Blocking probability and convergence analysis of an adaptive DQDB bandwidth balancing mechanism

Abstract

The authors present a distributed realization of an adaptive load-based bandwidth balancing mechanism (APR, adaptive proportional) for improving DQDB (distributed queue dual bus), fairness and performance. Under the APR strategy, bandwidth is assigned distributedly and dynamically to each station according to the current offered load distribution in the system. The major significance of the method is its simple realization in a distributed environment and its ability to achieve load-based fairness in throughput and blocking probability while at the same time reducing the wasted system bandwidth to a minimum. The APR scheme is shown to be fair in handling multi-priority traffic. In addition to the performance of APR, the system convergence behavior is also analyzed.<>