java-performance

Optimize JVM performance through profiling, GC tuning, and memory analysis.

Overview

This skill covers JVM performance optimization including garbage collection tuning, memory analysis, CPU profiling, and benchmarking with JMH.

When to Use This Skill

Use when you need to:

• Tune GC for low latency or throughput
• Profile CPU hotspots
• Analyze memory leaks
• Benchmark code performance
• Optimize container settings

Quick Reference

GC Presets

# High-throughput
-XX:+UseG1GC
-XX:MaxGCPauseMillis=200
-Xms4g -Xmx4g
-XX:+AlwaysPreTouch

# Low-latency
-XX:+UseZGC
-XX:+ZGenerational
-Xms8g -Xmx8g

# Memory-constrained
-XX:+UseSerialGC
-Xms512m -Xmx512m
-XX:+UseCompressedOops

# Container-optimized
-XX:+UseContainerSupport
-XX:MaxRAMPercentage=75.0
-XX:+ExitOnOutOfMemoryError

Profiling Commands

# Thread dump
jstack -l <pid> > threaddump.txt

# Heap dump
jmap -dump:format=b,file=heap.hprof <pid>

# GC analysis
jstat -gcutil <pid> 1000 10

# Flight recording
jcmd <pid> JFR.start duration=60s filename=app.jfr

# Async profiler
./profiler.sh -d 30 -f profile.html <pid>

JMH Benchmark

@BenchmarkMode(Mode.Throughput)
@Warmup(iterations = 3, time = 1)
@Measurement(iterations = 5, time = 1)
@State(Scope.Benchmark)
public class MyBenchmark {

    @Benchmark
    public void testMethod(Blackhole bh) {
        bh.consume(compute());
    }
}

GC Comparison

| GC | Latency | Throughput | Heap Size | |----|---------|------------|-----------| | G1 | Medium | High | 4-32GB | | ZGC | Very Low | Medium | 8GB-16TB | | Shenandoah | Very Low | Medium | 8GB+ | | Parallel | High | Very High | Any |

Troubleshooting

| Problem | Cause | Solution | |---------|-------|----------| | GC thrashing | Heap too small | Increase heap | | High latency | GC pauses | Switch to ZGC | | Memory leak | Object retention | Heap dump + MAT | | CPU spikes | Hot loops | Profile + optimize |

Usage

Skill("java-performance")