Content-Priority-Aware Chunk Scheduling Over Swarm-Based P2P Live Streaming System: From Theoretical Analysis to Practical Design

Abstract

In this work, we revisit two chunk, which is the smallest video data unit for scheduling and transmission, scheduling policies that are essentially mutually exclusive but beneficial for swarm-based P2P live streaming systems. The first is content-diversified oriented (cd-oriented) policy, which regards each chunk of equal importance and schedules chunks to be sent in a near-random fashion. With this approach, peers hold different parts of stream content and contribute their available bandwidth to the system. The second is importance-first oriented (if-oriented) policy, which gives each chunk a content-dependent priority, usually in a rate-distortion (RD) sense, and first schedules the highest-priority chunk to be sent. In doing so, important chunks are more likely to be successfully received before their playback; the reconstructed video quality is thus enhanced under poor network conditions. We successfully identify a simple methodology, which operates on the data availability domain, to leverage both policies. (Data availability here means the set of data units that the user can get from its source(s)). This allows us to deploy dynamic strategy switch scheduling in practical systems to further improve the received video quality of each peer. Simulation results show that our data-availability driven dynamic strategy switch not only overcomes the drawbacks of the two individual policies but also retains the benefits of both. Most importantly, it bridges the gap between rate-distortion analysis on compressed video and P2P content delivery research.