Load testing
Load testing is running a system under the volume of concurrent users or requests it is expected to handle and measuring how it behaves: response times, error rate and resource use. The aim is to confirm that normal peak demand is served acceptably before real users arrive, not to find the point where the system breaks.
The longer version
A load test replaces real users with scripted ones. Each virtual user follows a realistic journey, such as logging in, searching and checking out, with pauses between steps like a person would take. The tool ramps the number of users up to the target, holds it there, and records response times by percentile, the share of requests that fail, and what the servers and the database were doing at the time. The target usually comes from analytics or from the traffic a launch is expected to bring.
Load testing is the baseline the other performance types vary. Endurance testing holds a similar load for hours or days, spike testing jumps it abruptly, and scalability testing changes the load or the resources between runs to see how capacity responds. Pushing past the expected volume until something fails is usually called stress testing, and it answers a different question from whether normal peak demand is served.
Traffic models go wrong in familiar ways. A script that hits one cheap page proves little about a site where most visitors search, and a test run against a small staging environment says little about production unless the two are alike in size and configuration. Test data matters as well, since a cache that is warm on identical requests flatters every number.
Common questions
What does a load test measure?
A load test records response times, usually as percentiles such as the median and the 95th, along with throughput in requests per second, the error rate, and resource use on the servers and database. Percentiles matter because an average hides the slow requests that real users notice.
How many users should a load test simulate?
The number should match the expected peak, taken from analytics for an existing product or from launch forecasts for a new one. Concurrent users are not the same as registered users, so the figure to estimate is how many are active in the busiest period.
Is load testing the same as stress testing?
Load testing stays within the volume the system is expected to handle and checks that it performs acceptably there. Stress testing goes beyond expected volume until the system degrades or fails, to learn where the limit is and how the failure looks.
When should load testing happen?
Load testing belongs before a launch or a campaign that will raise traffic, and after changes that affect performance, such as a new database schema or a framework upgrade. Teams with a pipeline often run a smaller, repeatable load test on each release to catch a regression early.
Can load testing run against production?
Production load testing is possible, but the risk is that the test itself degrades service for real customers and pollutes real data. Most teams test in a production-like environment, and those that test live do it with limits, off-peak timing and test accounts that can be told apart from real ones.
Bug reports that arrive ready to triage.
Collect replay, console, network and browser context with the Free plan.