节点文献

网络并行计算网格化平台的构建研究

Research on the Implement of Grid Parallel Computing Platform

【作者】 罗伟刚

【导师】 李代平;

【作者基本信息】 广东工业大学 , 计算机软件与理论, 2006, 硕士

【摘要】 高性能应用需求的迅猛发展,使得单台高性能计算机已经不能胜任一些超大规模应用问题的解决。网络的发展和分布式计算的提出,使得通过多机互连的并行计算来解决大型的计算问题已成为许多人的首要选择。目前,分布式计算已经发展到了一个新阶段,网格计算已被视为下一代分布式计算的标志。网格是借鉴电力网的概念提出来的,它的最终目的是希望给使用者提供与具体的计算设施无关的通用的计算能力,让用户在使用网格计算能力时,就如同现在使用电力一样方便。 目前,国内外都致力于研究网格计算,主要研究系统核心运转以及组织各部分有效工作的底层问题。而关于并行计算的研究则已较为普遍,主要集中于应用领域及计算性能等方面。将网格计算的概念应用到并行计算方面的工作则比较少,这是本文的立意所在。 本文提出了一种分布式环境下的高性能计算平台——并行计算网格化平台(Grid Parallel Platform,GPP),该平台将网格资源自发现及利用的优点移植到并行计算中,自动构建好用户所需的并行计算环境,并且将用户所使用的计算程序自动发送到各节点机中,使用户可以透明的使用并行计算,免去了并行计算中许多繁琐的配置步骤,做并行计算就像做本机的单机运算一样方便。因此,网络并行计算的网格化平台的构建研究不仅有利于网络并行计算的推广与普遍应用。也提出了让传统的并行计算向网格方向发展的一条可能之路。 本文首先对网络并行计算的体系结构、通信机制以及构建方式进行分析研究;其次,研究了网格计算中的资源搜索和利用机制;接下来对本系统的功能要求、平台定位进行了详尽的分析;然后给出了本系统的设计过程;最后描述了系统的实现过程,并利用一个地质勘探中电法勘测的工程算例,分析了系统的效率。

【Abstract】 Nowadays, a single high-performance computer cannot suit for the development of high-performance application. With the development of computer network and distributed system, using parallel computing technology to solve complicated problem has become any people’ s first choice. Currently, distributed computing has developed to grid computing stage. Grid computing has been considered to be the next generation of distributed computing. The concept of grid is represented in the light of electric power grid, its aims to provide with computing ability regardless the special computing facilities to the users. And people can use grid computing in a convenient way just like they use electric force.Grid computing is emerging as one of the hottest research topics all over the world, most of correspond projects are concentrated on the operation of its core system and how to organize the low layer’ s components in an effectively way. Research about parallel computing has become very popular, and it focuses on the application of parallel computing and calculating performance. But few projects introduce the feature of grid to parallel computing, it is the innovations of this paper.The paper presents a new high performance computing platform in the distributed environment—Grid Parallel Platform (GPP). GPP introduces the basic purpose of Grid to the Parallel Computing. Parallel computing system can discover and add new resource automatically by GPP. And GPP can also help the user send the computing program to all machine-nodes automatically. GPP avoids many tedious configuration steps for the users. People can use parallel computing just like using single computer. The researching of GPP not only promotes the development of network parallel computing, but also submits a new possible road of integrating parallel computing and grid computing.Firstly, the thesis introduces the architecture, communication mechanism and building of parallel computing system. Secondly, it studies the technology of resource discovery and adding in grid computing. Thirdly, it brings forward the system’ s location and function requirements. Then, i t introduces the design of GPP. At last, the paper describes the implement of the platform, and it puts forward a use case to test the platform’ s speedup and parallel efficiency.

  • 【分类号】TP393.02
  • 【被引频次】10
  • 【下载频次】411
节点文献中: 

本文链接的文献网络图示:

本文的引文网络