Modeling And Performance Evaluation Of Self-Similar Traffic

The discovery of the self-similar characteristic of network traffic has great influence on network traffic modeling, performance evaluation and network control. Traditional Poisson-based models of network traffic are based on the hypothesis of Markov which has the nature of short-range dependence (SRD). Recent traffic analysis from various packet networks shows that network traffic processes exhibit ubiquitous properties of self-similarity and long range dependence (LRD), i.e. of correlation over a wide range of time scales. LRD exists on multiple time scales and has great influences on network performances such as delay, jitter, cell loss rate and throughput on the large time scale. Modeling and performance analysis of self-similar traffic becomes Current hotspot.In this thesis, the problems of network modeling and performance evaluation of self-similar traffic are studied with depth. Firstly, several mathematical definitions of self-similarity are given. Some mathematical and physical features describing the self-similar processes are described. The methods of modeling and generation of the self-similar traffic are discussed. The performance of these models is analyzed. The influence of network performance on self-similariy is studied through simulation.Though many traffic models have been proprosed, this thesis focus on the problems of FBM and FARIMA -based traffic modeling and performance analysis. Mathematical model is provided, and its theory results are derived.In addition, the validity and efficiency of these results are conformed through simulation based on OPNET, so as to test the precision and adaptability of existing self-similar model and find out the best traffic model for givven applications.The FBM traffic model which holds self-similarity is used to study the performance of the G/D/1 queuing model. Based on the buffer overflow rate given by Norros, the asymptotic analytic expression of the network performance with FBM input is obtained. The variance of the system performance indices, such as packet loss probability, effective bandwidth, average delay and queue length, with the model parameters, namely Hurst index, buffer size, utilization ,variance and load of the traffic, is studied through theoretical analysis and simulation. The results show that beside Hurst index, several other parameters, such as buffer size, utilization ,variance and load of the traffic have also great influence on system performance, some influence are even greater than that of Hurst index. The traditional concept which only consist the influence of Hurst index is not overall, it may be sometimes misleading. The study results also reveal that the performance of the FBM model has obviously the characteristic of time scale, the dominative factors of large and small time scale are deffernent, and there is a state change or abrupt changes between small time scale and large time scale.Measurement also shows that network traffic exhibits properties of short-range and long-range dependence. Short-range and long-range dependence have great impact on network performance. FARIMA (p, d, q) model is a good traffic model capable of capturing both long-range and short-range behavior of network traffic. In this paper, FARIMA(p,d,q) model is used to model, generate traffic and estimate parameters which fit the actual traffic trace. The results of analysis and simulation demonstrate that FARIMA model fits real multimedia traffic very good. The short-range dependence is the ascendant with small buffer, while network performance reduces lower in long-range traffic than short-range traffic. These results are very important for the future research of the network performance.