Output details
11 - Computer Science and Informatics
Lancaster University
Characterising and exploiting workloads of highly interactive video-on-demand
<06> This paper presents the results of a 12-months longitudinal study of video-on-demand user behaviour, in the context of a large-scale interactive VoD testbed involving real users viewing popular sporting and musical content. As key contribution identifies user behaviour that can be exploited by content distributors to improve user experience and to reduce the load on the Content Delivery Network. The results challenge the general assumption, still commonly made in VoD circles, of a "start-to-finish" model. For example, we demonstrate the benefits of dynamic bookmark placement and interactivity-aware pre-fetching and replication in increasing user satisfaction while reducing network load.