Os Labs / Quality
QualityAdvanced~8h
p99 Load Test with a Latency Budget Gate
A scripted load test that drives realistic concurrent traffic against an API, measures p50/p95/p99 latency and error rate under load, and enforces a hard performance budget as an automated CI gate — proven by intentionally introducing a slow code path and watching the gate fail.
O que isso prova
A frase que você pode defender em uma entrevista.
Linha do currículo
Built an automated p99 latency budget gate (k6 + CI) that load-tests critical API endpoints under realistic concurrency and fails the pipeline on SLO violations, catching a deliberately introduced N+1 query regression before merge.
- can design and run realistic load tests, not just single-request benchmarks
- understands percentile latency vs average latency and why it matters
- turns a performance target into an enforced, automated gate
O briefing
O que você constrói, passo a passo.
- 01Write a k6 (or equivalent) load test script simulating realistic ramping concurrent users against at least 2 critical endpoints
- 02Capture p50, p95, and p99 latency plus error rate per endpoint from each run
- 03Define an explicit latency budget (e.g. p99 < 300ms, error rate < 1%) as a machine-checked threshold, not a manual read of a dashboard
- 04Fail the CI job with a non-zero exit code and a readable summary when any endpoint exceeds its budget
- 05Produce a human-readable report (HTML or markdown) summarizing latency distribution per run, kept as a build artifact
- 06Demonstrate the gate catching a real regression: introduce a slow path (e.g. N+1 query or added synchronous delay) and show the run go red
A prova
Está pronto quando estes passarem.
Load test run against the healthy baseline passes the budget gate with p99 latency reported under threshold
automated test
Load test run against a deliberately regressed slow path fails the budget gate with a clear per-endpoint breakdown of which threshold was violated
automated test
CI job artifact (HTML/markdown report) showing p50/p95/p99 and error rate for both the passing and failing runs
artifact
Code review confirms the budget thresholds are configurable and the gate logic can't silently pass on a k6 script error
reviewed
Stack
k6Node.js/Express (or existing API)GitHub ActionsGrafana or k6 HTML report
Método Sage
frame → decide → prove
Você mantém
k6 load test suite with a CI-enforced p99 latency budget gate and before/after regression reports