When analyzing computer networks with the goal of optimizing its performance and detecting anomalies, an important piece of information is the generation of network traffic in terms of the distribution of inter-departure times of requests. This work analyzes network traffic captured at the Faculty of Computer and Information Science, University of Ljubljana. Our goal was to find the theoretical distribution that most closely resembles the experimentally obtained distribution of inter-departure times. The captured traffic was analyzed in several steps, where the data was divided according to time intervals and traffic generators. To determine the most suitable distribution, we used the Kolmogorov-Smirnov metric along with QQ plots. The results obtained led us to reject our original hypothesis that the inter-departure times follow a Pareto distribution. The most suitable distribution was found to be the Pareto distribution of the second kind.
|
