教育资源网格中的副本管理策略研究

【作者】 姜浩然；

【导师】 刘方爱；

【作者基本信息】 山东师范大学 ， 计算机应用技术， 2012， 硕士

【摘要】 随着计算机技术的普及和数字化校园的发展，全国各地中小学都积累了大量的教育资源，包括精品课件，视频教程，实验报告等等。但是，不同的学校以自己的方式对这些资源进行管理，学校间的资源共享度不高，形成了一个个的信息孤岛。教育资源网格就是为了对教育资源进行有效共享而提出的，它将不同的学校看作是一个个的虚拟组织，以网格的思想对教育资源进行管理，为教育资源的有效共享提供了一个高效的平台，避免了资源浪费和软件的重复开发。作为数据网格的一种，副本管理技术在教育资源网格中同样起到了非常重要的作用。一方面，副本的存在有效提高了资源的可用性，避免了因节点离开网格而造成的资源不可用现象；另一方面，副本技术使得对于某个文件的请求分散到各个副本节点，而不是集中于同一个节点，在一定程度上均衡了网格负载；第三，副本技术能够通过在请求节点本地或附近节点建立副本的方式缩短数据传输的距离，提高数据访问效率。教育资源网格的独特特点也决定了在教育资源网格环境下研究副本管理技术具有非常重要的意义。本文针对教育资源网格环境中的副本管理问题，主要做了如下工作：首先，研究了教育资源网格环境和副本管理技术，阐述了教育资源网格环境下副本管理策略研究的意义。首先介绍了网格和数据网格的概念、特点、体系结构和研究现状等知识，并在此基础上介绍了教育资源网格独有的特点和体系结构，对研究环境作出了比较清晰的描述。接下来阐述了副本管理技术的概念和特点，对副本管理所包含的各个功能模块都进行了详细的介绍。其次，提出一种基于节点负载的副本创建策略。针对教育资源网格的存储处理能力低，带宽分布不均的特点，提出了一种基于节点负载的副本放置策略。该策略根据节点所在管理域的不同分成域内和域间副本请求两种情况，对于域内副本请求，副本在负载较小的节点进行创建，以平衡域内节点负载；而对域间副本请求，则在请求域的管理节点上创建副本，以减少副本远程传输的次数。实验表明，在教育资源网格环境下，该策略能够有效提高作业的平均执行时间，改善教育资源网格的性能。第三，提出一种基于模糊多属性决策的副本选择策略。该策略综合副本的可用性、安全性、响应时间和花费作为评价副本好坏的标准集，更加关注副本的服务质量。由于该标准集中各个属性很难进行定量分析和度量，本文使用三角模糊数来表示各个属性值，并且，各个属性的评价标准和物理量纲不同，甚至相互冲突和制约。为了使得副本选择更加客观和公正，在均衡各个属性值的基础上做出最优选择，本文使用多属性决策机制来解决副本选择问题，为副本选择策略提出了一种新的解决方案。最后本文通过实例说明了该策略的可行性。最后，对OptorSim模拟器进行了深入的研究，并对模拟器自带的经典算法进行了改进。本文在介绍OptorSim的体系结构，主要框架和类以及自带的几种经典副本管理算法的基础上，分析了OptorSim中自带的LRU和LFU算法的不足，提出了LRU和LFU算法的改进，并在OptorSim中进行了实现，得出了实验结果。通过与原算法的实验结果进行比较，证明了改进后的两种算法在性能上有明显的提高。更多还原

【Abstract】 With the popularization of computer technology and the development of digital campus,primary and middle schools of our country accumulate a large number of educational resources,including Excellent course, video tutorials, experimental reports and so on. However, differentschools manages these resources on their own way. Resource sharing among schools is not high,which forms a separated information island. Education resource grid is proposed in order toefficiently share educational resources. It views different schools as virtual organizations andmanage educational resources based on the grid, which provides an efficient platform for theeducation resource sharing and avoids the waste of resources and software redevelopment.As a kind of data grid, replica management technology also plays a very important role ineducation resource grid. On the one hand, the existence of replicas efficiently improvesavailability of resources and avoids the phenomenon that the resources are not available owing tothe nodes’leaving. On the other hand, replica technology makes the request for a certain file notconcentrate on the same node, which balances the network load to some extent. Thirdly, replicatechnology improves the efficiency of data access, the setting up of replica can reduce thedistance of data transmission. Unique characteristics of educational resources grid suggestresearching replica management technology has very important significance in educationalresource grid environment. According to the replica management problems existed in theeducation resource grid, the paper mainly do the following works.First, the paper studies educational resource grid environment and replica managementtechnology and elaborates the significance of educational resource grid replica managementstrategy. The paper introduces the concept,features,system structure and research status of gridand data grid and so on, analyzes unique characteristics and system structure of educationalresource grid and clearly describes the research environment. Then, the paper elaborates theconcepts and features of replica management and introduces the various functional modulesincluded in the copy management in detail.Second, the paper proposes a replica creation strategy based on the node load. Afterspecificly analyzing features of model of educational resources grid, the paper proposes a replicasetting strategy based on node load. According to the situation where the request node is, thestrategy is divided into replica request in domains and replica request among domains. In thesame domain, replica creation is carried out based on the nodes whose load is small to balancethe nodes load. However, for the quest among domains, replica is created on the managementnodes to reduce the price of remote transmission of replica. The experiments suggest that, in education resource grid model environment, the strategy can improve the average execution timeof jobs.Third, the paper proposes a replica selection strategy using fuzzy multi-attributedecision-making. The strategy utilizes availability, security, response time and cost as selectioncriteria between different replicas. In order to evaluate the attributes more convenience, we usetriangular fuzzy numbers to describe the attributes. The selection criteria(availability, security,response time and cost) are heterogeneous and cannot be aggregated with each other for theselection, furthermore, the criteria may contradict one against other. In order to select the replicawith fairness and satisfy the users, the paper considers the problem as a multi-attribute decisionmaking problem,which provides a new solution for replica selection.Finally, an example toillustrate the feasibility of the strategy .At last, the paper deeply studies the OptorSim simulator and improves the typical algorithmof the simulator. In order to verify the validation of replica creation strategy, the paper choosesOptorsim as experimental tools to deeply study. The architecture of OptorSim, main frame,theclass and several classical replica management algorithm are introduced. Through analyzing thedefect of LRU and LFU of OptorSim in realization, LRU and LFU are improved. Throughcomparing with the original algorithm, the experimental results suggest the performance of theimproved algorithm is raised.更多还原