In flash sales or exam logins, each user sends only one request. To test how your system handles this burst of simultaneous traffic, configure Performance Testing (PTS) so that every virtual user (VU) runs the test scenario exactly once at full concurrency.
The following steps use a JMeter-based PTS scenario where each VU sends a single request.
Why this configuration works
A single-request-per-user test relies on three settings working together:
All VUs launch simultaneously. Setting Starting Percentage to 100% sends every VU at once instead of ramping up gradually. This matches the sudden spike of a real flash sale.
Each VU iterates once. Setting Number of Loops to 1 means each VU sends exactly one request and then exits.
Fixed concurrency. The Manual Adjustment traffic model lets you set an exact VU count rather than relying on auto-scaling. This gives you direct control over how many simultaneous users hit your system.
Procedure
Log on to the PTS console, choose , and then click JMeter.
In the Load Settings section, configure the following parameters:
Parameter Value Purpose Traffic Model Manual Adjustment Direct control over VU count Max VUs Target concurrency (for example, 1,000) Total simultaneous users Starting Percentage 100% All VUs launch at once to simulate burst traffic Specify Loop Yes Enables fixed iteration count Number of Loops 1 Each VU runs the scenario once Set the Test Duration based on your expected response time. The minimum duration is one minute. Choose a duration long enough for all VUs to complete their single request. For example, if each request takes about 5 seconds, a one-minute duration provides sufficient margin.

Verify completion and stop the test
PTS does not automatically stop the test after all VUs complete their request. Check the concurrency line chart to confirm completion:
Open the real-time monitoring dashboard.
When the concurrency value drops to 0, all VUs have finished.
Manually stop the test.
Do not end the test before concurrency reaches 0. Stopping too early may cut off in-flight requests and produce incomplete results.