This document proposes a design for a concurrent cache that spreads cached data across a cluster of computers. It separates persistent storage from cache storage and implements cached objects as processes to allow for runtime configurability. Each cached datum operates as an independent server process, allowing cached data to be accessed and replaced in parallel. The cache manager process monitors datum processes and maintains a location index. This implementation allows experimentation with cache size, replacement policies, and other performance aspects independently of the underlying data store.