考虑擦除编码可靠视频流三步式内凸逼近优化

随着视频流服务的扩展和改进,云视频得到快速发展。考虑在分布式系统上的视频流,使用擦除码进行编码以提高可靠性。利用对擦除编码块的选择的有序统计量来对每个视频片段的编码块的下载时间进行表征,并获得不同视频片段的回放时间。利用失速持续时间上的矩生成函数的界限来界定平均失速持续时间。基于有序统计量的矩生成函数边界也用于界定失速持续时间尾概率,其确定失速时间大于预定义概率。针对最终用户的重要体验质量(QoE)度量,选取平均失速持续时间和失速持续时间尾概率,构造优化问题,以联合最小化在视频内容的放置和访问期间在所有请求上平均的两个QoE度量的凸组合,可利用有效的迭代算法求解非凸问题。数值结果表明,相比于选取的对比算法,基于云视频的QoE度量可显著性能提升。

THREE-STEP INNER CONVEX APPROXIMATION OPTIMIZATION FOR RELIABLE VIDEO STREAM WITH ERASURE CODING

With the expansion and improvement of video streaming services,cloud video has developed rapidly. This paper considered video stream in distributed systems,and used erasure codes to encode to improve reliability. The download time of each video clip was characterized by the ordered statistics of the selection of erased blocks,and the playback time of different video clips was obtained. The average stall duration was defined by the limit of moment generating function on the stall duration. The boundary of moment generating function based on ordered statistics was also used to define the tail probability of stall duration,which determined that the stall time was greater than the predefined probability. Aiming at the important quality of experience (QoE) metric of end users,the average stall duration and the tail probability of stall duration were selected to construct an optimization problem to jointly minimize the convex combination of two QoE metrics averaged on all requests during video content placement and access. We used efficient iterative algorithm to solve non-convex problems. The numerical results show that QoE metrics based on cloud video can significantly improve the performance compared with the selected comparison algorithm.