medium10 pts · first answer countsrate-limitsreliability

Why add jitter to backoff

Hundreds of workers hit a rate limit at the same moment. Each retries with exponential backoff (1s, 2s, 4s...). Why should you add random jitter to the wait times?

A.Jitter makes each retried request use fewer tokens, so it costs lessB.Without it, workers retry at the same moments and hit the limit together againC.The HTTP standard requires random waits after every 429 responseD.With jitter you can safely ignore the Retry-After header the server sends

← Savings from a cached prefix Tracing a retry helper →