Statistical analysis of network traffic inter-arrival

Self-similarity with Long Range Dependence (LRD) in network traffic has been modeled using aggregation from On-Off sources. The model requires that each of the On and Off period distribution follow Power-tail distribution like Pareto distribution. This paper characterizes of burst inter-arrival time of network traffic using Bellcore Morristown Laboratory Data and an Academic Institute LAN trace. It is shown that, Log-Normal distribution is a better fit for the data compared to Pareto. Current work analyzes statistically both the data from spatial as well as temporal view point.