云平台可维护性研究

发布时间：2018-06-26 14:09

本文选题：云平台 + 可维护性　；参考：《浙江大学》2015年硕士论文

【摘要】：因云计算的兴盛与普及,越来越多的企业选择引入云计算技术改造原有的企业数据中心,一些企业则开始租赁大厂商的云服务。学术界和业界研究人员投入了相当多的精力,专注于提高云平台的安全性和资源利用率,但对于云平台的可维护性,却很少有人涉及。本文通过搭建DartCloud云平台的实践和思考,研究了云平台的可维护性,主要包括：云平台的快速部署及监控系统的搭建、镜像管理策略的定制和对业务扩展性的支持。本文工作得到了863项目“互联网语言翻译系统研制”的资助和支持。本文工作和贡献如下：●对云平台的可维护性做了全面阐述,并在云平台的快速部署及监控系统的搭建、镜像管理策略的定制和对业务扩展性的支持这三方面做了深入的探讨。●分析了OpenNebula和Xen的实现原理,给出了从物理裸机到一个云平台的搭建指南,针对现有部署效率低下的问题,提出并实现了批量和自动化部署策略,加速平台搭建和部署,提高了云平台扩展性和部署效率。●实现了基于Ganglia的云资源监控管理系统,提供各虚拟机、各物理服务器的CPU利用率等指标,为性能调优提供基本依据。●提出并实现了一个完整的镜像管理解决方案,其中主要包括分级镜像权限控制、镜像版本管理和基于增量复制策略的镜像自动化管理三部分。●通过实现的DartCloud资源监控系统采集实验数据,深入分析了云平台中scale-up和scale-out两种资源扩展方法在大数据并行计算应用上的性能差异,为综合提高资源利用率和减少成本的性能调优管理提供了依据。
[Abstract]:Because of the prosperity and popularity of cloud computing, more and more enterprises choose to introduce cloud computing technology to transform the original enterprise data center, and some enterprises begin to lease cloud services from large manufacturers. Academic and industry researchers have devoted considerable attention to improving the security and resource utilization of cloud platforms, but few people have been involved in the maintainability of cloud platforms. Through the practice and thinking of constructing DartCloud cloud platform, this paper studies the maintainability of cloud platform, including the rapid deployment of cloud platform and the construction of monitoring system, the customization of mirror image management strategy and the support of business expansibility. This work is supported by 863 project, Internet language Translation system Development. The work and contributions of this paper are as follows: the maintainability of cloud platform is expounded comprehensively, and the rapid deployment of cloud platform and the construction of monitoring system are introduced. The customization of mirror management policy and the support for business extensibility are discussed in depth. The implementation principle of OpenNebula and Xen is analyzed, and the construction guide from physical bare machine to a cloud platform is given. Aiming at the problem of low deployment efficiency, a batch and automatic deployment strategy is proposed and implemented, which accelerates platform construction and deployment, improves the scalability and deployment efficiency of cloud platform, and implements cloud resource monitoring and management system based on ganglia. The CPU utilization ratio of each virtual machine and physical server is provided to provide the basic basis for performance tuning. A complete mirror management solution is proposed and implemented, which mainly includes hierarchical mirror permission control. Image version management and image automation management based on incremental replication strategy. Through the implementation of DartCloud resource monitoring system to collect experimental data, In this paper, the performance differences between scale-up and scale-out in big data parallel computing are analyzed, which provides a basis for improving resource utilization and reducing cost.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP393.09

【相似文献】