Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
SlideShare a Scribd company logo

Sergii Baidachnyi ITEM 2018

•
•
ITEM
ITEM

This document discusses Azure Machine Learning services for data scientists. It provides an overview of Azure Machine Learning Studio for building and deploying machine learning models with over 100 modules. Numbers show hundreds of thousands of deployed models serving billions of requests. It also discusses Azure Batch AI for scalable machine learning training without managing infrastructure, and Azure Databricks for Apache Spark as a managed service on Azure. The document outlines the machine learning development lifecycle supported in Azure and tools for experimentation, model management, and operationalization of models.

1 of 28
Download to read offline
Azure Machine Learning for Data Scientists Sergii Baidachnyi Principal Software Engineer Microsoft sbaydach@microsoft.com @sbaidachni
Looking back
Offering Platform for emerging data scientists to graphically build and deploy experiments Key Value Props • Rapid experiment composition • > 100 easily configured modules for data prep, training, evaluation • Extensibility through R & Python • Serverless training and deployment Numbers • 100’s of thousands of deployed models serving billions of requests Azure Machine Learning Studio
Azure Batch AI
Infrastructure Can Get in Your Way Clusters • Provision GPUs • Install drivers and software • Interactive use Scheduling • Queue work • Prioritize jobs • Start MPI • Monitor • Handle failures Data • Scale access to training data • Output logs & models • Secure & compliant Cost • Scale up and down • Share reserved instances • Low priority Workflow • Choose efficient hardware • Tooling integration • Laptop to cloud
• Managed Service • Supports Role Based Access Control • Run any toolkit (CNTK, Tensorflow, Caffee/Caffee2, Chainer, Keras, …) • Run experiments in Parallel • Run in Containers or directly on VM • Support various Shared File Systems • Load based automatic scaling • Only Storage and compute cost. Service is free Azure Batch AI Service
Azure DataBricks Databricks Spark as a managed service on Azure
CONTROL EASE OF USE Azure Data Lake Store Azure Storage Any Hadoop technology, any distribution Workload optimized, managed clusters Data Engineering in a Job-as-a-service model Azure Marketplace HDP | CDH | MapR Azure Data Lake Analytics IaaS Clusters Managed Clusters Big Data as-a-service Azure HDInsight Frictionless & Optimized Spark clusters Azure Databricks BIGDATA STORAGE BIGDATA ANALYTICS ReducedAdministration IaaS and PaaS Big Data Analytics
Azure Databricks Microsoft Azure
Optimized Databricks Runtime Engine DATABRICKS I/O SERVERLESS Collaborative Workspace Cloud storage Data warehouses Hadoop storage IoT / streaming data Rest APIs Machine learning models BI tools Data exports Data warehouses Azure Databricks Enhance Productivity Deploy Production Jobs & Workflows APACHE SPARK MULTI-STAGE PIPELINES DATA ENGINEER JOB SCHEDULER NOTIFICATION & LOGS DATA SCIENTIST BUSINESS ANALYST Build on secure & trusted cloud Scale without limits Azure Databricks
Azure Databricks Cluster Architecture Azure DB for PostgreSQL Webapp Azure Compute Cluster Manager Databricks’ Azure Account User’s Azure Account Azure Compute Spark Driver Azure Compute Spark Worker Azure Compute Spark Worker Jobs FileSystem Service Spark History Server Log Daemon Log Daemon
Azure Databricks Core Artifacts Azure Databricks
Azure Machine Learning Experimentation and Management
Apps + insights Social LOB Graph IoT Image CRM INGEST STORE PREP & TRAIN MODEL & SERVE Data orchestration and monitoring Data lake and storage Hadoop/Spark/SQL and ML . IoT Azure Machine Learning The AI Development lifecycle
Local machine Scale up to DSVM Scale out with Spark on HDInsight Azure Batch AI (Coming Soon) ML Server (Coming Soon) Experiment Anywhere A ZURE ML EXPERIMENTATION Command line tools IDEs Notebooks in Workbench VS Code Tools for AI
Transparent Compute
Demo Experimentation Service
DOCKER Single node deployment (cloud/on-prem) Azure Container Service Azure IoT Edge Microsoft ML Server Spark clusters SQL Server (Coming Soon) Deploy Everywhere A ZURE ML MODEL MANAGEMENT
Model Management
Machine Learning Server
R Server Overview • Enhances upon open source R to scale to big data • Embraces combined open source and commercial innovations • Allows customers to get the support they trust • Microsoft innovations: • RevoScaleR • Parallelized, distributed algorithms • Microsoft Machine learning • Fast and Deep learning • Pretrained models • Custom parallel frameworks
ML Services Version 9.2 at a glance Platforms & Data Tools Languages Algorithms Data Sources Rattle Mrsdeploy RESTful API deployment Real-Time Scoring Visualization Tool Integration .csv Microsoft .XDF In-database deployment Operationalization Distributed Parallelized Algorithms: •RevoScaleR and RevoScalePy libraries •MicrosoftML library •Custom parallelization frameworks Open source R algorithms & visualizations: •CRAN •bioconductor Plus: •Deep Learning •Pretrained Models •Prebuilt Featurizers ODBC/JDBC
Looking forward
Data Science lifecycle •Primary stages: Lifecycle
TDSP objective Integrate DevOps with data science workflows to improve collaboration, quality, robustness and efficiency in data science projects o Infrastructure as Code (IaC) o Building o Testing o CI / CD o … o App performance monitoring
TDSP documentation: https://aka.ms/tdsp
Using TDSP within Azure Machine Learning
Questions? sbaydach@microsoft.com @sbaidachni

More Related Content

Sergii Baidachnyi ITEM 2018

  • 1. Azure Machine Learning for Data Scientists Sergii Baidachnyi Principal Software Engineer Microsoft sbaydach@microsoft.com @sbaidachni
  • 3. Offering Platform for emerging data scientists to graphically build and deploy experiments Key Value Props • Rapid experiment composition • > 100 easily configured modules for data prep, training, evaluation • Extensibility through R & Python • Serverless training and deployment Numbers • 100’s of thousands of deployed models serving billions of requests Azure Machine Learning Studio
  • 5. Infrastructure Can Get in Your Way Clusters • Provision GPUs • Install drivers and software • Interactive use Scheduling • Queue work • Prioritize jobs • Start MPI • Monitor • Handle failures Data • Scale access to training data • Output logs & models • Secure & compliant Cost • Scale up and down • Share reserved instances • Low priority Workflow • Choose efficient hardware • Tooling integration • Laptop to cloud
  • 6. • Managed Service • Supports Role Based Access Control • Run any toolkit (CNTK, Tensorflow, Caffee/Caffee2, Chainer, Keras, …) • Run experiments in Parallel • Run in Containers or directly on VM • Support various Shared File Systems • Load based automatic scaling • Only Storage and compute cost. Service is free Azure Batch AI Service
  • 7. Azure DataBricks Databricks Spark as a managed service on Azure
  • 8. CONTROL EASE OF USE Azure Data Lake Store Azure Storage Any Hadoop technology, any distribution Workload optimized, managed clusters Data Engineering in a Job-as-a-service model Azure Marketplace HDP | CDH | MapR Azure Data Lake Analytics IaaS Clusters Managed Clusters Big Data as-a-service Azure HDInsight Frictionless & Optimized Spark clusters Azure Databricks BIGDATA STORAGE BIGDATA ANALYTICS ReducedAdministration IaaS and PaaS Big Data Analytics
  • 10. Optimized Databricks Runtime Engine DATABRICKS I/O SERVERLESS Collaborative Workspace Cloud storage Data warehouses Hadoop storage IoT / streaming data Rest APIs Machine learning models BI tools Data exports Data warehouses Azure Databricks Enhance Productivity Deploy Production Jobs & Workflows APACHE SPARK MULTI-STAGE PIPELINES DATA ENGINEER JOB SCHEDULER NOTIFICATION & LOGS DATA SCIENTIST BUSINESS ANALYST Build on secure & trusted cloud Scale without limits Azure Databricks
  • 11. Azure Databricks Cluster Architecture Azure DB for PostgreSQL Webapp Azure Compute Cluster Manager Databricks’ Azure Account User’s Azure Account Azure Compute Spark Driver Azure Compute Spark Worker Azure Compute Spark Worker Jobs FileSystem Service Spark History Server Log Daemon Log Daemon
  • 12. Azure Databricks Core Artifacts Azure Databricks
  • 14. Apps + insights Social LOB Graph IoT Image CRM INGEST STORE PREP & TRAIN MODEL & SERVE Data orchestration and monitoring Data lake and storage Hadoop/Spark/SQL and ML . IoT Azure Machine Learning The AI Development lifecycle
  • 15. Local machine Scale up to DSVM Scale out with Spark on HDInsight Azure Batch AI (Coming Soon) ML Server (Coming Soon) Experiment Anywhere A ZURE ML EXPERIMENTATION Command line tools IDEs Notebooks in Workbench VS Code Tools for AI
  • 18. DOCKER Single node deployment (cloud/on-prem) Azure Container Service Azure IoT Edge Microsoft ML Server Spark clusters SQL Server (Coming Soon) Deploy Everywhere A ZURE ML MODEL MANAGEMENT
  • 21. R Server Overview • Enhances upon open source R to scale to big data • Embraces combined open source and commercial innovations • Allows customers to get the support they trust • Microsoft innovations: • RevoScaleR • Parallelized, distributed algorithms • Microsoft Machine learning • Fast and Deep learning • Pretrained models • Custom parallel frameworks
  • 22. ML Services Version 9.2 at a glance Platforms & Data Tools Languages Algorithms Data Sources Rattle Mrsdeploy RESTful API deployment Real-Time Scoring Visualization Tool Integration .csv Microsoft .XDF In-database deployment Operationalization Distributed Parallelized Algorithms: •RevoScaleR and RevoScalePy libraries •MicrosoftML library •Custom parallelization frameworks Open source R algorithms & visualizations: •CRAN •bioconductor Plus: •Deep Learning •Pretrained Models •Prebuilt Featurizers ODBC/JDBC
  • 25. TDSP objective Integrate DevOps with data science workflows to improve collaboration, quality, robustness and efficiency in data science projects o Infrastructure as Code (IaC) o Building o Testing o CI / CD o … o App performance monitoring
  • 27. Using TDSP within Azure Machine Learning