Skip to Main Content
Several traffic measurements reports have shown the presence of persistent correlations in current networ Its. The essence of this behavior can be captured by several classes of stochastic processes. The M/G/infin process plays a key role for this purpose. In addition to its theoretical simplicity, it has been successfully used to model widely different traffic sources. In this letter, we obtain the parametric form of the spectral density of an M/G/infin-based process, able to model both the long-term and the short-term correlation behavior of traffic, and we check the robustness of the Whittle estimator taking this process as the underlying one.