The Use of Long Range Dependence for Network Congestion Prediction

Traffic measurements from communication networks have shown that network traffic is exhibiting self similar as well as long range dependence properties. In telecommunication networks, congestion events tend to persist, producing large delays and packet loss resulting in performance degradation. In order to guarantee quality of service to diverse Internet services, congestion prediction became a fundamental objective of network management algorithms. High performance predictors are required that are efficient and simple to implement. In this paper we use a number of ON/OFF sources to generate long range dependence, self-similar network traffic. We propose a novel model for congestion prediction, by using the mean time spent ON for each node as an indicator of which node is causing congestion. The simulation results have shown that applying the algorithm can provide better performance in terms of the delay and the mean queue length.