摘要:一般而言，读写分离技术可以解决当前大数据环境下的读写速度失配的部分问题，但是现有的读写分离技术主要是针对同构数据库的解决方案.由于存储结构的不一致，由行式存储数据库和列式存储数据库构成的异构分布式数据库系统相较于同构分布式数据库系统在数据同步的过程中就会面临格式转换、同步速度不匹配等诸多难题.提出了基于MySQL二进制日志（Binlog）进行SQL还原的方法TD-Reduction，设计并实现了Binlog解析器BinParser和Binlog还原器BinReducer，它们可以基于Mixed格式的Binlog，针对不同的事件（event）进行日志的解析，并依据相应的规则进行还原，生成可执行的SQL语句.综合以上技术，实现了一款分布式数据库数据同步工具Cynomys.在实验环境中，Cynomys表现出较好的性能.该方法适用于所有具有类似Binlog机制的其他异构数据库之间进行数据同步.



Abstract:In general, the read-write separation technology can solve some of the problems on mismatch between read and write in the current big data environment, but most of the current read-write separation technology are prepared for homogeneous database. Due to the inconsistent storage structure, heterogeneous distributed database systems composed of a row storage database and a columnar storage database will face many difficulties like format conversion and mismatch of synchronization speed in data synchronization compared to a homogeneous distributed database system. This study proposes the use of MySQL binary log to perform the TD-Reduction of SQL. It designs and implements Binlog parser BinParser and Binlog restorer BinReducer, which based on the mixed format. Different events perform log parsing and restoring according to the corresponding rules to generate executable SQL statements. Based on the above techniques, this study has implemented Cynomys, a distributed database data synchronization tool. In the experimental environment, Cynomys has shown sound performance. The method is suitable for data synchronization between all other heterogeneous databases with a similar mechanism like Binlog.



PDF全文下载地址:

http://jos.org.cn/jos/article/pdf/5694