load-stress-test

To find the breaking point of HTTP and gRPC services, you can run load and stress tests using an open arrival-rate model. This approach identifies exact throughput ceilings and latency degradation under sustained traffic.

An open arrival-rate model in load testing generates precise requests per second independent of system response times. It avoids coordinated omission to ensure realistic traffic patterns when measuring capacity limits.

You can use k6 for SLO gatekeeping in CI/CD by defining machine-checkable thresholds for p95, p99 latency, and error rates. The performance suite fails the pipeline if services regress under target requests per second.

Yes, you can simulate traffic spikes for capacity planning by configuring stress tests against staging environments. This validates whether your service maintains its p99 latency SLO and identifies bottlenecks like database connection pools.

The best way to enforce performance thresholds as code is applying machine-checkable SLO thresholds to your load testing suite. This automatically fails CI/CD quality gating scenarios when error rates or latency metrics regress.

Coordinated omission skews stress testing results because load generation pauses when the system slows down, hiding true latency degradation. Using an open arrival-rate model prevents this by maintaining constant requests per second.