Semicolony home
Search guides and simulators ⌘K
Performance

Cache hit ratio

Fraction of lookups answered from the cache.

Main article: How caching works

In plain terms

Because misses are much slower than hits, small changes matter: with 1 ms hits and 50 ms misses, raising the hit ratio from 90% to 99% cuts the average time per request by three quarters.

Origin

The metric is as old as caching itself. Modern systems target 95%+ as the healthy threshold; below 80% suggests the cache key design is wrong.

Where it shows up in production
  • Facebook TAO Documented in their 2013 paper as targeting >99% read hit ratio on the social graph.
  • CDN PoPs A cache hit ratio reported per object size; 90%+ for static, 50–70% for highly personalised.
On Semicolony
Handbook: caching strategies Handbook: caching strategies
Sources & further reading
Found this useful?

Published