edercnj/finops

Knowledge Pack: FinOps

Purpose

Provides FinOps (Financial Operations) practices for cloud cost management in {{LANGUAGE}} {{FRAMEWORK}} projects. Covers resource rightsizing, cost allocation, spot/preemptible instances, reserved capacity, cost alerting, and cloud-specific optimization tools.

Quick Reference

See references/cost-optimization-checklist.md for a layer-by-layer cost optimization checklist (compute, storage, network, database).

See references/tagging-strategy-template.md for a tagging strategy template with mandatory and optional tags.

Resource Rightsizing

Align provisioned resources with actual usage to eliminate waste.

• CPU/Memory Analysis: Compare actual usage (P95/P99) against provisioned capacity; target 60-80% utilization
• Recommendation Engines: AWS Compute Optimizer, Azure Advisor, GCP Recommender provide automated rightsizing suggestions
• Rightsizing Cadence: Monthly review of resource utilization metrics; adjust after sustained usage patterns (minimum 14 days of data)
• Container Resource Limits: Set CPU requests/limits and memory requests/limits based on observed usage; avoid unbounded containers
• Vertical vs Horizontal: Prefer horizontal scaling (more small instances) over vertical scaling (fewer large instances) for cost efficiency

Cost Allocation

Track and attribute cloud costs to teams, services, and environments.

• Tagging Strategy: Mandatory tags: team, environment, service, cost-center; optional: project, owner, expiry
• Cost Centers and Ownership: Each service must have a designated cost owner who reviews monthly spend
• Showback vs Chargeback: Showback (visibility into costs per team) as first step; chargeback (billing teams) for mature organizations
• Untagged Resource Detection: Automated policies to detect and flag untagged resources; enforce tagging at provisioning time
• Budget Boundaries: Define per-team and per-environment budgets; alert at 80% and 100% thresholds

Spot/Preemptible Instances

Leverage discounted compute for fault-tolerant workloads.

• Interruption Handling: Implement graceful shutdown (SIGTERM handler), checkpointing for long-running tasks, and state externalization
• Workload Suitability: Best for stateless services, batch processing, CI/CD pipelines, and fault-tolerant workloads
• Spot Fleet Strategies: Diversify across instance types and availability zones to reduce interruption probability
• Fallback to On-Demand: Configure automatic fallback to on-demand instances when spot capacity is unavailable
• Cost Savings: Typical 60-90% discount compared to on-demand pricing

Reserved Capacity

Commit to usage for predictable workloads to reduce costs.

• RI vs Savings Plans: Reserved Instances lock to instance type; Savings Plans offer flexibility across instance families
• Break-Even Analysis: Typically 7-9 months for 1-year reservations; calculate based on current on-demand spend
• Commitment Period: 1-year (lower discount, lower risk) vs 3-year (higher discount, higher commitment)
• Coverage Recommendations: Reserve 70-80% of baseline capacity; use on-demand/spot for variable load
• Review Cadence: Quarterly review of reservation utilization and coverage; adjust for workload changes

Cost Alerting

Proactive monitoring and notification for cloud spend.

• Budget Alerts: Per team, per environment, per service; alert at 50%, 80%, 100% of budget
• Anomaly Detection: Detect unexpected cost spikes (>20% day-over-day or >50% week-over-week increase)
• Billing Threshold Notifications: Absolute threshold alerts for total account spend
• Daily/Weekly Cost Reports: Automated reports showing spend breakdown, trends, and top cost drivers
• Forecast Alerts: Notify when projected monthly spend exceeds budget based on current burn rate

FinOps Practices

Organizational practices for cloud financial management.

• Cost Review Cadence: Weekly team review (15 min), monthly organization review (30 min), quarterly strategic review
• Optimization Sprints: Dedicate 10-20% of sprint capacity to cost reduction tasks
• Unit Economics: Track cost per transaction, cost per user, cost per request; set targets and monitor trends
• Waste Elimination: Identify and remove unused resources (unattached volumes, idle load balancers, orphaned snapshots)
• Architecture Cost Reviews: Include cost impact assessment in architecture decision records (ADRs)

Cloud-Specific Cost Tools

AWS

• Cost Explorer: Visualize and analyze spend patterns over time
• Trusted Advisor: Automated checks for cost optimization opportunities
• Savings Plans: Flexible commitment-based discounts across EC2, Lambda, Fargate
• Compute Optimizer: ML-based recommendations for EC2, EBS, Lambda right-sizing

Azure

• Cost Management: Budget tracking, cost analysis, and optimization recommendations
• Advisor: Personalized best practices for cost, security, reliability, performance
• Reservations: Reserved VM Instances, SQL Database, Cosmos DB, and more

GCP

• Billing Console: Cost breakdown, budgets, and alerts
• Recommender: ML-based recommendations for VM right-sizing and idle resource cleanup
• Committed Use Discounts: 1-year and 3-year commitments for compute and memory

Related Knowledge Packs

| Pack | Relationship | |------|-------------| | infrastructure | Infrastructure patterns and 12-Factor App principles | | observability | Metrics and monitoring for cost-aware observability |