Aggregated Network Traffic Modeling based on Superstatistical Approach with Account of Long-Term Dependence and Non-Stationary Dynamics Effects

Abstract

A superstatistical approach that takes into account the long-term correlation and the non-stationary dynamics is proposed fo r modelling aggregated traffic with non-stationary dynamics. By means of queuing system simulation, it is shown that traditional approximation based on Kingman's formula underestimates the average sojourn time by up to two decades at high utilization. On the contrary, the use of alternative superstatistical model taking into account the longterm correlation, this underestimation can be reduced by more than one decade.

Keywords

• сетевой трафик
• долговременная зависимость
• производительность смо
• суперстатистики