节点文献
基于移动代理的网格资源监控技术的研究
The Research of Grid Resource Monitoring Technology Based on Mobile Agent
【作者】 方娟;
【导师】 张书杰;
【作者基本信息】 北京工业大学 , 计算机应用技术, 2005, 博士
【摘要】 网格是伴随着互联网技术迅速发展起来的,最初是专门针对复杂科学计算应用的一种新型计算模式,这种计算模式把整个网络整合成一台巨大的超级计算机。目前对网格技术的各个方面的研究都正在发展之中。网格的核心优势在于它能发现、分配并协商使用网络可以达到的范围内的各种能力——它们可以是计算机提供的计算服务、软件提供的应用服务、网络提供的带宽或者是存储系统提供的存储空间。网格的资源管理就是定位某种能力、协调对它的使用、使用该能力并监控它的状态。网格环境下资源管理和本地系统中资源管理的主要区别在于网格环境所管理的资源跨越多个管理域,这种分布造成了相似的资源在进行配置和管理时所产生的异构性问题。因此,网格资源管理的大部分早期工作主要集中于解决异构性问题,如制定标准的资源管理协议以及标准的资源和任务需求描述机制。本文主要开展了网格资源监控方面的研究。网格监控是用来度量和显示网格组件在某一时刻的状态。在应用端所有组件都必须监控,包括软件(如应用程序、服务、中间件和操作系统)、终端主机的硬件(如CPU、磁盘、内存和网络接口)和网络(如路由器、交换机或端到端的路径)。监控可以用在很多方面,包括状态检查、故障检测、性能协调和调试等。对监控数据的使用有故障处理和错误检测、性能分析和调整、指导调度决策、收集数据以提高下次运行性能、计算执行自适应、调试、审计和入侵检测等方面。本论文中对网格资源的监控提出并设计了一种新的方法,其中引入了移动Agent 技术和LDAP 技术,这两种技术的引入使网格的监控充分发挥了移动Agent 和LDAP 的优势。由于移动Agent 本身的特点,它可以克服网络延迟,而网格环境中需要对用户的请求做出实时响应,移动Agent 是由中央处理器派遣到系统局部,直接执行控制器的指令,从而消除网络延迟所带来的隐患。另外移动Agent 具有感知运行环境和对其变化做出自主反应的能力,多个Agent 可以拥有在网络上各主机之间合理分布的能力,以维持解决特定问题的最优配置。总之,移动Agent 在网格环境中充分发挥它的优点,解决了网格环境中一些可能出现的问题。LDAP 提供了一个访问控制的安全模型,根据身份信息对提出的访问请求进行控制。安全机制中一个重要的内容就是备份和恢复,在LDAP 中提供了主从备份服务器的机制,它方便和经济地实现了备份和恢复的功能。本文另外还深入研究了网格节点的模拟工具MicroGrid,并设计和实现了虚拟网格环境和多个网格节点,为今后网格技术的研究和实验打下了良好的基础。遗传算法是当今影响最广泛的进化计算方法之一。遗传算法模拟了达尔文的自然
【Abstract】 Grid is developed with Internet Technology. Originally, as a new computing mode, it is devoted to complex scientific computing, by combining all networks to a large supercomputer. The grid research is in progress. Grid is predominant in discovering, allocating and arranging all kinds of capacities for network i.e. computing service, software application service, network bandwidth, or storage space. Grid resource management is to locate certain capacity, harmonize and apply it, and in the meantime monitor its state. Compared to local resource management, the grid resource management spans many managing domains, which will bring heterologous problems when similar resources are configured and managed. Therefore, the initial work grid resource management concentrates on is to solve heterologous problems, such as to constitute standard resource management protocol, and standard resource and job requirement description mechanism. This dissertation concentrates on grid monitoring technology. Grid monitoring is to measure and display the state of grid components. All the components on application end must be monitored, including software, hardware of mainframe and network. Monitor can be used in many places, such as state checking, fault detection, performance negotiation and debug. Monitoring data can be used to process fault and check error, analyze and adjust performance, guide schedule, gather data thus to improve next performance, and inbreak detection. This dissertation has brought forward and designed a new method to monitor grid resources, in which mobile agent technology and LDAP technology are introduced considering the advantage of the two technologies. Mobile agent can overcome the network delay, thus it can be dispatched from central process unit to system local to run the command of control unit directly in order to avoid the hidden trouble brought by network delay for users’requirement must be responded in real time in grid environment. In addition, mobile agent is able to apperceive the running environment, and react to its change automatically. Multi-agent possesses the ability of distributing among the mainframes in the network, and can optimize configuration. In a word, mobile agent can exert its advantages and solve possible problems in grid environment. The other technology LDAP can provide a secure model to control access requirement according to identity information. One important character of security mechanism is to backup and restore. Principal and subordinate backup server mechanism in LDAP helps carry out the backup and restore function conveniently and economically. This dissertation has also studied the simulated tool of grid nodes, Microgrid, and designed virtual grid environment and many grid nodes. In addition, genetic algorithm, as one of the most
【Key words】 Grid computing; Grid monitoring; Mobile agent; Genetic algorithm;