Luca BertonThe Kubernetes Cost Problem
Kubernetes makes it easy to deploy workloads — and easy to waste money. The average Kubernetes cluster runs at 20-35% resource utilisation, meaning 65-80% of cloud spend on compute is wasted. For regulated enterprises running dozens of clusters across multiple environments, this waste compounds into millions per year.
Kubernetes FinOps brings financial accountability to container infrastructure: who's spending what, where's the waste, and how do we fix it without compromising performance or compliance.
The FinOps Framework for Kubernetes
1. Inform — Cost Visibility
- Cost allocation: Map every dollar of cloud spend to a team, project, and environment using Kubernetes labels
- Shared cost distribution: Allocate cluster overhead (control plane, monitoring, networking) fairly across teams
- Idle cost identification: Separate requested-but-unused resources from actually-utilised resources
- Multi-cloud visibility: Unified cost view across AWS EKS, Azure AKS, GCP GKE, and on-premise clusters
2. Optimise — Reduce Waste
- Right-sizing: Adjust resource requests/limits based on actual usage (most pods request 3-5x what they use)
- Autoscaling: HPA (horizontal pod autoscaler), VPA (vertical pod autoscaler), KEDA (event-driven) — scale based on demand
- Spot/preemptible instances: Run non-critical and stateless workloads on spot instances (60-90% savings)
- Node optimisation: Right-size node pools, use cluster autoscaler aggressively, consider ARM instances
- Storage cleanup: Identify and delete orphaned PVCs, unused snapshots, and oversized volumes
3. Operate — Governance
- Showback: Show teams what they're spending (visibility, no billing)
- Chargeback: Bill teams for their actual consumption (financial accountability)
- Budgets and alerts: Set spend budgets per team/namespace with alerts at 80% and 100%
- Policy enforcement: Require resource requests/limits (via Kyverno/OPA), maximum pod counts, namespace quotas
Tools Comparison
| Tool | Type | Key Strength | Cost |
|---|---|---|---|
| OpenCost | CNCF Sandbox, open source | Real-time cost allocation | Free |
| Kubecost | Commercial (OpenCost fork) | Savings recommendations, governance | Free tier + paid |
| CAST AI | Automation platform | Automated right-sizing + spot | % of savings |
| ScaleOps | Automation platform | Autonomous resource management | Enterprise pricing |
📘 Book
Kubernetes Recipes
Practical guide for container orchestration and deployment — hands-on patterns you can use today.
View on Amazon →Quick Wins
- Enforce resource requests/limits via policy — prevents unbounded resource consumption
- Deploy OpenCost or Kubecost — visibility is the foundation; you can't optimise what you can't see
- Right-size the top 10 workloads — typically captures 60-80% of the savings opportunity
- Enable cluster autoscaler with aggressive scale-down (10 minutes unneeded → scale down)
- Move dev/staging to spot instances — immediate 60%+ savings with minimal risk
FinOps
Kubernetes
cost management
cost optimisation
Kubecost
regulated enterprises
cloud costs