基于Hadoop视频转码缓存策略的研究  

Caching strategies based on video transcoding using Hadoop

作　　者：毕莎莎[1] 陈清华[1,2,3] 

机构地区：[1]南京工业大学电子与信息工程学院,江苏南京210009 [2]浙江清华长三角研究院,浙江嘉兴314006 [3]北京大学软件与微电子学院,北京102600

出　　处：《计算机工程与设计》2015年第3期683-686,692,共5页Computer Engineering and Design

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)

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。In real-time cloud transcoding system,the reused rate of transcoded resources is not high.To address this problem,the caching mechanism was introduced and a caching strategy was proposed.Videos were cached by cloud storage space and utility function value was calculated on the basis of popularities of video,request interval,video size,the probability to be accessed in the future,etc.Videos were replaced in accordance with much less utility value when cache space was not enough.Simulation results demonstrate that compared with the traditional algorithms,for example,LRU,LFU and FIFO,this strategy increased cache hit rate by 2.6%,4.3% and 9.2% respectively.It can not only shortens the delivery latency,but also reduces the response time to some certain extent.

关 键 词：实时转码 缓存机制 替换策略 云存储 效能函数 命中率 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]