[Exception Request] DRA: ResourceClaim Support for Workloads (KEP-5729)

[Jon Huhn]

Hello,

I’m requesting a code freeze exception for KEP-5729: DRA: ResourceClaim Support for Workloads.

Enhancement name: DRA: ResourceClaim Support for Workloads

Enhancement status: alpha

SIG: Scheduling

k/enhancements repo issue #: #5729

PR #’s: #136989

Additional time needed: 3 days (21 March)

Reason this enhancement is critical for this milestone:

Managing multi-host workloads is necessary for AI/ML. This feature builds on the workload aware scheduling to enabling lifecycle management of the related resources along with the PodGroup. It is important to understand how these resources will be managed and getting this to alpha state helps us iron out any gaps and additional requirements.

Risks from adding code late:

Low risk. The relevant tests on the PR have been stable. All of the functionality is gated on the DRAWorkloadResourceClaims feature gate.

Risks from cutting enhancement: 

This is a key feature for the future of workload-aware scheduling. This feature is needed to unblock experiments for critical advancements in large-scale AI/ML workload orchestration in Kubernetes.

Thanks,

Jon Huhn

[John Belamaric]

+1 this has most of the approvals it needs and we would really like to prevent it slipping a whole release 

--
You received this message because you are subscribed to the Google Groups "wg-device-management" group.
To unsubscribe from this group and stop receiving emails from it, send an email to wg-device-manage...@kubernetes.io.
To view this discussion visit https://groups.google.com/a/kubernetes.io/d/msgid/wg-device-management/LV0PR21MB650712CB0165B0E00DAD882A9F4EA%40LV0PR21MB6507.namprd21.prod.outlook.com.

[Dominik Marciński]

+1

This feature is important for developing workload-aware scheduling in future releases.

To view this discussion visit https://groups.google.com/a/kubernetes.io/d/msgid/wg-device-management/CAC_RkjyedQtDg7SQN1y1ZVU8-zTA6syGb9wOfMjVhW6SxOp1rQ%40mail.gmail.com.