CSR-SMA: Measurement, Modeling and Analysis of Large Complex Data Centers

CSR-SMA：大型复杂数据中心的测量、建模和分析

• 批准号: 0720616
• 负责人: Prashant Shenoy
• 金额: $ 20万
• 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 2011-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0720616&HistoricalAwards=false
• 关键词: CSR; SMA; Measurement; Modeling; Analysis

The complexity of enterprise data centers has grown steadily in recentyears, primarily due to continuous evolution resulting from smallincremental changes to system hardware and software. This projectfocuses on a modeling and analysis approach for improving themanageability of complex data centers. The project focuses on (i)scalable, adaptive monitoring techniques for very large data centers,(ii) multi-scale modeling of complex data-center applications, and(iii) model-driven analysis to predict system performance, bottlenecksand capacity requirements. The adaptive monitoring approach involvesthe use of statistical data mining and learning to dynamically varywhat systems are monitored and when. The multi-scale models involvecapturing system behavior at large, medium and small scales bymodeling application interactions at different levels of abstraction:at the data center level, at the application level, and at theindividual server level. Model-driven analysis brings togetheradaptive monitoring and the multi-scale models for capacityprovisioning, bottleneck detection and performance prediction. Themodeling and analysis techniques will be validated using real-worlddata center applications running on a Linux-based data center testbed.Broader impacts include source code release and sharing of anonymizedtrace data to enable further experimentation by other researchers andthe participation of undergraduate students in this research viasummer REU projects.

近年来，企业数据中心的复杂性稳步增长，这主要是由于系统硬件和软件的微小增量变化导致的持续演进。 该项目重点关注用于提高复杂数据中心的可管理性的建模和分析方法。 该项目重点关注 (i) 超大型数据中心的可扩展、自适应监控技术，(ii) 复杂数据中心应用程序的多尺度建模，以及 (iii) 用于预测系统性能、瓶颈和容量需求的模型驱动分析。 自适应监控方法涉及使用统计数据挖掘和学习来动态改变监控的系统和时间。 多尺度模型涉及通过在不同抽象级别（数据中心级别、应用程序级别和单个服务器级别）对应用程序交互进行建模来捕获大、中、小规模的系统行为。模型驱动的分析将自适应监控和多尺度模型结合在一起，用于容量配置、瓶颈检测和性能预测。建模和分析技术将使用在基于 Linux 的数据中心测试台上运行的真实数据中心应用程序进行验证。更广泛的影响包括源代码发布和匿名跟踪数据共享，以便其他研究人员进行进一步实验，以及本科生通过夏季 REU 项目参与这项研究。