Java Profiler Skill

Overview

Professional Java profiling skill integrating industry-leading tools for JVM performance analysis, memory management, and thread debugging in enterprise environments.

Capabilities

1. JVM Performance Analysis

• YourKit Java Profiler: Deep profiling for distributed systems
• JProfiler: Advanced UI with real-time metrics
• VisualVM: Built-in monitoring and profiling
• Java Flight Recorder (JFR): Low-overhead production profiling

2. Memory & Heap Analysis

• Heap dump analysis
• Memory leak detection
• Garbage collection tuning
• Object allocation tracking

3. Thread & Concurrency

• Thread contention detection
• Deadlock identification
• CPU sampling per thread
• Wait time analysis

4. Database & I/O

• JDBC query profiling
• Connection pool monitoring
• File I/O performance
• Network latency tracking

Usage

YourKit Profiler

# Attach to running JVM
java -agentpath:/path/to/yjpagent.so MyApplication

# CPU profiling
# Memory allocation tracking
# Thread monitoring

VisualVM (Built-in)

# Launch VisualVM
jvisualvm

# Connect to local/remote JVM
# Monitor: CPU, Memory, Threads, Classes
# Profile: CPU sampling, Memory profiling

Java Flight Recorder

# Start JFR recording
java -XX:StartFlightRecording=duration=60s,filename=recording.jfr MyApp

# Analyze with JMC
jmc recording.jfr

JProfiler Integration

# Remote profiling
java -agentpath:/path/to/jprofilerti.so=port=8849 MyApplication

# Connect JProfiler GUI to port 8849

Integration Scripts

jvm_monitor.sh

Automated JVM health monitoring:

#!/bin/bash
# Monitor JVM performance metrics

PID=$(jps | grep MyApplication | awk '{print $1}')

echo "=== JVM Memory Usage ==="
jmap -heap $PID

echo "=== Thread Dump ==="
jstack $PID

echo "=== GC Statistics ==="
jstat -gcutil $PID 1000 5

heap_analyzer.sh

Heap dump analysis automation:

#!/bin/bash
# Generate and analyze heap dump

PID=$1
DUMP_FILE="heap_dump_$(date +%Y%m%d_%H%M%S).hprof"

jmap -dump:live,format=b,file=$DUMP_FILE $PID
echo "Heap dump created: $DUMP_FILE"

# Analyze with jhat or VisualVM
jhat $DUMP_FILE

Performance Tuning Guidelines

JVM Arguments for Profiling

-XX:+UseG1GC                          # Use G1 garbage collector
-XX:MaxGCPauseMillis=200              # Target max GC pause
-Xms2g -Xmx4g                         # Heap size (min/max)
-XX:+HeapDumpOnOutOfMemoryError       # Dump on OOM
-XX:HeapDumpPath=/tmp/heapdumps       # Dump location
-XX:+PrintGCDetails                   # GC logging
-XX:+PrintGCDateStamps                # Timestamp GC logs

Common Performance Issues

1. Memory Leaks

• Growing heap usage over time
• Increasing full GC frequency
• OutOfMemoryError exceptions

Detection: Monitor heap growth pattern, analyze retained objects

2. Thread Contention

• High CPU wait time
• Deadlocks and livelocks
• Poor scalability under load

Detection: Thread dumps, lock contention analysis

3. Database Performance

• N+1 query problems
• Missing connection pool configuration
• Unoptimized JDBC batch sizes

Detection: SQL logging, JDBC profiling

4. GC Overhead

• Long GC pause times
• Frequent full GCs
• High GC CPU usage

Detection: GC logs analysis, pause time distribution

Best Practices

1. Start with Low-Overhead Tools: JFR for production, VisualVM for development
2. Profile Realistic Workloads: Use production-like data volumes
3. Focus on Hot Spots: Optimize methods consuming >10% CPU time
4. Monitor GC Impact: Aim for <1% CPU time in GC
5. Analyze Thread Dumps: Take 3-5 dumps, 10 seconds apart
6. Use Sampling Mode initially, switch to instrumentation for deep analysis

Requirements

# Built-in tools (JDK 8+)
jvisualvm, jconsole, jmc, jstat, jmap, jstack, jhat

# Commercial tools (trial/licensed)
YourKit: https://www.yourkit.com/
JProfiler: https://www.ej-technologies.com/products/jprofiler/

Example Workflow

1. Baseline: Capture performance metrics under normal load
2. Profile: Run YourKit/JProfiler during load test
3. Analyze: Identify CPU hot spots and memory allocations
4. Optimize: Fix identified issues (algorithms, caching, pooling)
5. Verify: Re-profile to confirm improvements
6. Monitor: Set up continuous monitoring with JFR

Metrics to Track

• Throughput: Requests/second
• Latency: p50, p95, p99 response times
• Memory: Heap usage, allocation rate
• GC: Pause frequency and duration
• Threads: Active thread count, wait time
• CPU: Per-thread CPU time distribution