Tuning Pmax in RED Gateways for QoS Enhancement in Wireless Packet Switching Networks

Abstract

AbstractIn wireless packet switching networks, random early detection (RED) gateways are used to control the congestion formation. By computing mean queue length of the bottleneck link, the random early detection gateway detects the incipient congestion. The RED gateways notify the congestion formation to the source TCP connections either by setting a bit in the header of the acknowledgment packet or by dropping the incoming packets with a certain probability. Through this means RED gateways try to keep the queue length below the maximum threshold value. Tuning RED parameters is a challenging task for network engineers. In this paper, we tried to estimate a stable boundary for Pmax. We use stochastic models representing sender window dynamics, average queue length and packet dropping probability. The non-linear equations have been linearized using transformation technique. By constructing Hermite matrix for time-delay control system, analysis has been made to determine a stable boundary for Pmax. The performance of sender window, queue length, packet dropping probability and throughput has been analysed using statistical data and graphs obtained by MATLAB programming. The results revel that the queue length, sender window size and random packet dropping probability converge to smaller range over their values. A comparative study has been made by conducting experiments for assumed Pmax value (0.05) and for stable Pmax value. Application of stable Pmax value gives an excellent throughput and better network performance.KeywordsBottleneck linkCwnd packet droppingQueue lengthRED gatewayStabilityTCPRTT