提出了一种保证多volume数据一致性的远程复制机制。其借鉴数据库系统中事务处理的基本思想,将多个volume中相关联的更新作为一个原子事件向远程端复制,分析实现中如数据打包、故障恢复策略、I/O合并等关键问题,并给出了具体的实现方法。该远程复制机制解决了在基于存储层或基于逻辑卷方式下的大规模数据复制应用中,保证一组具有相关性数据在复制中的一致性和可用性问题。
This paper proposed a remote replication mechanism which could guarantee the consistency of data in muhi-volume by leveraging the regulations of transaction processing in the database management system. The system could take several correlated updates in multiple volumes at the primary site as an atomic event, and replicated it to the secondary site. In addition, the paper analyzed the key techniques in this mechanism, such as data packing, policy of fault recovery, I/O merging, and discussed the implementation. With this mechanism, the consistency and availability of data could be guaranteed in a group of volumes which had relevancy and the relevancy had to be maintained during the replication operation. This mechanism can be used to solve the mass data replication problem in a storage system at the block level (including physical and logical level).