9. Cost Optimization

"There is no cloud. It's just someone else's computer — and they're billing you for it."

What is Cost Optimization?

Managing cloud spending to maximize value — not just cutting costs, but getting the most reliability per dollar.

FinOps Principles

Inform

• Visibility into spending
• Allocation to teams/products
• Forecasting

Optimize

• Right-sizing
• Reserved capacity
• Spot/preemptible instances
• Waste elimination

Operate

• Showback/chargeback
• Budgets and alerts
• Continuous optimization

Key Concepts

Unit Economics

Cost per request
Cost per user
Cost per transaction
Cost per GB stored

Cloud Pricing Models

Cost Allocation

• Tag everything
• Allocate to teams/products
• Track untagged resources

Topics

• ☐ Cloud cost visibility
• ☐ Resource tagging strategy
• ☐ Right-sizing analysis
• ☐ Reserved capacity planning
• ☐ Spot instance strategies
• ☐ Storage optimization
• ☐ Network cost reduction
• ☐ Idle resource cleanup
• ☐ Cost anomaly detection
• ☐ FinOps culture

Optimization Strategies

Compute

• Right-size instances (most are over-provisioned)
• Use spot for fault-tolerant workloads
• Auto-scale to demand
• Terminate unused resources
• Consider ARM instances (Graviton)

Storage

• Lifecycle policies (S3, EBS snapshots)
• Right storage tier (hot/warm/cold)
• Delete orphaned volumes
• Compress and deduplicate

Network

• Use VPC endpoints (avoid NAT gateway costs)
• Optimize cross-region traffic
• Cache at edge (CDN)
• Compress payloads

Database

• Right-size instances
• Reserved capacity for stable workloads
• Use serverless for variable workloads
• Archive old data

Cost Visibility

Tagging Strategy

Required tags:
- environment (dev/staging/prod)
- team (owning team)
- service (application name)
- cost-center (billing code)

Dashboards

• Daily/weekly cost trends
• Cost by team/service
• Cost anomalies
• Reserved utilization
• Savings opportunities

Tools

Anti-Patterns

• No cost visibility
• No tagging
• Reserved instances without utilization tracking
• Dev environments running 24/7
• Orphaned resources
• Storing everything forever
• Over-provisioning "just in case"

Quick Wins

1. Delete unused resources (EIPs, EBS volumes, old snapshots)
2. Right-size obviously over-provisioned instances
3. Shut down dev/test outside business hours
4. Enable S3 lifecycle policies
5. Review and cancel unused reserved capacity

Metrics

• Total cloud spend
• Cost per unit (request, user, transaction)
• Reserved instance utilization
• Spot instance ratio
• Waste (unused resources)
• Cost forecast accuracy

Reading

• Cloud FinOps (O'Reilly)
• AWS Well-Architected: Cost Optimization Pillar
• FinOps Foundation resources

Atoms

• Unit Economics of Infrastructure

Published expressions

• So you want to save cost on your cloud bill… -- the introductory public version of this pillar's cost-reduction stance.

People-substrate cross-cluster

Over-indexed virtues have a cost. The CPTSD-substrate operator is expensive: burnout, team breakage, remediation work, hiring churn. The substrate-cost view is the FinOps view applied to people. Hiring "maximum" instead of "calibrated" pays for the wound, not the person -- and the wound has running costs.

• Bridge essay: Interview Training as Applied Clinical Psychology -- Item 8 (over-indexing as compensatory; "maximum" hires the wound)
• Psychology: Caretaker syndrome -- labour cost externalised to the over-functioning person; downstream cost is the dependent team
• Psychology: Hyper-independence -- the compute cost of refusing delegation; doesn't show up on the dashboard until burnout or churn
• Psychology: Learned helplessness -- the downstream cost shape of over-functioning leadership; the team that can't self-serve has running coordination overhead
• Anchor: Ownership psychology -- the under/over/calibrated axis is the substrate FinOps reads off when hiring; over-indexed = hidden running cost, calibrated = sustainable spend