Skip to content
Ameen Abozaid
Englishالعربية
MID-046

Performance testing

  • 6 Lessons
  • 3 h practice (estimate)
Start module
On this page 8 topics

Overview

Topics

  • k6 or Gatling
  • load
  • stress
  • spike
  • soak testing

Learn first

3 h of practice (estimate)

Learning outcomes

  1. Distinguish load, stress, spike, and soak tests and what each reveals.
  2. Write k6 (or Gatling) scenarios with realistic traffic mixes and think time.
  3. Report latency percentiles and throughput, and avoid coordinated omission.
  4. Correlate test results with profiles and metrics to find the bottleneck.

Lessons

  1. 01Kinds of performance tests
  2. 02Writing k6 scenarios
  3. 03Workload models
  4. 04Percentiles and coordinated omission
  5. 05From test result to bottleneck
  6. 06Demo: a soak test that finds a slow leak

Lab and exercises

Lab

MID-046-LAB1 · Shop under load

Outcomes: 1–4 · Time: ~3 h Run load, spike, and 2-hour soak tests of checkout from a separate machine with a realistic traffic mix and data set. Write a report naming the first bottleneck. Acceptance: report with percentiles, graphs, and evidence.

Exercises (assessment: report review)

IDOutcomeTask
MID-046-EX011For four scenarios (flash sale, nightly leak, capacity planning, recovery after overload), pick the test type and justify it.
MID-046-EX022Write a k6 script for browse, add to cart, and checkout with think time and thresholds.
MID-046-EX032Make the script use varied users and SKUs from a data file.
MID-046-EX043Rewrite a closed-loop scenario as an arrival-rate scenario and explain the change in the results.
MID-046-EX053Given a latency histogram, report p50, p95, p99 and explain why the average misleads.
MID-046-EX064Given a graph of p95 and pool-wait metrics, state the bottleneck hypothesis and the next measurement.
MID-046-EX071–4Review a sample report: list what is missing (percentiles, load generator location, data realism, evidence).
Show solutions
  • EX01 Flash sale: spike; nightly leak: soak; capacity planning: stress or load to saturation; recovery: stress with a drop back to normal.
  • EX02 options.thresholds for p95 and failure rate; a default function with three requests and sleep calls.
  • EX03 SharedArray loaded from JSON/CSV and a random pick per iteration.
  • EX04 executor: 'constant-arrival-rate' with preAllocatedVUs; the slow system now shows rising latency and dropped iterations instead of hidden lower throughput.
  • EX05 State the three percentiles; the average sits between them and hides the tail.
  • EX06 Hypothesis: connection pool saturation; measure pool wait, active connections, and slow queries; then retest after adjusting one setting.
  • EX07 Missing items are fixed one by one; reject reports that quote only averages or ran on the same machine as the system under test.

Outcome 1 is also checked by the lab (three test types) and quiz Q01-Q02.

Quiz

Outcomes: O1 test kinds · O2 k6 scenarios and workloads · O3 percentiles and coordinated omission · O4 finding the bottleneck.

A self-check for you: nothing is graded, saved or sent anywhere. Choose an answer, then check it. You can try again as often as you like.

  1. 1Which test finds slow leaks?O1
  2. 2A spike test simulates:O1
  3. 3In k6, thresholds are used to:O2
  4. 4Think time (sleep) is added because:O2
  5. 5For public web traffic the better model is:O2
  6. 6Average latency is a poor summary because:O3
  7. 7Coordinated omission means:O3
  8. 8Percentiles from several machines should be:O3
  9. 9After finding a bottleneck you should:O4
  10. 10The load generator should run:O4

Practical part

Seven exercises MID-046-EX01 to EX07 plus the load, spike, and soak lab. Pass: at least five exercises and a report with percentiles, graphs, and evidence of the first bottleneck.

Common mistakes

Averages instead of percentiles; testing from the same machine; unrealistic data sets.

Assessment

Report review.

Career relevance

Capacity planning (SRC-004) and the MID gate.

Type to search levels, phases, modules and lessons.