• ↑↓ pour naviguer
  • pour ouvrir
  • pour sélectionner
  • ⌘ ⌥ ↵ pour ouvrir dans un panneau
  • ←→ pour naviguer
  • esc pour rejeter
⌘ '
raccourcis clavier

The Problem

After a collision, if all devices retry immediately, they’ll collide again. A randomized waiting mechanism is needed to reduce the probability of repeated collisions.

Core Idea

An algorithm where the maximum random wait time doubles after each successive collision, reducing retry attempts when the network is congested.

How It Works

  1. First collision: wait random time between 0 and 1 slot time
  2. Second collision: wait random time between 0 and 2 slot times
  3. Third collision: wait random time between 0 and 4 slot times
  4. Nth collision: wait random time between 0 and 2^N slot times (capped at 1024)
  5. After successful transmission, the backoff counter resets

Visual Explanation

G Collision Collision Backoff Random wait: 0 to 2^N slots Collision->Backoff Success Success? Backoff->Success Success->Collision No, N++ Reset N=0 Reset N=0 Success->Reset N=0 Yes

Key Properties

  • Reduces collision probability under high load
  • Wait time grows exponentially with repeated collisions
  • Capped at maximum backoff (e.g., 1024 slots in Ethernet)
  • Used in CSMA/CD and some wireless protocols

Connections

  • Built from: CD — uses this backoff algorithm
  • Related: Collision — triggers backoff
  • Related: Jam Signal — sent before backoff
  • Related: Random Access — broader category of protocols using backoff

Edge Cases & Gotchas

  • Maximum backoff limit prevents excessive wait times
  • Many collisions can still cause long delays (exponential growth)
  • Not used in modern full-duplex Ethernet (no collisions to back off from)