面向云计算弹性扩展的规则自动生成与优化

发布时间：2018-03-15 03:17

本文选题：云计算　切入点：弹性　出处：《上海交通大学》2013年硕士论文　论文类型：学位论文

【摘要】：云计算的相关课题是近几年的研究热点，作为一种新型的计算范式，她倡导将计算能力以服务的形式加以传递。随着相关技术的进步和成熟，这种理念变得越来越实际，不少行业巨头都相继投入大量资本展开相关实践。弹性扩展能力是云计算的重要特性之一，它也是保证云服务能否成功实施的关键。正是因为它，，云平台能根据目标系统的运行特征动态为其调配资源，以确保系统的服务质量符合契约（SLA）规定，并保证资源利用率在较高的水平。实现云平台弹性扩展机制的重要步骤是为目标系统构建准确的资源需求模型，以便根据系统运行特征参数值估算其所需的资源数目，从而为其增添资源或从中释放回收资源。目前的研究工作主要采用实时在线的方法为目标系统微调相关配置参数或调整虚拟资源的配额，本文认为这些方法各自存在一定局限性，容易出现“抖动”现象。本文提出了一种基于神经网络相关知识体系来实现云平台弹性扩展机制的方法框架。该方法借助神经网络构建目标系统的资源需求模型，由此生成弹性扩展规则，并进一步指示云平台为目标系统执行相应的资源调配操作。同时，本文发现单一固定的弹性扩展规则，其作用效果很难得到持续的保证，本文为此设计了一套扩展规则的优化机制和方法。为验证本文所提方法框架的实效性，文章设计实现了一系列模拟实验，并通过对比静态扩展规则来分析生成的弹性扩展规则的作用效果。实验结果表明，本文提出的扩展规则生成方法行之有效，由此生成的弹性扩展规则能保证目标系统的服务质量（系统响应时间）在至少93%的时间里符合SLA规定，相比静态扩展规则只能保证系统服务质量在86%甚至更少的时间里符合要求。此外，文章设计的实验也证明了本文提出的弹性扩展规则优化方法的可用性。本文工作以IaaS的视角展开，并以多层架构的web系统作为目标系统的代表性案例，以此展开相关分析和研究。
[Abstract]:Cloud computing is a hot research topic in recent years. As a new computing paradigm, she advocates the transfer of computing power in the form of services. With the development and maturity of related technologies, this concept becomes more and more practical. Many industry giants have invested a lot of capital to carry out related practices. Flexibility expansion is one of the important features of cloud computing, and it is also the key to ensure the successful implementation of cloud services. In order to ensure that the quality of service of the system conforms to the stipulations of the contract SLAs, and to ensure that the utilization of resources is at a higher level, the important step to realize the elastic expansion mechanism of cloud platform is to build an accurate resource requirement model for the target system. In order to estimate the number of resources required by the system based on its operating characteristic parameters, The current research mainly uses real-time on-line method to fine-tune the related configuration parameters or adjust the quota of virtual resources for the target system. This paper holds that each of these methods has some limitations. The phenomenon of "jitter" is easy to occur. In this paper, a method framework based on neural network related knowledge system is proposed to realize the elastic expansion mechanism of cloud platform. This method uses neural network to construct the resource requirement model of the target system. The elastic expansion rules are generated, and the cloud platform is further instructed to perform the corresponding resource allocation operations for the target system. At the same time, it is found that the effect of the single fixed elastic expansion rule is difficult to ensure continuously. In this paper, a set of optimization mechanisms and methods for extended rules are designed. In order to verify the effectiveness of the proposed framework, a series of simulation experiments are designed and implemented, and the effects of the generated elastic expansion rules are analyzed by comparing the static expansion rules. The experimental results show that, The extended rule generation method proposed in this paper is effective, and the elastic extended rule can guarantee the quality of service (system response time) of the target system in at least 93% times in accordance with the SLA regulations. Compared with static extension rules, it can only guarantee the quality of service of the system in 86% or less time. In addition, the experiments designed in this paper also prove the availability of the flexible extension rule optimization method proposed in this paper. In this paper, the IaaS perspective and the multi-tier web system as the representative case of the target system, so as to carry out the relevant analysis and research.
【学位授予单位】：上海交通大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP183

【相似文献】