Skip to Main Content
Several successful commercial P2P-TV applications are already available. Unfortunately, some algorithms and protocols they adopt are unknown, since many follow a closed and proprietary design. This calls for tools and methodologies that allow the investigation of the application behavior. In this paper, we present a novel approach to analyze the graph properties and the traffic generated by P2P-TV applications run by customers in operative networks. The proposed methodology allows us to distinguish and investigate three different graphs: (i) the social networks that link users based on their interest, (ii) the overlay networks created by peers that are watching the same channel, and (iii) the distribution networks that involve the subset of peers that are contributing to the video distribution. We apply this methodology to the traffic collected for more than one year from three national ISPs in Europe, where SopCast is the largely preferred application. Considering users' behavior, we uncover the attitude to use the P2P-TV application mainly to follow live sport events. P2P-TV systems have then to deal with both flash crowd and sudden peer departures that happen at the beginning and end of an event. Furthermore, zapping among channels offering the same event is also relevant. SopCast deals with this by implementing a very robust and greedy overlay topology discovery process in which more than 170 peers are contacted every 60 s. Considering video distribution, we provide evidence that SopCast implements algorithms that restrict traffic within Autonomous System boundaries. Still, high bandwidth peers must be present to supply the necessary upload capacity to sustain the video service.