Telecommunication companies believe that the video-on-demand market will expand heavily in the near future. The service providers are interested in the deployment of video-on-demand (VoD) systems with large numbers of videos. As the number of videos and customers increases at a VoD system, two problems are faced. The first problem is the manual video allocation of multiple copies of videos to the disks, called video placement. The second problem is where to forward a newly arrived request to play a video so that the dynamic load amongst disks or arrays of disks is balanced. In this paper, we first formulate the video placement as an INLP optimization problem and note that the video placement problem to distribute the number of copies of each video to disks or arrays of disks is NP-hard. We then propose greedy video placement and disk load balancing algorithms to minimize the static and dynamic loads of disks respectively. The simulation result shows that the proposed greedy video placement algorithm along with the disk load balancing algorithm evenly distributes the loads among the disks.
展开▼
