Providing NPR-Style Time-Shifted Streaming in P2P Systems

Abstract

Abstract-Digital video recorder (DVR) style and nonprerecording (NPR) style are two possible implementations for P2P-based time-shifted streaming, but existing P2P streaming solutions are not suitable for implementing the NPR method. Since peers can view any arbitrary video segments which have been broadcasted, they might cause severe video quality problems and the server bandwidth consumption can become high. In this paper, we focus on minimizing the server bandwidth consumption to maintain smooth streaming service in NPR-style P2P-based time-shifted streaming. To reduce the server cost, peers prefetch segments which are not required for their current viewing. Hence even if they are viewing different parts of the video, they can exchange segments with one another. However, segment prefetching competes for bandwidth with ordinary segment fetching, and it might bring negative impact. A good prefetching solution should not affect a peer's viewing experience. We formulate the problem of finding a prefetching solution as an optimization problem, with the objective to minimize the server bandwidth consumption. Then we propose a heuristic algorithm by decomposing the global optimization problem into a set of smaller problems. Each peer runs this algorithm to determine which segments to prefetch and how to serve other peers. Simulation experiments demonstrate that our design provides P2P-based time-shifted streaming at low server bandwidth consumption.