网格中数据复制的一致性研究  

Replica consistency of data in grid

作　　者：刘彩燕[1] 白尚旺[1] 

机构地区：[1]太原科技大学软件工程研究所,山西太原030024

出　　处：《计算机工程与设计》2006年第17期3163-3164,3177,共3页Computer Engineering and Design

基　　金：山西省自然科学基金项目(200209)。

摘　　要：数据是网格中的一类重要资源,它具有可移动、可复制、可缓存等特点。随着网格中的数据集在数量上和规模上的不断增长,为提高访问速度和减少访问延迟,需要把网格数据传输到距离访问者比较近的位置,并在不同的网格结点上复制和存储数据集副本,使数据访问服务达到更好的性能和可靠性。网格中存在异构数据资源,通讯延迟和资源失效等问题,因此确保副本数据的一致性是一项极具挑战性的任务。通过对基于网格中间件系统的复制一致性服务及其存在的问题进行分析后,有针对性地提出了一种体系结构设计方案。In Grids, data is one kind of important resources. It can be moved, replicated and cached. As datasets increase in scale and number, in order to improve access speed and reduce access latency, the grid files need to be transferred to the position that is close to the user, and these datasets are replicated. To achieve better performance and reliability, the copies ofdatasets are stored on different grid nodes. Guaranteeing consistency among replicated data objects in grids is a challenging task, given heterogeneous data resources, communication delays, and likelihood ofresource failures. The replica consistency service based on the middleware system is analysed and the main issues are summarized in the replica consistency problem. Furthermore, a high-level architectural design for a replica consistency service is laid out.

关 键 词：网格计算 GLOBUS 元数据 复制一致性服务 体系结构 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]