Collaborative Streaming System Design as an Underlay to Cellular Networks

Media content sharing over a peer-to-peer (P2P) network is promising because of its distributive nature and scalability to network size. Recently device-to-device (D2D) communication as an underlay to cellular networks has been introduced as a technology component to the Long Term Evolution Advanced. The short range of D2D transmission exploits the reuse of frequency within a cell. In this work, we propose a collaborative streaming system where geographically close peers form groups and then share the content with others through a D2D multicast mechanism. We present the optimal grouping problem as the stable admission assignment problem. We also propose a distributive algorithm that forms groups of suitable sizes to tackle the retransmission problem. We present a mathematical model for peer grouping process that finds the relationship between two important measurements, the number of groups and group sizes, affecting the bandwidth savings and streaming rate. However, D2D transmitters may cause a significant amount of interference with the primary cellular network when radio resources are shared between them. We propose a heuristic radio resource allocation algorithm that can lessen interference to the primary cellular network with coordination from an LTE access point. We apply the theory of epidemiology to derive an analytical model for chunk distribution of live streaming content. The model finds buffer occupancy probability and provides insight on relationships among peer population size, buffer length and playback smoothness. We also propose a hybrid chunk selection method that is more scalable to the peer population in buffer space and also achieves better playback continuity by distributing urgent chunks first. We conduct extensive simulations to support our claims.