Repository navigation
[meru-staging][experiment] Raise CXPLAT_MAX_MTU to 3900 - #6497
Draft
Stephen L Peters (StephenLPeters) wants to merge 2 commits into
Draft
Stephen L Peters (StephenLPeters) wants to merge 2 commits into
Stephen L Peters (StephenLPeters) wants to merge 2 commits into
Conversation
…oft#6363) ## Description See microsoft#6362 for the description. This PR removes unnecessary stream scanning when logging is disabled. ## Benchmark Result Benchmark client setup: ``` ./build/tcam/bin/Release/secnetperf -target:localhost -conns:1 -streams:1000 -rstream:1 -run:10s -up:512 -down:4kb -plat:1 ``` Original code: ``` Result: 11383 RPS, Latency,us 0th: 10525, 50th: 86901, 90th: 88417, 99th: 123274, 99.9th: 132490, 99.99th: 135393, 99.999th: 137258, 99.9999th: 137265, Max: 137265 ``` Patched code: ``` Result: 118139 RPS, Latency,us 0th: 6319, 50th: 8515, 90th: 8892, 99th: 12190, 99.9th: 15922, 99.99th: 25233, 99.999th: 25400, 99.9999th: 25450, Max: 2545 ``` Here's the result of secnetperf with different number of streams | Concurrent streams | Before: p99 latency | After: p99 latency | Before: RPS | After: RPS | |---:|---:|---:|---:|---:| | 1 | 0.088 ms | 0.087 ms | 13,131 | 13,183 | | 10 | 0.199 ms | 0.191 ms | 78,792 | 81,551 | | 100 | 1.341 ms | 0.976 ms | 130,504 | 144,176 | | 1000 | 23.484 ms | 10.773 ms | 52,733 | 133,512 | ## Testing This PR does not need additional test code. --------- Co-authored-by: Guillaume Hetier <hetier.guillaume@gmail.com> (cherry picked from commit 47504f1)
Allow DPLPMTUD to probe up to Azure's Mellanox VF MTU limit when MaximumMtu is configured above 1500. The default MaximumMtu stays 1500, and IP_PMTUDISC_PROBE keeps oversized probes from fragmenting on hosts with a smaller MTU. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: d84afdd6-7b22-401b-9969-3c9c762497df
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Experiment, not for merge. Builds on #6492.
Raises
CXPLAT_MAX_MTUfrom 1500 to 3900 so DPLPMTUD can probe up to Azure's Mellanox VF MTU limit when an application configuresMaximumMtuabove 1500. Today,QuicSettingApplyclampsMaximumMtutoCXPLAT_MAX_MTU, and the Linux datapath reportsCXPLAT_MAX_MTUas the local MTU, so larger packets are impossible regardless of host configuration.MaximumMtu(QUIC_DPLPMTUD_DEFAULT_MAX_MTU) stays 1500, so applications that don't opt in are unchanged.IP_PMTUDISC_PROBE, so probes larger than a host's MTU fail instead of fragmenting.MAX_UDP_PAYLOAD_LENGTH, and therefore the small (non-GRO) I/O buffers, grow with the constant.Used to measure per-packet CPU savings for Meru storage replication on Azure VMs with accelerated networking.