Serverless is dead, long live to Serverless!

Some time ago a piece of news shocked the entire IT world: the Prime Video team abandoned the serverless microservices architecture to embrace the good old monolith. Many commentators quickly shared the news, issuing their verdict: serverless is dead!

But... is it really like that? In reality, things are a little bit different and if we can read the original post without any bias, we could even say that the whole story is excellent news for the serverless world and a great example of evolutionary architecture.

Together, we will delve into its essence and discover why it could be a rebirth rather than the final act.

Transcript

5. GAME OVER?
6. What are microservices?
7. Class 1 Class 2 Your 1st project’s scope Class 3
8. Benefits All in one Easy to develop Simple debugging Simple test E2E Fast deploy
9. Drawbacks (on big project) Slower dev. speed Global scaling Cascading failure Tech. stack blocked Difficult to evolve
10. Microservices Service1 project’s scope Class A Class B Class C Service2 project’s scope Class D Class E Class F
11. Microservices Service1 Service2
12. Benefits Agility (CI/CD, DevOps) Reliable Flexibility (scaling, stack,…) Better team collaboration Easy to evolve
13. Are microservices the best choice for your next project?
14. Are microservices the best choice for your next project? IT DEPENDS
15. What is serverless?
16. Serverless computing is a cloud computing execution model in which the cloud provider allocates machine resources on demand, taking care of the servers on behalf of their customers. High availability Elasticity Pay-as-you-go Fault tolerant
17. Is serverless the best choice for your next project?
18. Is serverless the best choice for your next project? IT DEPENDS
19. It always depends!
20. Bid producer Bid capture Bid tracking Bid analytics
21. Bid producer Bid capture Bid tracking Bid analytics Topic (pub/sub) Extensibility, Decoupling Data access, homogeneous contracts
22. Bid producer Bid capture Bid tracking Bid analytics Queues (point-to-point) Data persistence, heterogeneous contracts Poor exstensibility, low decoupling
23. “There are no right or wrong answers in architecture only trade-offs.”
24. But let's get back to the blog post…
25. «Prime Video set up a tool to monitor every stream viewed by customers.»
26. «Our service consists of three major components…» Media converter Orchestrator Defect detector
27. How micro is enough? One week replacement Two pizza’s team Low coupling High cohesion
28. «the degree of interdependence between modules» Coupling
29. Many forms of coupling Code Temporal Implementation
30. «the degree of interdependence between modules» High (bad) Coupling Low (good)
31. Cohesion «the degree to which the elements inside a module belong together» Low cohesion for functionality
32. Cohesion «the degree to which the elements inside a module belong together» High cohesion for functionality
33. «We realized that distributed approach wasn’t bringing a lot of benefits in our specific use case, so we packed all of the components into a single process» Orchestrator Media converter Defect detector Monitoring service
34. From three microservices to one
35. When the context changes, good architectures evolves!
36. «We designed our initial solution as a distributed system using serverless components, which was a good choice for building the service quickly.»
37. Business (needs) as usual
38. «However, the way we used some components caused us to hit a hard scaling limit […]»
39. Are you Prime Video?
40. «Our service performed multiple state transitions [and] AWS Step Functions charges users per state transition.»
41. Do the math!
42. «[…] the overall cost of all the building blocks was too high to accept the solution at a large scale.»
43. Do not be scared by break-even* (*) Unless it's necessary!
44. Stop talking about Lockin, call it switching cost!
45. Take away Don’t fall in love With your solutions Start simple Evolve Think broader
46. I am AWS specialist in VAR Group Mariano Calandra Work with us Blog: marianocalandra.medium.com LinkedIn: www.linkedin.com/in/marianocalandra
47. THANK YOU
48. Q&A
49. Further reading • Scaling up the Prime Video audio/video monitoring service and reducing costs by 90% – AWS Blog • Is serverless overpriced? What can we learn from the PrimeVideo team? – Yan Cui • Even Amazon can't make sense of serverless or microservices – DH Hansson • Monoliths are not dinosaurs – W. Wogels • Foundamentals of Software Architectures – M. Richards & N. Ford • Building Microservices – S. Newman