信息中心网络缓存资源管理与应用研究

发布时间：2018-05-01 10:10

本文选题：信息中心网络 + 无线网络　；参考：《北京邮电大学》2017年博士论文

【摘要】：信息中心网络(Information-Centric Networking,ICN)改变了现有IP (Internet Protocol,网际协议)网络的格局,针对目前互联网承载内容的飞速发展、用户访问网络的主要行为之一演变成对海量内容的获取这种现象,ICN以请求和获取信息为基本行为模式,而非传统的端到端连接,从而有效的进行内容分发。秉承骨干网络简洁化、边缘网络智能化的原则,骨干网的设计采用分布式ICN网络,局域网的设计采用集中式网络模式。其中,分布式ICN体系架构参考了当前IP网络的沙漏模型,以内容块为细腰,取代传统的IP细腰,既可以方便全球互联,又支持网络层以外各个层次的繁荣发展。它包含2种数据分组:兴趣分组(interest packet)和数据分组(data packet)。其路由器节点还具备三个数据结构:待定兴趣表(pending interest table,PIT )、转发信息表(forwarding information base, FIB )和内容缓存库(content store, CS)。该设计可以实现请求的聚合以及缓存的可重用。本文主要针对固网ICN缓存冗余、固网ICN缓存资源效用、如何将ICN缓存技术的优势应用到无线网络并解决其资源分配等问题展开研究,研究与实现ICN原型系统的开源代码ccnx和仿真平台NS-3。具体研究内容和主要创新工作包括以下几点:(1)提出了基于缓存状态相关性的ICN缓存决定策略。考虑到ICN缓存的目的是为了有效的减少流量冗余,然而在网络中的不同位置缓存内容会导致缓存冗余,因此放置内容采用怎样的缓存策略能够优化缓存资源的分配成为本文的一个研究重点。缓存策略包括缓存替换和缓存决定两部分。本文从节点主动选择是否缓存到达的数据包着手,通过分析路径缓存能力、不同位置节点的替换频率、节点间是否缓存内容的相互影响等因素,提出了一种基于缓存状态相关性的ICN缓存决定策略。通过以上几个因素的综合考虑,得到节点缓存数据包的概率,根据此概率来分配网络中的缓存资源,从而有效的消除缓存冗余、提高网络中的内容多样性、提高缓存节点的平均利用率、并极大地降低了内容服务器的负载。优化缓存资源分配的同时,还通过节点之间的协作保证路径节点具有一定的缓存规律。(2)提出了基于内容差异化的ICN缓存资源自动分配和管理机制。ICN最初的设想是利用网内缓存资源来换取带宽资源。然而现有研究中缓存的内容是不加区分的,没有考虑到差异化(用户对不同类型内容的请求率不同,如视频类)的因素,并且实际部署当中ICN的缓存大小有限。为了更好的满足当前海量内容的需求且有效的利用有限的缓存资源,本文首先将ICN中的用户流量按照一定的优先级分成若干类,并利用层次分析法将这些类内容定量的给出不同的缓存权重,进一步利用控制论的理论来建立节点自管控模型,基于上述的分类方法和管控模型,节点将为不同类的内容分配不同大小的缓存空间,差异化的为不同类的内容提供不同的服务质量(如缓存命中,请求时延小等)。通过控制环路的每次调整,来实时改变不同类内容的缓存空间,以期达到预设想要的性能。引入控制环路这一想法,不仅可以维持每个节点缓存空间的恒定,还能够保证性能的稳定性,使得网络运营商可以帮助内容运营商在骨干网中管理和控制每个缓存节点内存储的内容,为内容运营商提供不同级别的内容/服务加速。(3)提出了融合ICN缓存资源的无线网络架构及资源分配模型。目前关于ICN缓存技术在无线网络中的应用、边缘计算与无线网络的结合、软件定义网络在无线网络中的应用等方面的研究日渐增加,然而对于用户而言,通信、缓存、计算设备都是底层资源,用户并不关注资源是如何布局,只关注用户体验。考虑到不同种类的资源能够为用户解决不同的应用需求,并且结合当前无线网络的不同异构无线接入方式彼此之间存在互补的特性。本文为了给用户提供更好的体验并且完善多元化网络的能力,除了传统的通信能力,一方面增加ICN的通信设备缓存能力,另一方面增加边缘通信设备计算能力。通过软件定义的方式来管理和控制各式各样的异构无线模式,以及多样性的底层资源,进而程控不同接入方式下的不同资源分配,既保证了用户侧的时效性,又达到的绿色节能的效果。进一步,在该融合了 ICN缓存资源的无线网络架构模型下,针对当前的网络情况和用户需要的不同底层资源情况,为用户选择不同的接入方式,保证网络资源分配均匀、节省能耗、用户对应用时效性等方面用户体验的需求。
[Abstract]:Information-Centric Networking (ICN) has changed the pattern of existing IP (Internet Protocol, Internet Protocol) network. In view of the rapid development of the content of the Internet, one of the main behaviors of the user's access to the network becomes the phenomenon of obtaining the mass content, and ICN is the basic behavior model for the request and acquisition of information. And not the traditional end to end connection, so as to effectively carry on the content distribution. Adhering to the principle of the backbone network succinct and the intelligent edge network, the design of backbone network adopts distributed ICN network, the design of the LAN adopts centralized network mode. The distributed ICN architecture refers to the hourglass model of the current IP network and the inner volume block. For fine waist, instead of the traditional IP waist, it can facilitate the global interconnection and support the prosperity and development of various levels outside the network layer. It contains 2 data groups: interest packet and data packet. The router node also has three data structures: the undetermined interest table (pending interest table, PIT), forwarding letter. Forwarding information base (FIB) and content cache Library (content store, CS). This design can achieve the aggregation of requests and the reusable caching. This paper mainly focuses on the caching redundancy of fixed network ICN, the utility of the fixed network ICN caching resource, how to apply the advantage of the ICN caching technology to the wireless network and solve the problem of resource allocation. The research, the research and implementation of the open source code ccnx and the simulation platform of the ICN prototype system, the specific research content and the main innovation work include the following points: (1) a ICN cache decision strategy based on caching state relevance is proposed. The purpose of the ICN cache is to effectively reduce traffic redundancy, however, in different locations in the network Caching content can lead to caching redundancy, so how to optimize the allocation of caching resources with what kind of caching strategy is used is a focus of this paper. Caching strategy includes two parts: cache replacement and caching decision. A ICN cache decision strategy based on caching state relevance is proposed for the substitution frequency of the same location nodes and the mutual influence of caching content among nodes. Through the comprehensive consideration of the above factors, the probability of the node caching packets is obtained, and the cache resources in the network are allocated according to this probability, thus effectively eliminating the caching resources in the network. Caching redundancy can improve the content diversity in the network, improve the average utilization of caching nodes, and greatly reduce the load of the content server. At the same time, the allocation of caching resources is optimized, and the path nodes are guaranteed to have certain caching rules through the collaboration between nodes. (2) an automatic ICN cache resource based on content differentiation is proposed. The initial idea of the allocation and management mechanism.ICN is to use the network cache resources to exchange bandwidth resources. However, the content of the existing research is not differentiated, and does not consider the difference (the user's request rate for different types of content, such as the video class), and the actual deployment of the ICN is limited in size. In order to meet the needs of the current mass content and effectively utilize the limited cache resources, this paper first divides the user traffic in ICN into several classes according to a certain priority, and uses the analytic hierarchy process to give the different caching weights of these contents quantificationally, and further uses the theory of control theory to establish the node self control model. In the above classification method and control model, nodes will allocate different sizes of cache space for different classes of content, and provide different quality of service for different classes of content (such as cache hit, request delay, etc.). By controlling each adjustment of the loop, the cache space of different classes of content can be changed to achieve the presupposition. The idea of introducing the control loop can not only maintain the constant cache space for each node, but also ensure the stability of the performance, so that the network operators can help the content operators manage and control the contents stored in each cache node in the backbone network, and provide different levels of content / service for the content operators. (3) (3) the wireless network architecture and resource allocation model are proposed. The application of ICN caching technology in wireless network, the combination of edge computing and wireless network, and the application of software defined network in wireless network are increasing. However, for users, communication, caching and computing are used. Devices are all underlying resources. Users do not pay attention to how resources are laid out and only focus on user experience. Considering different kinds of resources can solve different application requirements for users, and there are complementary characteristics to each other in the current wireless network with different heterogeneous wireless access methods. This article provides a better experience for users. And the ability to improve the diversity of the network, in addition to the traditional communication ability, on the one hand increase the ICN communication equipment caching capability, on the other hand increase the computing power of the edge communication equipment. Through the software definition to manage and control a variety of heterogeneous wireless modes, and the diversity of the underlying resources, and then program control different access parties. The distribution of different resources under the model not only ensures the efficiency of the user side, but also achieves the effect of green energy saving. Further, under the wireless network architecture model of the ICN caching resource, we can choose different access methods for users to ensure the distribution of network resources for the users according to the current network situation and the different underlying resource conditions required by the users. Uniform, saving energy consumption, user needs for user experience in timeliness and application.

【学位授予单位】：北京邮电大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TP393.0

【参考文献】