NOTE
Linux Memory Performance Tuning
Diagnosing resident memory, reclaim, page cache, swap, faults, NUMA, and application allocation behavior.
This is a historical learning note and may contain outdated or incomplete understanding.
Memory tuning starts by separating useful page cache from anonymous/application memory and by checking whether there is actual reclaim pressure. Observe RSS, working set, page faults, reclaim, swap activity, cgroup pressure, and OOM events.
Application allocators and runtimes may retain memory after logical objects are freed, so process RSS can remain high without a leak. Profile object retention/allocations before changing kernel reclaim or runtime GC settings.
NUMA locality and huge pages can matter for large specialized workloads, but require measurement.