Optimal Multi-Tree Peer-to-Peer Video Streaming

Peer-to-peer (P2P) is a promising design paradigm for developing large-scale video distribution systems. In particular, P2P streaming systems have shown to deliver good overall video quality to its users. However, most proposed systems evaluate video quality considering mainly delay and other network related metrics, while considering spatial quality or video resolution as a fixed parameter. Evaluation of these systems usually assume that video encoding rates are much lower than upload bandwidth capacity of peers. Therefore offering lower overall video quality than could be sustained by the system. In this paper, we consider the inherent tradeoff between delay and video resolution, both of which are fundamental in determining overall video quality. In particular, we propose a multi-tree dissemination mechanism that explores and controls this tradeoff. We consider a scenario where the video encoding rate is set to match the upload bandwidth capacity of peers. When assuming the multi-tree approach introduces no overheads, we show that the optimal delay can be constant for any fixed number of peers. We also model and evaluate the system considering the more realistic case where overheads are present. Our results indicates that it is possible to deliver good quality video to a large number of peers with a small number of pieces per block and relative small block sizes, while fully utilizing the upload bandwidth capacity of peers.