MENTAL MODEL · M6173

Exponential Backoff

TechnicalHigh supportAlgorithms

Included

Updated 2026-08-05

Loading revision record…

INTRODUCTION

English translation pending.

CORE DEFINITION

In networking and distributed systems, exponential backoff delays a retry by a base interval multiplied by an increasing power after each failure, typically one second, then two, then four, until the attempt succeeds or a cap is reached. Randomized jitter is usually added so that many clients do not retry at the same instant. The strategy reduces congestion, avoids livelock, and is the standard congestion-friendly retry policy used in TCP retransmission and API clients.

SCAFFOLDING EFFECT

psychology

Reduce cognitive load

- Retreat gracefully: after a rejection, lengthen the interval before the next attempt instead of retrying at once. - Add jitter: randomize the delay so parallel attempts do not collide at the same moment. - Cap the wait: set a maximum interval and attempt count so backoff never starves the request.

anchor

Anchor fast decisions

Retrying immediately after a failure adds load exactly when the system is already saturated, which deepens the congestion and raises the failure rate. Increasing the delay after each failure gives the system time to recover and reduces total attempts during the overload. Jitter decorrelates many independent clients, so their retries spread across time instead of arriving in a synchronized burst.

MINIMUM ACTION

In progress 0/1

Practice this model in one real situation:

Set a short initial delay before the first retry. 2. Double the waiting delay after each subsequent failed attempt. 3. Add a random jitter value to every computed delay. 4. Cap both the maximum delay and the total attempt count. 5. Escalate or fail explicitly when the cap is reached.

Check to track your progress (stored locally)

Learning progress0%

account_treeGenealogyexpand_more

menu_bookReferencesexpand_more

Source support: Explicit

link
en.wikipedia.orghttps://en.wikipedia.org/wiki/Exponential_backoffZH · Explicit
verified

RELATED MODELS