Fairness Improvement About Random Early Detection Algorithm

Along with the reform and development of Internet technology,network congestion is sure to follow on.To solve this problem,AQM algorithm is a key technology.Therefore,as the most widely used algorithm of AQM,RED algorithm is drawing more and more attention.But a great number of studies have revealed that there is the fairness problem of bandwidth resources in RED algorithm,which cannot effectively control non-adaptive flow,sharing a larger possession of bandwidth resources.In this way,adaptive flow is easily affected in the unfair sharing of bandwidth resources.This dissertation analyzes the current RED algorithm and the review of the fairness improvement of AQM algorithm,then,a method is given to improve Max-Mini fairness of RED,which presents a new fairness improved algorithm, namely DF-RED.Here,we had carried out the research as following three aspects:(1) In this dissertation,a dynamic mechanism for fairness assessment is presented.It can periodically give an observation in the actual bandwidth occupation of different traffic flows,and dynamically adjusting its fairness,therefore,to increase the packet loss rate of the higher bandwidth flows and to ensure the quantity of the lower bandwidth flows as possible,and finally,achieving the fairness among different flows.(2) In this dissertation,dynamic fairness improvement based mechanism for calculating drop probability is presented.It gives higher bandwidth flows much more punishment,and lets other lower bandwidth flows get more resource. Meanwhile,this new method can also reserve the traditional RED's benefit of congestion anticipation when it works in fair situation.(3) This dissertation also optimizes cross-layer design of DF-RED on the basis of IEEE 802.11e.Through comparative analysis the fairness of RED and DropTail in the wireless environment,drawing a conclusion that cross-layer optimized DF-RED algorithm can also possess better fairness safeguards in the wireless LAN environment.