# Policy Engine Bench

Synthetic workload that measures end-to-end policy evaluation throughput against large advisory snapshots.

## Scenarios

`config.json` defines the default scenario (`policy_eval_baseline`) which generates **100 000 components** with **1 000 000 findings** derived from deterministic seeds. The harness measures:

- Latency (mean/p95/max milliseconds)
- Throughput (findings per second)
- Managed heap allocations (megabytes)

## Running locally

```bash
dotnet run \
  --project src/Bench/StellaOps.Bench/PolicyEngine/StellaOps.Bench.PolicyEngine/StellaOps.Bench.PolicyEngine.csproj \
  -- \
  --csv out/policy-bench.csv \
  --json out/policy-bench.json \
  --prometheus out/policy-bench.prom
```

The run exits non-zero if latency exceeds configured thresholds, throughput falls below the configured floor, or allocations surpass the budget. Baselines live in `baseline.csv` and can be regenerated whenever new steady-state metrics are established.