... to flood Internet access and backbone ISPs with massive amounts of new traffic. We recently measured 200,000 IPTV users for a single program, receiving at an aggregate simultaneous rate of 100 gigabits/second. Although many architectures are possible for IPTV video distribution, several chunkdriven P2P architectures have been successfully deployed in the Internet. In order to gain insight into chunk-driven P2P IPTV systems and the traffic loads they place on ISPs, we have undertaken an in-depth measurement study of one of the most popular IPTV systems, namely, PPLive. We have developed a dedicated PPLive crawler, which enables us to study the global characteristics of the chunk-driven PPLive system. We have also collected extensive packet traces for various different measurement scenarios, including both campus access network and residential access networks. The measurement results obtained through these platforms bring important insights into IPTV user behavior, P2P IPTV traffic overhead and redundancy, peer partnership characteristics, P2P IPTV viewing quality, and P2P IPTV design principles.

This paper presents results from our measurement and modeling efforts on the large-scale peer-to-peer (p2p) overlay graphs spanned by the PPLive system which is arguably the most popular and largest multimedia streaming p2p system today. We believe that our findings can be used to understand large-scale p2p streaming systems for future planning of resource usage, and to provide useful and practical hints for future design of large-scale p2p streaming systems. Unlike other previous studies on PPLive, which focused on either network-centric or user-centric measurements of the system, our study is unique in (a) focusing on PPLive overlay-specific characteristics, and (b) being the first to derive mathematical models for its distributions of channel population size and session length. Our studies also reveal characteristics of multimedia streaming p2p overlays that are markedly different from existing file-sharing p2p overlays. Specifically, we find that: (1) Small PPLive overlays (as many as 500 nodes) are similar to random graphs in structure, (2) Average degree of a peer in the overlay (i.e., its out-degree) is independent of channel population size, (3) The availability correlation between PPLive peer pairs is bimodal, i.e., some pairs have highly correlated availability, while others have no correlation, (4) Unlike p2p file-sharing users, PPLive peers are impatient, (5) Session lengths (discretized, per channel) are typically geometrically distributed, (6) Channel Population Size variations are larger than in p2p file-sharing networks, yet they can be fitted with polynomial mathematical models. We conclude with a series of suggestions on how our findings can improve IPTV future design. 1.

Emerging large scale Internet applications such as IPTV, VOD and File Sharing base their infrastructure on P2P technology. Yet, the characteristic fluctuational throughput of source peers affect the QOS of such applications which might be reflected by a reduced download rate in file sharing or even worse- annoying freezes in a streaming service. A significant factor for the unstable supply of source peers is the behavior of other processes running on the source peer that consume bandwidth resources. In this paper we present Collabrium- a collaborative solution that employs a machine learning approach to actively predict load in the uplink of source peers and alert their clients to replace their source. Experiments on home machines demonstrated successful predictions of upcoming loads and Collabrium learned the behavior of popular heavy bandwidth consuming protocols such as eMule & BitTorrent correctly with no prior knowledge. 1.

In theory, peer-to-peer (P2P) based streaming designs and simulations provide a promising alternative to serverbased streaming systems both in cost and scalability. In practice however, implementations of P2P based IPTV and VOD failed to provide a satisfying QoS as the characteristic fluctuational throughput of a peer’s uplink leads to frequent annoying hiccups, substantial delays and latency for those who download from it. A significant factor for the unstable throughput of peers ’ uplink is the behavior of other processes running on the source peer that consume bandwidth resources. In this paper we propose Maxtream- a machine learning based solution that actively predicts load in the uplink of streaming peers and coordinates source peers exchanges between peers that suffer from buffer underrun and peers that enjoy satisfactory buffer size for coping with future problems. Simulation and experiments have shown that the solution successfully predicts upcoming load in popular protocols and can improve the QoS in existing P2P streaming networks. 1.

Résumé — Les réseaux pair-à-pair (P2P) jouent aujourd'hui un rôle primordial dans la livraison de la vidéo à grande échelle. La perte des paquets sur les réseaux IP, peut affecter défavorablement la qualité de la vidéo reçue. Différents mécanismes de réparation de perte sont utilisés aujourd'hui. Ces mécanismes ont été proposés initialement pour des applications n'ayant pas de contraintes de délai de réparation. Ils ne peuvent pas, donc, garantir la qualité de la vidéo en cas de perte. Nous proposons dans ce papier un nouveau mécanisme de réparation de perte. Il permet d'optimiser la qualité de la vidéo transmise sur les réseaux P2P. Son principe consiste à demander les paquets perdus à un pair sélectionné aléatoirement parmi les pairs du réseau ayant ces paquets. Ce pair ne sera pas systématiquement leur fournisseur initial. Ce mécanisme vise à augmenter la probabilité de choisir un pair disponible pour faire la retransmission. Ceci peut augmenter la probabilité de recevoir les paquets retransmis avant leur temps de visualisation. Ce qui améliore la qualité de la vidéo reçue. Nos études par simulations ont montré l'efficacité de ce mécanisme par rapport aux mécanismes actuels de retransmission. Mots-clès: réseau pair-à-pair, transmission vidéo, perte des paquets, retransmission de paquets. I.

Abstract — This paper explores how to remotely monitor network-wide quality in mesh-pull P2P live streaming systems. Peers in such systems advertise to each other buffer maps which summarize the chunks of data that they currently have cached and make available for sharing. We show how buffer maps can be exploited to monitor network-wide quality. We show that information provided in a peer’s advertised buffer map correlates to that peer’s viewing-continuity and startup latency. Given this correlation, we can remotely harvest buffer maps from many peers and then process the harvested buffer maps to estimate ongoing quality. After having developed this methodology, we apply it to a popular P2P live streaming system, namely, PPLive. To harvest buffer maps, we build a buffer-map crawler and also deploy passive sniffing nodes. We process the harvested buffer maps and present results for network-wide playback continuity, startup latency, playback lags among peers, and chunk propagation. The results show that this methodology can provide reasonable estimates of ongoing quality throughout the network. Index Terms— I.