How to handle cloud failure.
1 like286 views
The document discusses how to handle cloud failures when using Amazon AWS. It recommends: 1) Designing infrastructure to be redundant across multiple availability zones and regions to prevent outages from single failures. 2) Decoupling applications so that they are stateless and can degrade gracefully or auto-recover from failures. 3) Automating infrastructure, monitoring, scaling and recovery processes to minimize downtime from failures. 4) Continuously testing infrastructure resilience through failure simulation tools like ChaosMonkey. 5) Weighing costs and risks of different redundancy and automation strategies for an organization's specific needs.
How to handle cloud failure.
- 1. How to handle cloud failure Grzegorz Kochan | grzegorz@adtaily.com 1
- 2. subtitle text Jak sobie radzić z awarią w chmurach Grzegorz Kochan | grzegorz@adtaily.com 2
- 3. AdTaily on Amazon AWS 1,5 bilion widget pageviews monthly 100 000 ad clicks daily 35 thousand registered publishers 15 thousand advertisers over 1500 requests per second over 150 mbit data per second Grzegorz Kochan | grzegorz@adtaily.com 3
- 4. Startup in the cloud why? Availability Pricing API Scalability Simplicity Grzegorz Kochan | grzegorz@adtaily.com 4
- 5. But things can go wrong TechCrunch “Amazon EC2 goes down, taking with it Reddit, Foursquare and Quora” - April 2011 „Down Goes The Internet… Again. Amazon EC2 Outage Takes Down Foursquare, Instagram, Quora, Reddit, Etc” - August 2011 Grzegorz Kochan | grzegorz@adtaily.com 5
- 6. Design for failure Grzegorz Kochan | grzegorz@adtaily.com 6
- 7. Amazon AWS Geographical map US - Oregon EU Ireland Asia Pacific: Tokyo Availability Zones: Availability Zones: Availability Zones: us-west-1a eu-west-1b ap-southeast-1a us-west-1b eu-west-1b ap-southeast-1b us-west-1c eu-west-1c US - N. California US - Virginia Asia Pacific: Availability Zones: Availability Zones: Singapore us-west-2a Availability Zones: us-east-1a us-west-2b ap-southeast-1a us-east-1b ap-southeast-1b us-east-1c us-east-1d Grzegorz Kochan | grzegorz@adtaily.com 7
- 8. Replicate, duplicate and balance Multi server Availability Zone Single server setup us-east-1a setup Grzegorz Kochan | grzegorz@adtaily.com 8
- 9. Distribute multi A-Zone architecture Availability Zone Availability Zone us-east-1a us-east-1b Grzegorz Kochan | grzegorz@adtaily.com 9
- 10. Distribute more multi Region architecture US East Region US West Region Grzegorz Kochan | grzegorz@adtaily.com 10
- 11. Design for failure application decoupling Stateless services Gracefull degradation Die fast and alone Auto recover Backup and scale independently Grzegorz Kochan | grzegorz@adtaily.com 11
- 12. Decoupling example Monitoring & Alerting Ecommerce (CloudWatch & SNS) Application AutoScaling - shopping cart - process orders - process payments Product Catalog #2 #3 - generate invoices Messaging (SQS) - send emails Order Processor #2 #3 Payment Processor #2 Invoice generator #2 Email Sender #2 ... #2 #3 #4 #5 Grzegorz Kochan | grzegorz@adtaily.com 12
- 13. Automate everything Infrastructure - custom AMIs, Chef, Puppet Monitoring - CloudWatch Scaling and recoverying - AutoScaling Fail and recover constantly - ChaosMonkey by NetFlix Grzegorz Kochan | grzegorz@adtaily.com 13
- 14. Be rational weight the risks and costs Grzegorz Kochan | grzegorz@adtaily.com 14
- 15. More info on Amazon AWS http://aws.amazon.com/architecture Grzegorz Kochan | grzegorz@adtaily.com 15
- 16. Questions? Grzegorz Kochan CTO & VP of Products w AdTaily email: grzegorz@adtaily.com www.adtaily.pl facebook.com/adtaily http://adtai.ly/TechCamp1 Grzegorz Kochan | grzegorz@adtaily.com 16