一种分布式网格计算框架以及在大规模分子动力学模拟中的应用  

Distributed Grid Computing Framework and Application in Molecular Dynamics Simulation

作　　者：王文睿[1] 陈国良[1] 邢利荣[2] 陈华平[1] 孙广中[1] 单久龙[1] 

机构地区：[1]中国科学技术大学国家高性能计算中心,安徽合肥230027 [2]上海电力学院数理系,上海201300

出　　处：《小型微型计算机系统》2006年第7期1326-1330,共5页Journal of Chinese Computer Systems

基　　金：国家高技术研究发展"八六三"计划基金项目(2002AA104560)资助;教育部归国留学人员基金资助.

摘　　要：分子动力学(m o lecu lar dynam ics)模拟蛋白质等大分子内原子间的相互作用.蛋白质折叠所需的时间通常在微秒(10-6s)量级,而进行模拟的时间步长在飞秒(10-15s)量级,并且每步需要计算大量的相互作用(O(n2),n为原子数),以致于无法模拟足够长时间的折叠过程.现今在满足精确度的需求下没有更好的模拟算法.最近,生物学家研究了一种分布式的动力学方法,使得可以利用分布在In ternet上的计算机进行并行模拟成为可能.本文的目标是设计并实现在分布式P 2P和网格计算环境等多种异构计算资源下进行动力学模拟的可靠框架,以便更大限度地利用计算资源,加快计算过程.我们基于Java和W eb serv ice技术,已经实现了对应用透明的计算框架,并已将它扩展到我们的网格计算环境.实验表明分子动力学模拟程序在该框架下运行良好.Molecular dynamics performs simulation of the interactions between large number of atoms in the molecular of protein. The time of the folding process of protein is number of 10^-6 second, while the time step of the simulation is the number of 10^-15 second, and in every step, computing large amount （O（n^2）, n is the number of atoms）of interaction is needed. Therefore, the overall folding process cannot be simulated. No better algorithm can match the requirement of precision. The biologist then develop a distributed process of dynamics, leveraging the computing resources on the internet to make parallel simulation. The goal of this paper is to design and develop a reliable simulation framework for molecular dynamics in various environment such as the P2P and the Grid environment. It can utilize computing resources on a higher level and fasten the computing. We have implemented such framework based on Java and Web services technology and deployed it to our Grid platform. The experiment shows that the simulation program runs well in this framework.

关 键 词：网格计算 分子动力学模拟 分布式计算 P2P 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]