Caching

A cache answers in under a millisecond what a database answers in several, which makes it the first part most systems add when reads grow. You will trace a cache-aside read step by step, choose how a cached value is kept fresh, predict what LRU eviction removes, and see how a Bloom filter stops lookups for keys that do not exist. The topic then turns to the two ways a cache takes a system down: one that grows without limit until the process dies, and a popular key whose expiry sends every request to the database at once. Distributed caches and their replication are left to the subject on distributed data.