CC* Compute: A high-performance GPU cluster for accelerated research

计算能力的指数增长和高性能计算范例的出现已经彻底改变了科学和工程的所有领域。图形处理单元(GPU)硬件是一种高度并行的协处理器，最初设计用于在视频游戏中生成3D场景，在过去十年中越来越多地被利用来显著加速科学计算工作负载。该项目是为了收购由24个最先进的NVIDIA Tesla V100 32 GB GPU组成的GPU计算集群，该GPU具有快速的GPU间通信。该资源位于加州大学圣巴巴拉分校(UCSB)，研究人员可通过连接到太平洋研究平台/Nautilus联邦系统网络从校园和外部访问该资源。该设施的初步研究活动跨越计算领域，包括：用于预测复杂聚合物溶液配方的结构和热力学性质的新型多尺度分子模拟；涉及晶体取向索引和实时仪器反馈控制的材料表征推力；以及开发可扩展的神经体系结构搜索框架，用于自动生成机器学习的科学应用的深度神经网络模型。该集群为培训下一代计算科学家掌握最新的GPU计算技术提供了重要资源。本科生、高中生和K-12教师也将通过现有的校园赞助项目获得机会：教师研究经验(RET)、加州少数群体参与联盟(CAMP)和科学与工程伙伴关系中心(CSEP)。这些计划旨在提供培训并增加STEM领域中代表性不足的学生数量。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

The exponential growth of computing power and the emergence of high-performance computing paradigms has revolutionized all fields of science and engineering. Graphics processing unit (GPU) hardware, a type of highly parallel co-processor originally designed for generating 3D scenes in video games, has been increasingly leveraged over the last decade to dramatically accelerate scientific computing workloads. This project is for the acquisition of a GPU compute cluster consisting of 24 state-of-the-art NVIDIA Tesla V100 32 GB GPUs with fast inter-GPU communication. The resource is housed at the University of California, Santa Barbara (UCSB), and is accessible to researchers across campus and externally through a connection to the Pacific Research Platform/Nautilus federated systems network.Initial research activities on the facility span the computational realm, including: a new type of multi-scale molecular simulation for predicting structural and thermodynamic properties of complex polymeric solution formulations; a materials characterization thrust involving crystal orientation indexing with real-time instrument feedback control; and the development of a scalable Neural Architecture Search framework for automatic generation of Deep Neural Network models for scientific applications of machine learning. The cluster provides a significant resource for educating the next generation of computational scientists in the latest GPU-computing techniques. Undergraduates, high-school students, and K-12 teachers will also have access via existing campus-sponsored programs: Research Experience for Teachers (RET), California Alliance for Minority Participation (CAMP), and the Center for Science and Engineering Partnerships (CSEP). These programs serve to provide training and increase the number of under-represented students in STEM fields.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.