CC* Compute: GPU HPC Cluster Partition for Research, Education, and Student Success

CC* 计算：GPU HPC 集群分区促进研究、教育和学生的成功

• 批准号: 2201435
• 负责人: Stephen Wheat
• 金额: $ 39.94万
• 依托单位: Oral Roberts University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-05-15 至 2025-04-30
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2201435&HistoricalAwards=false
• 关键词: CC; Compute; GPU; HPC; Cluster

This award is funded in whole or in part under the American Rescue Plan Act of 2021 (Public Law 117-2).Oral Roberts University aims to deploy a GPU-based server partition called Eli, and integrate it with Titan, a campus wide resource. Eli will address the GPU and storage needs of computing for research, research training, and education. The proposed hardware is 6 GPU Compute Nodes with 2 Intel 32-Core Xeon Gold CPUs having 16 x 32G DRAM, 4 NVIDIA Tesla A30 GPUs, 1 SATA SSD boot drive (960GB) and 2 NVMe Solid State Drives (4TB each). It also includes a meta data Server, storage server, and NFS node.This project addresses two areas of improvement in the Oral Roberts University (ORU) Research Computing and Analytics (ORCA) facility, providing GPU resources and adding much needed parallel file system (PFS) resources. These new resources are deployed as a third cluster partition, named Eli, in ORCA’s existing Titan cluster. Along with Eli’s 100/200Gb HDR InfiniBand fabric, the Eli PFS supports native communication from each of Titan’s other partitions with 100GbE OmniPath and 40GbE QDR InfiniBand. Eli’s GPU nodes has potential to address a local and regional void of GPU capacity for traditional HPC codes (especially VASP and GAMESS) as well as emerging Data Science codes. VASP and GAMESS consume many core hours. The GPUs will accelerate these efforts while freeing previously used resources to other applications. With ORU’s Data Science degree concentrations, nearly half of the new ORU research efforts are Data Science related, driving demand for tools such as TensorFlow and Natural Language Processing tools. Eli’s PFS provides the much-needed capacity and performance to support the existing and emerging science drivers. While serving ORCA’s current users, Eli’s augmentation of Titan provides new researchers additional motivation to embrace HPC/AI methods. Titan/Eli is freely available for use by all Oklahoma academic institutions. The ORCA management team actively evangelizes HPC/AI to nascent researchers and provides opportunities for substantial undergraduate research experiences. Success is measured by system utilization, availability, and increased researcher participation. This project aligns with ORU’s NSF CC*-funded 100GbE connectivity to the OneOklahoma Friction Free Network (OFFN), a science DMZ for 19 Oklahoma academic institutions and is funded through the collaborative efforts of the Office of Advanced Cyberinfrastructure (OAC) and the Established Program to Stimulate Competitive Research (EPSCoR).This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

该奖项全部或部分由2021年美国救援计划法案（公法117-2）资助。Oral Roberts University旨在部署一个名为Eli的基于GPU的服务器分区，并将其与校园资源Titan集成。Eli将满足研究、研究培训和教育计算的GPU和存储需求。建议的硬件是6个GPU计算节点，配备2个英特尔32核至强金牌CPU（16 x 32 G DRAM）、4个NVIDIA Tesla A30 GPU、1个SATA SSD靴子驱动器（960 GB）和2个NVMe固态驱动器（每个4 TB）。它还包括一个Meta数据服务器，存储服务器和NFS节点。该项目解决了Oral Roberts大学（ORU）研究计算和分析（ORCA）设施的两个改进领域，提供GPU资源和添加急需的并行文件系统（PFS）资源。这些新资源作为第三个集群分区部署在ORCA现有的Titan集群中，名为Eli。 沿着Eli的100/200 Gb HDR InfiniBand结构，Eli PFS支持与Titan的每个其他分区进行本地通信，这些分区具有100 GbE OmniPath和40 GbE QDR InfiniBand。Eli的GPU节点有潜力解决传统HPC代码（特别是VASP和GAMESS）以及新兴数据科学代码的本地和区域GPU容量不足的问题。VASP和GAMESS消耗了大量的核心时间。GPU将加速这些工作，同时将以前使用的资源释放给其他应用程序。随着ORU的数据科学学位集中，近一半的新ORU研究工作与数据科学相关，推动了对TensorFlow和自然语言处理工具等工具的需求。Eli的PFS提供了急需的容量和性能，以支持现有和新兴的科学驱动因素。在为ORCA的当前用户提供服务的同时，Eli对Titan的增强为新的研究人员提供了接受HPC/AI方法的额外动力。 Titan/Eli可供所有俄克拉荷马州学术机构免费使用。 ORCA管理团队积极向新生研究人员宣传HPC/AI，并提供大量本科生研究经验的机会。成功是衡量系统利用率，可用性和增加研究人员的参与。该项目与ORU的NSF CC* 资助的100 GbE连接到OneOklahoma无摩擦网络（OFFN），为19个俄克拉荷马州学术机构提供科学DMZ，并通过高级网络基础设施办公室（OAC）和刺激竞争研究的既定计划（EPSCoR）的合作努力提供资金该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。