Load Model & Traffic Shape
What Is Ramp-Up in Load Testing?
Ramp-up is the period over which a test climbs from zero to its target load, adding virtual users gradually instead of starting all of them at once.
Also known as: Ramp Up, Ramp-Up Period
Why gradual beats instant
Starting every virtual user in the same instant tests something no production system experiences. The first seconds of such a run measure cold caches, unwarmed JIT compilation, an empty connection pool, and autoscalers that have not reacted yet — all at once, and all attributed to your application.
A ramp separates those effects in time. Warm-up costs land early and visibly, and the numbers from the steady-state period that follows describe the system you actually deployed.
The ramp is the measurement
Treating ramp-up as setup to be skipped over throws away the most informative part of a run. During a ramp, load rises smoothly while response time stays flat — until it doesn’t. The bend in that curve is your saturation point, and reading it is the entire premise of a breakpoint test.
Plot achieved throughput against active VUs. While the line is straight, you have headroom. Where it flattens, something has run out.
Choosing a ramp duration
There is no universal figure, but there is a rule of thumb: the ramp should be long enough for every layer that reacts to load to finish reacting.
- Long enough to fill connection pools and warm caches — usually tens of seconds at minimum.
- Long enough for autoscaling to complete a full cycle, if you are testing an autoscaled system. That is often several minutes.
- Short enough that the ramp is a small fraction of total run time, so steady state still dominates the result.
- Deliberately violent when a sudden surge is the thing you want to test — that is a spike test, not a ramp.