Adaptive Active Queue Management Algorithm

In the past few decades, the Internet has evolved from an academic network founded by the U.S.Goverment to a global commercial network.With the rapid development of computer networks, more restrictive demands are made on them. Nowadays more and more real time services such as voice, image and critical data are transmitted on networks, and all of them are sensitive to bandwidth, transmission delay and delay jitter. This situation enrichen the network resources, but at the same time puts forward the questione of how to assure the quality of service in network. In the case of most connections linked to a bottle-necked node at the same time, networks that attempt to deliver more data than their capacity will experience congestion, causing buffers to overflow and leading to undesirable data losses. So, the primary task of improve network resource utilization and improve quality of service is design effective algorithms to control network congestion.Based on systematical introduction of TCP congestion control and AQM congestion control, some typical AQM algorithms have been summarized, and some of these algorithms are simulated under many kinds of network environment so sa to compare the validity and robustness of each algorithm by NS-2 in this paper.Due to the parameters of PI and PID algorithms are pre-selected, once selected will not change , it is not well adapted to the changing reality of the complexity of the network system. In addition, PI and PID algorithm parameters setting are difficult.a parameters adaptive PI algorithm (API) and a parameters fuzzy adaptive PID algorithm (FLA-PID) are proposed in this paper, these two algorithms can update the parameters according to network congestion in real time. And then carried out simulation, analysis and comparison of API algorithm and FLA-PID algorithm. It is found that under different environments API algorithm is better than PI algorithm, Simulations under many various network environments show that FLA-PID algorithm can stabilized the queue length to the target value and has good stabilization and robustness.