role-architect:capacity-planning

Capacity Planning

When to use

• Estimating infrastructure needs for a new service before launch
• Sizing CPU, memory, storage, and network for a given traffic profile
• Planning for seasonal peaks, product launches, or viral traffic events
• Calculating how much runway remains before current infrastructure reaches capacity
• Correlating load test results to real capacity thresholds and SLA boundaries
• Integrating continuous load testing into CI/CD pipelines

Core principles

1. DAU to QPS is the foundation — business metrics drive infrastructure math, not gut feel
2. Plan for 2x, stress-test for 5x — expected case determines the design; worst case tests it
3. Shortest runway is the binding constraint — CPU, memory, storage, and network all need their own runway calculation
4. Revalidate quarterly — growth projections are always wrong; the question is by how much
5. Load tests are perishable — capacity baselines stale within a release cycle without continuous testing

Reference Files

• references/traffic-and-resource-sizing.md — DAU-to-QPS conversion, peak multipliers by app type, traffic composition (read/write ratios), CPU/memory/storage/network sizing rules, target utilization headroom percentages, and linear/exponential/step-function growth modeling
• references/peak-load-and-load-testing.md — seasonal and event-driven peak planning, thundering herd mitigation, infrastructure runway formula and minimum thresholds, load test baseline establishment at current/2x/projected peak, capacity ceiling identification, and continuous load testing integration with CI/CD