MRI: Acquisition of a HPC System for Data-Driven Discovery in Science and Engineering

建议#：12-29081PI(S)：Somani，Arun K；Aluru，Srinivas；Fox，Rodney O；Gordon，Mark S；Takle，Eugene S研究所：爱荷华州立大学标题：磁共振成像：收购用于科学和工程领域数据驱动发现的HPC系统建议的项目：该项目购买一台HPC仪器(具有大容量存储和快速Infiniband网络的HPC集群)，旨在支持来自8个部门的17个项目，涉及广泛的计算学科，包括生物科学、生态学、流体动力学、地球和大气科学、材料科学和能源系统。包括图形处理器和对大规模存储单元的重视构成了拟议仪器的主要新奇之处。拟议的研究涉及并行体系结构的算法开发和计算建模的组合，同时在生物、材料、能源和气候科学方面追求引人注目的应用，即：-生物科学。将开发生物信息学工具，重点研究下一代测序仪的纠错算法、重新测序、基因组组装、全基因组关联、生物网络干扰分析和代谢组学。-解决重大挑战问题的多尺度方法。可以解决的方法？巨大的挑战？例如，大气气溶胶形成的模拟和新材料的设计。将使用粗粒化，从计算昂贵的高水平量子力学方法开始，将高水平势映射到新的势要简单得多。-计算流体动力学建模。新的HPC平台将使流体力学和多相流的尖端研究成为可能。具有流体和表面反应的多相流的粒子分辨直接数值模拟将是此类模拟的第一次。算法的发展将在喷雾、泡状流和气固两相流的设备规模模拟中有广泛的应用，这些应用程序使用基于求积的矩方法来处理固相。土地利用变化和区域气候极端的耦合动力学。长期目标是整合政策和气候预测模型，以捕捉政策驱动的农业土地利用变化和区域气候之间的动态耦合，包括新的气候和区域农业预测系统和模拟。更广泛的影响：区域和国家都应感受到影响。在国家一级，该工具应启动拟议的计算算法的变革性进展，并将以开放源码的形式提供给更广泛的研究界。由这些算法实现的模拟将产生广泛的国家和社会影响，从气候变化情景到风力发电，再到植物生物技术和改进的动物育种。在区域层面，拟议中的高性能计算集群将极大地丰富S研究所的研究基础设施。该仪器的使用将被纳入高级课程，并将分配时间培训本科生、研究生和博士后研究员计算建模和算法开发。HPC集群将为主要是本科院校提供时间，再加上积极的招聘计划，应该有助于吸引女性、代表性不足的少数族裔和第一代大学生，否则他们可能不会被鼓励进入该机构。

Proposal #: 12-29081PI(s): Somani, Arun K; Aluru, Srinivas; Fox, Rodney O; Gordon, Mark S; Takle, Eugene SInstitution: Iowa State UniversityTitle: MRI: Acquisition of a HPC System for Data-Driven Discovery in Science and EngineeringProject Proposed:This project, acquire an HPC instrument (HPC cluster with large storage and a fast Infiniband network), aims to support 17 projects from 8 departments in a broad range of computational disciplines, including bioscience, ecology, fluid dynamics, earth and atmospheric science, materials science, and energy systems. The inclusion of GPUs and emphasis on large-scale memory units constitutes the key novelty of the proposed instrument. The proposed research involves a mix of algorithm development for parallel architectures and computational modeling, while pursuing compelling applications in biological, material, energy and climate sciences, i.e.:- Biosciences. Bioinformatics tools will be developed to focus on research such as error-correcting algorithms for next-gen sequencers, resequencing, genome assembly, genome-wide association, biological network interference analysis, and metabolomics.- Multiscale methods for grand challenge problems. Methods that can address ?grand challenge? problems, such as simulation of atmospheric aerosol formation and design of new materials. Coarse-graining will be used, starting with high-level quantum mechanics methods that are computationally expensive and mapping the high-level potential onto a new potential is much simpler.- Computational fluid dynamics modeling. The new HPC platform will enable cutting edge research in fluid mechanics and multiphase flows. Particle-resolved direct numerical simulations of multiphase flow with fluid and surface reactions will be first-of-its-kind simulations. Algorithmic developments will have broad applications in sprays, bubbly flows and device-scale simulations of gas-solid flow applications that employ quadrature-based moment methods to treat the solid phase.- Coupled dynamics of land use change and regional climate extremes. The long-term goal is to integrate policy and climate projection models to capture dynamic coupling between policy-driven agricultural land use change and regional climate, including the novel climate and regional agricultural projection systems and simulations. Broader Impacts: The impact should be felt both regionally and nationally. At the national level, the instrument should initiate transformative advances in computational algorithms proposed and will be made available to the broader research community in the form of open-source codes. Simulations made possible by these algorithms will have broad national and societal impact ranging from climate change scenarios to wind power generation to plant biotechnology and improved animal breeding. At the regional level the proposed HPC cluster will greatly enrich the institution?s research infrastructure. Use of the instrument will be incorporated into advanced courses and time will be allocated to train undergraduates, graduate students, and postdoctoral fellows in computational modeling and algorithm development. The HPC cluster will make time available to primarily undergraduate institutions and, coupled with active recruitment plans, should help attract women, underrepresented minorities, and first generation college students, who might otherwise not be encouraged to attend the institution.