面向信息中心网络的协作缓存策略研究

发布时间：2018-07-09 22:43

本文选题：信息中心网络 + 网内缓存　；参考：《河南科技大学》2017年硕士论文

【摘要】：随着网络中数据流量的不断增加,传统的以主机为中心的网络体系架构正在面临着一系列的挑战。一方面,网络流量过载容易导致访问过程中出现较大的时延;另一方面,网络数据存在大量的冗余但能被重复利用的部分相当小。为适应海量、异质信息的访问,信息中心网络(ICN)采用以信息为中心的网络通信模式取代传统的以主机为中心的网络通信模式,其泛在化、透明化的网内缓存能够存储经过转发的内容,提高了用户获取内容的效率和网络资源的利用率。ICN中缓存策略的好坏直接影响着网络的传输性能,现有的缓存方案普遍存在着缓存冗余过多、缓存利用率低以及缓存内容差异率低等问题。为此,本文基于细粒度的缓存管理,提出了一种基于内容分块流行度和缓存收益的缓存策略(PGBCS),在一定程度上提升了网络的传输性能。主要研究内容及创新之处如下:1.基于细粒度的缓存管理,提出了一种通过比较节点缓存的综合收益寻找最佳放置节点的放置方法。首先,现有缓存策略的研究对象大多是内容文件,这无法满足ICN缓存细粒度化的特征,因此本文将内容对象细化至chunk级别,实现了细粒度的缓存要求;其次,基于内容分块研究其流行度,文中为每个内容分块设定了一个价值指标,流行度越大的内容分块其价值越大;最后,在寻找内容块的最佳放置节点时,综合度量影响缓存收益的几个因素,找出能够使整体缓存收益最大化的放置节点。2.提出了一种基于内容分块价值比较的缓存替换方法。节点缓存空间中内容块的价值随着用户的访问频率动态变化,为了使流行度高的内容分块有较高的缓存价值,文中设定内容块的价值随着被访问的次数呈指数增长,且随着时间的推移呈指数衰减。当内容块的价值低于设定的阈值时,将内容块的标识设置成“可替换”,并将其所占用的空间视为空闲。当缓存空间已满时,新到达的内容块通过价值比较优先替换节点中价值最低的内容块。文中首先对信息中心网络及网内缓存的研究进行了综述,并在此基础上,详细介绍了PGBCS缓存机制。为了验证PGBCS的性能优势,本文选取了四种具有代表性的缓存方案,分别是LCE,RCOne,ProbCache和Betw。通过缓存命中率、用户获取内容跳数、内容差异率等几个方面进行比较。实验表明,相比于上述四种缓存方案,PGBCS在引入较小的额外流量的情况下,能够较大的提高缓存性能。
[Abstract]:With the increasing of data traffic in the network, the traditional host-centric network architecture is facing a series of challenges. On the one hand, the overloading of network traffic easily leads to a large delay in the access process; on the other hand, there is a large amount of redundancy in the network data, but the part that can be reused is quite small. In order to adapt to the massive and heterogeneous information access, the information center network (ICN) adopts the information-centric network communication mode instead of the traditional host-centered network communication mode, which is ubiquitous. Transparent intra-network cache can store the forwarded content, which improves the efficiency of the user to obtain the content and the utilization of network resources. The cache policy in ICN directly affects the transmission performance of the network. The existing cache schemes generally have many problems, such as excessive cache redundancy, low cache utilization and low cache content difference rate. Therefore, based on fine-grained cache management, this paper proposes a cache policy (PGBCS) based on content partitioning popularity and cache income, which improves the transmission performance of the network to a certain extent. The main contents and innovations are as follows: 1. Based on fine-grained cache management, this paper proposes a method to find the best placement node by comparing the comprehensive benefits of node cache. First of all, most of the existing cache policy research objects are content files, which can not meet the characteristics of fine-grained chunk cache, so this paper refines the content object to chunk level to achieve fine-grained caching requirements. Based on the research on the popularity of content block, a value index is set for each content block. The greater the popularity of content block, the greater its value. Finally, when looking for the best placement node of content block, Synthetically measures several factors that affect the cache income, and finds out the placement node. 2. 2 which can maximize the overall cache revenue. A cache replacement method based on content block value comparison is proposed. The value of content block in node cache space changes dynamically with the user's access frequency. In order to make the content block with high popularity have higher cache value, the value of content block in this paper increases exponentially with the number of visits. And with the passage of time, the decay is exponential. When the value of a content block is below the set threshold, the identity of the content block is set to "replaceable" and the space it occupies is considered idle. When the cache space is full, the newly arrived content block replaces the lowest-value content block in the node by value priority. In this paper, firstly, the research of information center network and cache in network is reviewed, and on this basis, the mechanism of PGBCS cache is introduced in detail. In order to verify the performance advantages of PGBCS, this paper selects four representative caching schemes, namely, LCE-RCOne-ProbCache and Betw. Through cache hit rate, user get content hops, content difference rate and so on several aspects to carry on the comparison. The experimental results show that PGBCS can greatly improve the cache performance compared with the four cache schemes mentioned above.
【学位授予单位】：河南科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.0

【相似文献】