Repository navigation
feat(cc): put ABE (RFC 8511) behind a pref. - #4090
Merged
Merged
Conversation
omansfeld
requested review from
KershawChang,
larseggert,
martinthomson and
mxinden
as code owners
October 16, 2026 17:11
Contributor
There was a problem hiding this comment.
🟢 Approval recommended
The setting is consistently wired, preserves existing behavior by default, and has focused regression coverage; tests were not executed.
0 open findings
What changed in this PR
Makes Alternative Backoff with ECN (ABE) configurable for CUBIC, supporting Firefox performance experiments while preserving the enabled default.
Changes:
- Adds an ABE connection setting and
--no-abeCLI flag. - Propagates the setting to all CUBIC slow-start variants.
- Tests configuration toggling and congestion-window reductions.
| File | Description |
|---|---|
| neqo-transport/src/sender.rs | Passes the ABE setting to CUBIC. |
| neqo-transport/src/connection/params.rs | Adds the setting, default, and toggle test. |
| neqo-transport/src/cc/tests/mod.rs | Supports configurable CUBIC test instances. |
| neqo-transport/src/cc/tests/cubic.rs | Tests backoff behavior with ABE disabled. |
| neqo-transport/src/cc/cubic.rs | Gates alternative ECN backoff on the setting. |
| neqo-bin/src/lib.rs | Adds and wires --no-abe. |
🧠 Review effort: Balanced
💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #4090 +/- ##
==========================================
- Coverage 97.14% 96.99% -0.16%
==========================================
Files 113 118 +5
Lines 41500 41869 +369
Branches 41500 41869 +369
==========================================
+ Hits 40314 40609 +295
- Misses 1172 1236 +64
- Partials 14 24 +10
Flags with carried forward coverage won't be shown. Click here to find out more.
|
Contributor
Merging this PR will regress 2 benchmarks
Warning Please fix the performance issues or acknowledge them on CodSpeed. Performance Changes
Tip Investigate this regression by commenting Comparing 
|
Collaborator
|
Is ABE not orthogonal to whether CUBIC or Reno are used? |
Collaborator
Author
It is in theory, but we currently only implement it for cubic. That’s why I put the pref on cubic instead of on the congestion controller. The diff would be bigger if it was a layer above because of changing function calls etc. In general I don’t think it is worth it to implement ABE in Reno right now. We don’t use Reno and the constant needs to be separately tuned from cubic. |
Collaborator
|
I wonder if there is any point in keeping the Reno code around... |
larseggert
approved these changes
Oct 19, 2026
Collaborator
Author
I think it can be useful as a barebones RFC 9002/new reno reference. But right now we have some performance optimizations (e.g. spurious recovery) active for both, and some only for Cubic. So I agree that state is not that useful. If we decided to rip out everything but Cubic and also that we won't be doing any non-loss cc anytime soon (thus no need for the current I'll open an issue to discuss/track this. |
This will enable us to do experiments studying ABE performance on Firefox. See https://bugzilla.mozilla.org/show_bug.cgi?id=2062270 for more info on my planned ECN performance study.
omansfeld
force-pushed
the
ecn-abe-pref
branch
from
October 19, 2026 09:57
4fdf636 to
7cd4a09
Compare
omansfeld
enabled auto-merge
October 19, 2026 10:03
Collaborator
Author
There is #3453. I added this discussion to it. |
Contributor
Client/server transfer resultsPerformance differences relative to ad7d629. Transfer of 33554432 bytes over loopback, min. 100 runs. All unit-less numbers are in milliseconds.
Table above only shows statistically significant changes. See all results below. All resultsTransfer of 33554432 bytes over loopback, min. 100 runs. All unit-less numbers are in milliseconds.
Download data for |
Contributor
Benchmark resultsNo significant performance differences relative to ad7d629. All resultstransfer/1-conn/1-100mb-req (aka. Upload)/mtu-1500 time: [50.306 ms 50.329 ms 50.354 ms]
thrpt: [1.9394 GiB/s 1.9403 GiB/s 1.9413 GiB/s]
Found 3 outliers among 100 measurements (3.00%)
1 (1.00%) low mild
1 (1.00%) high mild
1 (1.00%) high severetransfer/1-conn/1-100mb-resp (aka. Download)/mtu-1500 time: [52.037 ms 52.066 ms 52.096 ms]
thrpt: [1.8746 GiB/s 1.8756 GiB/s 1.8767 GiB/s]
Found 1 outliers among 100 measurements (1.00%)
1 (1.00%) high mildtransfer/1-conn/1-1b-resp (aka. HPS)/mtu-1500 time: [2.9774 ms 2.9795 ms 2.9815 ms]
thrpt: [335.40 B/s 335.63 B/s 335.86 B/s]
Found 1 outliers among 100 measurements (1.00%)
1 (1.00%) high mildtransfer/1-conn/10_000-parallel-1b-resp (aka. RPS)/mtu-1500 time: [67.421 ms 67.758 ms 68.105 ms]
thrpt: [146.83 Kelem/s 147.58 Kelem/s 148.32 Kelem/s]
Found 1 outliers among 100 measurements (1.00%)
1 (1.00%) high mildstreams-flow-controlled/walltime/1-streams/each-4194304-bytes time: [10.632 ms 10.635 ms 10.638 ms]
thrpt: [376.03 MiB/s 376.13 MiB/s 376.23 MiB/s]
Found 2 outliers among 100 measurements (2.00%)
2 (2.00%) high mildstreams-flow-controlled/walltime/10-streams/each-1048576-bytes time: [28.260 ms 28.270 ms 28.280 ms]
thrpt: [353.60 MiB/s 353.73 MiB/s 353.86 MiB/s]
Found 1 outliers among 100 measurements (1.00%)
1 (1.00%) high mildstreams/walltime/1-streams/each-1000-bytes time: [132.29 µs 132.85 µs 133.55 µs]
thrpt: [7.1407 MiB/s 7.1786 MiB/s 7.2088 MiB/s]
Found 2 outliers among 100 measurements (2.00%)
1 (1.00%) high mild
1 (1.00%) high severestreams/walltime/1000-streams/each-1-bytes time: [3.7571 ms 3.7685 ms 3.7797 ms]
thrpt: [258.37 KiB/s 259.14 KiB/s 259.93 KiB/s]streams/walltime/1000-streams/each-1000-bytes time: [12.529 ms 12.536 ms 12.544 ms]
thrpt: [76.029 MiB/s 76.074 MiB/s 76.120 MiB/s]
Found 1 outliers among 100 measurements (1.00%)
1 (1.00%) low mildtransfer/walltime/pacing-false/same-seed time: [1.9996 ms 2.0006 ms 2.0018 ms]
thrpt: [1.9514 GiB/s 1.9526 GiB/s 1.9535 GiB/s]transfer/walltime/pacing-false/varying-seeds time: [2.0204 ms 2.0213 ms 2.0224 ms]
thrpt: [1.9315 GiB/s 1.9326 GiB/s 1.9334 GiB/s]transfer/walltime/pacing-true/same-seed time: [2.0787 ms 2.0813 ms 2.0844 ms]
thrpt: [1.8740 GiB/s 1.8768 GiB/s 1.8791 GiB/s]transfer/walltime/pacing-true/varying-seeds time: [2.0170 ms 2.0181 ms 2.0193 ms]
thrpt: [1.9344 GiB/s 1.9356 GiB/s 1.9367 GiB/s]Instructions per cycleCriterion reported no significant timing changes. All benchmarks
Profiles for profiler.firefox.com (64)
Download data for |
Contributor
Failed Interop TestsQUIC Interop Runner, client vs. server, differences relative to
All resultsSucceeded Interop TestsQUIC Interop Runner, client vs. server neqo-pr as client
neqo-pr as server
Unsupported Interop TestsQUIC Interop Runner, client vs. server neqo-pr as client
neqo-pr as server
|
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.
This will enable us to do experiments studying ABE performance on Firefox.
See https://bugzilla.mozilla.org/show_bug.cgi?id=2062270 for more info on my planned ECN performance study.