A sliding-window-based graph model for dynamic characteristics analysis of Internet traffic

With the rapid growth of Internet traffic and applications, user behaviors on Internet become increasingly complicated. It is essential to analyze changes of user behaviors on Internet for service provisioning and network management. In this paper, based on large-scale real flow data collected from ISP network, we propose a sliding-window-based graph model (SWG) from the perspective of complex network, to study changes of interactions of end-hosts in a day for four applications, i.e. P2P Download, Video, Web and Instant Messaging (IM). We experimentally configure key parameters of our model, i.e. window size and sliding time interval, and prove SWG valid. Furthermore, by investigating changes of assortativity in a day, we find that end-hosts´ interactions of the four applications have different changes with time.