Cloud Native JVM. Cloud Compiler
summary
Proposal: move the heavyweight JIT out of each container into a shared JIT Cloud Compilation Service that compiles hot methods centrally and returns binaries matching local guards and profiles. This lowers per-JVM native CPU and memory during warmup, reduces container sizes, and enables reuse of speculative optimizations across nodes. Prototype work uses OpenJDK Graal MCI with Azul Prime / Zulu and shows 30-50% faster code in experiments; IBM JIT Server is a similar precedent. Network latency and bandwidth are minor compared with compilation cost.
AI-written summary · about
chapters (6)
- 0:00Introduction
- 6:21Cloud Native JVM philosophy
- 15:57How JIT compilers work
- 20:09Cost of heavy optimizations
- 39:55Cloud Native compilation
- 48:44Q&A
related talks
45mCloud Native JVM. Cloud Runtime
41mThe Cloud Native Compiler: JIT-as-a-Service
47mCloud Native JVMs, parts I and II: Fleet Profiling and Optimization
24mJITServer: Disaggregated Caching JIT Compiler for the JVM in the Cloud
29mThe Next Frontier in Open Source Java Compilers: Just-in-Time Compilation as a Service
01h 03mKeeping Your Java Hot.
33mCloud Native Java with OpenJ9: Fast, lean and definitely mean
50mAOT or JIT: Faster Startup or Faster Code?