本文提出一种用于分布式视频点播服务系统的速率分配算法和实现框架。在考虑负载均衡与满足服务器带宽限制的前提下,速率分配能将以失真为度量的全局视频质量最优化。确定各服务结点的输出码率被建模成一个最优化问题:优化目标是最小化多个视频总失真,限制条件为最低可接受视频质量、服务器带宽限制以及发送速率对客户端TCP流量的友好性。根据优化算法的计算结果,视频节目使用可变码率编码,由多个视频分发服务器同时向客户端传输;同时服务器检测与客户端之间端到端链路的时变特性和估计网络拥塞情况,使优化速率分配结果适应网络状态的变化。本方案应用跨层设计的思想,将应用层视频质量建模为网络层参数的函数,当网络状态波动时,动态调整编码码率和发送速率以适应带宽变化,提前避免链路拥塞,降低接收端解码器丢包率,从而使总体视频质量得到提升。
An optimal rate allocation framework for Video on Demand (VoD) system with multiple distributed servers is proposed in the paper. The video quality in aspect of distortion rate is maximized with the consideration of server load balance and bandwidth limitation. The bit-rate allocation is considered as an optimization problem subject to the constraints of video quality, link capacity and TCP-friendliness. In the proposed VoD framework, the video packets encoded with variable bit-rate are efficiently distributed to multiple servers for load balancing and timevarying characteristics of end-to-end link to clients are adapted to eliminate the network congestion. The encode rate of each video and streaming rate on each server is dynamically adjusted according to the link state and video characteristics so that the total video distortion in all clients can be minimized. Compared with existing schemes, our scheme applies the cross-layer design and expresses the application layer's video quality as the function of network layer's parameters. The result of the performance evaluation shows that the optimal rate allocation framework can benefits from proactively lowering the bandwidth cost while network degradation is detected, hereby avoid network congestion and promote the system capacity as well as the overall video quality.