FinOps - Cloud Cost as an Engineering Discipline

Learning Objectives

• Understand the FinOps framework and the cultural shift it requires
• Implement cost visibility with granular tagging and allocation
• Identify and eliminate waste: idle resources, oversized instances, unused storage
• Configure showback and chargeback models for team-level accountability
• Establish unit economics: cost per request, cost per deployment, cost per customer

Requirements

You are required to implement a FinOps practice for the infrastructure built throughout the Advanced level:

1. Cost Visibility and Tagging
   • Design and enforce a tagging strategy across all cloud resources:
     • Required tags: team, environment, service, cost-center, owner
     • Implement tag enforcement via AWS Config Rules, Azure Policy, or GCP Organization Policies
     • Audit existing resources for missing tags and remediate
   • Configure cloud cost explorer or a third-party tool (Infracost, OpenCost, or Kubecost):
     • Break down costs by team, environment, and service
     • Identify the top 5 most expensive resources
     • Generate a weekly cost report
2. Kubernetes Cost Allocation with Kubecost
   • Deploy Kubecost on your cluster
   • Configure namespace-level cost allocation
   • Identify the most expensive workloads per namespace
   • Configure cost alerts: alert when a namespace exceeds its monthly budget
   • Document the cost of running your full DevOpsDiary platform per day
3. Waste Elimination
   • Audit and eliminate:
     • Idle EC2/VMs (running but no traffic for 7+ days)
     • Unattached EBS volumes, unused Elastic IPs, orphaned snapshots
     • Oversized instances: use AWS Compute Optimizer or equivalent recommendations
     • Over-provisioned Kubernetes resources: compare requests vs. actual usage with VPA recommendations
   • Document every eliminated resource: what it was, monthly cost, why it existed
   • Calculate total monthly savings achieved
4. Rightsizing and Autoscaling
   • Implement Kubernetes Vertical Pod Autoscaler (VPA) in recommendation mode
   • Review VPA recommendations and apply rightsized requests/limits
   • Configure Kubernetes Cluster Autoscaler or Karpenter to scale nodes based on actual demand
   • Implement KEDA (Kubernetes Event-Driven Autoscaling) for a workload that scales to zero when idle
   • Measure cost reduction before and after autoscaling implementation
5. Showback, Chargeback, and Unit Economics
   • Implement a showback model: generate a monthly report showing each team's cloud spend
   • Design a chargeback model: define how costs would be allocated back to teams or products
   • Calculate unit economics for your platform:
     • Cost per deployment (CI/CD pipeline run)
     • Cost per 1,000 HTTP requests served
     • Cost per developer onboarded via the IDP
   • Create a FinOps dashboard in Grafana showing spend trends, waste, and unit costs
   • Write a finops-runbook.md documenting the monthly cost review process

Stretch Goals

• Implement spot/preemptible instances for non-critical workloads and measure savings
• Configure Reserved Instances or Savings Plans for baseline workloads and calculate break-even
• Integrate Infracost into the CI pipeline to show cost impact of every Terraform PR

Deliverables

• Tagging strategy document and enforcement policy configuration
• Kubecost deployment with namespace cost allocation configured
• Waste audit report: resources eliminated and monthly savings
• Autoscaling configuration (VPA recommendations applied + Cluster Autoscaler or Karpenter)
• FinOps dashboard showing spend trends and unit economics
• finops-runbook.md for monthly cost review

References

Books

• Cloud FinOps - J.R. Storment & Mike Fuller
• The FinOps Book - FinOps Foundation

Courses and Certifications

• FinOps Certified Practitioner - FinOps Foundation
• AWS Cost Optimization - AWS Training
• Google Cloud Cost Management

Tools and Documentation

• Kubecost Documentation
• OpenCost - CNCF
• Karpenter - Node Autoscaling
• KEDA - Event-Driven Autoscaling
• Infracost - Cost in CI/CD
• AWS Compute Optimizer
• FinOps Framework

Once you complete this task you will treat cloud cost as an engineering metric - not a finance problem - and will be able to reduce waste, allocate costs fairly, and justify infrastructure decisions in terms of business value.