Early detection of interactive user experience degradation in self-similar traffic

Since the last years of the past decade, it is well known that chaotic traffic produced by a large number of workstations that access to the Internet is self-similar, provided that the bandwidth is enough so that there is no significant regulation on packet flow. In this paper, we propose the Logscale Diagram study as a leading tool in connection quality evaluation and monitoring. This method provides an objective way to assess interactive user experience (subjective parameter), either the network is deployed or just simulated. To support this, we use empirical customized NS-3 simulator´s classes and libraries, to show that the study of variations in packet inter-arrival time standard deviation (information that is present in Logscale Diagrams) can be useful in early detection of connection quality decrease, previous to congestion.