Generation of Synthetic Network Traffic Series Using a Transformed Autoregressive Model Based Adaptive Algorithm

Abstract

Modeling, testing and simulation of communications systems require to generate synthetic traffic series that present characteristics and behaviors as similar as possible to those of real network traffic traces. In this paper, we propose an adaptive algorithm to estimate the parameters of the Transformed Autoregressive Moving Average (TARMA) model in order to capture the autocorrelation function and the cumulative density function of the desired network traffic trace. Different from other works involving TARMA, we consider the adaptive estimation of its parameters. We compare the performance of the proposed on-line modeling approach to those of the Autoregressive Moving Average Model (ARMA) and of the on-batch Transformed Model in terms of mean, variance, moments, autocorrelation and probability density function. A transmission link composed of a single server with buffer is also simulated, which proves the efficiency of the proposed model in describing real traffic traces. The simulations carried out in this work show that the adaptive TARMA model outperforms in general the other considered autoregressive models.