dHCPS: Decentralized Hierarchically Clustered P2P Video Streaming

Peer-to-Peer (P2P) based live video streaming service has been gaining popularity. The new generation of P2P live streaming systems not only attracts a large number of viewers, but also increasingly provides better video quality by adopting advanced coding techniques and supporting higher streaming rates. It is, therefore, crucial to design scalable P2P streaming solutions that offer excellent user Quality of Experience (QoE) in order to prevail in a competitive market place. In this paper, we propose a decentralized Hierarchically Clustered P2P Video Streaming (dHCPS) solution that can accommodate large viewer populations and support streaming rates approaching the optimal upper bound. In dHCPS, the peers are grouped into clusters and a hierarchy is formed among peer clusters. Peers retrieve video data from the source server in a hierarchical fashion. dHCPS employs a distributed P2P streaming design with minimum coordination overhead among peers. We propose a queuebased peer uploading algorithm that adaptively converges to the optimal operating point in a dynamic network environment. The prototype of dHCPS is implemented. Experiments over the PlanetLab demonstrate the effectiveness of dHCPS.