Kubernetes
Container cost allocation down to namespace, deployment, and team level, with rightsizing and idle-pod recommendations from measured container utilisation.
At a glance
| Route | /kubernetes |
| Group | Analyze |
| Page permission | integrations |
| Active detectors | 13 |
What it is
Container cost allocation and workload optimisation. It attributes cluster cost down to namespace, deployment, and team level, and produces rightsizing, idle-pod, and autoscaling recommendations from measured container utilisation.
Who it is for
Platform teams running Kubernetes, and FinOps practitioners who need to allocate shared cluster cost to the teams actually consuming it.
How it works
Three ingestion paths
Thirteen active Kubernetes detectors run against that data — including two that convert the attribution pipeline's own rightsizing and idle-pod recommendations into first-class findings in the Opportunity Queue, alongside every other provider.
Features
- Namespace, deployment, and team-level cost attribution
- Three ingestion paths: Kubecost, OpenCost, or the bundled in-cluster agent
- Container-level CPU and memory request-versus-usage analysis
- Idle pod detection
- Rightsizing recommendations from measured container utilisation
- Autoscaling recommendations
- DaemonSet overhead analysis
- Spot migration candidate identification
- Orphaned persistent volume detection
- Findings routed into the Opportunity Queue alongside every other provider
How to use it
Choose an ingestion path
Kubecost or OpenCost is simplest where already deployed. The bundled agent is the lightest option otherwise.
Reconcile namespace attribution against total cluster cost
If these do not match, everything downstream is wrong.
Map namespaces to teams
Using Cost Allocation and Ownership. This is what makes cluster cost chargeable.
Work the rightsizing findings
Request-versus-usage mismatch is nearly always the largest single source of cluster waste — over-requesting is endemic to Kubernetes because the cost of getting it wrong is invisible.
Address idle pods and orphaned persistent volumes
Pure waste with negligible risk. Easy early wins.
Evaluate spot migration candidates
For workloads that tolerate interruption.
Why it matters
Kubernetes is where cost attribution usually fails completely. A cluster appears on the bill as a handful of large instances with no indication of which team or service consumed them — so it is either allocated by a crude split, or not at all.
Namespace and workload-level attribution makes cluster cost chargeable, and request-versus-usage analysis addresses the structural over-requesting that is endemic to Kubernetes.
Connects to
- Ingestion configured in Integrations
- Attribution feeds Cost Allocation and Unit Economics
- Findings reach Opportunity Queue
- The in-cluster agent is documented in Kubernetes Agent
API Monitoring
The invoice's view of the API layer — spend across gateways, the isolated cost of error traffic, and pricing-model waste detection.
Predictive Analytics
Forecasting with two competing models and automatic selection, honest error metrics, risk scenario modelling, and change-impact simulation.