Modeling the Content Popularity Evolution in Video-on-Demand Systems

The simulation and testing of Video-on-Demand (VoD) services require the generation of realistic content request patterns to emulate a virtual user base. The efficiency of these services depend on the popularity distribution of the video library, thus the traffic generators have to mimic the statistical properties of real life video requests. In this paper the connection among the content popularity descriptors of a generic VoD service is investigated. We provide an analytical model for the relationships among the most important popularity descriptors, such as the ordered long term popularity of the whole video library, the popularity evolutions and the initial popularity of the individual contents. Beyond the theoretical interest, our method provides a simple way of generating realistic request patterns for simulating or testing media servers.