Workload Capacity Modeler

Turn workload assumptions into inspectable ranges and estimates. Produce a state_patch; never replace the full design state or present a single guessed number as fact.

Contract

Read the bundled references/contract-v1.md and references/handoff-v1.md. Preserve design_id, stable requirement/assumption IDs, and contract version 1.0. Attach every EST-* record to at least one FR-*, NFR-*, or ASM-*.

Procedure

1. Extract request rate, read/write ratio, payload size, active users, concurrency, growth, retention, availability, and geography. Mark missing inputs as assumptions with ranges.
2. Create baseline, expected, peak, and stress scenarios. Keep average and peak rates separate and state the time window for each.
3. For every estimate, show formula, inputs, unit, horizon, source, confidence, and sensitivity impact. Keep bytes, bits, seconds, requests, and events dimensionally consistent.
4. Model storage as logical payload plus index, metadata, replication, and operational overhead. Model bandwidth in bits/s or bytes/s consistently and identify egress versus ingress.
5. Model concurrency from arrival rate and latency assumptions; state whether the result is a Little's Law estimate, a limit, or an observed measurement.
6. Evaluate cache capacity and hit-rate effects only when the access pattern supports caching. Do not infer cache correctness from cache presence.
7. Rank at least the three variables with the largest outcome impact. Use intervals where input uncertainty is material.
8. Recommend the next capability without directly calling it. Emit the standard handoff envelope.