Valkey vs. Redis: interesting results on io-threads scaling

Copying the summary here.

Redis-benchmark, GET commands 64 clients, bare metal Xeon Gold 6312U 24/48vCPU, 128G; 10M keys DB size, 256 bytes each.

Summary: - Redis does scale! and performance is fine on low-cores infra - Valkey scales much better with significantly lower latency - Both have limitations due to single-main-thread architecture (having>8 vcpus doesn't make much sense - use several services per box or cluster mode)