Azure Functions - the evolution of microservices platform or marketing gibberish?
- 1. Azure Functions - the evolution of microservices platform or marketing gibberish? Anton Boyko Microsoft Azure MVP me@boykoant.pro
- 2. Serverless? Population of homeless servers is growing
- 3. They asked me for NoSQL, so I dropped the database. No database, NoSQL!
- 4. Benefits of serverless Manage less Micro-pricing Ease of scale (up & down)
- 5. Reserved Agent Job 1 Job 2 Job 3 Serverless Job 1 Job 2 Job 3 Agent 1 Agent 2 • Fixed CPU • Fixed RAM • On-demand CPU • On-demand RAM
- 6. Serverless offerings Microsoft Azure • Azure Functions • Azure Logic Apps • Azure Container Instances AWS • AWS Lambda • AWS Fargate Google cloud • Cloud Functions • Cloud Run
- 7. Azure Functions Event-driven on-demand custom code execution.
- 8. I have a “code” I have an “event data”
- 9. I have a “code” I have an “event data” BOOM!
- 10. C# F# Java NodeJS Python PowerShell … HTTP(S) Timer Queue Storage … Azure Functions Stateless Small Single unit of work
- 12. Impediments Why aren’t we there yet?
- 13. Capacity What will it cost to run a serverless app for 10K users?
- 14. Capacity planning • Q: What will it cost to run a serverless app for 10K users? • A: How will you calculate what it will cost to run a non-serverless app? • A: What will it cost to wash all the windows in Seattle?
- 15. Capacity calculation example Facts • Each user spent 5 minutes on average on your website daily. • Each user reads 10 articles daily. • Each user generates 15 requests daily (10 articles + 5 utility requests). • Each request takes 1 second on average. Conclusions • Each user consumes 1 sec per request. • Each user consumes 15 secs per day. • Each user consumes 450 secs per month. • 10K users consumes 4.5M secs per month.
- 16. DurabilityHow to orchestrate workflows? Sagas?
- 17. Durability – fan out / fan in
- 19. Consumption example – fan out / fan in 1 sec 3 sec 3 sec 3 sec 2 sec 6 sec Orchestrate Fan out Fan in Response
- 20. Durability – long running async HTTP API
- 23. Durability – human interaction
- 25. Performance How to get predictable and consistent performance level?
- 26. Performance Consumption • 1 CPU core • Up to 1.5 Gb RAM • Cold start • Scale down to 0 instances Premium • Up to 4 CPU cores • Up to 14 Gb RAM • Pre-warmed instance • Scale down to 1 instance Azure Function Plan switcher https://aka.ms/premium-plan-sample Understanding serverless cold start https://azure.microsoft.com/en-us/blog/understanding-serverless-cold-start/
- 28. Performance – cold start
- 29. Performance – warm start
- 30. KEDA
- 32. DEMO Azure Functions in Docker
- 33. Conclusion
- 35. Questioning the question mark Anton Boyko Microsoft Azure MVP me@boykoant.pro
- 36. Appendix 1 – code examples https://github.com/TeamSpark-Learning/azure-functions