Load Testing Interview Questions and Answers

Sample questions

A few questions from each level with the correct answer and an explanation.

Load Testing · Junior

1. What is the main goal of a load test?
   ▸Show the answer

   Correct answer: Check how the system behaves under the expected number of users

   A load test applies the traffic you expect in production, usually the normal peak, and checks that response times and error rates stay within targets.

2. How does a stress test differ from a load test?
   ▸Show the answer

   Correct answer: It pushes past the expected load to see where and how the system breaks

   A load test stays at the expected level. A stress test goes above it to find the breaking point and to see whether the system degrades gracefully and recovers.

3. What does a spike test check?
   ▸Show the answer

   Correct answer: How the system copes with a sudden, sharp jump in traffic

   A spike test jumps from low to very high load in seconds, like a flash sale or a push notification, and checks whether the system survives and recovers.

4. What is a soak (endurance) test for?
   ▸Show the answer

   Correct answer: Finding problems that appear only after hours of steady load

   A soak test keeps a normal load for hours. It reveals memory leaks, growing queues, filling disks and connections that are never released.

Load Testing · Middle

1. What is the key difference between a closed and an open workload model?
   ▸Show the answer

   Correct answer: Closed: new requests wait for old ones. Open: arrivals don't depend on responses

   In a closed model a fixed number of users each send the next request after the previous one returns. In an open model requests arrive at a set rate, like real internet traffic, whatever the response time.

2. Why can a closed model hide a slowdown?
   ▸Show the answer

   Correct answer: When the server slows, users send fewer requests, so load drops by itself

   With 100 VUs and a 1 s response you get ~100 RPS. If responses take 5 s, you get ~20 RPS. The system gets relief that real traffic would never give it.

3. Which k6 executor keeps a fixed request rate regardless of response times?
   ▸Show the answer

   Correct answer: constant-arrival-rate

   Arrival-rate executors start iterations at the set rate and add VUs from the pool as needed. constant-vus is a closed model.

4. What does a non-zero dropped_iterations metric tell you in an arrival-rate test?
   ▸Show the answer

   Correct answer: k6 had no free VU to start an iteration, so the target rate wasn't reached

   If all VUs are busy and maxVUs is reached, k6 skips iterations. Raise maxVUs or note that the system was too slow to keep the rate.

Load Testing · Senior

1. You build a workload model for a public API. What is the best source of the arrival pattern?
   ▸Show the answer

   Correct answer: Production access logs: rates per endpoint, peaks and bursts

   Logs show the real request mix, daily peaks and how bursty traffic is. A model based on them answers questions about real risk.

2. When is a closed workload model the right choice?
   ▸Show the answer

   Correct answer: When clients really are a fixed pool, e.g. N workers or a call center

   If the real system has a fixed number of clients that wait for responses, a closed model matches it. Public traffic arrives independently, so an open model fits there.

3. How does wrk2 avoid coordinated omission?
   ▸Show the answer

   Correct answer: It sends at a constant rate and measures from the intended send time

   Latency counts from when the request should have been sent by schedule. A stall then shows up as the delay real users would see.

4. A page calls 100 backend services in parallel, each with p99 = 1 s. Roughly what share of page loads see at least one 1 s+ call?
   ▸Show the answer

   Correct answer: About 63%

   1 − 0.99^100 ≈ 0.63. Fan-out turns a rare tail into a common one, which is why tail latency matters so much in large systems.