Improvement of data sharing efficacy of P2P streaming mobile networks for news-broadcast-on-demand services

Data sharing efficacy of P2P streaming mobile networks is sensitive to the content access paradigm of applications. This paper considers a news-broadcast-on-demand (NBOD) application, where news clips are categorized into several news channels according to their categories, and each user (subscriber) is allowed to set her/his own watching policy (such as favorite topics, recommended most, news sources, recency, etc.). The server compiles a playlist of news according to the user's watching policy and then automatically delivers the news clips on that playlist to the user. The key issue is how to orchestrate news-clip delivery schedules of different users so that the inter-peer sharing gain is maximized. We formulate a P2P-NBOD optimization problem and propose two heuristic algorithms for it, termed, the clip-based scheduling (CS) for the short-term optimization with lower computational cost and the session-based scheduling (SS) for long-term optimization with higher computational cost. Through packet-level simulation, we evaluate these two schemes with respect to cache buffer size and startup delay tolerance. The results reveal that employing CS or SS can improve data sharing efficacy about 5% or 10%, respectively, over the one without any scheduling. CS can achieve over 70% traffic offloading with small (9-sec) cache buffer. It seems to be a good choice to implement due to its much lower computation complexity.