Comparison of Effective Bandwidth Estimation Methods for Data Networks

The purpose of this work is to apply techniques to estimate the Effective Bandwidth, from traffic traces, for the Generalized Markov Fluid Model in data networks. This model is assumed because it is versatile in describing traffic fluctuations. The concept of Effective Bandwidth proposed by Kelly is used to measure the channel occupancy of each source. Since the estimation techniques we will use require prior knowledge of the number of clustering clusters, the Silhouette algorithm is used as a first step to determine the number of classes of the modulating chain involved in the model. Using that optimal number of clusters, the Kernel Estimation and Gaussian Mixture Models techniques are used to estimate the model parameters. After that, the performance of the proposed methods is analyzed using simulated traffic traces generated by Markov Chain Monte Carlo algorithms.