I benchmarked identical Go and Rust servers under heavy load:
- 15,000 concurrent users
- 15 minutes sustained load
- Real Hetzner K3s cluster (not localhost)
- Separate nodes for apps, monitoring, and load testing
- k6 for load generation, Prometheus + Grafana for metrics
Both servers do the same thing: 100 iterations of SHA256 hashing, return JSON.
Go: standard library only
Rust: Hyper + Tokio + sha2
I tracked p50, p90, p95, p99 latencies in real-time.
The results were not what I expected.
huseyinbabal|26 days ago
Code is on GitHub if you want to run it yourself: https://github.com/huseyinbabal/benchmarks/tree/main/rust-se... Curious what you all think.