Issues on movie allocation in distributed video-on-demand systems

We study the allocation of movies in a distributed video-on-demand system that provides movies at will to users from a number of video servers scattered geographically. Each user is associated to one of the servers (the local server), but it can still watch transparently any movie located at any server. In our study, we take into consideration the user demand for the movies, the storage limitation of the video servers, and the communication cost experienced (to the service provider) for movies located remotely with respect to the users watching them. Our objective is to find how many copies of the movies should one make and where to place them so the communication cost is minimized.