LocalTree: An Efficient Algorithm for Mobile Peer-to-Peer Live Streaming

Abstract—Mobile peer-to-peer (P2P) technique is a promising approach to provide live video streaming. In mobile P2P streaming, mobile devices relay their received packets in a multi-hop manner by means of broadcasting in a secondary channel (such as Wi-Fi or bluetooth). We consider the general scenario that both wireless channel and nodes may not be reliable, i.e., packets may be lost in channel transmission and a node may forward its received packets probabilistically (due to, for examples, its power and incentive concerns). Under this scenario, we study how to achieve broadcasting minimizing the energy consumption in the network while meeting a certain stream quality requirement (in terms of received packet loss rate). We first formulate the problem under study, and propose and study a distributed algorithm called LocalTree which addresses the problem. LocalTree takes advantage of stable clusters of users to optimize the construction of its streaming overlay. It combines the strengths of both tree-based and mesh-based algorithms, and is simple and effective. Simulation results show that LocalTree performs similar to tree-based algorithm for a rather stable network, and exhibits robustness similar to unstructured algorithm if the network is dynamic. I