How can you optimize the memory footprint of a native image executable?

A handful of build- and run-time levers reliably shrink both the executable's disk size and its runtime memory usage.

1. Trim reachable code: unused dependencies and overly broad reflection configs pull in classes the analysis then has to keep and compile, so removing them shrinks the binary directly.
2. Choose a lean GC: Epsilon GC (no collection at all, for very short-lived processes) or Serial GC use far less memory overhead than G1, at the cost of G1's better scalability for larger heaps.
3. Strip debug info for production builds (omit -g/keep it only for local debugging), since embedded debug symbols add meaningfully to binary size.
4. Tune heap flags at build time (-R:MaxHeapSize) and launch time (-Xmx) to cap growth rather than relying on defaults sized for a general-purpose workload.
5. Use Quick Build Mode's opposite - the fully optimized build (default, not -Ob) for production, since it also does more aggressive dead-code elimination than the quick-build path.

Measuring with -H:+PrintImageHeapPartitionSizes during the build shows exactly what's taking up space, which is far more reliable than guessing.