Garbage Collection in Java: The Performance Benefits of Upgrading
summary
Upgrading past JDK8 yields concrete wins: SPECjbb2015 shows parallel and G1 throughput up ~20–30% and latency scores improve more than 50% versus JDK8; normalized P99 GC pauses drop ~40% (parallel) and even more for G1. ZGC with generational mode (added in JDK21, default generational in JDK24) reports pauses in the hundreds of microseconds and large throughput gains; ZGC uses more native metadata but lets many apps run with a smaller heap. Netflix halves pause time; LinkedIn cuts P99.9 by ~50%.
AI-written summary · about
chapters (19)
- 0:00Garbage collection in Java
- 1:08GC primer and heap management
- 5:28Choosing right GC
- 7:31Java GC overview
- 8:12Serial vs parallel collectors
- 10:21G1 GC features and benefits
- 12:27ZGC low-latency explained
- 14:36GC performance profiles
- 15:40Upgrading past JDK 8 benefits
- 17:48Benchmarking GC improvements
- 26:16Cassandra workload performance
- 29:45Netflix JDK upgrade success
- 32:19Mercado Libre JDK 21 gains
- 34:26LinkedIn generational ZGC
- 35:30Oracle Cloud storage gains
- 36:57Future G1 and ZGC improvements
- 37:41GC guidance resources
- 38:42Key takeaways
- 39:23Q&A on small heap and ZGC
related talks
50mGarbage Collection in Java: The progress since JDK 8
25mGarbage Collection Primer
48mZGC: A Decade of Redefining Java Performance
51mJava Performance Update: From JDK 21 to JDK 25
45mJava Garbage Collectors’ Current Performance Impact
01mThe evolution of Garbage Collection Algorithms in Java
53mCalibrate Garbage Collection on the Ground and Run Your Java App
01h 01mOpenJDK: In the new age of Concurrent Garbage Collectors