A Robust Algorithm for the Membership Management of Super-peer Overlay

Abstract. Peer-to-Peer technologies have been widely applied for multimedia applications. The super-peer based approach provides an efficient way to run applications by exploring nodes' heterogeneity. In P2P live video streaming, even though the number of stable nodes is small, they have significant impact on the performance of the network. Thus, we present a super-peer-based overlay design, where stable nodes are assigned as super-peers that organize client nodes. A gossip-based super-peer selection algorithm (GSPS) is proposed to identify the stable nodes to be chosen as super-peers and to manage the client nodes (namely the membership management). The basic idea of the GSPS is: first, a set of super-peer candidates for a node is built based on the gossip, then the role of this node is identified and the corresponding operations are executed. Simulation results show that the GSPS is efficient in managing the super-peer overlay and robust to the failure of super-peers. Keywords: Peer-to-Peer, super-peer, gossip, robustness. Introduction The peer-to-peer (P2P) paradigm provides an effective approach to construct largescale systems with high robustness, mainly due to their inherent decentralization and redundant structures In these research studies, Random walk and Graph theory are utilized for the construction of the overlay. For the management of the overlay, the diameter, and the degree are two important qualities to consider. The super-peer-based overlay (such as Gnutell