MRI: Acquisition of a Shared Parallel High Performance Storage System to Enable Computational Science and Engineering

提案#：12-29573PI(S)：Hillegas，Curtis W.Carter，Emily A.；Pretorius，Frans；Spitkovsky，Anatoly；Wood，Eric F.机构：普林斯顿大学标题：MRI/Acq.：共享并行高性能存储系统，以实现计算科学和工程项目建议：该项目，获得高性能计算(HPC)存储系统，旨在提供所需的存储和I/O带宽，使研究能够推进到以前无法实现的新领域。具体地说，在天体物理学中，这种存储将允许增加维度，以便对激波形成进行全三维建模；在土木工程和环境工程中，它将能够分析时间和空间中更高分辨率的水状况数据；在机械和航空航天工程中，该仪器将有助于对生物燃料燃烧的速率常数进行严格的物理模拟，并对分子吸附有基本的了解；在物理学方面，它将能够进行引力波和致密物体合并的建模，考虑到更广泛的物理学范围。更广泛地说，该系统将使计算科学和工程中的许多学科的项目成为可能。由于数据存储和访问已成为阻碍TIGRESS系统研究人员的瓶颈，拟议的系统应该有助于消除瓶颈。考虑到数据将继续增长，我们选择了模块化存储系统设计，该设计将使系统能够随着数据海量的持续增加而在容量和性能上进行扩展。计划从NetApp购买基于硬件的1.5 PB存储系统，并由Comnetco将其与SGI的服务器集成。该系统将运行IBM通用并行文件系统(GPFS)，并在整个机构提供12 GB/S的并行性能-S虎妞HPC系统。广泛的影响：该仪器将通过使研究人员易于在机构内共享数据来促进协作；此外，它将通过内置的Web服务器设施通过允许研究人员广泛共享他们的数据来促进国内和国际合作。作为一种研究工具，博士后、研究生和高级本科生，包括许多研究人员，没有充分代表少数群体，该仪器将作为一个培训平台，教学数据布局，管理和性能优化。

Proposal #: 12-29573PI(s): Hillegas, Curtis W. Carter, Emily A.; Pretorius, Frans; Spitkovsky, Anatoly; Wood, Eric F.Institution: Princeton UniversityTitle: MRI/Acq.: Shared Parallel High Performance Storage System to Enable Computational Science and EngineeringProject Proposed:This project, acquiring a High Performance Computing (HPC) storage system, aims to provide storage and I/O bandwidth required to enable advancement of research to new, previously unattainable areas. Specifically, in astrophysics the storage will allow an increase in dimensionality to perform full 3D modeling of shock formation; in civil and environmental engineering, it will enable the analysis of higher resolution water condition data in both time and space; in mechanical and aerospace engineering, the instrument will facilitate rigorous physical modeling of rate constants for biofuel combustion and fundamental understanding of molecular adsorptions; and in physics it will enable the modeling of gravitational waves and compact object mergers considering a broader range of physics. More generally, the system will enable projects across many disciplines in computational science and engineering.Since data storage and access have become a bottleneck hampering researchers in the TIGRESS systems, the proposed system should contribute to remove the bottleneck. Understanding that data growth will continue, a modular storage system design has been chosen that will allow the system to grow in capacity and performance as the data deluge continuous to mount. Planned is the purchase of a 1.5 PB storage system based on hardware from NetApp, integrated with servers from SGI by Comnetco. The system will run IBMs General Parallel File System (GPFS) and provide 12 GB/s parallel performance across the institution?s TIGRESS HPC systems.Broader Impacts: The instrument will facilitate collaboration by making it easy for researchers to share data within the institution; furthermore, it will foster collaboration nationally and internationally through the included web server facility by allowing researchers to broadly share their data. As a research tool available to postdocs, graduate students, and advanced undergraduate students, including many researchers underrepresented minority groups, the instrument will serve as a training platform, teaching data layout, management and performance optimization.