Inside Memcached vs Redis: Architecture, Memory, and Persistence

Overview

Memcached and Redis are in‑memory key‑value cache servers that accelerate data access by storing frequently used data in RAM, reducing database queries and improving performance under high load.

Service Model

Both run as independent processes (daemonizable) and provide network services via TCP (and optionally UDP). When the client and server reside on the same machine, Unix domain sockets can be used for inter‑process communication.

Event Model

Both use epoll for event notification on Linux (Redis can fall back to select / poll, and on BSD it can use kqueue). Redis runs a single‑threaded event loop, while Memcached employs a master‑worker multi‑threaded model. Redis stores client connections in a red‑black tree or an array indexed by file descriptor for O(1) lookup; Memcached uses a hash table with chaining and a background thread for slab expansion.