HTTP/2 Stress Tester

login

h2load fires sustained concurrent HTTP/2 requests to benchmark a server. A public, unauthenticated form for it is indistinguishable from a DDoS tool, so it's disabled. See the nghttp probe for a single read-only HTTP/2 negotiation.

This tool isn't available in the anonymous view. Sign in to access it.

Monitor this automatically

NetTests can run this check on a schedule, preserve historical results, compare changes over time, and alert you the moment something breaks.

Start monitoring free → See all monitoring products

Frequently Asked Questions

What does h2load benchmark?

h2load from the nghttp2 project sends many concurrent HTTP/2 requests and reports: total requests completed, requests per second, transfer rate, latency percentiles (min/max/mean/sd), and HTTP/2 stream statistics. It's the standard tool for HTTP/2 server performance benchmarking, comparable to ab (ApacheBench) for HTTP/1.1.

What HTTP/2 features make it faster than HTTP/1.1?

HTTP/2 multiplexes multiple requests over a single TCP connection (eliminating head-of-line blocking), compresses headers with HPACK (reducing overhead on repeat requests), and allows server push (sending resources before they're requested). Under h2load's concurrent load, these benefits are measurable compared to HTTP/1.1 benchmarks with the same concurrency.

What concurrency settings should I use?

h2load's -c flag sets concurrent clients and -m sets max concurrent streams per connection. A typical benchmark: h2load -n 10000 -c 100 -m 10 https://example.com/ — 10,000 total requests, 100 clients, 10 streams each. Start conservative and increase until you see latency degradation to find capacity limits.