KEP-2724: Add multi-level topology aware scheduling design#9243
KEP-2724: Add multi-level topology aware scheduling design#9243Huang-Wei wants to merge 1 commit intokubernetes-sigs:mainfrom
Conversation
k8s-ci-robot
added
release-note
✅ Deploy Preview for kubernetes-sigs-kueue canceled.
|
k8s-ci-robot
added
size/L
Add the design for multi-level TAS, which extends two-level scheduling to support N slice layers across deeper topology hierarchies (e.g., datacenter → block → rack → host).
Huang-Wei
force-pushed
the
tas-multi-layer-KEP
branch
from
588ff08 to
ba0b5bb
Compare
|
I assume the implementation will be a follow-up, so the release note for this PR should be dropped. |
|
/release-note-edit |
k8s-ci-robot
added
release-note-none
| - The value of `kueue.x-k8s.io/podset-slice-size` has to be a numeric value greater or equal | ||
| than 1. It has to evenly divide the size of a PodSet. | ||
| - The above 2 `podset-slice-*` rules apply to additional slice layers (`kueue.x-k8s.io/podset-slice-required-topology-[X]`) | ||
| as well. `[X]` can be up to `2`. |
There was a problem hiding this comment.
It this just a "sanity" limitation on the API, but the algorithm is generic? In other words, would the algo code need to be changed to support extra layers? Not a blocker, just curious on how generic the algo is.
There was a problem hiding this comment.
Yes, the algorithm is generic.
In other words, would the algo code need to be changed to support extra layers?
No core changes needed if we want to support extra layers.
I will submit the PR this week.
k8s-ci-robot
added
the
do-not-merge/hold
k8s-ci-robot
added
the
lgtm
|
LGTM label has been added. DetailsGit tree hash: 5ae36072db01f2dfaf51930b369c0d8b94a75294 |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: Huang-Wei, mimowo The full list of commands accepted by this bot can be found here. The pull request process is described here DetailsNeeds approval from an approver in each of these files:
Approvers can indicate their approval by writing |
k8s-ci-robot
added
the
approved
5 participants
What type of PR is this?
/kind feature
What this PR does / why we need it:
Add the design for multi-level TAS, which extends two-level scheduling to support N slice layers across deeper topology hierarchies (e.g., datacenter → block → rack → host).
KEP for #9046.
Which issue(s) this PR fixes:
Fixes #
Special notes for your reviewer:
Does this PR introduce a user-facing change?