Overview
Raw HTTP throughput benchmark across four languages. Uses wrk2 for coordinated load generation, measuring requests per second, latency percentiles, and error rates.
Test Configuration
- Server: 4 vCPU, 8GB RAM
- Load gen: wrk2 across 4 threads
- Payload: 256-byte JSON response
- Duration: 30s per concurrency level
Results (RPS at P99 < 10ms)
| Stack | 100 conns | 1K conns | 10K conns | |---|---|---|---| | Rust (Actix) | 185,000 | 142,000 | 89,000 | | Go (net/http) | 95,000 | 78,000 | 52,000 | | Node.js (Fastify) | 42,000 | 35,000 | 28,000 | | Python (uvicorn) | 8,200 | 7,100 | 5,400 |
Usage
docker-compose -f docker-compose.bench.yml up