Benchmarks

Results captured on 2026-04-24 using the release build.

google.com — 1000 requests, 10 workers, 100 concurrency¶

cargo run --release -- --target google.com --insecure \
  --success-status 2xx,3xx -n 1000 --workers 10 -c 100

Starting HTTP/3 load test:
  Target: google.com:443
  Host: google.com
  Path: /
  Workers: 10
  Concurrency per worker: 100
  Total requests: 1000
  Duration: 30s
  Insecure: true

Load test completed:
  Total time: 1.10s
  Total requests: 1000
  Successful requests: 1000
  Failed requests: 0
  Requests/sec: 907.55
  Completion reason: All 1000 requests completed

HTTP Status code breakdown:
  302: 3xx Redirect (1000)

Latency metrics (ms):
  Min:   283.63
  Max:  1052.91
  Avg:   690.64
  p50:   685.00
  p90:   854.11
  p95:   864.72
  p99:   930.92

Notes: - Google closes the QUIC connection after every response (one request per connection). - Each request pays a full TLS+QUIC handshake round trip (~300-1000ms to Google's servers). - Throughput of ~907 req/s is achieved by running 10 workers × 100 concurrent streams in parallel, saturating the reconnect pipeline. - Latency figures reflect handshake cost, not server processing time. A server that reuses connections will show much lower p50/p99.

google.com — 10000 requests, 100 workers, 1000 concurrency¶

cargo run --release -- --target google.com --insecure \
  --success-status 2xx,3xx -n 10000 --workers 100 -c 1000

Starting HTTP/3 load test:
  Target: google.com:443
  Host: google.com
  Path: /
  Workers: 100
  Concurrency per worker: 1000
  Total requests: 10000
  Duration: 30s
  Insecure: true

Load test completed:
  Total time: 8.62s
  Total requests: 10000
  Successful requests: 10000
  Failed requests: 0
  Requests/sec: 1159.92
  Completion reason: All 10000 requests completed

HTTP Status code breakdown:
  302: 3xx Redirect (10000)

Latency metrics (ms):
  Min:   308.18
  Max:  3806.45
  Avg:  1574.87
  p50:  1434.02
  p90:  2579.17
  p95:  2667.30
  p99:  2795.15

Notes: - 10× the requests at 10× the concurrency completes in 8.6s with zero failures. - Higher p99 (2.8s vs 0.9s) reflects queuing at the reconnect layer — each of the 100 workers is cycling through 1000 concurrent connections against a server that closes after every response. - Throughput scales from ~907 to ~1160 req/s as more workers absorb handshake latency in parallel.

Interpreting Results¶

Metric	What it tells you
`Requests/sec`	Overall throughput across all workers
`p50`	Median request latency — typical user experience
`p90` / `p95`	Tail latency — affects 1 in 10 / 1 in 20 requests
`p99`	Worst-tail latency — affects 1 in 100 requests
`Failed requests`	Requests that did not receive a valid response

A healthy server with persistent QUIC connections will show latency well under 100ms p99 and near-zero failed requests.