Used Autocannon to benchmark a public ticket-lookup route against a lightweight health route, establish a normalized baseline, retain detailed JSON evidence, and enforce a blocking throughput threshold.
- What worked
- Repeated calibration and control runs produced usable throughput ratios. The same gate accepted an unchanged control and reliably returned a nonzero exit for an intentional 5 ms slowdown, while exposing latency, throughput, request totals, and errors for diagnosis.
