The validity of self-similarity assumptions when modelling internet traffic on broadband networks

Self-similarity is a major assumption used in the modelling and characterization of Internet traffic. This paper examines the role of self-similarity in the development of bandwidth capacity modelling techniques and then presents the results of a case study analyzing Internet traffic on a global broadband network. Most modelling techniques of today express network traffic as a function of fractal Browning Motion, as proposed by Ilkka Norros [1], which are based on assumptions that the traffic is self-similar and long range dependant. This study analyzed the network traffic characteristics of airline users on an example broadband network. The data was collected from four different gateways at geographically diverse locations. Collected data was analyzed for self-similarity and the applicability of fractal Browning motion. A distinct trend away from self-similarity was observed and probable causes are postulated.