Supercomputing: Rutgers' Catalyst for Discovery at the Frontiers of Medicine

描述：新泽西医学和牙科大学(UMDNJ)和罗格斯大学联合通过了一项州立法机构法案，于2013年7月1日生效。这所新扩建的罗格斯大学将UMDNJ的医学和牙科专家整合为一个世界级的生物医学研究机构。我们在医学、生物学、遗传学、化学、数学、药理学、生物化学、化学和生化工程、分子生物学和产科等系的研究人员参与了资金充足的研究项目，这些项目需要大量的计算资源。这些研究人员总共得到了国立卫生研究院的资助，每年的直接成本总额为6000万美元。许多调查人员正在努力在我们目前掌握的有限计算资源上进行计算。其他研究人员无法充分发挥他们研究的全部潜力 由于缺乏足够的计算。例如，如果可以对系统进行长达1微秒的间隔建模，一些分子动力学模拟将产生更现实的结果，但我们目前的基础设施的限制是大多数系统每月数十纳秒。作为另一个例子，基因组研究人员有数百个外显子组和rna-seq样本需要处理和分析，但由于分析管道的瓶颈，研究进展缓慢。我们建议的群集具有2,800个计算核心、每个计算节点128 GB的RAM和465 TB的存储空间，采用最先进、高能效的群集配置。此外，非常快速的节点互连(FDR Infiniband)的使用为高度可扩展的问题(如分子动力学和对接)提供了适当的交换矩阵，同时支持执行数千个单独的串行作业。新的计算集群将加强我们研究人员的研究，使其达到现有资源无法达到的水平。这个计算集群 将为支持广泛的生物医学研究的大规模、集中管理的资源设定一个新的范例。

DESCRIPTION: The University of Medicine and Dentistry of New Jersey (UMDNJ) and Rutgers University were joined by an act of the State Legislature, effective July 1, 2013. This newly enlarged Rutgers University integrates the medical and dental specialists from UMDNJ into a world-class biomedical research institution. Our researchers across the departments of Medicine, Biology, Genetics, Chemistry, Math, Pharmacology, Biochemistry, Chemical and Biochemical Engineering, Molecular Biology and Obstetrics are involved in well-funded research projects that have a need for a large computational resource. In total, these researchers have 64 NIH-funded grants with total annual direct costs of > $60M. Many of the investigators are struggling to perform their calculations on the limited computational resources that are currently at our disposal. Other researchers are not able to realize the full potential of their research due to the lack of sufficient computation. For example, some molecular dynamics simulations will produce more realistic results if the system could be modeled for intervals of up to 1 microsecond, but the limit with our current infrastructure is tens of nanoseconds per month for most systems. As another example, genomics researchers have hundreds of exomes and RNA-seq samples that need to be processed and analyzed, but the research is moving at a slow pace due to the bottle-neck of the analysis pipeline. We are proposing a cluster with 2,800 computational cores, 128GB of RAM per compute node and 465TB of storage in a state-of- the-art, energy efficient, cluster configuration. The use, also, of very fast node interconnects (FDR Infiniband) provides the appropriate fabric for problems that are highly scalable, such as molecular dynamics and docking while at the same time supporting the execution of thousands of individual serial jobs. The new computational cluster will enhance the research of our investigators, taking it to levels unattainable with existing resources. This computational cluster will set a new paradigm for a large-scale, centrally managed resource supporting a broad spectrum of biomedical research.