计算网格环境下网格文件系统研究
A Study of GOS Grid File System in a Computational Grid Environment
【中文摘要】 为了将分布在不同地点的计算机资源共享起来,便于利用空闲的CPU或存储空间等资源,出现了网格的概念。人们希望方便的访问网格平台上的各种资源如计算资源、存储资源,科学仪器等。由于网格平台的计算机是分散的、异构的,当网格提供计算服务时所需要的输入输出数据不一定在同一个节点,这就需要有相适应的数据管理和传输平台做支持。在由计算网格节点和应用网格组成的国家网格计算环境下,由于网格节点内部用户认证和社区之间的访问控制机制限制,网格数据访问缺乏有效的统一的途径,给网格用户的使用带来了极大的不便。在这种情况下,急需一套对网格文件实施有效管理和访问的机制,实现网格资源的优化整合,方便用户的使用。本文在计算网格——中国国家网格基础系统软件特有GOS的环境下,设计并实现一套完整的网格文件系统,在本文中,研究了文件及访问路径的表达问题,提出了一个特殊的逻辑树,通过遍历逻辑树的各个节点就可以访问网格中的各个物理文件;同时研究了各个网格节点资源动态挂载和部署的问题,在GOS网格系统中,我们引入了Mount点的概念,通过对Mount点的管理来实现节点文件系统的动态部署,一个Mount点是一个包含节点信息的元信息,它通过GOS的Naming机制扩散的网格的每一个节点以供用户访问;还研究了文件在网格环境下的可靠文件传输问题,引入开源框架Quartz作为网格文件传输的调度器,实现对文件传输各个状态的监视,在错误发生时记录错误并在以后启动重传,从而保证传输任务的有效进行。通过大量的测试和使用,GOS网格文件系统已经成功融入到GOS体系中,为GOS提供了有效的文件访问解决方案。
【英文摘要】 In order to share computer resources which distribute in different locations and use idle CPU or storage space and other resources, we put forward the concept of grid. It is hoped that we can easy access to the grid platform, a variety of resources such as computing resources, storage resources, scientific instruments, etc. because grid platform is decentralized, heterogeneous, when the input and output data that grid computing services needed are not in the same node, We need data management and delivery platform to support this situation. In a computational Grid environment comprised of grid nodes and domain-specific applications, because of user authentication in the internal grid nodes and the restrictions of access control mechanism in the communities, the lack of an effective unified way to access the file of grid node is very inconvenience for grid user. In the GOS environment,an effective management and access mechanisms is needed for grid file, achieving optimal resource integration and convenience to use for user.In this paper, at the basis of China National Grid System Software GOS grid environment, designed and implemented a complete set of grid file system. This thesis has studied the following questions: Firstly, the problem of file and path-access expression, a particular logic tree has been raised, and all physical files can be accessed through the logic tree node. Secondly, the problems of resources dynamic deployment and mounting have been studied. In the grid system GOS, the concept of Mount-Point have been introduced into the system and completed the dynamic deployment of node file system. Mount point is a node of information that contains meta-information. It was spread to all gird node by Naming mechanism, so users can access it at their own node. Thirdly, the problem of Reliable File Transfer have been studied, open-source framework Quartz have been introduced into grid file transfer scheduling management. It watch the states of file transfer and start to retransmission when there is an error been recorded, ensuring the tasks of file transmission is reliant. Through many testing and using, GOS grid file system has been successfully integrated into the GOS and provided an effective file access solution.
【中文关键词】 计算网格; GOS; 网格文件系统; Mount机制; 命名空间; 可靠文件传输
【英文关键词】 Computational Grid; GOS; Grid File System; Mount mechanism ; Namespace ; Reliable File Transfer
计算网格环境下网格文件系统研究
摘要 3-4
Abstract 4-5
第一章 绪论 9-15
1.1
