NOTE
Redis Benchmarking
How to benchmark Redis without confusing synthetic command throughput with application capacity, including concurrency, payloads, pipelines, latency percentiles, and hot-key effects.
This is a historical learning note and may contain outdated or incomplete understanding.
1. redis-benchmark
Redis ships a benchmark utility useful for quick synthetic tests.
But a headline requests/second number is meaningful only with its configuration:
- command mix;
- key distribution;
- payload size;
- client concurrency;
- pipeline depth;
- network placement;
- TLS;
- persistence;
- cluster topology.
2. Measure Latency, Not Only Throughput
Track p50/p95/p99/p999 latency while increasing offered load. The sustainable capacity point is before tail latency and timeout/error rate become unacceptable.
3. Match Production
A real cache workload may contain:
- high hit-rate GETs;
- large values;
- hot keys;
- expirations/evictions;
- Lua scripts;
- replication and persistence.
Benchmark those patterns rather than only tiny SET/GET commands on localhost.
4. Avoid Benchmark Artifacts
Pipelining can produce spectacular throughput while changing request concurrency and batching semantics. Report whether pipelines were used.
Also monitor server CPU per core, memory, network, fork/fsync behavior, and client bottlenecks so you know what resource actually saturated.