节点文献
基于发布/订阅的网格数据库实时复制技术及实现
Techniques and Implementation of Publish/Subscribe-based Real-Time Replicate of Grid Database
【作者】 高卫中;
【导师】 程学旗;
【作者基本信息】 中国科学院研究生院(计算技术研究所) , 计算机应用技术, 2005, 硕士
【摘要】 随着网格技术和计算环境的发展,网格计算正从科学计算领域推向广阔的企业应用领域。数据网格中的网格数据库技术是企业应用中的一个重要研究课题。本文以一个企业的实际应用为背景,在OGSA-DAI4.0体系结构的基础上,开发设计了基于发布/订阅的网格数据库实时复制系统。系统从总体构架上采用了一种分片处理技术,通过把复制任务拆分成数据发布、数据监控、数据分发、数据导入等不同的子任务,并分配在不同的服务器上并行执行,有效地减少了数据访问和计算时间,达到了异构系统间实时数据加工复制的目的。系统的关键技术有:1)基于事务的分片数据更新机制;2)基于数据库的缓存机制;3)基于存储过程的分片计算技术;4)基于时间戳、启动阀值和信息通知的数据正确性保障机制;5)数据资源虚拟化机制;6)基于OGSA-DAI的数据访问、集成和传输机制;7)基于GT3信息通知的网格服务实时监控机制。经测试基于发布/订阅的复制系统比基于Oracle数据链路的Pull方式的数据复制系统具有更高的效率和更好的实时性,并满足了企业实际应用需要。
【Abstract】 With the development of grid technology and computing environment, grid computing is being pushed into the area of enterprise application from the area of science computing. It is an important area in the enterprise applications, the technology of grid database in the data grid.According to one practical application of an enterprise, we design the real-time data replicate system of grid database, which bases on OGSA-DAI4.0 and publish/subscribe model. We reduce the time of data access and compute by dividing the replicate task into subtasks,such as data publish, data monitor, data distribute and data update, then put them into different servers, so these subtasks can run parallel. In a word, the system uses the technology of splitting process.The key technologies include: 1) data-partition updating mechanism based on transaction; 2) cache mechanism based on database; 3) splitting compute technology based on stored procedure; 4) correct guarantee mechanism based on timestamp,threshold and notification; 5) data resource virtualization mechanism; 6) data access integration and transmit mechanism based on OGSA-DAI; 7) real-time monitor mechanism based on notification of GT3.Test results demonstrate that the system that bases on the Publish/Subscribe has higher efficiency and good real time property against the pull-method system that bases on the Oracle data linkage, and satisfies the requires of the enterprise.
【Key words】 Data Grid; Data Replicate; Publish/Subscribe; OGSA-DAI; Data Partition;
- 【网络出版投稿人】 中国科学院研究生院(计算技术研究所) 【网络出版年期】2007年 02期
- 【分类号】TP311.13
- 【被引频次】2
- 【下载频次】247