Performance Optimization

Introduction

Measure, don't guess. Java performance work without a profiler is folklore. Once you measure, the wins are usually in algorithms, allocation rate, and I/O — rarely in micro-tweaks.

Understanding the topic

The 80/20 of JVM perf wins:

• Right-size heap; pick a GC (ZGC for latency).
• Eliminate N+1 DB queries — most 'slow Java' is actually slow SQL.
• Use connection pools (HikariCP); set sane timeouts everywhere.
• Cache hot reads with Caffeine (size + TTL bounds, not unbounded).
• Avoid auto-boxing in tight loops (prefer int[] over List<Integer>).
• Use virtual threads for blocking I/O instead of growing thread pools.

Real-world use

War story: a checkout API hit 99p latency of 1.2s. JFR showed 70% of time in String.format inside the logger — every request logged a debug line at INFO level. Lowering the log level cut p99 to 80ms. Nothing about the business logic changed.

Best practices

• Profile in production with JFR — overhead is < 1%.
• Use async-profiler for CPU flame graphs.
• Benchmark with JMH, not System.nanoTime in a loop.

Purpose of this lesson

Master Performance Optimization so you can apply it confidently in production Java code, technical interviews, and code reviews.

Step-by-step explanation

1. Understand the core idea behind Performance Optimization.
2. Walk through the runnable example and tweak it in the playground.
3. Apply the pattern in a small Spring Boot or CLI exercise of your own.
4. Re-read the common mistakes and interview Q&A to lock the concept in.

Interactive workflow diagram

Debugging tips

• Read the full stack trace — Java's exception messages name the offending class and line.
• Reproduce in the smallest possible main() method before fixing in the real app.
• Use IntelliJ's debugger breakpoints and 'Evaluate Expression' rather than scattering System.out.

Optimization strategies

• Measure with JMH for micro-benchmarks — never trust System.currentTimeMillis() loops.
• Reduce allocation: reuse buffers, prefer primitives, watch boxing in hot paths.
• Cache wisely: Caffeine for in-process, Redis for shared.

Enterprise example

Teams at Netflix, Uber and Goldman Sachs apply Performance Optimization daily — usually wrapped behind Spring Boot services with observability hooks (Micrometer + OpenTelemetry).

Interview questions & answers

Summary

In this lesson you learned Performance Optimization — the concept, syntax, a runnable example, and the production pitfalls to avoid. Apply it in the playground before moving on.