Memory Efficient Java
summary
Claims that 70% of applications are memory‑bottlenecked and that allocation rates above >1 GB/s put you in a red zone that yields the largest wins. Shows concrete wins from reducing allocation pressure — 10–20% typical, up to 40% in some Scala workloads — and explains why: escape analysis, scalar replacement, TLAB behavior and promotion cause CPU stalls and tail latency. Demos use GC logs, VisualVM, Flight Recorder and JITWatch to find hot allocation sites and to verify fixes; off‑heap caching is warned to sometimes worsen churn.
AI-written summary · about
chapters (11)
- 0:00Introduction to memory issues
- 3:14Memory bottlenecks in Java apps
- 6:23Escape analysis and optimizations
- 8:53Heap structure and TLABs
- 13:29High allocation rate problems
- 17:42Live data set size impacts
- 21:15Escape analysis deep dive
- 23:23Demo setup delay
- 27:19Profiling allocation hotspots
- 31:34Optimizing index allocations
- 34:32Q&A and closing remarks
related talks
38mMemory Efficient Java
21mMemory Efficient Java
36mMemory Why You Need to Care
46mHow difficult can it be to write efficient code?
35mJava *is* Memory Efficient
40mLudicrous Speed: Designing For Performance on the JVM
55mThe Performance Engineer's Guide To Java HotSpot
19mA Multifaceted Memory Analysis of Java Benchmarks