Research On Loss Differentiation Algorithm And Congestion Control Mechanism In Heterogeneous Networks

With the rapid development of communication technology and Internet, wireless network hasbecome an important part of globlal communication system. Next Generation Network(NGN)certainly will be a single logical overlay network composed of many heterogeneous networkswith complex implementation. However, traditional transmission control mechanism simplycontributes all loss events to network congestion and blindly takes control strategy, that impactits performance dramatically in wired-wireless heterogeneous networks due to its complicatedloss reason.This paper focuses on improving performance of multimedia transmission control inwired-wireless heterogeneous networks based on Loss Differentiation Algorithm(LDA). Themain contents of the study are summarized as follows:First, general strategy to solve the problem is summarized in this paper, and typical methodsof distinguishing the reasons for packet loss are analysed, classified, and compared, whichprovides a good basis for further study.Second, this paper puts its emphasis on End-to-End LDAs due to its advantages of easyimplementation, high compatibility and configurability. This paper proposes a new LDA namedUBLDA(Utilization Based Loss Differentiation Algorithm), which has the ability of estimatingstate(congestion or not) of network by calculating utility parameter based on the interval ofpacket arrival and sending rate. UBLDA can choose appropriate differentiation scheme based onstate estimation, and integrates with TCP friendly transmission control protocol, making it a newcongestion control mechanism suitable for wired-wireless heterogeneous networks.Finally, simulation experiments in NS2(Network Simulator-2) are carried out. The resultsshow that the proposed algorithm has a promotion in the aspect of differentiation precision andresource utilization under variety of traffic/topology circumstances, which means that thealgorithm could, to some extent, mitigate the defects of high misclassification rate and lowversatility of existing End-to-End LDAs...