performance-validation-plan

Define baselines from your optimization phase, then run daily and weekly validation comparing production metrics like throughput, P95 latency, and error rate against those baselines. Use monitoring dashboards and APM tools with specific queries to measure each metric against acceptance criteria.

Track throughput, latency percentiles (P50, P95, P99), error rates, availability, CPU and memory usage, GC pause times, and database query performance. Each metric needs a baseline, a production target, and an acceptable range.

Use PromQL queries such as histogram_quantile for latency percentiles and rate-based queries for throughput and error rates, then compare results to documented baseline values. Deviations beyond defined thresholds trigger tiered degradation response procedures.

Classify the degradation by severity: minor (10-20% from baseline) gets logged and monitored, moderate (20-50%) triggers a P2 incident with investigation, and severe (>50%) triggers a P1 incident with immediate mitigation and possible rollback.

You need an optimization report with performance baselines, a testing strategy with SLA targets, a hypercare monitoring setup document, access to production metrics and APM tools, and business-defined acceptable performance thresholds.