W.C.: Understanding Sub-stream Scheduling in P2P Hybrid Live Streaming Systems

Abstract—As an emerging and promising streaming architec-ture, the P2P pull-push hybrid architecture has achieved great success in delivering live video traffic over the Internet. However, a formal study on the sub-stream scheduling problem, a key design issue in hybrid systems, is still lacking. In this paper, we propose a max-flow model for mathematical analysis of this problem. We find that the sub-stream scheduling schemes used in existing hybrid systems, including CoolStreaming+, GridMedia and LStreaming, individually solve one special case of the pro-posed max-flow model. Moreover, this model can also serve as a benchmark to assess the performance of these existing sub-stream scheduling schemes. Further, we propose a weighted max-flow scheme to address the issue of peer heterogeneity in scheduling sub-streams. This weighted max-flow scheme is implemented in a prototype called LStreaming+. Extensive simulation results show that all the existing sub-stream scheduling schemes have achieved almost their best obtainable performance, while LStreaming+ can perform even much better. Finally, we point out the benefits of combining the hybrid streaming architecture and layered coding, and we also investigate how to schedule sub-streams in hybrid layered streaming systems. Experiments show that, since the hybrid system can maximize throughput, minimize useless download and reduce watching fluctuation at the same time, it outperforms existing pull-based layered systems significantly. I