Operations research theory of stochastic service at fast food restaurants, banks, video-on-demand, and web access should be extensive. This paper is a study of the design of its video-on-demand system model. We design and built a specific mathematical model, and analysis of algorithms and test results. We propose video stream control method based on popularity, and the popularity of the program is divided into two types of popular and unpopular, analysis of the system instantaneous transition probability, and designs the popularity of video streaming bandwidth prediction formula. The results show that this method can effectively save server resources, and improve the level of service of video. This queuing analysis method can also be applied in other similar random service system.