Acquisition of an Advanced Workstation Cluster for High Performance Computing in Physics

采购用于物理高性能计算的先进工作站集群

• 批准号: 9977582
• 负责人: Stanley Hirschi
• 金额: $ 13.27万
• 依托单位: Central Michigan University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1999
• 资助国家: 美国
• 起止时间: 1999-09-01 至 2002-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9977582&HistoricalAwards=false
• 关键词: Acquisition; Advanced; Workstation; Cluster; Performance

9977582HirschiWith support from the Major Research Instrumentation program, a parallel array of high performance workstations will be acquired. These will be used to improve the scientific computational capability in the Department of Physics at Central Michigan University. Eight SMP workstations comprising sixteen DEC Alpha 21264 processors will be linked using a fast system area network (SAN). Parallelization on the array will be accomplished using MPI, the Message Passing Interface. This cluster will provide a theoretical peak performance of about 32 GFLOPS at an estimated cost of $6/MFLOPS. The cluster will meet critical hardware needs of three very active research groups in physics at Central Michigan University. These groups conduct computationally intensive research in materials and cluster physics (2 groups) and in nuclear structure physics (1 group). A polymer fluid dynamics group, which includes faculty from the departments of chemistry, physics, and mathematics, will also benefit from the use of this new capability. At present these groups use a variety of individual workstations, of which only one has a performance approaching 1 GFLOPS. Thus, the new array will increase the available peak computing power by at least a factor of 30. Such a performance enhancement will permit these groups at CMU to attack exciting new research problems involving larger and more complex systems. The planned parallel array of high performance workstations will also afford important new training opportunities for undergraduate and M.S. students at CMU in the area of parallel computing. A parallel array of high performance workstations will be acquired with support from the Major Research Instrumentation program. This array will provide increased computing power for both research projects and research training of M.S. level and undergraduate students, and thus will have a large impact on the infrastructure at CMU.

9977582Hirschi 在主要研究仪器计划的支持下，将获得高性能工作站的并行阵列。 这些将用于提高中央密歇根大学物理系的科学计算能力。 八个由 16 个 DEC Alpha 21264 处理器组成的 SMP 工作站将使用快速系统区域网络 (SAN) 连接起来。 阵列上的并行化将使用 MPI（消息传递接口）来完成。该集群将以 6 美元/MFLOPS 的估计成本提供约 32 GFLOPS 的理论峰值性能。 该集群将满足中央密歇根大学三个非常活跃的物理学研究小组的关键硬件需求。 这些小组在材料和团簇物理（2组）和核结构物理（1组）方面进行计算密集型研究。 由化学、物理和数学系的教员组成的聚合物流体动力学小组也将受益于这一新功能的使用。 目前，这些小组使用各种单独的工作站，其中只有一个的性能接近 1 GFLOPS。 因此，新阵列将使可用峰值计算能力至少提高 30 倍。这种性能增强将使 CMU 的这些团队能够解决涉及更大、更复杂系统的令人兴奋的新研究问题。计划中的高性能工作站并行阵列还将为本科生和硕士提供重要的新培训机会。 CMU 并行计算领域的学生。在主要研究仪器计划的支持下，将获得高性能工作站的并行阵列。 该阵列将为硕士学位的研究项目和研究培训提供增强的计算能力。水平和本科生，从而将对卡耐基梅隆大学的基础设施产生重大影响。