Low Latency Java in the Real World: LMAX Exchange & the Zing JVM
summary
LMAX Exchange runs its matching stack in Java on the Zing JVM and reports concrete reductions in jitter: a highly tuned core went from 4 ms blips every ~30 s to ~1 ms blips every ~2 h, and a less-tuned path fell from ~50 ms blips every 30 s to ~3 ms blips far less often. Built components remain idiomatic Java (including the Disruptor), tooling used includes HDR Histogram and JHiccup, and lessons include proactive Linux tuning (disable transparent hugepages) and measure-everywhere monitoring.
AI-written summary · about
chapters (7)
- 0:00Start
- 12:42Architecture
- 19:56Low Latency Java
- 27:10Systems on the Latency Path
- 39:32Page Cache Lock Contention
- 44:37Lessons Learned
- 50:10Summary and Wrap Up
related talks
35mJava at Speed: Building a Better JVM
01h 01mJava compilers: Why Java is one of the first options for Low latency apps
44mJava at Speed: Building a Better JVM.
52mAchieving Sub‑Millisecond Response Times in a Java Trading System on Linux
38mWelcome to the Jungle - A safari through the JVM landscape
41mBuilding Low Latency Microservices & Monoliths in Java
48mZGC: A Decade of Redefining Java Performance
47mLegal JVM Doping in Your Containers