Capacity planning ensures sufficient capacity and redundancy to serve projected future demand with the required availability. SRE must own capacity planning because capacity is critical to availability. ## Three Mandatory Steps 1. **Accurate organic demand forecast** — extends beyond lead time for acquiring capacity (natural product adoption) 2. **Inorganic demand incorporation** — feature launches, marketing campaigns, business-driven changes 3. **Regular load testing** — correlate raw capacity (servers, disks) to actual service capacity ## Provisioning = Change Management + Capacity Planning Provisioning must be: - **Quick** — capacity is expensive, don't hold it idle - **Correct** — or capacity doesn't work when needed - **Careful** — riskier than load shifting (involves new instances, config changes, load balancers, networking, validation) ## Efficiency Connection Resource use = f(demand, capacity, software efficiency). SREs predict demand, provision capacity, and can modify software — controlling all three levers of service cost. A slowdown equates to lost capacity; at the extreme, infinite slowness = service stopped. ## Cross-Domain Connections - [[Error Budget Model]] — capacity failures consume error budget - [[SRE Fifty Percent Ops Cap]] — provisioning is engineering work - [[SRE Change Management Seventy Percent Rule]] — provisioning is a change --- *Source: Site Reliability Engineering, Chapter 1 (Treynor Sloss, 2016)* *Extracted: 2026-03-25*