Akkurate Akka
Download as PPTX, PDF1 like117 views
The document discusses the actor model in Akka, highlighting its features for building highly concurrent, scalable, and fault-tolerant applications in Java and Scala. It emphasizes best practices, such as automatic load balancing and supervisor hierarchies for error management. The importance of adhering to the single responsibility principle and using immutable messages to maintain a responsive system is also mentioned.
Akkurate Akka
- 2. outline Problems to solve Actor Model Java & Scala actors Fault-tolerance Scaling, Remoting Reactiveness Some best practices
- 3. The Problem • highly concurrent • scalable • fault-tolerant
- 4. AKKA
- 6. Program at a higher level
- 7. Scaling up & out & down & .. whatever
- 8. Automatic Replication & Distribution for fault tolerance & scalability
- 10. Modules • akka-kernel • akka-actor • akka-camel • akka-remote • akka-cluster • akka-persistence • akka-stream • …
- 11. Actor Model Back into the 70’s actors could • perform • create another actors • send messages • become Carl Hewitt
- 12. Actor • one msg at a time • in one thread • msgs in order • small (~2.5M per Gb)
- 13. Core operations 0. Define 1. Create 2. Send 3. Become 4. Supervise
- 14. 0. Define
- 15. In Scala
- 16. 1. Create
- 17. Actor hierarchy
- 18. 2. Send
- 19. Full example
- 20. In Scala
- 21. Hot Swap
- 22. In Scala
- 23. ask
- 24. Futures
- 25. Supervision & Monitoring • failures happen • application should self-heal • each actor have supervisor • supervisor handles failures – RESUME – RESTART – STOP – ESCALATE
- 26. Failure management money snack not enough!FATAL ERROR FATAL ERROR supervisor
- 27. Supervisor hierarchies • One-for-one strategy • All-for-one strategy
- 28. Root actors
- 29. 4. Supervise
- 30. Death Watch
- 31. Failure management in Java
- 32. Load Balancing • scaling up – routers • scaling out – akka-cluster
- 33. Routers
- 34. From config
- 35. Routers • Random • Round-robin • Smallest messagebox • Broadcast • ScatterGatherFirstCompletedOf • …
- 37. Lookup
- 40. Actors should do only one thing • Single responsibility principle • Create specific supervisors • Keep the error kernel simple
- 41. Avoid blocking
- 42. Don’t expose actors • Avoid using this • Don’t close over variables • Use immutable messages with immutable data
- 43. E0 F