bechmark: scale up RPS to test data plane CPU performance by zhaohuabing · Pull Request #7810 · envoyproxy/gateway

zhaohuabing · 2025-12-24T05:19:35Z

What type of PR is this?
This PR adjusts the Envoy Gateway benchmark to measure data plane CPU performance using RPS instead of route count.

While route count is useful for evaluating the CPU performance patterns of the control plane(EG) , data plane(Envoy) CPU usage is primarily driven by request processing. Scaling load by RPS provides a more stable signal, and makes the results easier to interpret for data plane performance analysis.

This PR also:

Previously, failed Prometheus queries resulted in sample values being recorded as 0 and used to generate the report. These failed samples are now excluded from the min/max/mean calculations in the report.
Data-plane CPU sampling previously started before traffic was sent and continued after traffic had finished. It is now restricted to the actual traffic window, capturing CPU usage only while traffic is being sent.
Prometheus scraping was set to 10s, and BENCHMARK_DURATION was increased to 90s to ensure we collect enough samples for generating the reports.

The envoy proxy cpu metrics in the current benchmark report:

The envoy proxy cpu metrics in this PR:

Test Name	Envoy Gateway Container Memory (MiB) min/max/means	Envoy Gateway Process Memory (MiB) min/max/means	Envoy Gateway CPU (%) min/max/means	Averaged Envoy Proxy Memory (MiB) min/max/means	Averaged Envoy Proxy CPU (%) min/max/means
scaling up httproutes to 10 with 2 routes per hostname at 100 rps	122.74 / 134.74 / 132.60	30.68 / 43.92 / 37.77	0.45 / 1.20 / 0.60	4.16 / 22.77 / 22.14	5.79 / 9.93 / 9.10
scaling up httproutes to 50 with 10 routes per hostname at 300 rps	134.10 / 143.62 / 141.19	35.01 / 52.05 / 42.50	0.55 / 2.10 / 0.84	26.81 / 28.89 / 28.53	18.98 / 28.93 / 27.36
scaling up httproutes to 100 with 20 routes per hostname at 500 rps	140.60 / 148.92 / 145.47	37.59 / 49.14 / 42.69	0.65 / 2.45 / 1.03	32.87 / 34.96 / 34.72	45.10 / 46.27 / 45.50
scaling up httproutes to 300 with 60 routes per hostname at 800 rps	144.14 / 153.39 / 150.26	42.30 / 60.46 / 52.16	0.75 / 4.85 / 1.57	55.12 / 68.81 / 63.44	43.81 / 68.25 / 62.95
scaling up httproutes to 500 with 100 routes per hostname at 1000 rps	158.59 / 166.66 / 164.15	52.29 / 68.52 / 58.96	1.00 / 7.40 / 2.06	84.20 / 100.09 / 90.03	53.02 / 80.69 / 75.48
scaling up httproutes to 1000 with 200 routes per hostname at 2000 rps	173.98 / 188.18 / 186.47	65.00 / 87.07 / 77.12	1.05 / 13.50 / 3.09	140.79 / 141.78 / 141.43	63.11 / 99.49 / 91.75
scaling down httproutes to 500 with 100 routes per hostname at 1000 rps	168.58 / 181.48 / 173.22	54.95 / 78.56 / 67.08	1.00 / 1.35 / 1.16	140.78 / 141.30 / 141.13	54.45 / 81.40 / 76.96
scaling down httproutes to 300 with 60 routes per hostname at 800 rps	154.50 / 168.78 / 160.81	47.39 / 62.73 / 54.50	1.00 / 1.35 / 1.16	136.22 / 140.91 / 137.19	40.49 / 68.74 / 64.18
scaling down httproutes to 100 with 20 routes per hostname at 500 rps	142.30 / 163.45 / 151.32	38.12 / 65.46 / 47.57	1.00 / 1.40 / 1.16	134.57 / 136.95 / 134.93	29.56 / 48.22 / 43.26
scaling down httproutes to 50 with 10 routes per hostname at 300 rps	145.06 / 151.28 / 147.09	35.56 / 47.53 / 40.31	1.10 / 1.25 / 1.15	134.60 / 136.53 / 135.37	16.46 / 29.25 / 25.20
scaling down httproutes to 10 with 2 routes per hostname at 100 rps	144.98 / 151.25 / 146.45	34.93 / 49.85 / 42.12	1.00 / 1.30 / 1.14	136.50 / 136.53 / 136.51	5.30 / 9.77 / 9.04

cc @missBerg

codecov · 2025-12-24T05:33:42Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 72.73%. Comparing base (fe57c70) to head (f70ebad).
⚠️ Report is 2 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #7810      +/-   ##
==========================================
- Coverage   72.77%   72.73%   -0.04%     
==========================================
  Files         235      235              
  Lines       35092    35092              
==========================================
- Hits        25538    25525      -13     
- Misses       7738     7751      +13     
  Partials     1816     1816

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com>

jukie · 2025-12-28T01:58:12Z

/retest

charts/gateway-addons-helm/README.md

Signed-off-by: Huabing(Robin) Zhao <zhaohuabing@gmail.com>

…#7810) * Scale up RPS to test data plane CPU performance Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * set duration to 120s Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * discard invalid samples Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * change scrape interval to 10s Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * remove invalid cpu sampling data Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * reduce duration to 60 Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * fix benchmark end time Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * fix data plane benchmark start time Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * increase test time to get more samples Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * adjust rps for each scale Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * address comments Signed-off-by: Huabing(Robin) Zhao <zhaohuabing@gmail.com> --------- Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> Signed-off-by: Huabing(Robin) Zhao <zhaohuabing@gmail.com> Signed-off-by: Matt Miller <millermatt@outlook.com>

…#7810) * Scale up RPS to test data plane CPU performance Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * set duration to 120s Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * discard invalid samples Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * change scrape interval to 10s Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * remove invalid cpu sampling data Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * reduce duration to 60 Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * fix benchmark end time Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * fix data plane benchmark start time Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * increase test time to get more samples Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * adjust rps for each scale Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * address comments Signed-off-by: Huabing(Robin) Zhao <zhaohuabing@gmail.com> --------- Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> Signed-off-by: Huabing(Robin) Zhao <zhaohuabing@gmail.com> Signed-off-by: Rudrakh Panigrahi <rudrakh97@gmail.com>

* fix: set observedGeneration in envoy patch policy (#7715) * fix: set observedGeneration in envoy patch policy Signed-off-by: kkk777-7 <kota.kimura0725@gmail.com> * add release note Signed-off-by: kkk777-7 <kota.kimura0725@gmail.com> --------- Signed-off-by: kkk777-7 <kota.kimura0725@gmail.com> Signed-off-by: Rudrakh Panigrahi <rudrakh97@gmail.com> * fix: add validation for request buffer limit (#7687) Signed-off-by: kkk777-7 <kota.kimura0725@gmail.com> Signed-off-by: Rudrakh Panigrahi <rudrakh97@gmail.com> * fix: nil pointer error when applying BackendTrafficPolicy to HTTPRoute with no backendRefs (#7765) * fix: checking route section name in backend traffic policy Signed-off-by: kkk777-7 <kota.kimura0725@gmail.com> Signed-off-by: Rudrakh Panigrahi <rudrakh97@gmail.com> * fix: setting externalTrafficPolicy for NodePort service type (#7823) Signed-off-by: Rudrakh Panigrahi <rudrakh97@gmail.com> * fix: add indexing and processing for CRL references in ClientTrafficPolicies (#7829) Signed-off-by: Rudrakh Panigrahi <rudrakh97@gmail.com> * feat: change the benchmark report to json format (#6818) * benchmark json output Signed-off-by: zirain <zirain2009@gmail.com> * fix Signed-off-by: zirain <zirain2009@gmail.com> * fix Signed-off-by: zirain <zirain2009@gmail.com> * fix lint Signed-off-by: zirain <zirain2009@gmail.com> * fix Signed-off-by: zirain <zirain2009@gmail.com> * revert Signed-off-by: zirain <zirain2009@gmail.com> * fix seconds Signed-off-by: zirain <zirain2009@gmail.com> --------- Signed-off-by: zirain <zirain2009@gmail.com> Signed-off-by: Rudrakh Panigrahi <rudrakh97@gmail.com> * bechmark: scale up RPS to test data plane CPU performance (#7810) * Scale up RPS to test data plane CPU performance Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * set duration to 120s Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * discard invalid samples Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * change scrape interval to 10s Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * remove invalid cpu sampling data Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * reduce duration to 60 Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * fix benchmark end time Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * fix data plane benchmark start time Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * increase test time to get more samples Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * adjust rps for each scale Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> * address comments Signed-off-by: Huabing(Robin) Zhao <zhaohuabing@gmail.com> --------- Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> Signed-off-by: Huabing(Robin) Zhao <zhaohuabing@gmail.com> Signed-off-by: Rudrakh Panigrahi <rudrakh97@gmail.com> * fix: make port-forward worked for OTel collector on port 19001 (#7860) Signed-off-by: zirain <zirain2009@gmail.com> Signed-off-by: Rudrakh Panigrahi <rudrakh97@gmail.com> * chore: fix goroutine leak (#7880) fix goroutine leak Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> Signed-off-by: Rudrakh Panigrahi <rudrakh97@gmail.com> * fix gen-check Signed-off-by: Rudrakh Panigrahi <rudrakh97@gmail.com> --------- Signed-off-by: kkk777-7 <kota.kimura0725@gmail.com> Signed-off-by: Rudrakh Panigrahi <rudrakh97@gmail.com> Signed-off-by: zirain <zirain2009@gmail.com> Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com> Signed-off-by: Huabing(Robin) Zhao <zhaohuabing@gmail.com> Co-authored-by: Kota Kimura <86363983+kkk777-7@users.noreply.github.com> Co-authored-by: zirain <zirain2009@gmail.com> Co-authored-by: Huabing (Robin) Zhao <zhaohuabing@gmail.com>

zhaohuabing requested a review from a team as a code owner December 24, 2025 05:19

zhaohuabing marked this pull request as draft December 24, 2025 05:19

zhaohuabing changed the title ~~Scale up RPS to test data plane CPU performance~~ bechmark: scale up RPS to test data plane CPU performance Dec 24, 2025

zhaohuabing force-pushed the bechmark-scale-rps branch from 96ba7ff to cc9566e Compare December 24, 2025 05:24

zhaohuabing force-pushed the bechmark-scale-rps branch 16 times, most recently from fee920f to 1a6be88 Compare December 26, 2025 09:24

zhaohuabing marked this pull request as ready for review December 28, 2025 00:44

zhaohuabing added 8 commits December 28, 2025 08:44

Scale up RPS to test data plane CPU performance

2249a52

Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com>

set duration to 120s

004ae86

Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com>

discard invalid samples

1eace61

Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com>

change scrape interval to 10s

d5b431f

Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com>

remove invalid cpu sampling data

f1c6ac0

Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com>

reduce duration to 60

11f1a77

Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com>

fix benchmark end time

03070cd

Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com>

fix data plane benchmark start time

c39e258

Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com>

zhaohuabing added 2 commits December 28, 2025 08:44

increase test time to get more samples

d9b8dbb

Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com>

adjust rps for each scale

ed38304

Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com>

zhaohuabing force-pushed the bechmark-scale-rps branch from 1a6be88 to ed38304 Compare December 28, 2025 00:44

jukie previously approved these changes Dec 28, 2025

View reviewed changes

zhaohuabing requested a review from a team December 28, 2025 23:54

Merge branch 'main' into bechmark-scale-rps

f42b438

zirain reviewed Dec 29, 2025

View reviewed changes

charts/gateway-addons-helm/README.md Show resolved Hide resolved

zhaohuabing dismissed jukie’s stale review via ad69ceb December 29, 2025 09:30

address comments

f70ebad

Signed-off-by: Huabing(Robin) Zhao <zhaohuabing@gmail.com>

zhaohuabing force-pushed the bechmark-scale-rps branch from ad69ceb to f70ebad Compare December 29, 2025 09:34

zhaohuabing requested review from a team, jukie and zirain December 29, 2025 09:47

zirain approved these changes Dec 30, 2025

View reviewed changes

zhaohuabing added the cherrypick/release-v1.6.2 label Dec 30, 2025

arkodg approved these changes Dec 30, 2025

View reviewed changes

zhaohuabing merged commit 47a1dfa into envoyproxy:main Dec 30, 2025
95 of 101 checks passed

zhaohuabing deleted the bechmark-scale-rps branch December 30, 2025 03:08

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

bechmark: scale up RPS to test data plane CPU performance#7810

bechmark: scale up RPS to test data plane CPU performance#7810
zhaohuabing merged 12 commits intoenvoyproxy:mainfrom
zhaohuabing:bechmark-scale-rps

zhaohuabing commented Dec 24, 2025 •

edited

Loading

Uh oh!

codecov bot commented Dec 24, 2025 •

edited

Loading

Uh oh!

jukie commented Dec 28, 2025

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

zhaohuabing commented Dec 24, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

codecov bot commented Dec 24, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

jukie commented Dec 28, 2025

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

zhaohuabing commented Dec 24, 2025 •

edited

Loading

codecov bot commented Dec 24, 2025 •

edited

Loading