Exploring User-Based Scheduling Strategy for VoD with Frequent Seeks

Due to the dynamicity caused by users´ frequent VCR interactive operations, the problems of VoD service become complicated. Most P2P VoD solutions focused on service overlay optimization without considering segment popularity, correlation and user behaviors specially. In order to address that, we propose an novel P2P VoD framework called LUBStreaming, which focuses on exploring and learning user behaviors to improve VoD performance. With a carefully designed peer management overlay, peers are clustered according to their interests. Segment category and distribution scheme are utilized to optimize and reduce the latency of resuming playback after VCR operations. We also proposed a probabilistic pre-fetching algorithm which exploits user´s request-based associations efficiently. We studied the performance of LUBStreaming based on simulation and a prototype implementation. The results show that LUBStreaming improve quality of user experience significantly.