Load and throughput testing

Choose the workload and success contract before choosing a tool. Reuse the project's load runner and targets.

Establish the experiment

Record target, authorization, expected traffic, test data, duration, concurrency/rate limits, and cleanup. Derive latency/error thresholds from the service's requirements or an explicit experiment; example numbers are not universal SLAs.

Use existing k6 or Locust scripts when appropriate. Do not reject another established tool based on unrelated project history.

Choose the workload model

A single arrival-rate workload already warrants a named scenario. Closed models reduce offered load as responses slow, which can hide overload. For open models, bound allocated VUs and inspect dropped_iterations.