envoyproxy/envoy

quic: measure GRO performance and enable it via config

Open

#19,119 opened on Nov 29, 2021

View on GitHub
 (0 comments) (0 reactions) (0 assignees)C++ (5,373 forks)batch import
area/perfarea/quicarea/udphelp wanted

Repository metrics

Stars
 (27,997 stars)
PR merge metrics
 (Avg merge 8d) (378 merged PRs in 30d)

Description

UDP GRO is supposed to improve socket read performance on client side but currently our code performs very poorly in nighthawk client, by reading one packet per read while allocating 24k buffer. It will be worthwhile to invest the suitable traffic patten for GRO. Meanwhile this feature is disabled in https://github.com/envoyproxy/envoy/pull/19088 and it will be good to exposed it via a config knob after we know the suitable use case.

And there was a bug reported that the socket reading code allocating 24k buffer per read might lead to MSG_TRUNC and cause packets dropped occasionally. We should change the value to be 64k which is the max GRO receive buffer allowed by the kernel.

Contributor guide

Research direction
Investigate the GRO feature in the kernel and Envoy's QUIC socket reading code. Review the existing tests for GRO and the config mechanism. Measure performance with different buffer sizes.
Tech stack
cpp
Domain
backend
Issue type
Feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
1-2 days
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Clear
Prerequisites
C++Linux networkingEnvoy development setup
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
65