Design and implementation of distributed VOD system

Challenges of a video-on-demand technology

The development of Internet and WWW technology makes it easier for people to get information actively. More and more people are more willing to watch the program in time and actively, this trend is impacting the traditional one-way radio, the audience passively listening to the operating mode, forcing the radio and television system to the interactive direction of development, to achieve interactive VOD. But for the huge amount of audio and video data, its storage, transmission, a large number of concurrent access to the current Internet flow of text, image information is very different, these problems do not solve, will be difficult to achieve effective interactive on demand. In this paper, a distributed VOD system model is proposed based on the above problems.

Although many companies have developed video-on-demand software, such as Videocharger, Realserver, and so on, but most of them are park-level products, mainly for the users are enterprises, campus, etc., through the internal network of enterprises to provide users with services, can cover the region, The users of the service and the number of programs stored are relatively small. To meet the demands of a large number of users, companies such as Ncub and IBM ported their existing video-on-demand systems to supercomputing systems such as MPP or clusters. In general, however, these systems are organized in a centralized manner, with only one playback center in the system, and the video server in the playback center will directly serve the user's request when the user requests an on-demand program.

The traditional radio and television system covers a wide range, with a large number of users and programs, if a centralized video-on-demand system, the entire system will need a very high bandwidth backbone network. For example, the general MPEG-1 program needs the transmission rate of 1.5mb/s, if the system in parallel service 1000 users, then the backbone of the transmission rate of at least 1500mb/s, you can see that the construction of such a system is very expensive, and distributed video-on-demand system, is a more cost-effective system.

Two distributed VOD system

1 features of the program on demand

At present, the implementation and use of large-scale VOD system is relatively few, and it is impossible to statistic the exact feature of program Access. For VOD system, the mathematical statistical model of program on demand is mainly through DVD and video rental. The purchase situation obtained. In the process of statistical analysis, it is found that the program on demand is subject to the Zipf (zipf) rule, that is, the order of N movies is from large to small, and the access probability of Part I is =c/i,i=1,2,...,n, which 1/c=∑1/i,i=1,2,...,n. Figure 1 is the distribution of on-demand curves when the system has 100 movies.

Figure 1

Through this curve can be seen, program on demand has a lot of local, a large number of users to visit the former K-part program.

2 system Models

According to the feature of the program on demand, we propose a distributed VOD system with multiple playback centers to meet the needs of a wide range of users and programs, as shown in Figure 2.

Figure 2

The system is divided into two parts, the program center and the Broadcast center, the main function of the program center is to save all the programs in the system and provide data for the broadcast center. The playback center is composed of one or more video servers, the main function of which is to service the user on demand request, complete the program playback. There are some high on-demand programs in the video server within each zone, when a user on-demand program is responded to by a local playback center, if the user on-demand program is cached in the playback center, the playback center directly serves the user's request, otherwise the playback center will request the program center to provide the program that the user needs. According to the local characteristics of the program on demand, a large number of user requests will be completed by the playback center they belong to, which is very helpful to reduce the bandwidth demand of backbone networks.

The main idea of the model is to bring some program data with high on-demand rate to the user, compared with the centralized system, a large number of user requests will be completed locally, so that the program data in the transmission process does not need to pass the backbone network, thereby reducing the pressure of the backbone network, Its essence is through the relatively inexpensive video server system balance expensive high bandwidth backbone network construction, through this way can reduce the whole system cost effectively.