Multi-bottleneck Link Under Active Queue Management Algorithm

As the rapid expansion of the networked user group and emergence of new networkapplications, Internet traffic has been increasing dramatically, and network congestionappears to be a more and more critical problem. Although TCP congestion controlmechanism based on source nodes is effective in the prevention of congestion collapse, italso faces many new crises. Therefore, it is necessary for the network itself to participate incongestion control. As one of the most important congestion control strategy in thenetwork nodes, AQM (Active Queue Management) has become a hot research area incongestion control these years.In this paper, we first introduce several representative AQM algorithms. In order toreveal the AQM algorithms' performance, these algorithms are simulated using networksimulation software NS-2 in single bottleneck networks. More over, we pay our attentionon the multiple bottleneck networks, seeing about the acting of AQM algorithms undercomplex environment, then we give a general term on it.Due to the weakness of old AQM algorithms, a new AQM compensatory algorithmbased on queue length and mutative rate of qUeue length is proposed in this paper, throughimproving the old AQM algorithms, new AQM algorithms have better performance thanthe old ones. Simulations under various network environments show that the newalgorithm is able to stabilize the queue length to the target value, improve the systemrespond speed, and have good robustness and adaptability.