LearnGlossary

What is Kubernetes cluster utilization?

Updated 25 October 2026 · Published by Zoorik

Kubernetes cluster utilization is the share of your nodes’ CPU and memory that pods request or use. It’s Kubernetes resource utilization across the whole cluster.

Requests or use

  • By requests: the sum of pod requests, divided by node capacity (what the nodes can give pods).
  • By use: the CPU and memory in use, divided by the same capacity. Metrics Server collects it. Run kubectl top to see this Kubernetes cluster usage per node and pod.

Pods can use more than they request. If a container sets only a limit, Kubernetes uses that limit as its request, unless a default request applies.

The scheduler places pods by requests. A node whose requests fill it takes no new pods, even when real use is low.

Ten nodes at 60%

The Amazon EKS guide gives this illustration: 10 nodes, each at 60% utilization. So 40% of the capacity you run isn’t used.

With same-size nodes, that’s 4 nodes’ worth, but no node is empty.

No node is empty, yet 4 nodes' worth of capacity is not used Illustration from the Amazon EKS best practices guide. A row of 10 equal outlined node columns, labelled 10 nodes, each at 60% utilization. Each column is filled 60% used, in violet at the bottom, and 40% not used, in amber at the top, labelled 60% each and 40% each. Below, one bar on the same width, labelled The same 10 nodes, added up. Its left part reads used, 60%, 6 nodes' worth. Its right part, in amber, reads not used, 40%, 4 nodes' worth. No node is empty, yet 4 nodes' worth of capacity is not used 10 nodes, each at 60% utilizationnot used40% eachused60% eachThe same 10 nodes, added upused60%, 6 nodes' worthnot used40%, 4 nodes' worth
Illustration from the Amazon EKS best practices guide: 10 same-size nodes, each at 60% utilization, then the same capacity added up on one bar.

Why it matters for your bill

  • You pay for each node whole, used or not. AKS cost analysis calls the unused part “idle charges”.
  • Requests set too high leave unused capacity, “a large factor in total cluster costs”, says the EKS guide. That’s overprovisioning.
  • Coral rebalances CPU and memory requests on AKS and EKS to match real use. Then the same pods fit on fewer nodes, and Coral drains the spare ones so your autoscaler can remove them.
  • The free assessment works out your own figure.

Read next: Kubernetes overprovisioning: what it is and how to measure it

Sources: Kubernetes, “Resource Management for Pods and Containers” (opens in a new tab); Kubernetes, “Resource metrics pipeline” (opens in a new tab); Amazon Web Services, “Compute and Autoscaling” (opens in a new tab) (Amazon EKS Best Practices Guide); Microsoft, “Azure Kubernetes Service (AKS) cost analysis” (opens in a new tab).

Find your own number

One line in your cloud console. One PDF report with your estimated saving at list price.

Estimate your saving

Talk to the builders

Thirty minutes with the people who build Coral.

Book a demo (opens in a new tab)

Or write to founders@zoorik.com