Lessons Learned : Improving Performance and Resilience in Blibli.com Store Front

Slides from our talk about our journey in increasing performance using completable future, guava, and rx java at JVM Meetup #9 & Tech It Easy @ Blibli.com

Download the PDF for clickable links!

Transcript

1. Lesson Learned: Improving Performance and Resilience in Blibli.com Store Front By Frandy, Ardi, Nuridita, Bryan
2. Disclaimer Presentations are intended for educational purposes only and not to replace independent professional judgment. Statements of facts, views, opinions expressed do not necessarily reflect those of Blibli.com
3. Outline • Introduction • Performance Methodology • Implement CompletableFuture Java 8 • Implement Guava Cache • Implement RxJava • Result
4. Introduction Architecture Diagram
5. Introduction Performance Methodology 1. Monitoring (new relic, logs, etc) 2. Profiling (Xrebel) 3. Tuning
6. Monitoring New Relic
7. Monitoring Splunk
8. Profiling • Using XRebel
9. Profiling using XRebel • How we do – Narrow Profiling to specific end point that we want to improve – Identify which services that blocking / problematic • Possible Solution : – Move some services that high load to ajax – Make some services to asynchronous – Caching some services that high load – Make thread separated from tomcat
10. Multiple Services Call Problem Problem : • Multiple information taken from multiple services • Each service can take some time to give response Solution : • Make those service call (independent from other services) asynchronous using CompletableFuture
11. Multiple Services Call Problem
12. Completable Future • CompletableFutureis used to run tasks asynchronously • Implement Future and CompletionStage interface
13. Asynchronous Process Source: memory-alpha.wikia.com Case
14. Asynchronous Process Imagine that the captain is alone flying the ship, shooting, and evading enemies.
15. Asynchronous Process Source: www.rd.com and www.lacombecitycinemas.ca Case
16. Asynchronous Process Imagine that we had four employees checking tickets for one studio.
17. Completable Future A Future represents the result of an asynchronous computation. - Java documentation Future
18. Completable Future A stage of a possibly asynchronous computation, that performs an action or computes a value when another CompletionStage completes. - Java documentation Completion Stage
19. Completable Future A Future that may be explicitly completed (setting its value and status), and may be used as a CompletionStage, supporting dependent functions and actions that trigger upon its completion. - Java documentation Completable Future
20. Completable Future
21. High Load Problem to Backend Service Problem : • Store front serve content mainly from service • High response time during high concurrent request to backend services (cms, etc) • Mostly same content Solution : • Cache it using Guava!
22. Guava Cache Guava cache is a part of google guava core library and it provide an in memory cache. Why Guava cache ? - It provide us interface to populate cache and its behavior - We are willingly to spend our memory to improve the speed - We have many keys that called repetitively
23. Guava Cache • Pros – Improve speed – No need to execute same process with exact parameters – No need for network connection – Can handle huge content – Can handle complex cache key • Cons – Consume RAM – Local Cache, duplicate between nodes
24. Guava Cache Refresh • Different from eviction • Refreshing a key loads a new value for the key, possibly asynchronously • The old value (if any) is still returned while the key is being refreshed • If an exception is thrown while refreshing, the old value is kept, and the exception is logged and swallowed
25. Guava Cache Refresh • Pros – Return old value immediately and refresh a new value asynchronously – Keep the old value when refreshing fails • Cons – Not getting the newest value even the time expires (need to wait the refresh to successfully retrieve new value) – Value will always remain on RAM
26. Guava Cache Refresh • Code
27. Guava Cache Refresh • Code
28. Result - Our content API and search API are protected from high request - Blibli.com store front application resilience is improved. - Same repeated request doesn’t need to call the services
29. RxJava Implementation Problem : • High load + Slow Response = Out of Thread Solution : • Wrap the request using Rx
30. RxJava Implementation • Normal Code (sample)
31. RxJava Implementation • Using RxJava Code (sample)
32. RxJava Implementation • Using RxJava Code (sample)
33. RxJava Implementation • Using RxJava Code (sample)
34. RxJava Implementation • Using RxJava Code (sample)
35. RxJava Implementation
36. RxJava Implementation • Using RxJava Code (sample)
37. RxJava Implementation • Using RxJava Code (sample)
38. RxJava Implementation
39. Result • Can held 300-400% normal traffic • No Thread Exhaustion
40. Published Paper • Conference RAIEIC 2017 at Nov 21st • Lessons Learned In Applying Reactive Systems
41. Thank You