信息中心网络协作式缓存管理方法研究

发布时间：2018-10-24 17:03

【摘要】：随着计算机网络规模的扩张和用户对内容需求的不断增长,在网络中传递的内容对象数量日益增多。传统的“以主机为中心”的网络体系在安全性、移动性和内容传递效率上存在的缺陷日益突出,已经无法适应不断增长的信息访问需求。信息中心网络(Information-Centric Network,ICN)这种以信息为中心的网络架构应运而生。网络内嵌缓存是信息中心网络的一个重要优势,高效地管理网络缓存对于网络性能的提升有着重要的意义。信息中心网络的缓存管理分为协作式和非协作式,协作式的缓存管理需要网络中节点之间的高度配合。软件定义网络(Softeware Defined Network,SDN)作为重要的下一代网络架构,通过控制平面和数据平面的分离,实现了对网络的集中管控。将软件定义网络融入到信息中心网络的设计中,可以从全网的角度实现缓存管理。本文介绍了软件定义信息中心网络的实现方案,并在此基础上设计了协作式的缓存管理架构。通常,缓存管理需要将网络中内容流行度信息作为关键的输入之一,因而高性能的内容流行度统计方法对于缓存管理非常重要。本文提出了PopFilter内容流行度统计方法,实现了对内容请求的线速处理,降低了内容请求频次统计过程的复杂度,并将统计结果的解码处理工作交给运算能力强大的控制器,合理分配了交换机和控制器的运算能力。此外,与传统的内容流行度统计方法相比,该方法拥有更高的准确性。在本文的研究中,基于全网的协作式缓存管理以最小化网络中内容请求的代价为目标。首先在内容路由固定的前提下,本文引入了内容缓存本地收益的概念,其定义综合考虑了内容请求代价和内容的流行度。本文提出了基于内容缓存本地收益的缓存管理方法,仿真结果表明该方法可以在很大程度上减少网络中内容请求的代价。然后在此基础之上,本文提出一种更加灵活的缓存管理方法,即在内容路由可变的情况下,综合考虑内容路由和内容缓存的联合优化问题,并基于遗传算法来求解该问题,仿真表明内容路由和缓存的联合优化在减少网络中内容请求代价的问题上能取得更好的解决效果。
[Abstract]:With the expansion of computer network scale and the increasing demand for content, the number of content objects transferred in the network is increasing day by day. The traditional "host-centered" network system has more and more shortcomings in security, mobility and content delivery efficiency, which can not meet the increasing information access needs. Information Center Network (Information-Centric Network,ICN), an information-centric network architecture, emerged as the times require. Network inlay cache is an important advantage of information center network. It is very important to manage the network cache efficiently for the improvement of network performance. The cache management of information center network can be divided into cooperative and non-cooperative. The cooperative cache management needs a high degree of cooperation between nodes in the network. As an important next generation network architecture, the software defined network (Softeware Defined Network,SDN) realizes centralized control of the network by separating the control plane from the data plane. By integrating the software definition network into the design of the information center network, the cache management can be realized from the point of view of the whole network. This paper introduces the implementation of software defined information center network, and designs a cooperative cache management architecture. In general, cache management needs to take the content popularity information as one of the key inputs, so high performance content popularity statistics method is very important for cache management. In this paper, the method of PopFilter content popularity statistics is proposed, which realizes the line speed processing of content request, reduces the complexity of the process of content request frequency statistics, and hands over the decoding work of the statistical result to a controller with powerful operation ability. The operation power of switch and controller is allocated reasonably. In addition, compared with the traditional method of content popularity statistics, this method has higher accuracy. In this paper, the goal of collaborative cache management based on whole network is to minimize the cost of content request in the network. Firstly, under the premise of fixed content routing, this paper introduces the concept of local income of content cache, whose definition takes into account the cost of content request and the popularity of content. In this paper, a cache management method based on local revenue of content cache is proposed. The simulation results show that the method can greatly reduce the cost of content request in the network. Then, this paper proposes a more flexible cache management method, that is, considering the joint optimization of content routing and content cache in the case of variable content routing, and solving the problem based on genetic algorithm. Simulation results show that the joint optimization of content routing and caching can achieve a better solution to the problem of reducing the cost of content request in the network.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.0

【参考文献】