基于PEPA模型的内容分发网络的性能评估

发布时间：2018-11-02 07:03

【摘要】：随着信息技术的快速发展,互联网已成为了人类信息社会主要基础设施之一,但经过半个多世纪的发展,已经演变成了复杂的巨系统,随着用户规模的快速增长,资源并发访问显著,使得有限的网络资源与日益增长的用户需求之间的矛盾日益突出。同时各种新型应用和智能终端的出现,种类繁多的业务形态访问异质异构资源时,服务质量要求差异较大,传统的内容传递的方法难以满足要求,这些使得内容分发面临重大的挑战。内容分发网络(CDN)的思想产生于1998年,该技术通过内容的复制和用户就近访问的方式,缓解困扰着互联网内容提供商的瓶颈难题。经过多年的实践化探索,CDN技术逐渐受到广泛关注并得到了快速的发展,其提供的分发服务已成为互联网中的一项重要的组成部分,它以可扩展方式提供网络设施和运行机制,可以高效地实现内容和服务的分发,增强了用户的网络体验,有相关研究表明,通过对CDN的体系结构进行改进,性能进行优化,能够有效的应对内容分发所面临的一些挑战。首先,本文采用了一种适用于对系统进行建模分析的随机进程代数方法——性能评估进程代数(PEPA),它是一种专门对系统的相关性能进行评估的方法,本文通过该方法对内容分发网络和云化的内容分发网络中的各组件分别建立了相应的PEPA模型,用该方法建立的模型,可以更好地描述在内容分发网络中的各个组件之间的数据传输过程,这样就可以对该网络模型中的工作流程加以分析。本文在模拟在内容分发网络和云化的内容分发网络中用户访问的响应时间的同时,也给出了模拟响应时间的特殊算法,该特殊算法的核心思想是在给出的动作流的前提下,通过设置标记来表示完成的动作,使得标记的动作与动作流一一对应,直到一个完整的动作路径结束为止;同时本文也采用了流体逼近的方法,通过该方法,本文给出了内容分发网络和云化的内容分发网络中的相关组件的常微分方程(ODEs),通过这些方程可以计算出网络中的相关组件的吞吐量和利用率,通过对响应时间、吞吐量、利用率等相关性能量化地分析评估,可以对今后内容分发网络框架模型进行相应地改进。其次,本文对在内容分发网络及云化的内容分发网络中的模拟响应时间特殊算法进行改进,由于这些特殊算法有自身的局限性,所以就提出了模拟响应时间的一般算法,该一般算法可以在给出任意的动作流的情况下,模拟出相应的响应时间,在本文中就通过一个模型案例来用该算法模拟其响应时间。最后,本文在PEPA模型中验证了 Little定理,该验证的方法是根据Little定理在封闭的系统中的相关证明知识和PEPA的相关知识相结合的方法,所以今后系统在采用性能评估进程代数(PEPA)这个方法来建立模型时,可以根据该定理的理论依据来计算系统中的平均响应时间。
[Abstract]:With the rapid development of information technology, the Internet has become one of the main infrastructure of human information society, but after more than half a century of development, has evolved into a complex giant system, with the rapid growth of the scale of users, Resource concurrent access makes the contradiction between limited network resources and increasing user needs more and more prominent. At the same time, with the emergence of various new applications and intelligent terminals, when a variety of service forms access heterogeneous and heterogeneous resources, the quality of service requirements are quite different, and the traditional methods of content delivery are difficult to meet the requirements. This makes the distribution of content facing major challenges. The idea of content distribution network (CDN) originated in 1998. The technology alleviates the bottleneck problem of Internet content providers through the way of content replication and user access. After many years of practical exploration, CDN technology has been paid more and more attention and developed rapidly. The distribution service it provides has become an important part of the Internet. It provides network facilities and operation mechanisms in an extensible manner. It can realize the distribution of content and services efficiently and enhance the network experience of users. Some related studies show that by improving the architecture of CDN and optimizing its performance, it can effectively deal with some challenges faced by content distribution. First of all, this paper adopts a stochastic process algebra method, (PEPA), which is suitable for modeling and analyzing the system. It is a special method to evaluate the relativity of the system. In this paper, the corresponding PEPA models are established for each component of the content distribution network and the cloud content distribution network by this method. The process of data transmission between the components in the content distribution network can be better described, so that the workflow in the network model can be analyzed. This paper not only simulates the response time of user access in the content distribution network and the cloud content distribution network, but also gives a special algorithm for simulating the response time. The core idea of the special algorithm is based on the given action flow. By setting the mark to represent the completed action, the marked action corresponds to the action stream one by one until the end of a complete action path; At the same time, the method of fluid approximation is also used in this paper. Through this method, the ordinary differential equation (ODEs), of the components in the content distribution network and the cloud content distribution network is given. Through these equations, we can calculate the throughput and utilization ratio of the related components in the network, and analyze and evaluate the response time, throughput, utilization ratio and other related performance quantificationally. The framework model of content distribution network can be improved accordingly. Secondly, this paper improves the special algorithms of simulated response time in content distribution network and cloud content distribution network. Because these special algorithms have their own limitations, a general algorithm of simulating response time is proposed. The general algorithm can simulate the corresponding response time under the condition of giving any action flow. In this paper, a model case is used to simulate the response time of the algorithm. Finally, this paper verifies the Little theorem in the PEPA model. The method of verification is the combination of the relevant proof knowledge of the Little theorem in the closed system and the relevant knowledge of PEPA. So in the future, the system can calculate the average response time of the system according to the theoretical basis of the theorem when using the method of performance evaluation process algebra (PEPA) to build the model.
【学位授予单位】：扬州大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.0

【相似文献】