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

AWS Canberra WWPS Summit 2013 - AWS for Web Applications

•
•
Amazon Web Services
Amazon Web Services

Highly available and scalable web hosting can be complex and expensive. Learn how Amazon Web Services provides the reliable, scalable, secure, and high performance infrastructure required for web applications while enabling an elastic, scale out and scale down infrastructure to match IT costs in real time as customer traffic fluctuates.

1 of 34
2013 AWS WWPS Summit Canberra, Australia AWS for Web Applications Glenn Gore Sr Manager AWS
I am Barack Obama, Ask me anything Reddit Needed to Scale for a special guest • 2,987,307 pageviews on the day of the IAmA • President Obama’s user page received 428,004 pageviews on the day of the IAMA • Added 60 dedicated instance to handle the increased load • At peek transfering 48 MB/s to the internet
2013 AWS WWPS Summit, Canberra – May 23 While You’re Scaling • Architect for Failure – Failures do happen – Internet is unpredictable – People are unpredictable • Architect with Security – Security must happen – Security is multi-layered – Security can be automated while you scale
2013 AWS WWPS Summit, Canberra – May 23 Why Is Scale Important? Self Hosting Waste Customer Dissatisfaction Actual demand Predicted Demand Rigid Elastic Actual demand AWS
Web Application Scaling
2013 AWS WWPS Summit, Canberra – May 23 Loose coupling sets you free! • The looser they're coupled, the bigger they scale – Independent components – Design everything as a black box – Decouple interactions – Load-balance clusters Controller A Controller B Controller C Controller A Controller B Controller C Q Q Q Tight Coupling Use Amazon SQS as Buffers Loose Coupling
2013 AWS WWPS Summit, Canberra – May 23 Allows for Parallel Processing and Failure • Fan out • Use varied instance types • Use varied billing models • Protects against failure • Different workloads can scale independently
2013 AWS WWPS Summit, Canberra – May 23 Allows for Parallel Processing and Failure
2013 AWS WWPS Summit, Canberra – May 23 Go wide, not high! • Build for horizontal scale – Decrease request contention – Reduce capacity planning headaches – Requires a stateless application architecture – Allows for dynamic scalability Route 53 ELB Auto Scale EC2 S3 DynamoDBRDSCloudWatch Cloud Formation SQS SNS
But usually some state has to reside somewhere Cookies in browser Memory-resident session manager Session database Framework-provided session handler
So this store of state needs to be… Performant Scalable Reliable
Where should session state reside? Trigger auto- scaling policy Session State Service Not Here Here State must reside OUTSIDE the scope of the elements you wish to scale
And what do I build it on? The state service itself must be well architected
2013 AWS WWPS Summit, Canberra – May 23 Data Tier Scaling The bane of the Architect’s existence
2013 AWS WWPS Summit, Canberra – May 23 Vertical Scaling “We’re gonna need a bigger box” • Simplest approach • Can now leverage PIOPs • High I/O instances • Easy to change instance sizes • Will hit an endpoint eventually hi1.4xlarge m2.4xlarge m1.small
2013 AWS WWPS Summit, Canberra – May 23 Master/Slave Horizontal Scaling • Reasonably simple to adapt to • Can now leverage PIOPs • Easy to change instances sizes • Will hit an endpoint eventually
2013 AWS WWPS Summit, Canberra – May 23 Sharded Horizontal Scaling Hash Ring A BC D • More complex at the application layer • ORM support can help • No practical limit on scalability • Operation complexity/sophistication • Shard by function or key space • RDBMS or NoSQL
2013 AWS WWPS Summit, Canberra – May 23 Horizontal Scaling – Fully Managed DynamoDB • Provisioned throughput NoSQL database • Fast, predictable performance • Fully distributed, fault tolerant architecture • Considerations for non-uniform data Feature Details Provisioned throughput Dial up or down provisioned read/write capacity. Predictable performance Average single digit millisecond latencies from SSD-backed infrastructure. Strong consistency Be sure you are reading the most up to date values. Fault tolerant Data replicated across Availability Zones. Monitoring Integrated to CloudWatch. Secure Integrates with AWS Identity and Access Management (IAM). Elastic MapReduce Integrates with Elastic MapReduce for complex analytics on large datasets.
2013 AWS WWPS Summit, Canberra – May 23 Petabyte-Scale Data Warehousing Feature Details Optimized for Data Warehousing Redshift uses a variety of innovations to obtain very high query performance on datasets ranging in size from hundreds of gigabytes to a petabyte or more. Scalable Easily scale the number of nodes in your data warehouse up or down as your performance or capacity needs change Fault tolerant Data replicated across Availability Zones. Monitoring Integrated to CloudWatch. Secure Encrypt data in transit and at rest. Can also be run in VPC to isolate your data warehouse cluster. S3 intergration Loads data in parallel to each node from S3. Elastic MapReduce Integrates with ERM via Data Pipeline.
2013 AWS WWPS Summit, Canberra – May 23 AWS Application Management Solutions Elastic Beanstalk OpsWorks CloudFormation EC2 Convenience Control Higher-level Services Do it yourself
2013 AWS WWPS Summit, Canberra – May 23
2013 AWS WWPS Summit, Canberra – May 23 Summary • Awareness of the options is the first step to good design • Scaling is the ability to move the bottlenecks around to the least expensive part of the architecture • Loose coupling is good • AWS makes this easier – so your application is not a victim of its own success
Assessment of research impact Robyn Owens, DVC-Research UWA Class2Go and AWS Associate Professor David Glance
What is a MOOC? • Massive: no restrictions on numbers – from 1 to 160,000 • Open: free and open to anyone – no barriers to entry • Online: everything is online, special format videos, automated quizzes, peer assessment • Course: unit of 6 – 10 weeks in duration but can be self- paced
Background • First popularised in 2011 when Stanford Artificial Intelligence Course is put online and attracts 160,000 student enrollments • MOOC platforms developed Coursera from Stanford, edX from Harvard/MIT • Sebastian Thrun forms Udacity • Coursera now has 323 courses from 62 universities with 2.7m enrollments
AWS Canberra WWPS Summit 2013 - AWS for Web Applications
AWS Canberra WWPS Summit 2013 - AWS for Web Applications
Class2Go • Class2Go open source MOOC platform developed by Stanford University (Office of Vice Provost for Online Learning) • UWA is collaborating on development: – Modifying to support different universities – Updating management tools to support Fabric/Cuisine/Boto • Class2Go is being used to host UWA courses for UWA students and general public – 3 Courses so far – 2 more in July (licensed content from Stanford for one of them) – Possibility of Class2Go becoming 2nd Learning Management System at UWA • UWA is assisting Stanford in promoting Class2Go amongst other universities
Class2Go Software Architecture • Load balanced using ELB • Python Django Apache Web Application running on Ubuntu – Hosted on EC2 • MySQL database hosted on RDS • Uses video hosted on YouTube and downloaded versions transcoded using ffmpeg – Video and other assets stored on S3 • Communicates with utility servers for reports, transcoding, mail-outs using SQS • Mail handled using SES • DNS handled using Route 53 • Handling automated marking with Vagrant VMs
Class2Go Architecture
Class2Go Management • Original management via Chef on Opscode • Moved to simpler interface using Python Fabric/Cuisine/Boto – Uses SSH to manage machines – doesn’t require server – Tailored to roles of servers • Creating an App Server instance involves: – Provisioning an EC2 instance with the correct image – Updating Ubuntu – Installing Ubuntu/Python/Django packages – Installing app from github – Installing Apache and configuring – Adding instance to ELB – All run from command line
Class2Go Video Transcoding • Uploaded video is copied to YouTube • Saved to S3 • Job placed on SQS for utility server to pick up • Transcoded to smaller size • JPG images created from the video for thumbnails • Results stored in S3
General Observations • Generally it is very straightforward and functionality is reliable and simple to set up • API is relatively easy to use and makes management of software and systems environment straightforward • Use CloudWatch for alarms on services and monitor usage etc • Much more cost effective and administratively easier to use than going through IT department at the University
2013 AWS WWPS Summit Canberra, Australia

More Related Content

AWS Canberra WWPS Summit 2013 - AWS for Web Applications

  • 1. 2013 AWS WWPS Summit Canberra, Australia AWS for Web Applications Glenn Gore Sr Manager AWS
  • 2. I am Barack Obama, Ask me anything Reddit Needed to Scale for a special guest • 2,987,307 pageviews on the day of the IAmA • President Obama’s user page received 428,004 pageviews on the day of the IAMA • Added 60 dedicated instance to handle the increased load • At peek transfering 48 MB/s to the internet
  • 3. 2013 AWS WWPS Summit, Canberra – May 23 While You’re Scaling • Architect for Failure – Failures do happen – Internet is unpredictable – People are unpredictable • Architect with Security – Security must happen – Security is multi-layered – Security can be automated while you scale
  • 4. 2013 AWS WWPS Summit, Canberra – May 23 Why Is Scale Important? Self Hosting Waste Customer Dissatisfaction Actual demand Predicted Demand Rigid Elastic Actual demand AWS
  • 6. 2013 AWS WWPS Summit, Canberra – May 23 Loose coupling sets you free! • The looser they're coupled, the bigger they scale – Independent components – Design everything as a black box – Decouple interactions – Load-balance clusters Controller A Controller B Controller C Controller A Controller B Controller C Q Q Q Tight Coupling Use Amazon SQS as Buffers Loose Coupling
  • 7. 2013 AWS WWPS Summit, Canberra – May 23 Allows for Parallel Processing and Failure • Fan out • Use varied instance types • Use varied billing models • Protects against failure • Different workloads can scale independently
  • 8. 2013 AWS WWPS Summit, Canberra – May 23 Allows for Parallel Processing and Failure
  • 9. 2013 AWS WWPS Summit, Canberra – May 23 Go wide, not high! • Build for horizontal scale – Decrease request contention – Reduce capacity planning headaches – Requires a stateless application architecture – Allows for dynamic scalability Route 53 ELB Auto Scale EC2 S3 DynamoDBRDSCloudWatch Cloud Formation SQS SNS
  • 10. But usually some state has to reside somewhere Cookies in browser Memory-resident session manager Session database Framework-provided session handler
  • 11. So this store of state needs to be… Performant Scalable Reliable
  • 12. Where should session state reside? Trigger auto- scaling policy Session State Service Not Here Here State must reside OUTSIDE the scope of the elements you wish to scale
  • 13. And what do I build it on? The state service itself must be well architected
  • 14. 2013 AWS WWPS Summit, Canberra – May 23 Data Tier Scaling The bane of the Architect’s existence
  • 15. 2013 AWS WWPS Summit, Canberra – May 23 Vertical Scaling “We’re gonna need a bigger box” • Simplest approach • Can now leverage PIOPs • High I/O instances • Easy to change instance sizes • Will hit an endpoint eventually hi1.4xlarge m2.4xlarge m1.small
  • 16. 2013 AWS WWPS Summit, Canberra – May 23 Master/Slave Horizontal Scaling • Reasonably simple to adapt to • Can now leverage PIOPs • Easy to change instances sizes • Will hit an endpoint eventually
  • 17. 2013 AWS WWPS Summit, Canberra – May 23 Sharded Horizontal Scaling Hash Ring A BC D • More complex at the application layer • ORM support can help • No practical limit on scalability • Operation complexity/sophistication • Shard by function or key space • RDBMS or NoSQL
  • 18. 2013 AWS WWPS Summit, Canberra – May 23 Horizontal Scaling – Fully Managed DynamoDB • Provisioned throughput NoSQL database • Fast, predictable performance • Fully distributed, fault tolerant architecture • Considerations for non-uniform data Feature Details Provisioned throughput Dial up or down provisioned read/write capacity. Predictable performance Average single digit millisecond latencies from SSD-backed infrastructure. Strong consistency Be sure you are reading the most up to date values. Fault tolerant Data replicated across Availability Zones. Monitoring Integrated to CloudWatch. Secure Integrates with AWS Identity and Access Management (IAM). Elastic MapReduce Integrates with Elastic MapReduce for complex analytics on large datasets.
  • 19. 2013 AWS WWPS Summit, Canberra – May 23 Petabyte-Scale Data Warehousing Feature Details Optimized for Data Warehousing Redshift uses a variety of innovations to obtain very high query performance on datasets ranging in size from hundreds of gigabytes to a petabyte or more. Scalable Easily scale the number of nodes in your data warehouse up or down as your performance or capacity needs change Fault tolerant Data replicated across Availability Zones. Monitoring Integrated to CloudWatch. Secure Encrypt data in transit and at rest. Can also be run in VPC to isolate your data warehouse cluster. S3 intergration Loads data in parallel to each node from S3. Elastic MapReduce Integrates with ERM via Data Pipeline.
  • 20. 2013 AWS WWPS Summit, Canberra – May 23 AWS Application Management Solutions Elastic Beanstalk OpsWorks CloudFormation EC2 Convenience Control Higher-level Services Do it yourself
  • 21. 2013 AWS WWPS Summit, Canberra – May 23
  • 22. 2013 AWS WWPS Summit, Canberra – May 23 Summary • Awareness of the options is the first step to good design • Scaling is the ability to move the bottlenecks around to the least expensive part of the architecture • Loose coupling is good • AWS makes this easier – so your application is not a victim of its own success
  • 23. Assessment of research impact Robyn Owens, DVC-Research UWA Class2Go and AWS Associate Professor David Glance
  • 24. What is a MOOC? • Massive: no restrictions on numbers – from 1 to 160,000 • Open: free and open to anyone – no barriers to entry • Online: everything is online, special format videos, automated quizzes, peer assessment • Course: unit of 6 – 10 weeks in duration but can be self- paced
  • 25. Background • First popularised in 2011 when Stanford Artificial Intelligence Course is put online and attracts 160,000 student enrollments • MOOC platforms developed Coursera from Stanford, edX from Harvard/MIT • Sebastian Thrun forms Udacity • Coursera now has 323 courses from 62 universities with 2.7m enrollments
  • 28. Class2Go • Class2Go open source MOOC platform developed by Stanford University (Office of Vice Provost for Online Learning) • UWA is collaborating on development: – Modifying to support different universities – Updating management tools to support Fabric/Cuisine/Boto • Class2Go is being used to host UWA courses for UWA students and general public – 3 Courses so far – 2 more in July (licensed content from Stanford for one of them) – Possibility of Class2Go becoming 2nd Learning Management System at UWA • UWA is assisting Stanford in promoting Class2Go amongst other universities
  • 29. Class2Go Software Architecture • Load balanced using ELB • Python Django Apache Web Application running on Ubuntu – Hosted on EC2 • MySQL database hosted on RDS • Uses video hosted on YouTube and downloaded versions transcoded using ffmpeg – Video and other assets stored on S3 • Communicates with utility servers for reports, transcoding, mail-outs using SQS • Mail handled using SES • DNS handled using Route 53 • Handling automated marking with Vagrant VMs
  • 31. Class2Go Management • Original management via Chef on Opscode • Moved to simpler interface using Python Fabric/Cuisine/Boto – Uses SSH to manage machines – doesn’t require server – Tailored to roles of servers • Creating an App Server instance involves: – Provisioning an EC2 instance with the correct image – Updating Ubuntu – Installing Ubuntu/Python/Django packages – Installing app from github – Installing Apache and configuring – Adding instance to ELB – All run from command line
  • 32. Class2Go Video Transcoding • Uploaded video is copied to YouTube • Saved to S3 • Job placed on SQS for utility server to pick up • Transcoded to smaller size • JPG images created from the video for thumbnails • Results stored in S3
  • 33. General Observations • Generally it is very straightforward and functionality is reliable and simple to set up • API is relatively easy to use and makes management of software and systems environment straightforward • Use CloudWatch for alarms on services and monitor usage etc • Much more cost effective and administratively easier to use than going through IT department at the University
  • 34. 2013 AWS WWPS Summit Canberra, Australia