一种基于物理级的关系数据库数据复制模型研究

针对当前逻辑级数据库数据复制技术中存在的问题，本文提出一种基于物理级的数据库数据复制模型。该模型利用一个主备数据库复制组来确立其参照关系，通过对Oracle数据库在文件系统格式下、裸设备下以及对基于数据库特殊卷管理器格式下的Redo日志的解析，在日志变化的第一时间挖掘日志捕获数据变化片段，之后采用弱滚动校验和同步增量方式将数据片段并行同步到备端。备端进行日志片段的重组，修改日志头标志位，合成日志文件并与源端校验通过后注册到灾备库，然后将合成日志以自动或者手动方式写入灾备库，从而实现物理级数据库数据复制。通过测试对比分析，验证了该模型不但可以保证两端数据一致性，而且还可以降低数据复制延时。

Research on Relational Database Data Block Replication Model

 In view of the problems existing in the current logic level database replication technology, this paper presents a model of database data replication based on physical level. The model uses a database replication group to establish its reference relationship. By analyzing the Oracle log data in the file system format, the raw device and the redo log based on the database, the data fragments are synchronized to the end of the data segment. By end of log fragment reorganization, the log head sign is modified a synthetic log file and source side validation are registered to disaster libraries, and synthetic log is automatically or manually written to disaster libraries, so as to realize the physical database replication. By comparing and analyzing the test data, the model can not only ensure the data consistency, but also reduce the time delay of data replication.