📚 Chapter 14: Cost Optimization (TCO)

1. Architectural Strategy: Right-Sizing and Elasticity

A. Right-Sizing

• Definition: Continuously adjust CPU/RAM/storage to fit actual workload, removing over-provisioning.
• Mechanism: Requires observability; identify underutilized instances (e.g., <20% CPU) and scale down or choose cheaper families.

B. Leveraging Elasticity

• Definition: Auto-scale up/down to pay only for what you use.
• Mechanism: Prefer horizontal scaling (HPA/ASG); use Spot instances for tolerant batch/CI workloads.

C. Serverless for Variable Load

• Strategic Choice: Adopt FaaS for spiky traffic.
• Cost Benefit: Zero idle cost; massive TCO advantage vs provisioned compute.

2. Financial Strategy: Commitment and Discount Models

A. Reserved Instances (RIs)

• Mechanism: 1–3 year commitments for instance types.
• Benefit: 30–70% discount vs on-demand.
• Trade-off: Reduced flexibility if workload changes.

B. Savings Plans

• Mechanism: Commit to dollar spend per hour; flexible across types/regions/families.
• Benefit: Similar discounts with improved flexibility; generally preferred over RIs.

3. Optimizing Data and Storage Costs

A. Data Tiering and Lifecycle Management

• Mechanism: Move aging data to cheaper tiers via lifecycle policies.
• Example: Logs → Standard to Archival (Glacier/Archive) after 90 days.

B. Network Egress Cost Reduction

• Problem: Egress fees for cross-AZ/region traffic.
• Strategy: Co-locate services and DBs in same region/AZ; minimize cross-boundary transfers.

C. Database Optimization

• Indexing: Reduce CPU/I/O for managed DBs.
• Managed Services: Higher sticker price but lower TCO by eliminating DBA overhead.

4. Governance and FinOps Practices

• Showback/Chargeback: Report spend per team/product to drive accountability.
• Automated Cost Control: Budgets/alerts; auto-shutdown Dev/QA outside business hours.