MRI: Acquisition of a High Performance Computational Cluster for Yale University

提案号：CNS 08-21132PI(s): Girvin, Steve M. Basu, Sarbani；维克多·s·巴蒂斯塔；Ismail-Beigi,索拉博;机构：耶鲁大学纽黑文，CT 06520-8337标题：MRI/Acq。: Acq。项目建议：该项目为大规模计算系统获取仪器，使物理、生物和社会科学以及工程、语言学、统计学和许多其他领域的研究能够在高性能计算（HPC）中心的保护下进行。该仪器是生物医学高性能计算中心的补充。在信息技术服务（ITS）的支持下，这些中心被视为一个单一的虚拟单元。这项工作为物理科学、工程学、语言学、数学、生态学、经济学、心理学、统计学和生物科学等22个院系的52个研究项目提供服务。非常多样化的研究涵盖了广泛的主题，从宇宙学模拟到气候建模，到神经回路网络和内部动力学模拟，到经济建模，仅举几例。这项研究得到了超过35个现有项目和20多个未决项目的支持。研究计算科学家协助仪器，教授并行编程技术，优化当前代码，并编写新代码以支持单个项目。能够处理CPU和内存绑定任务的各种需求，以及大量存储和通信基础设施，很可能由一个由350个节点组成的大型CPU集群和一个由28个节点组成的大型内存集群来满足，每个节点8核共享16GB内存和共享7168 GB内存。这10倍多的cpu、增加的内存和大容量存储预计将被大量利用，并对研究和教育产生直接影响。更广泛的影响：该仪器利用了350多名教师、研究人员和博士后、研究生和本科生的个人研究。后者将能够培训和学习如何最佳地使用这种设施。将为本科生开发新的课程，重点是纳米尺度和生物系统的高性能计算。机构外展项目包括霍华德大学和泽维尔大学。这些大学的教员将通过暑期实习接受培训，并在HPC设施中分配时间。

Proposal #: CNS 08-21132PI(s): Girvin, Steve M. Basu, Sarbani; Batista, Victor S.; Ismail-Beigi, Sohrab; Smooke, Mitchell D.Institution: Yale University New Haven, CT 06520-8337Title: MRI/Acq.: Acq. of a High Performance Computational Cluster for Yale UniversityProject Proposed:This project, acquiring instrumentation for a large-scale computing system, enables research in the physical, biological, and social sciences, as well as engineering, linguistics, statistics, and many other fields, under the umbrella of a High Performance Computing (HPC) Center. The instrumentation complements the Biomedical HPC center. Supported by the Information Technology Services (ITS), these centers are considered a single virtual unit. This work services 52 research projects from faculty in 22 departments including physical sciences, engineering, linguistics, mathematics, ecology, economics, psychology, statistics, and the biological sciences. The very diverse research covers a wide range of topics from cosmological simulations to climate modeling, to neural circuit networks and simulation of internal dynamics, to economic modeling, to name a few. The research is supported by more than 35 current and more than 20 pending projects. Research computational scientists assist with the instrumentation, teach parallel programming techniques, optimize current code, and write new codes to support individual projects. The diverse needs capable of handling both CPU and memory bound tasks along with substantial mass storage and communication infrastructure will most likely be covered by a large CPU cluster consisting of 350 nodes with 16GB of memory shared by 8 cores per node and a large memory cluster consisting of 28 nodes sharing 7,168 GB memory. This 10-fold more CPUs, increased memory, and mass storage are expected to be heavily utilized and have immediate impact in research and education.Broader Impacts:The instrumentation leverages individual research of more than 350 faculty, research and postdoctoral staff, and graduate and undergraduate students. The latter will be able to train and learn how to optimally use such a facility. New courses will be developed for undergraduates focusing on HPC for nanoscale and biological systems. The institutional outreach program involves Howard and Xavier universities. Faculty from these universities will be trained via summer internships and allocated time in the HPC facilities.