chunkloris: hyper 1.5 + h2 0.4 (h2c)
on this page
part of the chunkloris per-chunk amplification survey. this page is the per-server record for hyper 1.5 + h2 0.4 (h2c) under http/2 (h2c) data frames.
at a glance
- server: hyper 1.5 + h2 0.4 (h2c)
hyper 1.5 / h2 0.4 - runtime: rustc 1.85
- ecosystem: rust
- concurrency model: event-loop
- parser: h2 crate (Rust)
- delivery granularity:
per-chunk - chunk-limit helper: none exposed by the framework
- verdict: batches correctly — the implementation coalesces wire units before waking the application, either via an explicit per-stream frame credit, a pipelined reader, or a similar batching primitive. mode b cpu cost is in the band you would expect from a per-
recv()batched delivery. - scaling exponent (mode a): 0.82 (wall time vs N, log-log slope across common cells)
- scaling exponent (mode b): 1.00
measurements
all cells run on a 1-vcpu docker container. cpu cost is derived from the target container’s cgroup v2 cpu.stat usage_usec delta around each cell.
| mode | N | wall (s) | server cpu % | µs / frame | basis | ok |
|---|---|---|---|---|---|---|
A-h2-bridge | 50,000 | 0.203 | 16.0 | 0.650 | server-cpu-cgroup | ✓ |
A-h2-bridge | 100,000 | 0.365 | 13.5 | 0.492 | server-cpu-cgroup | ✓ |
A-h2-bridge | 250,000 | 0.759 | 16.0 | 0.486 | server-cpu-cgroup | ✓ |
B-h2-paced-100us | 50,000 | 5.375 | 3.8 | 4.086 | server-cpu-cgroup | ✓ |
B-h2-paced-100us | 100,000 | 10.780 | 3.7 | 3.994 | server-cpu-cgroup | ✓ |
B-h2-paced-100us | 250,000 | 26.988 | 3.4 | 3.670 | server-cpu-cgroup | ✓ |
what this means
the implementation batches wire units before waking the application (either via an explicit per-stream frame credit, a pipelined reader, or an equivalent primitive). the cpu cost under paced mode b is in the band you would expect from a per-recv() batched delivery.
what to do today
- if this is an h2 origin, prefer a frontend that terminates h2 into h1 with
proxy_request_buffering onupstream. - consider imposing a per-stream
DATA-frame credit (count, not bytes) before forwarding the body to the application handler. - HTTP/2 byte-level flow control (
WINDOW_UPDATE) does not bound the number of frames; configure stream-frame-rate limits where the implementation exposes them.
reproducer
the full reproducer for this server is in the paper repo. the docker container pins hyper 1.5 + h2 0.4 (h2c) hyper 1.5 / h2 0.4 and constrains the test container to a single cpu (--cpus=1). the prober script implements mode a (bridge-coalesced) and mode b (paced 100 µs) per the methodology section.
see the draft pdf for the full per-framework discussion.