Short-term Mutual Dynamics in Campus Local Area Network Teletraffic

Statistical characteristics of the cooperative user activity were extracted from the local area network traffic patterns obtained at two different university campuses and investigated using three different mutual information metrics. The first two metrics were based on the cross-correlation function estimate and focused on the (normalized) magnitude and shift of its main peak, respectively, while the third metric implied instantaneous phase analysis based on the Hilbert transform. Statistically significant correlations between the total traffic amount received by a network node in a given time window and its cooperativity metrics have been revealed by regression analysis. Moreover, queuing system performance simulation indicated that the average sojourn time is also significantly correlated with the average cooperativity metrics. We believe that both algorithms and software developed for this study could be useful for an improved assessment of the statistical properties of the traffic patterns at different scales and for different network configurations.