文档介绍：华中科技大学
硕士学位论文
面向分布式仿真的对等网格资源发现系统设计与实现
姓名:曾铁
申请学位级别:硕士
专业:系统分析与集成
指导教师:王乘
20060430
摘要
网格计算自九十年代中期提出以来,一直是人们关注的热点,其目的是为了在
多个动态的虚拟组织之间实现资源的全面共享。其目标是宏伟和令人期待的,但目
前其发展重点还主要集中于理论研究、标准的制定和基础设施的构建阶段,研究成
果以 Globus 项目为代表,它提供了很多构建网格基础设施的中间件。而与网格最终
目标有所相似的对等计算,近年来在应用领域却取得了很大成功,它利用了大量与
普通用户直接相关的位于边缘的桌面系统来实现资源的共享。
本文的研究试图摆脱传统网格对高性能计算环境的依赖,将个人桌面资源融合
到网格环境中来,使其不仅是网格的客户端运行环境,使对等网络资源成为网格资
源的一个组成部分,并将其移植运用于仿真技术。
本论文主要工作包含如下几个方面:
(1)对网格体系结构和对等网络模型作了深入研究,主要考察了面向服务体系结
构的优点,对等网络模型存在的问题。
(2)在对比网格计算和对等计算的基础上,提出了一种以对等方式组织,包含多
层次超级节点,架构于网格基础设施之上的对等网格模型。对模型的优点进行了详
细说明。
(3)从资源发现的角度研究了模型的连通性存在的问题,并据此提出了几点对网
格基础设施的改进措施;结合对比分析,对模型的网络拓扑结构、消息路由机制、
资源查找机制进行了最优化选择,将缓存运用在查找机制上,并提出了虚拟组织名
字目录(VND),来方便虚拟组织的查找。
(4)将模型的思想运用于一个仿真实例的设计,并对实验结果进行了分析和比
较,证明了模型及资源发现机制的设计的合理性。

关键字:网格; 对等计算; 资源发现; 超级节点

I
Abstract
The Grid technology that emerges since mid 1990s is always on the people's focus. Its
aim is to provide the sharability between multiple dynamic anizations. Though
the ultimate target of the Grid is great and excited, the phase of its development lays in the
theoretic, standard and basic-infrastructure, which could be represented by the Globus
project that provides a number of middle-ware to Grid infrastructure. Be similar in the
ultimate target to the Grid, peer-to-peer technology gets great achievements in the
application area, which uses plenty of desktop resources that having direct concern with
normal users.
This thesis attempts to get rid of the yoke of the traditional Grid's reliance to the
High-puters environment and make use of the desktop resources in the
Grid, making them not only the client peer of the Grid, making the peer-to-work
resources a part of the whole Grid resources. And migrate it to the simulating technology.
This paper's work contain the following aspects:
Firstly, gives a research on the Grid architecture and models of peer-to-work,
mainly on the advantage of the SOA(Service Oriented Architectur