Dynamic matrix control for active queue management based on step response model

Most of existing active queue management (AQM) algorithms were designed on the fluid model of TCP/AQM dynamics which ignores the slow start and TCP timeout mechanism. Modeling TCP flows more precisely is very important to study AQM performance. To address this problem, this paper develops the step response model of TCP/AQM network based on the dumbbell topology in NS2. Then a novel AQM algorithm (DM-AQM) based on dynamic matrix control (DMC) is proposed in dynamic TCP networks. According to the predicted future queue length in data buffer, early packets at router are dropped reasonably so that the queue length reaches the desired value with minimal tracking error. The simulation results show that DM-AQM algorithm outperforms RED and PI in terms of stability and responsiveness.