MRI: Proteus++: Enabling Data-Intensive Computing at Drexel University

这笔拨款将获得、安装和维护proteus++，它由图形处理单元（GPU）节点和高内存节点组成。该项目为德雷塞尔大学带来了数据密集型计算（需要大内存和高吞吐量的计算）硬件。该计算将加强在精密医学（基因组学、绘制大脑和模拟分子）、先进制造业、环境建模和许多其他应用中需要大型数据集的研究。计算机还将通过课程和合作工作经验帮助培养学生的数据密集型计算能力，并通过让学生参与计算机研究以及整个地区的本科院校（包括传统黑人学院和大学和女子学院）利用这些资源和加强计算机教育来扩大对计算机的参与。研究和培训的成果不仅将推进基础研究，还将促进创新，使费城地区的科技、健康和生物产业受益。proteus++将支持数据密集型（需要大内存和高吞吐量）的计算发现，并通过提供新的科学能力和更快的计算速度的混合架构影响200多个用户。这个项目的智力价值来自于大量的研究课题，这些课题将被proteus++提供的纯粹的计算能力和大数据处理能力大大增强。由于混合计算和大内存计算，高通量基因组学最终可以从大量和多样化的查询中同时搜索成千上万的微生物基因组，尤其是涉及在各种环境中发现目前未知的微生物。在proteus++的混合架构下，增强的采样分子模拟和混合分子动力学/对接方法将大大提高速度，从而能够精确地观察罕见的生物分子事件过程，并研究前所未有的大量复杂蛋白质靶点。此外，材料行为和杂质的预测和模拟得益于协处理器架构改进了纺织品、医药和能源的材料设计。最后，通过gpu加速的蒙特卡罗模拟，了解大脑活动现在触手可及，这将改进利用大脑中光组织相互作用的技术。除了对德雷克塞尔大学的直接影响外，URCF能力的增强还促成了费城高性能计算联盟（PHPCC）的建立，这是德雷克塞尔大学和一些本地/区域本科院校之间的合作伙伴关系，目的是让更多的联盟成员，包括德雷克塞尔大学、教师、博士后和学生，接触到计算发现技术的力量和可能性。以讲习班和课程形式进行的能力建设培训将帮助新的和现有的PHPCC用户不仅了解当地大学研究计算设施资源，而且了解美国其他机构可用的先进计算资源。此外，高达200万核心学时将提供给非drexel PHPCC教师和学生使用。该奖项反映了美国国家科学基金会的法定使命，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。

This grant will acquire, install, and maintain Proteus++, which is composed of Graphical Processing Unit (GPU) nodes and a high memory node. This project brings data-intensive computing (computing requiring large memory and high-throughput) hardware to Drexel University. The computing will strengthen research requiring large datasets in precision medicine (genomics, mapping the brain, and simulating molecules), advancing manufacturing, environmental modeling, and many other applications. The computing will also help train students in data-intensive computing through classes and co-operative work experiences and broaden participation in computing by engaging students in computing research as well as undergraduate institutions (including Historically Black Colleges and Universities and women's colleges) throughout the region to use the resources and enhance computing education. The research and training that results will not only advance fundamental research but enable innovation that will benefit the tech, health, and biology-enabled industries in the Philadelphia region.Proteus++ will enable computational discovery that is data-intensive (needing large memory and high-throughput) and impact over 200 users by offering hybrid architectures that provide new scientific capabilities and enable faster computations. The intellectual merit of this project derives from a large collection of research topics that will be greatly enhanced by the sheer computational power and large-data processing that Proteus++ provides. Thanks to hybrid and large-memory computing, high-throughput genomics can finally enter the regime of searching simultaneously tens of thousands of microbial genomes from a large volume and diversity of queries, not least which involves discovery of currently unknown microorganisms in a variety of environments. Enhanced-sampling molecular simulations and hybrid molecular-dynamics/docking methods will significantly increase in speed under the hybrid architecture of Proteus++, enabling precise views of rare biomolecular event processes and investigation of an unprecedentedly large number of complex protein targets. Also, the prediction and simulation of material behavior and impurity benefits from co-processor architectures improving materials design for textiles, medicine, and energy. Finally, understanding brain activity is now within reach through GPU-accelerated Monte Carlo simulations, which will improve technologies that exploit light tissue interaction in the brain. Besides its direct impact on Drexel, the enhancement of URCF capabilities enables the creation of the Philadelphia High Performance Computing Consortium (PHPCC), a partnership among Drexel and a number of local/regional undergraduate-only institutions, for the purposes of exposing larger numbers of Consortium members, including Drexel, faculty, postdocs and students to the power and possibilities of computational discovery techniques. Capacity-building training in the form of workshops and courses will help new and existing PHPCC users learn not only about local University Research Computing Facility resources but also about advanced computational resources available at other institutions within the US. Furthermore, up to 2 million core-hours will be made available for use by non-Drexel PHPCC faculty and students.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.