A fast, simple and near-optimal content placement scheme for a large-scale VoD system

Video-on-Demand(VoD) service providers operate their own networks with large-scale serving nodes, each of which serves users in a certain geographic area. Efficient content placement can dramatically reduce the network latency and improve the user utilization. Content placement is a difficult problem though, due to the explosively increasing video library, the limited disk capacity and high diversity of request patterns. In this paper, we formulate the problem as an integer program, with the simple and practical objective of maximizing the aggregate amount of hitting data. We prove that this problem is NPhard. Jointly considering the size and popularity distributions of the video clips, we propose a fast greedy algorithm to solve it. Numerical results show that our algorithm is near-optimal (within 1-3% of optimal) and more efficient than existing solutions. We also evaluate our assumed distributions of video clips through a case study of the Youtube video data.