The drift power law distribution of the number of requests in Video-on-Demand

The study of user access behavior in Video-on-Demand (VoD) system is principally from the perspective of videos and user sessions. From the user's perspective, the distribution of requests for VoD has not been investigated. In this paper, the times of single user requested in a period are defined as the number of individual requests. Firstly, the number of individual requests is sorted in descending order, which appeared heavy tailed. Secondly, we fitted two datasets with two mathematical models respectively, and compared them under different criteria. Lastly, the result indicates that the drift power law outperforms the stretched exponential on individual requests fitting. We draw the conclusion that clients in the VoD system could be separated into different priorities from the user's perspective, based on the drift power law distribution. Finally, this paper also gives some advice on optimizing of the VoD system.