[PERF] Minimal API H2 - H3 performance difference

[MDA2AV]

We're benchmarking min apis accross H1, H2 and H3 and see a huge regression in performance between H2 and H3

Implementation

H2 load generator: h2load
H3 load generator: h2load-h3

H2 results
H3 results

Are there guidelines for H3 docker file? our dockerfile

Thanks

[github-actions]

Triage Summary

Area: area-networking — The issue centers on HTTP/2 vs HTTP/3 (QUIC) performance benchmarking, which falls under Kestrel's networking layer. Minimal API is used as the workload, but the protocol-level performance difference is a networking concern.

Type: performance — The author reports a large performance gap between H2 and H3 in their benchmarks and is asking whether this is expected or if there is a configuration issue (including a Docker question for H3).

Potential Duplicates

• None found — Searches for "HTTP/3 performance minimal API" and "H3 QUIC performance regression" did not surface closely matching open issues.

Notes

• The "regression" in the title appears to mean a performance difference between H2 and H3 (two different protocols) rather than the same behavior regressing across ASP.NET Core versions. No version-to-version regression signals are present.
• The author is also asking for best-practice Docker guidance for running HTTP/3, which may be better addressed via documentation or the Discussions tab.
• It may be worth clarifying whether the H3 test is running over a real QUIC/UDP path vs a loopback, and whether the Docker network configuration properly supports UDP (required for QUIC/HTTP/3).

Generated by Issue Triage Agent for dotnet/aspnetcore for issue #66400 · ● 322.6K · ◷