【作者】 曾芳；

【导师】 周英飚；

【作者基本信息】 华中科技大学 ， 计算机软件与理论， 2007， 硕士

【摘要】 数据复制是分布式数据库中提高系统可用性和可靠性的一项重要技术。长期以来,数据对象均使用静态复制方式,数据复制计划由分布式数据库管理者指定,该计划是固定的,直到管理者执行手动重新指定。然而,如果数据对象的读-写模式发生动态的、不可预期的变化时,静态的复制计划将导致严重的性能问题。理想的复制计划要能够根据读-写模式的变化自动地进行适当调整。因此,动态数据复制技术是分布式数据库系统面临的一个新的课题。在主复制方法的基础上,引入时间段限制,扩展成了一种具有自适应性的动态数据复制方法。采用“读一个写所有”更新策略,并假设在大多数情况,下一段时间内的读-写模式,能根据其紧跟着的上一段时间内的读-写模式进行预测。采用这种方法,可以通过比较一个时间段后,数据对象的本地读次数和远程写次数的大小,来自动调整该数据对象的复制计划。此外,借鉴了Microsoft SQL Server的复制模型,设计了由发布者、订阅者、发布、订阅、项目和数据源共同组成的复制模型;同时引入了“拉”式(Pull)订阅模型结构,实现了由订阅者启动复制过程完成数据库到数据库的复制;采用触发器技术,监视发布者的数据更改,解决了数据更改的捕捉问题;采用流行的可扩展标识语言(eXtensible Markup Language,XML)技术(包括创建与解析)很好地解决了复制数据更改的网络传播问题。更多还原

【Abstract】 Data replication is an important technology which is often used to improve the usability and dependability of distributed database system. Traditionally, Data objects are replicated in static way, their replication schemes are assigned by database administrator; the schemes are static and unaltered during the replication process unless they are changed by administrator. However, if the read-write patterns change dynamically, in unpredictable ways, a static replication scheme may lead to severe performance problem. So, dynamic data replication is a new subject of distributed database system.In this paper, we present a distributed algorithm for dynamic data replication of an object in a distributed database system. The algorithm dynamically adapts the replication scheme of an object to the pattern of read-write requests in the distributed database system. It works in the read-one-write-all context, which means a file read involves only one replica and the write update is propagated to all the replica; the read-write pattern during a time period is in most cases predictable based on the read-write pattern in the immediately preceding time periods.Referred to replication model of Microsoft SQL Server, our replication model has been developed as six parts including publisher, subscriber, publication, subscription, data item and data source. The introduction of PULL subscribing structure fulfills bidirectional replication between database and database which is started by subscriber.In addition, the usage of trigger can monitor the data changes in publisher’s database, thus the problem of catching the data changes has been worked out. And with the popular XML techniques (creating and parsing), transmitting replication data among networks has been better resolved.更多还原