基于Globus Toolkit和SunTM ONE Grid Engine构建计算网格

网格计算(grid computing)是近几年发展起来的一个崭新研究领域，引起国内外学术界及工业界的广泛关注。其目的是研究如何安全有效地将现有的各种计算资源(尤其是那些分布在Internet的异构网络中的计算资源)组织起来协同解决复杂的科学及工程计算问题。在化学信息学和生物信息学中最典型的应用是虚拟高通量筛选侯选药物分子。本文以两个Linux机群为基础，用开放源码的网格支持软件包Globus Toolkit 3．2及Sun^TM ONE Grid Engine 5．3成功构建了计算网格；并通过设计测试程序实现一次性提交多个作业(300个)以及分析作业在计算网格中各个节点的分配及运行情况，从而测试了计算网格的效率。结果表明，所构建的计算网格在保持原机群运行稳定、可靠的前提下，改进了系统资源的分配管理方式以及用户提交作业的方法，从整体上提高了网络计算资源的利用率，也同时方便了系统的管理。

A computational grid based on globus toolkit and SunTM ONE Grid Engine

Grid computing is a brand-new research field developed in recent years, which causes the extensive concerns in domestic and international academia and industrial circles. It is to study how to safely and effectively organize various kinds of existing computing resources, especially those distributed in nets of Internet with different constructions, to solve complicated calculation problems in science and technology in coordination. A typical application in chemoinformatics and bioinformatics is Virtual High-throughput Screening (vHTS) on millions of candidate drug molecules. In this paper, a computational grid is established on two Linux clusters with open source software packages Globus Toolkit 3.2 and Sun~(TM) ONE Grid Engine 5.3. In order to test its efficiency, a procedure is designed to submit 300 works at one time. And then the distribution and running situation of the works on each node of the grid are observed and analyzed. The result shows that the system does not affect the stability and reliability of the original system, and the efficiency of the computing is raised significantly. Furthermore, the management of the network resource allocation and the methods of job submitting are also improved.