Building-resilient-serverless-workloads-Navigating-through-failure

Presentation around serverless reliability

1. NORDICS DGI Byen’s CPH Conference 2024

2. Building resilient serverless workloads: Navigating through failures JIMMY DAHLQVIST | 2024-05-07

3. Thank You!

4. JIMMY DAHLQVIST Serverless enthusiast Head of AWS @ Sigma Technology Cloud Founder of serverless- handbook.com Blogging on jimmydqv.com AWS Ambassador | AWS Community Builder | User Group Leader § Hello, I'm

5. Resources https://serverless-handbook.com Architecture patterns Solutions Workshops

6. Agenda What is serverless and resiliency Architecting resilient system – Good practices Summary

7. What is serverless? Automatic and flexible scaling No capacity planning High Availability Pay-for-use billing

8. Service categories

9. What is resiliency? The ability for a software solution to handle the impact of problems, and recover from turbulent conditions, when other parts in the system fails.

10. “Everything fails all the time Dr. Werner Vogels , CTO, Amazon.com

11. Understand AWS Services Everything has a limit Understand how services work under the hood

12. Resiliency testing Chaos Engineering Amazon Fault Injector Service Start in QA Don’t forget about data

13. Web application

14. Do we need an immediate response?

15. Storage-First

16. Storage-First Data-centric design Durability and availability Scalable System Design Asynchronous processing

17. Storage-First Things to consider Architectural complexity Eventual consistency Design for idempotency Risk of over-optimization

18. Queue Load Leveling

19. Queue load leveling System stability Handle unexpected spikes Protect downstream resources

20. Decoupling

21. Decoupling

22. Retries Selfish Exponential backoff Users can make it worse

23. DLQ

24. DLQ

25. Retries with backoff and jitter No Jitter With Jitter Image: Amazon Architecture blog (https:// tinyurl.com /y48t2v4h)

26. Circuit breaker

27. Circuit breaker Half Open

28. Circuit breaker Avoid cascading failures Protect system resources Risk of early circuit break Good observability required

29. Put it all together

30. Notification Service Payment Service

31. What we talked about Design for failure Buffer and store messages first Process asynchronously Level the load Retry on failures Break if integrations are not healthy

32. @jimmydahlqvist dahlqvistjimmy https://serverless- handbook.com https:// jimmydqv.com THANK YOU

33. @jimmydahlqvist dahlqvistjimmy https://serverless- handbook.com https:// jimmydqv.com