specification: FinOps Framework
specificationVersion: '1.0'
schema: https://www.finops.org/framework/
alignedWith:
  framework: FinOps Foundation Framework
  frameworkUrl: https://www.finops.org/framework/
  dataSpec: FOCUS
  dataSpecVersion: '1.3'
  dataSpecUrl: https://focus.finops.org/focus-specification/v1-3/
provider: Containerd
providerId: containerd
publisherName: Cloud Native Computing Foundation (CNCF)
serviceCategory: Container Runtime / Open Source
created: '2026-05-04'
# Provenance stamped 2026-08-11: this artifact was written by the API Evangelist
# bulk sweep dated 2026-05-04, not harvested from the provider. See roadmap#35.
method: generated
modified: '2026-05-05'
reconciled: true
tags:
  - Cloud Native
  - Container Runtime
  - Kubernetes
  - OSS
  - FinOps
  - FOCUS
description: 'FOCUS-aligned FinOps for Containerd: CNCF-graduated OSS runtime with no per-call or license
  cost. FinOps focuses on the underlying compute / storage / network it consumes (Kubernetes node hours,
  registry-pull egress) rather than a vendor invoice from containerd itself.'
sources:
  - https://containerd.io
  - https://github.com/containerd/containerd
billingModel:
  pricingCategory: Open Source / Free
  billingFrequency: Not Applicable
  billingCurrency: USD
  chargeCategories:
    - Usage
focusColumns:
  ServiceName: Containerd
  ServiceCategory: Container Runtime
  ProviderName: Containerd Project
  PublisherName: Cloud Native Computing Foundation (CNCF)
  InvoiceIssuerName: Not Applicable
  BillingCurrency: USD
  ChargeCategory: Usage
meters:
  - name: node_hours
    description: Compute hours of nodes running containerd (proxy meter — billed by the underlying cloud
      provider, not by containerd)
    unit: instance-hour
    aggregation: sum
    dimensions:
      - cloud_provider
      - instance_type
      - cluster
  - name: image_pull_egress
    description: Egress from container registries during image pulls
    unit: GB
    aggregation: sum
    dimensions:
      - registry
      - region
  - name: container_runtime_count
    description: Containers / pods executed via the containerd CRI surface (operational meter, no charge)
    unit: container
    aggregation: count
    dimensions:
      - namespace
      - cluster
principles:
  - name: Visibility
    description: Containerd does not produce invoices; surface infrastructure cost via the underlying
      cloud provider's billing exports and via Prometheus metrics scraped from the kubelet / containerd.
  - name: Allocation
    description: Allocate via cluster / namespace / pod labels and cloud-provider node tags; align with
      Kubecost or OpenCost for per-workload mapping.
  - name: Optimization
    description: Reduce image-pull egress with registry mirrors / pull-through caches; tune `max_concurrent_downloads`;
      consider lazy-loading snapshotters (stargz) for large images; right-size node pools.
  - name: Accountability
    description: Platform / SRE teams own cluster-level spend; product teams own per-namespace footprint
      and image hygiene that drives storage and pull egress.
maintainers:
  - FN: Kin Lane
    email: kin@apievangelist.com