Open Spenhouet opened 3 years ago
This issue is stale because it has been open for 90 days with no
activity. It will automatically close after 30 more days of
inactivity. Reopen the issue with /reopen
. Mark the issue as
fresh by adding the comment /remove-lifecycle stale
.
/remove-lifecycle stale
This issue is stale because it has been open for 90 days with no
activity. It will automatically close after 30 more days of
inactivity. Reopen the issue with /reopen
. Mark the issue as
fresh by adding the comment /remove-lifecycle stale
.
/reopen /remove-lifecycle stale
@Spenhouet: Reopened this issue.
/triage accepted
This issue is stale because it has been open for 90 days with no
activity. It will automatically close after 30 more days of
inactivity. Reopen the issue with /reopen
. Mark the issue as
fresh by adding the comment /remove-lifecycle stale
.
Problem Given many events of the same type but for example for different groups. The group could be anything provided by the event producer e.g. a customer.
In our specific use-case we are limited by hardware resources (GPUs specifically). There might be 1000 requests by group 1 and only 10 requests by group 2. We would like the ability to apply some fairness factor so that group 2 will be scheduled with a higher priority than group 1.
Additional context (optional)
I'm not sure if something like this already exists. Kubernetes seems to have something in this direction: Pod Priority and Preemption If this does not exist yet, then I would like to make this a feature request.
It is also open where the interface for this is. Such a fairness factor could be automatically calculated and managed by Knative but could also just be provided by the event producer.