Visual Characterization of I/O System Behavior for High-End Computing

高端计算 I/O 系统行为的视觉表征

• 批准号: 0938114
• 负责人: Kwan-Liu Ma
• 金额: $ 47.06万
• 依托单位: University of California-Davis
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2009
• 资助国家: 美国
• 起止时间: 2009-09-15 至 2014-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0938114&HistoricalAwards=false
• 关键词: Visual; Characterization; System; Behavior; End

Modern supercomputers are complex, hierarchical systemsconsisting of huge numbers of cores, systems for disk storage,and nodes for I/O forwarding. These numbers continue to grow andthe need for tools to understand the behavior of the systemsoftware becomes paramount: without these tools it is impossibleto effectively tune that software, and high degrees of efficiencyis unattainable by applications. This project addresses thechallenge of understanding the behavior of complex systemsoftware on very large-scale compute platforms, like the currentpetascale computers. In particular, this project is developingsoftware infrastructure to provide end-to-end analysis andvisualization of I/O system software. Specifically, theobjectives are to develop, improve, and deploy (1) end-to-end,scalable tracing integrated into the I/O system (MPI-IO, I/Oforwarding, and file system); (2) information visualization toolsfor inspecting traces and extracting knowledge; (3) testingcomponents that drive this system to generate example patterns,including a component to generate anomalies; and (4) tutorialsand tools for helping other system software developersincorporate this analysis and visualization system into theirproduction software. The software and techniques developed inthis project will be directly applicable to and useful in othersystem software libraries which perform complex interactions onlarge systems.For further information see the project web site at the URL:http://vis.cs.ucdavis.edu/NSF/Jupiter

现代超级计算机是复杂的分层系统，由大量的核心、磁盘存储系统和I/O转发节点组成。 这些数字继续增长，对工具的需求，以了解系统软件的行为变得至关重要：没有这些工具，它是不可能有效地调优软件，高效率是无法实现的应用程序。这个项目解决了理解复杂系统软件在超大规模计算平台上的行为的挑战，比如当前的千万亿次计算机。特别是，该项目正在开发软件基础设施，以提供I/O系统软件的端到端分析和可视化。具体来说，目标是开发、改进和部署（1）集成到I/O系统中的端到端、可扩展的跟踪（2）信息可视化工具，用于检查跟踪和提取知识;（3）测试驱动该系统生成示例模式的组件，包括生成异常的组件;以及（4）用于帮助其他系统软件开发人员将该分析和可视化系统纳入其生产软件的教程和工具。本项目开发的软件和技术将直接应用于在大型系统上执行复杂交互的其他系统软件库。有关详细信息，请参阅项目网站：http://vis.cs.ucdavis.edu/NSF/Jupiter