While several deployed p2p file sharing overlays have been characterized in the literature, this paper shows that some of their conclusions may be false for p2p applications that stream media instead. Specifically, we undertake a crawler-based investigation of PPLive, the largest live multimedia streaming system in the world today. It is important to understand IPTV overlays like PPLive in order to enable the building of larger-scale media streaming overlays. Our task is challenging because PPLive is proprietary. PPLive has multiple channels, each channel with its own overlay, and a large fraction of these channels stream preset movie schedules. A human user may join any given channel, but the user's client machine could be used to relay feeds for other non-subscribed channels too. Popular PPLive channels contain several thousands of nodes.We crawl the real deployed PPLive network via both machines in a cluster at UIUC, and by using PlanetLab hosts. Our major findings are: (1) Unlike p2p file sharing users, PPLive peers are impatient, (2) Channel Size variations are larger than in p2p file sharing networks, (3) Average degree of a peer in the overlay (i.e., its out-degree) is independent of channel size, (4) Smaller PPLive overlays are similar to random graphs in structure, (5) The availability correlation between PPLive peer pairs is bimodal, i.e., some pairs have highly correlated availability, while others have no correlation. We believe these results point us towards taking seriously the nature of applications while designing and optimizing p2p overlays.
展开▼