MRI: Acquisition of a High Performance Computing System for Online Simulation

MRI：获取用于在线仿真的高性能计算系统

• 批准号: 0619838
• 负责人: Omar Ghattas
• 金额: $ 80万
• 依托单位: University of Texas at Austin
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2006
• 资助国家: 美国
• 起止时间: 2006-09-01 至 2009-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0619838&HistoricalAwards=false
• 关键词: MRI; Acquisition; Performance; Computing; System

This project, supporting research in online simulation for critical problems in many areas, aims at acquiring a high-performance computing system. Online simulation represents a new paradigm in the use of high-performance computing, comprising a tight and continuous interaction between data and simulation. The data may stream from instruments or repositories, or may stem from user interaction with, and steering of, an on-going computation. Often, the simulations must execute in real-time, that is, in time scales that enable simulation-based decision making. Online simulation demands a high-availability, high-throughput computational facility, -Dedicated to simulations with real-time and interactive needs;-Networked at gigabits speeds with advanced visualization facilities on the user end, and instruments and repositories on the data end; and-Sufficiently powerful to enable fast turn-around in support of high-fidelity simulation-based decision-making.Local servers typically support online simulation, as national supercomputer facilities are not configured to support the first two. The facility will be housed in ICES (Institute for Computational Engineering and Sciences), already established in modeling, large-scale simulation, and visualization of complex systems, and with an existing collaboration with the University of Texas-El Paso, a minority-serving university. Multiple real-time simulation efforts are planned, including-Storm surge modeling; -Online visualization of simulations of neuromuscular junctions (only real in the sense visualization is on line); -Instrumented oil fields; Image-driven surgery; -Simulation of multi-scale manufacturing processes (only real in that it needs to deal with multi-scale issues); -Genomic modeling; Real-time evacuation; and Cardiovascular simulation.

该项目支持许多领域关键问题的在线仿真研究，旨在获得高性能的计算系统。在线仿真代表了高性能计算使用的新范式，它包括数据和仿真之间紧密和连续的交互。数据可能来自仪器或存储库，也可能来自用户与正在进行的计算的交互和操作。通常，模拟必须实时执行，也就是说，在支持基于模拟的决策制定的时间尺度上。在线仿真需要高可用性、高吞吐量的计算设施，-致力于具有实时和交互需求的仿真；-以千兆速度联网，用户端有先进的可视化设施，数据端有仪器和存储库；并且足够强大，可以实现快速周转，支持高保真的基于仿真的决策。本地服务器通常支持在线模拟，因为国家超级计算机设施没有配置为支持前两种模拟。该设施将被安置在ICES（计算工程与科学研究所）中，该研究所已经建立了复杂系统的建模、大规模仿真和可视化，并与德克萨斯大学埃尔帕索分校（一所少数民族大学）合作。多个实时模拟工作正在计划中，包括：风暴潮建模；-神经肌肉连接模拟的在线可视化（仅在在线可视化的意义上是真实的）；-仪器化油田；Image-driven手术;-多尺度制造过程的仿真（只有在需要处理多尺度问题时才是真实的）；基因组建模;实时疏散;心血管模拟。