Java 9: The Quest for Very Large Heaps
summary
ActiveViam and Oracle co‑engineer JVM changes to run terabyte heaps, validating a 16 TB prototype credit‑risk app on Oracle M7. G1 GC gains focus on young collections and mixed/mark phases: large-array references are chunked to avoid buffer flooding, per‑thread bitmaps become a shared lock‑free bitmap using CAS, and shared queues go lock‑free. Resulting pauses drop dramatically—5–20× shorter young pauses and marking/mixed work up to an order of magnitude faster—allowing interactive read/write workloads on very large heaps.
AI-written summary · about
chapters (20)
- 0:00Intro
- 1:14Background
- 4:48Data in Memory
- 10:03Why Java
- 12:50Data structures
- 13:03NUMA
- 16:08Perspective
- 16:42GC Strategies
- 17:05G1 Architecture
- 17:49G1 GC Overview
- 18:07Young Generation
- 21:23Mixed Collection
- 22:50Concurrent Marking
- 25:52G1 Scale
- 29:09Threads getting flooded
- 30:45Benchmark
- 32:33Improved Marking Threads
- 35:00Improved Work Balance
- 35:54Conclusion
- 43:50Parallel Application
related talks
52m-XX:+UseG1GC
48mZGC: A Decade of Redefining Java Performance
01h 14mHave you really taken the time to know me: A G1 GC saga
44mJava at Speed: Building a Better JVM.
51mJava 10/11 Memory Management and GC
48mJava 9 Performance
35mJava at Speed: Building a Better JVM
49mTrash Talk: Exploring the JVM memory management