Maximizing streaming efficiency of multiple streams in peer-to-peer networks

Abstract Nowadays, multi-channel streaming and multi-file sharing are attracting increasing attention. To maximize the streaming efficiency of multiple streams, the multi-stream live streaming problem is studied from a practical point of view in this paper. The problem is modeled in a two-layer fully-connected network: the server network and the peer-to-peer (P2P) network. We first formulate the optimal non-forwarding (ONF) and optimal forwarding (OF) strategies into linear programs (LPs) to get the de facto upper bounds. Then, to facilitate the collaboration among servers while keeping as even load distribution as possible, we propose a novel water-filling (WF) algorithm that converts the original problem into solving a series of small linear systems (LSs). The WF algorithm proves to be of near-optimal efficiency and pretty low time complexity. Finally, we evaluate the obtained theoretical bounds, strategies and algorithm with key parameters, such as the bandwidths of the servers and P2P nodes, and the P2P network size. To expand the applicability, we also briefly discuss how to extend the proposed solutions to general networks. The flow-based method that we use is insightful for studying the intrinsic features of the problems related to multiple streams, and the obtained results are expected to be widely used for content distribution in datacenter networks (DCNs) and P2P networks.