Petascale Molecular Dynamics Data Processing System

描述（由申请人提供）：我们建议获得一个集成的计算机系统，用于大数据存储、高性能计算和最先进的可视化，该系统可连接到国家超级计算中心的当前和下一代太级到千万亿级计算机，以支持大规模和长期的膜过程、生物纳米技术设备、细胞力学、细胞能量转化和基因转录的生物分子模拟。100tb的模拟数据集将存储在容量为204tb的磁盘阵列上，并通过多个10gb /s网络访问。存储数据的高性能分析将通过两个大型共享内存（每个256千兆字节）服务器和图形处理单元加速（NVIDIA tesla具有4teraflop /s单精度浮点运算性能）实现，相当于每个处理器在预期应用程序中的实际性能为数百个内核。通过NVIDIA Quadro图形加速器，4gb内存、teraFLOP/s浮点性能和102gb /s内存带宽，该系统将驱动四个单用户可视化前端和一个最先进的980万像素多用户图形投影系统。该系统将支持美国国立卫生研究院伊利诺伊大学厄巴纳-香槟分校的研究人员正在进行的实验-计算合作项目，以原子细节解决涉及DNA解旋酶、脂蛋白组装、核糖体中的蛋白质合成、肌肉力量产生和弹性、电压门控离子通道、膜转运和生物能量学的单分子过程，以及纳米孔中的DNA单碱基对检测。这些项目在国家超级计算中心的万亿级-千万亿级计算机上进行模拟，生成10-100太字节的数据，这些数据需要在本地存储、分析和可视化。中心的技术进步导致数据量增加了1000倍，必须配备足够的本地数据分析设备。拟议的系统就是为了达到这个目的。首席研究员指导NIH大分子建模和生物信息学资源，该资源将运行该系统，是开发所需生物分子建模和分析软件的领导者，并且是新系统所采用技术的专家用户，特别是图形处理单元加速。

DESCRIPTION (provided by applicant): We propose to acquire an integrated computer system for large data storage, high-performance computation, and state-of-the-art visualization that links to present and next generation terascale- to-petascale computers at the national supercomputing centers to support large-scale and long- time biomolecular simulations of membrane processes, bionanotechnological devices, cell mechanics, cellular energy transformation, and gene transcription. Hundred-terabyte simulation datasets will be stored on a disk array with a capacity of 204 terabytes and accessed by multiple 10-gigabit/s networks. High-performance analysis of the stored data will be enabled through two large shared memory (256 gigabytes each) servers and through graphics processing unit acceleration (NVIDIA Teslas with 4 teraFLOP/s single precision floating point arithmetic performance) equivalent to hundreds of cores per processor in real performance on envisioned applications. Through NVIDIA Quadro graphics accelerators with 4 gigabytes of memory, teraFLOP/s floating point performance, and 102 gigabyte/s memory bandwidth, the system will drive four single-user visualization front ends and a state-of-the-art 9.8 million pixel multi-user graphics projection sys- tem. The system will support ongoing experimental-computational collaborative projects among NIH investigators at the University of Illinois at Urbana-Champaign to resolve, in atomic detail, single molecule processes involving DNA helicases, lipoprotein assembly, protein synthesis in the ribosome, muscle force generation and elasticity, voltage-gated ion channels, membrane trans- port, and bioenergetics, as well as DNA single base pair detection in nanopores. The projects carry out simulations on terascale-petascale computers at national supercomputing centers that generate 10-100 terabytes of data which need to be stored, analyzed, and visualized locally. Technological advances at the centers leading to 1,000-fold data size increases must be matched by adequate local data analysis equipment. The proposed system serves this purpose. The Principal Investigator directs the NIH Resource for Macromolecular Modeling and Bioinformatics that will operate the system, is a leader in developing the needed biomolecular modeling and analysis software, and is an expert user of the technologies employed by the new system, in particular graphics processing unit acceleration. PUBLIC HEALTH RELEVANCE: The proposed computational microscopy system will assist in devising treatments against cancers, muscle disorders, cardiovascular disorders, in fighting drug resistance, and in discovering new drugs, e.g., new antibiotics. Experiments on DNA, lipoproteins, protein synthesis, muscle function, and nerve activity are interpreted through computer simulations that provide a tom by a tom views of the body's molecular machines. These views reveal the chemical detail needed to develop new pharmacological treatments.