Distributed Cache System Design (Sharding, Eviction & Scale)

You’re building an in-memory cluster that feels like Redis or Memcached at scale: sub-millisecond reads, horizontal add/remove of nodes, and explicit policies for eviction, replication, and what “consistent” means when failures and partitions happen. The interview is less about memorizing RDB/AOF flags and more about partitioning, hot keys, and failure behavior.