Kompass HiberScale technology provisions nodes in advance and prewarms them by downloading container images and prestarting the OS. These nodes are then placed into AWS hibernation, where they remain ready to resume within seconds.
When a Pod gets unscheduled, like during a demand spike, the prewarmed state enables Pods to start running applications up to 5 times faster than native Kubernetes autoscalers.
This technology enables Kompass to power both Headroom Reduction and Spot Management—solutions that minimize infrastructure cost while preserving workload performance.
For more information about AWS hibernating, see Hibernate your Amazon EC2 instance.
The following video shows how HiberScale technology reduces cost in a comparative stress test:
HiberScale creates hibernated, On-Demand VMs:
in node pools linked to the protected workload
of multiple instance types (the same types that are configured in the node pools associated with the protected workload)
across multiple zones
HiberScale integrates seamlessly with Karpenter autoscaler or Cluster Autoscaler for launching and draining nodes.