Building and Tuning High Performance Java Platforms
summary
Conference analysis claims half of Java apps run with 2x over‑provisioning while many still miss SLAs, and recommends concrete GC and sizing rules: set -Xms = -Xmx, favor 4GB is the new 1GB for most web apps, prefer fewer large JVMs for latency‑sensitive middleware (example: eight 90GB JVMs serving 1TB in‑memory with 20M trades/day) and a single 360GB JVM can be viable. Uses ParNew + CMS and tuned parallel GC thread counts, advises scale up before scaling out and NUMA‑aware placement.
AI-written summary · about
chapters (28)
- 0:00Introduction and background
- 3:58Identifying the core problem
- 5:46Broken application mapping
- 8:36The future of CIOs
- 10:15Platform engineering
- 13:24Cloud native reality
- 15:31Bridging platforms
- 18:43JVM sizing rules
- 22:41The 4GB JVM standard
- 25:55JVM memory breakdown
- 30:06Multi-tier tuning
- 32:26Load balancer strategies
- 36:45Workload categories
- 38:51Consolidating applications
- 45:01Memory databases
- 47:39Sizing versus throughput
- 50:38Platform architecture
- 52:51Golden rules summary
- 55:24Numa architecture
- 59:37Scaling and certification
- 1:02:03Memory vs CPU scaling
- 1:05:49Scaling Mythbusters
- 1:08:47Specialized JVM tuning
- 1:11:56Garbage collection tuning
- 1:15:51GC tuning recipe
- 1:18:07Tuning outcomes
- 1:24:15Microservices consolidation
- 1:28:45Future of JVM and Q&A
related talks
01h 32mTuning Large Scale Java Platforms
01h 29mVirtualizing and Tuning Large Scale Java Applications
47mSecrets of Performance Tuning Java on Kubernetes
51mJava 10/11 Memory Management and GC
01h 28mThe dark art of performance tuning or how to become a performance hero ...
48mJava Memory Management Best Practices
53mCalibrate Garbage Collection on the Ground and Run Your Java App
47mManaging and monitoring a JVM