Add docs for KEP-4176 new static policy DistributeCPUsAcrossCores #47014
Conversation
k8s-ci-robot
added
do-not-merge/work-in-progress
👷 Deploy Preview for kubernetes-io-vnext-staging processing.
|
✅ Pull request preview available for checkingBuilt without sensitive environment variables
To edit notification comments on pull requests, go to your Netlify site configuration. |
|
/sig node |
k8s-ci-robot
added
the
sig/node
|
Hello @Jeffwan 👋 please take a look at Documenting for a release - PR Ready for Review to get your PR ready for review before Tuesday July 16th 2024 18:00 PST. Thank you! |
|
Hi @Jeffwan, a gentle reminder that tomorrow is the deadline for having your Docs PR ready for review. Please take a look at Documenting for a release - PR Ready for Review to get your PR ready for review before tomorrow, Tuesday, July 16th, 2024 18:00 PST. |
|
@Princesso Thanks! the code change has been merged and I will finish the doc change today. |
k8s-ci-robot
added
size/S
Jeffwan
changed the title
k8s-ci-robot
removed
the
do-not-merge/work-in-progress
|
@Princesso Please check the doc update. @ffromani is the KEP reviewer. Feel free to take a look.
|
content/en/docs/tasks/administer-cluster/cpu-management-policies.md
Outdated
Show resolved
Hide resolved
|
/cc |
content/en/docs/tasks/administer-cluster/cpu-management-policies.md
Outdated
Show resolved
Hide resolved
content/en/docs/tasks/administer-cluster/cpu-management-policies.md
Outdated
Show resolved
Hide resolved
k8s-ci-robot
added
the
lgtm
|
LGTM label has been added. Git tree hash: b55963009724add254afc0a13277db3841bca7bc
|
| might be less effective when the system is heavily loaded. Under such conditions, | ||
| the benefit of reducing contention diminishes. Conversely, default behavior | ||
| can help in reducing inter-core communication overhead, potentially providing | ||
| better performance under high load conditions. |
There was a problem hiding this comment.
While the virtual cores are backed by more physical cores, the application can potentially get performance boost thanks to more L2, L3 caches to use.
However, these physical cores may come from different sockets.
The assumption of better performance is only true when the following factors are met:
- there are heavy inter-thread communication in the application;
- the load is not very high, or else you may want to use more cores rather than fewer,
which is the newly added policy.
Am I understanding this correctly?
There was a problem hiding this comment.
Exactly. Our benchmarks show that this approach works well in low-load scenarios. We employ this strategy for serverless workloads, where it's uncommon for all applications to be busy simultaneously. This policy performs effectively in such environments.
We've also tested the potential impact of cross-socket issues and found no significant difference for our application. However, other applications may experience varying results. In the alpha version, we did not explicitly consider whether CPUs were across sockets or not. In future versions, we plan to improve this aspect and clarify the behavior.
|
LGTM! |
|
@tengqm Please take another look. I can explain more details if there's anything not clear yet. |
|
/lgtm |
|
@Princesso Please help approve the PR. I think all have finish reviewing the change. |
|
/approve |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: Princesso The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
k8s-ci-robot
added
the
approved
this is a placeholder PR for KEP-4176 kubernetes/enhancements#4176
/cc @Princesso