A research on modeling feedback-based data scheduling in P2P streaming system

Abstract

In order to evaluate the influence of neighbors' quality to P2P streaming system's performance during scheduling, this paper establishes a markov chain based differential function model to analyze the relationship between peer's average startup delay and probability neighbors have wanted data as well as neighbors' available upload bandwidth. This paper also proposes a scheduling feedback algorithm, which learns real-time information of neighbors' data ownership as well as service capacity during data transferring and feeds back to overlay to help its optimization. The algorithm could help overlay discover possible better Qos neighbors when scheduling. Numerical analysis and experiment shows better quality neighbors can improve system's average startup delay significantly during scheduling.