As the rapid development of Internet, network congestion has become an important issue. Recently AQM (Active Queue Manegement) has been a hotspot in studies on the end-to-end congestion control. By evaluating the network state and predicting the incipient congestion, it can drop packets purposefully, thus sources can be well informed the network state and then adjust its sending rate. As the only candidate algorithm for AQM recommended by IETF (Internet Engineering Task Force), the RED (Random Early Detection) algorithm is still not perfect in terms of respose time, stability and parameter setting. This paper revises the parameter setting of RED algorithm, and validates the revised algorithm using NS2 (Network Simulator Version Two). Main reaserch work includes the following:This thesis first works on the parameter setting issue of RED: the weight wq, the maximum drop probability maxp, the minimum threshold minth and the maximum threshold maxth. Then while leaving the basic idea intact, this paper revises parameter setting of the RED algorithm on scenarios with small RTT (Round Trip Time): revising the weight wq to better calculate the average queue lenth and revising the parameter interval to shorten Adaptive RED's respose time. Last, this paper validates the modified RED using NS2.The research in this thesis provides a valuable reference for further study on AQM algorithm; furthermore it is of great value to the RED deployment on routers. |