内容中心网络网内缓存策略研究

发布时间：2018-02-25 16:16

本文关键词： 内容中心网络网内缓存缓存决策策略缓存资源分配概率缓存请求影响度　出处：《北京邮电大学》2014年博士论文　论文类型：学位论文

【摘要】：随着流媒体等新业务在互联网流量中所占比重不断增加,以及云计算、物联网、社交网络等新型网络的出现,使得互联网应用从以面向主机的端到端通信模式为主转向以用户驱动的信息内容获取为主,而现有的覆盖网分发技术CDN、P2P在信息传送时,存在大量的重复内容传输,严重影响了传送效率。面对互联网在数据传输方面的可扩展性问题,以及在移动性、安全性等方面的诸多挑战。学术界提出了具有革命性的信息中心网络(ICN)体系结构,其中内容中心网络CCN则是一种具有里程碑式的ICN架构。CCN把内容当成最基本的实体,数据内容有一系列的内容块构成,每个内容块都有一个唯一的分级命名标识,并且每个内容块都由一个块级别兴趣包来请求。通过对内容统一标识进行定位、路由和传输。CCN从根本上改变了数据包的封装结构和寻址方式,从关注地址和主机到关注内容本身。 CCN架构的重要特征之一便是利用泛在化、透明化的网内缓存提高用户驱动的内容获取的传输效率和网络资源的利用率,网内缓存策略对于CCN的性能具有非常大的影响。然而,CCN缓存所具有的透明化、泛在化和细粒度化等新特征使传统的缓存理论、方法和模型均无法直接应用到CCN的缓存系统中,这对CCN缓存的相关研究均提出了新的挑战。本论文选择内容中心网络数据传输的可扩展性问题作为研究切入点,着重研究网内缓存最优决策策略和缓存资源分配这两个具体问题。这对于提高缓存系统性能,合理利用缓存和带宽资源具有重要的意义。本文的工作及贡献主要包括以下三点。 1.提出基于节点状态和内容需求度的网内缓存决策策略。首先,将网络节点状态作为主要衡量标准,提出基于节点介数和缓存替换率的缓存策略BetwRep。综合使用网络节点介数和节点缓存内容更替速率,作为缓存决策度量,通过权衡节点位置重要性和缓存内容时效性,实现回传内容的最佳放置。该策略既有效保证了内容尽量缓存在相对重要的节点上,又能通过节点的内容替换率来调控内容的缓存,使重要节点避免处于高频率的内容替换状态而导致系统性能下降。其次,综合考虑节点状态和内容需求度即节点内容的受欢迎程度,基于重要的节点尽量缓存流行度高的内容的思想,设计请求内容返回路径上节点的概率缓存策略ST-REQ。该策略通过计算内容请求命中时在路径上每个节点上被缓存的概率,内容返回时节点按各自的概率进行内容缓存。 2.提出基于灰色关联分析的多参数缓存决策策略。由于CCN中兴趣包请求具有聚合特性,相对重要的节点,可能收到的请求数较少,直接利用收到的请求数,无法衡量节点的重要性。针对这个问题本文定义了一个新型的网络状态度量参数——节点影响度RID,根据节点到达请求数、节点的位置以及请求的命中率来区分内容分发路径上节点的重要性。随后,综合考虑节点影响度、流量、兴趣包所经过的跳数和节点缓存更替率这四个网络系统参数,基于灰色关联分析决策法提出了多参数维度的缓存策略GRAD。缓存系统判决在路径上哪些节点需要缓存该返回内容时,既考虑节点的状态、重要性、拓扑信息,又要考虑网络中流量的分布。并且在GRAD决策策略中利用层次分析法来确定不同参数的权重。 3.提出分别基于请求影响度和能耗的节点缓存空间分配策略,并分析缓存判决策略对缓存资源分配的影响。为解决有限的缓存资源的分配问题,首先以本文中定义的节点影响度为比例系数,提出一种节点缓存资源的分配策略,相比节点缓存大小平均分配的方案,该策略降低了路由扩展度,并提高了网内命中率。随后,建立以总能耗为目标函数的缓存资源分配优化模型,该模型是一个整数线性规划问题。并利用遗传算法来求解,以获得(近似)最优的缓存决策策略,从而实现(近似)最优的缓存资源分配策略。最后,本文通过仿真分析了基于不同缓存判决策略的缓存资源分配方案对于缓存系统性能的影响。
[Abstract]:With streaming media and other new business accounts for increasing the proportion of Internet traffic, and cloud computing, networking, social networks and other new network, the Internet application from the host oriented end-to-end communication mode to the information content to user driven access, while the existing overlay network distribution technology CDN P2P in information transmission, transmission has a lot of topics, seriously affect the transmission efficiency.
In the face of the Internet data transmission scalability problems, and many challenges in mobility, security and other aspects. The academic circles put forward the revolutionary Information Center Network (ICN) architecture, which content center network CCN is a milepost type with ICN architecture.CCN content as the most basic entity. The data content of a series of content blocks, each block has a unique domain name logo, and each piece of content by a block level packet to request interest. Based on the content of unified logo positioning, routing and transmission of.CCN has changed the data package structure and addressing mode fundamentally. From the address and the host to focus on the content itself.
One of the important features of the CCN architecture is the use of ubiquitous network, using transparent cache to improve the user driven content to obtain the transmission efficiency and the rate of cyber source, network caching strategy has great effect on the performance of CCN. However, transparent CCN cache with the new features in the pan and fine-grained to cache the traditional theories, methods and models are not directly applied to the cache in the CCN system, the research of the CCN cache are presented new challenges.
The center content data transmission network scalability problem as the research starting point, focuses on the optimal decision strategy in network caching and distribution of the two specific problems to improve the cache cache resources. The performance of the system, is of great significance to the rational use of cache and bandwidth resources. The work and contributions of this paper include the following three point.
1. an intra network caching decision strategy based on node state and content requirement is proposed.
First of all, the network node status as the main measure, proposed replacement rate betweenness and cache replacement rate BetwRep. cache strategy integrated use of network nodes and node betweenness based on cache content, as caching decision metric, by weighing the importance of node location and storage content timeliness, the best placed to achieve return content. This strategy not only effectively to ensure the content as far as possible in the relative importance of node cache, and to regulate the content through the contents of a node cache replacement rate, to avoid an important node in the high frequency content and lead to system performance degradation to replace state.
Secondly, considering the node status and content requirements of the popularity of node content, important nodes to cache more popular content based on the idea of the design request probability caching strategy ST-REQ. content return nodes on the path of the strategy by calculating the content of each request hit on the path node cached content probability. Return the node cache content according to their probability.
2. a multi parameter caching strategy based on grey relational analysis is proposed.
The CCN request packet with interest polymerization characteristics, the relative importance of nodes, may receive the number of requests is less, the number of requests received directly by the importance of not measure nodes. This paper defines a new network node - state metric parameters influence degree RID, according to the node number of requests arrived, the hit rate of node the location and the request to distinguish the content distribution of nodes on the path of importance.
Then, considering the effect of node degree, flow through the interest packet hops and node cache replacement rate of the four system parameters, based on Grey Relational Analysis and decision method is proposed decision caching strategy GRAD. caching system of multi parameter dimension on the path which nodes to cache the content returned, not only consider the node state and the importance of topology information, but also consider the distribution of network traffic. And using the AHP to determine the weight of different parameters in the GRAD strategy.
3. a node cache spatial allocation strategy based on the request impact degree and energy consumption is proposed, and the impact of caching decision strategy on the allocation of caching resources is analyzed.
For the problem of distribution cache with limited resources, the impact of node in this paper firstly defined as proportional coefficient, proposed a resource allocation strategy of node cache, node cache size compared to the average allocation scheme, this strategy reduces the routing extension of network and improve the hit rate.
Then, a caching resource allocation optimization model with total energy consumption as the objective function is established. The model is an integer linear programming problem. Genetic algorithm is applied to obtain the (approximate) optimal cache decision strategy, so as to achieve (approximate) the optimal allocation strategy of caching resources.
Finally, the impact of caching resource allocation scheme based on different cache decision strategies on the performance of caching system is analyzed by simulation.

【学位授予单位】：北京邮电大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TP393.02

【参考文献】