Xin Jin and Yu-Kwong Kwok
With recent advancements in computing and communication technologies, peer-to-peer (P2P) architectures have been successfully deployed and extensively studied in both industry and research communities. In essence, P2P networks are self-organizing and scalable distributed systems to fully exploit computing, streaming, and storage capacities of geographically distributed client users [1‒3]. Numerous application scenarios of P2P overlays exist in the market, including both time-insensitive services (e.g., file sharing) [4‒6] and instantaneous streaming systems(e.g., live video broadcast) [7‒12]. Among them, scalable large-scale P2P multimedia streaming systems have penetrated into our daily lives by supplying real-time dissemination of media, including audio, video, massively multiplayer online games, and interactive social virtual worlds [13‒17]. In particular, video streaming hasevolved into a killer Internet application  and has predominantly occupied the Internet traffic .
To attain efficient video streaming, various overlay structures have evolved in the literature: single tree , multitree [21‒22], and unstructured meshes [23‒25]. Single tree-based structures are vulnerable to peer dynamics, and leaf nodes cannot contribute their streaming capacities. To this end, multiple-tree structures are proposed by dividing the media stream ...