Costfluent Docs
  • Pricing
  • Changelog
  • costfluent.com
  • Documentation
  • Guides
  • API Reference
Product
  • Pricing
  • Security
  • Changelog
Developers
  • API reference
  • Terraform
  • llms.txt
Legal
  • Privacy
  • Terms
  • Imprint
Get started
Concepts
Connect a source
Reporting
Allocation
Planning
Optimization
    InsightsSavingsCommitmentsKubernetes efficiency
Administration
Developers
Optimization

Kubernetes efficiency

See what each cluster, namespace and workload cost, how much of it was idle capacity, and how much of the CPU and memory it requested it used.

Kubernetes sits under Optimization in the product's sidebar. For the last 30 days it shows each cluster in the workspace, then the namespaces or workloads of the cluster you choose, with their cost and how much of what they requested they used.

Before you start

  • Any team role can view the page. It is available on every plan.
  • The figures come from the Kubernetes agent's reports (Kubernetes). The cluster's connection must be in the workspace. Cost and efficiency appear the morning after the agent first reports.

Read your cluster efficiency

  1. Open Kubernetes under Optimization.
  2. Read Clusters for each cluster's Cost, its Idle cost and its efficiency.
  3. Choose a cluster's name to break it down below.
  4. Choose By namespace or By workload. A workload is the pod's controller, such as Deployment/api, or Pod/<name> for a pod without one.

Each row shows:

  • CPU used and Memory used: measured usage over the request.
  • CPU peak and Memory peak: each container's highest daily usage, held all day, over the request. A peak well above the request is a workload that needs more than it asks for.
  • CPU requested (core-hours) and Memory requested (GiB-hours): what the namespace or workload requested over the window.

How it behaves

  • Idle is node capacity no container was allocated. It reads as Idle capacity and requests nothing, so it has no efficiency.
  • A container is charged for the larger of its request and its usage (Kubernetes), so a workload using more than it requested shows above 100%.
  • GPUs are not in the efficiency figures: the agent does not measure GPU usage.
  • Only the workspace's Kubernetes rows are read. The cloud cost of the same nodes stays out of these figures.
  • A workspace whose Kubernetes cost is in more than one currency shows no figures until currency conversion is on (Currency).
  • At most the 500 rows with the largest cost are listed, and the page says when more matched.

Related

  • Kubernetes
  • Explore costs
CommitmentsOrganization and workspaces
On this page
  • Before you start
  • Read your cluster efficiency
  • How it behaves
  • Related