Overview
Overview
Koku is an open source cost management platform designed to identify, report, and analyze cost and usage data across clouds, containers, and hybrid environments. It is the upstream project for Red Hat Lightspeed Cost Management.
What Koku Measures
Koku provides more than 40 cost dimensions out of the box, including:
- Compute: CPU requests, usage, and limits; memory requests, usage, and limits; node capacity and allocatable resources
- Storage: Persistent volume capacity, requests, and usage
- Network: Network-attributed costs
- GPU: NVIDIA GPU allocation and usage (including OpenShift AI / MaaS workloads)
- Virtualization: OpenShift Virtualization VM-level cost tracking
- Organizational: Cluster, node, namespace, project, and tag-based grouping and allocation
All metering and cost data is available via REST API for integration with any billing, ERP, or business intelligence system.
Supported Providers
| Provider | Description |
|---|---|
| Amazon Web Services (AWS) | Cost and Usage Reports (CUR), including private offers and Savings Plans |
| Microsoft Azure | Cost exports, including Azure Marketplace |
| Google Cloud Platform (GCP) | BigQuery billing exports |
| OpenShift Container Platform | Prometheus-based metering via the koku-metrics-operator |
| OCP on AWS (ROSA) | Correlated OpenShift + AWS costs with automatic subscription distribution |
| OCP on Azure (ARO) | Correlated OpenShift + Azure costs with automatic subscription distribution |
| OCP on GCP | Correlated OpenShift + GCP costs |
Multi-Architecture Support
Koku is the only FinOps tool that supports metering and costing of OpenShift across all architectures that OpenShift runs on:
- x86-64
- ARM (aarch64)
- IBM Z (s390x)
- IBM LinuxONE (s390x)
- IBM POWER (ppc64le)
Key Capabilities
Cost Models
Cost models define how raw usage is priced. Koku supports:
- Tiered rates — usage-based pricing for CPU, memory, and storage
- Tag-based rates — rates applied by tag key/value pairs
- Markup — percentage-based markup on infrastructure costs
- Cost distribution — overhead allocation across platform, worker, storage, network, and GPU dimensions
Forecasting
Built-in cost forecasting uses historical usage patterns to project future costs, helping teams plan budgets and identify spending trends.
Fine-Grained RBAC
Koku’s role-based access control allows operators to aggregate all clusters, VMs, namespaces, and projects in a single instance while restricting what each user — or each tenant — can see. A tenant given read access sees only their own namespaces and costs, never another tenant’s data.
Resource Optimization
Beyond cost tracking, Koku includes a resource optimization feature (ROS) that provides rightsizing recommendations for containers, deployments, and jobs. Two recommendation types are available: cost-optimized and performance-optimized.
Deployment Modes
Koku supports two deployment models:
- Cloud (SaaS) — Used by console.redhat.com. Heavy aggregation runs on Trino over Parquet files in S3; summary data lands in PostgreSQL for API queries.
- On-premise — Fully self-contained with PostgreSQL only (no Trino). Deployed via Helm using the cost-onprem-chart. All data stays inside the operator’s perimeter.