A Proposal of Traffic Model That Allows Estimating Throughput Mean Values

Abstract

The present paper is aimed at developing a multi-variable traffic model of a Wi-Fi data network that allows estimating throughput mean values. In order to construct the model, data corresponding to an 8-host wireless ad-hoc network were collected using a software package called Wire Shark, the network was specially designed for modeling purposes. Subsequently, the most convenient multi-variable models were estimated according to the traffic features extracted from the collected data. Results were the evaluated using a software package called STATA, leading to the establishment of significant explanatory variables for the model and its performance levels. For our Wi-Fi network, results show that the analyzed traffic exhibits self-similarity features. Additionally, model coefficients and their corresponding significance levels are shown in various tables. Finally, an explanatory multivariable model consisting of four variables was produced on the basis of ordinary least-squares methodologies (with a per-cent error of 22, 16). The findings suggest that the multi-variable traffic model produced in this study allows a reliable analysis of throughput mean values, however, the model is limited when predicting traffic values for data outside the selected estimation set.