. The drone events convey the physical location of a drone. The delivery events, on the
other hand, represent changes in the status of a delivery, which is a different business entity.

Using a service mesh

A service mesh is a software layer that handles service-to-service communication. Service meshes are designed to
address many of the concerns listed in the previous section, and to move responsibility for these concerns away
from the microservices themselves and into a shared layer. The service mesh acts as a proxy that intercepts
network communication between microservices in the cluster. Currently, the service mesh concept applies mainly
to container orchestrators, rather than serverless architectures.

NOTE
Service mesh is an example of the Ambassador pattern — a helper service that sends network requests on behalf of the
application.

Right now, the main options for a service mesh in Kubernetes are linkerd and Istio. Both of these technologies are
evolving rapidly. However, some features that both linkerd and Istio have in common include:
Load balancing at the session level, based on observed latencies or number of outstanding requests. This
can improve performance over the layer-4 load balancing that is provided by Kubernetes.
Layer-7 routing based on URL path, Host header, API version, or other application-level rules.
Retry of failed requests. A service mesh understands HTTP error codes, and can automatically retry failed
requests. You can configure that maximum number of retries, along with a timeout period in order to
bound the maximum latency.
Circuit breaking. If an instance consistently fails requests, the service mesh will temporarily mark it as
unavailable. After a backoff period, it will try the instance again. You can configure the circuit breaker based
on various criteria, such as the number of consecutive failures,
Service mesh captures metrics about interservice calls, such as the request volume, latency, error and
 success rates, and response sizes. The service mesh also enables distributed tracing by adding correlation
information for each hop in a request.
Mutual TLS Authentication for service-to-service calls.
Do you need a service mesh? It depends. Without a service mesh, you'll need to consider each of the challenges
mentioned at the beginning of this article. You can solve problems like retry, circuit breaker, and distributed tracing
without a service mesh, but a service mesh moves these concerns out of the individual services and into a
dedicated layer. On the other hand, a service mesh adds complexity to the setup and configuration of the cluster.
There may be performance implications, because requests now get routed through the service mesh proxy, and
because extra services are now running on every node in the cluster. You should do thorough performance and
load testing before deploying a service mesh in production.

Distributed transactions
A common challenge in microservices is correctly handling transactions that span multiple services. Often in this
scenario, the success of a transaction is all or nothing — if one of the participating services fails, the entire
transaction must fail.
There are two cases to consider:
A service may experience a transient failure such as a network timeout. These errors can often be resolved
simply by retrying the call. If the operation still fails after a certain number of attempts, it's considered a
nontransient failure.
A nontransient failure is any failure that's unlikely to go away by itself. Nontransient failures include normal
error conditions, such as invalid input. They also include unhandled exceptions in application code or a
process crashing. If this type of error occurs, the entire business transaction must be marked as a failure. It
may be necessary to undo other steps in the same transaction that already succeeded.
After a nontransient failure, the current transaction might be in a partially failed state, where one or more steps
already completed successfully. For example, if the Drone service already scheduled a drone, the drone must be
canceled. In that case, the application needs to undo the steps that succeeded, by using a Compensating
Transaction. In some cases, this must be done by an external system or even by a manual process.
If the logic for compensating transactions is complex, consider creating a separate service that is responsible for
this process. In the Drone Delivery application, the Scheduler service puts failed operations onto a dedicated
queue. A separate microservice, called the Supervisor, reads from this queue and calls a cancellation API on the
services that need to compensate. This is a variation of the Scheduler Agent Supervisor pattern. The Supervisor
service might take other actions as well, such as notify the user by text or email, or send an alert to an operations
dashboard.
The Scheduler service itself might fault (for example, because a node crashes). In that case, a new instance can spin
up and take over. However, any transactions that were already in progress must be resumed.
One approach is to save a checkpoint to a durable store after each step in the workflow is completed. If an instance
of the Scheduler service crashes in the middle of a transaction, a new instance can use the checkpoint to resume
where the previous instance left off. However, writing checkpoints can create a performance overhead.
Another option is to design all operations to be idempotent. An operation is idempotent if it can be called multiple
times without producing additional side-effects after the first call. Essentially, the downstream service should
ignore duplicate calls, which means the service must be able to detect duplicate calls. It's not always
straightforward to implement idempotent methods. For more information, see Idempotent operations.

Next steps
For microservices that talk directly to each other, it's important to create well-designed APIs.
API design
 Designing APIs for microservices
12/18/2020 • 12 minutes to read • Edit Online

Good API design is important in a microservices architecture, because all data exchange between services happens
either through messages or API calls. APIs must be efficient to avoid creating chatty I/O. Because services are
designed by teams working independently, APIs must have well-defined semantics and versioning schemes, so
that updates don't break other services.

It's important to distinguish between two types of API:

Public APIs that client applications call.
Backend APIs that are used for interservice communication.
These two use cases have somewhat different requirements. A public API must be compatible with client
applications, typically browser applications or native mobile applications. Most of the time, that means the public
API will use REST over HTTP. For the backend APIs, however, you need to take network performance into account.
Depending on the granularity of your services, interservice communication can result in a lot of network traffic.
Services can quickly become I/O bound. For that reason, considerations such as serialization speed and payload
size become more important. Some popular alternatives to using REST over HTTP include gRPC, Apache Avro, and
Apache Thrift. These protocols support binary serialization and are generally more efficient than HTTP.

Considerations
Here are some things to think about when choosing how to implement an API.
REST versus RPC . Consider the tradeoffs between using a REST-style interface versus an RPC-style interface.
REST models resources, which can be a natural way express your domain model. It defines a uniform
interface based on HTTP verbs, which encourages evolvability. It has well-defined semantics in terms of
idempotency, side effects, and response codes. And it enforces stateless communication, which improves
scalability.
RPC is more oriented around operations or commands. Because RPC interfaces look like local method calls,
it may lead you to design overly chatty APIs. However, that doesn't mean RPC must be chatty. It just means
you need to use care when designing the interface.
For a RESTful interface, the most common choice is REST over HTTP using JSON. For an RPC-style interface, there
are several popular frameworks, including gRPC, Apache Avro, and Apache Thrift.
Efficiency . Consider efficiency in terms of speed, memory, and payload size. Typically a gRPC-based interface is
faster than REST over HTTP.
Interface definition language (IDL) . An IDL is used to define the methods, parameters, and return values of an
API. An IDL can be used to generate client code, serialization code, and API documentation. IDLs can also be
consumed by API testing tools such as Postman. Frameworks such as gRPC, Avro, and Thrift define their own IDL
specifications. REST over HTTP does not have a standard IDL format, but a common choice is OpenAPI (formerly
Swagger). You can also create an HTTP REST API without using a formal definition language, but then you lose the
benefits of code generation and testing.
Serialization . How are objects serialized over the wire? Options include text-based formats (primarily JSON) and
binary formats such as protocol buffer. Binary formats are generally faster than text-based formats. However,
JSON has advantages in terms of interoperability, because most languages and frameworks support JSON
serialization. Some serialization formats require a fixed schema, and some require compiling a schema definition
file. In that case, you'll need to incorporate this step into your build process.
Framework and language suppor t . HTTP is supported in nearly every framework and language. gRPC, Avro,
and Thrift all have libraries for C++, C#, Java, and Python. Thrift and gRPC also support Go.
Compatibility and interoperability . If you choose a protocol like gRPC, you may need a protocol translation
layer between the public API and the back end. A gateway can perform that function. If you are using a service
mesh, consider which protocols are compatible with the service mesh. For example, linkerd has built-in support for
HTTP, Thrift, and gRPC.
Our baseline recommendation is to choose REST over HTTP unless you need the performance benefits of a binary
protocol. REST over HTTP requires no special libraries. It creates minimal coupling, because callers don't need a
client stub to communicate with the service. There is rich ecosystems of tools to support schema definitions,
testing, and monitoring of RESTful HTTP endpoints. Finally, HTTP is compatible with browser clients, so you don't
need a protocol translation layer between the client and the backend.
However, if you choose REST over HTTP, you should do performance and load testing early in the development
process, to validate whether it performs well enough for your scenario.

RESTful API design

There are many resources for designing RESTful APIs. Here are some that you might find helpful:
API design
API implementation
Microsoft REST API Guidelines
Here are some specific considerations to keep in mind.
Watch out for APIs that leak internal implementation details or simply mirror an internal database schema.
The API should model the domain. It's a contract between services, and ideally should only change when
new functionality is added, not just because you refactored some code or normalized a database table.
Different types of client, such as mobile application and desktop web browser, may require different
payload sizes or interaction patterns. Consider using the Backends for Frontends pattern to create separate
backends for each client, that expose an optimal interface for that client.
For operations with side effects, consider making them idempotent and implementing them as PUT
methods. That will enable safe retries and can improve resiliency. The article Interservice communication
 discuss this issue in more detail.
HTTP methods can have asynchronous semantics, where the method returns a response immediately, but
the service carries out the operation asynchronously. In that case, the method should return an HTTP 202
response code, which indicates the request was accepted for processing, but the processing is not yet
completed. For more information, see Asynchronous Request-Reply pattern.

Mapping REST to DDD patterns

Patterns such as entity, aggregate, and value object are designed to place certain constraints on the objects in your
domain model. In many discussions of DDD, the patterns are modeled using object-oriented (OO) language
concepts like constructors or property getters and setters. For example, value objects are supposed to be
immutable. In an OO programming language, you would enforce this by assigning the values in the constructor
and making the properties read-only:

export class Location {

readonly latitude: number;
readonly longitude: number;

constructor(latitude: number, longitude: number) {

if (latitude < -90 || latitude > 90) {
throw new RangeError('latitude must be between -90 and 90');
}
if (longitude < -180 || longitude > 180) {
throw new RangeError('longitude must be between -180 and 180');
}
this.latitude = latitude;
this.longitude = longitude;
}
}

These sorts of coding practices are particularly important when building a traditional monolithic application. With
a large code base, many subsystems might use the Location object, so it's important for the object to enforce
correct behavior.
Another example is the Repository pattern, which ensures that other parts of the application do not make direct
reads or writes to the data store:

In a microservices architecture, however, services don't share the same code base and don't share data stores.
Instead, they communicate through APIs. Consider the case where the Scheduler service requests information
about a drone from the Drone service. The Drone service has its internal model of a drone, expressed through
code. But the Scheduler doesn't see that. Instead, it gets back a representation of the drone entity — perhaps a
JSON object in an HTTP response.
The Scheduler service can't modify the Drone service's internal models, or write to the Drone service's data store.
That means the code that implements the Drone service has a smaller exposed surface area, compared with code
in a traditional monolith. If the Drone service defines a Location class, the scope of that class is limited — no other
service will directly consume the class.
For these reasons, this guidance doesn't focus much on coding practices as they relate to the tactical DDD patterns.
But it turns out that you can also model many of the DDD patterns through REST APIs.
For example:
Aggregates map naturally to resources in REST. For example, the Delivery aggregate would be exposed as a
resource by the Delivery API.
Aggregates are consistency boundaries. Operations on aggregates should never leave an aggregate in an
inconsistent state. Therefore, you should avoid creating APIs that allow a client to manipulate the internal
state of an aggregate. Instead, favor coarse-grained APIs that expose aggregates as resources.
Entities have unique identities. In REST, resources have unique identifiers in the form of URLs. Create
resource URLs that correspond to an entity's domain identity. The mapping from URL to domain identity
may be opaque to client.
Child entities of an aggregate can be reached by navigating from the root entity. If you follow HATEOAS
principles, child entities can be reached via links in the representation of the parent entity.
Because value objects are immutable, updates are performed by replacing the entire value object. In REST,
implement updates through PUT or PATCH requests.
A repository lets clients query, add, or remove objects in a collection, abstracting the details of the
underlying data store. In REST, a collection can be a distinct resource, with methods for querying the
collection or adding new entities to the collection.
When you design your APIs, think about how they express the domain model, not just the data inside the model,
but also the business operations and the constraints on the data.

DDD C O N C EP T REST EQ UIVA L EN T EXA M P L E

Aggregate Resource { "1":1234, "status":"pending"...

}

Identity URL https://delivery-

service/deliveries/1

Child entities Links { "href":

"/deliveries/1/confirmation" }

Update value objects PUT or PATCH PUT https://delivery-

service/deliveries/1/dropoff
 DDD C O N C EP T REST EQ UIVA L EN T EXA M P L E

Repository Collection https://delivery-

service/deliveries?status=pending

API versioning
An API is a contract between a service and clients or consumers of that service. If an API changes, there is a risk of
breaking clients that depend on the API, whether those are external clients or other microservices. Therefore, it's a
good idea to minimize the number of API changes that you make. Often, changes in the underlying
implementation don't require any changes to the API. Realistically, however, at some point you will want to add
new features or new capabilities that require changing an existing API.
Whenever possible, make API changes backward compatible. For example, avoid removing a field from a model,
because that can break clients that expect the field to be there. Adding a field does not break compatibility, because
clients should ignore any fields they don't understand in a response. However, the service must handle the case
where an older client omits the new field in a request.
Support versioning in your API contract. If you introduce a breaking API change, introduce a new API version.
Continue to support the previous version, and let clients select which version to call. There are a couple of ways to
do this. One is simply to expose both versions in the same service. Another option is to run two versions of the
service side-by-side, and route requests to one or the other version, based on HTTP routing rules.

The diagram has two parts. "Service supports two versions" shows the v1 Client and the v2 Client both pointing to
one Service. "Side-by-side deployment" shows the v1 Client pointing to a v1 Service, and the v2 Client pointing to
a v2 Service.
There's a cost to supporting multiple versions, in terms of developer time, testing, and operational overhead.
Therefore, it's good to deprecate old versions as quickly as possible. For internal APIs, the team that owns the API
can work with other teams to help them migrate to the new version. This is when having a cross-team governance
process is useful. For external (public) APIs, it can be harder to deprecate an API version, especially if the API is
consumed by third parties or by native client applications.
When a service implementation changes, it's useful to tag the change with a version. The version provides
important information when troubleshooting errors. It can be very helpful for root cause analysis to know exactly
which version of the service was called. Consider using semantic versioning for service versions. Semantic
versioning uses a MAJOR.MINOR.PATCH format. However, clients should only select an API by the major version
number, or possibly the minor version if there are significant (but non-breaking) changes between minor versions.
In other words, it's reasonable for clients to select between version 1 and version 2 of an API, but not to select
version 2.1.3. If you allow that level of granularity, you risk having to support a proliferation of versions.
For further discussion of API versioning, see Versioning a RESTful web API.

Idempotent operations
An operation is idempotent if it can be called multiple times without producing additional side-effects after the
first call. Idempotency can be a useful resiliency strategy, because it allows an upstream service to safely invoke an
operation multiple times. For a discussion of this point, see Distributed transactions.
The HTTP specification states that GET, PUT, and DELETE methods must be idempotent. POST methods are not
guaranteed to be idempotent. If a POST method creates a new resource, there is generally no guarantee that this
operation is idempotent. The specification defines idempotent this way:

A request method is considered "idempotent" if the intended effect on the server of multiple identical requests
with that method is the same as the effect for a single such request. (RFC 7231)

It's important to understand the difference between PUT and POST semantics when creating a new entity. In both
cases, the client sends a representation of an entity in the request body. But the meaning of the URI is different.
For a POST method, the URI represents a parent resource of the new entity, such as a collection. For
example, to create a new delivery, the URI might be /api/deliveries . The server creates the entity and
assigns it a new URI, such as /api/deliveries/39660 . This URI is returned in the Location header of the
response. Each time the client sends a request, the server will create a new entity with a new URI.
For a PUT method, the URI identifies the entity. If there already exists an entity with that URI, the server
replaces the existing entity with the version in the request. If no entity exists with that URI, the server creates
one. For example, suppose the client sends a PUT request to api/deliveries/39660 . Assuming there is no
delivery with that URI, the server creates a new one. Now if the client sends the same request again, the
server will replace the existing entity.
Here is the Delivery service's implementation of the PUT method.
 [HttpPut("{id}")]
[ProducesResponseType(typeof(Delivery), 201)]
[ProducesResponseType(typeof(void), 204)]
public async Task<IActionResult> Put([FromBody]Delivery delivery, string id)
{
logger.LogInformation("In Put action with delivery {Id}: {@DeliveryInfo}", id, delivery.ToLogInfo());
try
{
var internalDelivery = delivery.ToInternal();

// Create the new delivery entity.

await deliveryRepository.CreateAsync(internalDelivery);

// Create a delivery status event.

var deliveryStatusEvent = new DeliveryStatusEvent { DeliveryId = delivery.Id, Stage =
DeliveryEventType.Created };
await deliveryStatusEventRepository.AddAsync(deliveryStatusEvent);

// Return HTTP 201 (Created)

return CreatedAtRoute("GetDelivery", new { id= delivery.Id }, delivery);
}
catch (DuplicateResourceException)
{
// This method is mainly used to create deliveries. If the delivery already exists then update it.
logger.LogInformation("Updating resource with delivery id: {DeliveryId}", id);

var internalDelivery = delivery.ToInternal();

await deliveryRepository.UpdateAsync(id, internalDelivery);

// Return HTTP 204 (No Content)

return NoContent();
}
}

It's expected that most requests will create a new entity, so the method optimistically calls CreateAsync on the
repository object, and then handles any duplicate-resource exceptions by updating the resource instead.

Next steps
Learn about using an API gateway at the boundary between client applications and microservices.
API gateways
 Using API gateways in microservices
12/18/2020 • 5 minutes to read • Edit Online

In a microservices architecture, a client might interact with more than one front-end service. Given this fact, how
does a client know what endpoints to call? What happens when new services are introduced, or existing services
are refactored? How do services handle SSL termination, authentication, and other concerns? An API gateway can
help to address these challenges.

What is an API gateway?

An API gateway sits between clients and services. It acts as a reverse proxy, routing requests from clients to
services. It may also perform various cross-cutting tasks such as authentication, SSL termination, and rate limiting.
If you don't deploy a gateway, clients must send requests directly to front-end services. However, there are some
potential problems with exposing services directly to clients:
It can result in complex client code. The client must keep track of multiple endpoints, and handle failures in a
resilient way.
It creates coupling between the client and the backend. The client needs to know how the individual services
are decomposed. That makes it harder to maintain the client and also harder to refactor services.
A single operation might require calls to multiple services. That can result in multiple network round trips
between the client and the server, adding significant latency.
Each public-facing service must handle concerns such as authentication, SSL, and client rate limiting.
Services must expose a client-friendly protocol such as HTTP or WebSocket. This limits the choice of
communication protocols.
Services with public endpoints are a potential attack surface, and must be hardened.
A gateway helps to address these issues by decoupling clients from services. Gateways can perform a number of
different functions, and you may not need all of them. The functions can be grouped into the following design
patterns:
Gateway Routing. Use the gateway as a reverse proxy to route requests to one or more backend services, using
layer 7 routing. The gateway provides a single endpoint for clients, and helps to decouple clients from services.
Gateway Aggregation. Use the gateway to aggregate multiple individual requests into a single request. This
pattern applies when a single operation requires calls to multiple backend services. The client sends one request
to the gateway. The gateway dispatches requests to the various backend services, and then aggregates the results
and sends them back to the client. This helps to reduce chattiness between the client and the backend.
Gateway Offloading. Use the gateway to offload functionality from individual services to the gateway, particularly
cross-cutting concerns. It can be useful to consolidate these functions into one place, rather than making every
service responsible for implementing them. This is particularly true for features that requires specialized skills to
implement correctly, such as authentication and authorization.
Here are some examples of functionality that could be offloaded to a gateway:
SSL termination
Authentication
IP allow/block list
Client rate limiting (throttling)
Logging and monitoring
Response caching
Web application firewall
GZIP compression
Servicing static content

Choosing a gateway technology

Here are some options for implementing an API gateway in your application.
Reverse proxy ser ver . Nginx and HAProxy are popular reverse proxy servers that support features such
as load balancing, SSL, and layer 7 routing. They are both free, open-source products, with paid editions
that provide additional features and support options. Nginx and HAProxy are both mature products with
rich feature sets and high performance. You can extend them with third-party modules or by writing
custom scripts in Lua. Nginx also supports a JavaScript-based scripting module referred to as 'NGINX
JavaScript'. This module was formally named nginScript.
Ser vice mesh ingress controller . If you are using a service mesh such as linkerd or Istio, consider the
features that are provided by the ingress controller for that service mesh. For example, the Istio ingress
controller supports layer 7 routing, HTTP redirects, retries, and other features.
Azure Application Gateway. Application Gateway is a managed load balancing service that can perform
layer-7 routing and SSL termination. It also provides a web application firewall (WAF).
Azure API Management. API Management is a turnkey solution for publishing APIs to external and internal
customers. It provides features that are useful for managing a public-facing API, including rate limiting, IP
restrictions, and authentication using Azure Active Directory or other identity providers. API Management
doesn't perform any load balancing, so it should be used in conjunction with a load balancer such as
Application Gateway or a reverse proxy. For information about using API Management with Application
Gateway, see Integrate API Management in an internal VNet with Application Gateway.
When choosing a gateway technology, consider the following:
Features . The options listed above all support layer 7 routing, but support for other features will vary. Depending
on the features that you need, you might deploy more than one gateway.
Deployment . Azure Application Gateway and API Management are managed services. Nginx and HAProxy will
typically run in containers inside the cluster, but can also be deployed to dedicated VMs outside of the cluster. This
isolates the gateway from the rest of the workload, but incurs higher management overhead.
Management . When services are updated or new services are added, the gateway routing rules may need to be
updated. Consider how this process will be managed. Similar considerations apply to managing SSL certificates, IP
allow lists, and other aspects of configuration.

Deploying Nginx or HAProxy to Kubernetes

You can deploy Nginx or HAProxy to Kubernetes as a ReplicaSet or DaemonSet that specifies the Nginx or
HAProxy container image. Use a ConfigMap to store the configuration file for the proxy, and mount the ConfigMap
as a volume. Create a service of type LoadBalancer to expose the gateway through an Azure Load Balancer.
An alternative is to create an Ingress Controller. An Ingress Controller is a Kubernetes resource that deploys a load
balancer or reverse proxy server. Several implementations exist, including Nginx and HAProxy. A separate resource
called an Ingress defines settings for the Ingress Controller, such as routing rules and TLS certificates. That way,
you don't need to manage complex configuration files that are specific to a particular proxy server technology.
The gateway is a potential bottleneck or single point of failure in the system, so always deploy at least two replicas
for high availability. You may need to scale out the replicas further, depending on the load.
Also consider running the gateway on a dedicated set of nodes in the cluster. Benefits to this approach include:
Isolation. All inbound traffic goes to a fixed set of nodes, which can be isolated from backend services.
Stable configuration. If the gateway is misconfigured, the entire application may become unavailable.
Performance. You may want to use a specific VM configuration for the gateway for performance reasons.

Next steps
The previous articles have looked at the interfaces between microservices or between microservices and client
applications. By design, these interfaces treat each service as a opaque box. In particular, microservices should
never expose implementation details about how they manage data. That has implications for data integrity and
data consistency, explored in the next article.
Data considerations for microservices
 Data considerations for microservices
12/18/2020 • 7 minutes to read • Edit Online

This article describes considerations for managing data in a microservices architecture. Because every
microservice manages its own data, data integrity and data consistency are critical challenges.
A basic principle of microservices is that each service manages its own data. Two services should not share a data
store. Instead, each service is responsible for its own private data store, which other services cannot access directly.
The reason for this rule is to avoid unintentional coupling between services, which can result if services share the
same underlying data schemas. If there is a change to the data schema, the change must be coordinated across
every service that relies on that database. By isolating each service's data store, we can limit the scope of change,
and preserve the agility of truly independent deployments. Another reason is that each microservice may have its
own data models, queries, or read/write patterns. Using a shared data store limits each team's ability to optimize
data storage for their particular service.

This approach naturally leads to polyglot persistence — the use of multiple data storage technologies within a
single application. One service might require the schema-on-read capabilities of a document database. Another
might need the referential integrity provided by an RDBMS. Each team is free to make the best choice for their
service. For more about the general principle of polyglot persistence, see Use the best data store for the job.

NOTE
It's fine for services to share the same physical database server. The problem occurs when services share the same schema,
or read and write to the same set of database tables.

Challenges
Some challenges arise from this distributed approach to managing data. First, there may be redundancy across the
data stores, with the same item of data appearing in multiple places. For example, data might be stored as part of a
transaction, then stored elsewhere for analytics, reporting, or archiving. Duplicated or partitioned data can lead to
issues of data integrity and consistency. When data relationships span multiple services, you can't use traditional
data management techniques to enforce the relationships.
Traditional data modeling uses the rule of "one fact in one place." Every entity appears exactly once in the schema.
Other entities may hold references to it but not duplicate it. The obvious advantage to the traditional approach is
that updates are made in a single place, which avoids problems with data consistency. In a microservices
architecture, you have to consider how updates are propagated across services, and how to manage eventual
consistency when data appears in multiple places without strong consistency.

Approaches to managing data

There is no single approach that's correct in all cases, but here are some general guidelines for managing data in a
microservices architecture.
Embrace eventual consistency where possible. Understand the places in the system where you need strong
consistency or ACID transactions, and the places where eventual consistency is acceptable.
When you need strong consistency guarantees, one service may represent the source of truth for a given
entity, which is exposed through an API. Other services might hold their own copy of the data, or a subset of
the data, that is eventually consistent with the master data but not considered the source of truth. For
example, imagine an e-commerce system with a customer order service and a recommendation service.
The recommendation service might listen to events from the order service, but if a customer requests a
refund, it is the order service, not the recommendation service, that has the complete transaction history.
For transactions, use patterns such as Scheduler Agent Supervisor and Compensating Transaction to keep
data consistent across several services. You may need to store an additional piece of data that captures the
state of a unit of work that spans multiple services, to avoid partial failure among multiple services. For
example, keep a work item on a durable queue while a multi-step transaction is in progress.
Store only the data that a service needs. A service might only need a subset of information about a domain
entity. For example, in the Shipping bounded context, we need to know which customer is associated to a
particular delivery. But we don't need the customer's billing address — that's managed by the Accounts
bounded context. Thinking carefully about the domain, and using a DDD approach, can help here.
Consider whether your services are coherent and loosely coupled. If two services are continually
exchanging information with each other, resulting in chatty APIs, you may need to redraw your service
boundaries, by merging two services or refactoring their functionality.
Use an event driven architecture style. In this architecture style, a service publishes an event when there are
changes to its public models or entities. Interested services can subscribe to these events. For example,
another service could use the events to construct a materialized view of the data that is more suitable for
querying.
A service that owns events should publish a schema that can be used to automate serializing and
deserializing the events, to avoid tight coupling between publishers and subscribers. Consider JSON
schema or a framework like Microsoft Bond, Protobuf, or Avro.
At high scale, events can become a bottleneck on the system, so consider using aggregation or batching to
reduce the total load.

Example: Choosing data stores for the Drone Delivery application

The previous articles in this series discuss a drone delivery service as a running example. You can read more about
the scenario and the corresponding reference implementation here.
To recap, this application defines several microservices for scheduling deliveries by drone. When a user schedules
a new delivery, the client request includes information about the delivery, such as pickup and dropoff locations,
and about the package, such as size and weight. This information defines a unit of work.
The various backend services care about different portions of the information in the request, and also have
different read and write profiles.
Delivery service
The Delivery service stores information about every delivery that is currently scheduled or in progress. It listens
for events from the drones, and tracks the status of deliveries that are in progress. It also sends domain events
with delivery status updates.
It's expected that users will frequently check the status of a delivery while they are waiting for their package.
Therefore, the Delivery service requires a data store that emphasizes throughput (read and write) over long-term
storage. Also, the Delivery service does not perform any complex queries or analysis, it simply fetches the latest
status for a given delivery. The Delivery service team chose Azure Cache for Redis for its high read-write
performance. The information stored in Redis is relatively short-lived. Once a delivery is complete, the Delivery
History service is the system of record.
Delivery History service
The Delivery History service listens for delivery status events from the Delivery service. It stores this data in long-
term storage. There are two different use-cases for this historical data, which have different data storage
requirements.
The first scenario is aggregating the data for the purpose of data analytics, in order to optimize the business or
improve the quality of the service. Note that the Delivery History service doesn't perform the actual analysis of the
data. It's only responsible for the ingestion and storage. For this scenario, the storage must be optimized for data
analysis over a large set of data, using a schema-on-read approach to accommodate a variety of data sources.
Azure Data Lake Store is a good fit for this scenario. Data Lake Store is an Apache Hadoop file system compatible
with Hadoop Distributed File System (HDFS), and is tuned for performance for data analytics scenarios.
The other scenario is enabling users to look up the history of a delivery after the delivery is completed. Azure Data
Lake is not optimized for this scenario. For optimal performance, Microsoft recommends storing time-series data
in Data Lake in folders partitioned by date. (See Tuning Azure Data Lake Store for performance). However, that
structure is not optimal for looking up individual records by ID. Unless you also know the timestamp, a lookup by
ID requires scanning the entire collection. Therefore, the Delivery History service also stores a subset of the
historical data in Cosmos DB for quicker lookup. The records don't need to stay in Cosmos DB indefinitely. Older
deliveries can be archived — say, after a month. This could be done by running an occasional batch process.
Package service
The Package service stores information about all of the packages. The storage requirements for the Package are:
Long-term storage.
 Able to handle a high volume of packages, requiring high write throughput.
Support simple queries by package ID. No complex joins or requirements for referential integrity.
Because the package data is not relational, a document-oriented database is appropriate, and Cosmos DB can
achieve high throughput by using sharded collections. The team that works on the Package service is familiar with
the MEAN stack (MongoDB, Express.js, AngularJS, and Node.js), so they select the MongoDB API for Cosmos DB.
That lets them leverage their existing experience with MongoDB, while getting the benefits of Cosmos DB, which is
a managed Azure service.

Next steps
Learn about design patterns that can help mitigate some common challenges in a microservices architecture.
Design patterns for microservices
 Design patterns for microservices
12/18/2020 • 2 minutes to read • Edit Online

The goal of microservices is to increase the velocity of application releases, by decomposing the application into
small autonomous services that can be deployed independently. A microservices architecture also brings some
challenges. The design patterns shown here can help mitigate these challenges.

Ambassador can be used to offload common client connectivity tasks such as monitoring, logging, routing, and
security (such as TLS) in a language agnostic way. Ambassador services are often deployed as a sidecar (see
below).
Anti-corruption layer implements a façade between new and legacy applications, to ensure that the design of a
new application is not limited by dependencies on legacy systems.
Backends for Frontends creates separate backend services for different types of clients, such as desktop and
mobile. That way, a single backend service doesn't need to handle the conflicting requirements of various client
types. This pattern can help keep each microservice simple, by separating client-specific concerns.
Bulkhead isolates critical resources, such as connection pool, memory, and CPU, for each workload or service. By
using bulkheads, a single workload (or service) can't consume all of the resources, starving others. This pattern
increases the resiliency of the system by preventing cascading failures caused by one service.
Gateway Aggregation aggregates requests to multiple individual microservices into a single request, reducing
chattiness between consumers and services.
Gateway Offloading enables each microservice to offload shared service functionality, such as the use of SSL
certificates, to an API gateway.
Gateway Routing routes requests to multiple microservices using a single endpoint, so that consumers don't
need to manage many separate endpoints.
Sidecar deploys helper components of an application as a separate container or process to provide isolation and
encapsulation.
Strangler Fig supports incremental refactoring of an application, by gradually replacing specific pieces of
functionality with new services.
For the complete catalog of cloud design patterns on the Azure Architecture Center, see Cloud Design Patterns.
 Monitoring a microservices architecture in Azure
Kubernetes Service (AKS)
12/18/2020 • 14 minutes to read • Edit Online

This article describes best practices for monitoring a microservices application that runs on Azure Kubernetes
Service (AKS).
In any complex application, at some point something will go wrong. In a microservices application, you need to
track what's happening across dozens or even hundreds of services. To make sense of what's happening, you must
collect telemetry from the application. Telemetry can be divided into logs and metrics.
Logs are text-based records of events that occur while the application is running. They include things like
application logs (trace statements) or web server logs. Logs are primarily useful for forensics and root cause
analysis.
Metrics are numerical values that can be analyzed. You can use them to observe the system in real time (or close
to real time), or to analyze performance trends over time. To understand the system holistically, you must collect
metrics at various levels of the architecture, from the physical infrastructure to the application, including:
Node-level metrics, including CPU, memory, network, disk, and file system usage. System metrics help you
to understand resource allocation for each node in the cluster, and troubleshoot outliers.
Container metrics. For containerized applications, you need to collect metrics at the container level, not just
at the VM level.
Application metrics. This includes any metrics that are relevant to understanding the behavior of a service.
Examples include the number of queued inbound HTTP requests, request latency, or message queue length.
Applications can also create custom metrics that are specific to the domain, such as the number of business
transactions processed per minute.
Dependent ser vice metrics. Services may call external services or endpoints, such as managed PaaS
services or SaaS services. Third-party services may or may not provide any metrics. If not, you'll have to rely
on your own application metrics to track statistics for latency and error rate.

Monitoring cluster status

Use Azure Monitor to monitor the overall health of your clusters. The following screenshot shows a cluster with
critical errors in user-deployed pods.

From here, you can drill in further to find the issue. For example, if the pod status is ImagePullBackoff , it means
that Kubernetes could not pull the container image from the registry. This could be caused by an invalid container
tag or an authentication error trying to pull from the registry.
Note that a container crashing will put the container state into State = Waiting ,with Reason = CrashLoopBackOff
. For a typical scenario where a pod is part of a replica set and the retry policy is Always , this won't show as an
error in the cluster status. However, you can run queries or set up alerts for this condition. For more information,
see Understand AKS cluster performance with Azure Monitor for containers.

Metrics
We recommend using Azure Monitor to collect and view metrics for your AKS clusters and any other dependent
Azure services.
For cluster and container metrics, enable Azure Monitor for containers. When this feature is enabled, Azure
Monitor collects memory and processor metrics from controllers, nodes, and containers via the Kubernetes
metrics API. For more information about the metrics that are available through Azure Monitor for
containers, see Understand AKS cluster performance with Azure Monitor for containers.
Use Application Insights to collect application metrics. Application Insights is an extensible Application
Performance Management (APM) service. To use it, you install an instrumentation package in your
application. This package monitors the app and sends telemetry data to the Application Insights service. It
can also pull telemetry data from the host environment. The data is then sent to Azure Monitor. Application
Insights also provides built-in correlation and dependency tracking (see Distributed tracing, below).
Application Insights has a maximum throughput measured in events/second, and it throttles if the data rate
exceeds the limit. For details, see Application Insights limits. Create different Application Insights instances per
environment, so that dev/test environments don't compete against the production telemetry for quota.
A single operation may generate several telemetry events, so if the application experiences a high volume of traffic,
it is likely to get throttled. To mitigate this problem, you can perform sampling to reduce the telemetry traffic. The
tradeoff is that your metrics will be less precise. For more information, see Sampling in Application Insights. You
can also reduce the data volume by pre-aggregating metrics — that is, calculating statistical values such as average
and standard deviation, and sending those values instead of the raw telemetry. The following blog post describes
an approach to using Application Insights at scale: Azure Monitoring and Analytics at Scale.
If your data rate is high enough to trigger throttling, and sampling or aggregation are not acceptable, consider
exporting metrics to a time-series database such as Prometheus or InfluxDB running in the cluster.
InfluxDB is a push-based system. An agent needs to push the metrics. You can use TICK stack, to setup
monitoring of Kubernetes, and push it to InfluxDB using Telegraf, which is an agent for collecting and
reporting metrics. InfluxDB can be used for irregular events and string data types.
Prometheus is a pull-based system. It periodically scrapes metrics from configured locations. Prometheus
can scrape metrics generated by cAdvisor or kube-state-metrics. kube-state-metrics is a service that collects
metrics from the Kubernetes API server and makes them available to Prometheus (or a scraper that is
compatible with a Prometheus client endpoint). For system metrics, use Node exporter, which is a
Prometheus exporter for system metrics. Prometheus supports floating point data, but not string data, so it
is appropriate for system metrics but not logs. Kubernetes Metrics Server is a cluster-wide aggregator of
resource usage data.

Logging
Here are some of the general challenges of logging in a microservices application:
Understanding the end-to-end processing of a client request, where multiple services might be invoked to
handle a single request.
Consolidating logs from multiple services into a single aggregated view.
Parsing logs that come from multiple sources, which use their own logging schemas or have no particular
 schema. Logs may be generated by third-party components that you don't control.
Microservices architectures often generate a larger volume of logs than traditional monoliths, because there are
more services, network calls, and steps in a transaction. That means logging itself can be a performance or
resource bottleneck for the application.
There are some additional challenges for a Kubernetes-based architecture:
Containers can move around and be rescheduled.
Kubernetes has a networking abstraction that uses virtual IP addresses and port mappings.
In Kubernetes, the standard approach to logging is for a container to write logs to stdout and stderr. The container
engine redirects these streams to a logging driver. For ease of querying, and to prevent possible loss of log data if
a node crashes, the usual approach is to collect the logs from each node and send them to a central storage
location.
Azure Monitor integrates with AKS to support this approach. Azure Monitor collects container logs and sends them
to a Log Analytics workspace. From there, you can use the Kusto query language to write queries across the
aggregated logs. For example, here is a Kusto query to show the container logs for a specified pod:

let ContainerIdList = KubePodInventory

| where ClusterName =~ '<cluster-name>'
| where Name =~ '<pod-name>'
| distinct ContainerID;
ContainerLog
| where ContainerID in (ContainerIdList)

Azure Monitor is a managed service, and configuring an AKS cluster to use Azure Monitor is a simple configuration
switch in the CLI or Resource Manager template. (For more information, see How to enable Azure Monitor for
containers.) Another advantage of using Azure Monitoring is that it consolidates your AKS logs with other Azure
platform logs, providing a unified monitoring experience.
Azure Monitor is billed per gigabyte (GB) of data ingested into the service (see Azure Monitor pricing). At very high
volumes, cost may become a consideration. There are many open-source alternatives available for the Kubernetes
ecosystem. For example, many organizations use Fluentd with Elasticsearch . Fluentd is an open-source data
collector, and Elasticsearch is a document database that is for search. A challenge with these options is that they
require additional configuration and management of the cluster. For a production workload, you may need to
experiment with configuration settings. You'll also need to monitor the performance of the logging infrastructure.
Application Insights
For richer log data, we recommend instrumenting your code with Application Insights. This requires adding an
Application Insights package to your code and configuring your code to send logging statements to Application
Insights. The details depend on the platform, such as .NET, Java, or Node.js. The Application Insights package sends
telemetry data to Azure Monitor.
If you are using .NET Core, we recommend also using the Application Insights for Kubernetes library. This library
enriches Application Insights traces with additional information such as the container, node, pod, labels, and replica
set.
Advantages of this approach include:
Application Insights logs HTTP requests, including latency and result code.
Distributed tracing is enabled by default.
Traces include an operation ID, so you can match all traces for a particular operation.
Traces generated by Application Insights often have additional contextual information. For example, ASP.NET
traces are decorated with the action name and a category such as ControllerActionInvoker , which give you
insights into the ASP.NET request pipeline.
 Application Insights collects performance metrics for performance troubleshooting and optimization.
Considerations:
Application Insights throttles the telemetry if the data rate exceeds a maximum limit; for details, see Application
Insights limits. A single operation may generate several telemetry events, so if the application experiences a
high volume of traffic, it is likely to get throttled.
Because Application Insights batches data, it's possible to lose a batch if a process crashes with an unhandled
exception.
Application Insights is billed based on data volume. For more information, see Manage pricing and data volume
in Application Insights.

Structured logging
To make logs easier to parse, use structured logging where possible. Structured logging is approach where the
application writes logs in a structured format, such as JSON, rather than outputting unstructured text strings. There
are many structured logging libraries available. For example, here is a logging statement that uses the Serilog
library for .NET Core:

public async Task<IActionResult> Put([FromBody]Delivery delivery, string id)

{
logger.LogInformation("In Put action with delivery {Id}: {@DeliveryInfo}", id, delivery.ToLogInfo());

...
}

Here, the call to LogInformation includes an Id parameter and DeliveryInfo parameter. With structured logging,
these values are not interpolated into the message string. Instead, the log output will look something like this:

{"@t":"2019-06-13T00:57:09.9932697Z","@mt":"In Put action with delivery {Id}: {@DeliveryInfo}","Id":"36585f2d-

c1fa-4a3d-9e06-a7f40b7d04ef","DeliveryInfo":{...

This is a JSON string, where the "@t" field is a timestamp, "@mt" is the message string, and the remaining
key/value pairs are the parameters. Outputting JSON format makes it easier to query the data in a structured way.
For example, the following Log Analytics query, written in the Kusto query language, searches for instances of this
particular message from all containers named fabrikam-delivery :

traces
| where customDimensions.["Kubernetes.Container.Name"] == "fabrikam-delivery"
| where customDimensions.["{OriginalFormat}"] == "In Put action with delivery {Id}: {@DeliveryInfo}"
| project message, customDimensions["Id"], customDimensions["@DeliveryInfo"]

Viewing the result in the Azure portal shows that DeliveryInfo is a structured record that contains the serialized
representation of the DeliveryInfo model:
Here is the JSON from this example:

{
"Id": "36585f2d-c1fa-4a3d-9e06-a7f40b7d04ef",
"Owner": {
"UserId": "user id for logging",
"AccountId": "52dadf0c-0067-43e7-af76-86e32b48bc5e"
},
"Pickup": {
"Altitude": 0.29295161612934972,
"Latitude": 0.26815900219052985,
"Longitude": 0.79841844309047727
},
"Dropoff": {
"Altitude": 0.31507750848078986,
"Latitude": 0.753494655598651,
"Longitude": 0.89352830773849423
},
"Deadline": "string",
"Expedited": true,
"ConfirmationRequired": 0,
"DroneId": "AssignedDroneId01ba4d0b-c01a-4369-ba75-51bde0e76cc9"
}

The previous code snippet used the Serilog library, but structured logging libraries are available for other
languages as well. For example, here's an example using the SLF4J library for Java:

MDC.put("DeliveryId", deliveryId);

log.info("In schedule delivery action with delivery request {}", externalDelivery.toString());

Distributed tracing
A significant challenge of microservices is to understand the flow of events across services. A single transaction
may involve calls to multiple services. To reconstruct the entire sequence of steps, each service should propagate a
correlation ID that acts as a unique identifier for that operation. The correlation ID enables distributed tracing
across services.
The first service that receives a client request should generate the correlation ID. If the service makes an HTTP call
to another service, it puts the correlation ID in a request header. If the service sends an asynchronous message, it
puts the correlation ID into the message. Downstream services continue to propagate the correlation ID, so that it
flows through the entire system. In addition, all code that writes application metrics or log events should include
the correlation ID.
When service calls are correlated, you can calculate operational metrics such as the end-to-end latency for a
complete transaction, the number of successful transactions per second, and the percentage of failed transactions.
Including correlation IDs in application logs makes it possible to perform root cause analysis. If an operation fails,
you can find the log statements for all of the service calls that were part of the same operation.
We recommend using Application Insights for distributed tracing. The Application Insights SDK automatically
injects correlation context into HTTP headers, and includes the correlation ID in Application Insights logs. Some
services may still need to explicitly propagate the correlation headers, depending on the frameworks and libraries
being used. For more information, see Telemetry correlation in Application Insights.
Some additional considerations when implementing distributed tracing:
There is now a standard HTTP header for correlation IDs, a W3C proposal has been accepted as an official
recommendation recently. Your team should standardize on a custom header value. The choice may be
decided by your logging framework, such as Application Insights, or choice of service mesh.
For asynchronous messages, if your messaging infrastructure supports adding metadata to messages, you
should include the correlation ID as metadata. Otherwise, include it as part of the message schema. For
example, see Distributed tracing and correlation through Service Bus messaging.
Rather than a single opaque identifier, you might send a correlation context that includes richer information,
such as caller-callee relationships.
If you are using Istio or linkerd as a service mesh, these technologies automatically generate correlation
headers when HTTP calls are routed through the service mesh proxies. Services should forward the relevant
headers.
Example of distributed tracing
This example follows a distributed transaction through a set of microservices. The example is taken from a
reference implementation described here.

In this scenario, the distributed transaction has the following steps:

1. The Ingestion service puts a message on a Service Bus queue.
2. The Workflow service pulls the message from the queue.
3. The Workflow service calls three backend services to process the request (Drone Scheduler, Package, and
Delivery).
The following screenshot shows the application map for the Drone Delivery application. This map shows calls to
the public API endpoint that result in a workflow that involves five microservices.
The arrows from fabrikam-workflow and fabrikam-ingestion to a Service Bus queue show where the messages
are sent and received. You can't tell from the diagram which service is sending messages and which is receiving —
the arrows just show that both services are calling Service Bus — but this information is available in the details:

Because every call includes an operation ID, you can also view the end-to-end steps in a single transaction,
including timing information and the HTTP calls at each step. Here is the visualization of one such transaction:
This visualization shows the steps from the ingestion service to the queue, from the queue to the workflow service,
and from the workflow service to the other backend services. The last step is the Workflow service marking the
Service Bus message as completed.
Now here is an example when calls to a backend service were failing:

This shows that a large fraction (36%) of calls to the Drone Scheduler service failed during the period being
queried. In the end-to-end transaction view, it shows that an exception occurs when sending an HTTP PUT request
to the service.
Further drilling in, the exception turns out to be a socket exception, "No such device or address."

Fabrikam.Workflow.Service.Services.BackendServiceCallFailedException: No such device or address ---u003e

System.Net.Http.HttpRequestException: No such device or address ---u003e
System.Net.Sockets.SocketException: No such device or address

This is a hint that the backend service is not reachable. At this point, you might use kubectl to view the deployment
configuration. In this example, it turned out the service hostname was not resolving, due to an error in the
Kubernetes configuration files. The article Debug Services in the Kubernetes documentation has tips for
diagnosing this sort of error.
Here are some common causes of errors:
Code bugs. These might manifest as:
Exceptions. Look in the Application Insights logs to view the exception details.
Process crashing. Look at container and pod status, and view container logs or Application Insights
traces.
HTTP 5xx errors
Resource exhaustion:
Look for throttling (HTTP 429) or request timeouts.
Examine container metrics for CPU, memory, and disk
Look at the configurations for container and pod resource limits.
Service discovery. Examine the Kubernetes service configuration and port mappings.
API mismatch. Look for HTTP 400 errors. If APIs are versioned, look at which version is being called.
Error pulling a container image. Look at the pod specification. Also make sure the cluster is authorized to pull
from the container registry.
RBAC issues.

Next steps
Learn more about features in Azure Monitor that support monitoring of applications on AKS:
 Azure Monitor for containers overview
Understand AKS cluster performance with Azure Monitor for containers
For more information about using metrics for performance tuning, see see Performance tuning a distributed
application.
 CI/CD for microservices architectures
12/18/2020 • 8 minutes to read • Edit Online

Faster release cycles are one of the major advantages of microservices architectures. But without a good CI/CD
process, you won't achieve the agility that microservices promise. This article describes the challenges and
recommends some approaches to the problem.

What is CI/CD?
When we talk about CI/CD, we're really talking about several related processes: Continuous integration, continuous
delivery, and continuous deployment.
Continuous integration . Code changes are frequently merged into the main branch. Automated build and
test processes ensure that code in the main branch is always production-quality.
Continuous deliver y . Any code changes that pass the CI process are automatically published to a
production-like environment. Deployment into the live production environment may require manual
approval, but is otherwise automated. The goal is that your code should always be ready to deploy into
production.
Continuous deployment . Code changes that pass the previous two steps are automatically deployed into
production.
Here are some goals of a robust CI/CD process for a microservices architecture:
Each team can build and deploy the services that it owns independently, without affecting or disrupting
other teams.
Before a new version of a service is deployed to production, it gets deployed to dev/test/QA environments
for validation. Quality gates are enforced at each stage.
A new version of a service can be deployed side by side with the previous version.
Sufficient access control policies are in place.
For containerized workloads, you can trust the container images that are deployed to production.

Why a robust CI/CD pipeline matters

In a traditional monolithic application, there is a single build pipeline whose output is the application executable. All
development work feeds into this pipeline. If a high-priority bug is found, a fix must be integrated, tested, and
published, which can delay the release of new features. You can mitigate these problems by having well-factored
modules and using feature branches to minimize the impact of code changes. But as the application grows more
complex, and more features are added, the release process for a monolith tends to become more brittle and likely
to break.
Following the microservices philosophy, there should never be a long release train where every team has to get in
line. The team that builds service "A" can release an update at any time, without waiting for changes in service "B"
to be merged, tested, and deployed.
To achieve a high release velocity, your release pipeline must be automated and highly reliable to minimize risk. If
you release to production one or more times daily , regressions or service disruptions must be rare. At the same
time, if a bad update does get deployed, you must have a reliable way to quickly roll back or roll forward to a
previous version of a service.

Challenges
Many small independent code bases . Each team is responsible for building its own service, with its own
build pipeline. In some organizations, teams may use separate code repositories. Separate repositories can
lead to a situation where the knowledge of how to build the system is spread across teams, and nobody in
the organization knows how to deploy the entire application. For example, what happens in a disaster
recovery scenario, if you need to quickly deploy to a new cluster?
Mitigation : Have a unified and automated pipeline to build and deploy services, so that this knowledge is
not "hidden" within each team.
Multiple languages and frameworks . With each team using its own mix of technologies, it can be
difficult to create a single build process that works across the organization. The build process must be
flexible enough that every team can adapt it for their choice of language or framework.
Mitigation : Containerize the build process for each service. That way, the build system just needs to be able
to run the containers.
Integration and load testing . With teams releasing updates at their own pace, it can be challenging to
design robust end-to-end testing, especially when services have dependencies on other services. Moreover,
running a full production cluster can be expensive, so it's unlikely that every team will run its own full
cluster at production scales, just for testing.
Release management . Every team should be able to deploy an update to production. That doesn't mean
that every team member has permissions to do so. But having a centralized Release Manager role can
reduce the velocity of deployments.
Mitigation : The more that your CI/CD process is automated and reliable, the less there should be a need
for a central authority. That said, you might have different policies for releasing major feature updates
versus minor bug fixes. Being decentralized doesn't mean zero governance.
Ser vice updates . When you update a service to a new version, it shouldn't break other services that
depend on it.
Mitigation : Use deployment techniques such as blue-green or canary release for non-breaking changes.
For breaking API changes, deploy the new version side by side with the previous version. That way, services
that consume the previous API can be updated and tested for the new API. See Updating services, below.

Monorepo vs. multi-repo

Before creating a CI/CD workflow, you must know how the code base will be structured and managed.
Do teams work in separate repositories or in a monorepo (single repository)?
 What is your branching strategy?
Who can push code to production? Is there a release manager role?
The monorepo approach has been gaining favor but there are advantages and disadvantages to both.

M O N O REP O M ULT IP L E REP O S

Advantages Code sharing Clear ownership per team

Easier to standardize code and tooling Potentially fewer merge conflicts
Easier to refactor code Helps to enforce decoupling of
Discoverability - single view of the code microservices

Challenges Changes to shared code can affect Harder to share code

multiple microservices Harder to enforce coding standards
Greater potential for merge conflicts Dependency management
Tooling must scale to a large code base Diffuse code base, poor discoverability
Access control Lack of shared infrastructure
More complex deployment process

Updating services
There are various strategies for updating a service that's already in production. Here we discuss three common
options: Rolling update, blue-green deployment, and canary release.
Rolling updates
In a rolling update, you deploy new instances of a service, and the new instances start receiving requests right
away. As the new instances come up, the previous instances are removed.
Example. In Kubernetes, rolling updates are the default behavior when you update the pod spec for a Deployment.
The Deployment controller creates a new ReplicaSet for the updated pods. Then it scales up the new ReplicaSet
while scaling down the old one, to maintain the desired replica count. It doesn't delete old pods until the new ones
are ready. Kubernetes keeps a history of the update, so you can roll back an update if needed.
Example. Azure Service Fabric uses the rolling update strategy by default. This strategy is best suited for
deploying a version of a service with new features without changing existing APIs. Service Fabric starts an upgrade
deployment by updating the application type to a subset of the nodes or an update domain. It then rolls forward to
the next update domain until all domains are upgraded. If an upgrade domain fails to update, the application type
rolls back to the previous version across all domains. Be aware that an application type with multiple services (and
if all services are updated as part of one upgrade deployment) is prone to failure. If one service fails to update, the
entire application is rolled back to the previous version and the other services are not updated.
One challenge of rolling updates is that during the update process, a mix of old and new versions are running and
receiving traffic. During this period, any request could get routed to either of the two versions.
For breaking API changes, a good practice is to support both versions side by side, until all clients of the previous
version are updated. See API versioning.
Blue -green deployment
In a blue-green deployment, you deploy the new version alongside the previous version. After you validate the
new version, you switch all traffic at once from the previous version to the new version. After the switch, you
monitor the application for any problems. If something goes wrong, you can swap back to the old version.
Assuming there are no problems, you can delete the old version.
With a more traditional monolithic or N-tier application, blue-green deployment generally meant provisioning two
identical environments. You would deploy the new version to a staging environment, then redirect client traffic to
the staging environment — for example, by swapping VIP addresses. In a microservices architecture, updates
happen at the microservice level, so you would typically deploy the update into the same environment and use a
service discovery mechanism to swap.
Example . In Kubernetes, you don't need to provision a separate cluster to do blue-green deployments. Instead, you
can take advantage of selectors. Create a new Deployment resource with a new pod spec and a different set of
labels. Create this deployment, without deleting the previous deployment or modifying the service that points to it.
Once the new pods are running, you can update the service's selector to match the new deployment.
One drawback of blue-green deployment is that during the update, you are running twice as many pods for the
service (current and next). If the pods require a lot of CPU or memory resources, you may need to scale out the
cluster temporarily to handle the resource consumption.
Canary release
In a canary release, you roll out an updated version to a small number of clients. Then you monitor the behavior of
the new service before rolling it out to all clients. This lets you do a slow rollout in a controlled fashion, observe
real data, and spot problems before all customers are affected.
A canary release is more complex to manage than either blue-green or rolling update, because you must
dynamically route requests to different versions of the service.
Example . In Kubernetes, you can configure a Service to span two replica sets (one for each version) and adjust the
replica counts manually. However, this approach is rather coarse-grained, because of the way Kubernetes load
balances across pods. For example, if you have a total of 10 replicas, you can only shift traffic in 10% increments. If
you are using a service mesh, you can use the service mesh routing rules to implement a more sophisticated
canary release strategy.

Next steps
Learn specific CI/CD practices for microservices running on Kubernetes.
CI/CD for microservices on Kubernetes
 Building a CI/CD pipeline for microservices on
Kubernetes
12/18/2020 • 12 minutes to read • Edit Online

It can be challenging to create a reliable CI/CD process for a microservices architecture. Individual teams must be
able to release services quickly and reliably, without disrupting other teams or destabilizing the application as a
whole.
This article describes an example CI/CD pipeline for deploying microservices to Azure Kubernetes Service (AKS).
Every team and project is different, so don't take this article as a set of hard-and-fast rules. Instead, it's meant to be
a starting point for designing your own CI/CD process.
The example pipeline described here was created for a microservices reference implementation called the Drone
Delivery application, which you can find on GitHub. The application scenario is described here.
The goals of the pipeline can be summarized as follows:
Teams can build and deploy their services independently.
Code changes that pass the CI process are automatically deployed to a production-like environment.
Quality gates are enforced at each stage of the pipeline.
A new version of a service can be deployed side by side with the previous version.
For more background, see CI/CD for microservices architectures.

Assumptions
For purposes of this example, here are some assumptions about the development team and the code base:
The code repository is a monorepo, with folders organized by microservice.
The team's branching strategy is based on trunk-based development.
The team uses release branches to manage releases. Separate releases are created for each microservice.
The CI/CD process uses Azure Pipelines to build, test, and deploy the microservices to AKS.
The container images for each microservice are stored in Azure Container Registry.
The team uses Helm charts to package each microservice.
These assumptions drive many of the specific details of the CI/CD pipeline. However, the basic approach described
here be adapted for other processes, tools, and services, such as Jenkins or Docker Hub.

Validation builds
Suppose that a developer is working on a microservice called the Delivery Service. While developing a new
feature, the developer checks code into a feature branch. By convention, feature branches are named feature/* .
The build definition file includes a trigger that filters by the branch name and the source path:

trigger:
batch: true
branches:
include:
# for new release to production: release flow strategy
- release/delivery/v*
- refs/release/delivery/v*
- master
- feature/delivery/*
- topic/delivery/*
paths:
include:
- /src/shipping/delivery/

See the source file.

Using this approach, each team can have its own build pipeline. Only code that is checked into the
/src/shipping/delivery folder triggers a build of the Delivery Service. Pushing commits to a branch that matches
the filter triggers a CI build. At this point in the workflow, the CI build runs some minimal code verification:
1. Build the code.
2. Run unit tests.
The goal is to keep build times short, so the developer can get quick feedback. Once the feature is ready to merge
into master, the developer opens a PR. This triggers another CI build that performs some additional checks:
1. Build the code.
2. Run unit tests.
3. Build the runtime container image.
4. Run vulnerability scans on the image.

NOTE
In Azure DevOps Repos, you can define policies to protect branches. For example, the policy could require a successful CI
build plus a sign-off from an approver in order to merge into master.

Full CI/CD build

At some point, the team is ready to deploy a new version of the Delivery service. The release manager creates a
branch from master with this naming pattern: release/<microservice name>/<semver> . For example,
release/delivery/v1.0.2 .
Creation of this branch triggers a full CI build that runs all of the previous steps plus:
1. Push the container image to Azure Container Registry. The image is tagged with the version number taken from
the branch name.
2. Run helm package to package the Helm chart for the service. The chart is also tagged with a version number.
3. Push the Helm package to Container Registry.
Assuming this build succeeds, it triggers a deployment (CD) process using an Azure Pipelines release pipeline. This
pipeline has the following steps:
1. Deploy the Helm chart to a QA environment.
2. An approver signs off before the package moves to production. See Release deployment control using
approvals.
3. Retag the Docker image for the production namespace in Azure Container Registry. For example, if the current
tag is myrepo.azurecr.io/delivery:v1.0.2 , the production tag is myrepo.azurecr.io/prod/delivery:v1.0.2 .
4. Deploy the Helm chart to the production environment.
Even in a monorepo, these tasks can be scoped to individual microservices, so that teams can deploy with high
velocity. The process has some manual steps: Approving PRs, creating release branches, and approving
deployments into the production cluster. These steps are manual by policy — they could be automated if the
organization prefers.

Isolation of environments
You will have multiple environments where you deploy services, including environments for development, smoke
testing, integration testing, load testing, and finally production. These environments need some level of isolation. In
Kubernetes, you have a choice between physical isolation and logical isolation. Physical isolation means deploying
to separate clusters. Logical isolation uses namespaces and policies, as described earlier.
Our recommendation is to create a dedicated production cluster along with a separate cluster for your dev/test
environments. Use logical isolation to separate environments within the dev/test cluster. Services deployed to the
dev/test cluster should never have access to data stores that hold business data.

Build process
When possible, package your build process into a Docker container. That allows you to build your code artifacts
using Docker, without needing to configure the build environment on each build machine. A containerized build
process makes it easy to scale out the CI pipeline by adding new build agents. Also, any developer on the team can
build the code simply by running the build container.
By using multi-stage builds in Docker, you can define the build environment and the runtime image in a single
Dockerfile. For example, here's a Dockerfile that builds a .NET application:
 FROM mcr.microsoft.com/dotnet/core/runtime:3.1 AS base
WORKDIR /app

FROM mcr.microsoft.com/dotnet/core/sdk:3.1 AS build

WORKDIR /src/Fabrikam.Workflow.Service

COPY Fabrikam.Workflow.Service/Fabrikam.Workflow.Service.csproj .
RUN dotnet restore Fabrikam.Workflow.Service.csproj

COPY Fabrikam.Workflow.Service/. .
RUN dotnet build Fabrikam.Workflow.Service.csproj -c release -o /app --no-restore

FROM build AS testrunner

WORKDIR /src/tests

COPY Fabrikam.Workflow.Service.Tests/*.csproj .
RUN dotnet restore Fabrikam.Workflow.Service.Tests.csproj

COPY Fabrikam.Workflow.Service.Tests/. .
ENTRYPOINT ["dotnet", "test", "--logger:trx"]

FROM build AS publish

RUN dotnet publish Fabrikam.Workflow.Service.csproj -c Release -o /app

FROM base AS final

WORKDIR /app
COPY --from=publish /app .
ENTRYPOINT ["dotnet", "Fabrikam.Workflow.Service.dll"]

See the source file.

This Dockerfile defines several build stages. Notice that the stage named base uses the .NET runtime, while the
stage named build uses the full .NET SDK. The build stage is used to build the .NET project. But the final runtime
container is built from base , which contains just the runtime and is significantly smaller than the full SDK image.
Building a test runner
Another good practice is to run unit tests in the container. For example, here is part of a Docker file that builds a
test runner:

FROM build AS testrunner

WORKDIR /src/tests

COPY Fabrikam.Workflow.Service.Tests/*.csproj .
RUN dotnet restore Fabrikam.Workflow.Service.Tests.csproj

COPY Fabrikam.Workflow.Service.Tests/. .
ENTRYPOINT ["dotnet", "test", "--logger:trx"]

A developer can use this Docker file to run the tests locally:

docker build . -t delivery-test:1 --target=testrunner

docker run delivery-test:1

The CI pipeline should also run the tests as part of the build verification step.
Note that this file uses the Docker ENTRYPOINT command to run the tests, not the Docker RUN command.
If you use the RUN command, the tests run every time you build the image. By using ENTRYPOINT , the tests are
opt-in. They run only when you explicitly target the testrunner stage.
A failing test doesn't cause the Docker build command to fail. That way you can distinguish container build
 failures from test failures.
Test results can be saved to a mounted volume.
Container best practices
Here are some other best practices to consider for containers:
Define organization-wide conventions for container tags, versioning, and naming conventions for resources
deployed to the cluster (pods, services, and so on). That can make it easier to diagnose deployment issues.
During the development and test cycle, the CI/CD process will build many container images. Only some of
those images are candidates for release, and then only some of those release candidates will get promoted
to production. Have a clear versioning strategy, so that you know which images are currently deployed to
production, and can roll back to a previous version if necessary.
Always deploy specific container version tags, not latest .
Use namespaces in Azure Container Registry to isolate images that are approved for production from
images that are still being tested. Don't move an image into the production namespace until you're ready to
deploy it into production. If you combine this practice with semantic versioning of container images, it can
reduce the chance of accidentally deploying a version that wasn't approved for release.
Follow the principle of least privilege by running containers as a nonprivileged user. In Kubernetes, you can
create a pod security policy that prevents containers from running as root. See Prevent Pods From Running
With Root Privileges.

Helm charts
Consider using Helm to manage building and deploying services. Here are some of the features of Helm that help
with CI/CD:
Often a single microservice is defined by multiple Kubernetes objects. Helm allows these objects to be
packaged into a single Helm chart.
A chart can be deployed with a single Helm command, rather than a series of kubectl commands.
Charts are explicitly versioned. Use Helm to release a version, view releases, and roll back to a previous version.
Tracking updates and revisions, using semantic versioning, along with the ability to roll back to a previous
version.
Helm charts use templates to avoid duplicating information, such as labels and selectors, across many files.
Helm can manage dependencies between charts.
Charts can be stored in a Helm repository, such as Azure Container Registry, and integrated into the build
pipeline.
For more information about using Container Registry as a Helm repository, see Use Azure Container Registry as a
Helm repository for your application charts.

IMPORTANT
This feature is currently in preview. Previews are made available to you on the condition that you agree to the supplemental
terms of use. Some aspects of this feature may change prior to general availability (GA).

A single microservice may involve multiple Kubernetes configuration files. Updating a service can mean touching
all of these files to update selectors, labels, and image tags. Helm treats these as a single package called a chart and
allows you to easily update the YAML files by using variables. Helm uses a template language (based on Go
templates) to let you write parameterized YAML configuration files.
For example, here's part of a YAML file that defines a deployment:
 apiVersion: apps/v1
kind: Deployment
metadata:
name: {{ include "package.fullname" . | replace "." "" }}
labels:
app.kubernetes.io/name: {{ include "package.name" . }}
app.kubernetes.io/instance: {{ .Release.Name }}
annotations:
kubernetes.io/change-cause: {{ .Values.reason }}

...

spec:
containers:
- name: &package-container_name fabrikam-package
image: {{ .Values.dockerregistry }}/{{ .Values.image.repository }}:{{ .Values.image.tag }}
imagePullPolicy: {{ .Values.image.pullPolicy }}
env:
- name: LOG_LEVEL
value: {{ .Values.log.level }}

See the source file.

You can see that the deployment name, labels, and container spec all use template parameters, which are provided
at deployment time. For example, from the command line:

helm install $HELM_CHARTS/package/ \

--set image.tag=0.1.0 \
--set image.repository=package \
--set dockerregistry=$ACR_SERVER \
--namespace backend \
--name package-v0.1.0

Although your CI/CD pipeline could install a chart directly to Kubernetes, we recommend creating a chart archive
(.tgz file) and pushing the chart to a Helm repository such as Azure Container Registry. For more information, see
Package Docker-based apps in Helm charts in Azure Pipelines.
Consider deploying Helm to its own namespace and using role-based access control (RBAC) to restrict which
namespaces it can deploy to. For more information, see Role-based Access Control in the Helm documentation.
Revisions
Helm charts always have a version number, which must use semantic versioning. A chart can also have an
appVersion . This field is optional, and doesn't have to be related to the chart version. Some teams might want to
application versions separately from updates to the charts. But a simpler approach is to use one version number,
so there's a 1:1 relation between chart version and application version. That way, you can store one chart per
release and easily deploy the desired release:

helm install <package-chart-name> --version <desiredVersion>

Another good practice is to provide a change-cause annotation in the deployment template:

 apiVersion: apps/v1
kind: Deployment
metadata:
name: {{ include "delivery.fullname" . | replace "." "" }}
labels:
...
annotations:
kubernetes.io/change-cause: {{ .Values.reason }}

This lets you view the change-cause field for each revision, using the kubectl rollout history command. In the
previous example, the change-cause is provided as a Helm chart parameter.

kubectl rollout history deployments/delivery-v010 -n backend

deployment.extensions/delivery-v010
REVISION CHANGE-CAUSE
1 Initial deployment

You can also use the helm list command to view the revision history:

helm list

NAME REVISION UPDATED STATUS CHART APP VERSION

NAMESPACE
delivery-v0.1.0 1 Sun Apr 7 00:25:30 2020 DEPLOYED delivery-v0.1.0 v0.1.0 backend

Azure DevOps Pipeline

In Azure Pipelines, pipelines are divided into build pipelines and release pipelines. The build pipeline runs the CI
process and creates build artifacts. For a microservices architecture on Kubernetes, these artifacts are the container
images and Helm charts that define each microservice. The release pipeline runs that CD process that deploys a
microservice into a cluster.
Based on the CI flow described earlier in this article, a build pipeline might consist of the following tasks:
1. Build the test runner container.

- task: Docker@1
inputs:
azureSubscriptionEndpoint: $(AzureSubscription)
azureContainerRegistry: $(AzureContainerRegistry)
arguments: '--pull --target testrunner'
dockerFile: $(System.DefaultWorkingDirectory)/$(dockerFileName)
imageName: '$(imageName)-test'

2. Run the tests, by invoking docker run against the test runner container.
 - task: Docker@1
inputs:
azureSubscriptionEndpoint: $(AzureSubscription)
azureContainerRegistry: $(AzureContainerRegistry)
command: 'run'
containerName: testrunner
volumes: '$(System.DefaultWorkingDirectory)/TestResults:/app/tests/TestResults'
imageName: '$(imageName)-test'
runInBackground: false

3. Publish the test results. See Build an image.

- task: PublishTestResults@2
inputs:
testResultsFormat: 'VSTest'
testResultsFiles: 'TestResults/*.trx'
searchFolder: '$(System.DefaultWorkingDirectory)'
publishRunAttachments: true

4. Build the runtime container.

- task: Docker@1
inputs:
azureSubscriptionEndpoint: $(AzureSubscription)
azureContainerRegistry: $(AzureContainerRegistry)
dockerFile: $(System.DefaultWorkingDirectory)/$(dockerFileName)
includeLatestTag: false
imageName: '$(imageName)'

5. Push to the container to Azure Container Registry (or other container registry).

- task: Docker@1
inputs:
azureSubscriptionEndpoint: $(AzureSubscription)
azureContainerRegistry: $(AzureContainerRegistry)
command: 'Push an image'
imageName: '$(imageName)'
includeSourceTags: false

6. Package the Helm chart.

- task: HelmDeploy@0
inputs:
command: package
chartPath: $(chartPath)
chartVersion: $(Build.SourceBranchName)
arguments: '--app-version $(Build.SourceBranchName)'

7. Push the Helm package to Azure Container Registry (or other Helm repository).

task: AzureCLI@1
inputs:
azureSubscription: $(AzureSubscription)
scriptLocation: inlineScript
inlineScript: |
az acr helm push $(System.ArtifactsDirectory)/$(repositoryName)-$(Build.SourceBranchName).tgz --
name $(AzureContainerRegistry);
 See the source file.
The output from the CI pipeline is a production-ready container image and an updated Helm chart for the
microservice. At this point, the release pipeline can take over. It performs the following steps:
Deploy to dev/QA/staging environments.
Wait for an approver to approve or reject the deployment.
Retag the container image for release
Push the release tag to the container registry.
Upgrade the Helm chart in the production cluster.
For more information about creating a release pipeline, see Release pipelines, draft releases, and release options.
The following diagram shows the end-to-end CI/CD process described in this article:

Next steps
This article was based on a reference implementation that you can find on GitHub.
 Monoliths to microservices using domain-driven
design
12/18/2020 • 6 minutes to read • Edit Online

This article describes how to use domain-driven design (DDD) to migrate a monolithic application to microservices.
A monolithic application is typically an application system in which all of the relevant modules are packaged
together as a single deployable unit of execution. For example, it might be a Java Web Application (WAR) running
on Tomcat or an ASP.NET application running on IIS. A typical monolithic application uses a layered design, with
separate layers for UI, application logic, and data access.

These systems start small but tend to grow over time to meet business needs. At some point, as new features are
added, a monolithic application can begin to suffer from the following problems:
The individual parts of the system cannot be scaled independently, because they are tightly coupled.
It is hard to maintain the code, because of tight coupling and hidden dependencies.
Testing becomes harder, increasing the probability of introducing vulnerabilities.
These problems can become an obstacle to future growth and stability. Teams become wary of making changes,
especially if the original developers are no longer working on the project and design documents are sparse or
outdated.
Despite these limitations, a monolithic design can make sense as a starting point for an application. Monoliths are
often the quickest path to building a proof-of-concept or minimal viable product. In the early phases of
development, monoliths tend to be:
Easier to build, because there is a single shared code base.
Easier to debug, because the code runs within a single process and memory space.
Easier to reason about, because there are fewer moving parts.
As the application grows in complexity, however, these advantages can disappear. Large monoliths often become
progressively harder to build, debug, and reason about. At some point, the problems outweigh the benefits. This is
the point when it can make sense to migrate the application to a microservices architecture. Unlike monoliths,
microservices are typically decentralized, loosely coupled units of execution. The following diagram shows a typical
microservices architecture:
Migrating a monolith to a microservice requires significant time and investment to avoid failures or overruns. To
ensure that any migration is successful, it's good to understand both the benefits and also challenges that
microservices bring. The benefits include:
Services can evolve independently based on user needs.
Services can scale independently to meet user demand.
Over time, development cycles become faster as features can be released to market quicker.
Services are isolated and are more tolerant of failure.
A single service that fails will not bring down the entire application.
Testing becomes more coherent and consistent, using behavior-driven development.
For more information about the benefits and challenges of microservices, see Microservices architecture style.

Apply domain-driven design

Any migration strategy should allow teams to incrementally refactor the application into smaller services, while still
providing continuity of service to end users. Here's the general approach:
Stop adding functionality to the monolith.
Split the front end from the back end.
Decompose and decouple the monolith into a series of microservices.
To help facilitate this decomposition, a viable software development approach is to apply the principles of domain-
driven design (DDD).
Domain Driven Design (DDD) is a software development approach first introduced by Eric Evans. DDD requires a
good understanding of the domain for which the application will be written. The necessary domain knowledge to
create the application resides within the people who understand it — the domain experts.
The DDD approach can be applied retroactively to an existing application, as a way to begin decomposing the
application.
1. Start with a ubiquitous language, a common vocabulary that is shared between all stakeholders.
2. Identify the relevant modules in the monolithic application and apply the common vocabulary to them.
3. Define the domain models of the monolithic application. The domain model is an abstract model of the
business domain.
4. Define bounded contexts for the models. A bounded context is the boundary within a domain where a
particular domain model applies. Apply explicit boundaries with clearly defined models and responsibilities.
The bounded contexts identified in step 4 are candidates for refactoring into smaller microservices. The following
diagram shows the existing monolith with the bounded contexts overlaid:

For more information about using a DDD approach for microservices architectures, see Using domain analysis to
model microservices.

Use glue code (anti-corruption layer)

While this investigative work is carried out to inventory the monolithic application, new functionality can be added
by applying the principles of DDD as separate services. "Glue code" allows the monolithic application to proxy calls
to the new service to obtain new functionality.

The glue code (adapter pattern) effectively acts as an anti-corruption layer, ensuring that the new service is not
polluted by data models required by the monolithic application. The glue code helps to mediate interactions
between the two and ensures that only data required by the new service is passed to enable compatibility.
Through the process of refactoring, teams can inventory the monolithic application and identify candidates for
microservices refactoring while also establishing new functionality with new services.
For more information about anti-corruption layers, see Anti-Corruption Layer pattern.

Create a presentation layer

The next step is to separate the presentation layer from the backend layer. In a traditional n-tier application, the
application (business) layer tends to be the components that are core to the application and have domain logic
within them. These coarse-grained APIs interact with the data access layer to retrieve persisted data from within a
database. These APIs establish a natural boundary to the presentation tier, and help to decouple the presentation
tier into a separate application space.
The follow diagram shows the presentation layer (UI) split out from the application logic and data access layers.

This diagram also introduces another layer, the API gateway, that sits between the presentation layer and the
application logic. The API gateway is a façade layer that provides a consistent and uniform interface for the
presentation layer to interact with, while allowing downstream services to evolve independently, without affecting
the application. The API Gateway may use a technology such as Azure API Management, and allows the application
to interact in a RESTful manner.
The presentation tier can be developed in any language or framework that the team has expertise in, such as a
single page application or an MVC application. These applications interact with the microservices via the gateway,
using standard HTTP calls. For more information about API Gateways, see Using API gateways in microservices.

Start to retire the monolith

At this stage, the team can begin peeling away the monolithic application and slowly extract the services that have
been established by their bounded contexts into their own set of microservices. The microservices can expose a
RESTful interface for the application layer to interact with, through the API gateway, with glue code in place to
communicate with the monolith in specific circumstances.
As you continue to peel away the monolith, eventually there will come the point when it no longer needs to exist,
and the microservices have been successfully extracted from the monolith. At this point, the anti-corruption layer
(glue code) can safely be removed.
This approach is an example of the Strangler Fig pattern and allows for a controlled decomposition of a monolith
into a set of microservices. Over time, as existing functionality is moved into microservices, the monolith will shrink
in size and complexity, to the point that it no longer exists.

Next steps
When the application has been decomposed into constituent microservices, it becomes possible to use modern
orchestration tools such as Azure DevOps to manage the lifecycle of each service. For more information, see CI/CD
for microservices architectures.
 Modernize enterprise applications with Azure Service
Fabric
12/18/2020 • 28 minutes to read • Edit Online

This article provides guidelines for moving Windows applications to an Azure compute platform without rewriting.
This migration uses container support in Azure Service Fabric.
A typical approach for migrating existing workloads to the cloud is the lift-and-shift strategy. In IaaS virtual machine
(VM) migrations, you provision VMs with network and storage components and deploy the existing applications
onto those VMs. Unfortunately, lift-and-shift often results in overprovisioning and overpaying for compute
resources. Another approach is to move to PaaS platforms or refactor code into microservices and run in newer
serverless platforms. But those options typically involve changing existing code.
Containers and container orchestration offer improvements. Containerizing an existing application enables it to run
on a cluster with other applications. It provides tight control over resources, scaling, and shared monitoring and
DevOps.
Optimizing and provisioning the right amount of compute resources for containerization isn't trivial. Service
Fabric's orchestration allows an organization to migrate Windows and Linux applications to a runtime platform
without changing code and to scale the needs of the application without overprovisioning VMs. The result is better
density, better hardware use, simplified operations, and overall lower cloud-compute costs.
An enterprise can use Service Fabric as a platform to run a large set of existing Windows-based web applications
with improved density, monitoring, consistency, and DevOps, all within a secure extended private network in the
cloud. The principle is to use Docker and Service Fabric's containerization support that packages and hosts existing
web applications on a shared cluster with shared monitoring and operations, in order to maximize cloud compute
resources for the ideal performance-to-cost ratio.
This article describes the processes, capabilities, and Service Fabric features that enable containerizing in an optimal
environment for a large enterprise. The guidance is scoped to web applications and Windows containers. Before
reading this article, get familiar with core Windows container and Service Fabric concepts. For more information,
see:
Create your first Service Fabric container application on Windows
Service Fabric terminology overview
Service Fabric best practices overview

Resources
Sample: Modernization templates and scripts.
The repo has these resources:
An example Azure Resource Manager template to bring up an Azure Service Fabric cluster.
A reverse-proxy solution for brokering web requests into the Service Fabric cluster to the destination containers.
Sample Service Fabric application configuration and scripts that show the use of placement, resource
constraints, and autoscaling.
Sample scripts and Docker files that build and package an existing web application.
Customize the templates in this repo for your cluster. The templates implement the best practices described in this
article.
Evaluate requirements
Before containerizing existing applications, evaluate requirements. Select applications that are right for migration,
choose the right developer workstation, and determine network requirements.
Application selection
First, determine the type of applications that are best suited for a containerized platform, full virtual machines, and
pure PaaS environment. The application could be a shared application that is built with Service Fabric to share
Windows Server hosts across various containerized applications. Each Service Fabric host can run multiple different
applications running in isolated Windows containers.
Consider creating a set of criteria to determine such applications. Here are some example criteria of containerized
Windows applications in Service Fabric.
HTTP/HTTPS web and application tiers without database dependency.
Stateless web applications.
Built with .NET Framework versions 3.5 and later.
Do not have hardware dependency or access device drivers.
Applications can run on Windows Server 2016 and later versions.
All dependencies can be containerized, such as are most .NET assemblies, WCF, COM+.

NOTE
Dependencies that cannot be containerized include MSMQ (Currently supported in preview releases of Windows
Server Core post 1709).

Applications can compile and build in Visual Studio.

For the web applications, databases, and other required servers (such as Active Directory) exist outside the Service
Fabric cluster in IaaS VMs, PaaS, or on-premises.
Developer workstation requirements
From an application development perspective, determine the workstation requirements.
Docker for Windows is required for developers to containerize and test their applications prior to deployment.
Visual Studio Docker support is required. Standardize on the latest version of Visual Studio for the best Docker
compatibility.
If workstations don't have enough hardware resources to oversee those requirements, use Azure compute
resources for speed and productivity gains. An option is the Azure DevTest Labs Service. Docker for Windows,
and Visual Studio 2019 require a minimum of 8 GB of memory.
Networking requirements
Service Fabric orchestration provides a platform for hosting, deploying, scaling, and operating applications at
enterprise scale. Most large enterprises that use Azure:
Extend their corporate network with a private address space to an Azure subscription. use either ExpressRoute or
a Site-to-Site VPN to provide secure on-premises connectivity.
Want to control inbound and outbound network traffic through third-party firewall appliances and/or Azure
Network Security Group rules.
Want tight control over the address space requirements and subnets.
Service Fabric is suitable as a containerization platform. It plugs into an existing cloud infrastructure and doesn't
require open public ingress endpoints. You just need to carve out the necessary address space for Service Fabric's
IP address requirements. For details, see the Service Fabric Networking section in this article.

Containerize existing Windows applications

After you've determined the applications that meet the selection criteria, containerize them into Docker images. The
result is containerized .NET web application running in IIS where all tiers run in one container.

NOTE
You can use multiple containers; one per tier.

Here are the basic steps for containerizing an application.

1. Open the project in Visual Studio.
2. Make sure the project compiles and runs locally on the developer workstation.
3. Add a Dockerfile to the project. This Dockerfile example shows a basic .NET MVC application.

FROM mcr.microsoft.com/dotnet/framework/aspnet:4.8
ADD PublishOutput/ /inetpub/wwwroot

# add a certificate and configure SSL

SHELL ["powershell","-command"]
RUN New-Item -ItemType Directory C:\install
ADD installwebsite.ps1 c:\install\
ADD wildcard.pfx c:\install\
# Powershell script to configure SSL on the website
RUN c:\install\configurewebsite

# plugin into SF healthcheck ensuring the container website is running

HEALTHCHECK --interval=30s --timeout=30s --start-period=60s --retries=3 CMD curl -f http://localhost/ ||
exit 1

4. Test locally by using Docker For Windows. The application must successfully run in a Docker container by
using the Visual Studio debug experience. For more information, see Deploy a .NET app using Docker
Compose.
5. Build (if needed), tag, and push the tested image to a container registry, like the Azure Container Registry
service. This example uses an existing Azure Container Registry named MyAcr and Docker build/tag/push to
build/deploy appA to the registry.

docker login myacr.azurecr.io -u myacr -p <pwd>

docker build -t appa .
docker tag appa myacr.azurecr.io/appa:1.0
docker push myacr.azurecr.io/appa:1.0

The image is tagged with a version number that Service Fabric references when it deploys and versions the
container. Azure DevOps encapsulates and executes the manual Docker build/tag/push process. DevOps details are
described in the DevOps and CI/CD section.

NOTE
In the preceding example, the base image is "mcr.microsoft.com/dotnet/framework/aspnet:4.8" from the Microsoft Container
Registry.

Here are some considerations about the base images:

 The base image could be a locked-down custom enterprise image that enforces enterprise requirements. For a
shared application, isolation boundaries can be created through credentials or by using separate registry. It's
recommended that enterprise-supported docker images be kept separately and stored in an isolated container
registry.
Avoid storing the registry login credentials in configuration files. Instead, use (role-based access control) RBAC
and Azure Active Directory service principals with Azure Container Registry. Provide read-only access to
registries depending on your enterprise requirements.
For information about running an IIS ASP.net MVC application in a Windows container, see Migrating ASP.NET MVC
Applications to Windows Containers.

Service Fabric cluster configuration for enterprise deployments

To deploy a Service Fabric cluster, start with the sample Azure Resource Manager template in this GitHub Repo and
customize it to fit your requirements. You also deploy a cluster through the Azure portal, but that option should be
used for development/test provisioning.
Service Fabric node types
A Service Fabric cluster contains one or more node types. A node type typically maps to an Azure virtual machine
scale set which is a set of one or more VM instances with same properties. The cluster must have at least one node
type-primary node type, which runs the Service Fabric system services. Do not run your application container on
the primary node type because the container can compete with the system services for resources. Consider,
designing a cluster with two or more node types depending on the types of applications. For example, if you have
two node types, run HTTP web and application tier containers on a non-primary node type.
This example infrastructure uses two Service Fabric node types: application and primary. You can scale in or scale
out the scale set individually. It doesn't require extensive planning and testing up front to determine the correct
application node type size (scale set instance count), because the actual size can grow and reduce in real time
driven by monitoring and metrics.

In the preceding image, the application node type has three nodes that run three containerized applications A, B,
and C. The containers can scale on demand based on CPU and memory usage. The hosts can scale out as more
applications are deployed to the cluster. Azure Load Balancer has a public IP. The primary nodes are in an internal
subnet. It has a load balancer that isn't publicly exposed.
Placement constraints
Use placement constraints that target the non-primary node type to reserve the primary node type for system
services.
This approach enables you to configure each scale set with a separate virtual network subnet. A unique subnet for
each node type uses network security group rules for inbound/outbound access to and from the subnet and
controls network flow. You can configure the primary node type with a private load balancer so that external traffic
can't access Service Fabric management and application deployment. If the application node type wants to expose
application endpoints publicly, configure it with a separate load balancer with security configuration. For an
example template that uses Windows Service with NSG rule and multiple node types, see 7 Node, 3 node type
secure Windows Service Fabric Cluster with NSG.
A scale set associated with a node type can reliably scale out to 100 VM instances by using a single placement
group as applications are added to the cluster. The primary node type often doesn't need as many instances and
can run with 5-7 nodes, depending on durability and reliability requirements.
Service Fabric networking
Service Fabric supports with two networking modes for containerized applications; nat and Open. For large
enterprise clusters that host multiple applications, use the Open mode. For more information, see Container
Networking and Constraints.
nat
By default, the cluster brings up containers by using a NAT-bridge mode to the host VM. The NAT bridge
routes requests over a defined port to the container. With this mode, only one IP address is needed per host
VM for the host's primary NIC.
To route traffic to each application container, a unique port is exposed through the load balancer. However,
that port is exposed to end users. If you don't want the port exposed, provide a URL rewrite mechanism.
Rewrite the application domain name with a unique application port. Traffic is routed to the load balancer
that front ends the cluster. One option for the rewrite mechanism is Azure Application Gateway.
Another benefit of this approach is simplistic load balancing with Azure Load Balancer. The load balancer's
probe mechanism balance traffic across the VM instances that are running the application's containers.
Open
The Open mode assigns an IP address to each running container on the host VM from the cluster's virtual
network subnet. Each host is pre-allocated with a set of IP addresses. Each container on the host is assigned
an IP from the virtual network range. You can configure this mode in the cluster Azure Resource Manager
template during cluster creation. The example infrastructure demonstrates the Open mode.
Benefits of the Open mode:
Makes connecting to application containers simple.
Provides application traceability that is, the assigned enterprise-friendly IP is constant for the life of the
container.
Is efficient with Windows containers.
There are downsides:
The number of IP addresses must be set aside during cluster creation and that number is fixed. For example, if
10 IP addresses are assigned to each host, the host supports up to 9 application containers; one IP is reserved
for the host VM's primary NIC, the remaining addresses for each container. The number of IP addresses to
configure for each host is determined based on the hardware size for the node type, and the maximum number
of containers on each host. Both factors depend on application size and needs. If you need more containers in
the cluster, add more application node type VM instances. That is because you can't change the IP number per
 instance without rebuilding the cluster.
You need a reverse proxy to route traffic to the correct destination container. The Service Fabric DNS Service can
be used by the reverse proxy to look up the application container and rewrite the HTTP request to this container.
That solution is more complex to implement than the nat mode.
For more information, see Service Fabric container networking modes.
Service Fabric runtime and Windows versions
Choose the version of Windows Server you want to run as the infrastructure host OS. The choice depends on the
testing and readiness at the time of first deployment. Supported versions are Windows Server with Containers
1607, 1709, 1803.
Also choose a version of the Service Fabric runtime to start the cluster. For a new containerization initiative, use the
latest version (Service Fabric 6.4 or later).
Other considerations:
The host OS running Service Fabric can be built by using locked down enterprise images with antivirus and
malware enforcement.
The host OS for all the Service Fabric cluster VMs can be domain joined. Domain joining isn't a requirement and
can add complexity to any Azure virtual machine. However, there are benefits. For example, if an application
requires Windows-integrated security to connect to domain-joined resources, then the domain service account
is typically used at the process level. The account is used to execute the application instead of connection string
credentials and secrets. Windows containers do not currently support direct domain joining but can be
configured to use Group Managed Storage Accounts (gMSA). The gMSA capability is supported by Service
Fabric for Windows-integrated security requirements. Each containerized application in the cluster can have its
own gMSA. gMSA requires the Service Fabric host VMs that run containers to be Active Directory domain
joined.
DNS service
Service Fabric has an internal DNS service that maps a containerized application name to its location in the cluster.
For the Open mode, the application's IP address is used instead. This service is enabled on the cluster. If you name
each service with a DNS name, traffic is routed to the application by using a reverse proxy. For information about
reverse proxy, see the Reverse proxy for inbound traffic section.
Monitoring and diagnostics
The Service Fabric Log Analytics workspace and Service Fabric solution provide detailed information about cluster
events. For information about setting it up, see Configure Azure Monitor logs to collect cluster events.
Install the [monitoring agent] for Azure Log Analytics in the Service Fabric cluster. You can then use the container
monitoring solution and view the running containers in the cluster.
Use Docker performance statistics to monitor memory and CPU use for each container.
Install the Azure Diagnostics extension that collects diagnostic data on a deployed application. For more
information, see What is Azure Diagnostics extension.
Unused container images
Container images are downloaded to each Service Fabric host and can consume space on the host disk. To free up
disk space, consider image pruning to remove images that are no longer referenced and used by running
containers. Configure this option in the Host section of the cluster manifest.
Secrets and certificates management with Key Vault
Avoid hardcoded secrets in the template. Modify the Azure Resource Manager template to add certificates (cluster
certificate) and secrets (host OS admin password) from Azure Key Vault during cluster deployment. To safely store
and retrieve secrets:
1. Create a Key Vault in a separate resource group in the same region as the cluster.
2. Protect the Key Vault with RBAC.
3. Load the cluster certificate and cluster password in Key Vault.
4. Reference this external Key Vault from the Azure Resource Manager template.
Cluster capacity planning
For an enterprise production cluster, start with 5-7 primary nodes, depending on the intended size of the overall
cluster. For more information, see Service Fabric cluster capacity planning considerations.
For a large cluster that hosts stateless web applications, the typical hardware size for the primary node type can be
Standard_D2s_v3. For the primary node type, a 5-node cluster with the Silver durability tier should scale to
containers across 50 application VM instances. Consider doing simulation tests and add more node types as
partitioning needs become apparent. Also perform tests to determine the VM size for application node type and the
number of running containers on each VM instance. Typically, 10 or more containers run on each VM, but that
number is dependent on these factors:
Resource needs, such as CPU and memory use, of the containerized applications. Web applications tend to be
more CPU intensive than memory intensive.
The compute size and available resources of the chosen node VM for the application node type.
The number of IP addresses configured per node. Each container requires an IP with Open mode. Given 15
containers on a VM, 16 IP addresses must be allocated for each VM in the application node type.
Resource needs of the underline OS. The recommendation is to leave 25% of the resources on the VM for OS
system processes.
Here are some recommendations: Choose recent Windows Server builds because older builds of Windows Server
(1607) use larger container image sizes compared to newer versions (1709 and later).
Select the most appropriate VM compute size and initial application node type VM count. Standard_D8s_v3,
Standard_D16s_v3, and Standard_DSv32_v3 are example compute sizes that are known to work well for this tier for
running containerized web applications.
Use premium storage for the production cluster because container images must be read quickly from disk. Attach
one data disk to each VM and move the Docker EE installation to that disk. That way there is enough log space for
each running container. A standard VM with the default 127-GB OS disk fills up because of container logs. A 500 GB
to 1 TB data disk is recommended per VM for the application node type scale set.
Availability and resiliency planning
Service Fabric spreads VM instances across fault and update domains to ensure maximum availability and
resilience. For an enterprise-scale production cluster, as a minimum target the Silver tier (5-node cluster) for
reliability and durability when hosting containerized stateless web applications.
Scale planning
There are two aspects to consider:
Scale in or scale out the scale set associated with the application node type by adding or removing VM
instances. VMs can be added and removed automatically through autoscaling or manually through scripts. For
an enterprise cluster, do not add or remove nodes frequently. It is important that you disable nodes in Service
Fabric before deleting them. Do not delete seed nodes from your cluster. Monitor the cluster VMs with alerts to
trigger when VM resources exceed threshold values.
Scale the application's container count across the scale set instances in the application node type.
For the application node type, start with the minimal required nodes to support the containerized applications and
ensure high availability. Have extra nodes in the cluster. A node can be removed from the cluster for maintenance
and its running containers can be temporarily moved to other nodes. The cluster can grow statically using the Add-
AzureRmSer viceFabricNode cmdlet, or dynamically by using scale set autoscale capability.
Starting with Service Fabric 6.2, application containers can autoscale individually through configuration. The
configuration is based on Docker statistics for CPU and memory use on the host. Together, those capabilities can
optimize the compute cost.
Rather than overprovisioning a VM to run an application, deploy a container to the cluster, monitor it, and then
scale out with additional containers, as necessary. Choosing the right size is easier because Docker statistics are
used to determine the number of containers.
In the example infrastructure, application A has two containers deployed across the cluster that divide load. This
approach allows the application and container combination to be adjusted later for optimization.

NOTE
Docker statistics showing individual container resource utilization is sent to Log Analytics and can be analyzed in Azure
Monitor.

Service Fabric offers constant monitoring and heath checks across the cluster. If a node is unhealthy, applications
on that node automatically move to a healthy node and the bad node stops receiving requests. Regardless of the
number of containers hosting an application, Service Fabric ensures that the application is healthy and running.
For an application that is infrequently used and can be offline, run it in the cluster with just one container instance
(such as application B and C). Service Fabric makes sure that the application is up and healthy during upgrades or
when the container needs to move to a new VM. Heath checking can reduce cost compared to hosting that
application on two redundant and overprovisioned VMs in the traditional IaaS model.

Container networking and constraints

Use the Open mode for hosting containerized web applications in the cluster. After deployment, the application is
immediately discoverable through the Service Fabric DNS service. The DNS service is a name-value lookup
between a configured service name and the resultant IP address of a container hosting the application.
To route web requests to an application, use an ingress reverse proxy. If application containers listen on different
ports (AppA port 8080, AppB on 8081), the default host NAT bridge works without issues. Azure Load Balancer
probes route the traffic appropriately. However, if you want incoming traffic over SSL/443 routed to one port for all
applications, use a reverse proxy to route traffic appropriately.
Reverse proxy for inbound traffic
Service Fabric has a built-in reverse proxy but is limited in its feature set. Therefore, deploy a different reverse
proxy. An option is the IIS Application Request Routing (ARR) extension for IIS hosted web applications. The ARR
can be deployed to a container and configured to take inbound requests and route them to the appropriate
application container. In this example, the ARR uses a NAT bridge over port 80/443, accepts all inbound web traffic,
inspects the traffic, looks up the destination container using Service Fabric DNS service, and rewrites the request to
the destination container. The traffic can be secured with SSL to the destination container. Follow the IIS Application
Request Routing sample for building an ARR reverse proxy. For information, see Using the Application Request
Routing Module.
Here is the network flow for the example infrastructure.
The key aspect of the ingress reverse proxy is inspecting inbound traffic and rewriting that traffic to the destination
container.
For example, application A is registered with the Service Fabric DNS service with the domain name:
appA.container.myorg.com. External users access the application with https://appA.myorg.com . Use public or
organizational DNS and register appA.myorg.com to point to the public IP for the application node type.
1. Requests for appA.myorg.com are routed to the Service Fabric cluster and handed off to the ARR container
listening on port 443. Service Fabric and Azure Load Balancer set that configuration value when the ARR
container is deployed.
2. When ARR gets the request, it has a condition to look for any request with the pattern='..*', and its action
rewrites the request to https://{C:1}.container.{C:2}.{C:3}/{REQUEST_URI}. Because the ARR is running in the
cluster, the Service Fabric DNS service is invoked. The service returns the destination container IP address.
3. The request is routed to the destination container. Certificates can be used for the initial request to ARR and the
rewrite to the destination container.
Here is an example ApplicationManifest.xml for Container A in the example infrastructure.
 <?xml version="1.0" encoding="utf-8"?>
<ApplicationManifest ApplicationTypeName="sfapp02Type"
ApplicationTypeVersion="1.0.0"
xmlns="http://schemas.microsoft.com/2011/01/fabric"
xmlns:xsd="http://www.w3.org/2001/XMLSchema"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance">
<Parameters>
<Parameter Name="appAsvc_InstanceCount" DefaultValue="2" />
</Parameters>
<ServiceManifestImport>
<ServiceManifestRef ServiceManifestName="appAsvcPkg" ServiceManifestVersion="1.0.0" />
<ConfigOverrides />
<Policies>
<ContainerHostPolicies CodePackageRef="Code" Isolation="default">
<!--SecurityOption Value="credentialspec=file://container_gmsa1.json"/-->
<RepositoryCredentials AccountName="myacr" Password=" " PasswordEncrypted="false"/>
<NetworkConfig NetworkType="Open"/>
</ContainerHostPolicies>
<ServicePackageResourceGovernancePolicy CpuCores="1" MemoryInMB="1024" />
</Policies>
</ServiceManifestImport>
<DefaultServices>
<Service Name="appAsvc" ServicePackageActivationMode="ExclusiveProcess"
ServiceDnsName="appA.container.myorg.com">
<StatelessService ServiceTypeName="appAsvcType" InstanceCount="[appAsvc_InstanceCount]">
<SingletonPartition />
<PlacementConstraints>(NodeTypeName==applicationNT)</PlacementConstraints>
</StatelessService>
</Service>
</DefaultServices>
</ApplicationManifest>

Uses the Open mode for the containers.

Registers the application domain name appA.container.myorg.com with the Azure DNS service.
Optionally configures the container to use an Active Directory gMSA (commented).
Uses placement constraints to deploy the container to the application node type, named applicationNT. It
instructs Service Fabric to run the container in the correct node type in the secured network subnet.
Optionally applies resource constraints. Each container is resource governed to use 1 vCPU and 1 GB of
memory on the VM host. Setting a resource governance policy is recommended because Service Fabric uses the
policy to distribute containers across the cluster, as opposed to the default even distribution of containers across
the cluster.
Here is the example ServiceManifest.xml for the containerized application appA.
 <?xml version="1.0" encoding="utf-8"?>
<ServiceManifest Name="appAsvcPkg"
Version="1.0.0"
xmlns="http://schemas.microsoft.com/2011/01/fabric"
xmlns:xsd="http://www.w3.org/2001/XMLSchema"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance">
<ServiceTypes>
<StatelessServiceType ServiceTypeName="appAsvcType" UseImplicitHost="true" />
</ServiceTypes>
<CodePackage Name="Code" Version="1.0.0">
<EntryPoint>
<ContainerHost>
<ImageName>myacr.azurecr.io/appa:1.0</ImageName>
</ContainerHost>
</EntryPoint>
</CodePackage>
<ConfigPackage Name="Config" Version="1.0.0" />
<Resources>
<Endpoints>
<Endpoint Name="appAsvcTypeEndpoint1" UriScheme="http" Port="80" Protocol="http" CodePackageRef="Code"/>
<Endpoint Name="appAsvcTypeEndpoint2" UriScheme="https" Port="443" Protocol="http"
CodePackageRef="Code"/>
</Endpoints>
</Resources>
</ServiceManifest>

The application is configured to listen on ports 80 and 443. By using the Open mode and reverse proxy, all
applications can share the same ports.
The application is containerized to an image named: myacr.azurecr.io/appa:1.0. Service Fabric invokes the Docker
daemon to pull down the image when the application is deployed. Service Fabric handles all interactions with
Docker.
The reverse proxy container uses similar manifests but isn't configured to use the Open mode. You can update
containers by versioning the Docker image, then redeploying the versioned Service Fabric package with the Star t-
Ser viceFabricApplicationUpgrade cmdlet.
For information about manifests, see Service Fabric application and service manifests.
Environmental configuration
Do not hardcode configuration values in the container image by using environment variables to pass values to a
container. A DevOps pipeline can build a container image, test in a test environment, promote to staging (or pre-
production), and promote to production. Do not rebuild an image for each environment.
Docker can pass environment variables directly to a container when starting one. In this example, Docker passes
the eShopTitle variable to the eshopweb container:

docker run -p 80:80 -d --name eshoptest -e eShopTitle=SomeName eshopweb:1.0

In a Service Fabric cluster, Service Fabric controls Docker execution and lists environment variables in the
ServiceManifest. Those variables are passed automatically when Service Fabric runs the container. You can override
the variables in ApplicationManifest.xml by using the EnvironmentOverrides element, which can be parameterized
and built from Visual Studio publish profiles for each environment.
For information about specifying environment variables in, see How to specify environment variables for services
in Service Fabric.

Security considerations
Here are some articles about container security:
Azure Service Fabric security
Service Fabric application and service security
Set up an encryption certificate and encrypt secrets on Windows clusters

Logging and monitoring

Monitoring and logging are critical to operational success and is achieved through integration with Azure Monitor
and Log Analytics.
Monitor the Service Fabric cluster and each executing containers by using the scale set extension agent for Log
Analytics and its associated Container Monitoring Solution. Make sure that you configure and install the extension
agent and solution during cluster creation. Docker statistics for container CPU and memory utilization are sent to
Log Analytics and can be queried for proactive monitoring and alerting. Set up proactive alerts through Azure
Monitor. Here are metrics that you should monitor.
High CPU and memory utilization of a container.
Container count hitting a per VM threshold.
HTTP error codes and count.
Docker Enterprise Edition (Docker EE) up or down on the host VM.
Response time of a service.
Host VM-based alerts on CPU, memory, disk, file consumption.
These Service Fabric virtual machine scale set extensions are installed on a typical node that results in logging.

ServiceFabricNode. Links the node to a storage account (support log) for tracing support. This log is used when
a ticket is opened.
IaaSDiagnostics. Collects platform events, such as ServiceFabricSystemEventTable, and stores that data in a blob
storage account (app log). The account is consumed in Log Analytics.
MicrosoftMonitoringAgent. Contains all the performance data such as Docker statistics. The data (such as
ContainerInventory and ContainerLog) is sent to Log Analytics.
Application log
If your containerized application runs in a shared cluster, you can get logs such as IIS and custom logs from the
container into Log Analytics. This option is recommended because of speed, scalability, and the ability to handle
large amounts of unstructured data.
Set up log rotation through Docker to keep the logs size manageable. For more information, see Rotating Docker
Logs - Keeping your overlay folder small.
Here are two approaches for getting application logs into Log Analytics.
Use the existing Container Monitoring Solution installed in Log Analytics. The solution automatically sends
data from the container log directories on the VM (C:\ProgramData\docker\windowsfilter*) to the
configured Log Analytics workspace. Each container creates a directory underneath the \WindowsFilter path,
and the contents are streamed to Log Analytics from MicrosoftMonitoringAgent on the VM. This way you
can send application logs to a shared directory(s) in the container and relay the logs by using Docker Logs to
the monitored container log folders.
1. Write a process script that runs in the container periodically and analyzes log files.
2. Monitor the log file changes in the shared folder and write the log changes to the command window
where Docker Logs can capture the information outside the container.
Each container records any output sent to the command line of a container. Access the output outside the
container, which is automatically executed by Container Monitoring Solution.

docker logs <ContainerID>

Move Docker to an attached VM data disk with enough storage to make sure the OS drive doesn't fill up with
container logs.
The automatic Container Monitoring Solution sends all logs to a single Log Analytics workspace. Different
containerized applications running on the same host send application logs to that shared workspace. If you
need to isolate logs such that each containerized application sends the log to a separate workspace, supply a
custom solution. That content is outside the scope of this article.
Mount external storage to each running container by using a file management service such as Azure Files.
The container logs are sent to the external storage location and don't take up disk space on the host VM.
You don't need, a data disk attached to each VM to hold Docker logs; move Docker Enterprise to the data
disk.
Create a job to monitor the Azure Files location and send logs to the appropriate Log Analytics workspace
for each installed application. The job doesn't need to run in the container. It just observes the Azure Files
location.

DevOps and CI/CD

Application containerization ensures consistency. It makes sure all Service Fabric-hosted applications use the latest
approved corporate image and provides an automatable image updating process that is consistent through
DevOps. Azure Pipelines provides the automation process. For more information, see Tutorial: Deploy an
application with CI/CD to a Service Fabric cluster.
 An enterprise may want to control the base container images in a centralized registry. The preceding workflow
shows one image registry. There could be multiple registries that are used to share central IT-built enterprise
images with application teams. One way to centralize control is for the central IT registry to allow application
teams with read-only access to the enterprise base image repository. Application teams each have their own
container registry with their Docker files and build off the central IT base image repository.
There are various third-party image scanning tools that can plug into this process on push/pulls from the Azure
Container Registry. Those solutions are available in Azure Marketplace and referenced in the Azure portal
Container Registry blade. For example, Aqua and TwistLock. After the source code is pushed to a git-based
repository, set up CI/CD by creating an Azure DevOps build definition, selecting the source repository, and
choosing the Azure Ser vice Fabric Application and Docker Suppor t template.

The template sets up the build process and tasks for CI/CD by building and containerizing the application, pushing
the container image to a container registry (Azure Container Registry is the default), and deploying the Service
Fabric application with the containerized services to the cluster. Each application code change creates a version of
the code and an updated containerized image. Service Fabric's rolling upgrade feature deploys service upgrades
gracefully.

Here is an example of a build starting the full DevOps process on an Azure-provided hosted build agent. Some
enterprises may require the build agents to run internally within their private Azure virtual network corporate
network. Set up a Windows build agent VM and instruct Azure DevOps to use the private VM for building and
deploying code. For information about using custom build agents, see Self-hosted Windows agents.
Conclusion
Here is the summary of best practices:
Before containerizing existing applications, selecting applications that are suitable for this migration, choose the
right developer workstation, and determine network requirements.
Do not run your application container on the primary node type. Instead, configure the cluster with two or more
node types and run the application tier containers on a non-primary node type. Use placement constraints that
target the non-primary node type to reserve the primary node type for system services.
Use the Open networking mode.
Use an ingress reverse proxy, such as the IIS Application Request Routing. The reverse proxy inspects inbound
traffic and rewrites the traffic to the destination container.
Do not hardcode configuration values in the container image and use environment variables to pass values to a
container.
Monitor the application, platform events, and infrastructure metrics by using IaasDiagnostics and
MicrosoftMonitoringAgent extensions. View the logs in Application Insights and Log Analytics.
Use the latest approved corporate image and provide an automatable image updating process that is consistent
through DevOps.

Next steps
Get the latest version of the tools you need for containerizing, such as Visual Studio and Docker for Windows.
Customize these templates to meet your requirements. Sample: Modernization templates and scripts.
 Migrate an Azure Cloud Services application to
Azure Service Fabric
12/18/2020 • 18 minutes to read • Edit Online

Sample code
This article describes migrating an application from Azure Cloud Services to Azure Service Fabric. It focuses on
architectural decisions and recommended practices.
For this project, we started with a Cloud Services application called Surveys and ported it to Service Fabric. The
goal was to migrate the application with as few changes as possible. Later in the article, we show how to optimize
the application for Service Fabric.
Before reading this article, it will be useful to understand the basics of Service Fabric. See Overview of Azure
Service Fabric

About the Surveys application

A fictional company named Tailspin created an application called the Surveys app that allows customers to create
surveys. After a customer signs up for the application, users can create and publish surveys, and collect the results
for analysis. The application includes a public website where people can take a survey. Read more about the original
Tailspin scenario here.
Now Tailspin wants to move the Surveys application to a microservices architecture, using Service Fabric running
on Azure. Because the application is already deployed as a Cloud Services application, Tailspin adopts a multi-phase
approach:
1. Port the cloud services to Service Fabric, while minimizing changes to the application.
2. Optimize the application for Service Fabric, by moving to a microservices architecture.
In a real-world project, it's likely that both stages would overlap. While porting to Service Fabric, you would also
start to rearchitect the application into micro-services. Later you might refine the architecture further, perhaps
dividing coarse-grained services into smaller services.
The application code is available on GitHub. This repo contains both the Cloud Services application and the Service
Fabric version.

Why Service Fabric?

Service Fabric is a good fit for this project, because most of the features needed in a distributed system are built
into Service Fabric, including:
Cluster management . Service Fabric automatically handles node failover, health monitoring, and other cluster
management functions.
Horizontal scaling . When you add nodes to a Service Fabric cluster, the application automatically scales, as
services are distributed across the new nodes.
Ser vice discover y . Service Fabric provides a discovery service that can resolve the endpoint for a named
service.
Stateless and stateful ser vices . Stateful services use reliable collections, which can take the place of a cache
or queue, and can be partitioned.
 Application lifecycle management . Services can be upgraded independently and without application
downtime.
Ser vice orchestration across a cluster of machines.
Higher density for optimizing resource consumption. A single node can host multiple services.
Service Fabric is used by various Microsoft services, including Azure SQL Database, Cosmos DB, Azure Event Hubs,
and others, making it a proven platform for building distributed cloud applications.

Comparing Cloud Services with Service Fabric

The following table summarizes some of the important differences between Cloud Services and Service Fabric
applications. For a more in-depth discussion, see Learn about the differences between Cloud Services and Service
Fabric before migrating applications.

A REA C LO UD SERVIC ES SERVIC E FA B RIC

Application composition Roles Services

Density One role instance per VM Multiple services in a single node

Minimum number of nodes 2 per role 5 per cluster, for production

deployments

State management Stateless Stateless or stateful*

Hosting Azure Cloud or on-premises

Web hosting IIS** Self-hosting

Deployment model Classic deployment model Resource Manager

Packaging Cloud service package files (.cspkg) Application and service packages

Application update VIP swap or rolling update Rolling update

Autoscaling Built-in service Virtual machine scale sets for auto scale
out

Debugging Local emulator Local cluster

* Stateful services use reliable collections to store state across replicas, so that all reads are local to the nodes in the
cluster. Writes are replicated across nodes for reliability. Stateless services can have external state, using a database
or other external storage.
** Worker roles can also self-host ASP.NET Web API using OWIN.

The Surveys application on Cloud Services

The following diagram shows the architecture of the Surveys application running on Cloud Services.
The application consists of two web roles and a worker role.
The Tailspin.Web web role hosts an ASP.NET website that Tailspin customers use to create and manage
surveys. Customers also use this website to sign up for the application and manage their subscriptions.
Finally, Tailspin administrators can use it to see the list of tenants and manage tenant data.
The Tailspin.Web.Sur vey.Public web role hosts an ASP.NET website where people can take the surveys
that Tailspin customers publish.
The Tailspin.Workers.Sur vey worker role does background processing. The web roles put work items
onto a queue, and the worker role processes the items. Two background tasks are defined: Exporting survey
answers to Azure SQL Database, and calculating statistics for survey answers.
In addition to Cloud Services, the Surveys application uses some other Azure services:
Azure Storage to store surveys, surveys answers, and tenant information.
Azure Cache for Redis to cache some of the data that is stored in Azure Storage, for faster read access.
Azure Active Director y (Azure AD) to authenticate customers and Tailspin administrators.
Azure SQL Database to store the survey answers for analysis.

Moving to Service Fabric

As mentioned, the goal of this phase was migrating to Service Fabric with the minimum necessary changes. To that
end, we created stateless services corresponding to each cloud service role in the original application:
Intentionally, this architecture is similar to the original application. However, the diagram hides some important
differences. In the rest of this article, we'll explore those differences.

Converting the cloud service roles to services

For the initial migration, Tailspin followed the steps outlined in Guide to converting Web and Worker Roles to
Service Fabric stateless services.
Creating the web front-end services
In Service Fabric, a service runs inside a process created by the Service Fabric runtime. For a web front end, that
means the service is not running inside IIS. Instead, the service must host a web server. This approach is called self-
hosting, because the code that runs inside the process acts as the web server host.
The original Surveys application uses ASP.NET MVC. Because ASP.NET MVC cannot be self-hosted in Service Fabric,
Tailspin considered the following migration options:
Port the web roles to ASP.NET Core, which can be self-hosted.
Convert the web site into a single-page application (SPA) that calls a web API implemented using ASP.NET Web
API. This would have required a complete redesign of the web front end.
Keep the existing ASP.NET MVC code and deploy IIS in a Windows Server container to Service Fabric. This
approach would require little or no code change.
The first option, porting to ASP.NET Core, allowed Tailspin to take advantage of the latest features in ASP.NET Core.
To do the conversion, Tailspin followed the steps described in Migrating From ASP.NET MVC to ASP.NET Core MVC.

NOTE
When using ASP.NET Core with Kestrel, you should place a reverse proxy in front of Kestrel to handle traffic from the Internet,
for security reasons. For more information, see Kestrel web server implementation in ASP.NET Core. The section Deploying the
application describes a recommended Azure deployment.

HTTP listeners
In Cloud Services, a web or worker role exposes an HTTP endpoint by declaring it in the service definition file. A
web role must have at least one endpoint.
 <!-- Cloud service endpoint -->
<Endpoints>
<InputEndpoint name="HttpIn" protocol="http" port="80" />
</Endpoints>

Similarly, Service Fabric endpoints are declared in a service manifest:

<!-- Service Fabric endpoint -->

<Endpoints>
<Endpoint Protocol="http" Name="ServiceEndpoint" Type="Input" Port="8002" />
</Endpoints>

Unlike a cloud service role, Service Fabric services can be colocated within the same node. Therefore, every service
must listen on a distinct port. Later in this article, we'll discuss how client requests on port 80 or port 443 get
routed to the correct port for the service.
A service must explicitly create listeners for each endpoint. The reason is that Service Fabric is agnostic about
communication stacks. For more information, see Build a web service front end for your application using ASP.NET
Core.

Packaging and configuration

A cloud service contains the following configuration and package files:

F IL E DESC RIP T IO N

Service definition (.csdef) Settings used by Azure to configure the cloud service. Defines
the roles, endpoints, startup tasks, and the names of
configuration settings.

Service configuration (.cscfg) Per-deployment settings, including the number of role

instances, endpoint port numbers, and the values of
configuration settings.

Service package (.cspkg) Contains the application code and configurations, and the
service definition file.

There is one .csdef file for the entire application. You can have multiple .cscfg files for different environments, such
as local, test, or production. When the service is running, you can update the .cscfg but not the .csdef. For more
information, see What is the Cloud Service model and how do I package it?
Service Fabric has a similar division between a service definition and service settings, but the structure is more
granular. To understand Service Fabric's configuration model, it helps to understand how a Service Fabric
application is packaged. Here is the structure:

Application package
- Service packages
- Code package
- Configuration package
- Data package (optional)

The application package is what you deploy. It contains one or more service packages. A service package contains
code, configuration, and data packages. The code package contains the binaries for the services, and the
configuration package contains configuration settings. This model allows you to upgrade individual services
without redeploying the entire application. It also lets you update just the configuration settings, without
redeploying the code or restarting the service.
A Service Fabric application contains the following configuration files:

F IL E LO C AT IO N DESC RIP T IO N

ApplicationManifest.xml Application package Defines the services that compose the

application.

ServiceManifest.xml Service package Describes one or more services.

Settings.xml Configuration package Contains configuration settings for the

services defined in the service package.

For more information, see Model an application in Service Fabric.

To support different configuration settings for multiple environments, use the following approach, described in
Manage application parameters for multiple environments:
1. Define the setting in the Setting.xml file for the service.
2. In the application manifest, define an override for the setting.
3. Put environment-specific settings into application parameter files.

Deploying the application

Whereas Azure Cloud Services is a managed service, Service Fabric is a runtime. You can create Service Fabric
clusters in many environments, including Azure and on premises. The following diagram shows a recommended
deployment for Azure:

The Service Fabric cluster is deployed to a virtual machine scale set. Scale sets are an Azure Compute resource that
can be used to deploy and manage a set of identical VMs.
As mentioned, it's recommended to place the Kestrel web server behind a reverse proxy for security reasons. This
diagram shows Azure Application Gateway, which is an Azure service that offers various layer 7 load-balancing
capabilities. It acts as a reverse-proxy service, terminating the client connection and forwarding requests to back-
end endpoints. You might use a different reverse proxy solution, such as nginx.
Layer 7 routing
In the original Surveys application, one web role listened on port 80, and the other web role listened on port 443.
 P UB L IC SIT E SURVEY M A N A GEM EN T SIT E

http://tailspin.cloudapp.net https://tailspin.cloudapp.net

Another option is to use layer 7 routing. In this approach, different URL paths get routed to different port numbers
on the back end. For example, the public site might use URL paths starting with /public/ .
Options for layer 7 routing include:
Use Application Gateway.
Use a network virtual appliance (NVA), such as nginx.
Write a custom gateway as a stateless service.
Consider this approach if you have two or more services with public HTTP endpoints, but want them to appear as
one site with a single domain name.

One approach that we don't recommend is allowing external clients to send requests through the Service Fabric
reverse proxy. Although this is possible, the reverse proxy is intended for service-to-service communication.
Opening it to external clients exposes any service running in the cluster that has an HTTP endpoint.

Node types and placement constraints

In the deployment shown above, all the services run on all the nodes. However, you can also group services, so that
certain services run only on particular nodes within the cluster. Reasons to use this approach include:
Run some services on different VM types. For example, some services might be compute-intensive or require
GPUs. You can have a mix of VM types in your Service Fabric cluster.
Isolate front-end services from back-end services, for security reasons. All the front-end services will run on one
set of nodes, and the back-end services will run on different nodes in the same cluster.
Different scale requirements. Some services might need to run on more nodes than other services. For example,
if you define front-end nodes and back-end nodes, each set can be scaled independently.
The following diagram shows a cluster that separates front-end and back-end services:

To implement this approach:

1. When you create the cluster, define two or more node types.
2. For each service, use placement constraints to assign the service to a node type.
When you deploy to Azure, each node type is deployed to a separate virtual machine scale set. The Service Fabric
cluster spans all node types. For more information, see The relationship between Service Fabric node types and
virtual machine scale sets.

If a cluster has multiple node types, one node type is designated as the primary node type. Service Fabric
runtime services, such as the Cluster Management Service, run on the primary node type. Provision at least 5
nodes for the primary node type in a production environment. The other node type should have at least 2
nodes.

Configuring and managing the cluster

Clusters must be secured to prevent unauthorized users from connecting to your cluster. It is recommended to use
Azure AD to authenticate clients, and X.509 certificates for node-to-node security. For more information, see
Service Fabric cluster security scenarios.
To configure a public HTTPS endpoint, see Specify resources in a service manifest.
You can scale out the application by adding VMs to the cluster. Virtual machine scale sets support autoscaling using
autoscale rules based on performance counters. For more information, see Scale a Service Fabric cluster in or out
using autoscale rules.
While the cluster is running, collect logs from all the nodes in a central location. For more information, see Collect
logs by using Azure Diagnostics.

Refactor the application

After the application is ported to Service Fabric, the next step is to refactor it to a more granular architecture.
Tailspin's motivation for refactoring is to make it easier to develop, build, and deploy the Surveys application. By
decomposing the existing web and worker roles to a more granular architecture, Tailspin wants to remove the
existing tightly coupled communication and data dependencies between these roles.
Tailspin sees other benefits in moving the Surveys application to a more granular architecture:
Each service can be packaged into independent projects with a scope small enough to be managed by a small
team.
Each service can be independently versioned and deployed.
Each service can be implemented using the best technology for that service. For example, a service fabric cluster
can include services built using different versions of the .Net Frameworks, Java, or other languages such as C or
C++.
Each service can be independently scaled to respond to increases and decreases in load.
The source code for the refactored version of the app is available on GitHub.

Design considerations
The following diagram shows the architecture of the Surveys application refactored to a more granular
architecture:
Tailspin.Web is a stateless service self-hosting an ASP.NET MVC application that Tailspin customers visit to create
surveys and view survey results. This service shares most of its code with the Tailspin.Web service from the ported
Service Fabric application. As mentioned earlier, this service uses ASP.NET core and switches from using Kestrel as
web frontend to implementing a WebListener.
Tailspin.Web.Sur vey.Public is a stateless service also self-hosting an ASP.NET MVC site. Users visit this site to
select surveys from a list and then fill them out. This service shares most of its code with the
Tailspin.Web.Survey.Public service from the ported Service Fabric application. This service also uses ASP.NET Core
and also switches from using Kestrel as web frontend to implementing a WebListener.
Tailspin.Sur veyResponseSer vice is a stateful service that stores survey answers in Azure Blob Storage. It also
merges answers into the survey analysis data. The service is implemented as a stateful service because it uses a
ReliableConcurrentQueue to process survey answers in batches. This functionality was originally implemented in
the Tailspin.AnswerAnalysisService service in the ported Service Fabric application.
Tailspin.Sur veyManagementSer vice is a stateless service that stores and retrieves surveys and survey
questions. The service uses Azure Blob storage. This functionality was also originally implemented in the data
access components of the Tailspin.Web and Tailspin.Web.Survey.Public services in the ported Service Fabric
application. Tailspin refactored the original functionality into this service to allow it to scale independently.
Tailspin.Sur veyAnswerSer vice is a stateless service that retrieves survey answers and survey analysis. The
service also uses Azure Blob storage. This functionality was also originally implemented in the data access
components of the Tailspin.Web service in the ported Service Fabric application. Tailspin refactored the original
functionality into this service because it expects less load and wants to use fewer instances to conserve resources.
Tailspin.Sur veyAnalysisSer vice is a stateless service that persists survey answer summary data in a Redis cache
for quick retrieval. This service is called by the Tailspin.SurveyResponseService each time a survey is answered and
the new survey answer data is merged in the summary data. This service includes the functionality originally
implemented in the Tailspin.AnswerAnalysisService service from the ported Service Fabric application.

Stateless versus stateful services

Azure Service Fabric supports the following programming models:
The guest executable model allows any executable to be packaged as a service and deployed to a Service Fabric
cluster. Service Fabric orchestrates and manages execution of the guest executable.
The container model allows for deployment of services in container images. Service Fabric supports creation
 and management of containers on top of Linux kernel containers as well as Windows Server containers.
The reliable services programming model allows for the creation of stateless or stateful services that integrate
with all Service Fabric platform features. Stateful services allow for replicated state to be stored in the Service
Fabric cluster. Stateless services do not.
The reliable actors programming model allows for the creation of services that implement the virtual actor
pattern.
All the services in the Surveys application are stateless reliable services, except for the
Tailspin.SurveyResponseService service. This service implements a ReliableConcurrentQueue to process survey
answers when they are received. Responses in the ReliableConcurrentQueue are saved into Azure Blob Storage and
passed to the Tailspin.SurveyAnalysisService for analysis. Tailspin chooses a ReliableConcurrentQueue because
responses do not require strict first-in-first-out (FIFO) ordering provided by a queue such as Azure Service Bus. A
ReliableConcurrentQueue is also designed to deliver high throughput and low latency for queue and dequeue
operations.
Operations to persist dequeued items from a ReliableConcurrentQueue should ideally be idempotent. If an
exception is thrown during the processing of an item from the queue, the same item may be processed more than
once. In the Surveys application, the operation to merge survey answers to the Tailspin.SurveyAnalysisService is
not idempotent because Tailspin decided that the survey analysis data is only a current snapshot of the analysis
data and does not need to be consistent. The survey answers saved to Azure Blob Storage are eventually consistent,
so the survey final analysis can always be recalculated correctly from this data.

Communication framework
Each service in the Surveys application communicates using a RESTful web API. RESTful APIs offer the following
benefits:
Ease of use: each service is built using ASP.NET Core MVC, which natively supports the creation of Web APIs.
Security: While each service does not require SSL, Tailspin could require each service to do so.
Versioning: clients can be written and tested against a specific version of a web API.
Services in the Survey application use the reverse proxy implemented by Service Fabric. Reverse proxy is a service
that runs on each node in the Service Fabric cluster and provides endpoint resolution, automatic retry, and handles
other types of connection failures. To use the reverse proxy, each RESTful API call to a specific service is made using
a predefined reverse proxy port. For example, if the reverse proxy port has been set to 19081 , a call to the
Tailspin.SurveyAnswerService can be made as follows:

static SurveyAnswerService()
{
httpClient = new HttpClient
{
BaseAddress = new Uri("http://localhost:19081/Tailspin/SurveyAnswerService/")
};
}

To enable reverse proxy, specify a reverse proxy port during creation of the Service Fabric cluster. For more
information, see reverse proxy in Azure Service Fabric.

Performance considerations
Tailspin created the ASP.NET Core services for Tailspin.Web and Tailspin.Web.Surveys.Public using Visual Studio
templates. By default, these templates include logging to the console. Logging to the console may be done during
development and debugging, but all logging to the console should be removed when the application is deployed to
production.
 NOTE
For more information about setting up monitoring and diagnostics for Service Fabric applications running in production, see
monitoring and diagnostics for Azure Service Fabric.

For example, the following lines in startup.cs for each of the web front end services should be commented out:

// This method gets called by the runtime. Use this method to configure the HTTP request pipeline.
public void Configure(IApplicationBuilder app, IHostingEnvironment env, ILoggerFactory loggerFactory)
{
//loggerFactory.AddConsole(Configuration.GetSection("Logging"));
//loggerFactory.AddDebug();

app.UseMvc();
}

NOTE
These lines may be conditionally excluded when Visual Studio is set to release when publishing.

Finally, when Tailspin deploys the Tailspin application to production, they switch Visual Studio to release mode.

Deployment considerations
The refactored Surveys application is composed of five stateless services and one stateful service, so cluster
planning is limited to determining the correct VM size and number of nodes. In the applicationmanifest.xml file that
describes the cluster, Tailspin sets the InstanceCount attribute of the StatelessService tag to -1 for each of the
services. A value of -1 directs Service Fabric to create an instance of the service on each node in the cluster.

NOTE
Stateful services require the additional step of planning the correct number of partitions and replicas for their data.

Tailspin deploys the cluster using the Azure portal. The Service Fabric Cluster resource type deploys all of the
necessary infrastructure, including virtual machine scale sets and a load balancer. The recommended VM sizes are
displayed in the Azure portal during the provisioning process for the Service Fabric cluster. Because the VMs are
deployed in a virtual machine scale set, they can be both scaled up and out as user load increases.

Next steps
The Surveys application code is available on GitHub.
If you are just getting started with Azure Service Fabric, first set up your development environment then download
the latest Azure SDK and the Azure Service Fabric SDK. The SDK includes the OneBox cluster manager so you can
deploy and test the Surveys application locally with full F5 debugging.
 Serverless Functions overview
12/18/2020 • 2 minutes to read • Edit Online

Serverless architecture evolves cloud platforms toward pure cloud-native code by abstracting code from the
infrastructure that it needs to run. Azure Functions is a serverless compute option that supports functions, small
pieces of code that do single things.
Benefits of using serverless architectures with Functions applications include:
The Azure infrastructure automatically provides all the updated servers that applications need to keep running
at scale.
Compute resources allocate dynamically, and instantly autoscale to meet elastic demands. Serverless doesn't
mean "no server," but "less server," because servers run only as needed.
Micro-billing saves costs by charging only for the compute resources and duration the code uses to execute.
Function bindings streamline integration by providing declarative access to a wide variety of Azure and third-
party services.
Functions are event-driven. An external event like an HTTP web request, message, schedule, or change in data
triggers the function code. A Functions application doesn't code the trigger, only the response to the trigger. With a
lower barrier to entry, developers can focus on business logic, rather than writing code to handle infrastructure
concerns like messaging.
Azure Functions is a managed service in Azure and Azure Stack. The open source Functions runtime works in many
environments, including Kubernetes, Azure IoT Edge, on-premises, and other clouds.
Serverless and Functions require new ways of thinking and new approaches to building applications. They aren't
the right solutions for every problem. For example serverless Functions scenarios, see Reference architectures.

Implementation steps
Successful implementation of serverless technologies with Azure Functions requires the following actions:
Decide and plan
Architects and technical decision makers (TDMs) perform application assessment, conduct or attend
technical workshops and trainings, run proof of concept (PoC) or pilot projects, and conduct architectural
designs sessions as necessary.
Develop and deploy apps
Developers implement serverless Functions app development patterns and practices, configure DevOps
pipelines, and employ site reliability engineering (SRE) best practices.
Manage operations
IT professionals identify hosting configurations, future-proof scalability by automating infrastructure
provisioning, and maintain availability by planning for business continuity and disaster recovery.
Secure apps
Security professionals handle Azure Functions security essentials, secure the hosting setup, and provide
application security guidance.
Related resources
To learn more about serverless technology, see the Azure serverless documentation.
To learn more about Azure Functions, see the Azure Functions documentation.
For help with choosing a compute technology, see Choose an Azure compute service for your application.
 Serverless Functions reference architectures
12/18/2020 • 4 minutes to read • Edit Online

A reference architecture is a template of required components and the technical requirements to implement them.
A reference architecture isn't custom-built for a customer solution, but is a high-level scenario based on extensive
experience. Before designing a serverless solution, use a reference architecture to visualize an ideal technical
architecture, then blend and integrate it into your environment.

Common serverless architecture patterns

Common serverless architecture patterns include:
Serverless APIs, mobile and web backends.
Event and stream processing, Internet of Things (IoT) data processing, big data and machine learning pipelines.
Integration and enterprise service bus to connect line-of-business systems, publish and subscribe (Pub/Sub) to
business events.
Automation and digital transformation and process automation.
Middleware, software-as-a-Service (SaaS) like Dynamics, and big data projects.

Web application backends

Retail scenario: Pick up online orders from a queue,
process them, and store the resulting data in a database

Mobile application backends

Financial services scenario: Colleagues use mobile banking
to reimburse each other for lunch. Whoever paid for lunch
requests payment through a mobile app, triggering a
notification on colleagues' phones.

IoT-connected backends
Manufacturing scenario: A manufacturing company uses
IoT to monitor its machines. Functions detects anomalous
data and triggers a message to the service department
when repair is required.

Conversational bot processing

Hospitality scenario: Customers ask for available vacation
accommodations on their smartphones. A serverless bot
deciphers requests and returns vacation options.
 Real-time file processing
Healthcare scenario: The solution securely uploads patient
records as PDF files. The solution then decomposes the
data, processes it using OCR detection, and adds it to a
database for easy queries.

Real-time stream processing

Independent software vendor (ISV) scenario: A massive
cloud app collects huge amounts of telemetry data. The
app processes that data in near real-time and stores it in
a database for use in an analytics dashboard.

Scheduled task automation

Financial services scenario: The app analyzes a customer
database for duplicate entries every 15 minutes, to avoid
sending out multiple communications to the same
customers.

Extending SaaS applications

Professional services scenario: A SaaS solution provides
extensibility through webhooks, which Functions can
implement to automate certain workflows.

Featured serverless reference architectures

The following featured serverless reference architectures walk through specific scenarios. See the linked articles
for architectural diagrams and details.
Serverless microservices
The serverless microservices reference architecture walks you through designing, developing, and delivering the
Rideshare application by Relecloud, a fictitious company. You get hands-on instructions for configuring and
deploying all the architectural components, with helpful information about each component.
Serverless web application and event processing with Azure Functions
This two-part solution describes a hypothetical drone delivery system. Drones send in-flight status to the cloud,
which stores these messages for later use. A web application allows users to retrieve the messages to get the latest
device status.
You can download the code for this solution from GitHub.
The article Code walkthrough: Serverless application with Azure Functions walks you through the code and the
design processes.
Event-based cloud automation
Automating workflows and repetitive tasks on the cloud can dramatically improve a DevOps team's productivity. A
serverless model is best suited for event-driven automation scenarios. This event-based automation reference
architecture illustrates two cloud automation scenarios: cost center tagging and throttling response.
Multicloud with Serverless Framework
The Serverless Framework architecture describes how the Microsoft Commercial Software Engineering (CSE) team
partnered with a global retailer to deploy a highly-available serverless solution across both Azure and Amazon
Web Services (AWS) cloud platforms, using the Serverless Framework.

More serverless Functions reference architectures

The following sections list other serverless and Azure Functions-related reference architectures and scenarios.
General
Serverless application architectures using Event Grid
Serverless apps using Cosmos DB
Serverless event processing using Azure Functions
Serverless web application on Azure
Serverless Asynchronous Multiplayer Reference Architecture
Instant Broadcasting on Serverless Architecture
Building a telehealth system on Azure
Custom Data Sovereignty & Data Gravity Requirements
Sharing location in real time using low-cost serverless Azure services
Web and mobile backend
An e-commerce front end
Architect scalable e-commerce web app
Improve scalability in an Azure web application
Uploading and CDN-preloading static content with Azure Functions
Cross Cloud Scaling Architecture
Social App for Mobile and Web with Authentication
AI + Machine Learning
Image classification for insurance claims
Personalized Offers
Personalized marketing solutions
Speech transcription with Azure Cognitive Services
Training a Model with AzureML and Azure Functions
Customer Reviews App with Cognitive Services
Enterprise-grade conversational bot
AI at the Edge
Mass ingestion and analysis of news feeds on Azure
HIPPA and HITRUST compliant health data AI
Intelligent Experiences On Containers (AKS, Functions, Keda)
Data and analytics
Application integration using Event Grid
Mass ingestion and analysis of news feeds
Tier Applications & Data for Analytics
Operational analysis and driving process efficiency
IoT
Azure IoT reference (SQL DB)
Azure IoT reference (Cosmos DB)
 IoT using Cosmos DB
Facilities management powered by mixed reality and IoT
Complementary Code Pattern for Azure IoT Edge Modules & Cloud Applications
Gaming
Custom Game Server Scaling
Non-real Time Dashboard
In-editor Debugging Telemetry
Multiplayer Serverless Matchmaker
Advanced leaderboard for large scale
Relational Leaderboard
Content Moderation
Text Translation
Text to Speech
Gaming using Cosmos DB
Automation
Smart scaling for Azure Scale Set with Azure Functions
 Plan for serverless Functions
12/18/2020 • 2 minutes to read • Edit Online

To plan for moving an application to a serverless Azure Functions architecture, a technical decision maker (TDM) or
architect:
Verifies the application's characteristics and business requirements.
Determines the application's suitability for serverless Azure Functions.
Transforms business requirements into functional and other requirements.
Planning activities also include assessing technical team readiness, providing or attending workshops and training,
and conducting architectural design reviews, proofs of concept, pilots, and technical implementations.
TDMs and architects may perform one or more of the following activities:
Execute an application assessment. Evaluate the main aspects of the application to determine how
complex and risky it is to rearchitect through application modernization, or rebuild a new cloud-native
application. See Application assessment.
Attend or promote technical workshops and training. Host a Serverless workshop or CloudHack, or
enjoy many other training and learning opportunities for serverless technologies, Azure Functions, app
modernization, and cloud-native apps. See Technical workshops and training.
Identify and execute a Proof of Concept (PoC) or pilot, or technical implementation. Deliver a
PoC, pilot, or technical implementation to provide evidence that serverless Azure Functions can solve a
team's business problems. Showing teams how to modernize or build new cloud-native applications to their
specifications can accelerate deployment to production. See PoC or pilot.
Conduct architectural design sessions. An architectural design session (ADS) is an in-depth discussion
on how a new solution will blend into the environment. ADSs validate business requirements and transform
them to functional and other requirements.

Next steps
For example scenarios that use serverless architectures with Azure Functions, see Serverless reference
architectures.
To move forward with serverless Azure Functions implementation, see the following resources:
Application development and deployment
Azure Functions app operations
Azure Functions app security
 Application assessment
12/18/2020 • 6 minutes to read • Edit Online

Cloud rationalization is the process of evaluating applications to determine the best way to migrate or modernize
them for the cloud.
Rationalization methods include:
Rehost . Also known as a lift and shift migration, rehost moves a current application to the cloud with minimal
change.
Refactor . Slightly refactoring an application to fit platform-as-a-service (PaaS)-based options can reduce
operational costs.
Rearchitect . Rearchitect aging applications that aren't compatible with cloud components, or cloud-compatible
applications that would realize cost and operational efficiencies by rearchitecting into a cloud-native solution.
Rebuild . If the changes or costs to carry an application forward are too great, consider creating a new cloud-
native code base. Rebuild is especially appropriate for applications that previously met business needs, but are
now unsupported or misaligned with current business processes.
Before you decide on an appropriate strategy, analyze the current application to determine the risk and complexity
of each method. Consider application lifecycle, technology, infrastructure, performance, and operations and
monitoring. For multitier architectures, evaluate the presentation tier, service tier, integrations tier, and data tier.
The following checklists evaluate an application to determine the complexity and risk of rearchitecting or
rebuilding.

Complexity and risk

Each of the following factors adds to complexity, risk, or both.
Architecture
Define the high-level architecture, such as web application, web services, data storage, or caching.

FA C TO R C O M P L EXIT Y RISK

Application components don't translate ✔ ✔

directly to Azure.

The application needs code changes to ✔ ✔

run in Azure.

The application needs major, complex ✔ ✔

code changes to run in Azure.

Business drivers
Older applications might require extensive changes to get to the cloud.

FA C TO R C O M P L EXIT Y RISK

This application has been around for ✔

more than three years.
 FA C TO R C O M P L EXIT Y RISK

This application is business critical. ✔

There are technology blockers for ✔

migration.

There are business blockers for ✔

migration.

This application has compliance ✔

requirements.

The application is subject to country- ✔

specific data requirements.

The application is publicly accessible. ✔ ✔

Technology
FA C TO R C O M P L EXIT Y RISK

This is not a web-based application, and ✔

isn't hosted on a web server.

The app isn't hosted in Windows IIS ✔

The app isn't hosted on Linux ✔

The application is hosted in a web farm, ✔ ✔

and requires multiple servers to host
the web components.

The application requires third-party ✔ ✔

software to be installed on the servers.

The application is hosted in a single ✔

datacenter, and operations are
performed in a single location.

The application accesses the server's ✔

registry.

The application sends emails, and needs ✔

access to an SMTP server.

This isn't a .NET application. ✔

The application uses SQL Server as its ✔

data store.

The application stores data on local ✔

disks, and needs access to the disks to
operate properly.
 FA C TO R C O M P L EXIT Y RISK

The application uses Windows Services ✔

to process asynchronous operations, or
needs external services to process data
or operations.

Deployment
When assessing deployment requirements, consider:
Number of daily users
Average number of concurrent users
Expected traffic
Bandwidth in Gbps
Requests per second
Amount of memory needed
You can reduce deployment risk by storing code under source control in a version control system such as Git,
Azure DevOps Server, or SVN.

FA C TO R C O M P L EXIT Y RISK

Leveraging existing code and data is a ✔ ✔

#1 priority.

The application code isn't under source ✔

control.

There's no automated build process like ✔

Azure DevOps Server or Jenkins.

There's no automated release process ✔ ✔

to deploy the application.

The application has a Service Level ✔

Agreement (SLA) that dictates the
amount of expected downtime.

The application experiences peak or ✔

variable usage times or loads.

The web application saves its session ✔

state in process, rather than an external
data store.

Operations
FA C TO R C O M P L EXIT Y RISK

The application doesn't have a well- ✔

established instrumentation strategy or
standard instrumentation framework.
 FA C TO R C O M P L EXIT Y RISK

The application doesn't use monitoring ✔

tools, and the operations team doesn't
monitor the app's performance.

The application has measured SLA in ✔

place, and the operations team
monitors the application's performance.

The application writes to a log store, ✔

event log, log file, log database, or
Application Insights.

The application doesn't write to a log ✔

store, event log, log file, log database,
or Application Insights.

The application isn't part of the ✔

organization's disaster recovery plan.

Security
FA C TO R C O M P L EXIT Y RISK

The application uses Active Directory to ✔ ✔

authenticate users.

The organization hasn't yet configured ✔

Azure Active Directory (Azure AD), or
hasn't configured Azure AD Connect to
synchronize on-premises AD with Azure
AD.

The application requires access to on- ✔

premises resources, which will require
VPN connectivity from Azure.

The organization hasn't yet configured ✔ ✔

a VPN connection between Azure and
their on-premises environment.

The application requires a SSL certificate ✔ ✔

to run.

Results
Count your application's Complexity and Risk checkmarks.
The expected level of complexity to migrate or modernize the application to Azure is: Total Complexity/25 .
The expected risk involved is: Total Risk/19 .
For both complexity and risk, a score of <0.3 = low, <0.7 = medium, >0.7 = high.

Refactor, rearchitect, or rebuild

To rationalize whether to rehost, refactor, rearchitect, or rebuild your application, consider the following points.
Many of these factors also contribute to complexity and risk.
Determine whether the application components can translate directly to Azure. If so, you don't need code changes
to move the application to Azure, and could use rehost or refactor strategies. If not, you need to rewrite code, so
you need to rearchitect or rebuild.
If the app does need code changes, determine the complexity and extent of the needed changes. Minor changes
might allow for rearchitecting, while major changes may require rebuilding.
Rehost or refactor
If leveraging existing code and data is a top priority, consider a refactor strategy rather than rearchitecting
or rebuilding.
If you have pressing timelines like datacenter shutdown or contract expiration, end-of-life licensing, or
mergers or acquisitions, the fastest way to get the application to Azure might be to rehost, followed by
refactoring to take advantage of cloud capabilities.
Rearchitect or rebuild
If there are applications serving similar needs in your portfolio, this might be an opportunity to rearchitect
or rebuild the entire solution.
If you want to implement multi-tier or microservices architecture for a monolithic app, you must rearchitect
or rebuild the app. If you don't mind retaining the monolithic structure, you might be able to rehost or
refactor.
Rearchitect or rebuild the app to take advantage of cloud capabilities if you plan to update the app more
often than yearly, if the app has peak or variable usage times, or if you expect the app to handle high traffic.
To decide between rearchitecting or rebuilding, assess the following factors. The largest scoring result indicates
your best strategy.

FA C TO R REA RC H IT EC T REB UIL D

There are other applications serving ✔ ✔

similar needs in your portfolio.

The application needs minor code ✔

changes to run in Azure.

The application needs major, complex ✔

code changes to run in Azure.

It's important to leverage existing code. ✔

You want to move a monolithic ✔

application to multi-tier architecture.

You want to move a monolithic ✔ ✔

application to a microservices
architecture.

You expect this app to add ✔

breakthrough capabilities like AI, IoT, or
bots.

Among functionality, cost, ✔

infrastructure, and processes,
functionality is the least efficient aspect
of this application.
FA C TO R REA RC H IT EC T REB UIL D

The application requires third-party ✔

software installed on the servers.

The application accesses the server's ✔

registry.

The application sends emails and needs ✔

access to an SMTP server.

The application uses SQL Server as its ✔

data store.

The application stores data on local ✔

disks, and needs access to the disks to
run properly.

The application uses Windows services ✔

to process asynchronous operations, or
needs external services to process data
or operations.

A web application saves its session ✔

state in process, rather than to an
external data store.

The app has peak and variable usage ✔ ✔

times and loads.

You expect the application to handle ✔ ✔

high traffic.
 Technical workshops and training
12/18/2020 • 3 minutes to read • Edit Online

The workshops, classes, and learning materials in this article provide technical training for serverless architectures
with Azure Functions. These resources help you and your team or customers understand and implement
application modernization and cloud-native apps.

Technical workshops
The Microsoft Cloud Workshop (MCW) program provides workshops you can host to foster cloud learning and
adoption. Each workshop includes presentation decks, trainer and student guides, and hands-on lab guides.
Contribute your own content and feedback to add to a robust database of training guides for deploying advanced
Azure workloads on the Microsoft Cloud Platform.
Workshops related to application development workloads include:
Serverless APIs in Azure. Set of entry-level exercises, which cover the basics of building and managing
serverless APIs in Microsoft Azure - with Azure Functions, Azure API Management, and Azure Application
Insights.
Serverless architecture. Implement a series of Azure Functions that independently scale and break down
business logic to discrete components, allowing customers to pay only for the services they use.
App modernization. Design a modernization plan to move services from on-premises to the cloud by
leveraging cloud, web, and mobile services, secured by Azure Active Directory.
Modern cloud apps. Deploy, configure, and implement an end-to-end secure and Payment Card Industry (PCI)
compliant solution for e-commerce, based on Azure App Services, Azure Active Directory, and Azure DevOps.
Cloud-native applications. Using DevOps best practices, build a proof of concept (PoC) to transform a platform-
as-a-service (PaaS) application to a container-based application with multi-tenant web app hosting.
Continuous delivery in Azure DevOps. Set up and configure continuous delivery (CD) in Azure to reduce
manual errors, using Azure Resource Manager templates, Azure DevOps, and Git repositories for source
control.

Instructor-led training
Course AZ-204: Developing solutions for Microsoft Azure teaches developers how to create end-to-end solutions
in Microsoft Azure. Students learn how to implement Azure compute solutions, create Azure Functions, implement
and manage web apps, develop solutions utilizing Azure Storage, implement authentication and authorization, and
secure their solutions by using Azure Key Vault and managed identities. Students also learn to connect to and
consume Azure and third-party services, and include event- and message-based models in their solutions. The
course also covers monitoring, troubleshooting, and optimizing Azure solutions.

Serverless OpenHack
The Serverless OpenHack simulates a real-world scenario where a company wants to utilize serverless services to
build and release an API to integrate into their distributor's application. This OpenHack lets attendees quickly build
and deploy Azure serverless solutions with cutting-edge compute services like Azure Functions, Logic Apps, Event
Grid, Service Bus, Event Hubs, and Cosmos DB. The OpenHack also covers related technologies like API
Management, Azure DevOps or GitHub, Application Insights, Dynamics 365/Microsoft 365, and Cognitive APIs.
OpenHack attendees build serverless functions, web APIs, and a CI/CD pipeline to support them. They implement
further serverless technologies to integrate line of business (LOB) app workflows, process user and data telemetry,
and create key progress indicator (KPI)-aligned reports. By the end of the OpenHack, attendees have built out a full
serverless technical solution that can create workflows between systems and handle events, files, and data
ingestion.
Microsoft customer projects inspired these OpenHack challenges:
Configure the developer environment.
Create your first serverless function and workflow.
Build APIs to support business needs.
Deploy a management layer for APIs and monitoring usage.
Build a LOB workflow process.
Process large amounts of unstructured file data.
Process large amounts of incoming event data.
Implement a publisher/subscriber messaging pattern and virtual network integration.
Conduct sentiment analysis.
Perform data aggregation, analysis, and reporting.
To attend an OpenHack, register at https://openhack.microsoft.com. For enterprises with many engineers,
Microsoft can request and organize a dedicated Serverless OpenHack.

Microsoft Learn
Microsoft Learn is a free, online training platform that provides interactive learning for Microsoft products. The
goal is to improve proficiency with fun, guided, hands-on content that's specific to your role and goals. Learning
paths are collections of modules that are organized around specific roles like developer, architect, or system admin,
or technologies like Azure Web Apps, Azure Functions, or Azure SQL DB. Learning paths provide understanding of
different aspects of the technology or role.
Learning paths about serverless apps and Azure Functions include:
Create serverless applications. Learn how to leverage functions to execute server-side logic and build serverless
architectures.
Architect message brokering and serverless applications in Azure. Learn how to create reliable messaging for
your applications, and how to take advantage of serverless application services in Azure.
Search all Functions-related learning paths.

Hands-on labs and how-to guides

Build a serverless web app, from the Build 2018 conference
Build a Serverless IoT Solution with Python Azure Functions and SignalR

Next steps
Execute an application assessment
Identify and execute a PoC or Pilot project
 Proof of Concept or pilot
12/18/2020 • 4 minutes to read • Edit Online

When driving a technical and security decision for your company or customer, a Proof of Concept (PoC) or pilot is
an opportunity to deliver evidence that the proposed solution solves the business problems. The PoC or pilot
increases the likelihood of a successful adoption.
A PoC:
Demonstrates that a business model or idea is feasible and will work to solve the business problem
Usually involves one to three features or capabilities
Can be in one or multiple technologies
Is usually geared toward a particular scenario, and proves what the customer needs to know to make the
technical or security decision
Is used only as a demonstration and won't go into production
Is IT-driven and enablement-driven
A pilot:
Is a test run or trial of a proposed action or product
Lasts longer than a PoC, often weeks or months
Has a higher return on investment (ROI) than a PoC
Builds in a pre-production or trial environment, with the intent that it will then go into production
Is adoption-driven and consumption-driven

PoC and pilot best practices

Be aware of compliance issues when working in a customer's environment, and make sure your actions are always
legal and compliant.
Touching or altering the customer's environment usually requires a contract, and may involve a partner or
Microsoft services. Without a contract, your company may be liable for issues or damages.
Governance may require Legal department approval. Your company may not be able to give intellectual
property (IP) away for free. You may need a legal contract or contracts to specify whether your company or the
customer pays for the IP.
Get disclosure guidance when dealing with non-disclosure agreements (NDAs), product roadmaps, NDA
features, or anything not released to the general public.
In a pilot, don't use a trial Microsoft Developers Network (MSDN) environment, or any environment that you
own.
Use properly-licensed software, and ask the opportunity owner to make sure to handle software licensing
correctly.
The customer, partner, or your company may pay for the PoC or pilot. Depending on the size of the contract, the
ROI, and the cost of sale, one group may cover it all, or a combination of all three parties may cover the cost.
Ensure that your company or customer has some investment in the PoC or pilot. If they don't, this can be a red flag
signaling that your company or customer doesn't yet see value in the solution.

PoC and pilot process

The Technical Decision Maker (TDM) is responsible for driving an adoption decision. The TDM is responsible for
ensuring that the right partners and resources are involved in a PoC or pilot. As a TDM, make sure you're aware of
the partners in your product and service area or region. Be aware of their key service offerings around your
product service area.
Planning
Consider the following health questions:
Do you have a good technical plan, including key decision makers and Microsoft potential?
Can you deliver the needed assurance without a PoC?
Should you switch to a pilot?
What are the detailed scope and decision criteria your team or customer agreed to?
If you meet the criteria, will your company or customer buy or deploy the solution?
Do the following tasks:
Analyze risk.
Evaluate the setting.
Do the preparation.
Consider workloads and human resources.
Present PoC or pilot health status.
Fulfill technical prerequisites.
Define the go/no go decision.
Create a final project plan specification.
Execution
For the execution phase:
Determine who kicks off the presentation.
Schedule the meeting in the morning, if possible.
Prepare the demos and slides.
Conduct a dry run to refine the presentation.
Get feedback.
Involve your company or customer team.
Complete the win/lose statement.
Debriefing
During the debriefing phase, consider:
Whether criteria were met or not met
Stakeholder investment
Initiating deployment
Finding a partner and training
Lessons learned
Corrections or extensions of PoC or pilot guidance
Archiving of valuable deliverables

Change management
Change management uses tested methods and techniques to avoid errors and minimize impact when
administering change.
Ideally, a pilot includes a cross-section of users, to address any potential issues or problems that arise. Users may
be comfortable and familiar with their old technology, and have difficulty moving into new technical solutions.
Change management keeps this in mind, and helps the user understand the reasons behind the change and the
impact the change will make.
This understanding is part of a pilot, and addresses everyone who has a stake in the project. A pilot is better than a
PoC, because the customer is more involved, so they're more likely to implement the change.
The pilot includes a detailed follow up through surveys or focus groups. The feedback can prove and improve the
change.

Next steps
Execute an application assessment
Promote a technical workshop or training
Code a technical implementation with the team or customer

Related resources
Prosci® change management training
 Application development and deployment
12/18/2020 • 4 minutes to read • Edit Online

To develop and deploy serverless applications with Azure Functions, examine patterns and practices, configure
DevOps pipelines, and implement site reliability engineering (SRE) best practices.
For detailed information about serverless architectures and Azure Functions, see:
Serverless apps: Architecture, patterns, and Azure implementation
Azure Serverless Computing Cookbook
Example serverless reference architectures

Planning
To plan app development and deployment:
1. Prepare development environment and set up workflow.
2. Structure projects to support Azure Functions app development.
3. Identify app triggers, bindings, and configuration requirements.
Understand event-driven architecture
A different event triggers every function in a serverless Functions project. For more information about event-
driven architectures, see:
Event-driven architecture style.
Event-driven design patterns to enhance existing applications using Azure Functions
Prepare development environment
Set up your development workflow and environment with the tools to create Functions. For details about
development tools and Functions code project structure, see:
Code and test Azure Functions locally
Develop Azure Functions by using Visual Studio Code
Develop Azure Functions using Visual Studio
Work with Azure Functions Core Tools
Folder structure

Development
Decide on the development language to use. Azure Functions supports C#, F#, PowerShell, JavaScript, TypeScript,
Java, and Python. All of a project's Functions must be in the same language. For more information, see Supported
languages in Azure Functions.
Define triggers and bindings
A trigger invokes a Function, and every Function must have exactly one trigger. Binding to a Function declaratively
connects another resource to the Function. For more information about Functions triggers and bindings, see:
Azure Functions triggers and bindings concepts
Execute an Azure Function with triggers
Chain Azure Functions together using input and output bindings
Create the Functions application
Functions follow the single responsibility principle: do only one thing. For more information about Functions
development, see:
Azure Functions developers guide
Create serverless applications
Strategies for testing your code in Azure Functions
Functions best practices
Use Durable Functions for stateful workflows
Durable Functions in Azure Functions let you define stateful workflows in a serverless environment by writing
orchestrator functions, and stateful entities by writing entity functions. Durable Functions manage state,
checkpoints, and restarts, allowing you to focus on business logic. For more information, see What are Durable
Functions.
Understand and address cold starts
If the number of serverless host instances scales down to zero, the next request has the added latency of restarting
the Function app, called a cold start. To minimize the performance impact of cold starts, reduce dependencies that
the Functions app needs to load on startup, and use as few large, synchronous calls and operations as possible.
For more information about autoscaling and cold starts, see Serverless Functions operations.
Manage application secrets
For security, don't store credentials in application code. To use Azure Key Vault with Azure Functions to store and
retrieve keys and credentials, see Use Key Vault references for App Service and Azure Functions.
For more information about serverless Functions application security, see Serverless Functions security.

Deployment
To prepare serverless Functions application for production, make sure you can:
Fulfill application resource requirements.
Monitor all aspects of the application.
Diagnose and troubleshoot application issues.
Deploy new application versions without affecting production systems.
Define deployment technology
Decide on deployment technology, and organize scheduled releases. For more information about how Functions
app deployment enables reliable, zero-downtime upgrades, see Deployment technologies in Azure Functions.
Avoid using too many resource connections
Functions in a Functions app share resources, including connections to HTTPS, databases, and services such as
Azure Storage. When many Functions are running concurrently, it's possible to run out of available connections.
For more information, see Manage connections in Azure Functions.
Configure logging, alerting, and application monitoring
Application Insights in Azure Monitor collects log, performance, and error data. Application Insights automatically
detects performance anomalies, and includes powerful analytics tools to help diagnose issues and understand
function usage.
For more information about application monitoring and logging, see:
Monitor Azure Functions
Monitoring Azure Functions with Azure Monitor Logs
Application Insights for Azure Functions supported features
Diagnose and troubleshoot issues
Learn how to effectively use diagnostics for troubleshooting in proactive and problem-first scenarios. For more
information, see:
Keep your Azure App Service and Azure Functions apps healthy and happy
Troubleshoot error: "Azure Functions Runtime is unreachable"
Deploy applications using an automated pipeline and DevOps
Full automation of all steps from code commit to production deployment lets teams focus on building code, and
removes the overhead and potential human error of manual steps. Deploying new code is quicker and less risky,
helping teams become more agile, more productive, and more confident about their code.
For more information about DevOps and continuous deployment (CD), see:
Continuous deployment for Azure Functions
Continuous delivery by using Azure DevOps
Continuous delivery by using GitHub Action

Optimization
Once the application is in production, prepare for scaling and implement site reliability engineering (SRE).
Ensure optimal scalability
For information about factors that impact Functions app scalability, see:
Scalability best practices
Performance and scale in Durable Functions
Implement SRE practices
Site Reliability Engineering (SRE) is a proven approach to maintaining crucial system and application reliability,
while iterating at the speed the marketplace demands. For more information, see:
Introduction to Site Reliability Engineering (SRE)
DevOps at Microsoft: Game streaming SRE

Next steps
For hands-on serverless Functions app development and deployment walkthroughs, see:
Serverless Functions code walkthrough
CI/CD for a serverless frontend
For an engineering playbook to help teams and customers successfully implement serverless Functions projects,
see the Code-With Customer/Partner Engineering Playbook.
 Code walkthrough: Serverless application with
Functions
12/18/2020 • 18 minutes to read • Edit Online

Serverless models abstract code from the underlying compute infrastructure, allowing developers to focus on
business logic without extensive setup. Serverless code reduces costs, because you pay only for the code execution
resources and duration.
The serverless event-driven model fits situations where a certain event triggers a defined action. For example,
receiving an incoming device message triggers storage for later use, or a database update triggers some further
processing.
To help you explore Azure serverless technologies in Azure, Microsoft developed and tested a serverless
application that uses Azure Functions. This article walks through the code for the serverless Functions solution, and
describes design decisions, implementation details, and some of the "gotchas" you might encounter.

Explore the solution

The two-part solution describes a hypothetical drone delivery system. Drones send in-flight status to the cloud,
which stores these messages for later use. A web app lets users retrieve the messages to get the latest status of the
devices.
You can download the code for this solution from GitHub.
This walkthrough assumes basic familiarity with the following technologies:
Azure Functions
Azure Event Hubs
.NET Core
You don't need to be an expert in Functions or Event Hubs, but you should understand their features at a high level.
Here are some good resources to get started:
An introduction to Azure Functions
Features and terminology in Azure Event Hubs

Understand the scenario

Fabrikam manages a fleet of drones for a drone delivery service. The application consists of two main functional
areas:
Event ingestion . During flight, drones send status messages to a cloud endpoint. The application ingests
and processes these messages, and writes the results to a back-end database (Cosmos DB). The devices
send messages in protocol buffer (protobuf) format. Protobuf is an efficient, self-describing serialization
format.
 These messages contain partial updates. At a fixed interval, each drone sends a "key frame" message that
contains all of the status fields. Between key frames, the status messages only include fields that changed
since the last message. This behavior is typical of many IoT devices that need to conserve bandwidth and
power.
Web app . A web application allows users to look up a device and query the device's last-known status.
Users must sign into the application and authenticate with Azure Active Directory (Azure AD). The
application only allows requests from users who are authorized to access the app.
Here's a screenshot of the web app, showing the result of a query:

Design the application

Fabrikam has decided to use Azure Functions to implement the application business logic. Azure Functions is an
example of "Functions as a Service" (FaaS). In this computing model, a function is a piece of code that is deployed
to the cloud and runs in a hosting environment. This hosting environment completely abstracts the servers that
run the code.
Why choose a serverless approach?
A serverless architecture with Functions is an example of an event-driven architecture. The function code is a
triggered by some event that's external to the function — in this case, either a message from a drone, or an HTTP
request from a client application. With a function app, you don't need to write any code for the trigger. You only
write the code that runs in response to the trigger. That means you can focus on your business logic, rather than
writing a lot of code to handle infrastructure concerns like messaging.
There are also some operational advantages to using a serverless architecture:
There is no need to manage servers.
Compute resources are allocated dynamically as needed.
You are charged only for the compute resources used to execute your code.
The compute resources scale on demand based on traffic.
Architecture
The following diagram shows the high-level architecture of the application:

In one data flow, arrows show messages going from Devices to Event Hubs and triggering the function app. From
the app, one arrow shows dead-letter messages going to a storage queue, and another arrow shows writing to
Azure Cosmos DB. In another data flow, arrows show the client web app getting static files from Blob storage static
web hosting, through a CDN. Another arrow shows the client HTTP request going through API Management. From
API Management, one arrow shows the function app triggering and reading data from Azure Cosmos DB. Another
arrow shows authentication through Azure AD. A User also signs in to Azure AD.
Event ingestion:
1. Drone messages are ingested by Azure Event Hubs.
2. Event Hubs produces a stream of events that contain the message data.
3. These events trigger an Azure Functions app to process them.
4. The results are stored in Cosmos DB.
Web app:
1. Static files are served by CDN from Blob storage.
2. A user signs into the web app using Azure AD.
3. Azure API Management acts as a gateway that exposes a REST API endpoint.
4. HTTP requests from the client trigger an Azure Functions app that reads from Cosmos DB and returns the
result.
This application is based on two reference architectures, corresponding to the two functional blocks described
above:
Serverless event processing using Azure Functions
Serverless web application on Azure
You can read those articles to learn more about the high-level architecture, the Azure services that are used in the
solution, and considerations for scalability, security, and reliability.

Drone telemetry function

Let's start by looking at the function that processes drone messages from Event Hubs. The function is defined in a
class named RawTelemetryFunction :
 namespace DroneTelemetryFunctionApp
{
public class RawTelemetryFunction
{
private readonly ITelemetryProcessor telemetryProcessor;
private readonly IStateChangeProcessor stateChangeProcessor;
private readonly TelemetryClient telemetryClient;

public RawTelemetryFunction(ITelemetryProcessor telemetryProcessor, IStateChangeProcessor

stateChangeProcessor, TelemetryClient telemetryClient)
{
this.telemetryProcessor = telemetryProcessor;
this.stateChangeProcessor = stateChangeProcessor;
this.telemetryClient = telemetryClient;
}
}
...
}

This class has several dependencies, which are injected into the constructor using dependency injection:
The ITelemetryProcessor and IStateChangeProcessor interfaces define two helper objects. As we'll see, these
objects do most of the work.
The TelemetryClient is part of the Application Insights SDK. It is used to send custom application metrics to
Application Insights.
Later, we'll look at how to configure the dependency injection. For now, just assume these dependencies exist.

Configure the Event Hubs trigger

The logic in the function is implemented as an asynchronous method named RunAsync . Here is the method
signature:

[FunctionName("RawTelemetryFunction")]
[StorageAccount("DeadLetterStorage")]
public async Task RunAsync(
[EventHubTrigger("%EventHubName%", Connection = "EventHubConnection", ConsumerGroup
="%EventHubConsumerGroup%")]EventData[] messages,
[Queue("deadletterqueue")] IAsyncCollector<DeadLetterMessage> deadLetterMessages,
ILogger logger)
{
// implementation goes here
}

The method takes the following parameters:

messages is an array of event hub messages.
deadLetterMessages is an Azure Storage Queue, used for storing dead letter messages.
logging provides a logging interface, for writing application logs. These logs are sent to Azure Monitor.

The EventHubTrigger attribute on the messages parameter configures the trigger. The properties of the attribute
specify an event hub name, a connection string, and a consumer group. (A consumer group is an isolated view of
the Event Hubs event stream. This abstraction allows for multiple consumers of the same event hub.)
Notice the percent signs (%) in some of the attribute properties. These indicate that the property specifies the
name of an app setting, and the actual value is taken from that app setting at run time. Otherwise, without percent
signs, the property gives the literal value.
The Connection property is an exception. This property always specifies an app setting name, never a literal value,
so the percent sign is not needed. The reason for this distinction is that a connection string is secret and should
never be checked into source code.
While the other two properties (event hub name and consumer group) are not sensitive data like a connection
string, it's still better to put them into app settings, rather than hard coding. That way, they can be updated without
recompiling the app.
For more information about configuring this trigger, see Azure Event Hubs bindings for Azure Functions.

Message processing logic

Here's the implementation of the RawTelemetryFunction.RunAsync method that processes a batch of messages:

[FunctionName("RawTelemetryFunction")]
[StorageAccount("DeadLetterStorage")]
public async Task RunAsync(
[EventHubTrigger("%EventHubName%", Connection = "EventHubConnection", ConsumerGroup
="%EventHubConsumerGroup%")]EventData[] messages,
[Queue("deadletterqueue")] IAsyncCollector<DeadLetterMessage> deadLetterMessages,
ILogger logger)
{
telemetryClient.GetMetric("EventHubMessageBatchSize").TrackValue(messages.Length);

foreach (var message in messages)

{
DeviceState deviceState = null;

try
{
deviceState = telemetryProcessor.Deserialize(message.Body.Array, logger);

try
{
await stateChangeProcessor.UpdateState(deviceState, logger);
}
catch (Exception ex)
{
logger.LogError(ex, "Error updating status document", deviceState);
await deadLetterMessages.AddAsync(new DeadLetterMessage { Exception = ex, EventData = message,
DeviceState = deviceState });
}
}
catch (Exception ex)
{
logger.LogError(ex, "Error deserializing message", message.SystemProperties.PartitionKey,
message.SystemProperties.SequenceNumber);
await deadLetterMessages.AddAsync(new DeadLetterMessage { Exception = ex, EventData = message });
}
}
}

When the function is invoked, the messages parameter contains an array of messages from the event hub.
Processing messages in batches will generally yield better performance than reading one message at a time.
However, you have to make sure the function is resilient and handles failures and exceptions gracefully. Otherwise,
if the function throws an unhandled exception in the middle of a batch, you might lose the remaining messages.
This consideration is discussed in more detail in the section Error handling.
But if you ignore the exception handling, the processing logic for each message is simple:
1. Call ITelemetryProcessor.Deserialize to deserialize the message that contains a device state change.
2. Call IStateChangeProcessor.UpdateState to process the state change.
Let's look at these two methods in more detail, starting with the Deserialize method.
Deserialize method
The TelemetryProcess.Deserialize method takes a byte array that contains the message payload. It deserializes this
payload and returns a DeviceState object, which represents the state of a drone. The state may represent a partial
update, containing just the delta from the last-known state. Therefore, the method needs to handle null fields in
the deserialized payload.

public class TelemetryProcessor : ITelemetryProcessor

{
private readonly ITelemetrySerializer<DroneState> serializer;

public TelemetryProcessor(ITelemetrySerializer<DroneState> serializer)

{
this.serializer = serializer;
}

public DeviceState Deserialize(byte[] payload, ILogger log)

{
DroneState restored = serializer.Deserialize(payload);

log.LogInformation("Deserialize message for device ID {DeviceId}", restored.DeviceId);

var deviceState = new DeviceState();

deviceState.DeviceId = restored.DeviceId;

if (restored.Battery != null)
{
deviceState.Battery = restored.Battery;
}
if (restored.FlightMode != null)
{
deviceState.FlightMode = (int)restored.FlightMode;
}
if (restored.Position != null)
{
deviceState.Latitude = restored.Position.Value.Latitude;
deviceState.Longitude = restored.Position.Value.Longitude;
deviceState.Altitude = restored.Position.Value.Altitude;
}
if (restored.Health != null)
{
deviceState.AccelerometerOK = restored.Health.Value.AccelerometerOK;
deviceState.GyrometerOK = restored.Health.Value.GyrometerOK;
deviceState.MagnetometerOK = restored.Health.Value.MagnetometerOK;
}
return deviceState;
}
}

This method uses another helper interface, ITelemetrySerializer<T> , to deserialize the raw message. The results
are then transformed into a POCO model that is easier to work with. This design helps to isolate the processing
logic from the serialization implementation details. The ITelemetrySerializer<T> interface is defined in a shared
library, which is also used by the device simulator to generate simulated device events and send them to Event
Hubs.
 using System;

namespace Serverless.Serialization
{
public interface ITelemetrySerializer<T>
{
T Deserialize(byte[] message);

ArraySegment<byte> Serialize(T message);

}
}

UpdateState method
The StateChangeProcessor.UpdateState method applies the state changes. The last-known state for each drone is
stored as a JSON document in Cosmos DB. Because the drones send partial updates, the application can't simply
overwrite the document when it gets an update. Instead, it needs to fetch the previous state, merge the fields, and
then perform an upsert operation.
 public class StateChangeProcessor : IStateChangeProcessor
{
private IDocumentClient client;
private readonly string cosmosDBDatabase;
private readonly string cosmosDBCollection;

public StateChangeProcessor(IDocumentClient client, IOptions<StateChangeProcessorOptions> options)

{
this.client = client;
this.cosmosDBDatabase = options.Value.COSMOSDB_DATABASE_NAME;
this.cosmosDBCollection = options.Value.COSMOSDB_DATABASE_COL;
}

public async Task<ResourceResponse<Document>> UpdateState(DeviceState source, ILogger log)

{
log.LogInformation("Processing change message for device ID {DeviceId}", source.DeviceId);

DeviceState target = null;

try
{
var response = await client.ReadDocumentAsync(UriFactory.CreateDocumentUri(cosmosDBDatabase,
cosmosDBCollection, source.DeviceId),
new RequestOptions { PartitionKey = new
PartitionKey(source.DeviceId) });

target = (DeviceState)(dynamic)response.Resource;

// Merge properties
target.Battery = source.Battery ?? target.Battery;
target.FlightMode = source.FlightMode ?? target.FlightMode;
target.Latitude = source.Latitude ?? target.Latitude;
target.Longitude = source.Longitude ?? target.Longitude;
target.Altitude = source.Altitude ?? target.Altitude;
target.AccelerometerOK = source.AccelerometerOK ?? target.AccelerometerOK;
target.GyrometerOK = source.GyrometerOK ?? target.GyrometerOK;
target.MagnetometerOK = source.MagnetometerOK ?? target.MagnetometerOK;
}
catch (DocumentClientException ex)
{
if (ex.StatusCode == System.Net.HttpStatusCode.NotFound)
{
target = source;
}
}

var collectionLink = UriFactory.CreateDocumentCollectionUri(cosmosDBDatabase, cosmosDBCollection);

return await client.UpsertDocumentAsync(collectionLink, target);
}
}

This code uses the IDocumentClient interface to fetch a document from Cosmos DB. If the document exists, the
new state values are merged into the existing document. Otherwise, a new document is created. Both cases are
handled by the UpsertDocumentAsync method.
This code is optimized for the case where the document already exists and can be merged. On the first telemetry
message from a given drone, the ReadDocumentAsync method will throw an exception, because there is no
document for that drone. After the first message, the document will be available.
Notice that this class uses dependency injection to inject the IDocumentClient for Cosmos DB and an IOptions<T>
with configuration settings. We'll see how to set up the dependency injection later.
 NOTE
Azure Functions supports an output binding for Cosmos DB. This binding lets the function app write documents in Cosmos
DB without any code. However, the output binding won't work for this particular scenario, because of the custom upsert
logic that's needed.

Error handling
As mentioned earlier, the RawTelemetryFunction function app processes a batch of messages in a loop. That means
the function needs to handle any exceptions gracefully and continue processing the rest of the batch. Otherwise,
messages might get dropped.
If an exception is encountered when processing a message, the function puts the message onto a dead-letter
queue:

catch (Exception ex)

{
logger.LogError(ex, "Error deserializing message", message.SystemProperties.PartitionKey,
message.SystemProperties.SequenceNumber);
await deadLetterMessages.AddAsync(new DeadLetterMessage { Exception = ex, EventData = message });
}

The dead-letter queue is defined using an output binding to a storage queue:

[FunctionName("RawTelemetryFunction")]
[StorageAccount("DeadLetterStorage")] // App setting that holds the connection string
public async Task RunAsync(
[EventHubTrigger("%EventHubName%", Connection = "EventHubConnection", ConsumerGroup
="%EventHubConsumerGroup%")]EventData[] messages,
[Queue("deadletterqueue")] IAsyncCollector<DeadLetterMessage> deadLetterMessages, // output binding
ILogger logger)

Here the Queue attribute specifies the output binding, and the StorageAccount attribute specifies the name of an
app setting that holds the connection string for the storage account.
Deployment tip: In the Resource Manager template that creates the storage account, you can automatically
populate an app setting with the connection string. The trick is to use the listKeys function.
Here is the section of the template that creates the storage account for the queue:

{
"name": "[variables('droneTelemetryDeadLetterStorageQueueAccountName')]",
"type": "Microsoft.Storage/storageAccounts",
"location": "[resourceGroup().location]",
"apiVersion": "2017-10-01",
"sku": {
"name": "[parameters('storageAccountType')]"
},

Here is the section of the template that creates the function app.
 {
"apiVersion": "2015-08-01",
"type": "Microsoft.Web/sites",
"name": "[variables('droneTelemetryFunctionAppName')]",
"location": "[resourceGroup().location]",
"tags": {
"displayName": "Drone Telemetry Function App"
},
"kind": "functionapp",
"dependsOn": [
"[resourceId('Microsoft.Web/serverfarms', variables('hostingPlanName'))]",
...
],
"properties": {
"serverFarmId": "[resourceId('Microsoft.Web/serverfarms', variables('hostingPlanName'))]",
"siteConfig": {
"appSettings": [
{
"name": "DeadLetterStorage",
"value": "[concat('DefaultEndpointsProtocol=https;AccountName=',
variables('droneTelemetryDeadLetterStorageQueueAccountName'), ';AccountKey=',
listKeys(variables('droneTelemetryDeadLetterStorageQueueAccountId'),'2015-05-01-preview').key1)]"
},
...

This defines an app setting named DeadLetterStorage whose value is populated using the listKeys function. It's
important to make the function app resource depend on the storage account resource (see the dependsOn
element). This guarantees that the storage account is created first and the connection string is available.

Setting up dependency injection

The following code sets up dependency injection for the RawTelemetryFunction function:

[assembly: FunctionsStartup(typeof(DroneTelemetryFunctionApp.Startup))]

namespace DroneTelemetryFunctionApp
{
public class Startup : FunctionsStartup
{
public override void Configure(IFunctionsHostBuilder builder)
{
builder.Services.AddOptions<StateChangeProcessorOptions>()
.Configure<IConfiguration>((configSection, configuration) =>
{
configuration.Bind(configSection);
});

builder.Services.AddTransient<ITelemetrySerializer<DroneState>, TelemetrySerializer<DroneState>>
();
builder.Services.AddTransient<ITelemetryProcessor, TelemetryProcessor>();
builder.Services.AddTransient<IStateChangeProcessor, StateChangeProcessor>();

builder.Services.AddSingleton<IDocumentClient>(ctx => {
var config = ctx.GetService<IConfiguration>();
var cosmosDBEndpoint = config.GetValue<string>("CosmosDBEndpoint");
var cosmosDBKey = config.GetValue<string>("CosmosDBKey");
return new DocumentClient(new Uri(cosmosDBEndpoint), cosmosDBKey);
});
}
}
}
Azure Functions written for .NET can use the ASP.NET Core dependency injection framework. The basic idea is that
you declare a startup method for your assembly. The method takes an IFunctionsHostBuilder interface, which is
used to declare the dependencies for DI. You do this by calling Add* method on the Services object. When you
add a dependency, you specify its lifetime:
Transient objects are created each time they're requested.
Scoped objects are created once per function execution.
Singleton objects are reused across function executions, within the lifetime of the function host.
In this example, the TelemetryProcessor and StateChangeProcessor objects are declared as transient. This is
appropriate for lightweight, stateless services. The DocumentClient class, on the other hand, should be a singleton
for best performance. For more information, see Performance tips for Azure Cosmos DB and .NET.
If you refer back to the code for the RawTelemetryFunction, you'll see there another dependency that doesn't
appear in DI setup code, namely the TelemetryClient class that is used to log application metrics. The Functions
runtime automatically registers this class into the DI container, so you don't need to register it explicitly.
For more information about DI in Azure Functions, see the following articles:
Use dependency injection in .NET Azure Functions
Dependency injection in ASP.NET Core
Passing configuration settings in DI
Sometimes an object must be initialized with some configuration values. Generally, these settings should come
from app settings or (in the case of secrets) from Azure Key Vault.
There are two examples in this application. First, the DocumentClient class takes a Cosmos DB service endpoint and
key. For this object, the application registers a lambda that will be invoked by the DI container. This lambda uses the
IConfiguration interface to read the configuration values:

builder.Services.AddSingleton<IDocumentClient>(ctx => {
var config = ctx.GetService<IConfiguration>();
var cosmosDBEndpoint = config.GetValue<string>("CosmosDBEndpoint");
var cosmosDBKey = config.GetValue<string>("CosmosDBKey");
return new DocumentClient(new Uri(cosmosDBEndpoint), cosmosDBKey);
});

The second example is the StateChangeProcessor class. For this object, we use an approach called the options
pattern. Here's how it works:
1. Define a class T that contains your configuration settings. In this case, the Cosmos DB database name and
collection name.

public class StateChangeProcessorOptions

{
public string COSMOSDB_DATABASE_NAME { get; set; }
public string COSMOSDB_DATABASE_COL { get; set; }
}

2. Add the class T as an options class for DI.

builder.Services.AddOptions<StateChangeProcessorOptions>()
.Configure<IConfiguration>((configSection, configuration) =>
{
configuration.Bind(configSection);
});
3. In the constructor of the class that is being configured, include an IOptions<T> parameter.

public StateChangeProcessor(IDocumentClient client, IOptions<StateChangeProcessorOptions> options)

The DI system will automatically populate the options class with configuration values and pass this to the
constructor.
There are several advantages of this approach:
Decouple the class from the source of the configuration values.
Easily set up different configuration sources, such as environment variables or JSON configuration files.
Simplify unit testing.
Use a strongly typed options class, which is less error prone than just passing in scalar values.

GetStatus function
The other Functions app in this solution implements a simple REST API to get the last-known status of a drone. This
function is defined in a class named GetStatusFunction . Here is the complete code for the function:
 using Microsoft.AspNetCore.Http;
using Microsoft.AspNetCore.Mvc;
using Microsoft.Azure.WebJobs;
using Microsoft.Azure.WebJobs.Extensions.Http;
using Microsoft.Extensions.Logging;
using System.Security.Claims;
using System.Threading.Tasks;

namespace DroneStatusFunctionApp
{
public static class GetStatusFunction
{
public const string GetDeviceStatusRoleName = "GetStatus";

[FunctionName("GetStatusFunction")]
public static IActionResult Run(
[HttpTrigger(AuthorizationLevel.Function, "get", Route = null)]HttpRequest req,
[CosmosDB(
databaseName: "%COSMOSDB_DATABASE_NAME%",
collectionName: "%COSMOSDB_DATABASE_COL%",
ConnectionStringSetting = "COSMOSDB_CONNECTION_STRING",
Id = "{Query.deviceId}",
PartitionKey = "{Query.deviceId}")] dynamic deviceStatus,
ClaimsPrincipal principal,
ILogger log)
{
log.LogInformation("Processing GetStatus request.");

if (!principal.IsAuthorizedByRoles(new[] { GetDeviceStatusRoleName }, log))

{
return new UnauthorizedResult();
}

string deviceId = req.Query["deviceId"];

if (deviceId == null)
{
return new BadRequestObjectResult("Missing DeviceId");
}

if (deviceStatus == null)
{
return new NotFoundResult();
}
else
{
return new OkObjectResult(deviceStatus);
}
}
}
}

This function uses an HTTP trigger to process an HTTP GET request. The function uses a Cosmos DB input binding
to fetch the requested document. One consideration is that this binding will run before the authorization logic is
performed inside the function. If an unauthorized user requests a document, the function binding will still fetch the
document. Then the authorization code will return a 401, so the user won't see the document. Whether this
behavior is acceptable may depend on your requirements. For example, this approach might make it harder to
audit data access for sensitive data.

Authentication and authorization

The web app uses Azure AD to authenticate users. Because the app is a single-page application (SPA) running in the
browser, the implicit grant flow is appropriate:
1. The web app redirects the user to the identity provider (in this case, Azure AD).
2. The user enters their credentials.
3. The identity provider redirects back to the web app with an access token.
4. The web app sends a request to the web API and includes the access token in the Authorization header.

A Function application can be configured to authenticate users with zero code. For more information, see
Authentication and authorization in Azure App Service.
Authorization, on the other hand, generally requires some business logic. Azure AD supports claims based
authentication. In this model, a user's identity is represented as a set of claims that come from the identity provider.
A claim can be any piece of information about the user, such as their name or email address.
The access token contains a subset of user claims. Among these are any application roles that the user is assigned
to.
The principal parameter of the function is a ClaimsPrincipal object that contains the claims from the access
token. Each claim is a key/value pair of claim type and claim value. The application uses these to authorize the
request.
The following extension method tests whether a ClaimsPrincipal object contains a set of roles. It returns false if
any of the specified roles is missing. If this method returns false, the function returns HTTP 401 (Unauthorized).

namespace DroneStatusFunctionApp
{
public static class ClaimsPrincipalAuthorizationExtensions
{
public static bool IsAuthorizedByRoles(
this ClaimsPrincipal principal,
string[] roles,
ILogger log)
{
var principalRoles = new HashSet<string>(principal.Claims.Where(kvp => kvp.Type ==
"roles").Select(kvp => kvp.Value));
var missingRoles = roles.Where(r => !principalRoles.Contains(r)).ToArray();
if (missingRoles.Length > 0)
{
log.LogWarning("The principal does not have the required {roles}", string.Join(", ",
missingRoles));
return false;
}

return true;
}
}
}
For more information about authentication and authorization in this application, see the Security considerations
section of the reference architecture.

Next steps
Once you get a feel for how this reference solution works, learn best practices and recommendations for similar
solutions.
For a serverless event ingestion solution, see Serverless event processing using Azure Functions.
For a serverless web app, see Serverless web application on Azure.
Azure Functions is just one Azure compute option. For help with choosing a compute technology, see Choose an
Azure compute service for your application.

Related resources
For in-depth discussion on developing serverless solutions on premises as well as in the cloud, read Serverless
apps: Architecture, patterns, and Azure implementation.
Read more about the Event-driven architecture style.
 CI/CD for serverless application frontend on Azure
12/18/2020 • 11 minutes to read • Edit Online

Serverless computing abstracts the servers, infrastructure, and operating systems, allowing developers to focus on
application development. A robust CI/CD or Continuous Integration/Continuous Delivery of such applications
allows companies to ship fully tested and integrated software versions within minutes of development. It provides
a backbone of modern DevOps environment.
What does CI/CD actually stand for?
Continuous Integration allows development teams to integrate code changes in a shared repository almost
instantaneously. This ability, coupled with automated build and testing before the changes are integrated,
ensures that only fully functional application code is available for deployment.
Continuous Delivery allows changes in the source code, configuration, content, and other artifacts to be
delivered to production, and ready to be deployed to end-users, as quickly and safely as possible. The process
keeps the code in a deployable state at all times. A special case of this is Continuous Deployment, which
includes actual deployment to end users.
This article discusses a CI/CD pipeline for the web frontend of a serverless reference implementation. This pipeline
is developed using Azure services. The web frontend demonstrates a modern web application, with client-side
JavaScript, reusable server-side APIs, and pre-built Markup, alternatively called Jamstack. You can find the code in
this GitHub repository. The readme describes the steps to download, build, and deploy the application.
The following diagram describes the CI/CD pipeline used in this sample frontend:

This article does not discuss the backend deployment.

Prerequisites
To work with this sample application, make sure you have the following:
A GitHub account.
An Azure account. If you don't have one, you can try out a free Azure account.
An Azure DevOps organization. If you don't have one, you can try out a basic plan, which includes DevOps
services such as Azure Pipelines.

Use an online version control system

Version control systems keep track and control changes in your source code. Keeping your source code in an online
version control system allows multiple development teams to collaborate. It is also easier to maintain than a
traditional version control on premises. These online systems can be easily integrated with leading CI/CD services.
You get the ability to create and maintain the source code in multiple directories, along with build and
configuration files, in what is called a repository.
The project files for this sample application are kept in GitHub. If you don't have a GitHub account, read this
documentation to get started with GitHub repositories.

Automate your build and deploy

Using a CI/CD service such as Azure Pipelines can help you to automate the build and deploy processes. You can
create multiple stages in the pipeline, each stage running based on the result of the previous one. The stages can
run in either a Windows or Linux container. The script must make sure the tools and environments are set properly
in the container. Azure Pipelines can run a variety of build tools, and can work with quite a few online version
control systems.
Integrate build tools
Modern build tools can simplify your build process, and provide functionality such as pre-configuration,
minification of the JavaScript files, and static site generation. Static site generators can build markup files before
they are deployed to the hosting servers, resulting in a fast user experience. You can select from a variety of these
tools, based on the type of your application's programming language and platform, as well as additional
functionality needed. This article provides a list of popular build tools for a modern application.
The sample is a React application, built using Gatsby.js, which is a static site generator and front-end development
framework. These tools can be run locally during development and testing phases, and then integrated with Azure
Pipelines for the final deployment.
The sample uses the Gatsby file gatsby-ssr.js for rendering, and gatsby-config.js for site configuration. Gatsby
converts all JavaScript files under the pages subdirectory of the src folder to HTML files. Additional components
go in the components subdirectory. The sample also uses the gatsby-plugin-typescript plugin that allows using
TypeScript for type safety, instead of JavaScript.
For more information about setting up a Gatsby project, see the official Gatsby documentation.
Automate builds
Automating the build process reduces the human errors that can be introduced in manual processes. The file
azure-pipelines.yml includes the script for a two-stage automation. The Readme for this project describes the steps
required to set up the automation pipeline using Azure Pipelines. The following subsections show how the pipeline
stages are configured.
Build stage
Since the Azure Pipeline is integrated with GitHub repository, any changes in the tracked directory of the main
branch trigger the first stage of the pipeline, the build stage:

trigger:
batch: true
branches:
include:
- main
paths:
include:
- src/ClientApp

The following snippet illustrates the start of the build stage, which starts an Ubuntu container to run this stage.
 stages:
- stage: Build
jobs:
- job: WebsiteBuild
displayName: Build Fabrikam Drone Status app
pool:
vmImage: 'Ubuntu-16.04'
continueOnError: false
steps:

This is followed by tasks and scripts required to successfully build the project. These include the following:
Installing Node.js and setting up environment variables,
Installing and running Gatsby.js that builds the static website:

- script: |
cd src/ClientApp
npm install
npx gatsby build
displayName: 'gatsby build'

installing and running a compression tool named brotli, to compress the built files before deployment:

- script: |
cd src/ClientApp/public
sudo apt-get install brotli --install-suggests --no-install-recommends -q --assume-yes
for f in $(find . -type f \( -iname '*.html' -o -iname '*.map' -o -iname '*.js' -o -iname
'*.json' \)); do brotli $f -Z -j -f -v && mv ${f}.br $f; done
displayName: 'enable compression at origin level'

Computing the version of the current build for cache management,

Publishing the built files for use by the deploy stage:

- task: PublishPipelineArtifact@1
inputs:
targetPath: 'src/ClientApp/public'
artifactName: 'drop'

A successful completion of the build stage tears down the Ubuntu environment, and triggers the deploy stage in
the pipeline.
Deploy stage
The deploy stage runs in a new Ubuntu container:

- stage: Deploy
jobs:
- deployment: WebsiteDeploy
displayName: Deploy Fabrikam Drone Status app
pool:
vmImage: 'Ubuntu-16.04'
environment: 'fabrikamdronestatus-prod'
strategy:
runOnce:
deploy:
steps:
This stage includes various deployment tasks and scripts to:
Download the build artifacts to the container (which happens automatically as a consequence of using
PublishPipelineArtifact in the build stage),
Record the build release version, and update in the GitHub repository,
Upload the website files to Blob Storage, in a new folder corresponding to the new version, and
Change the CDN to point to this new folder.
The last two steps together replicate a cache purge, since older folders are no longer accessible by the CDN edge
servers. The following snippet shows how this is achieved:

- script: |
az login --service-principal -u $(azureArmClientId) -p $(azureArmClientSecret) --tenant
$(azureArmTenantId)
# upload content to container versioned folder
az storage blob upload-batch -s "$(Pipeline.Workspace)/drop" --destination
"\$web\$(releaseSemVer)" --account-name $(azureStorageAccountName) --content-encoding br --pattern "*.html" --
content-type "text/html"
az storage blob upload-batch -s "$(Pipeline.Workspace)/drop" --destination
"\$web\$(releaseSemVer)" --account-name $(azureStorageAccountName) --content-encoding br --pattern "*.js" --
content-type "application/javascript"
az storage blob upload-batch -s "$(Pipeline.Workspace)/drop" --destination
"\$web\$(releaseSemVer)" --account-name $(azureStorageAccountName) --content-encoding br --pattern "*.js.map"
--content-type "application/octet-stream"
az storage blob upload-batch -s "$(Pipeline.Workspace)/drop" --destination
"\$web\$(releaseSemVer)" --account-name $(azureStorageAccountName) --content-encoding br --pattern "*.json" --
content-type "application/json"
az storage blob upload-batch -s "$(Pipeline.Workspace)/drop" --destination
"\$web\$(releaseSemVer)" --account-name $(azureStorageAccountName) --pattern "*.txt" --content-type
"text/plain"
# target new version
az cdn endpoint update --resource-group $(azureResourceGroup) --profile-name $(azureCdnName) --
name $(azureCdnName) --origin-path '/$(releaseSemVer)'
AZURE_CDN_ENDPOINT_HOSTNAME=$(az cdn endpoint show --resource-group $(azureResourceGroup) --name
$(azureCdnName) --profile-name $(azureCdnName) --query hostName -o tsv)
echo "Azure CDN endpoint host ${AZURE_CDN_ENDPOINT_HOSTNAME}"
echo '##vso[task.setvariable variable=azureCndEndpointHost]'$AZURE_CDN_ENDPOINT_HOSTNAME
displayName: 'upload to Azure Storage static website hosting and purge Azure CDN endpoint'

Atomic deploys
Atomic deployment ensures that the users of your website or application always get the content corresponding to
the same version.
In the sample CI/CD pipeline, the website contents are deployed to the Blob storage, which acts as the origin server
for the Azure CDN. If the files are updated in the same root folder in the blob, the website will be served
inconsistently. Uploading to a new versioned folder as shown in the preceding section solves this problem. The
users either get all or nothing of the new successful build, since the CDN points to the new folder as the origin, only
after all files are successfully updated.
The advantages of this approach are as follows:
Since new content is not available to users until the CDN points to the new origin folder, it results in an atomic
deployment.
You can easily roll back to an older version of the website if necessary.
Since the origin can host multiple versions of the website side by side, you can fine-tune the deployment by
using techniques such as allowing preview to certain users before wider availability.

Host and distribute using the cloud

A content delivery network (CDN) is a set of distributed servers that speed up the content delivery to users over a
vast geographical area. Every user gets the content from the server nearest to them. The CDN accesses this content
from an origin server, and caches it to edge servers at strategic locations. The sample CI/CD in this article uses
Azure CDN, pointing to website content hosted on Azure Blob Storage as the origin server. The Blob storage is
configured for static website hosting. For a quick guide on how to use Azure CDN with Azure Blob Storage, read
Integrate an Azure storage account with Azure CDN.
The following are some strategies that can improve the CDN performance.
Compress the content
Compressing the files before serving improves file transfer speed and increases page-load performance for the
end users.
There are two ways to do this:
1. On the fly in the CDN edge ser vers . This improves bandwidth consumption by a significant percentage,
making the website considerably faster than without compression. It is relatively easy to enable this type of
compression on Azure CDN. Since it's controlled by the CDN, you cannot choose or configure the
compression tool. Use this compression if website performance is not a critical concern.
2. Pre-compressing before reaching the CDN , either on your origin server, or before reaching the origin.
Pre-compressing further improves the run time performance of your website, since it's done before being
fetched by the CDN. The sample CI/CD pipeline compresses the built files in the deployment stage, using
brotli, as shown in the build section above. The advantages of using pre-compression are as follows:
a. You can use any compression tool that you're comfortable with, and further fine-tune the compression.
CDNs may have limitations on the tools they use.
b. Some CDNs limit file sizes that can be compressed on the fly, causing a performance hit for larger files.
Pre-compression sets no limits on the file size.
 c. Pre-compressing in the deployment pipeline results in less storage required at the origin.
d. This is faster and more efficient than the CDN compression.
For more information, see Improve performance by compressing files in Azure CDN.
Dynamic site acceleration
Using CDN is optimal for static content, which can be safely cached at the edge servers. However, dynamic web
sites require the server to generate content based on user response. This type of content cannot be cached on the
edge, requiring a more involved end-to-end solution that can speed up the content delivery. Dynamic site
acceleration by Azure CDN is one such solution that measurably improves the performance of dynamic web pages.
This sample enables dynamic site acceleration as shown in this section of the readme.

Manage website cache

CDNs use caching to improve their performance. Configuring and managing this cache becomes an integral part of
your deployment pipeline. The Azure CDN documentation shows several ways to do this. You can set caching rules
on your CDN, as well as configure the time-to-live for the content in the origin server. Static web content can be
cached for a long duration, since it may not change too much over time. This reduces the overhead of accessing the
single origin server for every user request. For more information, see How caching works.
Cache purge
The CDN caches the website files at the edge servers until their time-to-live expires. These are updated for a new
client request, only after their time-to-live has expired. Purging your CDN cache is required to guarantee that every
user gets the latest live website files, especially if the deployment happens at the same CDN origin folder. Azure
CDN allows you to purge the cache from the Azure portal.
A better approach is to invalidate this cache by using versioning during deployment. An explicit cache purge or
expiration usually causes the CDN to retrieve newer versions of the web content. However, since the CDN always
points to the latest version of the deployment, it improves caching in the following manner:
1. The CDN validates the index.html against the origin, for every new website instance.
2. Except for the index.html and 404.html, all other website files are fingerprinted and cached for a year. This is
based on the assumption that resources such as images and videos, do not need frequent changes. If these files
are updated and rebuilt, their names are updated by a new fingerprint GUID. This results in an update to the
index.html with an updated reference to the changed resource file. The CDN then retrieves the updated
index.html, and since it does not find the reference resource file in its cache, it also retrieves the changed
resource files.
This ensures that the CDN always gets new updated files, and removes the need to purge the cache for a new build.

Next steps
Now that you understand the basics, follow this readme to set up and execute the CI/CD pipeline.
Learn best practices for using content delivery networks (CDNs)
 Serverless Functions app operations
12/18/2020 • 3 minutes to read • Edit Online

This article describes Azure operations considerations for serverless Functions applications. To support Functions
apps, operations personnel need to:
Understand and implement hosting configurations.
Future-proof scalability by automating infrastructure provisioning.
Maintain business continuity by meeting availability and disaster recovery requirements.

Planning
To plan operations, understand your workloads and their requirements, then design and configure the best options
for the requirements.
Choose a hosting option
The Azure Functions Runtime provides flexibility in hosting. Use the hosting plan comparison table to determine
the best choice for your requirements.
Azure Functions hosting plans
Each Azure Functions project deploys and runs in its own Functions app, which is the unit of scale and cost.
The three hosting plans available for Azure Functions are the Consumption plan, Premium plan, and
Dedicated (App Service) plan. The hosting plan determines scaling behavior, available resources, and
support for advanced features like virtual network connectivity.
Azure Kubernetes Service (AKS)
Kubernetes-based Functions provides the Functions Runtime in a Docker container with event-driven
scaling through Kubernetes-based Event Driven Autoscaling (KEDA).
For more information about hosting plans, see:
Azure Functions scale and hosting
Consumption plan
Premium plan
Dedicated (App Service) plan
Azure Functions on Kubernetes with KEDA
Azure subscription and service limits, quotas, and constraints
Understand scaling
The serverless Consumption and Premium hosting plans scale automatically, adding and removing Azure
Functions host instances based on the number of incoming events. Scaling can vary on several dimensions, and
behave differently based on plan, trigger, and code language.
For more information about scaling, see:
Understand scaling behaviors
Scalability best practices
Understand and address cold starts
If the number of host instances scales down to zero, the next request has the added latency of restarting the
Function app, called a cold start. Cold start is a large discussion point for serverless architectures, and a point of
ambiguity for Azure Functions.
The Premium hosting plan prevents cold starts by keeping some instances warm. Reducing dependencies and
using asynchronous operations in the Functions app also minimizes the impact of cold starts. However, availability
requirements may require running the app in a Dedicated hosting plan with Always on enabled. The Dedicated
plan uses dedicated virtual machines (VMs), so is not serverless.
For more information about cold start, see Understanding serverless cold start.
Identify storage considerations
Every Azure Functions app relies on Azure Storage for operations such as managing triggers and logging function
executions. When creating a Functions app, you must create or link to a general-purpose Azure Storage account
that supports Blob, Queue, and Table storage. For more information, see Storage considerations for Azure
Functions.
Identify network design considerations
Networking options let the Functions app restrict access, or access resources without using internet-routable
addresses. The hosting plans offer different levels of network isolation. Choose the option that best meets your
network isolation requirements. For more information, see Azure Functions networking options.

Production
To prepare the application for production, make sure you can easily redeploy the hosting plan, and apply scale-out
rules.
Automate hosting plan provisioning
With infrastructure as code, you can automate infrastructure provisioning. Automatic provisioning provides more
resiliency during disasters, and more agility to quickly redeploy the infrastructure as needed.
For more information on automated provisioning, see:
Automate resource deployment for your function app in Azure Functions
Terraform - Manages a Function App
Configure scale out options
Autoscale provides the right amount of running resources to handle application load. Autoscale adds resources to
handle increases in load, and saves money by removing resources that are idle.
For more information about autoscale options, see:
Premium Plan settings
App Service Plan settings

Optimization
When the application is in production, make sure that:
The hosting plan can scale to meet application demands.
There's a plan for business continuity, availability, and disaster recovery.
You can monitor hosting and application health and receive alerts.
Implement availability requirements
Azure Functions run in a specific region. To get higher availability, you can deploy the same Functions app to
multiple regions. In multiple regions, Functions can run in the active-active or active-passive availability pattern.
For more information about Azure Functions availability and disaster recovery, see:
 Azure Functions geo-disaster recovery
Disaster recovery and geo-distribution in Azure Durable Functions
Monitoring logging, application monitoring, and alerting
Application Insights and logs in Azure Monitor automatically collect log, performance, and error data and detect
performance anomalies. Azure Monitor includes powerful analytics tools to help diagnose issues and understand
function use. Application Insights help you continuously improve performance and usability.
For more information about monitoring and analyzing Azure Functions performance, see:
Monitor Azure Functions
Monitor Azure Functions with Azure Monitor logs
Application Insights for Azure Functions supported features

Next steps
Serverless application development and deployment
Azure Functions app security
 Serverless Functions security
12/18/2020 • 5 minutes to read • Edit Online

This article describes Azure services and activities security personnel can implement for serverless Functions.
These guidelines and resources help develop secure code and deploy secure applications to the cloud.

Planning
The primary goals of a secure serverless Azure Functions application environment are to protect running
applications, quickly identify and address security issues, and prevent future similar issues.
The OWASP Serverless Top 10 describes the most common serverless application security vulnerabilities, and
provides basic techniques to identify and protect against them.
In many ways, planning for secure development, deployment, and operation of serverless functions is much the
same as for any web-based or cloud hosted application. Azure App Service provides the hosting infrastructure for
your function apps. Securing Azure Functions article provides security strategies for running your function code,
and how App Service can help you secure your functions.
For more information about Azure security, best practices, and shared responsibilities, see:
Security in Azure App Service
Built-in security controls
Secure development best practices on Azure.
Security best practices for Azure solutions (PDF report)
Shared responsibilities for cloud computing (PDF report)

Deployment
To prepare serverless Functions applications for production, security personnel should:
Conduct regular code reviews to identify code and library vulnerabilities.
Define resource permissions that Functions needs to execute.
Configure network security rules for inbound and outbound communication.
Identify and classify sensitive data access.
The Azure Security Baseline for Azure Functions article contains more recommendations that will help you
improve the security posture of your deployment.
Keep code secure
Find security vulnerabilities and errors in code and manage security vulnerabilities in projects and dependencies.
For more information, see:
GitHub - Finding security vulnerabilities and errors in your code
GitHub - Managing security vulnerabilities in your project
GitHub - Managing vulnerabilities in your project's dependencies
Perform input validation
Different event sources like Blob storage, Cosmos DB NoSQL databases, event hubs, queues, or Graph events can
trigger serverless Functions. Injections aren't strictly limited to inputs coming directly from the API calls. Functions
may consume other input from the possible event sources.
In general, don't trust input or make any assumptions about its validity. Always use safe APIs that sanitize or
validate the input. If possible, use APIs that bind or parameterize variables, like using prepared statements for SQL
queries.
For more information, see:
Azure Functions Input Validation with FluentValidation
Security Frame: Input Validation Mitigations
HTTP Trigger Function Request Validation
How to validate request for Azure Functions
Secure HTTP endpoints for development, testing, and production
Azure Functions lets you use keys to make it harder to access your HTTP function endpoints. To fully secure your
function endpoints in production, consider implementing one of the following Function app-level security options:
Turn on App Service authentication and authorization for your Functions app. See Authorization keys.
Use Azure API Management (APIM) to authenticate requests. See Import an Azure Function App as an API in
Azure API Management.
Deploy your Functions app to an Azure App Service Environment (ASE).
Use an App Service Plan that restricts access, and implement Azure Front Door + WAF to handle your incoming
requests. See Create a Front Door for a highly available global web application.
For more information, see Secure an HTTP endpoint in production.
Set up role -based access control (RBAC )
Azure role-based access control (RBAC) has several built-in Azure roles that you can assign to users, groups,
service principals, and managed identities to control access to Azure resources. If the built-in roles don't meet your
organization's needs, you can create your own Azure custom roles.
Review each Functions app before deployment to identify excessive permissions. Carefully examine functions to
apply "least privilege" permissions, giving each function only what it needs to successfully execute.
Use RBAC to assign permissions to users, groups, and applications at a certain scope. The scope of a role
assignment can be a subscription, a resource group, or a single resource. Avoid using wildcards whenever
possible.
For more information about RBAC, see:
What is role-based access control (RBAC) for Azure resources?
Azure built-in roles
Azure role-based access control (RBAC)
Azure custom roles
Use managed identities and key vaults
A common challenge when building cloud applications is how to manage credentials for authenticating to cloud
services in your code. Credentials should never appear in application code, developer workstations, or source
control. Instead, use a key vault to store and retrieve keys and credentials. Azure Key Vault provides a way to
securely store credentials, secrets, and other keys. The code authenticates to Key Vault to retrieve the credentials.
For more information, see Use Key Vault references for App Service and Azure Functions.
Managed identities let Functions apps access resources like key vaults and storage accounts without requiring
specific access keys or connection strings. A full audit trail in the logs displays which identities execute requests to
resources. Use RBAC and managed identities to granularly control exactly what resources Azure Functions
applications can access.
For more information, see:
What are managed identities for Azure resources?
How to use managed identities for App Service and Azure Functions
Use shared access signature (SAS ) tokens to limit access to resources
A shared access signature (SAS) provides secure delegated access to resources in your storage account, without
compromising the security of your data. With a SAS, you have granular control over how a client can access your
data. You can control what resources the client may access, what permissions they have on those resources, and
how long the SAS is valid, among other parameters.
For more information, see Grant limited access to Azure Storage resources using shared access signatures (SAS).
Secure Blob storage
Identify and classify sensitive data, and minimize sensitive data storage to only what is necessary. For sensitive
data storage, add multi-factor authentication and data encryption in transit and at rest. Grant limited access to
Azure Storage resources using SAS tokens.
For more information, see Security recommendations for Blob storage.

Optimization
Once an application is in production, security personnel can help optimize workflow and prepare for scaling.
Use Azure Security Center and apply security recommendations
Azure Security Center is a security scanning solution for your application that identifies potential security
vulnerabilities and creates recommendations. The recommendations guide you to configure needed controls to
harden and protect your resources.
For more information, see:
Protect your applications with Azure Security Center
Security Center app recommendations
Enforce application governance policies
Apply centralized, consistent enforcements and safeguards to your application at scale. For more information, see
Azure Policy built-in policy definitions.

Next steps
Serverless application development and deployment
Azure Functions app operations
 DevOps Checklist
11/2/2020 • 14 minutes to read • Edit Online

DevOps is the integration of development, quality assurance, and IT operations into a unified culture and set of
processes for delivering software. Use this checklist as a starting point to assess your DevOps culture and process.

Culture
Ensure business alignment across organizations and teams. Conflicts over resources, purpose, goals, and
priorities within an organization can be a risk to successful operations. Ensure that the business, development, and
operations teams are all aligned.
Ensure the entire team understands the software lifecycle. Your team needs to understand the overall
lifecycle of the application, and which part of the lifecycle the application is currently in. This helps all team
members know what they should be doing now, and what they should be planning and preparing for in the future.
Reduce cycle time. Aim to minimize the time it takes to move from ideas to usable developed software. Limit the
size and scope of individual releases to keep the test burden low. Automate the build, test, configuration, and
deployment processes whenever possible. Clear any obstacles to communication among developers, and between
developers and operations.
Review and improve processes. Your processes and procedures, both automated and manual, are never final.
Set up regular reviews of current workflows, procedures, and documentation, with a goal of continual
improvement.
Do proactive planning. Proactively plan for failure. Have processes in place to quickly identify issues when they
occur, escalate to the correct team members to fix, and confirm resolution.
Learn from failures. Failures are inevitable, but it's important to learn from failures to avoid repeating them. If an
operational failure occurs, triage the issue, document the cause and solution, and share any lessons that were
learned. Whenever possible, update your build processes to automatically detect that kind of failure in the future.
Optimize for speed and collect data. Every planned improvement is a hypothesis. Work in the smallest
increments possible. Treat new ideas as experiments. Instrument the experiments so that you can collect
production data to assess their effectiveness. Be prepared to fail fast if the hypothesis is wrong.
Allow time for learning. Both failures and successes provide good opportunities for learning. Before moving on
to new projects, allow enough time to gather the important lessons, and make sure those lessons are absorbed by
your team. Also give the team the time to build skills, experiment, and learn about new tools and techniques.
Document operations. Document all tools, processes, and automated tasks with the same level of quality as
your product code. Document the current design and architecture of any systems you support, along with recovery
processes and other maintenance procedures. Focus on the steps you actually perform, not theoretically optimal
processes. Regularly review and update the documentation. For code, make sure that meaningful comments are
included, especially in public APIs, and use tools to automatically generate code documentation whenever possible.
Share knowledge. Documentation is only useful if people know that it exists and can find it. Ensure the
documentation is organized and easily discoverable. Be creative: Use brown bags (informal presentations), videos,
or newsletters to share knowledge.

Development
Provide developers with production-like environments. If development and test environments don't match
the production environment, it is hard to test and diagnose problems. Therefore, keep development and test
environments as close to the production environment as possible. Make sure that test data is consistent with the
data used in production, even if it's sample data and not real production data (for privacy or compliance reasons).
Plan to generate and anonymize sample test data.
Ensure that all authorized team members can provision infrastructure and deploy the application.
Setting up production-like resources and deploying the application should not involve complicated manual tasks
or detailed technical knowledge of the system. Anyone with the right permissions should be able to create or
deploy production-like resources without going to the operations team.

This recommendation doesn't imply that anyone can push live updates to the production deployment. It's
about reducing friction for the development and QA teams to create production-like environments.

Instrument the application for insight. To understand the health of your application, you need to know how
it's performing and whether it's experiencing any errors or problems. Always include instrumentation as a design
requirement, and build the instrumentation into the application from the start. Instrumentation must include event
logging for root cause analysis, but also telemetry and metrics to monitor the overall health and usage of the
application.
Track your technical debt. In many projects, release schedules can get prioritized over code quality to one
degree or another. Always keep track when this occurs. Document any shortcuts or other suboptimal
implementations, and schedule time in the future to revisit these issues.
Consider pushing updates directly to production. To reduce the overall release cycle time, consider pushing
properly tested code commits directly to production. Use feature toggles to control which features are enabled.
This allows you to move from development to release quickly, using the toggles to enable or disable features.
Toggles are also useful when performing tests such as canary releases, where a particular feature is deployed to a
subset of the production environment.

Testing
Automate testing. Manually testing software is tedious and susceptible to error. Automate common testing tasks
and integrate the tests into your build processes. Automated testing ensures consistent test coverage and
reproducibility. Integrated UI tests should also be performed by an automated tool. Azure offers development and
test resources that can help you configure and execute testing. For more information, see Development and test.
Test for failures. If a system can't connect to a service, how does it respond? Can it recover once the service is
available again? Make fault injection testing a standard part of review on test and staging environments. When
your test process and practices are mature, consider running these tests in production.
Test in production. The release process doesn't end with deployment to production. Have tests in place to ensure
that deployed code works as expected. For deployments that are infrequently updated, schedule production testing
as a regular part of maintenance.
Automate performance testing to identify performance issues early. The impact of a serious performance
issue can be as severe as a bug in the code. While automated functional tests can prevent application bugs, they
might not detect performance problems. Define acceptable performance goals for metrics like latency, load times,
and resource usage. Include automated performance tests in your release pipeline, to make sure the application
meets those goals.
Perform capacity testing. An application might work fine under test conditions, and then have problems in
production due to scale or resource limitations. Always define the maximum expected capacity and usage limits.
Test to make sure the application can handle those limits, but also test what happens when those limits are
exceeded. Capacity testing should be performed at regular intervals.
After the initial release, you should run performance and capacity tests whenever updates are made to production
code. Use historical data to fine-tune tests and to determine what types of tests need to be performed.
Perform automated security penetration testing. Ensuring your application is secure is as important as
testing any other functionality. Make automated penetration testing a standard part of the build and deployment
process. Schedule regular security tests and vulnerability scanning on deployed applications, monitoring for open
ports, endpoints, and attacks. Automated testing does not remove the need for in-depth security reviews at regular
intervals.
Perform automated business continuity testing. Develop tests for large-scale business continuity, including
backup recovery and failover. Set up automated processes to perform these tests regularly.

Release
Automate deployments. Automate deploying the application to test, staging, and production environments.
Automation enables faster and more reliable deployments, and ensures consistent deployments to any supported
environment. It removes the risk of human error caused by manual deployments. It also makes it easy to schedule
releases for convenient times, to minimize any effects of potential downtime. Have systems in place to detect any
problems during rollout, and have an automated way to roll forward fixes or roll back changes.
Use continuous integration. Continuous integration (CI) is the practice of merging all developer code into a
central codebase on a regular schedule, and then automatically performing standard build and test processes. CI
ensures that an entire team can work on a codebase at the same time without having conflicts. It also ensures that
code defects are found as early as possible. Preferably, the CI process should run every time that code is
committed or checked in. At the very least, it should run once per day.

Consider adopting a trunk based development model. In this model, developers commit to a single branch (the
trunk). There is a requirement that commits never break the build. This model facilitates CI, because all feature
work is done in the trunk, and any merge conflicts are resolved when the commit happens.

Consider using continuous deliver y. Continuous delivery (CD) is the practice of ensuring that code is always
ready to deploy, by automatically building, testing, and deploying code to production-like environments. Adding
continuous delivery to create a full CI/CD pipeline will help you detect code defects as soon as possible, and
ensures that properly tested updates can be released in a very short time.

Continuous deployment is an additional process that automatically takes any updates that have passed
through the CI/CD pipeline and deploys them into production. Continuous deployment requires robust
automatic testing and advanced process planning, and may not be appropriate for all teams.

Make small incremental changes. Large code changes have a greater potential to introduce bugs. Whenever
possible, keep changes small. This limits the potential effects of each change, and makes it easier to understand
and debug any issues.
Control exposure to changes. Make sure you're in control of when updates are visible to your end users.
Consider using feature toggles to control when features are enabled for end users.
Implement release management strategies to reduce deployment risk . Deploying an application update
to production always entails some risk. To minimize this risk, use strategies such as canary releases or blue-green
deployments to deploy updates to a subset of users. Confirm the update works as expected, and then roll the
update out to the rest of the system.
Document all changes. Minor updates and configuration changes can be a source of confusion and versioning
conflict. Always keep a clear record of any changes, no matter how small. Log everything that changes, including
patches applied, policy changes, and configuration changes. (Don't include sensitive data in these logs. For
example, log that a credential was updated, and who made the change, but don't record the updated credentials.)
The record of the changes should be visible to the entire team.
Consider making infrastructure immutable. Immutable infrastructure is the principle that you shouldn't
modify infrastructure after it's deployed to production. Otherwise, you can get into a state where ad hoc changes
have been applied, making it hard to know exactly what changed. Immutable infrastructure works by replacing
entire servers as part of any new deployment. This allows the code and the hosting environment to be tested and
deployed as a block. Once deployed, infrastructure components aren't modified until the next build and deploy
cycle.

Monitoring
Make systems obser vable. The operations team should always have clear visibility into the health and status of
a system or service. Set up external health endpoints to monitor status, and ensure that applications are coded to
instrument the operations metrics. Use a common and consistent schema that helps you correlate events across
systems. Azure Diagnostics and Application Insights are the standard method of tracking the health and status of
Azure resources. Microsoft Operation Management Suite also provides centralized monitoring and management
for cloud or hybrid solutions.
Aggregate and correlate logs and metrics . A properly instrumented telemetry system will provide a large
amount of raw performance data and event logs. Make sure that telemetry and log data is processed and
correlated in a short period of time, so that operations staff always have an up-to-date picture of system health.
Organize and display data in ways that give a cohesive view of any issues, so that whenever possible it's clear
when events are related to one another.

Consult your corporate retention policy for requirements on how data is processed and how long it should be
stored.

Implement automated aler ts and notifications. Set up monitoring tools like Azure Monitor to detect patterns
or conditions that indicate potential or current issues, and send alerts to the team members who can address the
issues. Tune the alerts to avoid false positives.
Monitor assets and resources for expirations. Some resources and assets, such as certificates, expire after a
given amount of time. Make sure to track which assets expire, when they expire, and what services or features
depend on them. Use automated processes to monitor these assets. Notify the operations team before an asset
expires, and escalate if expiration threatens to disrupt the application.

Management
Automate operations tasks. Manually handling repetitive operations processes is error-prone. Automate these
tasks whenever possible to ensure consistent execution and quality. Code that implements the automation should
be versioned in source control. As with any other code, automation tools must be tested.
Take an infrastructure-as-code approach to provisioning. Minimize the amount of manual configuration
needed to provision resources. Instead, use scripts and Azure Resource Manager templates. Keep the scripts and
templates in source control, like any other code you maintain.
Consider using containers. Containers provide a standard package-based interface for deploying applications.
Using containers, an application is deployed using self-contained packages that include any software,
dependencies, and files needed to run the application, which greatly simplifies the deployment process.
Containers also create an abstraction layer between the application and the underlying operating system, which
provides consistency across environments. This abstraction can also isolate a container from other processes or
applications running on a host.
Implement resiliency and self-healing. Resiliency is the ability of an application to recover from failures.
Strategies for resiliency include retrying transient failures, and failing over to a secondary instance or even another
region. For more information, see Designing reliable Azure applications . Instrument your applications so that
issues are reported immediately and you can manage outages or other system failures.
Have an operations manual. An operations manual or runbook documents the procedures and management
information needed for operations staff to maintain a system. Also document any operations scenarios and
mitigation plans that might come into play during a failure or other disruption to your service. Create this
documentation during the development process, and keep it up to date afterwards. This is a living document, and
should be reviewed, tested, and improved regularly.
Shared documentation is critical. Encourage team members to contribute and share knowledge. The entire team
should have access to documents. Make it easy for anyone on the team to help keep documents updated.
Document on-call procedures. Make sure on-call duties, schedules, and procedures are documented and
shared to all team members. Keep this information up-to-date at all times.
Document escalation procedures for third-par ty dependencies. If your application depends on external
third-party services that you don't directly control, you must have a plan to deal with outages. Create
documentation for your planned mitigation processes. Include support contacts and escalation paths.
Use configuration management. Configuration changes should be planned, visible to operations, and recorded.
This could take the form of a configuration management database, or a configuration-as-code approach.
Configuration should be audited regularly to ensure that what's expected is actually in place.
Get an Azure suppor t plan and understand the process. Azure offers a number of support plans. Determine
the right plan for your needs, and make sure the entire team knows how to use it. Team members should
understand the details of the plan, how the support process works, and how to open a support ticket with Azure. If
you are anticipating a high-scale event, Azure support can assist you with increasing your service limits. For more
information, see the Azure Support FAQs.
Follow least-privilege principles when granting access to resources. Carefully manage access to
resources. Access should be denied by default, unless a user is explicitly given access to a resource. Only grant a
user access to what they need to complete their tasks. Track user permissions and perform regular security audits.
Use role-based access control. Assigning user accounts and access to resources should not be a manual
process. Use role-based access control (RBAC) grant access based on Azure Active Directory identities and groups.
Use a bug tracking system to track issues. Without a good way to track issues, it's easy to miss items,
duplicate work, or introduce additional problems. Don't rely on informal person-to-person communication to track
the status of bugs. Use a bug tracking tool to record details about problems, assign resources to address them, and
provide an audit trail of progress and status.
Manage all resources in a change management system. All aspects of your DevOps process should be
included in a management and versioning system, so that changes can be easily tracked and audited. This includes
code, infrastructure, configuration, documentation, and scripts. Treat all these types of resources as code
throughout the test/build/review process.
Use checklists. Create operations checklists to ensure processes are followed. It's common to miss something in
a large manual, and following a checklist can force attention to details that might otherwise be overlooked.
Maintain the checklists, and continually look for ways to automate tasks and streamline processes.
For more about DevOps, see What is DevOps? on the Visual Studio site.
 Extend Azure Resource Manager template
functionality
12/18/2020 • 2 minutes to read • Edit Online

In 2016, the Microsoft patterns & practices team created a set of Azure Resource Manager template building blocks
with the goal of simplifying resource deployment. Each building block contains a set of prebuilt templates that
deploy sets of resources specified by separate parameter files.
The building block templates are designed to be combined together to create larger and more complex
deployments. For example, deploying a virtual machine in Azure requires a virtual network, storage accounts, and
other resources. The virtual network building block template deploys a virtual network and subnets. The virtual
machine building block template deploys storage accounts, network interfaces, and the actual VMs. You can then
create a script or template to call both building block templates with their corresponding parameter files to deploy
a complete architecture with one operation.
While developing the building block templates, p&p designed several concepts to extend Azure Resource Manager
template functionality. In this series, we will describe several of these concepts so you can use them in your own
templates.

NOTE
These articles assume you have an advanced understanding of Azure Resource Manager templates.
 Update a resource in an Azure Resource Manager
template
12/18/2020 • 3 minutes to read • Edit Online

There are some scenarios in which you need to update a resource during a deployment. You might encounter this
scenario when you cannot specify all the properties for a resource until other, dependent resources are created. For
example, if you create a backend pool for a load balancer, you might update the network interfaces (NICs) on your
virtual machines (VMs) to include them in the backend pool. And while Resource Manager supports updating
resources during deployment, you must design your template correctly to avoid errors and to ensure the
deployment is handled as an update.
First, you must reference the resource once in the template to create it and then reference the resource by the same
name to update it later. However, if two resources have the same name in a template, Resource Manager throws an
exception. To avoid this error, specify the updated resource in a second template that's either linked or included as a
subtemplate using the Microsoft.Resources/deployments resource type.
Second, you must either specify the name of the existing property to change or a new name for a property to add
in the nested template. You must also specify the original properties and their original values. If you fail to provide
the original properties and values, Resource Manager assumes you want to create a new resource and deletes the
original resource.

Example template
Let's look at an example template that demonstrates this. Our template deploys a virtual network named
firstVNet that has one subnet named firstSubnet . It then deploys a virtual network interface (NIC) named nic1
and associates it with our subnet. Then, a deployment resource named updateVNet includes a nested template that
updates our firstVNet resource by adding a second subnet named secondSubnet .

{
"$schema": "https://schema.management.azure.com/schemas/2015-01-01/deploymentTemplate.json#",
"contentVersion": "1.0.0.0",
"parameters": {},
"resources": [
{
"apiVersion": "2016-03-30",
"name": "firstVNet",
"location":"[resourceGroup().location]",
"type": "Microsoft.Network/virtualNetworks",
"properties": {
"addressSpace":{"addressPrefixes": [
"10.0.0.0/22"
]},
"subnets":[
{
"name":"firstSubnet",
"properties":{
"addressPrefix":"10.0.0.0/24"
}
}
]
}
},
{
"apiVersion": "2015-06-15",
"type":"Microsoft.Network/networkInterfaces",
 "name":"nic1",
"location":"[resourceGroup().location]",
"dependsOn": [
"firstVNet"
],
"properties": {
"ipConfigurations":[
{
"name":"ipconfig1",
"properties": {
"privateIPAllocationMethod":"Dynamic",
"subnet": {
"id": "
[concat(resourceId('Microsoft.Network/virtualNetworks','firstVNet'),'/subnets/firstSubnet')]"
}
}
}
]
}
},
{
"apiVersion": "2015-01-01",
"type": "Microsoft.Resources/deployments",
"name": "updateVNet",
"dependsOn": [
"nic1"
],
"properties": {
"mode": "Incremental",
"parameters": {},
"template": {
"$schema": "https://schema.management.azure.com/schemas/2015-01-01/deploymentTemplate.json#",
"contentVersion": "1.0.0.0",
"parameters": {},
"variables": {},
"resources": [
{
"apiVersion": "2016-03-30",
"name": "firstVNet",
"location":"[resourceGroup().location]",
"type": "Microsoft.Network/virtualNetworks",
"properties": {
"addressSpace": "[reference('firstVNet').addressSpace]",
"subnets":[
{
"name":"[reference('firstVNet').subnets[0].name]",
"properties":{
"addressPrefix":"
[reference('firstVNet').subnets[0].properties.addressPrefix]"
}
},
{
"name":"secondSubnet",
"properties":{
"addressPrefix":"10.0.1.0/24"
}
}
]
}
}
],
"outputs": {}
}
}
}
],
"outputs": {}
}
Let's take a look at the resource object for our firstVNet resource first. Notice that we specify again the settings
for our firstVNet in a nested template—this is because Resource Manager doesn't allow the same deployment
name within the same template and nested templates are considered to be a different template. By again specifying
our values for our firstSubnet resource, we are telling Resource Manager to update the existing resource instead
of deleting it and redeploying it. Finally, our new settings for secondSubnet are picked up during this update.

Try the template

An example template is available on GitHub. To deploy the template, run the following Azure CLI commands:

az group create --location <location> --name <resource-group-name>

az deployment group create -g <resource-group-name> \
--template-uri https://raw.githubusercontent.com/mspnp/template-examples/master/example1-update/deploy.json

Once deployment has finished, open the resource group you specified in the portal. You see a virtual network
named firstVNet and a NIC named nic1 . Click firstVNet , then click subnets . You see the firstSubnet that was
originally created, and you see the secondSubnet that was added in the updateVNet resource.

Then, go back to the resource group and click nic1 then click IP configurations . In the IP configurations
section, the subnet is set to firstSubnet (10.0.0.0/24) .

The original firstVNet has been updated instead of re-created. If firstVNet had been re-created, nic1 would not
be associated with firstVNet .

Next steps
Learn how deploy a resource based on a condition, such as whether a parameter value is present. See
Conditionally deploy a resource in an Azure Resource Manager template.
 Conditionally deploy a resource in an Azure Resource
Manager template
12/18/2020 • 3 minutes to read • Edit Online

There are some scenarios in which you need to design your template to deploy a resource based on a condition,
such as whether or not a parameter value is present. For example, your template may deploy a virtual network and
include parameters to specify other virtual networks for peering. If you've not specified any parameter values for
peering, you don't want Resource Manager to deploy the peering resource.
To accomplish this, use the condition element in the resource to test the length of your parameter array. If the
length is zero, return false to prevent deployment, but for all values greater than zero return true to allow
deployment.

Example template
Let's look at an example template that demonstrates this. Our template uses the condition element to control
deployment of the Microsoft.Network/virtualNetworks/virtualNetworkPeerings resource. This resource creates a
peering between two Azure Virtual Networks in the same region.
Let's take a look at each section of the template.
The parameters element defines a single parameter named virtualNetworkPeerings :

{
"$schema": "https://schema.management.azure.com/schemas/2015-01-01/deploymentTemplate.json#",
"contentVersion": "1.0.0.0",
"parameters": {
"virtualNetworkPeerings": {
"type": "array",
"defaultValue": []
}
},

Our virtualNetworkPeerings parameter is an array and has the following schema:

"virtualNetworkPeerings": [
{
"name": "firstVNet/peering1",
"properties": {
"remoteVirtualNetwork": {
"id": "[resourceId('Microsoft.Network/virtualNetworks','secondVNet')]"
},
"allowForwardedTraffic": true,
"allowGatewayTransit": true,
"useRemoteGateways": false
}
}
]

The properties in our parameter specify the settings related to peering virtual networks. We'll provide the values
for these properties when we specify the Microsoft.Network/virtualNetworks/virtualNetworkPeerings resource in
the resources section:
 "resources": [
{
"type": "Microsoft.Resources/deployments",
"apiVersion": "2017-05-10",
"name": "[concat('vnp-', copyIndex())]",
"condition": "[greater(length(parameters('virtualNetworkPeerings')), 0)]",
"dependsOn": [
"firstVNet", "secondVNet"
],
"copy": {
"name": "iterator",
"count": "[length(variables('peerings'))]",
"mode": "serial"
},
"properties": {
"mode": "Incremental",
"template": {
"$schema": "https://schema.management.azure.com/schemas/2015-01-01/deploymentTemplate.json#",
"contentVersion": "1.0.0.0",
"parameters": {
},
"variables": {
},
"resources": [
{
"type": "Microsoft.Network/virtualNetworks/virtualNetworkPeerings",
"apiVersion": "2016-06-01",
"location": "[resourceGroup().location]",
"name": "[variables('peerings')[copyIndex()].name]",
"properties": "[variables('peerings')[copyIndex()].properties]"
}
],
"outputs": {
}
}
}
}
]

There are a couple of things going on in this part of our template. First, the actual resource being deployed is an
inline template of type Microsoft.Resources/deployments that includes its own template that actually deploys the
Microsoft.Network/virtualNetworks/virtualNetworkPeerings .

Our name for the inline template is made unique by concatenating the current iteration of the copyIndex() with
the prefix vnp- .
The condition element specifies that our resource should be processed when the greater() function evaluates to
true . Here, we're testing if the virtualNetworkPeerings parameter array is greater() than zero. If it is, it evaluates
to true and the condition is satisfied. Otherwise, it's false .
Next, we specify our copy loop. It's a serial loop that means the loop is done in sequence, with each resource
waiting until the last resource has been deployed. The count property specifies the number of times the loop
iterates. Here, normally we'd set it to the length of the virtualNetworkPeerings array because it contains the
parameter objects specifying the resource we want to deploy. However, if we do that, validation will fail if the array
is empty because Resource Manager notices that we are attempting to access properties that do not exist. We can
work around this, however. Let's take a look at the variables we'll need:
 "variables": {
"workaround": {
"true": "[parameters('virtualNetworkPeerings')]",
"false": [{
"name": "workaround",
"properties": {}
}]
},
"peerings": "[variables('workaround')[string(greater(length(parameters('virtualNetworkPeerings')), 0))]]"
},

Our workaround variable includes two properties, one named true and one named false . The true property
evaluates to the value of the virtualNetworkPeerings parameter array. The false property evaluates to an empty
object including the named properties that Resource Manager expects to see—note that false is actually an array,
just as our virtualNetworkPeerings parameter is, which will satisfy validation.
Our peerings variable uses our workaround variable by once again testing if the length of the
virtualNetworkPeerings parameter array is greater than zero. If it is, the string evaluates to true and the
workaround variable evaluates to the virtualNetworkPeerings parameter array. Otherwise, it evaluates to false
and the workaround variable evaluates to our empty object in the first element of the array.
Now that we've worked around the validation issue, we can simply specify the deployment of the
Microsoft.Network/virtualNetworks/virtualNetworkPeerings resource in the nested template, passing the name and
properties from our virtualNetworkPeerings parameter array. You can see this in the template element nested in
the properties element of our resource.

Try the template

An example template is available on GitHub. To deploy the template, run the following Azure CLI commands:

az group create --location <location> --name <resource-group-name>

az deployment group create -g <resource-group-name> \
--template-uri https://raw.githubusercontent.com/mspnp/template-examples/master/example2-
conditional/deploy.json

Next steps
Use objects instead of scalar values as template parameters. See Use an object as a parameter in an Azure
Resource Manager template
 Use an object as a parameter in an Azure Resource
Manager template
12/18/2020 • 5 minutes to read • Edit Online

When you author Azure Resource Manager templates, you can either specify resource property values directly in
the template or define a parameter and provide values during deployment. It's fine to use a parameter for each
property value for small deployments, but there is a limit of 255 parameters per deployment. Once you get to
larger and more complex deployments you may run out of parameters.
One way to solve this problem is to use an object as a parameter instead of a value. To do this, define the
parameter in your template and specify a JSON object instead of a single value during deployment. Then, reference
the subproperties of the parameter using the parameter() function and dot operator in your template.
Let's take a look at an example that deploys a virtual network resource. First, let's specify a VNetSettings
parameter in our template and set the type to object :

...
"parameters": {
"VNetSettings":{"type":"object"}
},

Next, let's provide values for the VNetSettings object:

NOTE
To learn how to provide parameter values during deployment, see the parameters section of understand the structure and
syntax of Azure Resource Manager templates.

"parameters":{
"VNetSettings":{
"value":{
"name":"VNet1",
"addressPrefixes": [
{
"name": "firstPrefix",
"addressPrefix": "10.0.0.0/22"
}
],
"subnets":[
{
"name": "firstSubnet",
"addressPrefix": "10.0.0.0/24"
},
{
"name":"secondSubnet",
"addressPrefix":"10.0.1.0/24"
}
]
}
}
}

As you can see, our single parameter actually specifies three subproperties: name , addressPrefixes , and subnets .
Each of these subproperties either specifies a value or other subproperties. The result is that our single parameter
specifies all the values necessary to deploy our virtual network.
Now let's have a look at the rest of our template to see how the VNetSettings object is used:

...
"resources": [
{
"apiVersion": "2015-06-15",
"type": "Microsoft.Network/virtualNetworks",
"name": "[parameters('VNetSettings').name]",
"location":"[resourceGroup().location]",
"properties": {
"addressSpace":{
"addressPrefixes": [
"[parameters('VNetSettings').addressPrefixes[0].addressPrefix]"
]
},
"subnets":[
{
"name":"[parameters('VNetSettings').subnets[0].name]",
"properties": {
"addressPrefix": "[parameters('VNetSettings').subnets[0].addressPrefix]"
}
},
{
"name":"[parameters('VNetSettings').subnets[1].name]",
"properties": {
"addressPrefix": "[parameters('VNetSettings').subnets[1].addressPrefix]"
}
}
]
}
}
]

The values of our VNetSettings object are applied to the properties required by our virtual network resource
using the parameters() function with both the [] array indexer and the dot operator. This approach works if you
just want to statically apply the values of the parameter object to the resource. However, if you want to dynamically
assign an array of property values during deployment you can use a copy loop. To use a copy loop, you provide a
JSON array of resource property values and the copy loop dynamically applies the values to the resource's
properties.
There is one issue to be aware of if you use the dynamic approach. To demonstrate the issue, let's take a look at a
typical array of property values. In this example the values for our properties are stored in a variable. Notice we
have two arrays here—one named firstProperty and one named secondProperty .
 "variables": {
"firstProperty": [
{
"name": "A",
"type": "typeA"
},
{
"name": "B",
"type": "typeB"
},
{
"name": "C",
"type": "typeC"
}
],
"secondProperty": [
"one","two", "three"
]
}

Now let's take a look at the way we access the properties in the variable using a copy loop.

{
"$schema": "https://schema.management.azure.com/schemas/2015-01-01/deploymentTemplate.json#",
"contentVersion": "1.0.0.0",
...
"copy": {
"name": "copyLoop1",
"count": "[length(variables('firstProperty'))]"
},
...
"properties": {
"name": { "value": "[variables('firstProperty')[copyIndex()].name]" },
"type": { "value": "[variables('firstProperty')[copyIndex()].type]" },
"number": { "value": "[variables('secondProperty')[copyIndex()]]" }
}
}

The copyIndex() function returns the current iteration of the copy loop, and we use that as an index into each of
the two arrays simultaneously.
This works fine when the two arrays are the same length. The issue arises if you've made a mistake and the two
arrays are different lengths—in this case your template will fail validation during deployment. You can avoid this
issue by including all your properties in a single object, because it is much easier to see when a value is missing.
For example, let's take a look another parameter object in which each element of the propertyObject array is the
union of the firstProperty and secondProperty arrays from earlier.
 "variables": {
"propertyObject": [
{
"name": "A",
"type": "typeA",
"number": "one"
},
{
"name": "B",
"type": "typeB",
"number": "two"
},
{
"name": "C",
"type": "typeC"
}
]
}

Notice the third element in the array? It's missing the number property, but it's much easier to notice that you've
missed it when you're authoring the parameter values this way.

Using a property object in a copy loop

This approach becomes even more useful when combined with the [serial copy loop][azure-resource-manager-
create-multiple], particularly for deploying child resources.
To demonstrate this, let's look at a template that deploys a network security group (NSG) with two security rules.
First, let's take a look at our parameters. When we look at our template we'll see that we've defined one parameter
named networkSecurityGroupsSettings that includes an array named securityRules . This array contains two JSON
objects that specify a number of settings for a security rule.
 {
"$schema": "https://schema.management.azure.com/schemas/2015-01-01/deploymentParameters.json#",
"contentVersion": "1.0.0.0",
"parameters":{
"networkSecurityGroupsSettings": {
"value": {
"securityRules": [
{
"name": "RDPAllow",
"description": "allow RDP connections",
"direction": "Inbound",
"priority": 100,
"sourceAddressPrefix": "*",
"destinationAddressPrefix": "10.0.0.0/24",
"sourcePortRange": "*",
"destinationPortRange": "3389",
"access": "Allow",
"protocol": "Tcp"
},
{
"name": "HTTPAllow",
"description": "allow HTTP connections",
"direction": "Inbound",
"priority": 200,
"sourceAddressPrefix": "*",
"destinationAddressPrefix": "10.0.1.0/24",
"sourcePortRange": "*",
"destinationPortRange": "80",
"access": "Allow",
"protocol": "Tcp"
}
]
}
}
}
}

Now let's take a look at our template. Our first resource named NSG1 deploys the NSG. Our second resource
named loop-0 performs two functions: first, it dependsOn the NSG so its deployment doesn't begin until NSG1 is
completed, and it is the first iteration of the sequential loop. Our third resource is a nested template that deploys
our security rules using an object for its parameter values as in the last example.

{
"$schema": "https://schema.management.azure.com/schemas/2015-01-01/deploymentTemplate.json#",
"contentVersion": "1.0.0.0",
"parameters": {
"networkSecurityGroupsSettings": {"type":"object"}
},
"variables": {},
"resources": [
{
"apiVersion": "2015-06-15",
"type": "Microsoft.Network/networkSecurityGroups",
"name": "NSG1",
"location":"[resourceGroup().location]",
"properties": {
"securityRules":[]
}
},
{
"apiVersion": "2015-01-01",
"type": "Microsoft.Resources/deployments",
"name": "loop-0",
"dependsOn": [
"NSG1"
 ],
"properties": {
"mode":"Incremental",
"parameters":{},
"template": {
"$schema": "https://schema.management.azure.com/schemas/2015-01-01/deploymentTemplate.json#",
"contentVersion": "1.0.0.0",
"parameters": {},
"variables": {},
"resources": [],
"outputs": {}
}
}
},
{
"apiVersion": "2015-01-01",
"type": "Microsoft.Resources/deployments",
"name": "[concat('loop-', copyIndex(1))]",
"dependsOn": [
"[concat('loop-', copyIndex())]"
],
"copy": {
"name": "iterator",
"count": "[length(parameters('networkSecurityGroupsSettings').securityRules)]"
},
"properties": {
"mode": "Incremental",
"template": {
"$schema": "https://schema.management.azure.com/schemas/2015-01-01/deploymentTemplate.json#",
"contentVersion": "1.0.0.0",
"parameters": {},
"variables": {},
"resources": [
{
"name": "[concat('NSG1/' ,
parameters('networkSecurityGroupsSettings').securityRules[copyIndex()].name)]",
"type": "Microsoft.Network/networkSecurityGroups/securityRules",
"apiVersion": "2016-09-01",
"location":"[resourceGroup().location]",
"properties":{
"description": "
[parameters('networkSecurityGroupsSettings').securityRules[copyIndex()].description]",
"priority":"
[parameters('networkSecurityGroupsSettings').securityRules[copyIndex()].priority]",
"protocol":"
[parameters('networkSecurityGroupsSettings').securityRules[copyIndex()].protocol]",
"sourcePortRange": "
[parameters('networkSecurityGroupsSettings').securityRules[copyIndex()].sourcePortRange]",
"destinationPortRange": "
[parameters('networkSecurityGroupsSettings').securityRules[copyIndex()].destinationPortRange]",
"sourceAddressPrefix": "
[parameters('networkSecurityGroupsSettings').securityRules[copyIndex()].sourceAddressPrefix]",
"destinationAddressPrefix": "
[parameters('networkSecurityGroupsSettings').securityRules[copyIndex()].destinationAddressPrefix]",
"access":"
[parameters('networkSecurityGroupsSettings').securityRules[copyIndex()].access]",
"direction":"
[parameters('networkSecurityGroupsSettings').securityRules[copyIndex()].direction]"
}
}
],
"outputs": {}
}
}
}
],
"outputs": {}
}
Let's take a closer look at how we specify our property values in the securityRules child resource. All of our
properties are referenced using the parameter() function, and then we use the dot operator to reference our
securityRules array, indexed by the current value of the iteration. Finally, we use another dot operator to reference
the name of the object.

Try the template

An example template is available on GitHub. To deploy the template, clone the repo and run the following Azure
CLI commands:

git clone https://github.com/mspnp/template-examples.git

cd template-examples/example3-object-param
az group create --location <location> --name <resource-group-name>
az deployment group create -g <resource-group-name> \
--template-uri https://raw.githubusercontent.com/mspnp/template-examples/master/example3-object-
param/deploy.json \
--parameters deploy.parameters.json

Next steps
Learn how to create a template that iterates through an object array and transforms it into a JSON schema. See
Implement a property transformer and collector in an Azure Resource Manager template
 Implement a property transformer and collector in an
Azure Resource Manager template
12/18/2020 • 6 minutes to read • Edit Online

In use an object as a parameter in an Azure Resource Manager template, you learned how to store resource
property values in an object and apply them to a resource during deployment. While this is a very useful way to
manage your parameters, it still requires you to map the object's properties to resource properties each time you
use it in your template.
To work around this, you can implement a property transform and collector template that iterates your object array
and transforms it into the JSON schema expected by the resource.

IMPORTANT
This approach requires that you have a deep understanding of Resource Manager templates and functions.

Let's take a look at how we can implement a property collector and transformer with an example that deploys a
network security group. The diagram below shows the relationship between our templates and our resources
within those templates:

Our calling template includes two resources:

A template link that invokes our collector template .
The network security group resource to deploy.
Our collector template includes two resources:
An anchor resource.
A template link that invokes the transform template in a copy loop.
Our transform template includes a single resource: an empty template with a variable that transforms our
source JSON to the JSON schema expected by our network security group resource in the main template .

Parameter object
We'll be using our securityRules parameter object from objects as parameters. Our transform template will
transform each object in the securityRules array into the JSON schema expected by the network security group
resource in our calling template .
 {
"$schema": "https://schema.management.azure.com/schemas/2015-01-01/deploymentParameters.json#",
"contentVersion": "1.0.0.0",
"parameters": {
"networkSecurityGroupsSettings": {
"value": {
"securityRules": [
{
"name": "RDPAllow",
"description": "allow RDP connections",
"direction": "Inbound",
"priority": 100,
"sourceAddressPrefix": "*",
"destinationAddressPrefix": "10.0.0.0/24",
"sourcePortRange": "*",
"destinationPortRange": "3389",
"access": "Allow",
"protocol": "Tcp"
},
{
"name": "HTTPAllow",
"description": "allow HTTP connections",
"direction": "Inbound",
"priority": 200,
"sourceAddressPrefix": "*",
"destinationAddressPrefix": "10.0.1.0/24",
"sourcePortRange": "*",
"destinationPortRange": "80",
"access": "Allow",
"protocol": "Tcp"
}
]
}
}
}
}

Let's look at our transform template first.

Transform template
Our transform template includes two parameters that are passed from the collector template :
source is an object that receives one of the property value objects from the property array. In our example,
each object from the "securityRules" array will be passed in one at a time.
state is an array that receives the concatenated results of all the previous transforms. This is the collection of
transformed JSON.
Our parameters look like this:

{
"$schema": "https://schema.management.azure.com/schemas/2015-01-01/deploymentTemplate.json#",
"contentVersion": "1.0.0.0",
"parameters": {
"source": { "type": "object" },
"state": {
"type": "array",
"defaultValue": [ ]
}
},

Our template also defines a variable named instance . It performs the actual transform of our source object into
the required JSON schema:

"variables": {
"instance": [
{
"name": "[parameters('source').name]",
"properties":{
"description": "[parameters('source').description]",
"protocol": "[parameters('source').protocol]",
"sourcePortRange": "[parameters('source').sourcePortRange]",
"destinationPortRange": "[parameters('source').destinationPortRange]",
"sourceAddressPrefix": "[parameters('source').sourceAddressPrefix]",
"destinationAddressPrefix": "[parameters('source').destinationAddressPrefix]",
"access": "[parameters('source').access]",
"priority": "[parameters('source').priority]",
"direction": "[parameters('source').direction]"
}
}
]

},

Finally, the output of our template concatenates the collected transforms of our state parameter with the current
transform performed by our instance variable:

"resources": [],
"outputs": {
"collection": {
"type": "array",
"value": "[concat(parameters('state'), variables('instance'))]"
}

Next, let's take a look at our collector template to see how it passes in our parameter values.

Collector template
Our collector template includes three parameters:
source is our complete parameter object array. It's passed in by the calling template . This has the same name
as the source parameter in our transform template but there is one key difference that you may have
already noticed: this is the complete array, but we only pass one element of this array to the transform
template at a time.
transformTemplateUri is the URI of our transform template . We're defining it as a parameter here for
template reusability.
state is an initially empty array that we pass to our transform template . It stores the collection of
transformed parameter objects when the copy loop is complete.
Our parameters look like this:

"parameters": {
"source": { "type": "array" },
"transformTemplateUri": { "type": "string" },
"state": {
"type": "array",
"defaultValue": [ ]
}

Next, we define a variable named count . Its value is the length of the source parameter object array:
 "variables": {
"count": "[length(parameters('source'))]"
},

As you might suspect, we use it for the number of iterations in our copy loop.
Now let's take a look at our resources. We define two resources:
loop-0 is the zero-based resource for our copy loop.
loop- is concatenated with the result of the copyIndex(1) function to generate a unique iteration-based name
for our resource, starting with 1 .

Our resources look like this:

"resources": [
{
"type": "Microsoft.Resources/deployments",
"apiVersion": "2015-01-01",
"name": "loop-0",
"properties": {
"mode": "Incremental",
"parameters": { },
"template": {
"$schema": "https://schema.management.azure.com/schemas/2015-01-01/deploymentTemplate.json#",
"contentVersion": "1.0.0.0",
"parameters": { },
"variables": { },
"resources": [ ],
"outputs": {
"collection": {
"type": "array",
"value": "[parameters('state')]"
}
}
}
}
},
{
"type": "Microsoft.Resources/deployments",
"apiVersion": "2015-01-01",
"name": "[concat('loop-', copyindex(1))]",
"copy": {
"name": "iterator",
"count": "[variables('count')]",
"mode": "serial"
},
"dependsOn": [
"loop-0"
],
"properties": {
"mode": "Incremental",
"templateLink": { "uri": "[parameters('transformTemplateUri')]" },
"parameters": {
"source": { "value": "[parameters('source')[copyindex()]]" },
"state": { "value": "[reference(concat('loop-', copyindex())).outputs.collection.value]" }
}
}
}
],

Let's take a closer look at the parameters we're passing to our transform template in the nested template. Recall
from earlier that our source parameter passes the current object in the source parameter object array. The
state parameter is where the collection happens, because it takes the output of the previous iteration of our copy
loop—notice that the reference() function uses the copyIndex() function with no parameter to reference the
name of our previous linked template object—and passes it to the current iteration.

Finally, the output of our template returns the output of the last iteration of our transform template :

"outputs": {
"result": {
"type": "array",
"value": "[reference(concat('loop-', variables('count'))).outputs.collection.value]"
}
}

It may seem counterintuitive to return the output of the last iteration of our transform template to our calling
template because it appeared we were storing it in our source parameter. However, remember that it's the last
iteration of our transform template that holds the complete array of transformed property objects, and that's
what we want to return.
Finally, let's take a look at how to call the collector template from our calling template .

Calling template
Our calling template defines a single parameter named networkSecurityGroupsSettings :

...
"parameters": {
"networkSecurityGroupsSettings": {
"type": "object"
}

Next, our template defines a single variable named collectorTemplateUri :

"variables": {
"collectorTemplateUri": "[uri(deployment().properties.templateLink.uri, 'collector.template.json')]"
}

As you would expect, this is the URI for the collector template that will be used by our linked template resource:

{
"apiVersion": "2015-01-01",
"name": "collector",
"type": "Microsoft.Resources/deployments",
"properties": {
"mode": "Incremental",
"templateLink": {
"uri": "[variables('collectorTemplateUri')]",
"contentVersion": "1.0.0.0"
},
"parameters": {
"source" : {"value": "[parameters('networkSecurityGroupsSettings').securityRules]"},
"transformTemplateUri": { "value": "[uri(deployment().properties.templateLink.uri,
'transform.json')]"}
}
}
}

We pass two parameters to the collector template :

source is our property object array. In our example, it's our networkSecurityGroupsSettings parameter.
 transformTemplateUri is the variable we just defined with the URI of our collector template .
Finally, our Microsoft.Network/networkSecurityGroups resource directly assigns the output of the collector linked
template resource to its securityRules property:

{
"apiVersion": "2015-06-15",
"type": "Microsoft.Network/networkSecurityGroups",
"name": "networkSecurityGroup1",
"location": "[resourceGroup().location]",
"properties": {
"securityRules": "[reference('collector').outputs.result.value]"
}
}
],
"outputs": {
"instance":{
"type": "array",
"value": "[reference('collector').outputs.result.value]"
}

Try the template

An example template is available on GitHub. To deploy the template, clone the repo and run the following Azure CLI
commands:

git clone https://github.com/mspnp/template-examples.git

cd template-examples/example4-collector
az group create --location <location> --name <resource-group-name>
az deployment group create -g <resource-group-name> \
--template-uri https://raw.githubusercontent.com/mspnp/template-examples/master/example4-
collector/deploy.json \
--parameters deploy.parameters.json
 Building solutions for high availability using
Availability Zones
12/18/2020 • 7 minutes to read • Edit Online

Microsoft Azure global infrastructure is designed and constructed at every layer to deliver the highest levels of
redundancy and resiliency to its customers. Azure infrastructure is composed of geographies, regions, and
Availability Zones, which limit the blast radius of a failure and therefore limit potential impact to customer
applications and data. The Azure Availability Zones construct was developed to provide a software and networking
solution to protect against datacenter failures and to provide increased high availability (HA) to our customers.
Availability Zones are unique physical locations within an Azure region. Each zone is made up of one or more
datacenters with independent power, cooling, and networking. The physical separation of Availability Zones within
a region limits the impact to applications and data from zone failures, such as large-scale flooding, major storms
and superstorms, and other events that could disrupt site access, safe passage, extended utilities uptime, and the
availability of resources. Availability Zones and their associated datacenters are designed such that if one zone is
compromised, the services, capacity, and availability are supported by the other Availability Zones in the region.
Availability Zones can be used to spread a solution across multiple zones within a region, allowing for an
application to continue functioning when one zone fails. With Availability Zones, Azure offers industry best 99.99%
Virtual Machine (VM) uptime service-level agreement (SLA). Zone-redundant services replicate your services and
data across Availability Zones to protect from single points of failure.

For additional information on Availability Zones, including service support by region and pricing, refer to What are
Availability Zones in Azure? in Microsoft Azure Documentation.

Delivering reliability in Azure

Designing solutions that continue to function in spite of failure is key to improving the reliability of a solution. In
cloud-based solutions, building to survive failure is a shared responsibility. This can be viewed at three levels: a
resilient foundation, resilient services, and resilient applications. The foundation is the Microsoft investment in the
platform, including Availability Zones. On top of this foundation are the Azure services that customers can enable
to support high availability, such as zone-redundant storage (ZRS), which replicates data across zones. The
customer builds applications upon the enabled services supported by the foundation. The applications should be
architected to support resiliency.

Your applications
Your app or workload architecture
Resilient services
Azure capabilities you enable as needed
Resilient foundation
Azure capabilities built into the platform

When architecting for resilience, all three layers-foundation, services, and applications-should be considered to
achieve the highest level of reliability. Since a solution can be made up of many components, each component
should be designed for reliability.

Zonal vs. zone-redundant architecture

An Availability Zone in an Azure region is a combination of a fault domain and an update domain. For example, if
you create three or more VMs across three zones in an Azure region, your VMs are effectively distributed across
three fault domains and three update domains. The Azure platform recognizes this distribution across update
domains to take care that VMs in different zones are not updated at the same time.
Azure services supporting Availability Zones fall into two categories: zonal and zone redundant. Customer
workloads can be categorized to utilize either architecture scenario to meet application performance and durability
requirements.
With zonal architecture, a resource can be deployed to a specific, self-selected Availability Zone to achieve more
stringent latency or performance requirements. Resiliency is self-architected by replicating applications and data to
one or more zones within the region. You can choose specific Availability Zones for synchronous replication,
providing high availability, or asynchronous replication, providing backup or cost advantage. You can pin
resources-for example, virtual machines, managed disks, or standard IP addresses-to a specific zone, allowing for
increased resilience by having one or more instances of resources spread across zones.
With zone-redundant architecture, the Azure platform automatically replicates the resource and data across zones.
Microsoft manages the delivery of high availability, since Azure automatically replicates and distributes instances
within the region.
A failure to a zone affects zonal and zone-redundant services differently. In the case of a zone failure, the zonal
services in the failed zone become unavailable until the zone has recovered. By architecting your solutions to use
replicated VMs in zones, you can protect your applications and data from a zone becoming unavailable-for
example, due to a power outage. If one zone is compromised, replicated apps and data are instantly available in
another zone.
Zonal architecture applies to a specific resource, typically an infrastructure as a service (IaaS) resource, like a VM or
managed disk, as illustrated.
For example, zonal load balancer, VM, managed disks, VM scale sets.
In the illustration, each VM and load balancer (LB) are deployed to a specific zone.
With zone-redundant services, the distribution of the workload is a feature of the service and is handled by Azure.
Azure automatically replicates the resource across zones without requiring your intervention. ZRS, for example,
replicates the data across three zones so a zone failure does not impact the HA of the data.
The following illustration is of a zone-redundant load balancer.

For example, zone-redundant load balancer, Azure Application Gateway, Azure Service Bus, virtual private network
(VPN), zone-redundant storage, Azure ExpressRoute, Azure Event Hubs, Azure Cosmos DB.
A few resources, like the load balancer and subnets, support both zonal and zone-redundant deployments. An
important consideration in HA is distributing the traffic effectively across resources in the different Availability
Zones. For information on how Availability Zones apply to the load balancer resources for both zonal and zone-
redundant resources, refer to Standard Load Balancer and Availability Zones.
The following is a summary of the zonal (Z) and zone-redundant (ZR) Azure services.
 Analytics Management and governance
Azure Data Explorer (ZR) Azure Traffic Manager (ZR)
Azure Event Hubs (ZR) Networking
Compute Azure Load Balancer (Z, ZR)
Linux vir tual machines (Z) VPN gateway (ZR)
Windows vir tual machines (Z) Azure ExpressRoute (ZR)
Vir tual machine scale sets (Z, ZR) Azure Application Gateway (ZR)
Azure App Ser vice Environments (Z) Azure Firewall (ZR)
Containers Azure Vir tual WAN (ZR)
Azure Kubernetes Ser vice (AKS) (Z) Security
Azure Ser vice Fabric (Z) Azure Active Director y Domain Ser vices (ZR)
Databases Storage
Azure SQL Database (ZR) Azure Data Lake Storage (ZRS)
Azure Cache for Redis (Z, ZR) Azure Blob storage (ZRS, ZR)
Azure Cosmos DB (ZR) Azure Managed Disks (Z)
DevOps Additional capabilities
Azure DevOps (ZR) Azure Premium Files (ZRS)
Identity Zone-redundant storage (ZRS)
Azure Active Director y Domain Ser vices (ZR) Standard IP address (ZR)
Integration Azure Traffic Analytics (ZR)
Azure Event Grid (ZR)
Azure Ser vice Bus (ZR)

VMs supporting AZs: AV2-series, B-series, DSv2-series, DSv3-series, Dv2-series, Dv3-series, ESv3-series, Ev3-series,
F-series, FS-series, FSv2-series, M-series.
For a list of Azure services that support Availability Zones, per Azure region, refer to the Availability Zones
documentation.

SLA offered by Availability Zones

With Availability Zones, Azure offers industry best 99.99% VM uptime SLA. The full Azure SLA explains the
guaranteed availability of Azure as a whole.
The following diagram illustrates the different levels of HA offered by a single VM, Availability Sets, and Availability
Zones.
Using a VM workload as an example, a single VM has an SLA of 99.9%. This means the VM will be available 99.9%
of the time. Within a single datacenter, the use of Availability Sets can increase the level of SLA to 99.95% by
protecting a set of VMs, ensuring they will not all be on the same hardware. Within a region, VM workloads can be
distributed across Availability Zones to increase the SLA to 99.99%. For more information, refer to Availability
options for VMs in Azure.
Every organization has unique requirements, and you should design your applications to best meet your complex
business needs. Defining a target SLA will make it possible to evaluate whether the architecture meets your
business requirements. Some things to consider include:
What are the availability requirements?
How much downtime is acceptable?
How much will potential downtime cost your business?
How much should you invest in making the application highly available?
What are the data backup requirements?
What are the data replication requirements?
What are the monitoring requirements?
Does your application have specific latency requirements?
For additional guidance, refer to Microsoft Azure Well-Architected Framework define requirements.
Depending on the availability needs of an application, the cost and design complexity will vary. When building for a
VM workload, there will be a cost associated with each VM. For example, two VMs per zone across three active
zones will have a cost for six VMs. For pricing of VM workloads, refer to the Azure pricing calculator.
 IaaS: Web application with relational database
12/18/2020 • 12 minutes to read • Edit Online

Availability Zones are unique physical locations within an Azure region. Each zone is made up of one or more
datacenters with independent power, cooling, and networking. The physical separation of availability zones within a
region limits the impact to applications and data from zone failures, such as largescale flooding, major/super-
storms, and other events that would disrupt site access, safe passage, extended utilities uptime, and the availability
of resources. This reference architecture shows best practices for applying Availability Zones to a web application
and Microsoft SQL Server database hosted on virtual machines (VMs) known as a zonal deployment.
This approach is used in high availability (HA) scenarios with resiliency as the most significant concern. With HA
architecture there is a balance between high resilience, low latency, and cost. This architecture uses redundant
resources spread across zones to provide high resiliency. Traffic can be routed between zones to minimize the
impact of a zonal failure. If a zone does fail, resources in the zone absorb the traffic until the zone recovers. This
provides a high level of resilience.
This architecture provides an efficient use of resources as most of the resources are actively used. All resources are
used in handling the requests other than the passive SQL Server. The passive SQL Server only becomes active in
the case of a failure of the current SQL Server.
The zone-redundant application gateway and zone-redundant load balancer distribute the traffic to the available
resources. Cross-zone routing will have increased latency relative to traffic routed within the zone.

Download a Visio file of this architecture.

Architecture
The architecture uses resources spread across multiple zones to provide a high availability (HA) architecture for
hosting an Infrastructure as a Service (IaaS) web application and SQL Server database. A zone-redundant
application gateway routes traffic to virtual machines within the web tier. A zone-redundant load balancer routes
traffic from the virtual machines in the web tier to the active SQL Server. In the case of a zonal failure, the
application gateway will route to virtual machines in the other available zones.
If the active SQL Server becomes unavailable, either due to a zone failure or local failure, a passive SQL Server will
become the active SQL Server. The zone-redundant load balancer will detect the failover of SQL Server and route
traffic to the newly active SQL Server.
The following illustrates a failure of Zone 1.

The application gateway is zone-redundant and is not impacted by the failure of Zone 1. The application gateway
uses health probes to determine the available virtual machines. With Zone 1 unavailable, the zone-redundant
application gateway only routes traffic to the remaining two zones. The zone-redundant load balancer is also not
impacted by the failure of Zone 1 and uses health probes to determine the location of the active SQL Server. In this
example, the load balancer will detect that the active SQL Server is in Zone 3 and will then route traffic to the SQL
Server instance in Zone 3.
Spreading resources across Availability Zones also protects an application from planned maintenance. When VMs
are distributed across three Availability Zones, effectively they are spread across three update domains. The Azure
platform recognizes this distribution across update domains to ensure that VMs in different zones are not updated
at the same time.
By replicating VMs across Availability Zones, you can protect your applications and data from a zonal failure. This
allows Azure to support the industry best 99.99% VM uptime service-level agreement (SLA). See Building solutions
for high availability using Availability Zones.
The architecture has the following components.
General
Resource group . Resource groups are used to group Azure resources so they can be managed by lifetime,
owner, or other criteria.
Availability Zones . Availability Zones are physical locations within an Azure region. Each zone consists of
one or more datacenters with independent power, cooling, and networking. By placing VMs across zones, the
application becomes resilient to failures within a zone.
Networking and load balancing
Vir tual network and subnets . Every Azure VM is deployed into a virtual network (VNet) that can be
segmented into subnets. Create a separate subnet for each tier.
Azure Application Gateway . Azure Application Gateway is a layer 7 load balancer. In this architecture, a
 zone-redundant application gateway routes HTTP requests to the web front end. Application Gateway also
provides a Web Application Firewall (WAF) that protects the application from common exploits and
vulnerabilities. The v2 SKU of Application Gateway supports cross-zone redundancy. A single Application
Gateway deployment can run multiple instances of the gateway. For production workloads, run at least two
instances. For more information, see Autoscaling and Zone-redundant Application Gateway v2 and How
does Application Gateway support high availability and scalability?.
Azure Load Balancer . Azure Load Balancer is a layer 4 load balancer. In this architecture, a zone-redundant
Azure Standard Load Balancer directs network traffic from the web tier to SQL Server. Because a zone-
redundant load balancer is not pinned to a specific zone, the application will continue to distribute the
network traffic in the case of a zonal failure. A zone-redundant load balancer is used to provide availability in
the case the active SQL Server becomes unavailable. The Standard SKU of Azure Load Balancer supports
cross-zone redundancy. For more information, see Standard Load Balancer and Availability Zones.
Network security groups (NSGs) . A network security group is used to restrict network traffic within the
virtual network. In this architecture, the web tier only accepts traffic from the public IP endpoint. Additionally,
the database tier does not accept traffic from any subnet other than the web-tier subnet.
DDoS protection . The Azure platform provides protection against distributed denial of service (DDoS)
attacks. For additional protection, we recommend using DDoS Protection Standard, which has enhanced
DDoS mitigation features. See Security considerations.
Bastion . Azure Bastion provides secure and seamless Remote Desktop Protocol (RDP) and Secure Shell
(SSH) access to the VMs within the VNet. This provides access while limiting the exposed public IP addresses
of the VMs with the VNet. Azure Bastion provides a cost-effective alternative to a provisioned VM to
provide access to all VMs within the same virtual network.
Microsoft SQL Server
SQL Ser ver Always On availability group . SQL Server Always On availability group provides high
availability at the data tier by enabling replication and failover. It uses Windows Server Failover Cluster
(WSFC) technology for failover.
Cloud Witness . A failover cluster requires more than half of its nodes to be running, which is known as
having quorum. If the cluster has just two nodes, a network partition could cause each node to think it's the
primary node. In that case, you need a witness to break ties and establish quorum. A witness is a resource
such as a shared disk that can act as a tie breaker to establish quorum. Cloud Witness is a type of witness
that uses Azure Blob Storage. The Azure Blob Storage must use Zone Redundant Storage (ZRS) to not be
impacted by zonal failure.
To learn more about the concept of quorum, see Understanding cluster and pool quorum. For more
information about Cloud Witness, see Deploy a Cloud Witness for a Failover Cluster.

Recommendations
Your requirements might differ from the architecture described here. Use these recommendations as a starting
point.
For recommendations on configuring the VMs, see Run a Windows VM on Azure.
For more information about designing virtual networks and subnets, see Plan and design Azure Virtual Networks.
Network security groups
Use NSG rules to restrict traffic between tiers. In the architecture shown earlier, only the web tier can communicate
directly with the database tier. To enforce this rule, the database tier should block all incoming traffic except for the
web-tier subnet.
Deny all inbound traffic from the virtual network. (Use the VIRTUAL_NETWORK tag in the rule.)
 Allow inbound traffic from the web-tier subnet.
Allow inbound traffic from the database-tier subnet itself. This rule allows communication between the
database VMs, which is needed for database replication and failover.
Create rules 2 – 3 with higher priority than the first rule, so they override it.
SQL Server Always On availability groups
We recommend Always On availability groups for Microsoft SQL Server high availability. Other tiers connect to the
database through an availability group listener. The listener enables a SQL client to connect without knowing the
name of the physical instance of SQL Server. VMs that access the database must be joined to the domain. The client
(in this case, another tier) uses DNS to resolve the listener's virtual network name into IP addresses.
Configure the SQL Server Always On availability group as follows:
Create a Windows Server Failover Clustering (WSFC) cluster, a SQL Server Always On availability group, and
a primary replica. For more information, see Getting Started with Always On availability groups.
Create an internal load balancer with a static private IP address.
Create an availability group listener and map the listener's DNS name to the IP address of an internal load
balancer.
Create a load balancer rule for the SQL Server listening port (TCP port 1433 by default). The load balancer
rule must enable floating IP, also called Direct Server Return. This causes the VM to reply directly to the client,
which enables a direct connection to the primary replica.

NOTE
When floating IP is enabled, the front-end port number must be the same as the back-end port number in the load balancer
rule.

When a SQL client tries to connect, the load balancer routes the connection request to the primary replica. If there
is a failover to another replica, the load balancer automatically routes new requests to a new primary replica. For
more information, see Configure a load balancer for an availability group on Azure SQL Server VMs.
During a failover, existing client connections are closed. After the failover completes, new connections will be routed
to the new primary replica.
If your application makes significantly more reads than writes, you can offload some of the read-only queries to a
secondary replica. See Connect to a read-only replica.
Test your deployment by forcing a manual failover of the availability group.

Availability considerations
Availability Zones provide high resilience within a single region. If you need even higher availability, consider
replicating the application across two regions, using Azure Traffic Manager for failover. For more information, see
Run an N-tier application in multiple Azure regions for high availability.
Not all regions support Availability Zones, and not all VM sizes are supported in all zones. Run the following Azure
CLI command to find the supported zones for each VM size within a region:

az vm list-skus --resource-type virtualMachines --zone false --location

\<location\> \\

\--query "[].{Name:name, Zones:locationInfo[].zones[] \| join(','\@)}" -o tabl

Virtual machine scale sets automatically use placement groups, which act as an implicit availability set. For more
information about placement groups, see Working with large virtual machine scale sets.
Health probes
Azure Application Gateway and Azure Load Balancer both use health probes to monitor the availability of VM
instances.
Application Gateway always uses an HTTP probe.
Load Balancer can test either HTTP or TCP. Generally, if a VM runs an HTTP server, use an HTTP probe.
Otherwise, use TCP.
If a probe can't reach an instance within a timeout period, the gateway or load balancer stops sending traffic to that
VM. The probe continues to check and will return the VM to the back-end pool if the VM becomes available again.
HTTP probes send an HTTP GET request to a specified path and listen for an HTTP 200 response. This path can be
the root path ("/"), or a health-monitoring endpoint that implements some custom logic to check the health of the
application. The endpoint must allow anonymous HTTP requests.
For more information about health probes, see:
Load Balancer health probes
Application Gateway health monitoring overview
For considerations about designing a health probe endpoint, see Health Endpoint Monitoring pattern.

Cost considerations
Use the Azure Pricing Calculator to estimates costs. Here are some other considerations.
Virtual machine scale sets
Virtual machine scale sets are available on all Windows VM sizes. You are only charged for the Azure VMs you
deploy, and any additional underlying infrastructure resources consumed, such as storage and networking. There
are no incremental charges for the virtual machine scale sets service.
For single VMs pricing options, see Windows VMs pricing.
SQL Server
If you choose Azure SQL DBaaS, you can save on cost because you don't need to configure an Always On
availability group and domain controller machines. There are several deployment options starting from single
database up to managed instance, or elastic pools. For more information, see Azure SQL pricing.
For SQL server VMs pricing options, see SQL VMs pricing.
Azure Load Balancer
You are charged only for the number of configured load-balancing and outbound rules. Inbound NAT rules are free.
There is no hourly charge for the Standard Load Balancer when no rules are configured.
For more information, see the cost section in Azure Architecture Framework.
Azure Application Gateway
The Application Gateway should be provisioned with the v2 SKU and can span multiple Availability Zones. With the
v2 SKU, the pricing model is driven by consumption and has two components: Hourly fixed price and a
consumption-based cost.
For more information, see the pricing section in Autoscaling and Zone-redundant Application Gateway v2.
Security considerations
Virtual networks are a traffic isolation boundary in Azure. By default, VMs in one virtual network can't
communicate directly with VMs in a different virtual network. However, you can explicitly connect virtual networks
by using virtual network peering.
Network security groups
Use network security groups (NSGs) to restrict traffic to and from the internet. For more information, see Microsoft
cloud services and network security.
DMZ
Consider adding a network virtual appliance (NVA) to create a DMZ between the internet and the Azure virtual
network. NVA is a generic term for a virtual appliance that can perform network-related tasks, such as firewall,
packet inspection, auditing, and custom routing. For more information, see Network DMZ between Azure and an
on-premises datacenter.
Encryption
Encrypt sensitive data at rest and use Azure Key Vault to manage the database encryption keys. Key Vault can store
encryption keys in hardware security modules (HSMs). For more information, see Configure Azure Key Vault
Integration for SQL Server on Azure VMs. It's also recommended to store application secrets, such as database
connection strings, in Key Vault.
DDoS protection
The Azure platform provides basic DDoS protection by default. This basic protection is targeted at protecting the
Azure infrastructure. Although basic DDoS protection is automatically enabled, we recommend using DDoS
Protection Standard. Standard protection uses adaptive tuning, based on your application's network traffic patterns,
to detect threats. This allows it to apply mitigations against DDoS attacks that might go unnoticed by the
infrastructure-wide DDoS policies. Standard protection also provides alerting, telemetry, and analytics through
Azure Monitor. For more information, see Azure DDoS Protection: Best practices and reference architectures.

Next steps
Microsoft Learn module: Tour the N-tier architecture style
 Manage identity in multitenant applications
12/18/2020 • 3 minutes to read • Edit Online

This series of articles describes best practices for multitenancy, when using Azure AD for authentication and
identity management.

Sample code
When you're building a multitenant application, one of the first challenges is managing user identities, because
now every user belongs to a tenant. For example:
Users sign in with their organizational credentials.
Users should have access to their organization's data, but not data that belongs to other tenants.
An organization can sign up for the application, and then assign application roles to its members.
Azure Active Directory (Azure AD) has some great features that support all of these scenarios.
To accompany this series of articles, we created a complete end-to-end implementation of a multitenant
application. The articles reflect what we learned in the process of building the application. To get started with the
application, see the GitHub readme.

Introduction
Let's say you're writing an enterprise SaaS application to be hosted in the cloud. Of course, the application will have
users:

But those users belong to organizations:

Example: Tailspin sells subscriptions to its SaaS application. Contoso and Fabrikam sign up for the app. When Alice (
alice@contoso ) signs in, the application should know that Alice is part of Contoso.

Alice should have access to Contoso data.

Alice should not have access to Fabrikam data.
This guidance will show you how to manage user identities in a multitenant application, using Azure Active
Directory (Azure AD) to handle sign-in and authentication.

What is multitenancy?
A tenant is a group of users. In a SaaS application, the tenant is a subscriber or customer of the application.
Multitenancy is an architecture where multiple tenants share the same physical instance of the app. Although
tenants share physical resources (such as VMs or storage), each tenant gets its own logical instance of the app.
Typically, application data is shared among the users within a tenant, but not with other tenants.

Compare this architecture with a single-tenant architecture, where each tenant has a dedicated physical instance. In
a single-tenant architecture, you add tenants by spinning up new instances of the app.

Multitenancy and horizontal scaling

To achieve scale in the cloud, it's common to add more physical instances. This is known as horizontal scaling or
scaling out. Consider a web app. To handle more traffic, you can add more server VMs and put them behind a load
balancer. Each VM runs a separate physical instance of the web app.

Any request can be routed to any instance. Together, the system functions as a single logical instance. You can tear
down a VM or spin up a new VM, without affecting users. In this architecture, each physical instance is multitenant,
and you scale by adding more instances. If one instance goes down, it should not affect any tenant.

Identity in a multitenant app

In a multitenant app, you must consider users in the context of tenants.
Authentication
Users sign into the app with their organization credentials. They don't have to create new user profiles for the
app.
Users within the same organization are part of the same tenant.
When a user signs in, the application knows which tenant the user belongs to.
Authorization
When authorizing a user's actions (say, viewing a resource), the app must take into account the user's tenant.
Users might be assigned roles within the application, such as "Admin" or "Standard User". Role assignments
should be managed by the customer, not by the SaaS provider.
Example. Alice, an employee at Contoso, navigates to the application in her browser and clicks the "Log in" button.
She is redirected to a sign-in screen where she enters her corporate credentials (username and password). At this
point, she is logged into the app as alice@contoso.com . The application also knows that Alice is an admin user for
this application. Because she is an admin, she can see a list of all the resources that belong to Contoso. However,
she cannot view Fabrikam's resources, because she is an admin only within her tenant.
In this guidance, we'll look specifically at using Azure AD for identity management.
We assume the customer stores their user profiles in Azure AD (including Office365 and Dynamics CRM
tenants)
Customers with on-premises Active Directory can use Azure AD Connect to sync their on-premises Active
Directory with Azure AD. If a customer with on-premises Active Directory cannot use Azure AD Connect (due to
corporate IT policy or other reasons), the SaaS provider can federate with the customer's directory through
Active Directory Federation Services (AD FS). This option is described in Federating with a customer's AD FS.
This guidance does not consider other aspects of multitenancy such as data partitioning, per-tenant configuration,
and so forth.
Next
 The Tailspin scenario
12/18/2020 • 2 minutes to read • Edit Online

Sample code
Tailspin is a fictional company that is developing a SaaS application named Surveys. This application enables
organizations to create and publish online surveys.
An organization can sign up for the application.
After the organization is signed up, users can sign into the application with their organizational credentials.
Users can create, edit, and publish surveys.

NOTE
To get started with the application, see the GitHub readme.

Users can create, edit, and view surveys

An authenticated user can view all the surveys that he or she has created or has contributor rights to, and create
new surveys. Notice that the user is signed in with his organizational identity, bob@contoso.com .

This screenshot shows the Edit Survey page:

Users can also view any surveys created by other users within the same tenant.

Survey owners can invite contributors

When a user creates a survey, he or she can invite other people to be contributors on the survey. Contributors can
edit the survey, but cannot delete or publish it.
A user can add contributors from other tenants, which enables cross-tenant sharing of resources. In this
screenshot, Bob ( bob@contoso.com ) is adding Alice ( alice@fabrikam.com ) as a contributor to a survey that Bob
created.
When Alice logs in, she sees the survey listed under "Surveys I can contribute to".

Note that Alice signs into her own tenant, not as a guest of the Contoso tenant. Alice has contributor permissions
only for that survey — she cannot view other surveys from the Contoso tenant.

Architecture
The Surveys application consists of a web front end and a web API backend. Both are implemented using ASP.NET
Core.
The web application uses Azure Active Directory (Azure AD) to authenticate users. The web application also calls
Azure AD to get OAuth 2 access tokens for the Web API. Access tokens are cached in Azure Cache for Redis. The
cache enables multiple instances to share the same token cache (for example, in a server farm).
The diagram shows components in boxes, interacting with other components via two-way arrows. The Surveys
web application authenticates with Azure AD to get access tokens for the web API, and caches the tokens in the
Azure Cache for Redis access token cache.
Next
 Authenticate using Azure AD and OpenID Connect
12/18/2020 • 7 minutes to read • Edit Online

Sample code
The Surveys application uses the OpenID Connect (OIDC) protocol to authenticate users with Azure Active
Directory (Azure AD). The Surveys application uses ASP.NET Core, which has built-in middleware for OIDC. The
following diagram shows what happens when the user signs in, at a high level.

1. The user clicks the "sign in" button in the app. This action is handled by an MVC controller.
2. The MVC controller returns a ChallengeResult action.
3. The middleware intercepts the ChallengeResult and creates a 302 response, which redirects the user to the
Azure AD sign-in page.
4. The user authenticates with Azure AD.
5. Azure AD sends an ID token to the application.
6. The middleware validates the ID token. At this point, the user is now authenticated inside the application.
7. The middleware redirects the user back to application.

Register the app with Azure AD

To enable OpenID Connect, the SaaS provider registers the application inside their own Azure AD tenant.
To register the application, follow the steps in Quickstart: Register an application with the Microsoft identity
platform.
To enable this functionality in the sample Surveys application, see the GitHub readme. Note the following:
For a multitenant application, you must configure the multitenanted option explicitly. This enables other
organizations to access the application.
The reply URL is the URL where Azure AD will send OAuth 2.0 responses. When using the ASP.NET Core, this
needs to match the path that you configure in the authentication middleware (see next section).

Configure the auth middleware

This section describes how to configure the authentication middleware in ASP.NET Core for multitenant
authentication with OpenID Connect.
In your startup class, add the OpenID Connect middleware:

app.AddAuthentication().AddOpenIdConnect(options => {
options.ClientId = configOptions.AzureAd.ClientId;
options.ClientSecret = configOptions.AzureAd.ClientSecret; // for code flow
options.Authority = Constants.AuthEndpointPrefix;
options.ResponseType = OpenIdConnectResponseType.CodeIdToken;
options.PostLogoutRedirectUri = configOptions.AzureAd.PostLogoutRedirectUri;
options.SignInScheme = CookieAuthenticationDefaults.AuthenticationScheme;
options.TokenValidationParameters = new TokenValidationParameters { ValidateIssuer = false };
options.Events = new SurveyAuthenticationEvents(configOptions.AzureAd, loggerFactory);
});

Notice that some of the settings are taken from runtime configuration options. Here's what the middleware options
mean:
ClientId . The application's client ID, which you got when you registered the application in Azure AD.
Authority . For a multitenant application, set this to https://login.microsoftonline.com/common/ . This is the URL
for the Azure AD common endpoint, which enables users from any Azure AD tenant to sign in. For more
information about the common endpoint, see this blog post.
In TokenValidationParameters , set ValidateIssuer to false. That means the app will be responsible for
validating the issuer value in the ID token. (The middleware still validates the token itself.) For more information
about validating the issuer, see Issuer validation.
PostLogoutRedirectUri . Specify a URL to redirect users after the sign out. This should be a page that allows
anonymous requests — typically the home page.
SignInScheme . Set this to CookieAuthenticationDefaults.AuthenticationScheme . This setting means that after
the user is authenticated, the user claims are stored locally in a cookie. This cookie is how the user stays logged
in during the browser session.
Events. Event callbacks; see Authentication events.
Also add the Cookie Authentication middleware to the pipeline. This middleware is responsible for writing the user
claims to a cookie, and then reading the cookie during subsequent page loads.

app.AddAuthentication().AddCookie(options => {
options.AutomaticAuthenticate = true;
options.AutomaticChallenge = true;
options.AccessDeniedPath = "/Home/Forbidden";
options.CookieSecure = CookieSecurePolicy.Always;

// The default setting for cookie expiration is 14 days. SlidingExpiration is set to true by default
options.ExpireTimeSpan = TimeSpan.FromHours(1);
options.SlidingExpiration = true;
});

Initiate the authentication flow

To start the authentication flow in ASP.NET MVC, return a ChallengeResult from the controller:
 [AllowAnonymous]
public IActionResult SignIn()
{
return new ChallengeResult(
OpenIdConnectDefaults.AuthenticationScheme,
new AuthenticationProperties
{
IsPersistent = true,
RedirectUri = Url.Action("SignInCallback", "Account")
});
}

This causes the middleware to return a 302 (Found) response that redirects to the authentication endpoint.

User login sessions

As mentioned, when the user first signs in, the Cookie Authentication middleware writes the user claims to a
cookie. After that, HTTP requests are authenticated by reading the cookie.
By default, the cookie middleware writes a session cookie, which gets deleted once the user closes the browser. The
next time the user next visits the site, they will have to sign in again. However, if you set IsPersistent to true in the
ChallengeResult , the middleware writes a persistent cookie, so the user stays logged in after closing the browser.
You can configure the cookie expiration; see Controlling cookie options. Persistent cookies are more convenient for
the user, but may be inappropriate for some applications (say, a banking application) where you want the user to
sign in every time.

About the OpenID Connect middleware

The OpenID Connect middleware in ASP.NET hides most of the protocol details. This section contains some notes
about the implementation, that may be useful for understanding the protocol flow.
First, let's examine the authentication flow in terms of ASP.NET (ignoring the details of the OIDC protocol flow
between the app and Azure AD). The following diagram shows the process.

In this diagram, there are two MVC controllers. The Account controller handles sign-in requests, and the Home
controller serves up the home page.
 Here is the authentication process:
1. The user clicks the "Sign in" button, and the browser sends a GET request. For example: GET /Account/SignIn/ .
2. The account controller returns a ChallengeResult .
3. The OIDC middleware returns an HTTP 302 response, redirecting to Azure AD.
4. The browser sends the authentication request to Azure AD
5. The user signs in to Azure AD, and Azure AD sends back an authentication response.
6. The OIDC middleware creates a claims principal and passes it to the Cookie Authentication middleware.
7. The cookie middleware serializes the claims principal and sets a cookie.
8. The OIDC middleware redirects to the application's callback URL.
9. The browser follows the redirect, sending the cookie in the request.
10. The cookie middleware deserializes the cookie to a claims principal and sets HttpContext.User equal to the
claims principal. The request is routed to an MVC controller.
Authentication ticket
If authentication succeeds, the OIDC middleware creates an authentication ticket, which contains a claims principal
that holds the user's claims.

NOTE
Until the entire authentication flow is completed, HttpContext.User still holds an anonymous principal, not the
authenticated user. The anonymous principal has an empty claims collection. After authentication completes and the app
redirects, the cookie middleware deserializes the authentication cookie and sets HttpContext.User to a claims principal that
represents the authenticated user.

Authentication events
During the authentication process, the OpenID Connect middleware raises a series of events:
RedirectToIdentityProvider . Called right before the middleware redirects to the authentication endpoint. You
can use this event to modify the redirect URL; for example, to add request parameters. See Adding the admin
consent prompt for an example.
AuthorizationCodeReceived . Called with the authorization code.
TokenResponseReceived . Called after the middleware gets an access token from the IDP, but before it is
validated. Applies only to authorization code flow.
TokenValidated . Called after the middleware validates the ID token. At this point, the application has a set of
validated claims about the user. You can use this event to perform additional validation on the claims, or to
transform claims. See Working with claims.
UserInformationReceived . Called if the middleware gets the user profile from the user info endpoint. Applies
only to authorization code flow, and only when GetClaimsFromUserInfoEndpoint = true in the middleware
options.
TicketReceived . Called when authentication is completed. This is the last event, assuming that authentication
succeeds. After this event is handled, the user is signed into the app.
AuthenticationFailed . Called if authentication fails. Use this event to handle authentication failures — for
example, by redirecting to an error page.
To provide callbacks for these events, set the Events option on the middleware. There are two different ways to
declare the event handlers: Inline with lambdas, or in a class that derives from OpenIdConnectEvents . The
second approach is recommended if your event callbacks have any substantial logic, so they don't clutter your
startup class. Our reference implementation uses this approach.
OpenID Connect endpoints
Azure AD supports OpenID Connect Discovery, wherein the identity provider (IDP) returns a JSON metadata
document from a well-known endpoint. The metadata document contains information such as:
The URL of the authorization endpoint. This is where the app redirects to authenticate the user.
The URL of the "end session" endpoint, where the app goes to log out the user.
The URL to get the signing keys, which the client uses to validate the OIDC tokens that it gets from the IDP.
By default, the OIDC middleware knows how to fetch this metadata. Set the Authority option in the middleware,
and the middleware constructs the URL for the metadata. (You can override the metadata URL by setting the
MetadataAddress option.)
OpenID Connect flows
By default, the OIDC middleware uses hybrid flow with form post response mode.
Hybrid flow means the client can get an ID token and an authorization code in the same round-trip to the
authorization server.
Form post response mode means the authorization server uses an HTTP POST request to send the ID token and
authorization code to the app. The values are form-urlencoded (content type = "application/x-www-form-
urlencoded").
When the OIDC middleware redirects to the authorization endpoint, the redirect URL includes all of the query
string parameters needed by OIDC. For hybrid flow:
client_id. This value is set in the ClientId option.
scope = "openid profile", which means it's an OIDC request and we want the user's profile.
response_type = "code id_token". This specifies hybrid flow.
response_mode = "form_post". This specifies form post response.
To specify a different flow, set the ResponseType property on the options. For example:

app.AddAuthentication().AddOpenIdConnect(options =>
{
options.ResponseType = "code"; // Authorization code flow

// Other options
}

Next
 Work with claims-based identities
12/18/2020 • 5 minutes to read • Edit Online

Sample code

Claims in Azure AD
When a user signs in, Azure AD sends an ID token that contains a set of claims about the user. A claim is simply a
piece of information, expressed as a key/value pair. For example, email = bob@contoso.com . Claims have an issuer
— in this case, Azure AD — which is the entity that authenticates the user and creates the claims. You trust the
claims because you trust the issuer. (Conversely, if you don't trust the issuer, don't trust the claims!)
At a high level:
1. The user authenticates.
2. The Identity Provider (IDP) sends a set of claims.
3. The app normalizes or augments the claims (optional).
4. The app uses the claims to make authorization decisions.
In OpenID Connect, the set of claims that you get is controlled by the scope parameter of the authentication
request. However, Azure AD issues a limited set of claims through OpenID Connect; see Supported Token and
Claim Types. If you want more information about the user, you'll need to use the Azure AD Graph API.
Here are some of the claims from Azure AD that an app might typically care about:

C L A IM T Y P E IN ID TO K EN DESC RIP T IO N

aud Who the token was issued for. This will be the application's
client ID. Generally, you shouldn't need to worry about this
claim, because the middleware automatically validates it.
Example: "91464657-d17a-4327-91f3-2ed99386406f"

groups A list of Azure AD groups of which the user is a member.

Example:
["93e8f556-8661-4955-87b6-890bc043c30f", "fc781505-
18ef-4a31-a7d5-7d931d7b857e"]

iss The issuer of the OIDC token. Example:

https://sts.windows.net/b9bd2162-77ac-4fb2-8254-
5c36e9c0a9c4/

name The user's display name. Example: "Alice A."

oid The object identifier for the user in Azure AD. This value is the
immutable and non-reusable identifier of the user. Use this
value, not email, as a unique identifier for users; email
addresses can change. If you use the Azure AD Graph API in
your app, object ID is that value used to query profile
information. Example:
"59f9d2dc-995a-4ddf-915e-b3bb314a7fa4"

roles A list of app roles for the user. Example: ["SurveyCreator"]

 C L A IM T Y P E IN ID TO K EN DESC RIP T IO N

tid Tenant ID. This value is a unique identifier for the tenant in
Azure AD. Example:
"b9bd2162-77ac-4fb2-8254-5c36e9c0a9c4"

unique_name A human readable display name of the user. Example:

"alice@contoso.com"

upn User principal name. Example: "alice@contoso.com"

This table lists the claim types as they appear in the ID token. In ASP.NET Core, the OpenID Connect middleware
converts some of the claim types when it populates the Claims collection for the user principal:
oid > http://schemas.microsoft.com/identity/claims/objectidentifier
tid > http://schemas.microsoft.com/identity/claims/tenantid
unique_name > http://schemas.xmlsoap.org/ws/2005/05/identity/claims/name
upn > http://schemas.xmlsoap.org/ws/2005/05/identity/claims/upn

Claims transformations
During the authentication flow, you might want to modify the claims that you get from the IDP. In ASP.NET Core,
you can perform claims transformation inside of the AuthenticationValidated event from the OpenID Connect
middleware. (See Authentication events.)
Any claims that you add during AuthenticationValidated are stored in the session authentication cookie. They
don't get pushed back to Azure AD.
Here are some examples of claims transformation:
Claims normalization , or making claims consistent across users. This is particularly relevant if you are
getting claims from multiple IDPs, which might use different claim types for similar information. For
example, Azure AD sends a "upn" claim that contains the user's email. Other IDPs might send an "email"
claim. The following code converts the "upn" claim into an "email" claim:

var email = principal.FindFirst(ClaimTypes.Upn)?.Value;

if (!string.IsNullOrWhiteSpace(email))
{
identity.AddClaim(new Claim(ClaimTypes.Email, email));
}

Add default claim values for claims that aren't present — for example, assigning a user to a default role.
In some cases this can simplify authorization logic.
Add custom claim types with application-specific information about the user. For example, you might
store some information about the user in a database. You could add a custom claim with this information to
the authentication ticket. The claim is stored in a cookie, so you only need to get it from the database once
per login session. On the other hand, you also want to avoid creating excessively large cookies, so you need
to consider the trade-off between cookie size versus database lookups.
After the authentication flow is complete, the claims are available in HttpContext.User . At that point, you should
treat them as a read-only collection—for example, using them to make authorization decisions.

Issuer validation
In OpenID Connect, the issuer claim ("iss") identifies the IDP that issued the ID token. Part of the OIDC
authentication flow is to verify that the issuer claim matches the actual issuer. The OIDC middleware handles this
for you.
In Azure AD, the issuer value is unique per AD tenant ( https://sts.windows.net/<tenantID> ). Therefore, an
application should do an additional check, to make sure the issuer represents a tenant that is allowed to sign in to
the app.
For a single-tenant application, you can just check that the issuer is your own tenant. In fact, the OIDC middleware
does this automatically by default. In a multitenant app, you need to allow for multiple issuers, corresponding to
the different tenants. Here is a general approach to use:
In the OIDC middleware options, set ValidateIssuer to false. This turns off the automatic check.
When a tenant signs up, store the tenant and the issuer in your user DB.
Whenever a user signs in, look up the issuer in the database. If the issuer isn't found, it means that tenant hasn't
signed up. You can redirect them to a sign up page.
You could also block certain tenants; for example, for customers that didn't pay their subscription.
For a more detailed discussion, see Sign-up and tenant onboarding in a multitenant application.

Using claims for authorization

With claims, a user's identity is no longer a monolithic entity. For example, a user might have an email address,
phone number, birthday, gender, etc. Maybe the user's IDP stores all of this information. But when you authenticate
the user, you'll typically get a subset of these as claims. In this model, the user's identity is simply a bundle of
claims. When you make authorization decisions about a user, you will look for particular sets of claims. In other
words, the question "Can user X perform action Y" ultimately becomes "Does user X have claim Z".
Here are some basic patterns for checking claims.
To check that the user has a particular claim with a particular value:

if (User.HasClaim(ClaimTypes.Role, "Admin")) { ... }

This code checks whether the user has a Role claim with the value "Admin". It correctly handles the case
where the user has no Role claim or multiple Role claims.
The ClaimTypes class defines constants for commonly used claim types. However, you can use any string
value for the claim type.
To get a single value for a claim type, when you expect there to be at most one value:

string email = User.FindFirst(ClaimTypes.Email)?.Value;

To get all the values for a claim type:

IEnumerable<Claim> groups = User.FindAll("groups");

For more information, see Role-based and resource-based authorization in multitenant applications.
Next
 Tenant sign-up and onboarding
12/18/2020 • 6 minutes to read • Edit Online

Sample code
This article describes how to implement a sign-up process in a multitenant application, which allows a customer to
sign up their organization for your application.
There are several reasons to implement a sign-up process:
Allow an AD admin to consent for the customer's entire organization to use the application.
Collect credit card payment or other customer information.
Perform any one-time per-tenant setup needed by your application.

Admin consent and Azure AD permissions

In order to authenticate with Azure AD, an application needs access to the user's directory. At a minimum, the
application needs permission to read the user's profile. The first time that a user signs in, Azure AD shows a
consent page that lists the permissions being requested. By clicking Accept , the user grants permission to the
application.
By default, consent is granted on a per-user basis. Every user who signs in sees the consent page. However, Azure
AD also supports admin consent, which allows an AD administrator to consent for an entire organization.
When the admin consent flow is used, the consent page states that the AD admin is granting permission on behalf
of the entire tenant:

After the admin clicks Accept , other users within the same tenant can sign in, and Azure AD will skip the consent
screen.
Only an AD administrator can give admin consent, because it grants permission on behalf of the entire
organization. If a non-administrator tries to authenticate with the admin consent flow, Azure AD displays an error:
If the application requires additional permissions at a later point, the customer will need to sign up again and
consent to the updated permissions.

Implementing tenant sign-up

For the Tailspin Surveys application, we defined several requirements for the sign-up process:
A tenant must sign up before users can sign in.
Sign-up uses the admin consent flow.
Sign-up adds the user's tenant to the application database.
After a tenant signs up, the application shows an onboarding page.
In this section, we'll walk through our implementation of the sign-up process. It's important to understand that
"sign up" versus "sign in" is an application concept. During the authentication flow, Azure AD does not inherently
know whether the user is in process of signing up. It's up to the application to keep track of the context.
When an anonymous user visits the Surveys application, the user is shown two buttons, one to sign in, and one to
"enroll your company" (sign up).

These buttons invoke actions in the AccountController class.

The SignIn action returns a ChallengeResult , which causes the OpenID Connect middleware to redirect to the
authentication endpoint. This is the default way to trigger authentication in ASP.NET Core.
 [AllowAnonymous]
public IActionResult SignIn()
{
return new ChallengeResult(
OpenIdConnectDefaults.AuthenticationScheme,
new AuthenticationProperties
{
IsPersistent = true,
RedirectUri = Url.Action("SignInCallback", "Account")
});
}

Now compare the SignUp action:

[AllowAnonymous]
public IActionResult SignUp()
{
var state = new Dictionary<string, string> { { "signup", "true" }};
return new ChallengeResult(
OpenIdConnectDefaults.AuthenticationScheme,
new AuthenticationProperties(state)
{
RedirectUri = Url.Action(nameof(SignUpCallback), "Account")
});
}

Like SignIn , the SignUp action also returns a ChallengeResult . But this time, we add a piece of state information
to the AuthenticationProperties in the ChallengeResult :
signup: A Boolean flag, indicating that the user has started the sign-up process.
The state information in AuthenticationProperties gets added to the OpenID Connect state parameter, which
round trips during the authentication flow.

After the user authenticates in Azure AD and gets redirected back to the application, the authentication ticket
contains the state. We are using this fact to make sure the "signup" value persists across the entire authentication
flow.

Adding the admin consent prompt

In Azure AD, the admin consent flow is triggered by adding a "prompt" parameter to the query string in the
authentication request:
 /authorize?prompt=admin_consent&...

The Surveys application adds the prompt during the RedirectToAuthenticationEndpoint event. This event is called
right before the middleware redirects to the authentication endpoint.

public override Task RedirectToAuthenticationEndpoint(RedirectContext context)

{
if (context.IsSigningUp())
{
context.ProtocolMessage.Prompt = "admin_consent";
}

_logger.RedirectToIdentityProvider();
return Task.FromResult(0);
}

Setting ProtocolMessage.Prompt tells the middleware to add the "prompt" parameter to the authentication request.
Note that the prompt is only needed during sign-up. Regular sign-in should not include it. To distinguish between
them, we check for the signup value in the authentication state. The following extension method checks for this
condition:

internal static bool IsSigningUp(this BaseControlContext context)

{
Guard.ArgumentNotNull(context, nameof(context));

string signupValue;
// Check the HTTP context and convert to string
if ((context.Ticket == null) ||
(!context.Ticket.Properties.Items.TryGetValue("signup", out signupValue)))
{
return false;
}

// We have found the value, so see if it's valid

bool isSigningUp;
if (!bool.TryParse(signupValue, out isSigningUp))
{
// The value for signup is not a valid boolean, throw

throw new InvalidOperationException($"'{signupValue}' is an invalid boolean value");

}

return isSigningUp;
}

Registering a tenant
The Surveys application stores some information about each tenant and user in the application database.
In the Tenant table, IssuerValue is the value of the issuer claim for the tenant. For Azure AD, this is
https://sts.windows.net/<tentantID> and gives a unique value per tenant.

When a new tenant signs up, the Surveys application writes a tenant record to the database. This happens inside
the AuthenticationValidated event. (Don't do it before this event, because the ID token won't be validated yet, so
you can't trust the claim values. See Authentication.
Here is the relevant code from the Surveys application:

public override async Task TokenValidated(TokenValidatedContext context)

{
var principal = context.AuthenticationTicket.Principal;
var userId = principal.GetObjectIdentifierValue();
var tenantManager = context.HttpContext.RequestServices.GetService<TenantManager>();
var userManager = context.HttpContext.RequestServices.GetService<UserManager>();
var issuerValue = principal.GetIssuerValue();
_logger.AuthenticationValidated(userId, issuerValue);

// Normalize the claims first.

NormalizeClaims(principal);
var tenant = await tenantManager.FindByIssuerValueAsync(issuerValue)
.ConfigureAwait(false);

if (context.IsSigningUp())
{
if (tenant == null)
{
tenant = await SignUpTenantAsync(context, tenantManager)
.ConfigureAwait(false);
}

// In this case, we need to go ahead and set up the user signing us up.
await CreateOrUpdateUserAsync(context.Ticket, userManager, tenant)
.ConfigureAwait(false);
}
else
{
if (tenant == null)
{
_logger.UnregisteredUserSignInAttempted(userId, issuerValue);
throw new SecurityTokenValidationException($"Tenant {issuerValue} is not registered");
}

await CreateOrUpdateUserAsync(context.Ticket, userManager, tenant)

.ConfigureAwait(false);
}
}

This code does the following:

1. Check if the tenant's issuer value is already in the database. If the tenant has not signed up,
FindByIssuerValueAsync returns null.
2. If the user is signing up:
a. Add the tenant to the database ( SignUpTenantAsync ).
b. Add the authenticated user to the database ( CreateOrUpdateUserAsync ).
3. Otherwise complete the normal sign-in flow:
a. If the tenant's issuer was not found in the database, it means the tenant is not registered, and the
customer needs to sign up. In that case, throw an exception to cause the authentication to fail.
b. Otherwise, create a database record for this user, if there isn't one already ( CreateOrUpdateUserAsync ).

Here is the SignUpTenantAsync method that adds the tenant to the database.
 private async Task<Tenant> SignUpTenantAsync(BaseControlContext context, TenantManager tenantManager)
{
Guard.ArgumentNotNull(context, nameof(context));
Guard.ArgumentNotNull(tenantManager, nameof(tenantManager));

var principal = context.Ticket.Principal;

var issuerValue = principal.GetIssuerValue();
var tenant = new Tenant
{
IssuerValue = issuerValue,
Created = DateTimeOffset.UtcNow
};

try
{
await tenantManager.CreateAsync(tenant)
.ConfigureAwait(false);
}
catch(Exception ex)
{
_logger.SignUpTenantFailed(principal.GetObjectIdentifierValue(), issuerValue, ex);
throw;
}

return tenant;
}

Here is a summary of the entire sign-up flow in the Surveys application:

1. The user clicks the Sign Up button.
2. The AccountController.SignUp action returns a challenge result. The authentication state includes "signup"
value.
3. In the RedirectToAuthenticationEndpoint event, add the admin_consent prompt.
4. The OpenID Connect middleware redirects to Azure AD and the user authenticates.
5. In the AuthenticationValidated event, look for the "signup" state.
6. Add the tenant to the database.
Next
 Application roles
12/18/2020 • 5 minutes to read • Edit Online

Sample code
Application roles are used to assign permissions to users. For example, the Tailspin Surveys application defines the
following roles:
Administrator. Can perform all CRUD operations on any survey that belongs to that tenant.
Creator. Can create new surveys.
Reader. Can read any surveys that belong to that tenant.
You can see that roles ultimately get translated into permissions, during authorization. But the first question is how
to assign and manage roles. We identified three main options:
Azure AD App Roles
Azure AD security groups
Application role manager.

Roles using Azure AD App Roles

This is the approach that we used in the Tailspin Surveys app.
In this approach, The SaaS provider defines the application roles by adding them to the application manifest. After
a customer signs up, an admin for the customer's AD directory assigns users to the roles. When a user signs in, the
user's assigned roles are sent as claims.

NOTE
If the customer has Azure AD Premium, the admin can assign a security group to a role, and user members of the group will
inherit the app role. This is a convenient way to manage roles, because the group owner doesn't need to be an admin or app
owner.

Advantages of this approach:

Simple programming model.
Roles are specific to the application. The role claims for one application are not sent to another application.
If the customer removes the application from their Azure AD tenant, the roles go away.
The application doesn't need any extra Azure AD permissions, other than reading the user's profile.
Drawbacks:
Customers without Azure AD Premium cannot assign app roles to security groups. For these customers, all app
role assignments to users must be done by individually, by an administrator or an owner of the app.
If you have a backend web API which is separate from the web app, the app role assignments for the web app
don't apply to the web API. For more discussion of this point, see Securing a backend web API.
Implementation
Define the roles. The SaaS provider declares the app roles in the application manifest. For example, here is the
manifest entry for the Surveys app:
 "appRoles": [
{
"allowedMemberTypes": [
"User"
],
"description": "Creators can create Surveys",
"displayName": "SurveyCreator",
"id": "1b4f816e-5eaf-48b9-8613-7923830595ad",
"isEnabled": true,
"value": "SurveyCreator"
},
{
"allowedMemberTypes": [
"User"
],
"description": "Administrators can manage the Surveys in their tenant",
"displayName": "SurveyAdmin",
"id": "c20e145e-5459-4a6c-a074-b942bbd4cfe1",
"isEnabled": true,
"value": "SurveyAdmin"
}
],

The value property appears in the role claim. The id property is the unique identifier for the defined role. Always
generate a new GUID value for id .
Assign users . When a new customer signs up, the application is registered in the customer's Azure AD tenant. At
this point, an Azure AD admin for that tenant or an app owner (under Enterprise apps) can assign app roles to
users.

NOTE
As noted earlier, customers with Azure AD Premium can also assign app roles to security groups.

The following screenshot from the Azure portal shows users and groups for the Survey application. Admin and
Creator are groups, assigned the SurveyAdmin and SurveyCreator app roles, respectively. Alice is a user who was
assigned the SurveyAdmin app role directly. Bob and Charles are users that have not been directly assigned an app
role.

As shown in the following screenshot, Charles is part of the Admin group, so he inherits the SurveyAdmin role. In
the case of Bob, he has not been assigned an app role yet.
 NOTE
An alternative approach is for the application to assign app roles programmatically, using the Azure AD Graph API. However,
this requires the application to obtain write permissions for the customer's Azure AD directory, which is a high privilege that
is usually unnecessary.

Get role claims . When a user signs in, the application receives the user's assigned role(s) in a claim with type
http://schemas.microsoft.com/ws/2008/06/identity/claims/role (the roles claim in a JWT token).

A user can be assigned multiple roles, or no role. In your authorization code, don't assume the user has exactly one
role claim. Instead, write code that checks whether a particular claim value is present:

if (context.User.HasClaim(ClaimTypes.Role, "Admin")) { ... }

Roles using Azure AD security groups

In this approach, roles are represented as Azure AD security groups. The application assigns permissions to users
based on their security group memberships.
Advantages:
For customers who do not have Azure AD Premium, this approach enables the customer to use security groups
to manage role assignments.
Disadvantages:
Complexity. Because every tenant sends different group claims, the app must keep track of which security
groups correspond to which application roles, for each tenant.
As users belong to more groups, access tokens grow to include more claims. After a certain limit, Azure AD
includes an "overage" claim to limit the token size; see Microsoft identity platform access tokens. Application
roles avoid this issue because they are scoped to the specific application.
Implementation
In the application manifest, set the groupMembershipClaims property to "SecurityGroup". This is needed to get group
membership claims from Azure AD.

{
// ...
"groupMembershipClaims": "SecurityGroup",
}

When a new customer signs up, the application instructs the customer to create security groups for the roles
needed by the application. The customer then needs to enter the group object IDs into the application. The
application stores these in a table that maps group IDs to application roles, per tenant.

NOTE
Alternatively, the application could create the groups programmatically, using the Microsoft Graph API. This could be less
error prone, but requires the application to obtain privileged read/write permissions for the customer's directory. Many
customers might be unwilling to grant this level of access.

When a user signs in:

1. The application receives the user's groups as claims. The value of each claim is the object ID of a group.
2. Azure AD limits the number of groups sent in the token. If the number of groups exceeds this limit, Azure AD
sends a special "overage" claim. If that claim is present, the application must query the Azure AD Graph API to
get all of the groups to which that user belongs. For details, see [Authorization in Cloud Applications using AD
Groups], under the section titled "Groups claim overage".
3. The application looks up the object IDs in its own database, to find the corresponding application roles to assign
to the user.
4. The application adds a custom claim value to the user principal that expresses the application role. For example:
survey_role = "SurveyAdmin".

Authorization policies should use the custom role claim, not the group claim.

Roles using an application role manager

With this approach, application roles are not stored in Azure AD at all. Instead, the application stores the role
assignments for each user in its own DB — for example, using the RoleManager class in ASP.NET Identity.
Advantages:
The app has full control over the roles and user assignments.
Drawbacks:
More complex, harder to maintain.
Cannot use Azure AD security groups to manage role assignments.
Stores user information in the application database, where it can get out of sync with the tenant's Azure AD
directory, as users are added or removed.
Next
 Role-based and resource-based authorization
12/18/2020 • 5 minutes to read • Edit Online

Sample code
Our reference implementation is an ASP.NET Core application. In this article we'll look at two general approaches
to authorization, using the authorization APIs provided in ASP.NET Core.
Role-based authorization . Authorizing an action based on the roles assigned to a user. For example, some
actions require an administrator role.
Resource-based authorization . Authorizing an action based on a particular resource. For example, every
resource has an owner. The owner can delete the resource; other users cannot.
A typical app will employ a mix of both. For example, to delete a resource, the user must be the resource owner or
an admin.

Role-based authorization
The Tailspin Surveys application defines the following roles:
Administrator. Can perform all CRUD operations on any survey that belongs to that tenant.
Creator. Can create new surveys
Reader. Can read any surveys that belong to that tenant
Roles apply to users of the application. In the Surveys application, a user is either an administrator, creator, or
reader.
For a discussion of how to define and manage roles, see Application roles.
Regardless of how you manage the roles, your authorization code will look similar. ASP.NET Core has an
abstraction called authorization policies. With this feature, you define authorization policies in code, and then apply
those policies to controller actions. The policy is decoupled from the controller.
Create policies
To define a policy, first create a class that implements IAuthorizationRequirement . It's easiest to derive from
AuthorizationHandler . In the Handle method, examine the relevant claim(s).

Here is an example from the Tailspin Surveys application:

public class SurveyCreatorRequirement : AuthorizationHandler<SurveyCreatorRequirement>,

IAuthorizationRequirement
{
protected override Task HandleRequirementAsync(AuthorizationHandlerContext context,
SurveyCreatorRequirement requirement)
{
if (context.User.HasClaim(ClaimTypes.Role, Roles.SurveyAdmin) ||
context.User.HasClaim(ClaimTypes.Role, Roles.SurveyCreator))
{
context.Succeed(requirement);
}
return Task.FromResult(0);
}
}
This class defines the requirement for a user to create a new survey. The user must be in the SurveyAdmin or
SurveyCreator role.
In your startup class, define a named policy that includes one or more requirements. If there are multiple
requirements, the user must meet every requirement to be authorized. The following code defines two policies:

services.AddAuthorization(options =>
{
options.AddPolicy(PolicyNames.RequireSurveyCreator,
policy =>
{
policy.AddRequirements(new SurveyCreatorRequirement());
policy.RequireAuthenticatedUser(); // Adds DenyAnonymousAuthorizationRequirement
// By adding the CookieAuthenticationDefaults.AuthenticationScheme, if an authenticated
// user is not in the appropriate role, they will be redirected to a "forbidden" page.
policy.AddAuthenticationSchemes(CookieAuthenticationDefaults.AuthenticationScheme);
});

options.AddPolicy(PolicyNames.RequireSurveyAdmin,
policy =>
{
policy.AddRequirements(new SurveyAdminRequirement());
policy.RequireAuthenticatedUser();
policy.AddAuthenticationSchemes(CookieAuthenticationDefaults.AuthenticationScheme);
});
});

This code also sets the authentication scheme, which tells ASP.NET which authentication middleware should run if
authorization fails. In this case, we specify the cookie authentication middleware, because the cookie authentication
middleware can redirect the user to a "Forbidden" page. The location of the Forbidden page is set in the
AccessDeniedPath option for the cookie middleware; see Configuring the authentication middleware.

Authorize controller actions

Finally, to authorize an action in an MVC controller, set the policy in the Authorize attribute:

[Authorize(Policy = PolicyNames.RequireSurveyCreator)]
public IActionResult Create()
{
var survey = new SurveyDTO();
return View(survey);
}

In earlier versions of ASP.NET, you would set the Roles property on the attribute:

// old way
[Authorize(Roles = "SurveyCreator")]

This is still supported in ASP.NET Core, but it has some drawbacks compared with authorization policies:
It assumes a particular claim type. Policies can check for any claim type. Roles are just a type of claim.
The role name is hard-coded into the attribute. With policies, the authorization logic is all in one place, making
it easier to update or even load from configuration settings.
Policies enable more complex authorization decisions (for example, age >= 21) that can't be expressed by
simple role membership.

Resource-based authorization
Resource based authorization occurs whenever the authorization depends on a specific resource that will be
affected by an operation. In the Tailspin Surveys application, every survey has an owner and zero-to-many
contributors.
The owner can read, update, delete, publish, and unpublish the survey.
The owner can assign contributors to the survey.
Contributors can read and update the survey.
Note that "owner" and "contributor" are not application roles; they are stored per survey, in the application
database. To check whether a user can delete a survey, for example, the app checks whether the user is the owner
for that survey.
In ASP.NET Core, implement resource-based authorization by deriving from AuthorizationHandler and
overriding the Handle method.

public class SurveyAuthorizationHandler : AuthorizationHandler<OperationAuthorizationRequirement, Survey>

{
protected override void HandleRequirementAsync(AuthorizationHandlerContext context,
OperationAuthorizationRequirement operation, Survey resource)
{
}
}

Notice that this class is strongly typed for Survey objects. Register the class for DI on startup:

services.AddSingleton<IAuthorizationHandler>(factory =>
{
return new SurveyAuthorizationHandler();
});

To perform authorization checks, use the IAuthorizationSer vice interface, which you can inject into your
controllers. The following code checks whether a user can read a survey:

if (await _authorizationService.AuthorizeAsync(User, survey, Operations.Read) == false)

{
return StatusCode(403);
}

Because we pass in a Survey object, this call will invoke the SurveyAuthorizationHandler .
In your authorization code, a good approach is to aggregate all of the user's role-based and resource-based
permissions, then check the aggregate set against the desired operation. Here is an example from the Surveys app.
The application defines several permission types:
Admin
Contributor
Creator
Owner
Reader
The application also defines a set of possible operations on surveys:
Create
Read
Update
Delete
 Publish
Unpublish
The following code creates a list of permissions for a particular user and survey. Notice that this code looks at both
the user's app roles, and the owner/contributor fields in the survey.

public class SurveyAuthorizationHandler : AuthorizationHandler<OperationAuthorizationRequirement, Survey>

{
protected override Task HandleRequirementAsync(AuthorizationHandlerContext context,
OperationAuthorizationRequirement requirement, Survey resource)
{
var permissions = new List<UserPermissionType>();
int surveyTenantId = context.User.GetSurveyTenantIdValue();
int userId = context.User.GetSurveyUserIdValue();
string user = context.User.GetUserName();

if (resource.TenantId == surveyTenantId)
{
// Admin can do anything, as long as the resource belongs to the admin's tenant.
if (context.User.HasClaim(ClaimTypes.Role, Roles.SurveyAdmin))
{
context.Succeed(requirement);
return Task.FromResult(0);
}

if (context.User.HasClaim(ClaimTypes.Role, Roles.SurveyCreator))
{
permissions.Add(UserPermissionType.Creator);
}
else
{
permissions.Add(UserPermissionType.Reader);
}

if (resource.OwnerId == userId)
{
permissions.Add(UserPermissionType.Owner);
}
}
if (resource.Contributors != null && resource.Contributors.Any(x => x.UserId == userId))
{
permissions.Add(UserPermissionType.Contributor);
}

if (ValidateUserPermissions[requirement](permissions))
{
context.Succeed(requirement);
}
return Task.FromResult(0);
}
}

In a multitenant application, you must ensure that permissions don't "leak" to another tenant's data. In the Surveys
app, the Contributor permission is allowed across tenants—you can assign someone from another tenant as a
contributor. The other permission types are restricted to resources that belong to that user's tenant. To enforce this
requirement, the code checks the tenant ID before granting the permission. (The TenantId field as assigned when
the survey is created.)
The next step is to check the operation (such as read, update, or delete) against the permissions. The Surveys app
implements this step by using a lookup table of functions:
 static readonly Dictionary<OperationAuthorizationRequirement, Func<List<UserPermissionType>, bool>>
ValidateUserPermissions
= new Dictionary<OperationAuthorizationRequirement, Func<List<UserPermissionType>, bool>>

{
{ Operations.Create, x => x.Contains(UserPermissionType.Creator) },

{ Operations.Read, x => x.Contains(UserPermissionType.Creator) ||

x.Contains(UserPermissionType.Reader) ||
x.Contains(UserPermissionType.Contributor) ||
x.Contains(UserPermissionType.Owner) },

{ Operations.Update, x => x.Contains(UserPermissionType.Contributor) ||

x.Contains(UserPermissionType.Owner) },

{ Operations.Delete, x => x.Contains(UserPermissionType.Owner) },

{ Operations.Publish, x => x.Contains(UserPermissionType.Owner) },

{ Operations.UnPublish, x => x.Contains(UserPermissionType.Owner) }

};

Next
 Secure a backend web API for multitenant
applications
12/18/2020 • 8 minutes to read • Edit Online

Sample code
The Tailspin Surveys application uses a backend web API to manage CRUD operations on surveys. For example,
when a user clicks "My Surveys", the web application sends an HTTP request to the web API:

GET /users/{userId}/surveys

The web API returns a JSON object:

{
"Published":[],
"Own":[
{"Id":1,"Title":"Survey 1"},
{"Id":3,"Title":"Survey 3"},
],
"Contribute": [{"Id":8,"Title":"My survey"}]
}

The web API does not allow anonymous requests, so the web app must authenticate itself using OAuth 2 bearer
tokens.

NOTE
This is a server-to-server scenario. The application does not make any AJAX calls to the API from the browser client.

There are two main approaches you can take:

Delegated user identity. The web application authenticates with the user's identity.
Application identity. The web application authenticates with its client ID, using OAuth 2 client credential flow.
The Tailspin application implements delegated user identity. Here are the main differences:
Delegated user identity:
The bearer token sent to the web API contains the user identity.
The web API makes authorization decisions based on the user identity.
The web application needs to handle 403 (Forbidden) errors from the web API, if the user is not authorized to
perform an action.
Typically, the web application still makes some authorization decisions that affect UI, such as showing or hiding
UI elements).
The web API can potentially be used by untrusted clients, such as a JavaScript application or a native client
application.
Application identity:
The web API does not get information about the user.
 The web API cannot perform any authorization based on the user identity. All authorization decisions are made
by the web application.
The web API cannot be used by an untrusted client (JavaScript or native client application).
This approach may be somewhat simpler to implement, because there is no authorization logic in the Web API.
In either approach, the web application must get an access token, which is the credential needed to call the web
API.
For delegated user identity, the token has to come from an identity provider (IDP), such as Azure Active
Directory, which can issue a token on behalf of the user.
For client credentials, an application might get the token from the IDP or host its own token server. (But don't
write a token server from scratch; use a well-tested framework like IdentityServer4.) If you authenticate with
Azure AD, it's strongly recommended to get the access token from Azure AD, even with client credential flow.
The rest of this article assumes the application is authenticating with Azure AD.

A diagram that shows the web application requesting an access token from Azure AD and sending the token to the
web API.

Register the web API in Azure AD

In order for Azure AD to issue a bearer token for the web API, you need to configure some things in Azure AD.
1. Register the web API in Azure AD.
2. Add the client ID of the web app to the web API application manifest, in the knownClientApplications
property. See the GitHub readme for more information.
3. Give the web application permission to call the web API. In the Azure portal, you can set two types of
permissions: "Application Permissions" for application identity (client credential flow), or "Delegated
Permissions" for delegated user identity.

A screenshot of the Azure portal that shows the application permissions and delegated permissions.
Getting an access token
Before calling the web API, the web application gets an access token from Azure AD. In a .NET application, use the
Azure AD Authentication Library (ADAL) for .NET.
In the OAuth 2 authorization code flow, the application exchanges an authorization code for an access token. The
following code uses ADAL to get the access token. This code is called during the AuthorizationCodeReceived event.

// The OpenID Connect middleware sends this event when it gets the authorization code.
public override async Task AuthorizationCodeReceived(AuthorizationCodeReceivedContext context)
{
string authorizationCode = context.ProtocolMessage.Code;
string authority = "https://login.microsoftonline.com/" + tenantID
string resourceID = "https://tailspin.onmicrosoft.com/surveys.webapi" // App ID URI
ClientCredential credential = new ClientCredential(clientId, clientSecret);

AuthenticationContext authContext = new AuthenticationContext(authority, tokenCache);

AuthenticationResult authResult = await authContext.AcquireTokenByAuthorizationCodeAsync(
authorizationCode, new Uri(redirectUri), credential, resourceID);

// If successful, the token is in authResult.AccessToken

}

Here are the various parameters that are needed:

authority . Derived from the tenant ID of the signed in user. (Not the tenant ID of the SaaS provider)
authorizationCode . the auth code that you got back from the IDP.
clientId . The web application's client ID.
clientSecret . The web application's client secret.
redirectUri . The redirect URI that you set for OpenID Connect. This is where the IDP calls back with the token.
resourceID . The App ID URI of the web API, which you created when you registered the web API in Azure AD
tokenCache . An object that caches the access tokens. See Token caching.

If AcquireTokenByAuthorizationCodeAsync succeeds, ADAL caches the token. Later, you can get the token from the
cache by calling AcquireTokenSilentAsync:

AuthenticationContext authContext = new AuthenticationContext(authority, tokenCache);

var result = await authContext.AcquireTokenSilentAsync(resourceID, credential, new UserIdentifier(userId,
UserIdentifierType.UniqueId));

where userId is the user's object ID, which is found in the

http://schemas.microsoft.com/identity/claims/objectidentifier claim.

Using the access token to call the web API

Once you have the token, send it in the Authorization header of the HTTP requests to the web API.

Authorization: Bearer xxxxxxxxxx

The following extension method from the Surveys application sets the Authorization header on an HTTP request,
using the HttpClient class.
 public static async Task<HttpResponseMessage> SendRequestWithBearerTokenAsync(this HttpClient httpClient,
HttpMethod method, string path, object requestBody, string accessToken, CancellationToken ct)
{
var request = new HttpRequestMessage(method, path);
if (requestBody != null)
{
var json = JsonConvert.SerializeObject(requestBody, Formatting.None);
var content = new StringContent(json, Encoding.UTF8, "application/json");
request.Content = content;
}

request.Headers.Authorization = new AuthenticationHeaderValue("Bearer", accessToken);

request.Headers.Accept.Add(new MediaTypeWithQualityHeaderValue("application/json"));

var response = await httpClient.SendAsync(request, ct);

return response;
}

Authenticating in the web API

The web API has to authenticate the bearer token. In ASP.NET Core, you can use the
Microsoft.AspNet.Authentication.JwtBearer package. This package provides middleware that enables the
application to receive OpenID Connect bearer tokens.
Register the middleware in your web API Startup class.

public void Configure(IApplicationBuilder app, IHostingEnvironment env, ApplicationDbContext dbContext,

ILoggerFactory loggerFactory)
{
// ...

app.UseJwtBearerAuthentication(new JwtBearerOptions {
Audience = configOptions.AzureAd.WebApiResourceId,
Authority = Constants.AuthEndpointPrefix,
TokenValidationParameters = new TokenValidationParameters {
ValidateIssuer = false
},
Events= new SurveysJwtBearerEvents(loggerFactory.CreateLogger<SurveysJwtBearerEvents>())
});

// ...
}

Audience . Set this to the App ID URL for the web API, which you created when you registered the web API with
Azure AD.
Authority . For a multitenant application, set this to https://login.microsoftonline.com/common/ .
TokenValidationParameters . For a multitenant application, set ValidateIssuer to false. That means the
application will validate the issuer.
Events is a class that derives from JwtBearerEvents .
Issuer validation
Validate the token issuer in the JwtBearerEvents.TokenValidated event. The issuer is sent in the "iss" claim.
In the Surveys application, the web API doesn't handle tenant sign-up. Therefore, it just checks if the issuer is
already in the application database. If not, it throws an exception, which causes authentication to fail.
 public override async Task TokenValidated(TokenValidatedContext context)
{
var principal = context.Ticket.Principal;
var tenantManager = context.HttpContext.RequestServices.GetService<TenantManager>();
var userManager = context.HttpContext.RequestServices.GetService<UserManager>();
var issuerValue = principal.GetIssuerValue();
var tenant = await tenantManager.FindByIssuerValueAsync(issuerValue);

if (tenant == null)
{
// The caller was not from a trusted issuer. Throw to block the authentication flow.
throw new SecurityTokenValidationException();
}

var identity = principal.Identities.First();

// Add new claim for survey_userid

var registeredUser = await userManager.FindByObjectIdentifier(principal.GetObjectIdentifierValue());
identity.AddClaim(new Claim(SurveyClaimTypes.SurveyUserIdClaimType, registeredUser.Id.ToString()));
identity.AddClaim(new Claim(SurveyClaimTypes.SurveyTenantIdClaimType,
registeredUser.TenantId.ToString()));

// Add new claim for Email

var email = principal.FindFirst(ClaimTypes.Upn)?.Value;
if (!string.IsNullOrWhiteSpace(email))
{
identity.AddClaim(new Claim(ClaimTypes.Email, email));
}
}

As this example shows, you can also use the TokenValidated event to modify the claims. Remember that the
claims come directly from Azure AD. If the web application modifies the claims that it gets, those changes won't
show up in the bearer token that the web API receives. For more information, see Claims transformations.

Authorization
For a general discussion of authorization, see Role-based and resource-based authorization.
The JwtBearer middleware handles the authorization responses. For example, to restrict a controller action to
authenticated users, use the [Authorize] attribute and specify JwtBearerDefaults.AuthenticationScheme as
the authentication scheme:

[Authorize(ActiveAuthenticationSchemes = JwtBearerDefaults.AuthenticationScheme)]

This returns a 401 status code if the user is not authenticated.

To restrict a controller action by authorization policy, specify the policy name in the [Authorize] attribute:

[Authorize(Policy = PolicyNames.RequireSurveyCreator)]

This returns a 401 status code if the user is not authenticated, and 403 if the user is authenticated but not
authorized. Register the policy on startup:
 public void ConfigureServices(IServiceCollection services)
{
services.AddAuthorization(options =>
{
options.AddPolicy(PolicyNames.RequireSurveyCreator,
policy =>
{
policy.AddRequirements(new SurveyCreatorRequirement());
policy.RequireAuthenticatedUser(); // Adds DenyAnonymousAuthorizationRequirement
policy.AddAuthenticationSchemes(JwtBearerDefaults.AuthenticationScheme);
});
options.AddPolicy(PolicyNames.RequireSurveyAdmin,
policy =>
{
policy.AddRequirements(new SurveyAdminRequirement());
policy.RequireAuthenticatedUser(); // Adds DenyAnonymousAuthorizationRequirement
policy.AddAuthenticationSchemes(JwtBearerDefaults.AuthenticationScheme);
});
});

// ...
}

Protecting application secrets

It's common to have application settings that are sensitive and must be protected, such as:
Database connection strings
Passwords
Cryptographic keys
As a security best practice, you should never store these secrets in source control. It's too easy for them to leak —
even if your source code repository is private. And it's not just about keeping secrets from the general public. On
larger projects, you might want to restrict which developers and operators can access the production secrets.
(Settings for test or development environments are different.)
A more secure option is to store these secrets in Azure Key Vault. Key Vault is a cloud-hosted service for managing
cryptographic keys and other secrets. This article shows how to use Key Vault to store configuration settings for
your app.
In the Tailspin Surveys application, the following settings are secret:
The database connection string.
The Redis connection string.
The client secret for the web application.
The Surveys application loads configuration settings from the following places:
The appsettings.json file
The user secrets store (development environment only; for testing)
The hosting environment (app settings in Azure web apps)
Key Vault (when enabled)
Each of these overrides the previous one, so any settings stored in Key Vault take precedence.
 NOTE
By default, the Key Vault configuration provider is disabled. It's not needed for running the application locally. You would
enable it in a production deployment.

At startup, the application reads settings from every registered configuration provider, and uses them to populate
a strongly typed options object. For more information, see Using Options and configuration objects.
Next
 Cache access tokens
12/18/2020 • 4 minutes to read • Edit Online

Sample code
It's relatively expensive to get an OAuth access token, because it requires an HTTP request to the token endpoint.
Therefore, it's good to cache tokens whenever possible. The Azure AD Authentication Library (ADAL) automatically
caches tokens obtained from Azure AD, including refresh tokens.
ADAL provides a default token cache implementation. However, this token cache is intended for native client apps,
and is not suitable for web apps:
It is a static instance, and not thread safe.
It doesn't scale to large numbers of users, because tokens from all users go into the same dictionary.
It can't be shared across web servers in a farm.
Instead, you should implement a custom token cache that derives from the ADAL TokenCache class but is suitable
for a server environment and provides the desirable level of isolation between tokens for different users.
The TokenCache class stores a dictionary of tokens, indexed by issuer, resource, client ID, and user. A custom token
cache should write this dictionary to a backing store, such as a Redis cache.
In the Tailspin Surveys application, the DistributedTokenCache class implements the token cache. This
implementation uses the IDistributedCache abstraction from ASP.NET Core. That way, any IDistributedCache
implementation can be used as a backing store.
By default, the Surveys app uses a Redis cache.
For a single-instance web server, you could use the ASP.NET Core in-memory cache. (This is also a good option
for running the app locally during development.)
DistributedTokenCache stores the cache data as key/value pairs in the backing store. The key is the user ID plus
client ID, so the backing store holds separate cache data for each unique combination of user/client.
The backing store is partitioned by user. For each HTTP request, the tokens for that user are read from the backing
store and loaded into the TokenCache dictionary. If Redis is used as the backing store, every server instance in a
server farm reads/writes to the same cache, and this approach scales to many users.

Encrypting cached tokens

Tokens are sensitive data, because they grant access to a user's resources. (Moreover, unlike a user's password, you
can't just store a hash of the token.) Therefore, it's critical to protect tokens from being compromised. The Redis-
backed cache is protected by a password, but if someone obtains the password, they could get all of the cached
access tokens. For that reason, the DistributedTokenCache encrypts everything that it writes to the backing store.
Encryption is done using the ASP.NET Core data protection APIs.

NOTE
If you deploy to Azure Web Sites, the encryption keys are backed up to network storage and synchronized across all
machines (see Key management and lifetime). By default, keys are not encrypted when running in Azure Web Sites, but you
can enable encryption using an X.509 certificate.

DistributedTokenCache implementation
The DistributedTokenCache class derives from the ADAL TokenCache class.
In the constructor, the DistributedTokenCache class creates a key for the current user and loads the cache from the
backing store:

public DistributedTokenCache(
ClaimsPrincipal claimsPrincipal,
IDistributedCache distributedCache,
ILoggerFactory loggerFactory,
IDataProtectionProvider dataProtectionProvider)
: base()
{
_claimsPrincipal = claimsPrincipal;
_cacheKey = BuildCacheKey(_claimsPrincipal);
_distributedCache = distributedCache;
_logger = loggerFactory.CreateLogger<DistributedTokenCache>();
_protector = dataProtectionProvider.CreateProtector(typeof(DistributedTokenCache).FullName);
AfterAccess = AfterAccessNotification;
LoadFromCache();
}

The key is created by concatenating the user ID and client ID. Both of these are taken from claims found in the
user's ClaimsPrincipal :

private static string BuildCacheKey(ClaimsPrincipal claimsPrincipal)

{
string clientId = claimsPrincipal.FindFirstValue("aud", true);
return string.Format(
"UserId:{0}::ClientId:{1}",
claimsPrincipal.GetObjectIdentifierValue(),
clientId);
}

To load the cache data, read the serialized blob from the backing store, and call TokenCache.Deserialize to convert
the blob into cache data.
 private void LoadFromCache()
{
byte[] cacheData = _distributedCache.Get(_cacheKey);
if (cacheData != null)
{
this.Deserialize(_protector.Unprotect(cacheData));
}
}

Whenever ADAL accesses the cache, it fires an AfterAccess event. If the cache data has changed, the
HasStateChanged property is true. In that case, update the backing store to reflect the change, and then set
HasStateChanged to false.

public void AfterAccessNotification(TokenCacheNotificationArgs args)

{
if (this.HasStateChanged)
{
try
{
if (this.Count > 0)
{
_distributedCache.Set(_cacheKey, _protector.Protect(this.Serialize()));
}
else
{
// There are no tokens for this user/client, so remove the item from the cache.
_distributedCache.Remove(_cacheKey);
}
this.HasStateChanged = false;
}
catch (Exception exp)
{
_logger.WriteToCacheFailed(exp);
throw;
}
}
}

TokenCache sends two other events:

BeforeWrite . Called immediately before ADAL writes to the cache. You can use this to implement a concurrency
strategy
BeforeAccess . Called immediately before ADAL reads from the cache. Here you can reload the cache to get the
latest version.
In our case, we decided not to handle these two events.
For concurrency, last write wins. That's OK, because tokens are stored independently for each user + client, so a
conflict would only happen if the same user had two concurrent login sessions.
For reading, we load the cache on every request. Requests are short lived. If the cache gets modified in that
time, the next request will pick up the new value.
Next
 Use client assertion to get access tokens from Azure
AD
12/18/2020 • 2 minutes to read • Edit Online

Sample code
This article describes how to add client assertion to the Tailspin Surveys sample application.

Understanding client assertion in OpenID Connect

When using authorization code flow or hybrid flow in OpenID Connect, the client exchanges an authorization code
for an access token. During this step, the client has to authenticate itself to the server.

One way to authenticate the client is by using a client secret. That's how the Tailspin Surveys application is
configured by default.
Here is an example request from the client to the IDP, requesting an access token. Note the client_secret
parameter.

POST https://login.microsoftonline.com/b9bd2162xxx/oauth2/token HTTP/1.1

Content-Type: application/x-www-form-urlencoded

resource=https://tailspin.onmicrosoft.com/surveys.webapi
&client_id=87df91dc-63de-4765-8701-b59cc8bd9e11
&client_secret=i3Bf12Dn...
&grant_type=authorization_code
&code=PG8wJG6Y...

The secret is just a string, so you have to make sure not to leak the value. The best practice is to keep the client
secret out of source control. When you deploy to Azure, store the secret in an app setting.
However, anyone with access to the Azure subscription can view the app settings. Furthermore, there is always a
temptation to check secrets into source control (for example, in deployment scripts), share them by email, and so
on.
For additional security, you can use client assertion instead of a client secret. With client assertion, the client uses
an X.509 certificate to prove the token request came from the client. The client certificate is installed on the web
server. Generally, it will be easier to restrict access to the certificate, than to ensure that nobody inadvertently
reveals a client secret. For more information about configuring certificates in a web app, see Using Certificates in
Azure Websites Applications
Here is a token request using client assertion:

POST https://login.microsoftonline.com/b9bd2162xxx/oauth2/token HTTP/1.1

Content-Type: application/x-www-form-urlencoded

resource=https://tailspin.onmicrosoft.com/surveys.webapi
&client_id=87df91dc-63de-4765-8701-b59cc8bd9e11
&client_assertion_type=urn:ietf:params:oauth:client-assertion-type:jwt-bearer
&client_assertion=eyJhbGci...
&grant_type=authorization_code
&code= PG8wJG6Y...

Notice that the client_secret parameter is no longer used. Instead, the client_assertion parameter contains a
JWT token that was signed using the client certificate. The client_assertion_type parameter specifies the type of
assertion — in this case, JWT token. The server validates the JWT token. If the JWT token is invalid, the token
request returns an error.

NOTE
X.509 certificates are not the only form of client assertion; we focus on it here because it is supported by Azure AD.

At run time, the web application reads the certificate from the certificate store. The certificate must be installed on
the same machine as the web app.

Implementing client assertion

The Surveys application includes a helper class that creates a ClientAssertionCertificate that you can pass to the
AuthenticationContext.AcquireTokenSilentAsync method to acquire a token from Azure AD.

public class CertificateCredentialService : ICredentialService

{
private Lazy<Task<AdalCredential>> _credential;

public CertificateCredentialService(IOptions<ConfigurationOptions> options)

{
var aadOptions = options.Value?.AzureAd;
_credential = new Lazy<Task<AdalCredential>>(() =>
{
X509Certificate2 cert = CertificateUtility.FindCertificateByThumbprint(
aadOptions.Asymmetric.StoreName,
aadOptions.Asymmetric.StoreLocation,
aadOptions.Asymmetric.CertificateThumbprint,
aadOptions.Asymmetric.ValidationRequired);
string password = null;
var certBytes = CertificateUtility.ExportCertificateWithPrivateKey(cert, out password);
return Task.FromResult(new AdalCredential(new ClientAssertionCertificate(aadOptions.ClientId, new
X509Certificate2(certBytes, password))));
});
}

public async Task<AdalCredential> GetCredentialsAsync()

{
return await _credential.Value;
}
}
 NOTE
For more information on ID Tokens, please see this document.

Next
 Federate with a customer's AD FS
12/18/2020 • 6 minutes to read • Edit Online

This article describes how a multitenant SaaS application can support authentication via Active Directory
Federation Services (AD FS), in order to federate with a customer's AD FS.

Overview
Azure Active Directory (Azure AD) makes it easy to sign in users from Azure AD tenants, including Office365 and
Dynamics CRM Online customers. But what about customers who use on-premises Active Directory on a corporate
intranet?
One option is for these customers to sync their on-premises AD with Azure AD, using Azure AD Connect. However,
some customers may be unable to use this approach, due to corporate IT policy or other reasons. In that case,
another option is to federate through Active Directory Federation Services (AD FS).
To enable this scenario:
The customer must have an Internet-facing AD FS farm.
The SaaS provider deploys their own AD FS farm.
The customer and the SaaS provider must set up federation trust. This is a manual process.
There are three main roles in the trust relation:
The customer's AD FS is the account partner, responsible for authenticating users from the customer's AD,
and creating security tokens with user claims.
The SaaS provider's AD FS is the resource partner, which trusts the account partner and receives the user
claims.
The application is configured as a relying party (RP) in the SaaS provider's AD FS.

NOTE
In this article, we assume the application uses OpenID Connect as the authentication protocol. Another option is to use WS-
Federation.
For OpenID Connect, the SaaS provider must use AD FS 2016, running in Windows Server 2016. AD FS 3.0 does not
support OpenID Connect.
For an example of using WS-Federation with ASP.NET 4, see the active-directory-dotnet-webapp-wsfederation
sample.

Authentication flow
1. When the user clicks "sign in", the application redirects to an OpenID Connect endpoint on the SaaS provider's
AD FS.
2. The user enters his or her organizational user name (" alice@corp.contoso.com "). AD FS uses home realm
discovery to redirect to the customer's AD FS, where the user enters their credentials.
3. The customer's AD FS sends user claims to the SaaS provider's AD FS, using WF-Federation (or SAML).
4. Claims flow from AD FS to the app, using OpenID Connect. This requires a protocol transition from WS-
Federation.

Limitations
By default, the relying party application receives only a fixed set of claims available in the id_token, shown in the
following table. With AD FS 2016, you can customize the id_token in OpenID Connect scenarios. For more
information, see Custom ID Tokens in AD FS.

C L A IM DESC RIP T IO N

aud Audience. The application for which the claims were issued.

authenticationinstant Authentication instant. The time at which authentication

occurred.

c_hash Code hash value. This is a hash of the token contents.

exp Expiration time. The time after which the token will no longer
be accepted.

iat Issued at. The time when the token was issued.

iss Issuer. The value of this claim is always the resource partner's
AD FS.

name User name. Example: john@corp.fabrikam.com

nameidentifier Name identifier. The identifier for the name of the entity for
which the token was issued.

nonce Session nonce. A unique value generated by AD FS to help

prevent replay attacks.

upn User principal name (UPN). Example:

john@corp.fabrikam.com

pwd_exp Password expiration period. The number of seconds until the

user's password or a similar authentication secret, such as a
PIN. expires.
 NOTE
The "iss" claim contains the AD FS of the partner (typically, this claim will identify the SaaS provider as the issuer). It does not
identify the customer's AD FS. You can find the customer's domain as part of the UPN.

The rest of this article describes how to set up the trust relationship between the RP (the app) and the account
partner (the customer).

AD FS deployment
The SaaS provider can deploy AD FS either on-premises or on Azure VMs. For security and availability, the
following guidelines are important:
Deploy at least two AD FS servers and two AD FS proxy servers to achieve the best availability of the AD FS
service.
Domain controllers and AD FS servers should never be exposed directly to the Internet and should be in a
virtual network with direct access to them.
Web application proxies (previously AD FS proxies) must be used to publish AD FS servers to the Internet.
To set up a similar topology in Azure requires the use of virtual networks, network security groups, virtual
machines, and availability sets. For more details, see Guidelines for deploying Windows Server Active Directory on
Azure Virtual Machines.

Configure OpenID Connect authentication with AD FS

The SaaS provider must enable OpenID Connect between the application and AD FS. To do so, add an application
group in AD FS. You can find detailed instructions in this blog post, under "Setting up a Web App for OpenId
Connect sign in AD FS."
Next, configure the OpenID Connect middleware. The metadata endpoint is
https://domain/adfs/.well-known/openid-configuration , where domain is the SaaS provider's AD FS domain.

Typically you might combine this with other OpenID Connect endpoints (such as Azure AD). You'll need two
different sign-in buttons or some other way to distinguish them, so that the user is sent to the correct
authentication endpoint.

Configure the AD FS Resource Partner

The SaaS provider must do the following for each customer that wants to connect via AD FS:
1. Add a claims provider trust.
2. Add claims rules.
3. Enable home-realm discovery.
Here are the steps in more detail.
Add the claims provider trust
1. In Server Manager, click Tools , and then select AD FS Management .
2. In the console tree, under AD FS , right click Claims Provider Trusts . Select Add Claims Provider Trust .
3. Click Star t to start the wizard.
4. Select the option "Import data about the claims provider published online or on a local network". Enter the URI
of the customer's federation metadata endpoint. (Example:
https://contoso.com/FederationMetadata/2007-06/FederationMetadata.xml .) You will need to get this from the
customer.
5. Complete the wizard using the default options.
Edit claims rules
1. Right-click the newly added claims provider trust, and select Edit Claims Rules .
2. Click Add Rule .
3. Select "Pass Through or Filter an Incoming Claim" and click Next .

4. Enter a name for the rule.

5. Under "Incoming claim type", select UPN .
6. Select "Pass through all claim values".
7. Click Finish .
8. Repeat steps 2 - 7, and specify Anchor Claim Type for the incoming claim type.
9. Click OK to complete the wizard.
Enable home -realm discovery
Run the following PowerShell script:

Set-AdfsClaimsProviderTrust -TargetName "name" -OrganizationalAccountSuffix @("suffix")

where "name" is the friendly name of the claims provider trust, and "suffix" is the UPN suffix for the customer's AD
(example, "corp.fabrikam.com").
With this configuration, end users can type in their organizational account, and AD FS automatically selects the
corresponding claims provider. See Customizing the AD FS Sign-in Pages, under the section "Configure Identity
Provider to use certain email suffixes".

Configure the AD FS Account Partner

The customer must do the following:
1. Add a relying party (RP) trust.
2. Adds claims rules.
Add the RP trust
1. In Server Manager, click Tools , and then select AD FS Management .
2. In the console tree, under AD FS , right click Relying Par ty Trusts . Select Add Relying Par ty Trust .
3. Select Claims Aware and click Star t .
4. On the Select Data Source page, select the option "Import data about the claims provider published online or
on a local network". Enter the URI of the SaaS provider's federation metadata endpoint.

5. On the Specify Display Name page, enter any name.

6. On the Choose Access Control Policy page, choose a policy. You could permit everyone in the organization,
or choose a specific security group.
7. Enter any parameters required in the Policy box.
8. Click Next to complete the wizard.
Add claims rules
1. Right-click the newly added relying party trust, and select Edit Claim Issuance Policy .
2. Click Add Rule .
3. Select "Send LDAP Attributes as Claims" and click Next .
4. Enter a name for the rule, such as "UPN".
5. Under Attribute store , select Active Director y .
6. In the Mapping of LDAP attributes section:
Under LDAP Attribute , select User-Principal-Name .
Under Outgoing Claim Type , select UPN .
 7. Click Finish .
8. Click Add Rule again.
9. Select "Send Claims Using a Custom Rule" and click Next .
10. Enter a name for the rule, such as "Anchor Claim Type".
11. Under Custom rule , enter the following:

EXISTS([Type == "http://schemas.microsoft.com/ws/2014/01/identity/claims/anchorclaimtype"])=>
issue (Type = "http://schemas.microsoft.com/ws/2014/01/identity/claims/anchorclaimtype",
Value = "http://schemas.xmlsoap.org/ws/2005/05/identity/claims/upn");

This rule issues a claim of type anchorclaimtype . The claim tells the relying party to use UPN as the user's
immutable ID.
12. Click Finish .
13. Click OK to complete the wizard.
 Azure AD identity and access management for AWS
12/18/2020 • 16 minutes to read • Edit Online

Amazon Web Services (AWS) accounts that support critical workloads and highly sensitive information need
strong identity protection and access control. Azure Active Directory (Azure AD) is a cloud-based, comprehensive,
centralized identity and access management solution that can help secure and protect AWS accounts and
environments. Azure AD provides centralized single sign-on (SSO) and strong authentication through multi-factor
authentication (MFA) and Conditional Access policies. Azure AD supports AWS role-based identities and access
control.
Many organizations that use AWS already rely on Azure AD for Microsoft 365 or hybrid cloud identity and access
protection. These organizations can quickly and easily deploy Azure AD for their AWS accounts, often without
additional cost. Other, advanced Azure AD features like Privileged Identity Management (PIM) and Advanced
Identity Protection can help protect the most sensitive AWS accounts.
Azure AD easily integrates with other Microsoft security solutions like Microsoft Cloud App Security (MCAS) and
Azure Sentinel. For more information, see MCAS and Azure Sentinel for AWS. Microsoft security solutions are
extensible and have multiple levels of protection. Organizations can implement one or more of these solutions
along with other types of protection for a full security architecture that protects current and future AWS
deployments.
This article provides AWS identity architects, administrators, and security analysts with immediate insights and
detailed guidance for deploying Azure AD identity and access solutions for AWS. You can configure and test these
solutions without impacting your existing identity providers and AWS account users until you're ready to switch
over.

Architecture
AWS creates a separate Identity and Access Management (IAM) store for each account it creates. The following
diagram shows the standard setup for an AWS environment with a single AWS account:
The root user fully controls the AWS account, and delegates access to other identities. The AWS IAM principal
provides a unique identity for each role and user that needs to access the AWS account. AWS IAM can protect each
root, principal, and user account with a complex password and basic MFA.
Many organizations need more than one AWS account, resulting in identity silos that are complex to manage:

To allow centralized identity management and avoid having to manage multiple identities and passwords, most
organizations want to use single sign-on for platform resources. Some AWS customers rely on server-based
Microsoft Active Directory for SSO integration. Other customers invest in third-party solutions to synchronize or
federate their identities and provide SSO.
Azure AD provides centralized identity management with strong SSO authentication. Almost any app or platform
that follows common web authentication standards, including AWS, can use Azure AD for identity and access
management.
Many organizations already use Azure AD to assign and protect Microsoft 365 or hybrid cloud identities.
Employees use their Azure AD identities to access email, files, instant messaging, cloud applications, and on-
premises resources. You can quickly and easily integrate Azure AD with your AWS accounts to let administrators
and developers sign in to your AWS environments with their existing identities.
The following diagram shows how Azure AD can integrate with multiple AWS accounts to provide centralized
identity and access management:
Azure AD offers several capabilities for direct integration with AWS:
SSO across legacy, traditional, and modern authentication solutions.
MFA, including integration with several third-party solutions from Microsoft Intelligent Security Association
(MISA) partners.
Powerful Conditional Access features for strong authentication and strict governance. When users attempt to
gain access to the AWS Management Console or AWS resources, Azure AD uses both Conditional Access policy-
defined controls and risk-based assessments to manage authentication and authorization in real time.
Large-scale threat detection and automated response. Azure AD processes over 30 billion authentication
requests per day, along with trillions of signals about threats worldwide.
Privileged Access Management (PAM) to enable Just-In-Time (JIT) provisioning to specific resources.
Advanced Azure AD identity management
Other advanced Azure AD features can provide additional layers of control for the most sensitive AWS accounts.
Azure AD Premium P2 licenses include these advanced features:
Privileged Identity Management (PIM) to provide advanced controls for all delegated roles within
Azure and Microsoft 365. For example, instead of an administrator always using the Global Admin role, they
have permission to activate the role on demand. This permission deactivates after a set time limit like one
hour. PIM logs all activations, and has additional controls that can further restrict the activation capabilities.
PIM further protects your identity architecture by ensuring additional layers of governance and protection
before administrators can make changes.
You can expand PIM to any delegated permission by controlling access to custom groups, such as the ones
you created for access to AWS roles. For more information about deploying PIM, see Deploy Azure AD
Privileged Identity Management.
Advanced Identity Protection increases Azure AD sign-in security by monitoring user or session risk.
User risk defines the potential of the credentials being compromised, such as the user ID and password
appearing in a publicly released breach list. Session risk determines whether the sign-in activity comes from
a risky location, IP address, or other indicator of compromise. Both detection types draw on Microsoft's
comprehensive threat intelligence capabilities.
For more information about Advanced Identity Protection, see the Azure AD Identity Protection security
overview.
 Microsoft Defender for Identity protects identities and services running on Active Directory domain
controllers by monitoring all activity and threat signals. Defender for Identity identifies threats based on
real-life experience from investigations of customer breaches. Defender for Identity monitors user behavior
and recommends attack surface reductions to prevent advanced attacks like reconnaissance, lateral
movement, and domain dominance.
For more information about Defender for Identity, see What is Microsoft Defender for Identity.

Security recommendations
The following principles and guidelines are important for any cloud security solution:
Ensure that the organization can monitor, detect, and automatically protect user and programmatic access
into cloud environments.
Continually review current accounts to ensure identity and permission governance and control.
Follow least privilege and zero trust principles. Make sure that each user can access only the specific
resources they require, from trusted devices and known locations. Reduce the permissions of every
administrator and developer to provide only the rights they need for the role they're performing. Review
regularly.
Continuously monitor platform configuration changes, especially if they provide opportunities for privilege
escalation or attack persistence.
Prevent unauthorized data exfiltration by actively inspecting and controlling content.
Take advantage of solutions you might already own like Azure AD Premium P2 that can increase security
without additional expense.
Basic AWS account security
To ensure basic security hygiene for AWS accounts and resources:
Review the AWS security guidance at Best practices for securing AWS accounts and resources.
Reduce the risk of uploading and downloading malware and other malicious content by actively inspecting
all data transfers through the AWS Management Console. Content that uploads or downloads directly to
resources within the AWS platform, such as web servers or databases, might need additional protection.
Consider protecting access to other resources, including:
Resources created within the AWS account.
Specific workload platforms, like Windows Server, Linux Server, or containers.
Devices that administrators and developers use to access the AWS Management Console.
AWS IAM security
A key aspect of securing the AWS Management Console is controlling who can make sensitive configuration
changes. The AWS account root user has unrestricted access. The security team should fully control the root user
account to prevent it from signing in to the AWS Management Console or working with AWS resources.
To control the root user account:
Consider changing the root user sign-in credentials from an individual's email address to a service account that
the security team controls.
Make sure the root user account password is extremely complex, and enforce MFA for the root user.
Monitor logs for instances of the root user account being used to sign in.
Use the root user account only in emergencies.
Use Azure AD to implement delegated administrative access rather than using the root user for administrative
 tasks.
Clearly understand and review other AWS IAM account components for appropriate mapping and assignments.
By default, an AWS account has no IAM users until the root user creates one or more identities to delegate
access. A solution that synchronizes existing users from another identity system, such as Microsoft Active
Directory, can also automatically provision IAM users.
IAM policies provide delegated access rights to AWS account resources. AWS provides over 750 unique IAM
policies, and customers can also define custom policies.
IAM roles attach specific policies to identities. Roles are the way to administer role-based access control
(RBAC). The current solution uses External Identities to implement Azure AD identities by assuming IAM
roles.
IAM groups are also a way to administer RBAC. Instead of assigning IAM policies directly to individual IAM
users, create an IAM group, assign permissions by attaching one or more IAM policies, and add IAM users to
the group to inherit the appropriate access rights to resources.
Some IAM service accounts must continue to run in AWS IAM to provide programmatic access. Be sure to review
these accounts, securely store and restrict access to their security credentials, and rotate the credentials regularly.

Plan and prepare

To prepare for deployment of Azure security solutions, review and record current AWS account and Azure AD
information. If you have more than one AWS account deployed, repeat these steps for each account.
1. In the AWS Billing Management Console, record the following current AWS account information:
AWS Account Id , a unique identifier.
Account Name or root user.
Payment method , whether assigned to a credit card or a company billing agreement.
Alternate contacts who have access to AWS account information.
Security questions securely updated and recorded for emergency access.
AWS regions enabled or disabled to comply with data security policy.
2. In the AWS IAM Management Console, review and record the following AWS IAM components:
Groups that have been created, including detailed membership and role-based mapping policies
attached.
Users that have been created, including the Password age for user accounts, and the Access key age
for service accounts. Also confirm that MFA is enabled for each user.
Roles . There are two default service-linked roles, AWSSer viceRoleForSuppor t and
AWSSer viceRoleForTrustedAdvisor . Record any other roles, which are custom. These roles link to
permission policies, to use for mapping roles in Azure AD.
Policies . Out-of-the-box policies have AWS managed , Job function , or Customer managed in the
Type column. Record all other policies, which are custom. Also record where each policy is assigned,
from the entries in the Used as column.
Identity providers , to understand any existing Security Assertion Markup Language (SAML) identity
providers. Plan how to replace the existing identity providers with the single Azure AD identity provider.
3. In the Azure Active Directory portal, review the Azure AD tenant:
Assess Tenant information to see whether the tenant has an Azure AD Premium P1 or P2 license. A P2
license provides Advanced Azure AD identity management features.
Assess Enterprise applications to see whether any existing applications use the AWS application type,
as shown by http://aws.amazon.com/ in the Homepage URL column.
Plan Azure AD deployment
The Azure AD deployment procedures assume that Azure AD is already configured for the organization, such as for
a Microsoft 365 implementation. Accounts can be synchronized from an Active Directory domain, or can be cloud
accounts created directly in Azure AD.
Plan RBAC
If the AWS installation already uses IAM groups and IAM roles to delegate permissions for human and
programmatic access, you can map that existing structure to new Azure AD user accounts and security groups.
If the AWS account doesn't have a strong RBAC implementation, start by working on the most sensitive access:
1. Update the AWS account root user.
2. Review the AWS IAM users, groups, and roles that are attached to the IAM policy AdministratorAccess .
3. Work through the other assigned IAM policies, starting with policies that can modify, create, or delete
resources and other configuration items. You can identify policies in use by looking at the Used as column.
Plan migration
Azure AD centralizes all authentication and authorization. You can plan and configure user mapping and RBAC
without impacting administrators and developers until you're ready to enforce the new methods.
The high-level process for migrating from AWS IAM accounts to Azure AD is as follows. For detailed instructions,
see Deployment.
1. Map IAM policies to Azure AD roles, and use RBAC to map roles to security groups.
2. Replace each IAM user with an Azure AD user who is a member of the appropriate security groups to sign in
and gain appropriate permissions.
3. Test by asking each user to sign in to AWS with their Azure AD account and confirm that they have the
appropriate access level.
4. Once the user confirms Azure AD access, remove the AWS IAM user account. Repeat the process for each
user until they're all migrated.
For service accounts and programmatic access, use the same approach. Update each application that uses the
account to use an equivalent Azure AD user account instead.
Make sure any remaining AWS IAM users have very complex passwords with MFA enabled, or an access key that's
replaced regularly.
The following diagram shows an example of the configuration steps and final policy and role mapping across
Azure AD and AWS IAM:
Deploy Azure AD for AWS SSO
The following procedure implements SSO for the example roles AWS Administrators and AWS Developers .
Repeat this process for any additional roles you need.
This procedure covers the following steps:
1. Create a new Azure AD enterprise application.
2. Configure Azure AD SSO for AWS.
3. Enable Azure AD to provision AWS IAM roles.
4. Update role mapping.
5. Test Azure AD SSO into AWS Management Console.
The following links provide full detailed implementation steps and troubleshooting:
Microsoft Docs tutorial: Azure AD SSO integration with AWS
AWS tutorial: Azure AD to AWS SSO using the SCIM protocol
Create a new Azure AD enterprise application
AWS administrators and developers use an enterprise application to sign in to Azure AD for authentication, then
redirect to AWS for authorization and access to AWS resources. The simplest method to see the application is by
signing in to https://myapps.microsoft.com , but you can also publish the unique URL anywhere that provides easy
access.
Follow the instructions in Add Amazon Web Services (AWS) from the gallery to set up the enterprise application.
If there's more than one AWS account to administer, such as DevTest and Production, use a unique name for the
enterprise application that includes an identifier for the company and specific AWS account.
Configure Azure AD SSO for AWS
1. Follow the steps in Configure and test Azure AD SSO for AWS through step 2a, Create AWS test user. In step
2a, assign a group to access the application, instead of directly adding a user.
You'll create more than one role, so you can't complete these steps until you finish the AWS configuration.
However, create the following two Azure AD test users and two Azure AD groups now:
User 1: Test-AWSAdmin
User 2: Test-AWSDeveloper
Group 1: AWS-Account1-Administrators
Group 2: AWS-Account1-Developers
Don't add these users or groups to the enterprise application yet.
2. Follow the steps under Configure AWS SSO.
3. Repeat steps 7-10 to create each of the following two roles and assign different permissions based on their
scope of work:
IAM Role 1: AzureAD-Admins
IAM Role 2: AzureAD-Developers
Enable Azure AD to provision AWS IAM roles
In steps 11-20 of Configure AWS SSO, create a new IAM user that acts on behalf of the Azure AD provisioning
agent to allow synchronization of all available AWS IAM roles into Azure AD. AWS needs this IAM user to map
users to roles before they can sign in to the AWS Management Console.
 Make it easy to identify the components you create to support this integration. For example, name service
accounts with a standard naming convention like "Svc-".
Be sure to document all new items.
Make sure any new credentials include complex passwords that you store centrally for secure lifecycle
management.
Based on these configuration steps, you can diagram the interactions like this:

Update role mapping

Because you're using two roles, carry out the following additional steps:
1. Confirm that the provisioning agent can see at least two roles:

2. Go to Users and groups and select Add User .

3. Select AWS-Account1-Administrators .
4. Select the associated role.
5. Repeat the preceding steps for each group-role mapping. Once complete, you should have two Azure AD
groups correctly mapped to AWS IAM roles:

If you can't see or select a role, go back to the Provisioning page to confirm successful provisioning in the Azure
AD provisioning agent, and make sure the IAM User account has the correct permissions. You can also restart the
provisioning engine to attempt the import again:

Test Azure AD SSO into AWS Management Console

Test signing-in as each of the test users to confirm that the SSO works.
1. Launch a new private browser session to ensure that other stored credentials don't conflict with testing.
2. Go to https://myapps.microsoft.com , using the Test-AWSAdmin or Test-AWSDeveloper Azure AD user
account credentials you created previously.
3. You should see the new icon for the AWS Console app. Select the icon, and follow any authentication
prompts:

4. Once you're signed into the AWS Console, navigate the features to confirm that this account has the
appropriate delegated access.
5. Notice the naming format for the user sign-in session:
ROLE / UPN / AWS Account Number
You can use this user sign-in session information for tracking user sign-in activity in MCAS or Azure
Sentinel.

6. Sign out, and repeat the process for the other test user account to confirm the differences in role mapping
and permissions.
Enable Conditional Access
To create a new Conditional Access policy that requires MFA:
1. In the Azure portal, navigate to Azure AD > Security , and then select Conditional Access .
2. In the left navigation, select Policies .

3. Select New policy , and complete the form as follows:

Name : Enter AWS Console – MFA
Users and Groups : Select the two role groups you created earlier:
AWS-Account1-Administrators
AWS-Account1-Developers
Grant : Select Require multi-factor authentication
4. Set Enable policy to On .
5. Select Create . The policy takes effect immediately.
6. To test the Conditional Access policy, sign out of the testing accounts, open a new in-private browsing
session, and sign in with one of the role group accounts. You see the MFA prompt:
7. Complete the MFA setup process. It's best to use the mobile app for authentication, instead of relying on
SMS.

You might need to create several Conditional Access policies to meet business needs for strong authentication.
Consider the naming convention you use when creating the policies to ensure ease of identification and ongoing
maintenance. Also, unless MFA is already widely deployed, make sure the policy is scoped to impact only the
intended users. Other policies should cover other user groups' needs.
Once you enable Conditional Access, you can impose additional controls such as PAM and just-in-time (JIT)
provisioning. For more information, see What is automated SaaS app user provisioning in Azure AD.
If you have MCAS, you can use Conditional Access to configure MCAS session policies. For more information, see
Configure Azure AD session policies for AWS activities.
See also
For in-depth coverage and comparison of Azure and AWS features, see the Azure for AWS professionals content
set.
For security guidance from AWS, see Best practices for securing AWS accounts and resources.
For the latest Microsoft security information, see www.microsoft.com/security.
For full details of how to implement and manage Azure AD, see Securing Azure environments with Azure Active
Directory.
Security and identity on Azure and AWS.
AWS tutorial: Azure AD with IDP SSO.
Microsoft tutorial: SSO for AWS.
PIM deployment plan.
Identity protection security overview.
What is Microsoft Defender for Identity?
Connect AWS to Microsoft Cloud App Security.
How Cloud App Security helps protect your Amazon Web Services (AWS) environment.
 Vision with Azure IoT Edge
12/18/2020 • 4 minutes to read • Edit Online

Visual inspection of products, resources, and environments has been a core practice for most enterprises, and was
until recently, a very manual process. An individual, or a group of individuals, would be responsible for manually
inspecting the assets or the environment. Depending on the circumstances, this could become inefficient,
inaccurate, or both, due to human error and limitations.
To improve the efficacy of visual inspection, enterprises began turning to deep learning artificial neural networks
known as convolutional neural networks (or CNNs), to emulate human vision for analysis of images and video.
Today this is commonly called computer vision, or simply Vision AI. Artificial intelligence for image analytics spans
a wide variety of industries, including manufacturing, retail, healthcare, and the public sector, and an equally wide
area of use cases.
Vision for quality assurance - In manufacturing environments, Vision AI can be very helpful with quality
inspection of parts and processes with a high degree of accuracy and velocity. An enterprise pursuing this
path automates the inspection of a product for defects to answer questions such as:
Is the manufacturing process producing consistent results?
Is the product assembled properly?
Can there be an earlier notification of a defect to reduce waste?
How to leverage drift in the computer vision model to prescribe predictive maintenance?
Vision for safety - In any environment, safety is a fundamental concern for every enterprise, and the
reduction of risk is a driving force for adopting Vision AI. Automated monitoring of video feeds to scan for
potential safety issues provides critical time to respond to incidents, and opportunities to reduce exposure to
risk. Enterprises looking at Vision AI for this use case are commonly trying to answer questions such as:
How compliant is the workforce with using personal protective equipment?
How often are people entering unauthorized work zones?
Are products being stored in a safe manner?
Are there unreported close calls in a facility, or pedestrian/equipment “near misses”?

Why vision on the Edge

Over the past decade, computer vision has become a rapidly evolving area of focus for enterprises, as cloud-native
technologies such as containerization, have enabled portability and migration of this technology towards the
network edge. For instance, custom vision inference models trained in the cloud can be easily containerized for use
in an Azure IoT Edge runtime-enabled device.
The rationale behind migrating workloads from the cloud to the edge for Vision AI generally falls into two
categories – performance and cost.
On the performance side of the equation, exfiltrating large quantities of data can cause an unintended performance
strain on existing network infrastructure. Additionally, the latency of sending images and/or video streams to the
cloud to retrieve results may not meet the needs of the use case. For instance, a person straying into an
unauthorized area may require immediate intervention, and that scenario can affect latency when every second
counts. Positioning the inferencing model near the point of ingestion, allows for near real-time scoring of the
image. It also allows alerting to be performed either locally or through the cloud, depending on the network
topology.
In terms of cost, sending all of the data to the cloud for analysis could significantly impact the ROI or return on
investment of a Vision AI initiative. With Azure IoT Edge, a Vision AI module can be designed to only capture the
relevant images with a reasonable confidence level based on the scoring. This significantly limits the amount of
data being sent.

Camera considerations
Camera is understandably a very important component of an Azure IoT Edge Vision solution. To learn what
considerations should be taken for this component, proceed to Camera selection in Azure IoT Edge Vision.

Hardware acceleration
To bring AI to the edge, the edge hardware should be able to run the powerful AI algorithms. To know the hardware
capabilities required for IoT Edge Vision, proceed to Hardware acceleration in Azure IoT Edge Vision.

Machine learning
Machine learning can be challenging for the data on the edge, due to resource restrictions of edge devices, limited
energy budget, and low compute capabilities. See Machine learning and data science in Azure IoT Edge Vision to
understand the key considerations in designing the machine learning capabilities of your IoT Edge Vision solution.

Image storage
Your IoT Edge Vision solution cannot be complete without careful consideration of how and where the images
generated will be stored. Read Image storage and management in Azure IoT Edge Vision for a thorough discussion.

Alerts
Your IoT Edge device may need to respond to various alerts in its environment. See Alert persistence in Azure IoT
Edge Vision to understand the best practices in managing these alerts.

User interface
The user interface or UI of your IoT Edge Vision solution will vary based on the target user. The article User interface
in Azure IoT Edge Vision discusses the main UI considerations.

Next steps
This series of articles demonstrate how to build a complete vision workload using Azure IoT Edge devices. For
further information, you may refer to the product documentation as following:
Azure IoT Edge documentation
Tutorial: Perform image classification at the edge with Custom Vision Service
Azure Machine Learning documentation
Azure Kinect DK documentation
MMdnn tool
ONNX
 Camera selection in Azure IoT Edge Vision
12/18/2020 • 8 minutes to read • Edit Online

One of the most critical components in any AI Vision workload is selecting the right camera. The items being
identified by this camera must be presented in such a way that the artificial intelligence or machine learning
models can evaluate them correctly. An in-depth understanding of the different camera types is required to
understand this concept.

NOTE
There are different manufacturers for area , line , and smar t cameras. Instead of recommending any one vendor over
another, Microsoft recommends that you select a vendor that fits your specific needs.

Types of cameras
Area scan cameras
This camera type generates the traditional camera image, where a 2D image is captured and then sent over to the
Edge hardware to be evaluated. This camera typically has a matrix of pixel sensors.
As the name suggests, area scan cameras look at a large area and are great at detecting change in an area.
Examples of workloads that could use an area scan camera would be workplace safety, or detecting or counting
objects (people, animals, cars, and so on) in an environment.
Examples of manufacturers of area scan cameras are Basler, Axis, Sony, Bosch, FLIR, Allied Vision.
Line scan cameras
Unlike the area scan cameras, the line scan camera has a single row of linear pixel sensors. This allows the camera
to take one-pixel width images in quick successions, and then stitches them together into a video stream. This
video stream is then sent over to an Edge device for processing.
Line scan cameras are great for vision workloads where the items to be identified are either moving past the
camera, or need to be rotated to detect defects. The line scan camera would then be able to produce a continuous
image stream for evaluation. Examples of workloads that would work best with a line scan camera are:
an item defect detection on parts that are moved on a conveyer belt,
workloads that require spinning to see a cylindrical object, or
any workload that requires rotation.
Examples of manufacturers of line scan cameras are Basler, Teledyne Dalsa, Hamamatsu Corporation, DataLogic,
Vieworks, and Xenics.
Embedded smart cameras
This type of camera can use either an area scan or a line scan camera for capturing the images, although a line scan
smart camera is rare. An embedded smart camera can not only acquire an image, but can also process that image
as it is a self-contained stand-alone system. They typically have either an RS232 or an Ethernet port output, which
allows them to be integrated directly into a PLC or other IIoT interfaces.
Examples of manufacturers of embedded smart cameras are Basler, Lesuze Electronics.

Camera features
Sensor size
This is one of the most important factors to evaluate in any vision workload. A sensor is the hardware within a
camera that captures the light and converts into signals, which then produce an image. The sensor contains
millions of semiconducting photodetectors called photosites. A higher megapixel count does not always result in a
better image. For example, let’s look at two different sensor sizes for a 12-megapixel camera. Camera A has a ½
inch sensor with 12 million photosites and camera B has a 1-inch sensor with 12 million photosites. In the same
lighting conditions, the camera that has the 1-inch sensor will be cleaner and sharper. Many cameras typically used
in vision workloads have a sensor sized between ¼ inch to 1 inch. In some cases, much larger sensors might be
required.
If a camera has a choice between a larger sensor or a smaller sensor, some factors deciding why you might choose
the larger sensor are:
need for precision measurements,
lower light conditions,
shorter exposure times, or fast-moving items.
Resolution
This is another important factor to both line scan and area scan camera workloads. If your workload must identify
fine features, such as the writing on an IC chip, then you need greater resolution cameras. If your workload is trying
to detect a face, then higher resolution is required. And if you need to identify a vehicle from a distance, again a
higher resolution will be required.
Speed
Sensors come in two types- CCD and CMOS . If the vision workload requires high number of images to be captured
per second, then two factors will come into play. The first is how fast is the connection on the interface of the
camera. The second is what type of sensor it is. CMOS sensors have a direct readout from the photosites, because
of which they typically offer a higher frame rate.

NOTE
There are several other camera features to consider when selecting the correct camera for your vision workload. These
include lens selection, focal length, monochrome, color depth, stereo depth, triggers, physical size, and support. Sensor
manufacturers can help you understand the specific feature that your application may require.

Camera placement
The items that you are capturing in your vision workload will determine the location and angles that the camera
should be placed. The camera location can also affect the sensor type, lens type, and camera body type.
There are several different factors that can weigh into the overall decision for camera placement. Two of the most
critical ones are the lighting and the field of view.
Camera lighting
In a computer vision workload, lighting is a critical component to camera placement. There are several different
lighting conditions. While some of the lighting conditions would be useful for one vision workload, it might
produce an undesirable effect in another. Types of lighting that are commonly used in computer vision workloads
are:
Direct lighting: This is the most commonly used lighting condition. This light source is projected at the
object to be captured for evaluation.
Line lighting: This is a single array of lights that are most used with line scan camera applications. This
creates a single line of light at the focus of the camera.
 Diffused lighting: This type of lighting is used to illuminate an object but prevent harsh shadows and is
mostly used around specular objects.
Back lighting: This type of light source is used behind the object, producing a silhouette of the object. This
is most useful when taking measurements, edge detection, or object orientation.
Axial diffused lighting: This type of light source is often used with highly reflective objects, or to prevent
shadows on the part that will be captured for evaluation.
Custom Grid lighting: This is a structured lighting condition that lays out a grid of light on the object, the
intent is to have a known grid projection to then provide more accurate measurements of components,
parts, placement of items, and so on.
Strobe lighting: Strobe lighting is used for high speed moving parts. The strobe must be in sync with the
camera to take a freeze of the object for evaluation, this lighting helps to prevent motion blurring effect.
Dark Field lighting: This type of light source uses several lights in conjunction with different angles to the
part to be captured. For example, if the part is laying flat on a conveyor belt the lights would be placed at a
45-degree angle to it. This type of lighting is most useful when looking at highly reflective clear objects and
is most commonly used with lens scratch detections.
The figure below shows the angular placement of light:

Field of view
In a vision workload, you need to know the distance to the object that you are trying to evaluate. This also will play
a part in the camera selection, sensor selection, and lens configuration. Some of the components that make up the
field of view are:
Distance to object(s): For example, is the object being monitored with computer vision on a conveyor belt
and the camera is two feet above it, or is the object across a parking lot? As the distance changes, so does the
camera’s sensors and lens configurations.
Area of coverage: Is the area that the computer vision is trying to monitor small or large? This has direct
correlation to the camera’s resolution, lens, and sensor type.
Direction of the sun: If the computer vision workload is outside, such as monitoring a job construction site
for worker safety, will the camera be pointed in the sun at any time? Keep in mind that if the sun is casting a
shadow over the object that the vision workload is monitoring, items might be a bit obscured. Also, if the
camera is getting direct sunlight in the lens, the camera might be blinded until the angle of the sun changes.
Camera angle to the object(s): Angle of the camera to the object that the vision workload is monitoring is
also a critical component to think about. If the camera is too high, it might miss the details that the vision
 workload is trying to capture, and the same may be true if it is too low.

Communication interface
In building a computer vision workload, it is also important to understand how the system will interact with the
output of the camera. Below are a few of the standard ways that a camera will communicate to IoT Edge:
Real Time Streaming Protocol (RTSP): RTSP is a protocol that transfers real-time video data from a
device (in our case, the camera) to an endpoint device (Edge compute) directly over a TCP/IP connection. It
functions in a client-server application model that is at the application level in the network.
Open Network Video Interface Forum (ONVIF): A global and open industry forum that is developing
open standards for IP-based cameras. This standard is aimed at standardization of communication between
the IP camera and down stream systems, interoperability, and open source.
USB: Unlike RTSP and ONVIF, USB connected cameras connect over the Universal Serial Bus directly on the
Edge compute device. This is less complex, however, it limits the distance that the camera can be placed
away from the Edge compute.
Camera Serial Interface: CSI specification is from Mobile Industry Processor Interface (MIPI). This
interface describes how to communicate between a camera and a host processor.
There are several standards defined for CSI:
CSI-1 : This was the original standard that MIPI started with.
CSI-2 : This standard was released in 2005, and uses either D-PHY or C-PHY as physical layers options.
This is further divided into several layers:
Physical Layer (C-PHY, D-PHY)
Lane Merger layer
Low-Level Protocol Layer
Pixel to Byte Conversion Layer
Application layer
The specification was updated in 2017 to v2, which added support for RAW-24 color depth, Unified Serial
Link, and Smart Region of Interest.

Next steps
Now that you know the camera considerations for your IoT Edge Vision workload, proceed to setting up the right
hardware for your workload. Read Hardware acceleration in Azure IoT Edge Vision for more information.
 Hardware acceleration in Azure IoT Edge Vision
12/18/2020 • 2 minutes to read • Edit Online

Along with the camera selection, one of the other critical decisions in Vision on the Edge projects is hardware
acceleration.
The following sections describe the key components of the underlying hardware.

CPU
The Central Processing Unit (CPU) is your default compute for most processes running on a computer. It is
designed for general purpose compute. For some vision workloads where timing is not as critical, this might be a
good option. However, most workloads that involve critical timing, multiple camera streams, and/or high frame
rates, will require more specific hardware acceleration.

GPU
Many people are familiar with the Graphics Processing Unit (GPU) as this is the de-facto processor for any high-
end PC graphics card. In recent years, the GPU has been leveraged in high performance computer (HPC) scenarios,
in data mining, and in computer AI/ML workloads. The GPU’s massive potential of parallel computing can be used
in a vision workload to accelerate the processing of pixel data. The downside to a GPU is its higher power
consumption, which is a critical factor to consider for your vision workload.

FPGA
Field Programmable Gate Arrays are reconfigurable hardware accelerators. These powerful accelerators allow for
the growth of Deep Learning Neural networks, which are still evolving. These accelerators have millions of
programmable gates, hundreds of I/O pins, and an exceptional compute power in the trillions of tera-MAC’s. There
also many different libraries available for FPGAs that are optimized for vision workloads. Some of these libraries
also include preconfigured interfaces to connect to downstream cameras and devices. One area that FPGAs tend to
fall short on is floating point operations. However, manufacturers are currently working on this issue and have
made many improvements in this area.

ASIC
Application-Specific Integrated Circuit is by far the fastest accelerator on the market today. While they are the
fastest, they are the hardest to change as they are manufactured to function for a specific task. These custom chips
are gaining popularity due to size, power per watt performance, and IP protection. This is because the IP burned
into the ASIC accelerator is much harder to backwards engineer proprietary algorithms.

Next steps
Proceed to learn what considerations go into place for Machine learning and data science in Azure IoT Edge Vision.
 Machine learning and data science in Azure IoT Edge
Vision
12/18/2020 • 12 minutes to read • Edit Online

The process of designing the machine learning (ML) approach for a vision on the edge scenario is one of the
biggest challenges in the entire planning process. It is important to understand how to consider and think about
ML in the context of edge devices.
To begin using machine learning to address business problems and pain points, consider the following points:
Always consider first how to solve the problem without ML or with a simple ML algorithm.
Have a plan to test several ML architectures as they will have different capacities to "learn".
Have a system in place to collect new data from the device to retrain an ML model.
For poorly performing ML models, often a simple fix is to add more representative data to the training process
and ensure it has variability with all classes represented equally.
Remember, this is often an iterative process with both the choice of data and choice of architecture being
updated in the exploratory phase.
It is not an easy space and, for some, a very new way of thinking. It is a data driven process. Careful planning will
be critical to successful results especially on very constrained devices.
It is always critical to clearly define the problem to be solved as the data science and machine learning approach
will depend upon this. It is also very important to consider what type of data will be encountered in the edge
scenario as this will determine the kind of ML algorithm that should be used.
Even at the start, before training any models, real world data collection and examination will help this process
greatly and new ideas could even arise. This article will discuss data considerations in detail. Of course, the
equipment itself will help determine the ML approach with regard to device attributes like limited memory,
compute, and/or power consumption limits.
Fortunately, data science and machine learning are iterative processes, so if the ML model has poor performance,
there are many ways to address issues through experimentation. This article will also discuss considerations
around ML architecture choices. Often, there will be some trial and error involved as well.

Machine learning data

Both the source(s) and attributes of data will dictate how the intelligent edge system is built. For vision, it could be
images, videos, or even LiDAR, as the streaming signal. Regardless of the signal, when training an ML model and
using it to score new data (called inferencing), domain knowledge will be required. This includes experience in
designing and using ML algorithms or neural network architectures and expertise deploying them to the
specialized hardware. Below are a few considerations related to ML. However, it is recommended to gain some
deeper knowledge in order to open up more possibilities or find an ML expert with edge experience to help with
the project.
Collecting and using a balanced dataset is critical, and it should equally represent all classes or categories. When
the ML model is trained on a dataset, generally that dataset has been split into train, validate, and test subsets. The
purpose of these subsets is as follows:
The training dataset is used for the actual model training over many passes or iterations (often called epochs).
Throughout the training process, the model is spot-checked for how well it is doing on the validation dataset.
After a model is done training, the final step is to pass the test dataset through it and assess how well it did as a
 proxy to the real-world.

NOTE
Be wary of optimizing for the test dataset, in addition to the training dataset, once one test has been run. It might be good
to have a few different test datasets available.

Some good news is that in using deep learning, often costly and onerous feature engineering, featurizations, and
preprocessing can be avoided because of how deep learning works to find signal in noise better than traditional
ML. However, in deep learning, transformations may still be utilized to clean or reformat data for model input
during training as well as inference. The same capacity needs to be used in training and when the model is scoring
new data.
When advanced preprocessing is used such as denoising, adjusting brightness or contrast, or transformations like
RGB to HSV, it must be noted that this can dramatically change the model performance for the better or,
sometimes, for the worse. In general, it is part of the data science exploration process and sometimes it is
something that must be observed once the device and other components are placed in a real-world location.
After the hardware is installed into its permanent location, the incoming data stream should be monitored for data
drift.
Data drift is the deviation due to changes in the current data compared to the original. Data drift will often result
in a degradation in model performance (like accuracy), although this is not the only cause of decreased
performance (for example, hardware or camera failure).
There should be an allowance for data drift testing in the system. This new data should also be collected for
another round of training. The more representative data collected for training, the better the model will perform in
almost all cases. So, preparing for this kind of collection is always a good idea.
In addition to using data for training and inference, new data coming from the device could be used to monitor the
device, camera or other components for hardware degradation.
In summary, here are the key considerations:
Always use a balanced dataset with all classes represented equally.
The more representative data used to train a model, the better.
Have a system in place to collect new data from device to retrain.
Have a system in place to test for data drift.
Only run a test set through a new ML model once. If you iterate and retest on the same test set, this could cause
overfitting to the test set in addition to the training set.

Machine learning architecture choices

Machine learning (ML) architecture is the layout of the mathematical operations that process input into the desired
and actionable output. For instance, in deep learning this would be the number of layers and neurons in each layer
of a deep neural network as well as their arrangement. It is important to note that there is no guarantee that the
performance metric goal (such as, high enough accuracy) for any given ML architecture will be achieved. To
mitigate this, several different architectures should be considered. Often, two or three different architectures are
tried before a choice is made. Remember that this is often an iterative process; both the choice of data and the
choice of architecture may be updated in the exploratory phase of the development process.
It helps to understand the issues that can arise when training an ML model that may only be seen after training or,
even at the point of inferencing on device. Overfitting and underfitting are some of the common issues found
during the training and testing process.
 Overfitting - Overfitting can give a false sense of success because the performance metric (like accuracy)
might be very good when the input data looks like the training data. However, overfitting can occur when
the model fits to the training data too closely and cannot generalize well to new data. For example, it may
become apparent that the model only performs well indoors because the training data was from an indoor
setting.
Overfitting can be caused by following issues:
The model learned to focus on incorrect, non-representative features specifically found in the training
dataset.
The model architecture may have too many learnable parameters, correlated to the number of layers in a
neural network and units per layer. A model's memorization capacity is determined by the number of
learnable parameters.
Not enough complexity or variation is found in the training data.
The model is trained over too many iterations.
There may be other reasons for good performance in training and significantly worse performance in
validation and testing, which are out of scope for this article.
Underfitting - Underfitting happens when the model has generalized so well that it cannot tell the
difference between classes with confidence. For example, the training loss will still be unacceptably high.
Underfitting can be caused by following issues:
Not enough samples available in training data.
The model is trained for too few iterations, in other words, it's too generalized.
Other reasons related to the model not being able to recognize any objects, or poor recognition and loss
values during training. The assessment values used to direct the training process pass through a process
called optimization and weight updates.
There is a trade-off between too much capacity (such as, a large network or a large number of learnable
parameters) and too little capacity. Transfer learning happens when some network layers are set as not trainable, or
frozen. In this situation, increasing capacity would equate to opening up more layers earlier in the network versus
only using the last few layers in training, with the rest remaining frozen.
There is no hard and fast rule for determining number of layers for deep neural networks. So, sometimes several
model architectures must be evaluated within an ML task. However, in general, it is good to start with fewer layers
and/or parameters (such as, smaller networks) and gradually increase the complexity.
Some considerations when coming up with the best architecture choice will include the inference speed
requirements. These include an assessment and acceptance of the speed versus accuracy tradeoff. Often, a faster
inference speed is associated with lower performance. For example, accuracy, confidence or precision could suffer.
A discussion around requirements for the ML training and inferencing will be necessary based upon the
considerations above and any company-specific requirements. For instance, if the company policy allows open-
source solutions to be utilized, it will open up a great deal of ML algorithmic possibilities as most cutting edge ML
work is in the open-source domain.
In summary, here are the key considerations:
Keep an eye out for overfitting and underfitting.
Testing several ML architectures is often a good idea. This is an iterative process.
There will be a trade-off between too much network capacity and too little. However, it is better to start with too
little and build up from there.
There will be a trade-off between speed and your performance metric such as accuracy.
If the performance of the ML model is acceptable, the exploratory phase is complete. This is important to note,
as one can be tempted to iterate indefinitely.
Data science workflows
The data science process for edge deployments has a general pattern. After a clear data-driven problem statement
is formulated, the next steps generally include those shown in the figure below.

Data collection - Data collection or acquisition could be an online image search from a currently deployed
device, or other representative data source. Generally, the more data the better. In addition, the more variability,
the better the generalization.
Data labeling - If only hundreds of images need to be labeled, such as, when using transfer learning, it can be
done in-house. If tens of thousands of images need to be labeled, a vendor could be enlisted for both data
collection and labeling.
Train a model with ML framework - An ML framework such as TensorFlow or PyTorch (both with Python
and C++ APIs) will need to be chosen. Usually this depends upon what code samples are available in open-
source or in-house, as well as the experience of the ML practitioner. Azure ML may be used to train a model
using any ML framework and approach, as it is agnostic of framework and has Python and R bindings, and
many wrappers around popular frameworks.
Conver t the model for inferencing on device - Almost always, a model will need to be converted to work
with a particular runtime. Model conversion usually involves advantageous optimizations like faster inference
and smaller model footprints. This step differs for each ML framework and runtime. There are open-source
interoperability frameworks available such as ONNX and MMdnn.
 Build the solution for device - The solution is usually built on the same type of device as will be used in the
final deployment because binary files created system-specific.
Using runtime, deploy solution to device - Once a runtime is chosen, usually in conjunction with the ML
framework choice, the compiled solution may be deployed. The Azure IoT Runtime is a Docker-based system in
which the ML runtimes may be deployed as containers.
The diagram below shows a sample data science process where open-source tools may be leveraged for the data
science workflow. Data availability and type will drive most of the choices, including the devices/hardware chosen.

If a workflow already exists for the data scientists and app developers, a few other considerations may apply. First,
it is advised to have a code, model, and data versioning system in place. Secondly, an automation plan for code and
integration testing along with other aspects of the data science process, such as triggers, build/release process, and
so on, will help speed up the time to production and cultivate collaboration within the team.
The language of choice can help dictate what API or SDK is used for inferencing and training ML models. This in
turn will then dictate what type of ML model, what type(s) of device, what type of IoT Edge module, and so on, need
to be used. For example, PyTorch has a C++ API for inferencing and training, that works well in conjunction with
the OpenCV C++ API. If the app developer working on the deployment strategy is building a C++ application, or
has this experience, one might consider PyTorch or others (such as TensorFlow or CNTK) that have C++ inferencing
APIs.

Machine learning and data science in a nutshell

In summary, here are the key considerations:
Converting models also involves optimizations such as faster inference and smaller model footprints, critical for
very resource-constrained devices.
The solution will usually need to be built on a build-dedicated device, the same type of device to which the
solution will be deployed.
The language and framework of choice will depend upon both the ML practitioner's experience as well as what
is available in open-source.
The runtime of choice will depend upon the availability of the device and hardware acceleration for ML.
It is important to have a code, model, and data versioning system.

Next steps
Proceed to Image storage and management in Azure IoT Edge Vision article to learn how to properly store the
images created by your IoT Edge Vision solution.
 Image storage and management in Azure IoT Edge
Vision
12/18/2020 • 2 minutes to read • Edit Online

Storage and management of the images involved in a computer vision application is a critical function.
Some of the key considerations for managing these images are:
Ability to store all raw images during training with ease of retrieval for labeling.
Faster storage medium to avoid pipeline bottlenecks and loss.
Storage on the edge as well as in the cloud, as labeling activity can be performed at both places.
Categorization of images for easy retrieval.
Naming and tagging images to link them with inferred metadata.
The combination of Azure Blob Storage, Azure IoT Hub, and Azure IoT Edge allow several potential options for the
storage of image data, such as:
Use of the Azure IoT Edge Blob Storage module, which will automatically sync images to Azure Blob based on
policy.
Storing images to local host file system and uploading to Azure Blob service using a custom module.
Use of a local database to store images, which then are synced to the cloud database.

Typical storage workflow

The IoT Edge Blob Storage module is one of the most powerful and straightforward solutions, and our preferred
approach. A typical workflow using this module might be as follows:
1. Raw messages after ingestion are stored locally on the Edge Blob module, with time stamping and sequence
numbering to uniquely identify the image files.
2. Policy is set on the Edge Blob module for automatic upload to Azure Blob with ordering.
3. To conserve space on the Edge device, automatic deletion after certain time is configured along with retain
while uploading option to ensure all images get synced to the cloud.
4. Local categorization or domain and labeling is implemented using module that can read these images into the
UX. The label data is associated to the image URI along with the coordinates and category.
5. As label data needs to be saved, a local database is preferred to store this metadata, as it will allow easy lookup
for the UX and can be synced to the cloud using telemetry messages.
6. During scoring run, the model detects matching patterns and generates events of interest. This metadata is sent
to cloud via telemetry referring to the image URI and optionally stored in local database for edge UX. The
images continue to be stored to Edge Blob and synced with Azure Blob.

Next steps
How you respond to alerts generated by the AI model is crucial. Learn more about this in Alert persistence in Azure
IoT Edge Vision.
 Alert persistence in Azure IoT Edge Vision
12/18/2020 • 2 minutes to read • Edit Online

In the context of vision on edge, an alert is a response to an event triggered by the AI model. In other words, it is
the inferencing results. The type of event is determined by the training imparted to the model. These events are
separate from operational events raised by the processing pipeline and any event related to the health of the
runtime.

Types of alerts
Some of the common alerts types are:
Image classification
Movement detection
Direction of movement
Object detection
Count of objects
Total count of objects over period of time
Average count of objects over period of time
Alerts are required to be monitored as they drive certain actions. They are critical to operations, being time
sensitive in terms of processing, and are required to be logged for audit and further analysis.

Persistence of alerts
The alerts need to persist locally on the edge where they are raised and then passed on to the cloud for further
processing and storage. This ensures a quick local response and avoids losing critical alerts due to any transient
failures.
Some options to achieve this persistence and cloud syncing are:
Utilize built-in store and forward capability of IoT Edge runtime, which automatically gets synced with Azure IoT
Hub after a lost connectivity.
Persist alerts on host file system as log files, which can be synced periodically to a blob storage in the cloud.
Utilize Azure Blob Edge module, which will sync this data to Azure Blob in cloud based on policies that can be
configured.
Use local database on IoT Edge, such as SQL Edge for storing data, and sync with Azure SQL DB using SQL Data
Sync. Another lightweight database option is the SQLite.
The preferred option is to use the built-in store and forward capability of IoT Edge runtime. This is more suitable
for the alerts due to its time sensitivity, typically small messages sizes, and ease of use.

Next steps
Now you can proceed to work on the User interface in Azure IoT Edge Vision for your vision workload.
 User interface in Azure IoT Edge Vision
12/18/2020 • 9 minutes to read • Edit Online

The user interface requirements of an IoT solution will vary depending on the overall objectives. Four types of user
interfaces are commonly found in IoT solutions:
Administrator : Allows full access to device provisioning, device and solution configuration, user management,
and so on. These features could be provided as part of one solution or as separate solutions.
Consumer : Is only applicable to consumer solutions. They provide similar access to the operator's interface,
but limited to the devices owned by the user.
Operator : Provides centralized access to the operational components of the solution. It typically includes
device management, alerts monitoring, and configuration.
Analytics: Is an interactive dashboard which provides visualization of telemetry and other data analyses.
This article focuses on a simple operator’s user interface and visualization dashboard.

Technology options
Power BI - Power BI is a compelling option for our analytics and virtualization needs. It provides power
features to create customizable interactive dashboards. It also allows connectivity to many popular database
systems and services. It is available as a managed service and as a self-hosted package. The former is the
most popular and recommended option. With Power BI embedded in your solution, you could add
customer-facing reports, dashboards, and analytics in your own applications by using and branding Power
BI as your own. You can reduce developer resources by automating the monitoring, management, and
deployment of analytics, while getting full control of Power BI features and intelligent analytics.
Azure Maps - Another suitable technology for IoT visualizations is Azure Maps which allows you to create
location-aware web and mobile applications using simple and secure geospatial services, APIs, and SDKs in
Azure. You can deliver seamless experiences based on geospatial data with built-in location intelligence from
world-class mobility technology partners.
Azure App Ser vice - Azure App Service is a managed platform with powerful capabilities for building web
and mobile apps for many platforms and mobile devices. It allows developers to quickly build, deploy, and
scale web apps created with popular frameworks, such as .NET, .NET Core, Node.js, Java, PHP, Ruby, or
Python, in containers or running on any supported operating system. You can also meet rigorous,
enterprise-grade performance, security, and compliance requirements by using the fully managed platform
for your operational and monitoring tasks.
Azure SignalR Ser vice - For real-time data reporting, Azure SignalR Service, makes adding real-time
communications to your web application as simple as provisioning a service. An in-depth real-time
communications expertise is not required. It easily integrates with services such as Azure Functions, Azure
Active Directory, Azure Storage, Azure App Service, Azure Analytics, Power BI, Azure IoT, Azure Cognitive
Services, Azure Machine Learning, and more.
To secure your user interface solutions, Azure Active Director y (Azure AD) enterprise identity service provides
single sign-on and multi-factor authentication.
Now let's learn how to build the user interface for some common scenarios.

Scenario 1
Contoso Boards produces high-quality circuit boards used in computers. Their number one product is a
motherboard. Lately, they have been seeing an increase in issues with chip placement on the board. Through their
investigation, they have noticed that the circuit boards are getting placed incorrectly on the assembly line. They
need a way to identify if the circuit board is placed on the assembly line correctly. The data scientists at Contoso
Boards are most familiar with TensorFlow and would like to continue using it as their primary ML model structure.
Contoso Boards has several assembly lines that produce these mother boards. They would also like to centralize
the management of the entire solution.
Considerations in this scenario
Contoso Boards can ask themselves questions such as the following:
What are we analyzing?
Motherboard
Where are we going to view the motherboard from?
Assembly Line Conveyor belt
What camera do we need?
Area or line scan
Color or monochrome
CCD or CMOS sensor
Global or rolling shutter
Frame rate
Resolution
What type of lighting is needed?
Backlighting
Shade
Darkfield
How should the camera be mounted?
Top down
Side view
Angular
What hardware should be used?
CPU
FPGA
GPU
ASIC
Solution
To find the solution that will be useful for Contoso Boards, let's focus on the edge detection. We need to position a
camera directly above at 90 degrees and about 16 inches above the edge par t . Since the conveyer
system moves relatively slowly, we can use an area scan camera with a global shutter . For this use case, our
camera should capture about 30 frames per second . The resolution can be found using the formula of
Res=(Object Size) Divided by (details to be captured) . Based on this formula, Res=16”/8” gives 2MP in x and 4
in y, so we need a camera capable of 4MP . As for the sensor type, we are not fast moving, and really looking for
an edge detection, so a CMOS sensor is better suited. One of the more critical aspects for any vision workload is
lighting. In this application, Contoso Boards should choose to use a white diffused filter back light . This will
make the part look almost black and have a high amount of contrast for edge detection. When it comes to color
options for this application, it is better to be in black and white, as this is what will yield the sharpest edge for the AI
detection model. The data scientists are most familiar with TensorFlow, so learning ONNX or others would slow
down the time for development of the model. Also because there are several assembly lines that will use this
solution, and Contoso Boards would like a centrally managed edge solution, so Azure Stack Edge (with GPU
option) would work well here. Based on the workload, the fact that Contoso Boards already knows TensorFlow, and
this will be used on multiple assembly lines, GPU-based hardware would be the choice for hardware acceleration.
The following figure shows a sample of what the camera would see in this scenario:

Scenario 2
Contoso Shipping recently has had several pedestrian accidents at their loading docks. Most of the accidents are
happening when a truck leaves the loading dock, and the driver does not see a dock worker walking in front of the
truck. Contoso Shipping would like a solution that would watch for people, predict the direction of travel, and warn
the drivers of potential dangers of hitting the workers. The distance from the cameras to Contoso Shipping's server
room is too far for GigE connectivity, however they do have a large WIFI mesh that could be used. Most of the data
scientists at Contoso Shipping are familiar with Open-VINO and they would like to be able to reuse the models on
additional hardware in the future. The solution will also need to ensure that devices are operating as power-
efficiently as possible. Finally, Contoso Shipping needs a way to manage the solution remotely for updates.
Considerations in this scenario
Contoso Shipping can introspect by asking the following questions:
 What are we analyzing?
People and patterns of movement
Where are we going to view the people from?
The loading docks are 165 feet long.
Cameras will be placed 17 feet high to keep with city ordinances.
Cameras will need to be positioned 100 feet away from the front of the trucks.
Camera focus will need to be 10 feet behind the front of the truck, and 10 additional feet in front of the
truck, giving a 20 foot depth on focus.
What camera do we need?
Area or line scan
Color or monochrome
CCD or CMOS sensor
Global or rolling shutter
Frame rate
Resolution
What type of lighting is needed?
Backlighting
Shade
Darkfield
What hardware should be used?
CPU
FPGA
GPU
ASIC
How should the camera be mounted?
Top down
Side view
Angular
Solution
Based on the distance of the loading dock size, Contoso Shipping will require several cameras to cover the entire
dock. The zoning laws that Contoso Shipping must adhere to, require that the surveillance cameras cannot be
mounted higher that 20 feet. In this use case, the average size of a worker is 5 foot 8 inches. The solution must use
the least number of cameras possible.
Formula for field of view (FOV):

For example, look at the following images.

This image is taken with 480 horizontal pixels at 20 foot:
This image is taken with 5184 horizontal pixels at 20 foot:

The red square is shown to illustrate one pixel color.

These images demonstrate the issue with using the wrong resolution camera for a given use case. Lens can impact
the FOV. However, if the wrong sensor is used for that given use case, the results could be less than expected.
With the above in mind, when choosing a camera for the solution required for Contoso Shipping, we need to think
about how many cameras, and at what resolution, are needed to get the correct number of details to detect a
person. Since we are only trying to identify if a person is in the frame or not, our PPF does not need to be around
80, which is needed for facial identification; we can use somewhere around 15-20. That would place the FOV
around 16 feet. A 16-foot FOV would give us about 17.5 pixels per foot, which fits within our required PPF of 15-
20. This would mean that we need a 10MP camera that has a horizontal resolution of ~5184 pixels , and a
lens that would allow for an FOV of 16 feet . The cameras would need to be placed outside, and the choice of
sensor type should not allow for bloom. Bloom is when light hits the sensor and overloads the sensor, causing a
view of almost over-exposure or a white out kind of condition. CMOS is the sensor of choice here.
Contoso operates 24x7 and as such, needs to ensure that nighttime personnel are also protected. Monochrome
handles low light conditions much better compared to color. We are not looking to identify a person based on
color. And monochrome sensors are a little cheaper then color.
How many cameras will it take? Since we have figured out that our cameras can look at a 16 foot path, we can do
simple math. 165 foot dock divided by 16 foot FOV gives us 10.3125 cameras. So the solution would need 11
monochrome 5184 horizontal pixel (or 10MP) CMOS cameras with IPX67 housings or weather boxes .
The cameras would be mounted on 11 poles 100 feet from the trucks at 17f high. Based on the fact that the data
scientists are more familiar with Open-VINO , data models should be built in ONNX . As for what hardware should
be used, a device that can be connected over WIFI, and use as little power as possible, is required. Based on this,
they should look to an FPGA processor . Potentially an ASIC processor could also be utilized, but due to the nature
of how an ASIC processor works, it would not meet the requirement of being able to use the models on different
hardware in the future.

Next steps
This series of articles have demonstrated how to build a complete vision workload using Azure IoT Edge devices.
For further information, you may refer to the product documentation as following:
Azure IoT Edge documentation
Tutorial: Perform image classification at the edge with Custom Vision Service
Azure Machine Learning documentation
Azure Kinect DK documentation
MMdnn tool
ONNX
 Azure Industrial IoT Analytics Guidance
12/18/2020 • 11 minutes to read • Edit Online

This article series shows a recommended architecture for an Industrial IoT (IIoT) analytics solution on Azure using
PaaS (Platform as a service) components. Industrial IoT or IIoT is the application of Internet of Things in the
manufacturing industry.
An IIoT analytics solution can be used to build a variety of applications that provide:
Asset monitoring
Process dashboards
Overall equipment effectiveness (OEE)
Predictive maintenance
Forecasting
Such an IIoT analytics solution relies on real-time and historical data from industrial devices and control systems
located in discrete and process manufacturing facilities. These include PLCs (Programmable Logic Controller),
industrial equipment, SCADA (Supervisory Control and Data Acquisition) systems, MES (Manufacturing Execution
System), and Process Historians. The architecture covered by this series, includes guidance for connecting to all
these systems.
A modern IIoT analytics solution goes beyond moving existing industrial processes and tools to the cloud. It
involves transforming your operations and processes, embracing PaaS services, and leveraging the power of
machine learning and the intelligent edge to optimize industrial processes.
The following list shows some typical personas who would use the solution and how they would use this solution:
Plant Manager - responsible for the entire operations, production, and administrative tasks of the
manufacturing plant.
Production Manager - responsible for production of a certain number of components.
Process Engineer - responsible for designing, implementing, controlling, and optimizing industrial processes.
Operations Manager - responsible for overall efficiency of operation in terms of cost reduction, process time,
process improvement, and so on.
Data Scientist – responsible for building and training predictive Machine Learning models using historical
industrial telemetry.
The following architecture diagram shows the core subsystems that form an IIoT analytics solution.
 NOTE
This architecture represents an ingestion-only pattern. No control commands are sent back to the industrial systems or
devices.

The architecture consists of a number of subsystems and services, and makes use of the Azure Industrial IoT
components. Your own solution may not use all these services or may have additional services. This architecture
also lists alternative service options, where applicable.

IMPORTANT
This architecture includes some services marked as "Preview" or "Public Preview". Preview services are governed by
Supplemental Terms of Use for Microsoft Azure Previews.

Industrial Systems and Devices

In process manufacturing, industrial equipment (for example, flow monitors, pumps, and so on) is often
geographically dispersed and must be monitored remotely. Remote Terminal Units (RTUs) connect remote
equipment to a central SCADA system. RTUs work well in conditions where connectivity is intermittent, and no
reliable continuous power supply exists.
In discrete manufacturing, industrial equipment (for example, factory robots, conveyor systems, and so on) are
connected and controlled by a PLC. One or more PLCs may be connected to a central SCADA system using
protocols such as Modbus or other industrial protocols.
In some cases, the data from SCADA systems is forwarded and centralized in an MES or a historian software (also
known as process historian or operational historian). These historians are often located in IT controlled networks
and have some access to the internet.
Frequently, industrial equipment and SCADA systems are in a closed Process Control Network (PCN), behind one
or more firewalls, with no direct access to the internet. Historians often contain industrial data from multiple
facilities and are located outside of a PCN. So, connecting to a historian is often the path of least resistance, rather
than connecting to a SCADA, MES, or PLC. If a historian is not available, then connecting to an MES or SCADA
system would be the next logical choice.
The connection to the historian, MES, or SCADA system will depend on what protocols are available on that system.
Many systems now include Industry 4.0 standards such as OPC UA. Older systems may only support legacy
protocols such as Modbus, ODBC, or SOAP. If so, you will most likely require a protocol translation software
running on an intelligent edge device.

Intelligent Edge
Intelligent edge devices perform some data processing on the device itself or on a field gateway. In most industrial
scenarios, the industrial equipment cannot have additional software installed on it. This is why a field gateway is
required to connect the industrial equipment to the cloud.
Azure IoT Edge
To connect industrial equipment and systems to the cloud, we recommend using Azure IoT Edge as the field
gateway for:
Protocol and identity translation;
Edge processing and analytics; and
Adhering to network security policies (ISA 95, ISA 99).
Azure IoT Edge is a free, open source field gateway software that runs on a variety of supported hardware devices
or a virtual machine.
IoT Edge allows you to run edge workloads as Docker container modules. The modules can be developed in several
languages, with SDKs provided for Python, Node.js, C#, Java and C. Prebuilt Azure IoT Edge modules from
Microsoft and third-party partners are available from the Azure IoT Edge Marketplace.
Real-time industrial data is encrypted and streamed through Azure IoT Edge to Azure IoT Hub using AMQP 1.0 or
MQTT 3.1.1 protocols. IoT Edge can operate in offline or intermittent network conditions providing store and
for ward capabilities.

There are two system modules provided as part of IoT Edge runtime.
The EdgeAgent module is responsible for pulling down the container orchestration specification (manifest)
from the cloud, so that it knows which modules to run. Module configuration is provided as part of the module
twin.
The EdgeHub module manages the communication from the device to Azure IoT Hub, as well as the inter-
module communication. Messages are routed from one module to the next using JSON configuration.
Azure IoT Edge automatic deployments can be used to specify a standing configuration for new or existing devices.
This provides a single location for deployment configuration across thousands of Azure IoT Edge devices.
A number of third-party IoT Edge gateway devices are available from the Azure Certified for IoT Device Catalog.
 IMPORTANT
Proper hardware sizing of an IoT Edge gateway is important to ensure edge module performance. See the performance
considerations for this architecture.

Gateway Patterns
There are three patterns for connecting your devices to Azure via an IoT Edge field gateway (or virtual machine):
1. Transparent - Devices already have the capability to send messages to IoT Hub using AMQP or MQTT.
Instead of sending the messages directly to the hub, they instead send the messages to IoT Edge, which in
turn passes them on to IoT Hub. Each device has an identity and device twin in Azure IoT Hub.
2. Protocol Translation - Also known as an opaque gateway pattern. This pattern is often used to connect
older brownfield equipment (for example, Modbus) to Azure. Modules are deployed to Azure IoT Edge to
perform the protocol conversion. Devices must provide a unique identifier to the gateway.
3. Identity Translation - In this pattern, devices cannot communicate directly to IoT Hub (for example, OPC
UA Pub/Sub, BLE devices). The gateway is smart enough to understand the protocol used by the
downstream devices, provide them identity, and translate IoT Hub primitives. Each device has an identity and
device twin in Azure IoT Hub.
Although you can use any of these patterns in your IIoT Analytics Solution, your choice will be driven by which
protocol is installed on your industrial systems. For example, if your SCADA system supports ethernet/IP, you will
need to use a protocol translation software to convert ethernet/IP to MQTT or AMQP. See the Connecting to
Historians section for additional guidance.
IoT Edge gateways can be provisioned at scale using the Azure IoT Hub Device Provisioning Service (DPS). DPS is a
helper service for IoT Hub that enables zero-touch, just-in-time provisioning to the right IoT hub without requiring
human intervention, enabling customers to provision millions of devices in a secure and scalable manner.

OPC UA
OPC UA is the successor to OPC Classic (OPC DA, AE, HDA). The OPC UA standard is maintained by the OPC
Foundation. Microsoft has been a member of the OPC Foundation since 1996 and has supported OPC UA on Azure
since 2016.
Industry and domain-specific Information Models can be created based on the OPC UA Data Model. The
specifications of such Information Models (also called industry standard models since they typically address a
dedicated industry problem) are called Companion Specifications. The synergy of the OPC UA infrastructure to
exchange such industry information models enables interoperability at the semantic level. OPC UA can use a
number of transport protocols including MQTT, AMQP, and UADP.
Microsoft has developed open source Azure Industrial IoT components, based on OPC UA, which implement
identity translation pattern:
OPC Twin consists of microservices and an Azure IoT Edge module to connect the cloud and the factory
network. OPC Twin provides discovery, registration, and synchronous remote control of industrial devices
through REST APIs.
OPC Publisher is an Azure IoT Edge module that connects to existing OPC UA servers and publishes telemetry
data from OPC UA servers in OPC UA PubSub format, in both JSON and binary.
OPC Vault is a microservice that can configure, register, and manage certificate lifecycle for OPC UA server and
client applications in the cloud.
Discovery Services is an Azure IoT Edge module that supports network scanning and OPC UA discovery.
The Microsoft Azure IIoT solution also contains a number of services, REST APIs, deployment scripts, and
configuration tools that you can integrate into your IIoT analytics solution. These are open source and available on
GitHub.

Edge Workloads
The ability to run custom or third-party modules at the edge is important.
If you want to respond to emergencies as quickly as possible, you can run anomaly detection or Machine
Learning module in tight control loops at the edge.
If you want to reduce bandwidth costs and avoid transferring terabytes of raw data, you can clean and
aggregate the data locally then send only the insights to the cloud for analysis.
If you want to convert legacy industrial protocols, you can develop a custom module or purchase a third-party
module for protocol translation.
If you want to quickly respond to an event on the factory floor, you can use an edge module to detect that event
and another module to respond to it.
Microsoft and our partners have made available on the Azure Marketplace a number of edge modules, which can
be used in your IIoT analytics solution. Protocol and identity translation are the most common edge workloads
used within an IIoT analytics solution. In the future, expect to see other workloads such as closed loop control using
edge ML models.

Connecting to Historians
A common pattern when developing an IIoT analytics solution is to connect to a process historian and stream real-
time data from the historian to Azure IoT Hub. How this is done, will depend on which protocols are installed and
accessible (that is, not blocked by firewalls) on the historian.

P ROTO C O L AVA IL A B L E O N H ISTO RIA N O P T IO N S

OPC UA - Use Azure IoT Edge, along with OPC Publisher, OPC
Twin, and OPC Vault, to send OPC UA data over MQTT to
IoT Hub. OPC Twin also has support for OPC UA HDA
profile, useful for obtaining historical data.
- Use a third-party Azure IoT Edge OPC UA module to
send OPC UA data over MQTT to IoT Hub.

OPC DA - Use third-party software to convert OPC DA to OPC UA

and send OPC UA data to IoT Hub over MQTT. Or use
OPC Publisher, OPC Twin and OPC Vault to send OPC UA
data over MQTT to IoT Hub.

Web Service - Use a custom Azure IoT Edge HTTP module to poll the
web service.
- Use third-party software that supports HTTP to MQTT
3.1.1 or AMQP 1.0.

MQTT 3.1.1 (Can publish MQTT messages) - Connect historian directly to Azure IoT Hub using MQTT.
- Connect historian to Azure IoT Edge as a leaf device. See
Transparent Gateway pattern.
 P ROTO C O L AVA IL A B L E O N H ISTO RIA N O P T IO N S

Other - Use a custom Azure IoT Edge module.

- Use third-party software to convert to MQTT 3.1.1 or
AMQP 1.0.

A number of Microsoft partners have developed protocol and identity translation modules or solutions that are
available on the Azure Marketplace.
Some historian vendors also provide first-class capabilities to send data to Azure.

H ISTO RIA N O P T IO N S

OSIsoft PI PI Integrator for Azure

Honeywell Uniformance Cloud Historian

Once real time data streaming has been established between your historian and Azure IoT Hub, it is important to
export your historian's historical data and import it into your IIoT analytics solution. For guidance on how to
accomplish this, see Historical Data Ingestion.

Cloud Gateway
A cloud gateway provides a cloud hub for devices and field gateways to connect securely to the cloud and send
data. It also provides device management capabilities. For the cloud gateway, we recommend Azure IoT Hub. IoT
Hub is a hosted cloud service that ingests events from devices and IoT Edge gateways. IoT Hub provides secure
connectivity, event ingestion, bidirectional communication, and device management. When IoT Hub is combined
with the Azure Industrial IoT components, you can control your industrial devices using cloud-based REST APIs.
IoT Hub supports the following protocols:
MQTT 3.1.1,
MQTT over WebSockets,
AMQP 1.0,
AMQP over WebSockets, and
HTTPS.
If the industrial device or system supports any of these protocols, it can send data directly to IoT Hub. In most
industrial environments, this is not permissible because of PCN firewalls and network security policies (ISA 95, ISA
99). In such cases, an Azure IoT Edge field gateway can be installed in a DMZ between the PCN and the Internet.

Next steps
To learn the services recommended for this architecture, continue reading the series with Services in an IIoT
analytics solution.
 Services in an IIoT analytics solution
12/18/2020 • 19 minutes to read • Edit Online

Building on the architectural components in the recommended Azure Industrial IoT analytics solution, this article
discusses the subsystems and Azure services that can be used in such a solution. Your solution may not use all
these services or may have additional services.

Time Series Service

A time series service will be required to provide a warm and cold store for the JSON-based industrial time series
data. We recommend using Time Series Insights (or TSI) as the time series service for the following reasons:
Ability to ingest streaming data directly from Azure IoT Hub or Azure Event Hub.
Multi-layered storage solution with warm and cold analytics support. It provides you the option to route
data between warm and cold, for interactive analytics over warm data as well as operational intelligence
over decades of historical data.
A highly interactive warm analytics solution to perform frequent, and large number of queries over
shorter time span data.
A scalable, performant, and cost-optimized time series data lake based on Azure Storage allowing
customers to trend years' worth of time series data.
Asset hierarchy support that describes the domain and metadata associated with the derived and raw
signals from industrial assets and devices. Multiple hierarchies can be defined reflecting different
departments within your company. For instance, a factory hierarchy will look different from an Oil & Gas
hierarchy.
Flexible analytics platform to store historical time series data in your own Azure Storage account, thereby
allowing you to have ownership of your industrial IoT data. Data is stored in open source Apache Parquet
format that enables connectivity and interoperability across a variety of data scenarios including predictive
analytics, machine learning, and other custom computations done using familiar technologies including
Spark, Databricks, and Jupyter.
Rich analytics with enhanced query APIs and user experience that combines asset-based data insights with
rich, improvised data analytics with support for interpolation, scalar and aggregate functions, categorical
variables, scatter plots, and time-shifting time series signals for in-depth analysis.
Enterprise grade platform to support the scale, performance, security, and reliability needs of our industrial
IoT customers.
Ability to view and export the data in CSV format for further analysis.
Ability to share analysis across your organization. Current visualizations can be saved and shared.
Alternatively, the current state of each user's analysis is assigned a unique identifier. This identifier is placed
in the URL allowing users to easily share their analysis by placing the URL in emails, documents, etc. Because
TSI pricing is not seat-based, this democratizes the data by allowing anyone with access to the data to see it.
You can continue to use Time Series Insights with the old pricing model (S1, S2) for warm, unplanned data analysis.
However, we highly encourage you to use the updated offering (PAYG) as it offers many new capabilities as well as
Pay-As-You-Go pricing, cost savings, and flexibility to scale. Alternatively, Azure Data Explorer or Cosmos DB may
be used as a time series database if you plan to develop a custom time series service.
 NOTE
When connecting Time Series Insights with IoT Hub or Event Hub, ensure you select an appropriate Time Series ID. We
recommend using a SCADA tag name field or OPC UA node id (for example, nsu=http://msft/boiler;i=#####) if possible, as
these will map to leaf nodes in your Time Series Model.

The data in Time Series Insights is stored in your Azure Blob Storage account (bring your own storage account) in
Parquet file format. It is your data after all!
You can query your data in Time Series Insights using:
Time Series Insights Explorer
Query API
REST API
Power BI
Any of your favorite BI and analytics tools (for example, Spark, Databricks, Azure Notebooks) by accessing the
Parquet files in your Azure blob storage account

Microservices
Your IIoT analytics solution will require a number of microservices to perform functions such as:
Providing HTTP REST APIs to support your web application.
We recommend creating HTTP-triggered Azure Functions to implement your APIs.
Alternatively, you can develop and host your REST APIs using Azure Service Fabric or Azure Kubernetes
Service (AKS).
Providing an HTTP REST API interface to your factory floor OPC UA servers (for example, using Azure
Industrial IoT components consisting of OPC Publisher, OPC Twin and OPC Vault) to provide discovery,
registration, and remote control of industrial devices.
For hosting the Azure Industrial IoT microservices, we recommend using Azure Kubernetes Service (AKS).
See Deploying Azure Industrial IoT Platform to understand the various deployment options.
Performing data transformation such as converting binary payloads to JSON or differing JSON payloads to
a common, canonical format.
We recommend creating Azure Functions connected to IoT Hub to perform payload transformations.
Different industrial equipment vendors will send telemetry in different payload formats (JSON,
binary, and so on) and schemas. When possible, we recommend converting the different equipment
schemas to a common, canonical schema, ideally based on an industry standard.
If the message body is binary, use an Azure Function to convert the incoming messages to JSON and
send the converted messages back to IoT Hub or to Event Hub.
When the message body is binary, IoT Hub message routing cannot be used against the message
body, but can be used against message properties.
The Azure Industrial IoT components include the capability to decode OPC UA binary messages to
JSON.
A Data Ingest Administration service for updating the list of tags monitored by your IIoT analytics solution.
A Historical Data Ingestion service for importing historical data from your SCADA, MES, or historian into
your solution.
Your solution will likely involve additional microservices to satisfy the specific requirements of your IIoT analytics
solution. If your organization is new to building microservices, we recommend implementing custom
microservices using Azure Functions. Azure Functions is an event-driven serverless compute platform that can be
used to develop microservices and solve complex orchestration problems. It allows you to build and debug locally
(in several software languages) without additional setup, deploy and operate at scale in the cloud, and integrate
Azure services using triggers and bindings.
Both stateless and stateful microservices can be developed using Azure Functions. Azure Functions can use
Cosmos DB, Table Storage, Azure SQL, and other databases to store stateful information.
Alternatively, if your organization has a previous experience building container-based microservices, we
recommend you to also consider Azure Service Fabric or Azure Kubernetes Service (AKS). Refer to Microservices in
Azure for more information.
Regardless of your microservices platform choice, we recommend using Azure API Management to create
consistent and modern API gateways for your microservices. API Management helps abstract, publish, secure, and
version your APIs.
Data Ingest Administration
We recommend developing a Data Ingest Administration service to add/update the list of tags monitored by your
IIoT analytics solution.
SCADA tags are variables mapped to I/O addresses on a PLC or RTU. Tag names vary from organization to
organization but often follow a naming pattern. As an example, tag names for a pump with tag number 14P103
located in STN001 (Station 001), has these statuses:
STN001_14P103_RUN
STN001_14P103_STOP
STN001_14P103_TRIP
As new tags are created in your SCADA system, the IIoT analytics solution must become aware of these tags and
subscribe to them in order to begin collecting data from them. In some cases, the IIoT analytics solution may not
subscribe to certain tags as the data they contain may be irrelevant.
If your SCADA system supports OPC UA, new tags should appear as new NodeIDs in the OPC UA hierarchy. For
example, the above tag names may appear as:
ns=2;s= STN001_14P103_RUN
ns=2;s= STN001_14P103_STOP
ns=2;s= STN001_14P103_TRIP
We recommend developing a workflow that informs the administrators of the IIoT analytics solution when new
tags are created, or existing tags are edited in the SCADA system. At the end of the workflow, the OPC Publisher is
updated with the new/updated tags.
To accomplish this, we recommend developing a workflow that involves Power Apps, Logic Apps, and Azure
Functions, as follows:
The SCADA system operator can trigger the Logic Apps workflow using a Power Apps form whenever tags are
created or edited in the SCADA system.
Alternatively, Logic Apps connectors can monitor a table in the SCADA system database for tag changes.
The OPC UA Discovery service can be used to both find OPC UA servers and the tags and methods they
implement.
The Logic Apps workflow includes an approval step where the IIoT analytics solution owners can approve the
new/updated tags.
Once the new/updated tags are approved and a frequency assigned, the Logic App calls an Azure Function.
The Azure function calls the OPC Twin microservice, which directs the OPC Publisher module to subscribe to the
 new tags.
A sample can be found here.
If your solution involves third-party software, instead of OPC Publisher, configure the Azure Function to
call an API running on the third-party software either directly or using an IoT Hub Direct Method.
Alternatively, Microsoft Forms and Microsoft Flow can be used in place of Power Apps and Logic Apps.
Historical Data Ingestion
Years of historical data likely exists in your current SCADA, MES, or historian system. In most cases, you will want to
import your historical data into your IIoT analytics solution.
Loading historical data into your IIoT analytics solution consists of three steps:
1. Export your historical data.
a. Most SCADA, MES, or historian systems have some mechanism that allows you to export your historical
data, often as CSV files. Consult your system's documentation on how best to do this.
b. If there is no export option in your system, consult the system's documentation to determine if an API
exists. Some systems support HTTP REST APIs or OPC Historical Data Access (HDA). If so, build an
application or use a Microsoft partner solution that connects to the API, queries for the historical data,
and saves it to a file in formats such as CSV, Parquet, TSV, and so on.
2. Upload to Azure.
a. If the aggregate size of the exported data is small, you can upload the files to Azure Blob Storage over the
internet using Azcopy.
b. If the aggregate size of the exported data is large (tens or hundreds of TBs), consider using Azure
Import/Export Service or Azure Data Box to ship the files to the Azure region where your IIoT analytics
solution is deployed. Once received, the files will be imported into your Azure Storage account.
3. Import your data.
a. This step involves reading the files in your Azure Storage account, serializing the data as JSON, and
sending data as streaming events into Time Series Insights. We recommend using an Azure Function to
perform this.
b. Time Series Insights only supports IoT Hub and Event Hub as data sources. We recommend using an
Azure Function to send the events to a temporary Event Hub, which is connected to Time Series Insights.
c. Refer to How to shape JSON events and Supported JSON shapes for best practices on shaping your
JSON payload.
d. Make sure to use the same Time Series ID as you do for your streaming data.
e. Once this process is completed, the Event Hub and Azure Function may be deleted. This is an optional
step.

NOTE
Exporting large volumes of data from your industrial system (for example, SCADA or historian) may place a significant
performance load on that system, which can negatively impact operations. Consider exporting smaller batches of historical
data to minimize performance impacts.

Rules and Calculation Engine

Your IIoT analytics solution may need to perform near real-time (low latency) calculations and complex event
processing (or CEP) over streaming data, before it lands in a database. For example, calculating moving averages or
calculated tags. This is often referred to as a calculations engine. Your solution may also need to trigger actions (for
example, display an alert) based on the streaming data. This is referred to as a rules engine.
We recommend using Time Series Insights for simple calculations, at query time. The Time Series Model
introduced with Time Series Insights supports a number of formulas including: Avg, Min, Max, Sum, Count, First,
and Last. The formulas can be created and applied using the Time Series Insights APIs or Time Series Insights
Explorer user interface.
For example, a Production Manager may want to calculate the average number of widgets produced on a
manufacturing line, over a time interval, to ensure productivity goals are met. In this example, we would
recommend the Production Manager to use the Time Series Insights explorer interface to create and visualize the
calculation. Or if you have developed a custom web application, it can use the Time Series Insights APIs to create
the calculation, and the Azure Time Series Insights JavaScript SDK (or tsiclient) to display the data in your custom
web application.
For more advanced calculations and/or to implement a rules engine, we recommend using Azure Stream Analytics.
Azure Stream Analytics is a real-time analytics and complex event-processing engine, that is designed to analyze
and process high volumes of fast streaming data from multiple sources simultaneously. Patterns and relationships
can be identified in information extracted from a number of input sources including devices, sensors, click streams,
social media feeds, and applications. These patterns can be used to trigger actions and initiate workflows such
creating alerts, feeding information to a reporting tool, or storing transformed data for later use.
For example, a Process Engineer may want to implement a more complex calculation such as calculating the
standard deviation (SDEV) of the widgets produced across a number of production lines to determine when any
line is more than 2x beyond the mean over a period of time. In this example, we recommend using Stream
Analytics, with a custom web application. The Process Engineer authors the calculations using the custom web
application, which calls the Stream Analytics REST APIs to create and run these calculations (also known as Jobs).
The Job output can be sent to an Event Hub, connected to Time Series Insights, so the result can be visualized in
Time Series Insights explorer.
Similarly, for a Rules Engine, a custom web application can be developed that allows users to author alerts and
actions. The web application creates associated Jobs in Azure Stream Analytics using the Steam Analytics REST API.
To trigger actions, a Stream Analytics Job calls an Azure Function output. The Azure Function can call a Logic App or
Power Automate task that sends an Email alert or invokes Azure SignalR to display a message in the web
application.
Azure Stream Analytics supports processing events in CSV, JSON, and Avro data formats while Time Series Insights
supports JSON. If your payload does not meet these requirements, consider using an Azure Function to perform
data transformation prior to sending the data to Stream Analytics or Time Series Insights (using IoT Hub or Event
Hubs).
Azure Stream Analytics also supports reference data, a finite data set that is static or slowly changing in nature,
used to perform a lookup or to augment your data streams. A common scenario is exporting asset metadata from
your Enterprise Asset Management system and joining it with real-time data coming from those industrial devices.
Stream Analytics is also available as a module on the Azure IoT Edge runtime. This is useful for situations where
complex event processing needs to happen at the Edge. As an alternative to Azure Stream Analytics, near real-time
Calculation and Rules Engines may be implemented using Apache Spark Streaming on Azure Databricks.

Notifications
Since the IIoT analytics solution is not a control system, it does not require a complete Alarm Management system.
However, there will be cases where you will want the ability to detect conditions in the streaming data and generate
notifications or trigger workflows. Examples include:
temperature of a heat exchanger exceeding a configured limit, which changes the color of an icon in your web
application,
an error code sent from a pump, which triggers a work order in your ERP system, or
 the vibration of a motor exceeding limits, which triggers an email notification to an Operations Manager.
We recommend using Azure Stream Analytics to define and detect conditions in the streaming data (refer to the
rules engine mentioned earlier). For example, a Plant Manager implements an automated workflow that runs
whenever an error code is received from any equipment. In this example, your custom web application can use the
Stream Analytics REST API to provide a user interface for the Plant Manager to create and run a job that monitors
for specific error codes.
For defining an alert (email or SMS) or triggering a workflow, we recommend using Azure Logic Apps. Logic Apps
can be used to build automated, scalable workflows, business processes, and enterprise orchestrations to integrate
your equipment and data across cloud services and on-premises systems.
We recommend connecting Azure Stream Analytics with Azure Logic Apps using Azure Service Bus. In the previous
example, when an error code is detected by Stream Analytics, the job will send the error code to an Azure Service
Bus queue output. A Logic App will be triggered to run whenever a message is received on the queue. This Logic
App will then perform the workflow defined by the Plant Manager, which may involve creating a work order in
Dynamics 365 or SAP, or sending an email to maintenance technician. Your web application can use the Logic Apps
REST API to provide a user interface for the Plant Manager to author workflows or these can be built using the
Azure portal authoring experience.
To display visual alerts in your web application, we recommend creating an Azure Stream Analytics job to detect
specific events and send those to either:
An Event Hub output - Then connect the Event Hub to Time Series Insights. Use the Azure Time Series
Insights JavaScript SDK (tsiclient) to display the event in your web application.
or,
An Azure Functions output - Then develop an Azure Function that sends the events to your web application
using SignalR.
Operational alarms and events triggered on premise can also be ingested into Azure for reporting and to trigger
work orders, SMS messages, and emails.

Microsoft 365
The IIoT analytics solution can also include Microsoft 365 services to automate tasks and send notifications. The
following are a few examples:
Receive email alerts in Microsoft Outlook or post a message to a Microsoft Teams channel when a condition is
met in Azure Stream Analytics.
Receive notifications as part of an approval workflow triggered by a Power App or Microsoft Forms submission.
Create an item in a SharePoint list when an alert is triggered by a Logic App.
Notify a user or execute a workflow when a new tag is created in a SCADA system.

Machine Learning
Machine learning models can be trained using your historical industrial data, enabling you to add predictive
capabilities to your IIoT application. For example, your Data Scientists may be interested in using the IIoT analytics
solution to build and train models that can predict events on the factory floor or indicate when maintenance should
be conducted on an asset.
For building and training machine learning models, we recommend Azure Machine Learning. Azure Machine
Learning can connect to Time Series Insights data stored in your Azure Storage account. Using the data, you can
create and train forecasting models in Azure Machine Learning. Once a model has been trained, it can be deployed
as a web service on Azure (hosted on Azure Kubernetes Services or Azure Functions, for example) or to an Azure
IoT Edge field gateway.
For those new to machine learning or organizations without Data Scientists, we recommend starting with Azure
Cognitive Services. Azure Cognitive Services are APIs, SDKs, and services available to help you build intelligent
applications without having formal AI or data science skills or knowledge. Azure Cognitive Services enable you to
easily add cognitive features into your IIoT analytics solution. The goal of Azure Cognitive Services is to help you
create applications that can see, hear, speak, understand, and even begin to reason. The catalog of services within
Azure Cognitive Services can be categorized into five main pillars - Vision, Speech, Language, Web Search, and
Decision.

Asset Hierarchy
An asset hierarchy allows you to define hierarchies for classifying your asset, for example, Country > Location >
Facility > Room. They may also contain the relationship between your assets. Many organizations maintain asset
hierarchies within their industrial systems or within an Enterprise Asset Management (EAM) system.
The Time Series Model in Azure Time Series Insights provides asset hierarchy capabilities. Through the use of
Instances, Types and Hierarchies, you can store metadata about your industrial devices, as shown in the image
below.

If possible, we recommend exporting your existing asset hierarchy and importing it into Time Series Insights using
the Time Series Model APIs. We recommend periodically refreshing it as updates are made in your Enterprise Asset
Management system.
In the future, asset models will evolve to become digital twins, combining dynamic asset data (real-time telemetry),
static data (3D models, metadata from Asset Management Systems), and graph-based relationships, allowing the
digital twin to change in real-time along with the physical asset.
Azure Digital Twins is an Azure IoT service that provides the ability to:
Create comprehensive models of physical environments,
Create spatial intelligence graphs to model the relationships and interactions between people, places, and
devices,
Query data from a physical space rather than disparate sensors, and
Build reusable, highly scalable, spatially aware experiences that link streaming data across the physical and
digital world.

Business Process Integration

In some instances, you will want your IIoT analytics solution to perform actions based on insights from your
industrial data. This can include raising alarms, sending email, sending SMS messages, or triggering a workflow in
your line-of-business systems (for example, CRM, ERP, and so on). We recommend using Azure Logic Apps to
integrate your IIoT analytics solution with your line-of-business systems. Azure Logic Apps has a number of
connectors to business systems and Microsoft services such as:
Dynamics 365
SharePoint Online
Office 365 Outlook
Salesforce
SAP
For example, an error code from a pump is detected by an Azure Stream Analytics job. The job sends a message to
Azure Service Bus and triggers a Logic App to run. The Logic App sends an email notification to the Plant Manager
using the Office 365 Outlook connector. It then sends a message to your SAP S/4 HANA system using the SAP
connector, which creates a Service Order in SAP.

User Management
User management involves managing user profiles and controlling what actions a user can perform in your IIoT
analytics solution. For example, what asset data can a user view, or whether the user can create conditions and
alerts. This is frequently referred to as role-based access control (RBAC).
We recommend implementing role-based access control using the Microsoft identity platform along with Azure
Active Directory. In addition, the Azure PaaS services mentioned in this IIoT analytics solution can integrate directly
with Azure Active Directory, thereby ensuring security across your solution.
Your web application and custom microservices can also integrate with the Microsoft identity platform using
libraries such as Microsoft Authentication Library (or MSAL) and protocols such as OAuth 2.0 and OpenID Connect.
User management also involves operations such as:
creating a new user,
updating a user's profile, such as their location and phone number,
changing a user's password, and
disabling a user's account.
For these operations, we recommend using the Microsoft Graph.
Next steps
Data visualization is the backbone of a well-defined analytics system. Learn about the data visualization techniques
that you can use with the IIoT analytics solution recommended in this series.
 Data analysis in Azure Industrial IoT analytics solution
12/18/2020 • 4 minutes to read • Edit Online

This article shows you how to visualize the data collected by the Azure Industrial IoT analytics solution. You can
easily look for data trends using visual elements and dashboards, and use these trends to analyze the effectiveness
of your solution.

Visualization
There are many options for visualizing your industrial data. Your IIoT analytics solution may use some or all of
these options, depending on the personas using your solution.
For Process Engineers and other personas looking to perform ad-hoc analytics and trend visualizations, we
recommend using Azure Time Series Insights explorer.
For Plant Managers and other personas wanting to develop dashboards, we recommend using Power BI, and
connecting Power BI with your data in Time Series Insights using the Power BI connector. Using Power BI, these
users can also combine external data from your ERP, EAM, or other systems with the data in Time Series
Insights.
For advanced visualizations, such as schematic views and process graphics, we recommend a custom web
application.
For Data Scientists interested in using open source data analysis and visualization tools such as Python, Jupyter
Notebooks, and Matplotlib, we recommend Azure Notebooks.

Data Trends
The Azure Time Series Insights explorer is a web application that provides powerful data trending and visualization
capabilities that make it simple to explore and analyze billions of IIoT events simultaneously.
Time Series Insights Explorer is ideally suited to personas, such as a Process Engineer or Operations Manager, who
want to explore, analyze and visualize the raw data coming from your industrial systems. The insights gained from
exploring the raw data can help build Azure Stream Analytics jobs, which look for conditions in the data or perform
calculations over the data.
The Azure Time Series Insights explorer allows you to seamlessly explore both warm and cold data, or your
historical data, as demonstrated in the following figure.
Azure Time Series Insights explorer has a powerful yet intuitive user interface, as shown below.

Key features of Azure Time Series Insights explorer:

1. Environment panel - Displays all your Azure Time Series Insights environments.
2. Navigation bar - Lets you switch between the Analyze and Model pages.
3. Hierarchy tree and search panel - Lets you select and search for specific data elements to be charted.
4. Time series well - Shows all your currently selected data elements.
5. Char t panel - Displays your current working chart.
6. Timeline - Lets you modify your working time span.
7. App bar - Contains your user management options (such as current tenant), and allows you to change them
and language settings.

Dashboards
For some personas, such as a Plant Manager, dashboards containing factory or plant KPIs and visualizations are
more important than viewing the raw data. For such users, we recommend Power BI as the visualization solution.
You can connect Power BI with your data stored in Time Series Insights, providing you with powerful reporting and
dashboard capabilities over your industrial data, and allowing you to share insights and results across your
organization.

By connecting your data to Power BI, you can:

Perform correlations with other data sources supported by Power BI and access a host of different data
visualization options.
Create Power BI dashboards and reports using your Time Series Insight data and share them with your
organization.
Unlock data interoperability scenarios in a simple, easy-to-use manner, and get to insights faster than ever.
Modify Time Series Insights data within Power BI using the powerful Advanced Editor.

Schematic Views
For advanced visualizations, such as schematic views or process graphics, you may require a custom web
application. A custom web application also allows you to provide a single pane of glass user experience and other
advanced capabilities including:
a simplified and integrated authoring experience for Azure Stream Analytics jobs and Logic Apps,
displaying real-time data using process or custom visuals,
displaying KPIs and external data with embedded Power BI dashboards,
displaying visual alerts using SignalR, and
allowing administrators to add/remove users from the solution.
We recommend building a Single Page Application (SPA) using:
JavaScript, HTML5, and CSS3
Time Series Insights JavaScript SDK for displaying process or custom visuals with data from Time Series
Insights
MSAL.js to sign in users and acquire tokens to use with the Microsoft Graph
Azure App Services Web Apps to host the web application
Power BI to embed Power BI dashboards directly in the web app
Azure Maps to render map visualizations
Microsoft Graph SDK for JavaScript to integrate with Microsoft 365

Notebooks
One of the advantages of moving operational data to the cloud is to take advantage of modern big-data tool sets.
One of the most common tools used by Data Scientists for ad-hoc analysis of big data are Jupyter Notebooks.
Jupyter (formerly IPython) is an open-source project that lets you easily combine Markdown text, executable code,
persistent data, graphics, and visualizations onto a single, sharable canvas - the notebook. Production Engineers
should also consider learning Jupyter Notebooks technology to assist in analysis of plant events, finding
correlations, and so on. Jupyter Notebooks provide support for Python 2/3, R, and F# programming languages and
can connect to your Time Series Insights data stored Azure storage.

Next steps
Now that you have learned the architecture of an Azure IIoT analytics solution, read the architectural considerations
that improve the resiliency and efficiency of this architecture.
 Architectural Considerations in an IIoT Analytics
Solution
12/18/2020 • 3 minutes to read • Edit Online

The Microsoft Azure Well-Architected Framework describes some key tenets of a good architectural design.
Keeping in line with these tenets, this article describes the considerations in the reference Azure Industrial IoT
analytics solution that improve its performance and resiliency.

Performance Considerations
Azure PaaS Services
All Azure PaaS services have an ability to scale up and/or out. Some services will do this automatically (for
example, IoT Hub, Azure Functions in a Consumption Plan) while others can be scaled manually.
As you test your IIoT analytics solution, we recommend that you:
understand how each service scales (that is, the units of scale),
collect performance metrics and establish baselines, and
setup alerts when performance metrics exceed baselines.
All Azure PaaS services have a metrics blade that allows you to view service metrics, and configure conditions and
alerts, which are collected and displayed in Azure Monitor. We recommend enabling these features to ensure your
solution performs as expected.
IoT Edge
Azure IoT Edge gateway performance is impacted by:
the number of edge modules running and their performance requirements,
the number of messages processed by modules and EdgeHub,
Edge modules requiring GPU processing,
offline buffering of messages,
the gateway hardware, and
the gateway operating system.
We recommend real world testing and/or testing with simulated telemetry to understand the field gateway
hardware requirements for Azure IoT Edge. Conduct your initial testing using virtual machine where CPU, RAM,
disk can be easily adjusted. Once approximate hardware requirements are known, get your field gateway hardware
and conduct your testing again using actual hardware.
You should also test to ensure:
no messages are being lost between source (for example, historian) and destination (for example, Time Series
Insights),
acceptable message latency exists between the source and the destination,
that source timestamps are preserved, and
data accuracy is maintained, especially when performing data transformations.

Availability Considerations
IoT Edge
A single Azure IoT Edge field gateway can be a single point of failure between your SCADA, MES, or historian and
Azure IoT Hub. A failure can cause gaps in data in your IIoT analytics solution. To prevent this, IoT Edge can
integrate with your on-premise Kubernetes environment, using it as a resilient, highly available infrastructure layer.
For more information, see How to install IoT Edge on Kubernetes (Preview).

Network Considerations
IoT Edge and Firewalls
To maintain compliance with standards such as ISA 95 and ISA 99, industrial equipment is often installed in a
closed Process Control Network (PCN), behind firewalls, with no direct access to the Internet (see Purdue
networking model).
There are three options to connect to equipment installed in a PCN:
1. Connect to a higher-level system, such as a historian, located outside of the PCN.
2. Deploy an Azure IoT Edge device or virtual machine in a DMZ between the PCN and the internet.
a. The firewall between the DMZ and the PCN will need to allow inbound connections from the DMZ to the
appropriate system or device in the PCN.
b. There may be no internal DNS setup to resolve PCN names to IP addresses.
3. Deploy an Azure IoT Edge device or virtual machine in the PCN and configure IoT Edge to communicate with
the Internet through a Proxy server.
a. Additional IoT Edge setup and configuration are required. See Configure an IoT Edge device to
communicate through a proxy server.
b. The Proxy server may introduce a single point of failure and/or a performance bottleneck.
c. There may be no DNS setup in the PCN to resolve external names to IP addresses.
Azure IoT Edge will also require:
access to container registries, such as Docker Hub or Azure Container Registry, to download modules over
HTTPS,
access to DNS to resolve external FQDNs, and
ability to communicate with Azure IoT Hub using MQTT, MQTT over WebSockets, AMQP, or AMQP over
WebSockets.
For additional security, industrial firewalls can be configured to only allow traffic between IoT Edge and IoT Hub
using Service Tags. IP address prefixes of IoT Hub public endpoints are published periodically under the
AzureIoTHub service tag. Firewall administrators can programmatically retrieve the current list of service tags,
together with IP address range detail, and update their firewall configuration.

Next Steps
For a more detailed discussion of the recommended architecture and implementation choices, download
and read the Microsoft Azure IoT Reference Architecture pdf.
Azure Industrial IoT components, tutorials, and source code.
For detailed documentation of the various Azure IoT services, see Azure IoT Fundamentals.
 Add IP address spaces to peered virtual networks
12/18/2020 • 3 minutes to read • Edit Online

Many organizations deploy a virtual networking architecture that follows the Hub and Spoke model. At some point,
the hub virtual network might require additional IP address spaces. However, address ranges can't be added or
deleted from a virtual network's address space once it's peered with another virtual network. To add or remove
address ranges, delete the peering, add or remove the address ranges, then re-create the peering manually. The
scripts described in this article can make that process easier.

Single subscription
A single subscription use case, both hub and all spoke virtual networks are in the same subscription.

Multiple subscriptions
Another use case can be where the hub virtual network is in one subscription and all other spoke virtual networks
are in different subscriptions. The subscriptions are for a single Azure Active Directory tenant.
Considerations
Running the script will result in outage or disconnections between the Hub and Spoke virtual networks. Execute
it during an approved maintenance window.
Run Get-Module -ListAvailable Az to find the installed version. The script requires the Azure PowerShell module
version 1.0.0 or later. If you need to upgrade, see Install Azure PowerShell module.
If not already connected, run Connect-AzAccount to create a connection with Azure.
Consider assigning accounts, used for virtual network peering, to the Network Contributor role or a Custom
Role containing the necessary actions found under virtual network peering permissions.
Assign accounts used to add IP address spaces, to the Network Contributor role or a Custom Role containing the
necessary actions found under virtual network permissions.
The IP address space that you want to add to the hub virtual network must not overlap with any of the IP
address spaces of the spoke virtual networks that you intend to peer with the hub virtual network.

Add the IP address range

The script automatically removes all Virtual Network peerings from the Hub Virtual Network, adds an IP address
range prefix to the Hub Virtual Network based on Input parameters, adds the Virtual Network peerings back to the
Hub Virtual Network, and reconnects the Hub virtual network peerings to the existing Spoke virtual network
peerings. The script applies to single and multiple subscription hub and spoke topologies.

param (
# Address Prefix range (CIDR Notation, e.g., 10.0.0.0/24 or 2607:f000:0000:00::/64)
[Parameter(Mandatory = $true)]
[String[]]
$IPAddressRange,

# Hub VNet Subscription Name

[Parameter(Mandatory = $true)]
 [Parameter(Mandatory = $true)]
[String]
$HubVNetSubscriptionName,

# Hub VNet Resource Group Name

[Parameter(Mandatory = $true)]
[String]
$HubVNetRGName,

# Hub VNet Name

[Parameter(Mandatory = $true)]
[String]
$HubVNetName
)

#Set context to Hub VNet Subscription

Get-AzSubscription -SubscriptionName $HubVNetSubscriptionName | Set-AzContext
#end

#Get All Hub VNet Peerings and Hub VNet Object

$hubPeerings = Get-AzVirtualNetworkPeering -ResourceGroupName $HubVNetRGName -VirtualNetworkName $HubVNetName
$hubVNet = Get-AzVirtualNetwork -Name $HubVNetName -ResourceGroupName $HubVNetRGName
#end

#Remove All Hub VNet Peerings

Remove-AzVirtualNetworkPeering -VirtualNetworkName $HubVNetName -ResourceGroupName $HubVNetRGName -name
$hubPeerings.Name -Force
#end

#Add IP address range to the hub vnet

$hubVNet.AddressSpace.AddressPrefixes.Add($IPAddressRange)
#end

#Add $IPAddressRange to subnet

$subnet = $HUBvnet.subnets[0]
$subnet.addressprefix.add($IPAddressRange)
#end

#Apply configuration stored in $hubVnet

Set-AzVirtualNetwork -VirtualNetwork $hubVNet
#end

foreach ($vNetPeering in $hubPeerings)

{
# Get remote vnet name
$vNetFullId = $vNetPeering.RemoteVirtualNetwork.Id
$vNetName = $vNetFullId.Substring($vNetFullId.LastIndexOf('/') + 1)

# Pull remote vNet object

$vNetObj = Get-AzVirtualNetwork -Name $vNetName

# Get the peering from the remote vnet object

$peeringName = $vNetObj.VirtualNetworkPeerings.Where({$_.RemoteVirtualNetwork.Id -like
"*$($hubVNet.Name)"}).Name
$peering = Get-AzVirtualNetworkPeering -ResourceGroupName $vNetObj.ResourceGroupName -VirtualNetworkName
$vNetName -Name $peeringName

# Reset to initiated state

Set-AzVirtualNetworkPeering -VirtualNetworkPeering $peering

# Re-create peering on hub

Add-AzVirtualNetworkPeering -Name $vNetPeering.Name -VirtualNetwork $HubVNet -RemoteVirtualNetworkId
$vNetFullId -AllowGatewayTransit

}
Pricing
There is a nominal charge for ingress and egress traffic that utilizes a virtual network peering. There is no change to
existing pricing when adding an additional IP address space to an Azure virtual network. For more information, see
the pricing page.

Next steps
Learn more about managing Virtual Network peerings.
Learn more about managing IP Address ranges on Virtual Networks.
 Azure security solutions for AWS
12/18/2020 • 11 minutes to read • Edit Online

Microsoft offers several security solutions that can help secure and protect Amazon Web Services (AWS) accounts
and environments.
AWS organizations that use Azure Active Directory (Azure AD) for Microsoft 365 or hybrid cloud identity and
access protection can quickly and easily deploy Azure AD for AWS accounts, often without additional cost.
Other Microsoft security components can integrate with Azure AD to provide additional security for AWS accounts.
Microsoft Cloud App Security (MCAS) backs up Azure AD with session protection and user behavior monitoring.
Azure Sentinel integrates with Azure AD and MCAS to detect and automatically respond to threats against AWS
environments.
These Microsoft security solutions are extensible and have multiple levels of protection. Organizations can
implement one or more of these solutions along with other types of protection for a full security architecture that
protects current and future AWS deployments.
This article provides AWS identity architects, administrators, and security analysts with immediate insights and
detailed guidance for deploying several Microsoft security solutions.

Architecture
This diagram summarizes how AWS installations can benefit from key Microsoft security components:

Azure AD provides centralized single sign-on (SSO) and strong authentication through multi-factor
authentication (MFA) and Conditional Access. Azure AD supports AWS role-based identities and
authorization for access to AWS resources. For more information and detailed instructions, see Azure AD
identity and access management for AWS.
MCAS integrates with Azure AD Conditional Access to enforce additional restrictions, and monitors and
protects sessions after sign-in. MCAS uses user behavior analytics (UBA) and other AWS APIs to monitor
sessions and users and to support information protection.
Azure Sentinel integrates with MCAS and AWS to detect and automatically respond to threats. Azure
Sentinel monitors the AWS environment for misconfiguration, potential malware, and advanced threats to
 AWS identities, devices, applications, and data.
MCAS for visibility and control
Several users or roles making administrative changes can result in configuration drift away from intended security
architecture and standards. Security standards can also change over time. Security personnel must constantly and
consistently detect new risks, evaluate mitigation options, and update security architecture to prevent potential
breaches. Security management across multiple public cloud and private infrastructure environments can become
burdensome.
Microsoft Cloud App Security is a Cloud Access Security Broker (CASB) platform with capabilities for Cloud Security
Posture Management (CSPM). MCAS can connect to multiple cloud services and applications to collect security
logs, monitor user behavior, and impose restrictions that the platforms themselves may not offer.
MCAS provides several capabilities that can integrate with AWS for immediate benefits:
The MCAS app connector uses several AWS APIs, including UBA, to search for configuration issues and threats
on the AWS platform.
AWS Access Controls can enforce sign-in restrictions based on application, device, IP address, location,
registered ISP, and specific user attributes.
Session Controls for AWS block potential malware uploads or downloads based on Microsoft Threat Intelligence
or real-time content inspection.
Session controls can also use real-time content inspection and sensitive data detection to impose data loss
prevention (DLP) rules that prevent cut, copy, paste, or print operations.
MCAS is available standalone, or as part of Microsoft Enterprise Mobility + Security E5, which includes Azure AD
Premium P2. For more pricing and licensing information, see Enterprise Mobility + Security pricing options.
Azure Sentinel for advanced threat detection
Threats can come from a wide range of devices, applications, locations, and user types. Data loss prevention
requires inspecting content during upload or download, because post-mortem review may be too late. AWS
doesn't have native capabilities for device and application management, risk-based conditional access, session-
based controls, or inline UBA.
Azure Sentinel is a Security Information and Event Management (SIEM) and Security Orchestration, Automation,
and Response (SOAR) solution that centralizes and coordinates threat detection and response automation for
modern security operations. Azure Sentinel can monitor AWS accounts to compare events across multiple firewalls,
network devices, and servers. Azure Sentinel combines monitoring data with threat intelligence, analytics rules, and
machine learning to discover and respond to advanced attack techniques.
Connect both AWS and MCAS into Azure Sentinel to get MCAS alerts and run additional threat checks using
multiple Threat Intelligence feeds. Azure Sentinel can initiate a coordinated response outside of MCAS, integrate
with IT Service Management (ITSM) solutions, and retain data long term for compliance purposes.

Security recommendations
The following principles and guidelines are important for any cloud security solution:
Ensure that the organization can monitor, detect, and automatically protect user and programmatic access
into cloud environments.
Continually review current accounts to ensure identity and permission governance and control.
Follow least privilege and zero trust principles. Make sure that each user can access only the specific
resources they require, from trusted devices and known locations. Reduce the permissions of every
administrator and developer to provide only the rights they need for the role they're performing. Review
regularly.
 Continuously monitor platform configuration changes, especially if they provide opportunities for privilege
escalation or attack persistence.
Prevent unauthorized data exfiltration by actively inspecting and controlling content.
Take advantage of solutions you might already own like Azure AD Premium P2 that can increase security
without additional expense.
Basic AWS account security
To ensure basic security hygiene for AWS accounts and resources:
Review the AWS security guidance at Best practices for securing AWS accounts and resources.
Reduce the risk of uploading and downloading malware and other malicious content by actively inspecting
all data transfers through the AWS Management Console. Content that uploads or downloads directly to
resources within the AWS platform, such as web servers or databases, might need additional protection.
Consider protecting access to other resources, including:
Resources created within the AWS account.
Specific workload platforms, like Windows Server, Linux Server, or containers.
Devices that administrators and developers use to access the AWS Management Console.

Plan and prepare

To prepare for deployment of Azure security solutions, review and record current AWS and Azure AD account
information. If you have more than one AWS account deployed, repeat these steps for each account.
1. In the AWS Billing Management Console, record the following current AWS account information:
AWS Account Id , a unique identifier.
Account Name or root user.
Payment method , whether assigned to a credit card or a company billing agreement.
Alternate contacts who have access to AWS account information.
Security questions securely updated and recorded for emergency access.
AWS regions enabled or disabled to comply with data security policy.
2. In the Azure Active Directory portal, review the Azure AD tenant:
Assess Tenant information to see whether the tenant has an Azure AD Premium P1 or P2 license. A P2
license provides Advanced Azure AD identity management features.
Assess Enterprise applications to see whether any existing applications use the AWS application type,
as shown by http://aws.amazon.com/ in the Homepage URL column.

Deploy MCAS
Once you deploy the central management and strong authentication that modern identity and access management
require, you can implement MCAS to:
Collect security data and carry out threat detections for AWS accounts.
Implement advanced controls to mitigate risk and prevent data loss.
To deploy MCAS, you:
1. Add an MCAS app connector for AWS.
2. Configure MCAS monitoring policies for AWS activities.
3. Configure Azure AD session policies for AWS activities.
4. Test MCAS policies for AWS.
Add an AWS app connector
1. In the MCAS portal, expand Investigate and then select Connected apps .
2. On the App Connectors page, select the + and then select Amazon Web Ser vices from the list.
3. Use a unique name for the connector that includes an identifier for the company and specific AWS account,
for example Contoso-AWS-Account1.
4. Follow the instructions at Connect AWS to Microsoft Cloud App Security to create an appropriate AWS IAM
user.
a. Define a policy for restricted permissions.
b. Create a service account to use those permissions on behalf of the MCAS service.
c. Provide the credentials to the app connector.
The initial connection may take some time, depending on the AWS account log sizes. Upon completion, you see a
successful connection confirmation:

Configure MCAS monitoring policies for AWS activities

Once the app connector is enabled, MCAS shows new templates and options in the policy configuration builder.
You can create policies directly from the templates and modify them for your needs, or develop a policy without
using the templates.
To implement policies using the templates:
1. In the MCAS left navigation, expand Control and then select Templates .
2. Search for aws and review the available policy templates for AWS.

3. To use a template, select the + to the right of the template item.

4. Each policy type has different options. Review the configuration and save the policy. Repeat for each of the
templates.
 To use File policies, make sure the file monitoring setting is enabled in MCAS settings:

As MCAS detects alerts, it displays them on the Aler ts page in the MCAS portal:
Configure Azure AD session policies for AWS activities
Session policies are a powerful combination of Azure AD Conditional Access policies and MCAS reverse proxy
capability that provide real-time suspicious behavior monitoring and control.
1. In Azure AD, create a new Conditional Access policy with the following settings:
Name : Enter AWS Console – Session Controls
Users and Groups : Select the two role groups you created earlier:
AWS-Account1-Administrators
AWS-Account1-Developers
Cloud apps or actions : Select the enterprise application you created earlier, Contoso-AWS-Account
1
Session : Select Use Conditional Access App Control
2. Set Enable policy to On .
3. Select Create .
After you create the Azure AD Conditional Access policy, set up an MCAS Session Policy to control user behavior
during AWS sessions.
1. In the MCAS portal, expand Control and then select Policies .
2. On the Policies page, select Create policy and then select Session policy from the list.

3. On the Create session policy page, under Policy template , select Block upload of potential malware
(based on Microsoft Threat Intelligence) .
4. In the ACTIVITIES section, modify the activity filter to include App equal to Amazon Web Ser vices , and
 remove the default device selection.

5. Review the other settings, and then select Create .

Test MCAS policies for AWS
Test all policies regularly to ensure they're still effective and relevant. Here are a few recommended tests:
IAM Policy changes : This policy should trigger each time you attempt to modify the settings within AWS
IAM, such as creating the new IAM policy and account to use in the following Azure Sentinel section.
Console sign-in failures : Any failed attempts to sign in to one of the test accounts trigger this policy. The
alert details show that the attempt came from one of the Azure regional datacenters.
S3 bucket activity policy : Attempting to create a new AWS S3 storage account and set it to be publicly
available triggers the policy.
Malware detection policy : If you configured malware detection as a session policy, you can test it by
uploading a file to an AWS S3 storage account. You can download a safe test file from the European Institute
for Computer Anti-Virus Research (EICAR). The policy should immediately block you from uploading the file,
and you should see the alert trigger in the MCAS portal shortly afterwards.

Deploy Azure Sentinel

Connecting an AWS account and MCAS to Azure Sentinel enables monitoring capabilities that compare events
across multiple firewalls, network devices, and servers.
Enable the Azure Sentinel AWS connector
After you enable the Azure Sentinel Connector for AWS, you can monitor AWS incidents and data ingestion.
As with the MCAS configuration, this connection requires configuring AWS IAM to provide credentials and
permissions.
1. In AWS IAM, follow the steps at Connect Azure Sentinel to AWS CloudTrail.
2. To complete the configuration in the Azure portal, under Azure Sentinel > Data connectors , select the
Amazon Web Ser vices connector.
3. Select Open connector page .
4. Under Configuration , enter the Role ARN from the AWS IAM configuration in the Role to add field, and
select Add .
5. Select Next steps , and select the AWS Network Activities and AWS User Activities activities to
monitor.
6. Under Relevant analytic templates , select Create rule next to the AWS analytic templates you want to
enable.
7. Set up each rule, and select Create .
The following table shows the available rule templates for checking AWS entity behaviors and threat indicators. The
rule names describe their purpose, and the potential data sources list the data sources each rule can use.

A N A LY T IC T EM P L AT E N A M E DATA SO URC ES

Known IRIDIUM IP DNS, Azure Monitor, Cisco ASA, Palo Alto Networks, Azure AD,
Azure Activity, AWS

Full Admin policy created and then attached to Roles, Users, or AWS
Groups

Failed AzureAD logons but success logon to AWS Console Azure AD, AWS
 A N A LY T IC T EM P L AT E N A M E DATA SO URC ES

Failed AWS Console logons but success logon to AzureAD Azure AD, AWS

MFA disabled for a user Azure AD, AWS

Changes to AWS Security Group ingress and egress settings AWS

Monitor AWS Credential abuse or hijacking AWS

Changes to AWS Elastic Load Balancer security groups AWS

Changes to Amazon VPC settings AWS

New UserAgent observed in last 24 hours Microsoft 365, Azure Monitor, AWS

Login to AWS Management Console without MFA AWS

Changes to internet facing AWS RDS Database instances AWS

Changes made to AWS CloudTrail logs AWS

Threat Intelligence map IP entity to AWS CloudTrail Threat Intelligence Platforms, AWS

Enabled templates have an IN USE indicator on the connector details page:

Monitor AWS incidents

Azure Sentinel creates incidents based on the enabled analyses and detections. Each incident can include one or
more events, which reduces the overall number of investigations necessary to detect and respond to potential
threats.
Azure Sentinel shows incidents generated by MCAS, if connected, and incidents created by Azure Sentinel. The
Product names column shows the Incident source.
Check data ingestion
Check that data is continuously ingested into Azure Sentinel by regularly viewing the connector details. The
following graph shows a new connection:

If the data stops ingesting and the graph drops, check the credentials used to connect to the AWS account, and
check that AWS CloudTrail can still collect the events.

See also
For in-depth coverage and comparison of Azure and AWS features, see the Azure for AWS professionals content
set.
For security guidance from AWS, see Best practices for securing AWS accounts and resources.
For the latest Microsoft security information, see www.microsoft.com/security.
For full details of how to implement and manage Azure AD, see Securing Azure environments with Azure Active
Directory.
Azure AD identity and access management for AWS.
Connect AWS to Microsoft Cloud App Security.
How Cloud App Security helps protect your Amazon Web Services (AWS) environment.
Connect Azure Sentinel to AWS CloudTrail.
 Confidential computing on a healthcare platform
12/18/2020 • 6 minutes to read • Edit Online

When organizations collaborate, they share information. But most parties don't want to give other parties access to
all parts of the data. Mechanisms exist for safeguarding data at rest and in transit. However, encrypting data in use
poses different challenges. This article presents a solution that Azure confidential computing (ACC) offers for
encrypting in-use data.
By using confidential computing and containers, the solution provides a way for a provider-hosted application to
securely collaborate with a hospital and a third-party diagnostic provider. Azure Kubernetes Service (AKS) hosts
confidential computing nodes. Azure Attestation establishes trust with the diagnostic provider. By using these Azure
components, the architecture isolates the sensitive data of the hospital patients while the specific shared data is
being processed in the cloud. The hospital data is then inaccessible to the diagnostic provider. Through this
architecture, the provider-hosted application can also take advantage of advanced analytics. The diagnostic provider
makes these analytics available as confidential computing services of machine learning (ML) applications.

Potential use cases

Many industries protect their data by using confidential computing for these purposes:
Securing financial data
Protecting patient information
Running ML processes on sensitive information
Performing algorithms on encrypted datasets from many sources
Protecting container data and code integrity

Architecture

Diagram showing how data flows between three parties in a healthcare setting. Three rectangles represent the
three parties: a hospital, a medical platform, and a diagnostic provider. Each rectangle contains icons that represent
various components, such as a website, a client application, Azure Attestation, a web API, data storage, and a
runtime. The medical platform and diagnostic provider rectangles also contain smaller rectangles that represent
confidential nodes and A K S clusters. Arrows connect these components and show the flow of data. Numbered
callouts correspond to the steps that this article describes after the diagram.
The diagram outlines the architecture. Throughout the system:
Network communication is TLS encrypted in transit.
Azure Monitor tracks component performance, and Azure Container Registry (ACR) manages the solution's
containers.
The solution involves the following steps:
1. A clerk for Lamna Hospital opens the hospital's web portal, an Azure Blob Storage static website, to enter
patient data.
2. The clerk enters data into the hospital's web portal, which Contoso Medical Platform Ltd. powers with a
Python Flask–based web API. A confidential node in the SCONE confidential computing software protects the
patient data. SCONE works within an AKS cluster that has the Software Guard Extensions (SGX) enabled that
help run the container in an enclave.
3. The Python code is wrapped in SCONE SGX software. The solution deploys that code on an AKS confidential
node as a confidential container. The code stores the data in memory within a Redis cache (3a ), using Azure
Attestation to establish trust (3b ).
4. If the code doesn't find the patient data, it prompts the clerk to enter the sensitive information on a form. The
code then stores that information in Redis.
5. The Contoso application sends the protected patient data from its AKS cluster to an enclave in the Open
Neural Network Exchange (ONNX) runtime server. Fabrikam Diagnostic Provider hosts this confidential
inferencing server in a confidential node of the Fabrikam AKS cluster.
6. The Fabrikam machine learning–based application obtains the diagnostic results from the confidential
inferencing ONNX runtime server. The app sends these results back to the confidential node in the Contoso
AKS cluster.
7. The Contoso application sends the diagnostic results from the confidential node back to Lamna's client
application.
Components
Static website hosting in Blob Storage serves static content like HTML, CSS, JavaScript, and image files
directly from a storage container.
Azure Attestation is a unified solution that remotely verifies the trustworthiness of a platform. Azure
Attestation also remotely verifies the integrity of the binaries that run in the platform. Use Azure Attestation
to establish trust with the confidential application.
AKS Cluster simplifies the process of deploying a Kubernetes cluster.
Confidential computing nodes are hosted on a specific virtual machine series that can run sensitive
workloads on AKS within a hardware-based trusted execution environment (TEE) by allowing user-level code
to allocate private regions of memory, known as enclaves. Confidential computing nodes can support
confidential containers or enclave-aware containers.
SCONE platform is an Azure Partner independent software vendor (ISV) solution from Scontain.
Redis is an open-source, in-memory data structure store.
Secure Container Environment (SCONE) supports the execution of confidential applications in containers
that run inside a Kubernetes cluster.
Confidential Inferencing ONNX Runtime Server Enclave (ONNX RT - Enclave) is a host that restricts the ML
hosting party from accessing both the inferencing request and its corresponding response.
 Azure Monitor collects and analyzes app telemetry, such as performance metrics and activity logs.
ACR is a service that creates a managed registry. ACR builds, stores, and manages container images and can
store containerized machine learning models.
Alternatives
You can use Fortanix instead of SCONE to deploy confidential containers to use with your containerized
application. Fortanix provides the flexibility you need to run and manage the broadest set of applications:
existing applications, new enclave-native applications, and pre-packaged applications.
Graphene is a lightweight, open-source guest OS. Graphene can run a single Linux application in an isolated
environment with benefits comparable to running a complete OS. It has good tooling support for converting
existing Docker container applications to Graphene Shielded Containers (GSC).

Considerations
Azure confidential computing virtual machines (VMs) are available in 2nd-generation D family sizes for general
purpose needs. These sizes are known collectively as D-Series v2 or DCsv2 series. This scenario uses Intel SGX-
enabled DCs_v2-series virtual machines with Gen2 operating system (OS) images. But you can only deploy certain
sizes in certain regions. For more information, see Quickstart: Deploy an Azure Confidential Computing VM in the
Marketplace and Products available by region.

Deploy this scenario

Deploying this scenario involves the following high-level steps:
Deploy the confidential inferencing server on an existing SGX-enabled AKS Cluster. See the confidential
ONNX inference server project on GitHub for information on this step.
Configure Azure Attestation policies.
Deploy an SGX-enabled AKS cluster node pool.
Get access to curated confidential applications called SconeApps. SconeApps are available on a private
GitHub repository that's currently only available for commercial customers, through SCONE Standard
Edition. Go to the SCONE website and contact the company directly to get this service level.
Install and run SCONE services on your AKS cluster.
Install and test the Flask-based application on your AKS cluster.
Deploy and access the web client.
These steps focus on the enclave containers. A secured infrastructure would extend beyond this implementation
and include compliance requirements, such as added protections required by HIPAA.

Pricing
To explore the cost of running this scenario, use the Azure pricing calculator, which preconfigures all Azure services.
A sample cost profile is available for the Contoso Medical SaaS Platform, as pictured in the diagram. It includes the
following components:
System node pool and SGX node pool: no disks, all ephemeral
AKS Load Balancer
Azure Virtual Network: nominal
Azure Container Registry
 Storage account for single-page application (SPA)
The profile doesn't include the following components:
Azure Attestation Service: free
Azure Monitor Logs: usage based
SCONE ISV licensing
Compliance services required for solutions working with sensitive data, including:
Azure Security Center and Azure Defender for Kubernetes
Azure DDoS Protection: standard
Azure Firewall
Azure Application Gateway and Azure Web Application Firewall

Next steps
Learn more about Azure confidential computing.
See the confidential ONNX inference server project on GitHub.

Related resources
Confidential containers on AKS.
Official ONNX runtime website.
Confidential ONNX inference server (GitHub sample).
MobileCoin use case with anonymized blockchain data.
A sample brain segmentation image for use with the delineation function that invokes the confidential
inferencing server.