diskcaches

DiskCaches are caching mechanisms that store copies of data blocks on non-volatile storage to speed up future disk I/O. They sit between the primary storage and the requester, aiming to reduce latency and improve throughput by avoiding repeated reads or writes to slower devices.

Typical deployments include hardware-based caches on storage controllers, software caches that use solid-state drives as a fast tier, and operating system or database features that cache blocks or pages on disk as a secondary cache layer. Operation involves a multi-tier hierarchy: an in-memory cache (RAM), an on-disk cache, and the primary storage. On a request, the system checks RAM, then disk cache, then the backing store. Writes may be propagated to the disk cache and to the underlying storage depending on policy, such as write-through or write-back.

Common eviction policies include LRU, LFU, and time-based heuristics. Coherency and consistency are critical: caches must invalidate or update cached data on writes to ensure stale data isn't returned. Some disk caches are write-back and require power-loss protection or journaling to preserve data integrity.

Performance considerations include cache size, hit rate, and the speed of the backing storage, which determine the overall benefits. Disk caches can help where read locality exists or where write latency is costly. Limitations include risk of data loss on power failure, cache pollution, and management overhead. Security concerns involve protecting cached data through appropriate access controls and, if needed, encryption.

Examples and implementations include Linux dm-cache and bcache, Windows storage tiering features, hardware RAID controllers with flash caches, and database systems that implement on-disk caching layers. The term DiskCaches is used generically to describe any on-disk caching layer integrated into a storage stack.