Scaling notifications; notification-service V2

Scaling notifications; a look at how we evaluated technologies and moved from a low-performant python implementation to a scalable backend.

Transcript

1. Agenda • Introduction • Architecture • Ways of working • Benchmark • Integrations • Learnings
2. Introduction • Problem statement • Proposed solution • Project timelines • Objectives
3. Architecture • Language decision • Framework decision • Database decision • Application structure • Resiliency/Fault-tolerance • Caching • GraphQL
4. Language Decision
6. Framework Decision
8. Database Decision
10. Application Structure
11. Application Structure
12. Fault Tolerance
13. Caching
15. Ways of Working • Pair programming • TDD • Offsite sessions
16. Pair Programming
17. TDD
18. Benchmark • Why? • Approach • Results and Observations • Assumptions • Improvements
19. graphQL inbox 10/10s
20. graphQL inbox 50/10s
21. graphQL inbox 100/10s
22. graphQL inbox 200/10s
23. graphQL inbox 400/10s
24. POST inbox 10/10s
25. POST inbox 100/100s
26. POST inbox 200/20s
27. POST inbox 400/10s
28. Get Config 10/10s
29. Get Config 100/10s
30. Get Config 200/10s
31. Get Config 500/10s
32. Post Config 50/10s
33. Post Config 100/10s
34. Post Config 200/10s
35. Post Config 500/10s
36. V1 vs V2 vs Sub ~40ms response time on average
37. Integrations • Support HTTP • API Documentation • Roadmap
38. Learnings • Ease of learning Kotlin • Transferred knowledge • Pair programming worked really well • Challenges
39. References • Notification Service Rearchitecture • API Documentation • Ktor • Kotlin • CockroachDB • Vegeta benchmark • Notification Service Repository