How to use this jitter calculator
Enter the first wait, the cap and the retry count, pick an attempt, and the delay range, mean and standard deviation appear for full, equal and decorrelated jitter. Enter how many clients retry at the same moment and the tool measures how many land in one bucket compared with using no jitter at all.
How the figures are derived
Full draws uniformly from zero to the ceiling, equal from half the ceiling to the ceiling, and decorrelated from the base up to three times the previous delay. The three formulas and the comparison follow the AWS Architecture Blog article on backoff and jitter (checked October 2026). The standard deviation of a uniform range is its width divided by the square root of twelve.
Limits and cautions
The decorrelated figures and pile-up counts come from a random sample, so they shift on each run. No retry success rate or server load is predicted, and the fixed per-attempt delay table belongs to the exponential backoff calculator.
Frequently asked questions
The AWS article reports that full and equal make a similar number of calls, decorrelated makes more, and equal is the weakest. Full has the widest range and spreads a pile-up best.
Half of the wait is fixed, so its spread is only half that of full jitter. The pile-up scatters less and the total time grows longer.
The decorrelated figures and the pile-up counts are measured from a random sample. The ranges and means come from the formulas and do not move.