Supporting track

Performance Testing

Load profiles, SLAs and reading results like an engineer.

What this track covers

Load vs stress vs soakWorkload modellingThink time & pacingJMeter basicsk6 scriptingPercentiles vs averagesBottleneck analysisCI integration

A load test with clear SLAs

Define thresholds in the script so the test fails itself — no manual chart reading.

Why it is required: Performance tests without thresholds produce opinions, not gates.

Example: A k6 script with p95 latency and error-rate thresholds.

javascript

1export const options = {
2  stages: [
3    { duration: '2m', target: 200 },
4    { duration: '5m', target: 200 },
5    { duration: '2m', target: 0 },
6  ],
7  thresholds: {
8    http_req_duration: ['p(95)<800'],
9    http_req_failed: ['rate<0.01'],
10  },
11};

Expected result

The run fails automatically if p95 exceeds 800ms or errors exceed 1%.

Common mistakes

Reporting averages instead of p95/p99
No think time, producing unrealistic load
Testing on hardware unlike production

Why are percentiles more useful than averages?

Related interview questions

Open bank

How do you decide the load profile for a performance test?
Derive it from production analytics: peak concurrent users, transaction mix, think time and peak-hour distribution — never from a guessed round number.
Load, stress, soak and spike testing — what does each answer?
Load: does it meet SLAs at expected traffic. Stress: where does it break and how. Soak: does it degrade over hours (leaks). Spike: does it survive a sudden surge and recover.
Which metrics do you report, and why is average response time misleading?
Report p90/p95/p99 latency, throughput, error rate and concurrency together — averages hide the tail where real users suffer.
How do you build a realistic workload model?
Derive it from production data: transaction mix by volume, think times, session length, peak concurrency and data volumes — then validate the model against a known production hour.
You found a bottleneck. How do you analyse it?
Work from the outside in: confirm it in the client metrics, correlate with server APM traces, then isolate the layer — app CPU, GC, database, cache, external call, or network.
What is load testing, and why does it matter in Performance?
load testing is a core Performance concept used to make testing decisions more accurate, repeatable, and aligned with product risk.