An integrated Scheme to Distribute Segmented Streaming Media over Hierarchical Caches
An integrated Scheme to Distribute Segmented Streaming Media over Hierarchical Caches
复制标题
分层缓存上分段流媒体分布的综合方案
DOI：
--
复制
发表时间：
2002
期刊：
影响因子：
--
通讯作者：
Y. Yasuda
中科院分区：
文献类型：
--
作者：
Zhou Su;J. Katto;Takayuki Nishikawa;M. Murakami;T. Washizawa;Y. Yasuda
因此，在本文中，我们提出了一种用于流媒体的集成传输和缓存系统，其中每个流被分为多个段，并且这些段分布在分层分布的缓存服务器之间。摘要：互联网作为一种普遍的通信媒介的出现导致了流媒体的许多应用的兴起。然而，由于其独特的统计特性和用户观看模式，HTML 文件或图像等 Web 对象的传统传输和缓存方案无法有效地应用于音频和视频等流媒体。因此，在本文中，我们提出了一种具有基于分段的缓存和分层分布式代理的流媒体集成缓存方案。首先，将每个流划分为段，并考虑其缓存算法来确定如何有效地将段分配到不同级别的代理中。然后，通过引入两种段优先级，提出了段替换算法，以确定当缓存已满时应该替换哪些流和哪些段。最后，提出了一种网络友好的缓存方案，将流式缓存与普通网络对象的传统缓存集成。通过仿真验证了所提出算法的性能。首先，由于将整个流存储在单个代理缓存中效率低下甚至不可能，因此提出并比较了不同的基于段的缓存算法。所请求的流的一部分被缓存在本地缓存中，而流的其余部分将被缓存在上层代理缓存中。其次，按照惯例，当执行经典替换算法时，将相同的流行度分配给整个流，但是，每个不同的流具有不同的流行度，并且每个段具有不同的访问模式。因此，本文为每个段引入了两个优先级：一个反映其访问属性，另一个表示其在流中的位置信息。比较两种替换算法，以决定当缓存超出其限制时应删除哪些流的哪些段。两种算法之一在缓存中保留每个流的相同相对长度，而另一种算法在缓存中保留最常访问的段。
In this paper, we therefore propose an integrated delivery and caching system for streaming media, where each stream is divided into segments and these segments are distributed among hierarchically distributed cache servers. Abstract—The emergence of the Internet as a pervasive communication medium has led to the rise of many applications of streaming media. However, because of their distinct statistical properties and user viewing patterns, traditional delivery and caching schemes for web objects such as HTML files or images can not be efficiently applied to streaming media such as audio and video. In this paper, we therefore propose an integrated caching scheme for streaming media with segment-based caching and with hierarchically distributed proxies. Firstly, each stream is divided into segments and their caching algorithms are considered to determine how to distribute the segments into different level proxies efficiently. Then, by introducing two kinds of segment priorities, segments replacing algorithms are proposed to determine which stream and which segments should be replaced when the cache is full. Finally, a web friendly caching scheme is proposed to integrate the streaming caching with the conventional caching of normal web objects. Performance of the proposed algorithms is verified by carrying out simulations. Firstly, because storing the entire stream in a single proxy cache is inefficient or even impossible due to its large size, different segment-based caching algorithms are proposed and compared. A part of the requested stream is cached in a local cache, and the remainder of the stream will be cached in an upper proxy cache. Secondly, by convention, the same popularity was assigned to the whole stream when the classical replacing algorithm was carried out, however, each different stream has a different popularity and each segment has different access patterns. Therefore, in this paper, two priorities for each segment are introduced: one reflects its access property, and the other represents its position information in the stream. Two kinds of replacing algorithms are compared to decide which segments of which streams should be removed when the cache exceeds its limit. One of the two algorithms keeps the same relative length of each stream in the cache, while the other keeps the most accessed segments in the cache.