In praise of memcached

Category: Engineering post
Status: confirmed
Sources: jchri.st, discussion
Summary: A post dated 2026-06-02 argues for memcached over Redis as a pure cache on operational grounds: client libraries treat a connection failure as a cache miss and return a default, scaling uses client-side consistent hashing rather than built-in clustering (a failed node drops from the hash ring and is retried periodically), and the absence of disk persistence keeps the cache strictly ephemeral. The author frames Redis-as-cache as an antipattern because teams come to depend on its persistence and cannot later remove it.
Why it matters: It restates a clear boundary between an ephemeral cache and a datastore, a recurring source of coupling bugs when a cache is quietly treated as a database.