On the Influence of the Time Scale Range on Network Traffic Modeling Accuracy
Kristof Sleurs, Dagang Li, Emmanuel H. Van Lil, Antoine Van de Capelle · Lirias (KU Leuven) · 2009
Recent models for traffic flowing on computer networks employ a discrete representation of the data set. A mathematical description of the self-similar or multifractal behavior of a packet stream serves as the input to these models. Unfortunately these models are not able to accurately model this behavior on a wide time scale range. Discretisation issues that arise when using very small base time scales are an example hereof. This paper tries to quantify the consequences of decreasing the modeled time scale range. Decoupled experiments were conducted to test the effect of increasing the base modeling time scale and decreasing the upper time scale. For these scenarios, detailed queuing performance results are discussed.