移动Agent技术在对等计算中的应用研究

【作者】 王晓华；

【导师】 高德远；

【作者基本信息】 西北工业大学 ， 计算机系统结构， 2006， 硕士

【摘要】 随着高性能计算的不断发展,传统的C/S计算模式已经不能满足处理海量数据的要求。只有通过分布式系统,协调利用网络上闲置的处理机资源,共同完成处理任务,才能实现真正的高性能计算和无所不在的计算。对等计算(P2P)作为分布式计算的一个分支,具有主机利用效率高、网络体系结构灵活等特点。对等网络中参与计算的主机在完全透明的情况下通过相互协作通信实现资源的收集、任务分派和信息发送等功能。 移动Agent能在异构的网络环境中从一台主机迁移到另一台主机并完成用户指定的任务。将移动Agent和对等计算二者结合起来,有利于减少计算过程中的通信量,提高通信效率,充分利用网络资源,动态适应网络环境的变化。 本文详细分析移动Agent和对等计算的基本原理,提出一个基于对等网络通过移动Agent来实现的计算平台——MAPS。MAPS基于Jxta平台设计的,与编程语言、硬件系统无关,是开放式的。为了开发通用、灵活、高效的Agent构件,必须将Agent的数据和功能分离。文中针对Agent在P2P网络中的应用,设计Agent的基本结构。根据对Agent安全性能的要求,文中对公开数据进行完整性验证,私密数据采用双重加密的措施。这样既能提高数据的安全性,又不会降低加密的效率。在动态分析对等网络中移动Agent的运行过程时,根据Agent的状态以及状态的迁移,建立Agent生命周期模型,研究Agent的相关行为。为了提高移动Agent系统的可靠性,本文设计了动态停靠主机机制,在移动Agent通信出现故障时,采用停靠主机来最大限度的保障数据、代码以及中间运行结果的完整性。双重命名机制保证移动Agent的精确寻址。 对等计算的关键性问题是对等主机之间的通信机制以及可用信息的查询算法。MAPS采用了与Jxta一致的通信方式,由主机向系统中发布通告。通告采用XML格式,其他主机获取通告,根据约定的形式进行通信。在查询信息时,借鉴目前几种查询方式的优点,提出一种适合在对等网络中采用的双重索引查询(DDI)。查询是根据结点之间的有用性进行的,并且有用性也与主机节点间的最佳路由有一定的关联。 MAPS将移动Agent技术、Jxta通信机制以及对等网络查询机制综合在一起,提供了一个无所不在的高性能计算的平台。更多还原

【Abstract】 Accompany with the rapid development of high performance computing, the traditional C/S mode cannot fit the demand of numerous data processing. The high performance computing and ubiquitous computing can be realized by distributed computing system. It utilizes the resources of processor on the Internet and completes the project each other. As a part of DCS, P2P computing has some properties, such as high efficiency of host and the agility of the structure of the network infrastructure. The computers involving the computing correspond with each other in the transparent environment;accomplish the works such as information collecting, jobs detachment and data transmission.Mobile Agent can complete the assignment successfully by migrating from one computer to another on the heterogeneous network. The quantity of communication will be reduced if we combine the technology of Agent and P2P computing, it also can improve the communication efficiency, makes use of network resources and adapt the change of the network environment dynamicly.By discussing the principles of Agent and P2P computing, this paper promotes aplatform--MAPS, which implementes via Mobile Agent based on P2P network.MAPS is designed on the Jxta platform. It is independent of software and middleware, and is open sources. For developing the common, agile and high efficient agent component, we must separate the data and function of Agent. It starts with analyzing the properties of agent, and designs the major structure of agent at the basis of appling of P2P computing. According to the requirement of the secrurity about Agent, it verifies the intergrality of the public data and adopts the double encryption mechanism to the private data. By this way, it can improve the security of the data without reducing the efficiency of the data encryption. At the time of analyzing the Agent’s action dynamicly, we construct the life cycle model according to the Agent’s status and transformations. For the sake of improving the reliability of Agent System, we design the dynamic docking host mechanisim, if the correspondence between two agents is in trouble, the dynamic docking host mechanism can keep the intergrality of data, code and themptoary results mostly. The Agent can be accurately accessed with the help of double name system.The major problem of P2P computing is the correspondence of P2P network andhow to carry out the information query. The correspondence mechanism of P2P computing is the same as Jxta. The hosts propagate the advertisements in system, which is writien by XML type, and the others catch the advertisements and correspond with the very host. Acoording to the query mechanisms complement on P2P network commonly, we promote a new mechanism name distributed searching with double index (DDI), which is compatibly applied on P2P network. The querying is based on the availiablity between two nodes, which have some relationships with the best route of two nodes.MAPS combines the technology of Agent, Jxta correspondece mechanism and P2P network quering. It prompts a platform, which can complete the high performance computing and can be realize everywhere.更多还原