DevConf.US 2024 - Kubernetes Autoscale Everything!

Topics

• machine learning and autoscaler
  • https://github.com/kubernetes/autoscaler/blob/master/cluster-autoscaler/proposals/provisioning-request.md
  • how to handle GPUs
• how will CAS handle DRA?
• bottlenecks in CAS, where are they?
• moving running PODs, is this possible?
  • checkpointing is work in this direction
  • https://github.com/kubernetes/enhancements/pull/4305
• proactive vs reactive autoscaling
• how to scale based on other sources of data, eg bandwidth
  • https://keda.sh/

DevConf.CZ 2024 - Kubernetes Autoscale Everything!

Topics

• workload placement
  • consolidating resources
  • CEL
    • might be useful for admission webhooks replacement
  • might also be helpful pod scheduling readiness
    • https://kubernetes.io/docs/concepts/scheduling-eviction/pod-scheduling-readiness/
  • kueue project might also work
    • https://github.com/kubernetes-sigs/kueue
• autoscaling vs autoprovisioning
• VPA - in-place scaling, has anyone run this?
  • dynamic resource allocation, might be related in the future
    • https://kubernetes.io/docs/concepts/scheduling-eviction/dynamic-resource-allocation/
• karpenter and cluster-api
  • did they consider capi for a backend?
• karpenter with provisioning request
  • related topic, https://www.redhat.com/en/blog/how-full-is-my-cluster-part-6-proactive-node-autoscaling
• hybrid cloud
  • will we see kubernetes spanning across cloud providers?