What would you like to be added?

We would like to start a discussion and hopefully kick off an effort to pivot scheduling from the pod-based to workload-based. There are many things we are certain about, but there are also plenty of unanswered questions. We want to start from framing a problem, agreeing on goals and opening a discussion how to achieve those goals.

Please review [External] Workload-aware scheduling doc and share your thoughts there.

Please note that there are two separate tabs, the first focused on framing the problem, the second on drawing a vision, which should help us to imagine how things could be organized even if we don't know all the details yet.

Why is this needed?

There has been a lot of discussions and thoughts about making native scheduling workload-aware. It's needed not only to improve scheduling capabilities for workloads (finding a better placement), but also to make all necessary components aware of workloads to protect their integrity though their entire lifecycle. The second goal may be even more important as this is a big missing element in the Kubernetes ecosystem which even custom schedulers cannot address on their own.