A WTP-ARED Active Queue Management Algorithm Based On Traffic Predictability In Self-similar Network

With the popularity of network application and the increase of global communication business, effective control and management of network traffic are particularly important. For a long time, network traffic modeling and analysis based on the Poisson distribution and Markov process theory. However, in recent years plenty of observation studies about network traffic show that network traffic have a character of self-similarity which can not be described through traditional mathematic model. At the same time, the fact that self-similarity theory applied for the studies of network traffic features has paid much attention. Active Queue Management algorithm based on self-similar network traffic predictability is one of the research focuses, and is also the content of this paper.This thesis firstly describes the self-similar phenomenon of network traffic, and introduces the definition and features of self-similar. It is also give a detailed analysis of the effects of self-similarity on the network performance.The paper makes a comparative study of the two of the primary traffic predictability model: fractal prediction model and statistical forecasts model, and on the basis presents a method of the weighted average traffic forecast based on time-lag. Then with the purpose of managing and forecasting the future traffic, the paper sets up appropriate functions and dependencies for the data has been arrived, and analysis the inherent law of the data. Afterwards the paper analysis the two active queue management algorithms: RED algorithm and ARED algorithm, and on this base raises a theory of WTP-ARED (Weighted Traffic Predictability Adaptive RED algorithm) algorithm based on self-similar network traffic predictability. In the calculation of the average length of the queue, uses the above discussion about the weighted average traffic method to predict the future traffic, which would bring the self-similar property of the traffic into ARED algorithm, and making the active queue management mechanism combine with the future traffic for controlling the length of the queue better. The paper carries through some emulator experiments via NS2 platform for WTP-ARED algorithm about throughput, delay, delay jitter, packet loss, queue length, and other parameters. Final results showed that in the network, the algorithm can control the average queue length better, can maintain the stability of the queue, so as to enhance the network performance.