Load Model & Traffic Shape
What Is a Load Profile?
A load profile is the shape of load over time in a test — how many virtual users are active at each moment, from ramp-up through steady state to ramp-down.
Also known as: Load Pattern, Load Shape, Workload Model
The shape decides what you learn
Two tests with an identical peak VU count can answer completely different questions depending on how they get there and how long they stay. The profile, not the peak, determines which failure modes a run can expose at all.
- Ramp and hold — climb to target, hold, drain. The default for measuring steady-state capacity.
- Step — climb in flat plateaus. Makes it easy to attribute a change in response time to a specific load level.
- Spike — a near-vertical jump and an equally sharp drop. See spike test.
- Soak — a moderate level held for hours. See soak test.
- Sawtooth — repeated climbs and drops, which exercises autoscaling and cache warm-up over and over.
Building a profile from production data
The most defensible profile is traced from a real traffic graph. Pull a week of request rates, find the daily peak and the steepest rise into it, then scale both by whatever growth factor you are planning for. A profile that reproduces the shape of Monday 09:00 is worth more than an invented ramp with a rounder number at the top.
Reproduce the ratios rather than the raw volume when you are testing a scaled-down environment. Holding the peak-to-trough ratio and the rate of change constant keeps the profile meaningful even at a fraction of production traffic.
Do not forget the ramp-down
Most profiles stop at the peak, which skips half the story. Recovery is a real property: connection pools have to drain, queues have to empty, and autoscalers have to scale back in without dropping in-flight work.
Ending a run with a deliberate descent back to baseline load — and keeping it there for a few minutes — turns “did it survive the peak” into “did it come back”. Systems fail that second question more often than teams expect.