Disk replication (or disk mirroring) systems replicate the contents of a disk volume or file system from a primary site to a standby site.
The main benefits of disk replication are:
Redundancy of primary disk devices at a remote standby site
Zero or near-zero data loss, particularly with synchronous replication
The benefits of disk replication come at a relatively high cost, and there are significant disadvantages and gaps in the protection that disk replication provides:
A substantial cost of disk replication results from forcing standby system resources to go offline for periodic updates.
Disk replication requires exclusive write control of mirror devices at the standby site, which conflicts with the DBMS requirement for exclusive device write control, even when client applications cannot change data.
To avoid keeping standby systems completely idle or offline until a failover, local disk replication must be implemented at the standby site to make periodic “snapshots” of mirror devices available to the standby systems.
Standby system resources are hardware-dependent.
Disk replication is device-based, sending data from a primary device in blocks (or pages) to a mirror device. Hardware and operating systems at the primary site must be duplicated at the standby site, and they must be identical to the primary site configuration.
Data integrity (transactional consistency) between primary and standby databases cannot be guaranteed.
Disk replication updates mirror devices by block (or page) boundaries, with no knowledge of transaction boundaries. Because a single transaction can change several blocks, on several devices, the standby database can be corrupted if transmission from the primary site is interrupted before all of the affected blocks (on all of the affected devices) are received at the standby site.
Disk replication provides no protection from disk corruptions at the primary site, which are replicated block-for-block to mirror devices at the standby site.
Disk replication requires high network bandwidth for synchronous replication with acceptable application response time, particularly for databases with high transaction volumes.
