FinOps & Cost Management

FinOps is the discipline of running cloud spend like a budget — knowing where dollars go, attributing them to teams / features, and making informed tradeoffs.

Pre-deploy cost estimation

• ★ Infracost — show cost diff on Terraform / OpenTofu / Pulumi PRs. The default for "this PR will cost $X more / month."
• Pulumi Cost Estimation — built-in to Pulumi.
• Terraform Cloud Cost Estimation — built-in to TFC paid tiers.
• Spacelift / env0 / Scalr — IaC pipelines with cost estimation included.

Kubernetes cost attribution

• ★ OpenCost — open-source k8s cost monitoring; CNCF; the foundation. The default.
• ★ Kubecost — OpenCost-derived commercial product; richer UI / reports / alerts. Free tier available.
• Cast.ai — k8s autoscaler + cost optimizer; commercial.
• Karpenter — AWS open-source autoscaler; reduces cost via better scheduling.

Cloud cost dashboards

• AWS Cost Explorer + Cost Anomaly Detection — built-in.
• GCP Cloud Billing reports — built-in.
• Azure Cost Management — built-in.
• ★ Vantage — multi-cloud cost analytics; modern UI; generous free tier.
• CloudZero, Apptio Cloudability, Anodot — paid alternatives.
• Komiser (open source) — AWS / GCP / Azure inventory + cost dashboard.
• Resoto — open-source cloud asset graph + cost.
• Steampipe — query cloud as SQL; build your own cost dashboards.

Tagging / chargeback

• Cost allocation tags — every resource tagged with team, environment, cost-center, application. Make this a policy enforced by IaC scanners.
• Budgets and alerts at the account / project / namespace level.
• Showback / chargeback — most teams start with showback (visibility) before chargeback (actual cross-charging).

Reserved / savings plans

• AWS Savings Plans / Reserved Instances — commit to save 30–70%.
• GCP Committed Use Discounts — same idea.
• Azure Reservations — same.
• Spot / Preemptible instances — 60–90% off if you can handle interruption.
• Tools like Spot.io automate spot fallback.

Common high-impact wins

• ★ NAT Gateway — AWS's silent killer. Use S3 VPC endpoints; review traffic.
• ★ Egress — replace S3 with R2 (zero egress); review CloudFront usage.
• Idle resources — old EBS volumes, unused IPs, abandoned RDS / EKS / VPC. Komiser surfaces these.
• Over-provisioned RDS / EC2 — right-size after you have a few weeks of metrics.
• Logs / monitoring spend — Datadog / Splunk / CloudWatch can cost more than compute. Sample, archive, downsample.
• Dev / staging running 24/7 — auto-stop nights and weekends.

Patterns to adopt

• ★ Cost in code review. Infracost on every PR. People internalize cost as a normal review concern.
• Budget alerts at 50 / 80 / 100% of expected spend.
• Tag enforcement via policy — Kyverno / Gatekeeper / Cloud-native (AWS Config / GCP Policy / Azure Policy).
• Quarterly waste sweep. A few hours of scanning old resources usually finds a meaningful savings number.
• Right-size based on data, not gut. Use Cost Explorer / Kubecost / Vantage data.

Pick this if…

• PR-time cost estimation: Infracost.
• k8s cost attribution, free: OpenCost.
• k8s cost attribution, polished UI: Kubecost.
• Multi-cloud cost dashboard: Vantage.
• OSS asset / cost inventory: Komiser or Resoto.
• Reserved-capacity automation: Spot.io or your cloud's native savings plans.