Credit-Based Flow Control for Multihop Wireless Networks and Stochastic Petri Nets Analysis

Abstract

As the increase of data traffic is expected to grow faster than wireless capacity, dealing with congestion is unavoidable in the future. The capacity imbalance between wired and wireless links leaves a bottleneck on the wireless link while traffic keeps coming in through the bigger pipe. In current systems, the dynamics of a wireless channel and the variations in the higher priority real time traffic both make full link utilization impossible. Either the buffers are full, giving a full link but resulting in packet losses in the queues or the link is underutilized. Closed loop credit based flow control can solve this problem by avoiding any buffer overflow completely and providing enough packets in the buffers to ramp up immediately when the channel has remaining capacity. In this paper a model using generalized stochastic Petri nets (SPN) is introduced. By structural analysis the boundedness of memory is proven while a Markov state analysis of the SPN provides numeric performance results. The studied scenario includes two wireless relayed hops, a wired backhaul and traffic class separation.