performance-testing

Performance testing validates p95 latency against SLO thresholds by running controlled load tests that measure latency distribution, throughput, and error rate under defined workload conditions. You define load profiles and pass/fail criteria based on explicit SLOs.

Performance testing applies different load shapes to verify distinct operational properties: smoke tests check basic functionality, stress tests push capacity limits, and soak tests validate sustained throughput and resource saturation over extended periods.

Latency percentiles like p50, p95, p99, and p99.9 provide accurate user-experience guarantees during performance testing, whereas averages mask tail latencies and hide extreme outliers that impact real users under load.

Yes, performance testing requires production-equivalent staging environments so measurements accurately reflect the system you actually ship. Environment fidelity ensures latency and throughput data matches real-world performance behavior.

Yes, performance testing supports breakpoint testing using load generation tools like k6 and JMeter to identify failure modes and capacity limits. You establish SLO-driven pass/fail criteria to determine the exact system breaking point.

Performance testing uses spike testing instead of standard load testing when you need to validate system behavior under sudden, extreme traffic surges. Standard load testing verifies steady-state capacity, while spike testing exposes saturation and recovery failures.