Rate limited API - Breakscale

A limiter refuses excess traffic at the door. It serves slightly less, but what it does serve stays fast instead of queueing.

What to watch

The limiter does not add capacity. It refuses some requests quickly so the rest are answered quickly: at 600 offered it serves 200 at 36ms, where removing it serves 234 at 235ms. Saying no fast is what buys the predictable wait.

What it is made of

Component	Count
Client	1
Database	1
Rate limiter	1
Service	1

Under load

Measured by running this design through the simulator at multiples of its own offered rate of 150 requests a second. Twenty seconds of simulated time, one fixed seed, so the same numbers come out every time.

This design holds 150 requests a second with 0% errors. At 225 it is losing 5% of them.

Offered	Goodput	p99	Errors
150	148	81.8ms	0%
225	193	85.7ms	5.4%
300	197	81.9ms	34.9%
375	202	83.1ms	46.9%
450	202	71.7ms	53%
600	199	71.7ms	67.2%
900	200	77.3ms	77.9%
1200	200	80.5ms	83.3%

Open this example

Load Rate limited API from the Examples menu and drag the traffic slider yourself.