Speed JVM Performance with JWarmUp
summary
Describes Alibaba’s unified deployment approach for two Java application classes: large-scale eCommerce apps that reuse profiling data via a new compilation technology called JWarmUp, and smaller multi-tenant apps accelerated with an AppAOT (AOT variant) that packs many apps into one JVM. Explains the engineering choices to adapt these techniques to different scenarios and reports measured reductions in startup JIT CPU load (from 100% to 20%) and startup time (from 40s to 25s).
AI-written summary · about
chapters (5)
- 0:00Start
- 4:25JVM Warmup - Mock Requests
- 26:42Class Resolution - Class Init Chain
- 29:17Class resolution - multiple loaders
- 31:08Eager Class Resolution
related talks
44mOptimizing JVM at Alibaba, for e-commerce apps running on 100,000+ servers
56mKeeping Your #Java Hot by Solving the #JVM Warmup Problem
54mKeeping Your Java Hot by Solving the JVM Warmup Problem
46mKeeping Your Java Hot by Solving the JVM Warmup Problem
27mHow to Eliminate Java Performance Warmup
50mAOT or JIT: Faster Startup or Faster Code?
50mBinary Java: CDS, CRaC and AOT for faster JVM startup and warmup
45mWarming up the JVM: CRaC and other tricks