warm-up

Running traffic for a while before measuring, so caches, connection pools and compiled code reach their normal state.

In more detail

Measurements taken while a system is still cold are slower than normal and mislead comparisons. Discard the warm-up period from results, or report it separately.

Example

In this course: Measuring from a cold start mixes start-up cost into steady-state numbers and makes runs hard to compare. Most load tools let you ramp traffic up over the warm-up before holding a steady level.

Related terms

Taught in

References