How does garbage collection work in the JVM?

Garbage collection starts from roots such as thread stacks, static fields, JNI references and active class loaders. Anything reachable from those roots is live. Everything else can be reclaimed. Reference counting alone cannot handle cycles, so the JVM uses tracing.

Generational collection

• The young generation holds new objects in Eden plus two survivor spaces
• Minor collections copy live young objects between survivor spaces and promote long-lived ones
• The old generation holds promoted objects and is collected less frequently
• Most objects die young, so collecting the young generation frequently is efficient

G1 is the default collector on modern JDKs. It divides the heap into regions and targets a pause-time goal. ZGC and Shenandoah pursue very low pauses by doing most work concurrently. Serial and Parallel are still useful for small heaps or throughput-focused batch jobs.

List<byte[]> cache = new ArrayList<>();
for (int i = 0; i < 10_000; i++) {
    cache.add(new byte[1024]);
}
cache.clear();

Tuning starts with measurement, not guesswork. Enable GC logging, inspect pause distribution and allocation rate, then adjust heap size, collector choice or object lifetime. A larger heap can reduce collection frequency but increases worst-case pause and memory cost.