权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

Parallel Computer with High Memory Nodes

高内存节点并行计算机

基本信息

批准号：
8447324
负责人：
David W Piston
金额：
$ 59.75万
依托单位：
VANDERBILT UNIVERSITY
依托单位国家：
美国
项目类别：
财政年份：
2013
资助国家：
美国
起止时间：
2013-07-11 至 2014-07-10
项目状态：
已结题

项目摘要

DESCRIPTION (provided by applicant): We request funds to purchase 180 high-memory parallel computer nodes to support data and computationally intensive, NIH-funded, biomedical research at Vanderbilt University and Medical Center. The proposed system will facilitate many applications that require high memory/processor supercomputer performance at a small fraction of the cost. This system will be built aside our successful 4000-processor parallel computer known as the Advanced Computing Center for Research and Education (ACCRE) at Vanderbilt, which supports high-performance computing for research from elementary particle investigations using the Large Hadron Collider to model system genetics to human behavioral studies. Despite our successes with the existing cluster, many state-of-the-art biomedical computation projects have become memory hogs, and thus caused severe inefficiencies in our current system. These projects include macromolecular structure prediction, next-generation sequencing, multi-dimensional imaging, and proteomic identification. ACCRE users have access to the instrument and training through our established infrastructure. ACCRE is a grass-roots organization that was initially created by faculty from throughout the institution and later awarded the strong institutional support it enjoys today. It was initially established through a Major Research instrumentation Grant from NIH in 2004 along with Departments of Defense and Energy grants to the School of Engineering and Department of Physics. ACCRE was upgraded through another NIH Shared Instrumentation Grant awarded in 2009. ACCRE is managed by three co-directors, one each from the College of Arts and Science, School of Engineering, and School of Medicine (of which Dr. Piston is the co- director). In keeping with the tradition of bottom-up ownership of ACCRE, policy and procedures are set by the co-directors upon recommendations of the Faculty Advisory Board, which consists of 10 active PI users. The proposed parallel computer will be used directly by more than 20 NIH-funded researcher groups and its impact will reach many more investigators through outreach efforts of centralized facilities such as the Center for Structural Biology, the Center for Human Genetics Research, The Small Animal and Cell Imaging Resources, and the Proteomics Core Laboratory. In this manner, the equipment will support more than 100 NIH grants. The major users of the system have primary appointments in six different departments within the Vanderbilt University School of Medicine, College of Arts and Science, and School of Engineering. Research programs with unmet demand for high memory node computation are concentrated in structural biology, human genetics, imaging, and proteomics. These new high memory nodes will make it possible to carry out data and computationally intensive biomedical investigations otherwise not feasible thus significantly increasing the rate of scientific discovery in a variety of biomedical disciplins.

描述（由申请人提供）：我们请求资金购买 180 个高内存并行计算机节点，以支持范德比尔特大学和医学中心的数据和计算密集型、NIH 资助的生物医学研究。所提出的系统将以很小的成本促进许多需要高内存/处理器超级计算机性能的应用程序。该系统将建立在我们成功的 4000 处理器并行计算机（称为范德比尔特高级计算研究和教育中心 (ACCRE)）的旁边，该计算机支持高性能计算，从使用大型强子对撞机的基本粒子研究到系统遗传学建模再到人类行为研究。尽管我们在现有集群上取得了成功，但许多最先进的生物医学计算项目已经成为内存消耗大户，从而导致我们当前系统的效率严重低下。这些项目包括大分子结构预测、新一代测序、多维成像和蛋白质组鉴定。 ACCRE 用户可以通过我们现有的基础设施使用仪器并接受培训。 ACCRE 是一个草根组织，最初由整个机构的教师创建，后来获得了如今所享有的强有力的机构支持。它最初是通过 2004 年 NIH 的重大研究仪器拨款以及国防部和能源部向工程学院和物理系的拨款建立的。 ACCRE 通过 2009 年授予的另一项 NIH 共享仪器拨款进行了升级。ACCRE 由三名联合主任管理，分别来自艺术与科学学院、工程学院和医学院（其中 Piston 博士是联合主任）。为了与 ACCRE 自下而上所有权的传统保持一致，政策和程序由联合主任根据教职顾问委员会的建议制定，该委员会由 10 名活跃的 PI 用户组成。拟议的并行计算机将由 20 多个 NIH 资助的研究小组直接使用，通过结构生物学中心、人类遗传学研究中心、小动物和细胞成像资源以及蛋白质组学核心实验室等集中设施的外展工作，其影响将惠及更多研究人员。通过这种方式，该设备将支持超过 100 项 NIH 拨款。该系统的主要用户在范德比尔特大学医学院、文理学院和工程学院的六个不同部门担任主要职务。高内存节点计算需求未得到满足的研究项目集中在结构生物学、人类遗传学、成像和蛋白质组学领域。这些新的高内存节点将使执行数据和计算密集型生物医学研究成为可能，否则这些研究是不可行的，从而显着提高各种生物医学学科的科学发现率。