lock-striping

This article benchmarks six Go in-memory cache designs under different workloads, finding that a sharded map with 256 locks outperforms single mutex and RWMutex approaches, especially on multi-core systems.