Temporal-DHT and Its Application in P2P-VoD Systems

Peer-to-Peer (P2P) based Video-on-Demand (VoD) streaming services are gaining popularity. Meanwhile, the technical barriers of P2P-VoD systems still exist because of asynchronous and dynamic user interactivity where any user can jump to random locations in the video stream along with the support for Fast-Forward/Fast-Backward operations. An important component to address these issues lies in efficient neighbor discovery mechanism for searching and receiving desired data in a peer-to-peer fashion with less burden of the server. We advocate a Distributed Hash Table (DHT) based overlay for efficient lookup with its desirable features of provable scalability, robustness and decentralization for supporting Internet-scale P2P applications. However, one big challenge associated with a DHT-based overlay is how to accommodate a large number of update operations with the continuous change of peers' playing position. To solve this inadequacy, we propose a novel concept of temporal-DHT which exploits the temporal dynamics of the content to estimate the playing position of peers, thereby greatly reducing the update overhead. We also propose a new content management and discovery scheme by the skillful integration of: (1) random/out-of-order/static caching to support interactive operations and (2) playing/in-order/dynamic caching to maintain streaming efficiency and playback continuity. Extensive simulation results demonstrate that our mechanism achieves low messaging overhead with high streaming quality for P2P-VoD systems in a dynamic environment where peers asynchronously join/leave the network.