We Must Save the Last Giga of RAM
summary
A real incident response where a production Java system hit 96 GB RAM during year-end batch runs processing millions of letters. The team used Eclipse Memory Analyzer, JFR recordings, and pg_stat_statements to capture heap dumps and query stats, discovered 300 metadata attributes per record while only 12 were needed (saving ~1 GB), fixed N+1 queries with projected SELECTs and switched to streaming results with tuned fetchSize (saving ~3 GB), and removed unnecessary collections to reclaim a total of ~4 GB; the fixes were validated in production using streaming and closed projections.
AI-written summary · about
chapters (6)
- 0:00Introduction
- 14:14JVM analysis tools
- 31:15PostgreSQL optimization
- 41:31Streams and cache optimization
- 43:57Summary
- 44:39Q&A
related talks
39mMemory Efficient Java
21mMemory Efficient Java
51mA Healthy diet for your Java application
01h 32mTuning Large Scale Java Platforms
40mTrash Talk - Exploring the Memory Management in the JVM
53mPlugging Memory Leaks With The Chrome Dev Tools
51mIn-Memory Computing - The Big Picture
54mThe Dark Arts of Performance Tuning