In scope of peer-to-peer multimedia content delivery, we focus on:
BitTorrent has become a de-facto standard in peer-to-peer content distribution over the Internet. However, it suffers from one significant problem: the long-term availability of content. Our large-scale measurement studies reveal that 38% of torrents become unavailable within their first month, directly affecting over 1,59 million of the users from our traces. Through analysis of chunk availability it can be shown that the primary cause for unavailability is a lack of seeders. In this regard, cross-torrent collaboration offers a great potential to overcome this problem. Specifically, many users nowadays are active in multiple torrents; for instance, a user who joins torrent X and completed the download, might very well join another torrent Y at a later point in time. When the node comes online again to download torrent Y, it could theoretically persist as a replica for torrent X. As a consequence, this cross-torrent approach significantly increases the number of available online seeders per torrent and thus increases the long-term availability of content.
Peer-to-peer paradigm is an attractive way to realize video distribution through the Internet at low cost. By using the peer resources both the distribution costs and the scalability of the system can be improved drastically. A common solution is to organize peers in a mesh-based overlay where peers contribute their upload capacity and offload content provider's servers. However, it becomes difficult to achieve the service level guarantees comparable to well-dimensioned server-based systems because of high system dynamics when peers join and leave the video streams. Additionally, the resources of users are heterogeneous: some users have high bandwidth capacities and others quite low. This especially applies to the upload capacities, that are typically much lower than download capacities. This situation can be alleviated by moving from the today's best-effort systems to more controlled peer-to-peer overlays. Here peers collaborate to assure content availability and sufficient amount of online resources. In our collaborative caching overlay, peers exchange information about content popularity and availability to assure high hit rates and appropriate distribution of load among peers.
Although P2P video streaming has recently been a major research area, streaming over P2P is still best effort and suffers from lack of adaptation. Our research focuses on designing a P2P streaming system with an inherent support for adaptation. By leveraging on scalable video coding, a streaming system has the ability to adapt to different requirements and constraints that heterogeneous peers have in today’s Internet. In addition, we build on intelligent trackers, that not only monitor the quality of experience of peers, but also guide the peers to the best scheduling strategies to perfectly match QoS to resources. By this the P2P system is brought to self organization.