testing-load

Load goal (throughput/latency/stability) Key user scenarios Environment constraints and limits User profile and scenario distribution Base metrics and current baseline Allowed thresholds (SLA/SLO) Monitoring requirements (APM/logging) </input_requirements> Load reflects real traffic distribution Metrics and thresholds are agreed upfront Scenarios are isolated and repeatable Tests are split into warm-up/peak/stability Clearly separate load/stress/spike/soak tests Add think time and realistic pauses Metric collection is agreed with the infra team </test_design> Do not exceed environment limits without permission Record load configuration and environment Verify stability across repeated runs Identify degradations and saturation points Verify you are testing the correct service version Compare against baseline and record changes Synchronize timestamps and log correlation </execution_rules> p50/p95/p99 latency throughput (RPS) error rate resource usage (CPU/RAM/IO) latency for key endpoints Do not run load tests in production without permission Do not use real user data Do not change load parameters during the test without recording it Do not run tests without monitoring key metrics </do_not> Linear ramp-up to peak load Step load with stabilization periods Long soak test for stability </example_scenarios>