The main cost of using mirrored protection is in additional hardware. To achieve high availability and prevent data loss when a disk unit fails, you need mirrored protection for all the disk pools. This normally requires twice as many disk units. If you want continuous operation and prevention of data loss when a disk unit, I/O adapter, or I/O processor fails, you need duplicate I/O adapter and I/O processors. A model upgrade can be done to get nearly continuous operation and to prevent data loss when any of these failures occur, as well as the failure of a bus. If bus 1 fails, the system cannot continue to operate. Because bus failures are rare, and bus-level protection is not significantly greater than I/O processor-level protection, you may not find a model upgrade to be cost-effective for your protection needs.
Mirrored protection has a minimal effect on performance. If the buses, I/O processors, and I/O adapter are no more heavily loaded on a system with mirrored protection than they are on an equivalent system without mirrored protection, then the performance of the two systems should be approximately the same.
In deciding whether to use mirrored protection on your system, you must evaluate the cost of potential downtime against the cost of additional hardware, over the life of the system. The additional cost in performance or system complexity is typically negligible. You should also consider other availability and recovery alternatives, such as device parity protection. Mirrored protection normally requires twice as many storage units. For concurrent maintenance and higher availability on systems with mirrored protection, other disk-related hardware may be required.
Although mirrored protection can keep the system available after disk-related hardware failures occur, it is not a replacement for save procedures. There can be multiple types of disk-related hardware failures, or disasters (such as flood or sabotage) that require backup media.
Mirrored protection cannot keep your system available if the remaining storage unit in the mirrored pair fails before the first failing storage unit is repaired and mirrored protection is resumed. If two failed storage units are in different mirrored pairs, the system is still available and normal mirrored protection recovery is done because the mirrored pairs are not dependent on each other for recovery. If a second storage unit of the same mirrored pair fails, the failure may not result in a data loss. If the failure is limited to the disk electronics, or if the service representative can successfully use the save disk unit data function to recover all of the data, no data is lost.
If both storage units in a mirrored pair fail causing data loss, the entire disk pool is lost and all units in the disk pool are cleared. You must be prepared to restore your disk pool from the backup media and apply any journal changes.
When starting the mirrored protection operation, objects that are created on a preferred unit may be moved to another unit. The preferred unit may no longer exist after mirror protection is started.