基于OpenStack的业务云平台负载均衡策略的研究与实现

发布时间：2018-02-14 02:25

本文关键词： 云计算负载均衡负载预测任务调度　出处：《北京邮电大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着互联网的普及与发展,数据量不断增大,我们已经进入了数据量急剧膨胀的时代。云计算技术的出现大大缓解了数据的压力。云计算的一系列优势,如海量计算能力、廉价、按需使用等,为云计算带来了广阔的发展空间。但与此同时产生的大量的虚拟资源却变得难以管理和控制,用户对虚拟资源选择的不确定性非常容易造成资源节点负载失衡。针对这个问题,本文提出了一个具有负载均衡和动态扩展特性的资源调度框架。资源调度框架由4个组件组成：历史数据仓库、负载均衡器、扩展决策器和资源分配管理器。各个组件互相协作,最终在云环境下实现负载均衡、动态扩展的功能。其中,负载均衡器使用基于能力匹配的负载均衡策略,通过统计学趋势预测算法进行业务量预测,并以任务请求与虚拟机能力相匹配为原则将请求分配到合适的节点,从而实现负载均衡的特性。本文重点介绍了基于能力匹配的负载均衡策略的实现。该策略的实现需要负载预测、负载监控和任务调度三个组件。负载预测组件采用统计学中的趋势预测算法,根据历史数据对未来负载量进行预测,能够得到相对准确的预测结果；负载监控组件使用开源munin组件和collect组件实现对物理机和虚拟机资源的监控；对于任务调度组件,本文设计了能力匹配算法,将任务请求所需的计算资源与虚拟机的计算能力进行匹配,从而将任务请求分发到合适的虚拟机进行处理,充分利用计算资源,同时保证服务质量。最后,本文对该业务云平台资源调度框架进行了实验和测试,并将基于能力匹配的负载均衡策略与传统的负载均衡策略进行对比。实验结果表明,该框架能够在动态扩展的基础上实现负载均衡,基于能力匹配的负载均衡策略能够比传统的负载均衡策略更好地适应负载的动态变化,更合理地利用云中资源。
[Abstract]:With the popularization and development of the Internet, the amount of data is increasing, and we have entered the era of rapid expansion of data. The emergence of cloud computing technology has greatly alleviated the pressure of data, cloud computing has a series of advantages, such as the capacity of mass computing. Cheap, on-demand and so on, bring the cloud computing a broad space for development. But at the same time, a large number of virtual resources have become difficult to manage and control. The uncertainty of users' choice of virtual resources is very easy to cause resource node load imbalance. In order to solve this problem, a resource scheduling framework with load balancing and dynamic expansion is proposed in this paper. The resource scheduling framework consists of four components: historical data warehouse, load balancer, extended decision maker and resource allocation manager. The load balancer uses load balancing strategy based on capacity matching to predict traffic through statistical trend prediction algorithm and assigns the request to the appropriate node based on the matching of task request and virtual machine capability. Thus, the characteristic of load balancing is realized. This paper focuses on the implementation of load balancing strategy based on capability matching, which requires three components: load forecasting, load monitoring and task scheduling. According to the historical data to predict the future load, can get a relatively accurate prediction results; load monitoring components using open source munin components and collect components to monitor the physical machine and virtual machine resources; for task scheduling components, In this paper, a capacity matching algorithm is designed to match the computing resources required by the task request and the computing power of the virtual machine, so that the task request can be distributed to the appropriate virtual machine for processing, making full use of the computing resources and ensuring the quality of service at the same time. Finally, this paper tests and tests the resource scheduling framework of the service cloud platform, and compares the load balancing strategy based on capacity matching with the traditional load balancing strategy. The experimental results show that, This framework can realize load balancing on the basis of dynamic expansion. The load balancing strategy based on capacity matching can adapt to the dynamic change of load better than the traditional load balancing strategy and make more rational use of resources in the cloud.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.09

【参考文献】