Huge pages
Memory pages that are much larger than the standard 4 KB size — typically 2 MB or 1 GB. Fewer, larger pages reduce pressure on the TLB (the address-translation cache) and give better performance for memory-heavy workloads like databases and virtual machines.
Problem: the CPU caches virtual-to-physical address translation in the TLB; with 4 KB pages, huge amounts of memory require an enormous number of entries → TLB misses that cost performance. Huge pages map much more memory per entry → fewer misses, smaller page table. Use case: Postgres/Oracle/JVM with large heaps, KVM guests, HPC. Linux has explicit hugepages (pre-allocated) and Transparent Huge Pages (THP, automatic) — where THP ironically sometimes hurts databases (latency spikes) and is recommended off for e.g. Postgres/Redis. Trade-off: reduces fragmentation flexibility. A classic tuning knob for memory-heavy servers. Related to NUMA and swap.