Yac keeps the cache in a shared memory segment accessible to all PHP workers, so a read is a hash lookup in local memory — no cache server, no socket, no network.
Highlights
- Fast. No socket, no network, no global lock, so throughput scales with workers. Yac’s own benchmark measured 13x-70x higher ops/s than APCu and 170x-730x higher than local Memcached, depending on value size. On the author’s own site (www.laruence.com) that translated to page renders coming out 19% faster than with the Memcached drop-in.
- Nothing to operate. No cache server to install, secure, monitor or restart.
- Health you can read. Hit rate, hits and misses charted over Today / Yesterday / Last 7 days, with a diagnosis that also says when not to add memory.
- Fails soft. No extension, or
yac.enable=0, degrades to a per-request cache and the site keeps serving.
Best fit
Single-node or few-node installs where all PHP workers share a machine. For multi-server clusters needing cross-node consistency, a network cache (Memcached/Redis) suits better.
Screenshots
Dashboard widget: hit rate, uptime, key slots, values occupied and the 24-hour counters.
Tools → Yac Object Cache: the status bar, with the live shared-memory round trip.
Tools → Yac Object Cache: cache status — hit rate, hits and misses over Today / Yesterday / Last 7 days, with the capacity diagnosis on hover.
Tools → Yac Object Cache: shared memory contents — keys-by-group pie, occupancy totals and the largest (or hottest) entries, each clickable for the entry inspector.