chunkloris: quic-go http/3
on this page
part of the chunkloris per-chunk amplification survey. this page is the per-server record for quic-go HTTP/3 under http/3 data frames over quic.
at a glance
- server: quic-go HTTP/3
quic-go v0.54.0 - runtime: Go 1.23
- ecosystem: go
- concurrency model: n-m-scheduler
- parser: quic-go/http3
- delivery granularity:
per-frame - chunk-limit helper: none exposed by the framework
- verdict: per-frame — the parser/dispatcher boundary delivers one event per protocol frame (h2 / h3 DATA frame, or ws data frame). cpu cost under paced mode b is measurable per frame.
- scaling exponent (mode a): 0.94 (wall time vs N, log-log slope across common cells)
- scaling exponent (mode b): 1.01
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-h3-bridge | 50,000 | 0.120 | 61.6 | 1.479 | server-cpu-cgroup | ✓ |
A-h3-bridge | 100,000 | 0.260 | 39.7 | 1.031 | server-cpu-cgroup | ✓ |
A-h3-bridge | 250,000 | 0.550 | 37.7 | 0.830 | server-cpu-cgroup | ✓ |
B-h3-paced-100us | 50,000 | 7.785 | 19.1 | 29.652 | server-cpu-cgroup | ✓ |
B-h3-paced-100us | 100,000 | 15.390 | 19.8 | 30.433 | server-cpu-cgroup | ✓ |
B-h3-paced-100us | 250,000 | 39.401 | 21.1 | 33.193 | server-cpu-cgroup | ✓ |
parser path — source citations
- HTTP/3 server —
?→ source
what this means
the parser/dispatcher path on this server delivers one event per protocol frame (a http/3 data frames over quic DATA frame or ws frame), so an attacker who sends a request body as N one-byte frames consumes roughly N × (mode-b µs/frame) of server cpu on a single core.
what to do today
- HTTP/3 quic stream / connection flow control is byte-level; it does not bound DATA frame count.
- consider a per-stream DATA-frame credit at the h3 dispatcher before delivering payload to the application.
reproducer
the full reproducer for this server is in the paper repo. the docker container pins quic-go HTTP/3 quic-go v0.54.0 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.