Benchmarks
Benchmarking & Load Testing
How to run the benchmark suites and what each suite is good for.
Trypema ships with two performance views:
- Criterion microbenchmarks for per-operation cost
- the
trypema-stressharness for throughput, latency, and contention behavior
Which suite answers which question
| Question | Use |
|---|---|
| How expensive is one operation on this code path? | Criterion |
| How much throughput can this provider sustain under concurrency? | stress harness |
| What happens to p99 and p99.9 under load or bursts? | stress harness |
If you are new to the output, start with Benchmark Concepts.
Quick commands
make bench-local
make bench-redis
make stress-local
make stress-redis
make stress-help
Redis benchmark prerequisites
- Docker with Compose support
- a Redis instance reachable through
REDIS_URL
The project defaults to Redis on redis://127.0.0.1:16379/.
Reading the results
- Use
maxmode to find peak throughput. - Use
target-qpsmode when you care about tail latency under a fixed offered load. - Treat cross-library comparisons as directional, especially when semantics differ.