Analysis And Modeling Of The Cascade Property Of Network Traffic

Ever since 1993, when W.E.Leland formally introduced the concept of self-similarity of fractal into the study of network traffic in the field of telecommunication, the approach that the network traffic is characterized by scaling property has become feasible. It has and will go on to influence analysis of network performance, QoS control and network design. Although the following over ten year's research has discovered such important network traffic properties as self-similarity, LRD, heavy-tailed distribution and multifractal, researchers are still sceptical about the causes of these properties (e.g. the development from similarity to multifractal, the uncertain of estimation of LRD exponent, and so on), doubt their real existence, and, with the development of traffic content and network structure, feel uncertain about the viability of these properties. Therefore, it is very important to identify, measure and model those properties that do not change or change slowly .Due to ever-increasing complexity of the Internet, many researchers attempt to describe network traffic by dynamics, which they think is able to describe complex system such as turbulence. IDC is a new product of this attempt. Thus, on the basis of this model, this paper intends to explore the characterization of network traffic, which includes checking, analysis, estimation, and synthesis of IDC.The 2nd chapter of this paper summarizes concepts and models put forward by researchers from various academic backgrounds. After studying their theoretical backgrounds, basic principles and the correlation between each other, I have decided that the approaches of characterizing network traffic should fall into two categories: one is application of point process, renewal process, random walk, and of thereby derived classical probability and queuing theory . Models derived from Markov sequence/process should also belong to this category. Another is the gradual development of the fractal scaling processes, from self-similarity to multifractal, and then to IDC. The first category can vividly show what the net-...