Ctrl K
QA · Mid-level · Theory

A load test showed that at 500 requests per second, response time grew from 200 milliseconds to three seconds. How would you track down the cause, and what would you put in the report?

~4 min by voice

How to approach it

This checks whether the candidate can systematically find a bottleneck under load rather than just noting 'it's slow'. Interviewers look at how they validate the test itself, read the metrics, go through the layers of the system (resources, database, dependencies) and frame the conclusion relative to the SLA.

Have you seen this question in a real interview?
Answers0
Avg. score—
Strong—
Practiced0
Topics
Similar questionsA service running on G1 periodically has pauses of one to two seconds, and it's hurting latency. What would you do, and when is it worth looking at a different garbage collector? Mid-levelHow is reflow different from repaint, and why does frequent reflow hurt performance? JuniorHow do you track down a performance bottleneck in Python code? What's the most telling case from your own experience? SeniorWhat happens in the browser between a change to the DOM or styles and pixels appearing on screen? How do you use that to keep animations and scrolling from janking? SeniorYour Java service has to respond with a p99 of no more than ten milliseconds. How would you design it with the JVM's characteristics in mind, and how would you find the source of latency if the target isn't being met? Senior

Answer by voice

The interviewer will ask this question and listen. Think out loud and ask for a hint if you need one. When you're done, pause or tap “That's my answer”. In about a minute, your score and the points you covered show up here.

Sign in to answer

Best with headphones · ~4 min