D.K. Swift and C.H. Dagli (USA)
Broadband Network, the Internet, Modelling, Variability
Current methods for modelling network traffic bandwidth do not sufficiently take into account the impacts of traffic variability. The most common models for characterizing network traffic bandwidth are those based on the theories of Illka Norros using fractal Brownian motion [1] or those based on Frank Kelly’s theories using stochastic processes [2]. In both cases the accuracy of the model is dependant on values derived from measured Internet data. These measurements are often sampled at a limited number of points on the network. The task of collecting, evaluating, and constantly updating the data in order to maintain required accuracy is extremely difficult. Data characteristics vary greatly according to time, location on the network, applications in use, and the behavior of users. This paper details the results of a survey examining the degree of network traffic variability on an example broadband network. Impacts are evaluated and the need for adaptive modelling techniques that utilize computational intelligence and adaptive capabilities is proposed.
Important Links:
Go Back
