CSR-VCM: Dynamic Reconfiguration in Virtualized Data Centers

CSR-VCM：虚拟化数据中心的动态重新配置

• 批准号: 0720271
• 负责人: Prashant Shenoy
• 金额: --
• 依托单位: University of Massachusetts Amherst
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2007
• 资助国家: 美国
• 起止时间: 2007-09-01 至 2010-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0720271&HistoricalAwards=false
• 关键词: CSR; VCM; Dynamic; Reconfiguration; Virtualized

Modern data centers employ a virtualized architecture where application components run inside virtual machines and one or more virtual machines are mapped onto each physical server. This project focuses on the design of a virtualized data center that can dynamically remap physical to virtual resources and employ such dynamic reconfiguration capabilities to automate management tasks such as server consolidation, dynamic capacity provisioning, and system evolution. There are three areas of focus: low-overhead non-intrusive monitoring in a virtualized data center, agile reconfiguration for consolidation and provisioning, and the use of virtualized replicas for managed evolution. Non-intrusive monitoring involves a novel black-box approach that infers resource usage solely from external observations. Agile system reconfiguration utilizes these black-box measurements to automate reconfiguration tasks such as capacity provisioning and server consolidation using a combination of virtual machine migration and agile replication. The final technique exploits the ability to run concurrent application replicas inside virtual machine to ease hurdles faced in the evolution of the system from one version to another. A combination of system prototyping and experimental evaluation on a Linux/Xen cluster testbed running real-world data center applications will be employed to demonstrate the effectiveness of these methods.Broader impacts include source code release of the prototype to enable further experimentation by other researchers and participation of undergraduate students in this research via summer REU projects.

现代数据中心采用虚拟化架构，其中应用程序组件在虚拟机内运行，并且一个或多个虚拟机被映射到每个物理服务器上。该项目的重点是设计一个虚拟化数据中心，它可以动态地将物理资源重新映射到虚拟资源，并采用这种动态重新配置功能来自动执行管理任务，如服务器整合、动态容量配置和系统演进。有三个重点领域：虚拟化数据中心中的低开销非侵入式监控，用于整合和配置的敏捷重新配置，以及使用虚拟化副本进行托管演进。 非侵入式监控涉及一种新的黑箱方法，仅从外部观察推断资源使用情况。 敏捷系统重新配置利用这些黑盒测量来自动执行重新配置任务，例如使用虚拟机迁移和敏捷复制的组合进行容量调配和服务器整合。 最后一种技术利用在虚拟机内运行并发应用程序副本的能力，以缓解系统从一个版本到另一个版本的演变中所面临的障碍。 系统原型和实验评估相结合的Linux/Xen集群测试平台上运行的真实世界的数据中心应用程序将被用来证明这些方法的有效性。更广泛的影响包括原型的源代码发布，使其他研究人员和本科生通过夏季REU项目参与这项研究的进一步实验。