【作者】 孟然；

【导师】 蔡开裕；

【作者基本信息】 国防科学技术大学 ， 计算机科学与技术， 2006， 硕士

【摘要】 网格计算是当前分布计算研究领域的热点,其核心是网格资源管理,而网格资源发现则是网格资源管理中一个基本组成部分,它为网格资源调度寻找满足应用需求的各种资源。随着网格的快速发展,集中式的资源发现方式将会出现性能瓶颈。因此,网格的资源发现方式应该是非集中的方式,才能够适应大规模、动态的网格环境。数据网格是当前应用领域中的一类重要的网格,实现了广域范围内数据资源的共享以及数据处理能力的共享。本文基于数据网格的特点,提出了一种分布式层次化数据资源发现方法。具体地,本文做了以下几个方面的工作:1.层次化网格数据资源发现方法框架建立了网格数据资源发现的理论分析模型和分析指标,并与现有方法进行了比较,提出了一种层次化的网格数据资源发现方法框架。该方法框架由两个层次构成,即上层的数据资源类型索引网络,和下层不同类型数据的底层资源网络。基于这种框架,可以有效地限制资源发现请求以及资源信息更新消息的扩散范围,以达到提高资源发现效率的目的。2.基于元数据的资源类型索引网络提出了一种基于网格元数据的数据资源类型组织方法,并讨论了分布式元数据结点发现、覆盖拓扑构造与维护,以及元数据注册和元数据请求转发方法。3.基于局部信息的资源定位方法提出了一种基于局部信息的资源定位方法,用于在底层资源网络中搜索满足给定需求的全局较优的资源。我们还建立数学模型分析了该方法的效果,结果表明在一定条件下,该方法能够以较小的消息与时间开销在相当规模的数据网格中找到全局较优的资源。更多还原

【Abstract】 Nowadays, Grid Computing is a hot research topic in distributed computing area. Resource management is acknowledged as one of the main points of Grid Computing, and resource discovery is a basic issue in Grid resource management, which concerns discovering resources in Grid to meet the requirement of applications. With the rapid development of Grid, centralized Grid resource discovery schemes have potential scalability and performance problems. Therefore, in order to survive the dynamic and larger-scale Grid environment, resource discovery should be decentralized and should not rely on centralized control.Data Grid is one of most important Grids, and it provides sharing of both data. resources and the capability of data processing globally. The thesis thus focuses on the Data Grid, and presents a distributed and hierarchical data resource discovery approach. Its distinguished feature is the high efficiency of the resource discovery process while keeping completely distributed.In details, the thesis makes following contributions:1. The framework of the distributed and hierarchical data resource discovery approach.The thesis first constructs the theoretical model for data resource discovery. Then a hierarchical framework for data resource discovery, as well as the comparison with current methods, is presented. The presented framework may effectively control the propagation of messages.2. The metadata-based resource type index network.Different types of data resources are organized into different resource sub-networks based on their metadata. We also discuss the topics on distributed metadata nodes discovery, topology construction and maintenance, metadata registration, and metadata request forwarding.3. The resource locating method inside the underlaid data resource network based on local information.Such resource locating method is used to find out the optimal data resource inside the underlaid data resource network. An analysis model is built to study the effectiveness of the presented method, and results are also validated by simulation. It is shown that, this method can find resources with relative high qualities among all qualified resources in a big resource network with small number of hops.更多还原