Characterizing impulsive network traffic using truncated α-stable processes

It has been recently recognized that aggregated traffic in a variety of networks exhibits a similar impulsiveness over a wide range of aggregation levels, but approaches a Gaussian distribution in the limit as the aggregation level grows. Although several traffic models have been proposed in the past decade, their accuracy in simultaneously characterizing the above properties still needs to be further improved. In this letter, we propose a truncated α-stable process model which is able to capture the impulsiveness of observed network traffic as well as its tendency toward the Gaussian distribution with aggregation. An inherent physical mechanism is also proposed to give insight into the underlying meaning of the proposed model. Simulation results show that the proposed process achieves close agreement with real traffic and outperforms previous models.