A distributed video-on-demand system (DVoD) with multiple server-nodes is a cost-effective and fault-tolerant solution for a high scalable enterprise video-on-demand (VoD) system. However, such a server-oriented design is highly vulnerable to workload variations given that the service capacity is limited. Peer-to-Peer (P2P) has been introduced as an architectural solution with self-growing capacity. However, the characteristics of a pure P2P system such as the peer transient nature and high network overhead make this kind of architecture unsuitable for a fully interactive VoD system. In this paper, we propose a new efficient integrated VoD architecture, called DPn2Pm, that combines DVoD with a P2P system and multicast communications. The DVoD's server-nodes provide a minimum required quality of service (QoS) and the P2P system is able to offer the mechanism to increase the system service capacity according to client demands. Multicast communication, wherever it is possible, is effectively exploited by our P2P system. In our design, each client is able to send video information to a set of m clients using only one multicast channel. Furthermore, the collaboration mechanism is able to coordinate a set of clients to create one collaboration group to replace the server, providing an extensive, efficient and low network-overhead collaboration mechanism from n-peers to m-peers. Regardless of the video the client is watching, our P2P scheme allows every active client to collaborate with the server. The P2P scheme is complemented with recovery mechanisms that are able to replace the failed client before affecting the QoS, offering continuous playback. The proposed approach has been broadly evaluated, firstly using a mathematical model to derive the theoretical performance and secondly using a simulation environment to analyze the system's dynamic behavior, the VCR interaction impact and the client failures. Comparing DPn2Pm with other DVoD architectures and the most relevant P2P delivery policies, we show that our design is an improvement on previous solutions, providing a higher scalability. © 2010 Elsevier Inc. All rights reserved.
|Journal||Journal of Parallel and Distributed Computing|
|Publication status||Published - 1 Dec 2010|
- Content delivery network
- Distributed VoD