Research On Congestion Control Strategies Based-on RED

The key reason that results in the network congestion lies in the overload that users(end-system) put on the network.This overload results in the increase of datagram delay,the increase of packet drop probability,the performance in application system in upper layer getting worse, ete.The congestion control is taking measures to avoid or to respond the occurrence of congestion in the network nodes .The congestion control algorithms can be divided into the source algorithm (source algorithm) and link algorithm (link algorithm) two broad categories based on realization its position.At present, what in the source algorithm uses most widespread is in the TCP agreement congestion control algorithm,and link the algorithm research was concenta-ed in "Active Queue Management" (referred to AQM) area,RED (Random Early Detection) is a model for AQM algorithm.Research shows that RED has better performance than the traditional "to the end" (Drop Tail) algorithm, but it there are two main shortcomings, one is that RED is very sensitive regarding the parameter set,another question is with the increase in the number of the network " Flow ",( means one TCP connection),in the router the average queue size will increase gradually.Therefore, the researchers have made the improvement to it,ARED(Adaptive RED)is modification relativly small one algorithm.This article introduced the ARED algorithm's principle of design, analyed the algorithm performance, and discussed questions of parameter setting and so on.In view of the RED algorithm existence's insufficiency as well as the existing max_p adaptive algorithm, considered the role of packet dropping probability, link utilization, the average queue length in the parameters adaptive process, proposed one kind of grading adaptive model of RED parameters, GARED algorithm, to achieve the purposes of reducing packet dropping probability, improving link utilization.