add benchmark test for `ScheduleOne` #124728

chengjoey · 2024-05-07T14:12:08Z

What type of PR is this?

/kind cleanup
/kind flake

What this PR does / why we need it:

add benchmark test for ScheduleOne
testing 5000 pods and nodes the performance of selecting the fit node for the pod
#124709

Special notes for your reviewer: @alculquicondor @sanposhiho @AxeZhan

Does this PR introduce a user-facing change?

None

k8s-ci-robot · 2024-05-07T14:12:17Z

This issue is currently awaiting triage.

If a SIG or subproject determines this is a relevant issue, they will accept it by applying the triage/accepted label and provide further guidance.

The triage/accepted label can be added by org members by writing /triage accepted in a comment.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository.

k8s-ci-robot · 2024-05-07T14:12:18Z

Hi @chengjoey. Thanks for your PR.

I'm waiting for a kubernetes member to verify that this patch is reasonable to test. If it is, they should reply with /ok-to-test on its own line. Until that is done, I will not automatically test new commits in this PR, but the usual testing commands by org members will still work. Regular contributors should join the org to skip this step.

Once the patch is verified, the new status will be reflected by the ok-to-test label.

I understand the commands that are listed here.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository.

k8s-ci-robot · 2024-05-07T14:12:21Z

Welcome @chengjoey!

It looks like this is your first PR to kubernetes/kubernetes 🎉. Please refer to our pull request process documentation to help your PR have a smooth ride to approval.

You will be prompted by a bot to use commands during the review process. Do not be afraid to follow the prompts! It is okay to experiment. Here is the bot commands documentation.

You can also check if kubernetes/kubernetes has its own contribution guidelines.

You may want to refer to our testing guide if you run into trouble with your tests not passing.

If you are having difficulty getting your pull request seen, please follow the recommended escalation practices. Also, for tips and tricks in the contribution process you may want to read the Kubernetes contributor cheat sheet. We want to make sure your contribution gets all the attention it needs!

Thank you, and welcome to Kubernetes. 😃

k8s-ci-robot · 2024-05-07T14:12:30Z

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by: chengjoey
Once this PR has been reviewed and has the lgtm label, please assign alculquicondor for approval. For more information see the Kubernetes Code Review Process.

The full list of commands accepted by this bot can be found here.

Needs approval from an approver in each of these files:

pkg/scheduler/OWNERS

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

alculquicondor · 2024-05-07T14:56:46Z

pkg/scheduler/schedule_one_test.go

+// simulating a scenario with 5000 nodes and 5000 pods. at the same time, each pod corresponds to a node one by one
+// and the corresponding node is selected for each pod concurrently to test
+// the performance of findNodesThatFitPod under high concurrency
+func Benchmark_findNodesThatFitPod(b *testing.B) {


can you change this to test at scheduleOne level instead?

Also, maybe something like Benchmark_scheduleOne_onePodPerNode

ok, I'll give it a try

sanposhiho · 2024-05-08T08:34:40Z

Instead of having this benchmark test, can we somehow improve scheduler_perf (e.g., add test cases) so that it can catch the throughput improvement of preallocation?

testing 5000 pods and nodes the performance of selecting the fit node for the pod Signed-off-by: joey <[email protected]>

chengjoey · 2024-05-09T12:55:46Z

pkg/scheduler/schedule_one_test.go

+ start := time.Now()
+ logger := klog.FromContext(ctx)
+ // Send pods to be scheduled.
+ for _, p := range pods {
+ sched.SchedulingQueue.Add(logger, p)
+ }
+ wg.Wait()
+ cancel()
+ cost := time.Since(start).Seconds()
+ throughput := float64(nodeNum) / cost
+ if throughput < benchmarkThroughput {
+ b.Fatalf("ScheduleOne_onePodPerNode: %d nodes, %d pods, cost: %f seconds, throughput: %f pods/second, below the baseline throughput standard: %f pods/second",
+ nodeNum, nodeNum, cost, throughput, benchmarkThroughput)
+ }


@sanposhiho Is it possible to add benchmarkThroughput to this benchmark test, at least to verify that the code quality is not lower than a certain standard?

Of course, the throughput in the actual production environment will be affected by more factors such as network memory. I wonder if we can use kwok in e2e to help us better simulate the actual environment. The current benchmark test focuses more on code quality.

At least in this benchmark test, we can see that preallocation for NodeToStatusMap helps improve throughput.

I'm not sure why that needs to be this type of benchmark test. scheduler_perf is a basic tool we usually use to measure performance. It also has a monitoring visualization provided via perf-dash. I believe we should create this kind of benchmark tests only when it's somehow not possible to measure the performance within scheduler_perf. So, that's why I suggested we should try to seek the possibility of improving scheduler_perf first.
(Besides that, either way, IIUC, Benchmark_XXX tests aren't run in prow jobs.)

But, I like your idea of keeping a certain throughput by implementing as tests. However, for that idea too, I'd like to see something like:

we define a desired throughput for each test case of scheduler_perf.

implement something so that we easily notice the degradation based on ^. e.g., the CI fails in PRs, we get alerts in Slack if the scheduler of the master branch is slow, etc.

Starting a thread -
https://kubernetes.slack.com/archives/C09QZTRH7/p1715262959575039

k8s-ci-robot added needs-ok-to-test Indicates a PR that requires an org member to verify it is safe to test. needs-priority Indicates a PR lacks a `priority/foo` label and requires one. labels May 7, 2024

k8s-ci-robot added sig/scheduling Categorizes an issue or PR as relevant to SIG Scheduling. and removed do-not-merge/needs-sig Indicates an issue or PR lacks a `sig/foo` label and requires one. labels May 7, 2024

k8s-ci-robot requested review from denkensk and sanposhiho May 7, 2024 14:12

alculquicondor reviewed May 7, 2024

View reviewed changes

chengjoey force-pushed the benchmark-for-findNodesThatFitPod branch from 2b97feb to 1a477df Compare May 8, 2024 03:48

k8s-ci-robot added size/L Denotes a PR that changes 100-499 lines, ignoring generated files. and removed size/M Denotes a PR that changes 30-99 lines, ignoring generated files. labels May 8, 2024

chengjoey force-pushed the benchmark-for-findNodesThatFitPod branch from 1a477df to d8e2d8f Compare May 8, 2024 03:49

chengjoey changed the title ~~add benchmark test for findNodesThatFitPod~~ add benchmark test for ScheduleOne May 8, 2024

chengjoey force-pushed the benchmark-for-findNodesThatFitPod branch from d8e2d8f to ccaf6f5 Compare May 8, 2024 05:47

add benchmark test for ScheduleOne

bd58f0e

testing 5000 pods and nodes the performance of selecting the fit node for the pod Signed-off-by: joey <[email protected]>

chengjoey force-pushed the benchmark-for-findNodesThatFitPod branch from ccaf6f5 to bd58f0e Compare May 9, 2024 12:46

chengjoey commented May 9, 2024

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

add benchmark test for `ScheduleOne` #124728

add benchmark test for `ScheduleOne` #124728

chengjoey commented May 7, 2024 •

edited

k8s-ci-robot commented May 7, 2024

k8s-ci-robot commented May 7, 2024

k8s-ci-robot commented May 7, 2024

k8s-ci-robot commented May 7, 2024

alculquicondor May 7, 2024

alculquicondor May 7, 2024

chengjoey May 7, 2024

chengjoey May 8, 2024

sanposhiho commented May 8, 2024

chengjoey May 9, 2024

chengjoey May 9, 2024

sanposhiho May 9, 2024 •

edited

sanposhiho May 9, 2024

sanposhiho May 9, 2024

add benchmark test for ScheduleOne #124728

Are you sure you want to change the base?

add benchmark test for ScheduleOne #124728

Conversation

chengjoey commented May 7, 2024 • edited

What type of PR is this?

What this PR does / why we need it:

Special notes for your reviewer: @alculquicondor @sanposhiho @AxeZhan

Does this PR introduce a user-facing change?

k8s-ci-robot commented May 7, 2024

k8s-ci-robot commented May 7, 2024

k8s-ci-robot commented May 7, 2024

k8s-ci-robot commented May 7, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

sanposhiho commented May 8, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

sanposhiho May 9, 2024 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

add benchmark test for `ScheduleOne` #124728

add benchmark test for `ScheduleOne` #124728

chengjoey commented May 7, 2024 •

edited

sanposhiho May 9, 2024 •

edited