MRI: Acquisition of a High Performance Central Computing Facility at University of California Santa Barbara

该项目购买了一台计算机集群（小型超级计算机）以取代旧计算机，旨在满足多个领域的科学研究和教育的计算需求。该机器包括120节点Infiniband互连集群，用于高效的消息传递接口（MPI）并行处理，四个共享内存“胖节点”，每个节点的内存为1 TB，四个图形处理单元（GPU）节点围绕NVIDIA Tesla P100 1 GB GPU构建，以及四个英特尔Knight's Landing节点。这种混合系统将满足绝大多数校园研究人员的计算需求。该系统还将为需要大规模资源的用户提供服务，允许在超级计算机中心生产运行之前进行本地开发，原型设计和基准测试计算。该研究跨越多个部门，并支持许多校园中心的研究，包括AlloSphere研究设施，加州纳米系统研究所，聚合物和有机固体中心，地球研究所，理论物理Kavli研究所，海洋科学研究所，材料研究实验室和国家生态分析和合成中心。这一跨学科的、主要是合作性的研究将通过当地设施的存在而得到促进，而不存在与应用于超级计算中心相关的行政障碍和延迟。更广泛的影响：该集群预计将在教育下一代科学家、工程师和数学家方面发挥突出作用。NSF网络科学和大数据的综合研究生教育和研究培训（IGERT）计划也将采用该集群，该集群还将通过现有的赞助计划由本科生，高中和社区大学学生以及K-12教师使用。它还将有助于吸引其他学生和研究人员。此外，该设施将通过高级学位（UCLEADS）和博士学位课程的桥梁为加州大学卓越领导力课程的毕业生提供服务。这些计划有助于增加参与人数不足的学生。

This project, acquiring a computer cluster (mini-supercomputer) to replace the old one, aims to serve the computational needs facilitating scientific research and education in multiple areas. The machine includes 120 node Infiniband interconnected cluster for efficient message passing interface (MPI) parallel processing, four shared memory "fat nodes" with 1 Terabyte (TB) of memory/node, four graphic processing unit (GPU) nodes built around NVIDIA Tesla P100 1 Gigabyte (GB) GPUs, and four Intel Knight's Landing nodes. This blended system will serve the computational needs of the vast majority of campus researchers. This system will also service users needing large-scale resources by allowing development, prototyping, and benchmarking calculations locally, prior to production runs at supercomputer centers. The research enabled spans multiple departments and supports research in many campus centers including AlloSphere Research Facility, the California Nanosystems Institute, the Center of Polymers and Organic Solids, the Earth Research Institute, the Kavli Institute of Theoretical Physics, the Marine Science Institute, the Materials Research Laboratory, and the National Center for Ecological Analysis and Synthesis. This interdisciplinary and mainly collaborative research will be facilitated by the presence of an available local facility, without the administrative hurdles and delays associated with application to supercomputing centers.Broader Impacts:The cluster is expected to play a prominent role in educating the next generation of scientists, engineers, and mathematicians. The NSF Integrated Graduate Education and Research Traineeship (IGERT) program in Network Science and Big Data will also employ the cluster that will additionally be utilized by undergraduates, high school, and community college students and K-12 teachers via existing sponsored programs. It will also contribute to attract others students and researchers. Furthermore the facility will service graduate participants in the University of California Leadership Excellence through Advanced Degrees (UCLEADS) and the Bridges to Doctorate programs. These programs help increase the number of under-represented students involved.