The Implementation And Application Of High-performance Clusters For Computing Theoretical Chemistry.

The high-performance clusters are being used widely; it is normally applied on high-throughput calculation-intensive tasks. The applications generally include theoretical chemistry, quantum physics, weather predication, climate research, molecular modeling (computing the molecular structure and biological properties of compounds, polymers, polymers and crystals, etc.), and physical simulations (such as the simulation under the condition of aircraft wind tunnel, simulation of detonation, nuclear weapons research, and nuclear fusion research), etc.Chemistry has been developed into both experimental and theoretical science, theoretical chemistry can reveal the experimental results more deeply and could describe the nature of law and theory. Every critical chemistry breaken-out promotes and affects the discipline of chemistry and related disciplines, both theoretical and experimental chemistry boast the development of chemistry a lot. In 1998, The Chemistry Nobel Prize was awarded to two quantum chemists, W. Kohn and JA Pople. The methods in calculacal theoretical chemistry have become the integral part in the research and development of chemistry, physics, life sciences, and other basic science research and drug material industry. Nowadays, the significant progress in the study of theoretical chemistry is undergoing a revolutionary change in the whole discipline chemistry.Today, a high-performance computing application has become an important means for Theoretical and Computational Chemistry. China continues to increase investment in this field, from the earliest devices that can only be imported abroad, to domestic produced high performance computing devices. The high-performance clusters of our laboratory include from the domestic dawning 5000A, Lenovo 1800, and the Inspur Tissot, to imported SGI3800 and 3900. The application of high-performance computing in China is developing acceleratly, while the case of implementation of high-performance computing platform is pretty precious, especially with HP, IBM and other equipment from honored foreign companies even less. This paper is a new attempt in applying multiple IBM eServer P575 Minicomputer on the high-performance computing platform in theoretical chemistry.This article focuses on using IBM eServer P575 Minicomputer to implement the platform of Cluster, and installing, debugging, computing and computational comparing chemical softwares on the platform, as well as analyzing how to apply high-performance clusters of different structures more effectively in practical uses. This article reflects the author's major research achievements include the following:1. An in-depth and detailed description of the formation of multiple high-performance IBM eServer Power575 minicomputer cluster approach and it's the first time that the minicomputers IBM eServer Power5 processor family cases are used to form multiple clusters in China;2. Installation of AIX5L operating system on the cluster. The system is from the mainframe, and is of high integrity, manageability and high availability. GPFS file system is applied to improve and data storage stability in the software level, IBM LoadLeveler is used to schedule jobs and balance loads, to improve performance of job submission.3. Detailed description about software installation and configuration methods and practical experience on high-performance computing cluster applied for theoretical chemistry computing. Physical and chemical software are used widely, but due to the high price of commercial software which is as high as tens of thousands, many of which are for individual users (for example, the crack version), there is no detailed installation method. And even the few authorized users seldom applied them on a large cluster. This paper provides some useful references.4. Cases studies on the implementation of high-performance cluster system. The analysis of software usability udder the different usage of resources on the cluster, and the precious experimental conclusions have been achieved which will be beneficial for making more reasonable use of cluster resources and providing more cluster schedules .