Du réactif au service du pneu connecté - Thomas Segismont / Fabien Pomerol

Given @ SnowCamp 2020

Transcript

1. Du réactif au service du pneu connecté Thomas Segismont et Fabien Pomerol SnowCamp 2020
2. Fabien Pomerol Thomas Segismont @tsegismont @FabienPomerol
7. SENSORS
8. ON-BOARD ROUTER MANUAL DEVICES + SMARTPHONE
9. Building an API to : Collect devices data Process it Generate Insight
10. Services Use case? Tyre end of life prediction Security (low pressure detection) Fleet Management
11. 100 Million Assets > 4 Billion Device Events > 1K Req / sec Low resources footprint
12. Original Architecture Monolithic REST API Event Producer Web / Mobile Applications
13. Split Monolith into Microservices Rewrite using &
14. Ingestion Service Micro Services !
15. Ingestion Service Fleet Management Micro Services !
16. Ingestion Service Tyre Maintenance Micro Services ! Fleet Management
17. Tyre Maintenance Edge Service Ingestion Service Micro Services ! Fleet Management
18. Insights Service Ingestion Service Tyre Maintenance Edge Service Micro Services ! Fleet Management
19. Ingestion Service Alerting Tyre Maintenance Edge Service Insights Service Micro Services ! Fleet Management
20. Micro Services ! Ingestion Service Tyre Maintenance Edge Service Alerting Insights Service Fleet Management
21. Vert.x = ?
22. Blocking I/O don’t scale! Thread request #1 Thread request #2
23. Resources are scarce! “Cloud native” Virtual machines Containers
24. Event-loop thread Async I/O do scale! Async events
25. Reactive Systems Elasticity Resiliency Responsiveness
26. Reactive Asynchronous Event-driven Toolkit Fast Polyglot Versatile
27. Ingestion Flow Ingestions Service Event Producer TCP Socket HTTP:/ / MQTT
28. { demo }
29. Service Aggregation Web / Mobile Applications Edge Service Tyre Management Insights Service http:/ / http:/ / Composition
30. Callback Future CompletionStage Quasar RxJava / RxKotlin Kotlin coroutines Event
31. { demo }
32. Event Driven Microservices Ingestion Service Insights Service Alerting Service Events splits by topic Consume Pressure Events Publish Computed Insights Consume Insights
33. { demo }
34. Embracing Async Programming was not so easy for the team Vert.x is unopinionated: easy to be tempted to built Framework / Cathedral Vert.x the hard way!
35. Keep the Vert.x EventBus usage inside microservices No durability No back pressure
37. K8S Pod Sizing Distributed Locks Exactly Once Idempotence Distributed Cache Tracing A NEW WORLD OF PROBLEMS Data Denormalization Concurrency Control
38. You can’t survive without monitoring and tracing !
39. Thank you! Thomas Segismont et Fabien Pomerol SnowCamp 2020 Icons made by Freepik, Smashicons from www.flaticon.com