zswap

Zswap is a Linux kernel feature that provides a compressed swap cache in RAM. It sits between the memory manager and the swap subsystem, intercepting writes to swap and storing the resulting pages in a RAM-based pool in compressed form. The intended benefit is to reduce I/O to slow swap devices and to speed up swap-in and swap-out on systems with limited RAM or slow storage.

Operation: When a page is evicted from RAM to swap, zswap compresses the page and stores it in the in-memory pool. The pool has a configurable size; when full, the least-recently-used entries are evicted to the underlying swap device. On a swap-in, the kernel first checks the zswap pool; if the page is present, it is decompressed and returned. If not, the page is read from the backing swap device as usual.

Configuration and behavior: Zswap is tunable at runtime via sysfs, allowing enabling or disabling and selecting the compression algorithm (for example lz4, lzo, or zstd) and adjusting the pool size. Because the pool uses RAM, enabling zswap reduces available memory for other processes. The effectiveness depends on workload, compression ratio, and CPU overhead for compression.

Performance and limitations: In systems with slow storage or high swap activity, zswap can improve responsiveness and decrease disk wear by keeping more pages in RAM in compressed form. However, CPU time spent on compression and the RAM reserved for the pool can offset gains on memory-constrained or CPU-bound systems. Zswap may be less beneficial if many pages are seldom re-referenced.

History and scope: Zswap was introduced to provide a compressed swap cache and has matured across kernel releases, adding support for multiple compressors and tunables. It is supported by modern Linux distributions and integrates with the swap subsystem and the kernel memory manager when configured.