Diffserv-based Active Queue Management

With the rapid development of Internet, various network business have put forward higher and higher requirement to QOS, which result in the presentation of DiffServ model. The queue management, which relizes the function of classification and adjustment, is an important part of the edge router of DiffServ. The dissertation makes a series of research on the Active Queue Management algorithms of DiffServ.Random Early Detection(RED),which is representative algorithm of AQM,performs better than PQM(Passive Queue Management) on each performance index.The dissertation focuses on discussing and analysing the RED algorithm through NS2 simulator,we find that there are some shortcomings on parameter config and bandwidth about RED algorithm. Existing algorithms normally improve RED algorithm on these aspects.Marking idea is important to DiffServ model.Through the mark of different data flows, it produces different priorities and realizes different QOS. The dissertation applies Marking ideas to RED algorithm directly ,then improves queue management by marking special events.At the same time,basing above analysis of parameter config of RED algorithm,the dissertation applies a new method of dynamic parameter adjustment to RED algorithm. Combining with two methods, we present a new improved algorithm—SMRED algorithm, which is based on Marking ideas and dynamaic parameter adjusting.The dissertation tests the improved algorithm through the simulation experiments. The experiments show SMRED algorithm performs better than original algorithm on many performance indexes.The dissertation establishs some theory and experiment foundation, which make router offer better queue management and better net QOS.