TV-Anytime系统中动态自适应的影像对象映射

Dynamic Adaptive Video Objects Mapping in a Large-Scale TV-Anytime System

该文提出了两种不同的策略 ,来处理一个基于层次型服务器网络的大规模 TV- Anytim e系统中的动态影像映射问题 .基于预测的阶段性重映射策略是每隔一个时间周期决定服务器网络中的影像对象映射 ,以处理系统较长时间内的变化 .自适应的重分配策略则处理系统较短时间内的变化 ,该重分配策略基于不同影像对象的收益 /开销 ,其目标是尽大化服务器系统产生的收益 .模拟实验表明 。

Nowadays TV anytime has been widely applied in various fields. Because a number of high quality video stream requires large brand width, the technology of using a central server to provide video service to the large amount of users, is not feasible as well as being expensive. One of the viable solutions, hence, is to construct a distributed server network to support a large scale TV anytime system. Besides all the technical concerns involved in server network, the video mapping problem in server network is actually one of the key issues which determines the efficiency of the TV anytime system. This paper presents two various policies used for the effective mapping of videos in the runtime of a large scale TV Anytime system that is based on a hierarchical server network. A predictive assignment policy is presented to periodically determine the mapping of videos to the server network. The predictive assignment is intended to address long term variations in the system. It decides when, where and how often each video should be replaced. We employ a set of heuristic algorithms to solve this dynamic mapping problem with the aim to maximally utilize the resources in the system, in particular, to increase bps of each video requested by users. Furthermore, an adaptive reallocation policy is presented to handle the short term variations. This adaptive reallocation is based on revenue/cost of different videos. The objective is to maximize the overall revenue that the server system generates. We used a hierarchical architecture of server network assumed to be a 4 level 2 ary complete tree to test the performance. The simulation results demonstrate that these policies yield an efficient system and make the system more adaptive.