What is pacing in performance testing?

Pacing is the fixed cycle time allotted to each iteration of a virtual user. It controls how frequently each user repeats its transaction, and is the primary lever for hitting a target throughput without simply adding more users.

How is pacing calculated?

Pacing = (3600 / (target hourly throughput / number of virtual users)). The think time you add per iteration is that pacing value minus the average response time of one iteration.

Pacing Calculator

Free pacing calculator: compute the pacing and think time needed to hit a target throughput with a given number of virtual users and response time.

Target throughput

req / hour

Concurrent virtual users

VUs

Avg. response time per iteration

sec

—sec pacing

Required throughput / VU: — iter/hr
Cycle time per iteration: — sec
Think time to add: — sec

How pacing works

Pacing is the fixed cycle time you give each virtual user (VU) per iteration. It's the cleanest way to control throughput: instead of throwing more users at the system, you tune how often each user repeats its transaction.

The math:

throughput_per_vu = target_throughput / virtual_users   (iterations/hour)
cycle_time        = 3600 / throughput_per_vu             (seconds)
pacing            = cycle_time
think_time        = pacing − response_time

Worked example

Target 3,600 req/hour with 50 VUs and a 2 s response time. Each VU must do 72 iter/hour → a 50 s cycle. So set constant pacing of 50 s, which is 48 s of think time after each 2 s iteration.

If the response time ever exceeds the cycle time, the target is infeasible with that VU count — add users or make the system faster.

Discussions coming soon.

Comments are powered by Giscus (GitHub Discussions). Enable them by configuring GISCUS in src/consts.ts — see giscus.app.