SQL 2005 中提出了数据库镜像的概念。
这个跟 Clustering，Replication还是有区别的。
mirroring is at a database level, has fast failover and duplicates all hardware and storage. Clustering has shared storage, duplicates the entire instance, and has longer failover times. Consider replication for scaleout reporting.
有个技术聊天关于 Data Mirroring 的。
http://www.microsoft.com/technet/community/chats/trans/sql/sql_102704.mspx
看来ms 在数据库的高可用性方面确实有很多新的思想加进来。