Skip to content
redis-web Docs

Hiredis Compat Performance Harness

Goal

Section titled “Goal”

Track performance over time for optimization work. This harness is informational and does not currently enforce CI pass/fail thresholds.

Comparison Modes

Section titled “Comparison Modes”

Measure at minimum:

  • direct hiredis -> Redis TCP baseline
  • compat over redis-web WS raw path
  • compat over redis-web HTTP fallback path

Workload Categories

Section titled “Workload Categories”

Include representative workloads:

  • small GET/SET command loops
  • mixed read/write sequences
  • pipelined bursts
  • Pub/Sub message fanout scenarios

Reporting

Section titled “Reporting”

Capture per run:

  • p50/p95/p99 latency
  • throughput (ops/sec)
  • command mix and pipeline depth
  • transport mode and host environment

Store benchmark results as artifacts so maintainers can compare trendlines between commits.

Current harness script:

  • crates/redis-web/tests/bench-hiredis-compat.sh

Run:

Terminal window
make bench_hiredis_compat

With environment overrides:

Terminal window
HOST=127.0.0.1 PORT=7379 ITERATIONS=500 make bench_hiredis_compat

The harness prints timing summary per run, for example:

[compat-bench] Creating compat session on http://127.0.0.1:7379/__compat/session
[compat-bench] Running 500 SET/GET roundtrips
[compat-bench] elapsed_sec=12 iterations=500
[compat-bench] Cleaning up session

Caveats

Section titled “Caveats”

Do not compare runs collected on materially different hardware or background load conditions without clear labeling.