云服务系统中组件服务副本的关键问题研究

发布时间：2018-01-30 22:58

  本文关键词： 云服务系统 组件服务副本 吞吐量约束 拓扑匹配 灰色-马尔科夫预测 CloudSim　出处：《东北大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着云计算技术的日益成熟,云服务系统已经成为了一种重要的软件开发模式。在云服务系统应用中,组件服务被部署在不同的服务器或者服务器集群中。由于用户访问行为的不确定性,当用户访问量增加时,云服务系统可以动态增加组件服务副本,提高系统的吞吐量进而保证云服务系统性能。从而,如何放置及选择合适的组件服务副本以保证云服务系统的性能已经成为了目前研究的一个热点。本文首先分析了云服务系统中组件服务副本相关技术,针对于其中的组件服务副本数量估计、组件服务副本放置以及组件服务副本选择等三个关键问题展开了研究工作。针对于组件服务副本数量估计问题,提出了一个基于吞吐量约束的组件服务副本数量估计算法,该算法根据所建立的组件服务吞吐量聚合规则,面向云服务系统吞吐量约束保证,计算每个组件服务的吞吐量约束,进而得到每个组件服务副本的数量。针对组件服务副本放置问题,提出了一个基于图拓扑匹配的组件服务副本放置算法,该算法使用聚类技术获得组件服务和计算节点的拓扑结构,并通过匹配两种拓扑结构来进行组件服务副本放置。该算法可以有效的降低云应用的执行周期和延迟时间。针对组件服务副本选择问题,通过结合负载模型和灰色-马尔可夫预测技术提出了基于灰色-马尔可夫预测的组件服务副本选择算法。在进行组件服务副本选择时,对计算节点的负载进行预测,并从中选择负载最小的计算节点来进行组件服务副本选择。通过使用该算法,可以有效的提高组件服务副本的利用率。基于CloudSim云仿真软件搭建了一个仿真实验环境并开展了一系列实验,仿真实验结果表明了所提出的方案和算法的有效性。
[Abstract]:With the development of cloud computing technology, cloud service system has become an important software development model. Component services are deployed in different servers or server clusters. Because of the uncertainty of user access behavior, cloud service systems can dynamically increase replicas of component services when user traffic increases. Improve the throughput of the system and thus ensure the performance of the cloud service system. How to place and select the appropriate component service replica to ensure the performance of cloud service system has become a hot topic. Firstly, this paper analyzes the component service replica related technology in cloud service system. This paper focuses on the estimation of the number of component service replicas, the placement of component service replicas and the selection of component service replicas, and aims at estimating the number of component service replicas. An algorithm for estimating the number of component service replicas based on throughput constraints is proposed. According to the established aggregation rules of component service throughput, the algorithm is oriented to the throughput constraint assurance of cloud service systems. The throughput constraints of each component service are calculated, and then the number of replicas of each component service is obtained. A component service replica placement algorithm based on graph topology matching is proposed to address the problem of component service replica placement. The algorithm uses clustering technology to obtain the topology of component services and compute nodes. The algorithm can effectively reduce the execution period and delay time of the cloud application, aiming at the problem of component service replica selection. By combining the load model and the grey-Markov prediction technology, this paper proposes a component service replica selection algorithm based on the gray Markov prediction, which is used to select the component service replicas. The load of the computing node is predicted and the computing node with the smallest load is selected to select the copy of the component service. It can effectively improve the utilization rate of component service replicas. A simulation environment based on CloudSim cloud simulation software is built and a series of experiments are carried out. Simulation results show the effectiveness of the proposed scheme and algorithm.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.09
，

本文编号：1477465