Analysis and design of active queue management for TCP-RED congestion control strategies

This thesis investigates the problems of the Active Queue Management (AQM) techniques for congestion control in TCP networks. Random Early Detection (RED) and the RED-based strategies, which adopt the AQM approach, are evaluated through simulation. Two main problems of RED, and its variants, are considered. The first problem is the mismatch between the average and actual queue sizes. The second problem is the parameter configuration. This thesis proposes three new RED-based strategies and simulates them using the NS-2 simulator. These novel strategies are evaluated and com- pared with current RED based strategies. The proposed strategies are: Queue Sectors RED (QSRED), Risk Threshold RED (RTRED) and Weighted RED (WTRED). The performance of these strategies is evaluated using performance indicators such as: throughput, link utilization, packet loss and delay. QSRED divides the router buffer into equal subsectors and monitors the queue dynamics. The actual drop probability pa and maximum drop probability maxp are adjusted depending on the position of the actual and average queue sizes; q and avg respectively. Currently, RED maintains a maximum threshold maxth and minimum threshold min th. The second RED modification, RTRED, adds a third drop- ping level. This new dropping level is the risk threshold riskth which works with the actual and average queue sizes to detect the immediate congestion in gateways. Congestion reaction by RTRED is on time. The reaction to congestion is neither too early, to avoid unfair packet losses, nor too late to avoid packet dropping from time-outs. The third proposed RED modification, WTRED, adjusts the weight parameter wq dynamically, to reduce the mismatch between the average and actual queue size. WTRED also adjusts the maximum and minimum thresholds, to increase network throughput and link utilization. The simulation results demonstrate the shortcomings of RED and RED- based strategies. The results also show that QSRED, RTRED and WTRED achieve greater network performance over other strategies.