Acquisition of a Computer System for Biophysical and Computational Chemistry

要求拨款购买一台基于RISC架构的多处理器计算机，以取代凸式C220。所要求的设备将用于的项目包括：溶剂化蛋白质和液体的计算机模拟（R.M. Levy），双螺旋DNA的计算机模拟（W.K. Olson），生物分子结构和相互作用（h.m. Berman），共轭分子内旋的潜在表面（L. Goodman），电子结构计算中的介质效应（K. kroh - jespersen），核磁共振结构测定和蛋白质折叠（J. Baum），大分子配合物的结构与相互作用（M. Georgiadis）。对基于RISC的多处理器的需求是基于这些项目的计算密集型性质，以及对复杂生物分子系统计算结果的可视化需求。罗格斯大学因其生物物理化学研究项目的实力而享誉国际。“生物分子结构和动力学”是连接参与该提案的研究人员的一个中心主题。生物大分子的复杂性和庞大的体积使得分析这些系统的结构和动态特征需要大量的计算。这对于理论研究是正确的，包括使用计算机图形学的分子建模，溶液中生物分子的计算机模拟以及生物分子和共轭分子的电子特性的量子化学计算。提案中所代表的实验技术也是如此：大分子x射线晶体学和核磁共振波谱学。共同研究人员进行的个别项目都涉及非常大规模的计算。分子生物物理学和计算化学的共享计算设施主要设在化学系内。生物物理化学计算资源的骨干是凸C220双处理器。1986年，在美国国立卫生研究院共享仪器拨款的资助下，购买了凸Cl单处理器。1989年，在美国国家科学基金会（National Science Foundation）和罗格斯大学（Rutgers University）提供的仪器补助金的资助下，Convex升级为双处理器C220。在购买的时候，凸的矢量架构是相当显著的，提供10-50倍的性能提升比DEC VAX 780硬件，它取代了。在今天的计算环境中，随着RISC计算机工作站的可用性，凸的性能就不那么令人印象深刻了。目前，也许凸的唯一值得注意的属性是它的纯粹的物理尺寸，它的功率需求，以及高昂的维护成本。我们建议用多处理器RISC计算机系统代替凸C220。在授予补助金后，将对备选RISC系统进行竞争性评估，然后做出最终的供应商选择。可能的供应商包括Silicon Graphics， Hewlett Packard， IBM和DEC.目前，我们正在认真考虑Silicon Graphics Power Onyx，主要是因为它具有成本效益的浮点性能，但也因为它的图形功能。我们提出的预算是基于购买SGI Power Onyx系统，该系统具有8个R8000 90 Mhz处理器，2 gb高密度内存，现实引擎图形硬件和64 gb磁盘存储。

Funding is requested to purchase a multi-processor computer based on the RISC architecture to replace the Convex C220. The projects for which the requested equipment will be used include: Computer Simulations of Solvated Proteins and Liquids (R.M. Levy), Computer Simulations of Double Helical DNA (W.K. Olson), Biomolecular Structure and Interactions (H. M. Berman), Potential Surfaces for Internal Rotation in Conjugated Molecules (L. Goodman), Medium Effects in Electronic Structure Calculations (K. Krogh-Jespersen), NMR Structure Determination and Protein Folding (J. Baum), and Structure and Interactions of Macromolecular Complexes (M. Georgiadis). The need for the RISC based multi-processor is based on the computationally intensive nature of these projects, and the need for the visualization of the computed results concerning complex biomolecular systems. Rutgers University is internationally recognized for the strength of its research program in Biophysical Chemistry. A central theme connecting the investigators participating in this proposal is that of "Biomolecular Structure and Dynamics". The complexity and sheer size of biological macromolecules makes the analysis of the structural and dynamic features in these systems computationally intensive. This is true for the theoretical studies including molecular modeling using computer graphics, computer simulations of biomolecules in solution and quantum chemical calculations of electronic properties of biomolecules and conjugated molecules. This is also true of the experimental techniques that are represented in the proposal: macromolecular x-ray crystallography and NMR spectroscopy. The individual projects carried out by the coinvestigators all involve very large scale computations. The shared computational facilities for molecular biophysics and computational chemistry are housed primarily within the chemistry department. The backbone of the computational resources for biophysical chemistry is the Convex C220 dual processor. In 1986 a Convex Cl uniprocessor was purchased with funding from a National Institutes of Health Shared Instrumentation Grant. In 1989, with funding provided by an instrumentation grant from the National Science Foundation and by Rutgers University, the Convex was upgraded to a dual processor C220. At the time of its purchase, the vector architecture of the Convex was quite remarkable, providing 10-50 times the performance gains over the DEC VAX 780 hardware which it replaced. In the context of today's computing environment, with the availability of RISC computer workstations, the performance of the Convex is much less impressive. At present, perhaps the only remaining remarkable attributes of the Convex are its sheer physical size, its power requirements, and exhorbitant maintenance costs. We propose to replace the Convex C220 with a multi-processor RISC computer system. A final choice of vendor will be made after competitive evaluations of alternative RISC systems is carried out after the grant award. Possible vendors include Silicon Graphics, Hewlett Packard, IBM and DEC. At this time, we are seriously considering the Silicon Graphics Power Onyx primarily because of its cost effective floating point performance, but also because of its graphics features. Our proposed budget is based on the purchase of an SGI Power Onyx system with eight R8000 90 Mhz processors, 2 GByte high density memory, Reality Engine graphics hardware, and 64 GBytes of magnetic disk storage.