Admission Control For Distributed Video On Demand System With Scalable Videos

Nowadays, Internet applications are becoming increasingly broad and diverse. Among them, online video service has became an integral part of the daily entertainment for people. In recent years, the rapid development of video terminal devices and network infrastructure challenges online video on demand systems. In order to adapt to varied terminals and heterogeneous network access environment, scalable video coding becomes a promising solution. Scalable video coding encodes video into several subset bit-streams called layers, different layers are given different transmission priorities and are delivered separately. In this way, scalable video coding provides variable quality of service for a single video, and well adapts to the requirements from different users and changeable network conditions.The current mainstream network video on demand systems adopt distributed architecture, which provides high stability and scalability. However, the limited resources still cause a performance bottleneck. Admission control can greatly improve the performance of a system under limited cost. It is an important mechanism to protect the user service quality and improve the utilization rate of the resources. However, the current admission control schemes are mainly oriented to single bit-rate videos and only make 0-1 decision when request arrives. Therefore, they cannot be applied to systems with scalable videos directly. Providing scalable videos in a distributed video on demand system brings many additional problems, such as the adjustment of quality of service according to varying terminal capabilities or network conditions, the ensurance of providing acceptable video quality while avoiding network congestion etc. This paper proposes a novel admission control scheme designed for distributed video on demand system with scalable videos and solves the optimal admission control strategy. The major achievement and contribution include the following aspects:1. Analyze the characteristics of distributed system architecture and scalable video, define user behavior and system response, and propose an admission control procedure for distributed video on demand system supporting scalable videos. When user requests arrive, a 0-n decision is made to decide whether to access, which node to serve, as well as the number of enhancement layers transmitted. When a user exits, take action at the current node to improve the quality of service online.2. Establish a partially observable Markov decision process for the admission control procedure proposed. The six(?)elements of the model are presented, through which the evolution of the system is described. In this paper, we adopt a parameterized randomized policy. And the optimization objective is the long term average performance. The policy gradient formula is given, and the parameters in the formula are solved by the method of sample trajectory simulation. The optimal policy can be iteratively solved.3. Design a simulation experiment to verify the effectiveness of the admission control scheme proposed. Set the iteration initial value to greedy strategy which is widely used in current video on demand systems. The result shows that our method improve the performance of the system.