System Design Cases
FinOps and Cost Optimization
FinOps & Cost Optimization concept page with rightsize, spot fleet, S3 lifecycle tiering, egress cost shock scenarios. Includes ADR on when to invest in cost optimization vs feature work.
FinOps: cost, value and reliability in one loop
FinOps is a collaborative operating practice, not a periodic bill-cutting task. Engineering, finance, product, procurement and leadership iterate through Inform, Optimize and Operate with timely, attributable data and explicit business-value and risk trade-offs.
Unit-safe formulas
- Compute:
cost = instance_count * billed_hours * effective_rate_per_hour. - Origin requests:
total_requests * (1 - cache_hit_ratio)when the measured ratio and request population use the same scope. - Viewer delivery:
viewer_GB * CDN_viewer_rate; cache hits do not make viewer bytes zero. - Unit economics:
allocated_period_cost / successful_business_unitsfor the same period and scope. Define the zero-unit case and quality criteria.
Provider rates vary by region, tier, date and contract, so keep the price catalog versioned. Model every billed network leg. With an AWS origin, origin-to-CloudFront transfer may be free while CloudFront-to-viewer data and requests remain billable.
Optimization safety
Rightsizing and scheduling reduce usage; commitments reduce eligible rates. Commitments create utilization and term risk, so cover a risk-adjusted stable baseline and measure unused commitment, not a headline discount. Every change has an expected saving, SLO/security/capacity guardrail, limited rollout, observation window and rollback. Cost anomalies first need attribution; blindly turning off production can cost more than the anomaly.