Redis Cache Cluster Solutions: Dynomite, Twemproxy, and Codis

Because a single Redis server has limited memory management capacity and excessive memory usage degrades performance, many organizations adopt proxy‑based Redis clustering solutions to improve cache performance and availability; this article summarizes the approaches taken by Netflix, Twitter, and the Chinese company Wandoujia.

1. Netflix’s Open‑Source Dynamo Implementation Dynomite

Dynomite is Netflix’s open‑source, C/C++‑implemented proxy that turns Redis and Memcached into a distributed database, also supporting persistent back‑ends such as MySQL, BerkeleyDB, and LevelDB, and offering simple, efficient, cross‑data‑center replication under the Apache License 2.0.

2. Twitter’s Redis/Memcached Proxy Service Twemproxy

Twemproxy, written in C, is a lightweight Redis proxy that consolidates multiple Redis or Memcached instances behind a single endpoint; it automatically removes failed nodes and re‑adds them when they recover, incurring at most about 20 % performance overhead compared with direct Redis access, and is released under the Apache License 2.0.

3. Wandoujia’s Redis Cluster Solution Codis

Codis, developed by Wandoujia using Go and C, is a proxy‑based Redis clustering solution fully compatible with Twemproxy; it handles request forwarding, online data migration, and presents a single logical Redis service to clients, and is distributed under the MIT License.

Other companies such as Sina, Baidu, Taobao, and Tencent have internal, non‑open‑source Redis cluster solutions; until Redis’s official production‑grade clustering is released, the three solutions above are worth considering for production use.