Backoffalgoritmen

Backoffalgoritmen refers to a family of methods used to schedule retry attempts after a failed access or request in order to reduce contention and overload on shared resources. They are widely used in semiconductor networks, computer networks, and distributed systems to minimize repeated collisions and thundering herd effects, while balancing responsiveness and throughput.

In networking, backoff algorithms coordinate access to a common communication channel. In Ethernet networks using CSMA/CD, after a collision a station waits for a random number of time slots before attempting to retransmit. The waiting window typically grows with each subsequent collision, a mechanism often described as binary exponential backoff: the number of possible backoff slots doubles after each collision, up to a maximum limit. This reduces the chance of repeated collisions as traffic increases.

In wireless local area networks, such as those using IEEE 802.11, contention-based access employs a similar idea. Each station selects a backoff counter from a contention window that is adjusted after successes or collisions. The counter is decremented while the channel is idle, and a transmission occurs when it reaches zero. The size of the window is bounded by minimum and maximum values (CWmin and CWmax) to prevent excessively long delays.

Backoff concepts also appear in transport protocols like TCP, where retransmission timeouts can be increased using exponential backoff to curb network congestion. Variants include linear backoff and adaptive or p-persistent strategies that adjust the waiting period based on observed conditions.

The main trade-off in backoff algorithms is between responsiveness and collision/overload avoidance. Properly tuned backoff helps sustain throughput under contention while avoiding large delays during peak traffic. Backoff ideas are also applied outside network contexts to mitigate simultaneous retries in distributed services and other retry-heavy systems.