CloudDevelopersDevOpsNews

Kubecost Spot Commander Analyzes Cluster Workloads For Spot-Readiness

0

Kubecost has announced the release of Kubecost Spot Commander, a tool that supports organizations currently using – or looking to take advantage of – spot or preemptible nodes in Kubernetes clusters running across AWS, GCP, or Azure. Spot Commander determines which workloads are safe to run on spot and selects a new set of nodes to handle the split of spot-ready and non-spot-ready workloads.

Kubecost Spot Commander analyzes the workloads running on a cluster to determine spot-readiness using Kubecost’s Spot Checklist feature. The tool harnesses information from the Kubernetes API and heuristics to pinpoint workloads that are ready.

Once spot-ready workloads have been identified, Kubecost leverages its extensive knowledge (see what information is available via its Allocation API on GitHub) of an organization’s workload usage patterns and the node capacities and pricing of their cloud provider.

With this data, Spot Commander intelligently suggests the least expensive node types and quantities to support spot-ready and non-spot-ready workloads. The tool can also calculate overall savings for adopting the new cluster configuration by calculating the current run-rate of a cluster’s nodes and comparing it with the approximate run-rate of the new configuration.

To realize the estimated savings, Spot Commander updates the cluster configuration to include new spot nodes and reduce the number of on-demand nodes. Doing so ensures that only spot-ready workloads are scheduled on spot nodes.

Webb Brown, CEO and co-founder of Kubecost, said: “Kubecost Spot Commander is built to enable Kubernetes domain experts to more efficiently consider which workloads can be optimized as spot and preemptible nodes. We believe that the Kubecost recommendations delivered through Spot Commander, which include estimated savings, will motivate more infrastructure teams to start realizing significant cost reductions.”

Don't miss out great stories, subscribe to our newsletter.

Login/Sign up