MRI: Development of a System for Interactive Analysis and Visualization of Multi-Terabyte Datasets

MRI：开发用于多 TB 数据集交互式分析和可视化的系统

• 批准号: 0421423
• 负责人: Paul Woodward
• 金额: --
• 依托单位: University of Minnesota-Twin Cities
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2004
• 资助国家: 美国
• 起止时间: 2004-09-01 至 2007-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0421423&HistoricalAwards=false
• 关键词: MRI; Development; System; Interactive; Analysis

This project, creating a new cluster computer design targeted at interactive exploration of datasets tens of terabytes in scale, supports analysis and visualization of massive datasets. Research efforts concentrate mainly on genomics and bioinformatics, cosmology and geophysics, and computational fluid dynamic (turbulence and stellar convections). For the purpose of rendering, in near real-time, datasets whose render information requires up to 2TB of simultaneously active storage, a large system consisting of 80 dual processor nodes with a massive amount of disk space will be used for the visualization task. The visualization system will allow slightly more leeway for exploratory viewing of interactions among multiple variables at multiple time slices. Motivated by the need to interactively explore voluminous datasets, the system supports interactive analysis and visualization that follows any interactively specified and/or modified path through the 4 dimensions examining all the data along that path by streaming this data, potentially replicated at each of 80 network nodes, into the memories at a rate that matches the node's ability to perform significant analysis and/or visualization operations on the data and to display the results for the user at high resolution on viewing screens as large as the LCSE's 13 Mpixels Power Wall. Volume visualization and various kinds of pattern matching and cross correlation target this data. The design goals require significant research in scientific visualization, data management, data storage and cluster data sharing, as well as prority-based job scheduling.Broader Impacts: The development of this system involves many students and postdocs in the collaborating disciplines. The system will be used in instruction as well as in research leading to advance degrees. The computer science department at Fond du Lac Tribal and community College in Cloquet Minnesota will also benefit from this project due to the existing collaboration of the Laboratory for Computational Science and Engineering (LCSE) with this Native American serving college. LCSE has also shared its technology developments and software. Once again, the developed software will be made available over the Web in Open source format.

该项目创建了一种新的集群计算机设计，旨在对数十 TB 规模的数据集进行交互式探索，支持海量数据集的分析和可视化。研究工作主要集中在基因组学和生物信息学、宇宙学和地球物理学以及计算流体动力学（湍流和恒星对流）。为了近乎实时地渲染数据集，其渲染信息需要高达 2TB 的同时活动存储，由 80 个双处理器节点组成的大型系统以及大量磁盘空间将用于可视化任务。可视化系统将为探索性地查看多个时间片上多个变量之间的相互作用提供更多的余地。出于交互式探索大量数据集的需求，该系统支持交互式分析和可视化，可遵循任何交互式指定和/或修改的 4 个维度路径，通过流式传输这些数据来检查该路径上的所有数据，可能在 80 个网络中的每个网络进行复制节点以与节点对数据执行重要分析和/或可视化操作的能力相匹配的速率进入存储器，并在与 LCSE 13 一样大的屏幕上以高分辨率向用户显示结果Mpixels 电源墙。体积可视化以及各种模式匹配和互相关都以这些数据为目标。设计目标需要在科学可视化、数据管理、数据存储和集群数据共享以及基于优先级的作业调度方面进行大量研究。更广泛的影响：该系统的开发涉及合作学科的许多学生和博士后。该系统将用于教学以及获得高级学位的研究。由于计算科学与工程实验室 (LCSE) 与这所美国原住民服务学院的现有合作，明尼苏达州 Cloquet 的 Fond du Lac 部落和社区学院的计算机科学系也将从该项目中受益。 LCSE 还分享了其技术开发和软件。开发的软件将再次以开源格式通过网络提供。