【作者】 王璿；

【导师】 孔令富；

【作者基本信息】 燕山大学 ， 电路与系统， 2005， 博士

【摘要】 网格计算技术是当前分布式计算研究领域的热点。它将地理上分布的、异构的各种资源通过高速网络连接起来,实现高性能联合计算,共同完成重大应用项目。目前,在关乎国计民生的各个行业中都已感受到了网格计算技术带来的革命性影响。然而,将网格计算技术引入特定应用领域是一项实施性很强的技术,其相关的理论研究和技术实现还很薄弱。因此,面向特定应用领域需求,展开对网格计算环境基本理论、技术和实施方法的研究,具有重要的理论意义和实用价值。本文利用网格计算环境提供的强大计算能力和广域共享的资源,解决了曲面重构应用中重构速度慢、资源共享能力不足等问题,并构建了曲面重构网格应用原型系统。取得的相关成果为解决其它同类应用问题提供了理论依据和实现方法。首先,比较了集群计算系统与网格计算系统在技术实现上的异同;面向曲面重构应用领域需求,借助GT3设计了基于OGSA的网格应用系统体系结构;围绕如何实现当前网格资源和用户请求之间建立最佳映射的中心议题,提出了一种保障服务质量的网格服务匹配模型,并采用非线性规划思想详细刻画了资源描述、服务发现、服务选择和容错等网格构建技术。其次,针对现有资源描述机制中对服务性能描述不足的问题,定义了四个与曲面重构应用性能相关的QoS参数,在现有服务描述机制的基础上进行了QoS参数的扩展,并采用WSDL规范实现了重构资源的封装。第三,针对现有服务发现机制在支持QoS及大规模查找方面存在的不足,结合P2P系统和人类社区的构建方式,设计了基于对等服务社区的非集中式服务发现框架;提出了社区内的两阶段服务发布和更新的思想以及社区间的有偏消息转发协议;在此基础上,提出了基于两阶段匹配的服务发现算法,通过仿真实验验证了算法的先进性;为保证服务发现机制及时、准确地获取有效的服务信息,设计并实现了网格监控服务。第四,针对网格任务执行特点,引入分形理论将基于QoS参数的服务选择问题转换为多目标决策问题,从而简化问题求解过程;为解决价格与响应时间两个性能目标不相容的问题,提出了满足价格-时间均衡的服务选择算法,通过仿真实验验证了算法的有效性;针对服务请求的模糊特性,给出了QoS参数的模糊隶属度和重要性的度量方法,提出了模糊集下满足多目标均衡的服务选择算法;依据所定义的服务可用性参数和服务等价关系,设计并部署了网格备份服务,使得服务匹配过程具备了一定的容错能力。最后,以曲面重构为应用背景,将基于6-DOF并联机构的双目主动视觉曲面重构平台与网格技术相结合,实现了曲面重构网格应用原型系统;通过所开发的网格门户调用部署的重构服务对原型系统进行了测试。更多还原

【Abstract】 Grid computing technology is a hot research field of distributed computing. It connects all kinds of resources which are geographical distributed and heterogeneous by high-speed network, to achieve high performance federation computing and complete major application projects together. Now, it brings the revolutionary influence to various domains of the national economy and the people’s livelihood. However, it’s still a strong implementation of the technology to relate the grid technology to specific the application area. The related theoretic research is still weak. So, our project does researches on the basic theories, technologies and implemental methods for certain application area. It has theoretical and practical significance. This dissertation utilizes powerful computing abilities and vast resources of grid to improve the speed of shape-reconstruction and share the shape-reconstruction resources. And the prototype is built. The related productions offer theoretical foundations and practical techniques to same kind applications.Firstly, the similarities and differences of implemental methods are compared about cluster computing and grid computing. To meet the requirement of shape-reconstruction application, the architecture of the shape-reconstruction application oriented grid computing platform based on OGSA and GT3 is proposed. Surrounding a core topic that is how to realize the best mapping between grid resources and user’s request, a service matching model which enables QoS-guaranteed is defined. And some implementation technologies such as resource description, service discovery, service selection and fault-tolerant are depicted with nonlinear programming idea.Secondly, to resolve the deficiency of service performance description in existed resource description mechanism, four QoS parameters related with the shape-reconstruction application are defined. Some extended works are done for existed resource description mechanism. And the shape-reconstruction resource is encapsulated with WSDL standard.Thirdly, existed service discovery mechanism lacks the ability of QoS-guaranteed and large-scale distributed searching. Combining with P2P and the people communities, the decentralized grid service discovery frame based on service community is constructed. The idea of two-stage information release and update is proposed interior of community. And the flexible protocol is designed to guarantee the high efficient of messages forwarding among communities. The services discovery algorithm based on two-stage matching is proposed and the advantage of the algorithm is validated through emulation experimentation. The grid monitoring service is designed and deployed for gaining the available service information in time.Fourthly, aiming at the traits of grid task executing, the problem of QoS-based service selection is transformed into the multi-goals optimal problem using the fractal theory. To resolve the antipathy problem of two performance objects, the price-time balance service selection algorithms is proposed and the advantage of the algorithm is validated through emulation experimentation. Due to fuzzy traits of service request, the measurements for fuzzy subjection and QoS weightiness is given, and the multi-goals balanced service selection algorithm is proposed to get well service. According as the service availability and the service equipollent relationship, a spare service is designed and deployed. It makes the service matching process possess the fault-tolerant ability.Finally, combining with the SR-BAV platform on 6-DOF parallel robot and grid technologies, the shape reconstruction grid application prototype system is constructed. The grid portal is designed to use grid services conveniently. The validity of prototype shows by example tests.更多还原