节点文献

高可用性附网存储集群的研究与实现

Research and Implementation of High Availability NAS Cluster

【作者】 吴振宇

【导师】 谢长生;

【作者基本信息】 华中科技大学 , 计算机系统结构, 2004, 硕士

【摘要】 在对NAS高可用性集群的关键技术进行分析的基础上,深入研究了NAS高可用性集群的体系结构、单一系统映像技术等方面的内容。初步设计实现了一个双节点的NAS高可用性集群系统的原型,同时该系统具有良好的可扩展性。 NAS高可用性集群系统设计中面临的首要问题是系统的体系结构。分析了传统NAS集群方案的基本原理、体系结构,建立了了一种高性能的NAS集群方案模型。它对用户采用分散服务,对全局源数据采用集中管理,发挥了分散和集中的各自优势,并详细讨论了它的实现原理。单一系统映像技术是建立NAS高可用性集群的关键技术之一,由于NAS集群主要用于数据存储,所以在实现单一系统映像时与I/O性能相关的技术是重点。说明了单一系统映像所提供的服务,详细讨论了实现单一I/O空间,单一入口点,单一文件层次结构的技术方案。NAS集群中的高可用系统软件是整个NAS高可用集群系统的核心部分。该系统软件在WindowsNT系统环境下设计实现,其核心的实现形式为WindowsNT下的系统服务。通过一系列的监测代理,实现了对系统状态的监测。整个系统实现了完全自动的故障接管。当一个节点上运行的应用出现故障后,系统能在很短的时间内将其自动转移到另外的节点上。

【Abstract】 Based on the analysis of high availability technique of computer system,a research on NAS cluster’s architecture, single system image is done.A prototype of two nodes NAS cluster is designed and implemented, and it will have good scalability. To design a NAS cluster, system architecture is the first issue that has to be solved. By analyzing the basic theory and architecture of traditional cluster solutions and insightful study on the advantages and disadvantages of traditonal cluster solutions, designed a high performance NAS cluster model,which provides distributed servers for customers while centralizedly managing the global metadata,thus make the most use of the advantages of both distribution and centralization.and the method of implementation is discussed in detail.Single system image is the key technology of designing NAS cluster. because data store is the main use of NAS cluster, focused on the technology included I/O performance.discussed the main server of single system image.The desigen of single I/O space,single entry,single file hiberarchy is given.High availability software is the kernel of the NAS cluster system .the software is implemented under windows server environments. The implementation of the kernel is windows server.Through a series of monitor agent , can monitor the whole system status.When the application has problem on a node ,the system can automatically transfer the application to another node in a short minute.discussed the design of the kernel module and monitor module,and the detail implementation is given.

  • 【分类号】TP333
  • 【被引频次】2
  • 【下载频次】197
节点文献中: