Java at Speed: Building a Better JVM.
summary
Zing JVM replaces hotspot collectors with the C4 garbage collector (continuous concurrent compacting) to remove stop the world compactions and report near-zero GC pauses even on multi‑TB heaps up to 20 TB, uses a Falcon JIT (LLVM, AVX2) backend to generate wide vectorized code for big speedups, and records/rehydrates profiles with ReadyNow! so restarts reach ~98% previous performance within a few transactions; compile stashing cuts JIT replay time by up to 80% and jHiccup latency traces show milliseconds vs seconds for comparable workloads.
AI-written summary · about
chapters (24)
- 0:00Intro
- 1:52JVM Performance Graph: Ideal
- 2:17JVM Performance Graph: Reality
- 6:32Big JVM Challenges
- 8:22Azul Zing JVM
- 10:43Zing System Tools
- 14:45Loaded Value Barrier
- 17:29Concurrent Mark Phase
- 20:13Relocation Phase
- 21:30Quick Release
- 22:51Remapping Phase
- 24:15Zing: Big Heaps, No Problem
- 25:17Non-Zing GC Tuning Options
- 25:41GC Tuning Used To Be Hard
- 27:49Measuring Platform Performance
- 29:31Big Heap, Small Latency
- 31:49Small Heap, Small Latency
- 32:32Adaptive Compilation Challenges
- 33:02Advancing Adaptive Compilation
- 35:14More Complex Code Example
- 35:42Traditional JVM JIT
- 36:27Falcon JIT
- 37:34ReadyNow! Warmup Elimination
- 41:35The Zing JVM
related talks
35mJava at Speed: Building a Better JVM
01h 32mUnderstanding Java Garbage Collection and what you can do about it
51mJava at Speed: Getting the Most Out of Modern Hardware
48mZGC: A Decade of Redefining Java Performance
11mJVM internals — bytecode, JIT, GC, speculative optimization
21mFaster Data With Faster Java
48mJava 9 Performance
01h 03mKeeping Your Java Hot.