Standardize how we implement in-memory caches