Traffic Modelling with Stochastic Lindenmayer Systems

Abstract:
Traffic characteristics in communication systems have changed massively in the past. However, these changes are ongoing as this is a continuous process caused by constantly changing technologies, services and usage patterns. Simultaneously, research enabled a refined understanding of traffic characteristics. An important and traffic invariant characteristic is self-similarity and long range dependence. This is also present in the Internet backbone, impacting particularly performance evaluation studies due to its significantly slower decrease of correlations/ variability. Therefore, it essential that traffic models reflect the relevant characteristics of real traffic appearing in current networks and provide additional flexibility for further changes. Lindenmayer System (L-System) based traffic models provide such a flexibility. In this paper, we extend the L-System model to the bit-rate level. We show the good modelling properties of the extended L-System model on basis of artificial traces and real traces of a broad-band user access rate scenario. We describe the developed modifications of the fitting process to accommodate the model extension.