Differentiated WRED algorithm for wireless mesh networks

Abstract

As wireless multimedia communication for next-generation broadband wireless Internet access deploying WMN, one emerging challenge is meeting the Quality of Service (QoS) requirements with efficient resource utilization. Different QoS guarantees are required for real-time and non-real-time traffics such as delay constraints for voice traffic and throughput for delay insensitive data applications. Quality of services experienced by the users in wireless mesh networks, i.e., end-to-end delays, bandwidth utilization, and packet losses, is heavily dependent upon the effective congestion avoidance techniques employed in the wireless nodes. Congestion can be detected by computing the average queue size and either dropping the packet or marking special bits in the packet header when average queue buffer usage crosses the threshold. Several algorithms and packet dropping probabilities are discussed in Floyd and Jacobson (IEEE/ACM Transactions on Networking, 1(4), 397–413, 1993, [1]), Chaegwon Lim et al. (A Weighted RED for Alleviating Starvation Problem in Wireless Mesh Networks, IEEE, 2008, [2]). In this paper, we discuss the existing average queue length computation and packet drop methods employed and propose improvements to consider dynamic buffer allocation for Wireless Random Early Detection (WRED) algorithm.