基于Hadoop视频转码缓存策略的研究

发布时间：2018-07-29 15:15

【摘要】：针对实时云转码系统中已转码的媒体文件重新使用率不高的情况,引入转码缓存机制,提出一种缓存替换策略。采用云存储空间缓存视频,根据视频资源的流行度、请求间隔、视频大小和未来被访问概率计算视频的效能函数值,当缓存空间不足时替换出效能值较小的视频。实验结果表明,与传统置换算法LRU、LFU和FIFO相比,该策略的资源请求命中率分别提高了2.6%,4.3%和9.2%,在一定程度上缩短了视频分发延迟,减少了用户响应时间。
[Abstract]:In order to reduce the reutilization rate of transcoded media files in real time cloud transcoding system, a cache replacement strategy is proposed by introducing the mechanism of transcoding cache. Cloud storage space is used to cache video. According to the popularity of video resource, request interval, video size and future access probability, the video performance function value is calculated. When the cache space is insufficient, the video with small performance value is replaced. The experimental results show that compared with the traditional permutation algorithms LRULFU and FIFO, the hit rate of resource request of this strategy is increased by 2.6% and 9.2%, respectively, which shortens the video distribution delay and reduces the user response time to some extent.
【作者单位】：南京工业大学电子与信息工程学院;浙江清华长三角研究院;北京大学软件与微电子学院;
【基金】：江苏省基础研究计划基金项目(BK2010137) 浙江省十二五重大科技专项基金项目(2012C01037-1)
【分类号】：TN919.81;TP333

【参考文献】