【作者】 施晓烨；

【导师】 王汝传；

【作者基本信息】 南京邮电大学 ， 计算机软件与理论， 2011， 硕士

【摘要】 信息爆炸带来了对数据存储及访问速度的空前要求,存储规模越来越大,管理也越来越复杂,同时对存储的可扩展及可靠性也带来了更高的挑战。为了解决这些问题,数据网格应运而生。数据网格正是一个以数据为主要资源的网格系统,它将网络上海量的、分散的、独立的、异构的储存系统组织成一个可靠、安全的逻辑意义上的整体,进行统一的管理,从而为用户提供透明的、高效的、高可靠的服务。数据网格中的副本技术是必不可少的。副本技术主要包括:副本创建、副本选择、副本定位以及副本一致性维护。其中,副本创建的优劣将直接影响到网格的系统性能,因此必须结合环境特点在合适的节点上建立副本;在副本创建之后,则需要副本的选择和定位机制来获取最优副本;同时由于网格的动态性,副本一致性维护也是副本管理中的重要组成部分,因为这直接影响到副本管理的性能和正确性。本文针对数据网格中副本的关键技术,做了以下几方面的研究:首先,改进了传统的副本创建算法,提出了改进的最佳副本创建算法;其次,提出了适合当前环境的副本定位及一致性维护方法以及给出了基于副本访问代价的副本选择算法;最后,实现了副本管理系统。更多还原

【Abstract】 The information explosion has brought about an unprecedented demand in the speed of data storing and accessing. The storage scale is growing soaring larger, and management is becoming increasingly complex as well, all of these have caused greater challenges to the storage scalability and reliability. To solve all these difficulties, the data grid technology comes into being at this time. The data grid is a major resource based data-grid system that organizes the huge, decentralized, independent, heterogeneous storage system from the internet into a reliable, secure integration of logical meaning, and manages them unified, aiming at providing the users with transparent, efficient, highly reliable service.The replica technology of data grid is essential, such as: replica creation, replica selection, replica location, replica consistency maintenance. Firstly, among them, the quality of replica creation will directly affect the system performance of information grid, so a good foundation must be laid according to the characteristics of environment to create replica at the appropriate nodes. After the replica creation, it’s necessary to use the technology of replica selection and replica location to find the best replica. At the same time, replica consistency maintenance is also a copy of an important part of the replica management, which will directly impact on the performance and correctness of replica management.This thesis researches several aspects on key technologies of replica in the data grid: at first, we advance the traditional algorithm and propose the advanced best replica algorithm; secondly, accordingly to the environment, we propose suitable methods of replica location and replica consistency maintenance and raise the technology of replica selection based on replica cost. At last, we finish the replica management system.更多还原