Dynamic P-Persistent Backoff for Higher Efficiency and Implicit Prioritization

This article studies the efficiency of backoff algorithms. The fraction of channel time devoted to successful transmissions is maximized when the stations choose the optimal transmission probability. The binary exponential backoff algorithm does not come close to optimal channel efficiency, thus a new backoff mechanism that attains near-optimal efficiency is proposed. This algorithm is called Dynamic-P-Persistent backoff and is based on the observation that, under optimal efficiency conditions, the fraction of channel slots busy with collisions is constant. The stations monitor the channel to estimate the fraction of collision slots and adjust their transmission probabilities consequently. As opposed to previous backoff proposals, DPP does not require any estimation of the number of concurrent active stations. Further, DPP offers implicit prioritization that reduces the delay of real time and interactive traffic while maintaining optimal throughput for background traffic.