↝ OWC / MacSales.com... ↜
↝ diglloyd Deal Finder... ↜
Buy other stuff at Amazon.com...
Performance as RAID-1 Mirror
The unit was tested in a RAID-1 mirror configuration, a fault tolerant mode where the two drives are identical and real-time copies of each other.
In RAID-1 mirror mode, one drive can fail with no loss of data or functionality (replace the failed drive ASAP so that the spare can be rebuilt to match the remaining drive).
Total capacity as a RAID-1 mirror with dual 3TB drives is of course 3TB, since each drive is identical to the other.
For professionals who cannot afford to be “down” due to drive failure, a RAID-1 mirror is a smart move. A choice of RAID-5 or RAID-6 is also good, but this is best done with units offering four bays for RAID-5, and six or eight drive bays for RAID-6.
Hardware vs software
One advantage of hardware RAID-1 mirror is that half the data is sent over the bus; with software RAID-1, data has to be sent to two “recipients” (each drive in the mirror). This distinction exists for writes only of course.
Mirror to single volumes
An unexpected plus of the OWC Mercury Elite Pro Dual: if the unit is switched from Mirror mode to Independent mode, the volumes are not destroyed; instead one gets two identical volumes on the desktop. Nice.
This has certain potentially useful applications: install two drives, set to Mirror mode. Then make a backup or similar, then set to Independent mode. Remove the drives and store separately. This is faster than making two backups onto two single drives. It would be interesting to see a dual-drive Voyager Q variant with this idea.
Performance as mirror
Average for entire capacity: 147, 155 write/read MB/sec
Performance as a mirror matches the speed of a single drive, which is as much as one can hope for for writes.
Read performance of a 2-drive RAID-1 mirror in theory ought to be nearly double the speed, but this is not happening*. But your author has yet to see any mirror solution that delivers the full promise of 2X performance for mirror reads, or even comes close, so in context this is not a negative relative to other solutions.
* DiskTester was used in a mode with 4MB writes and reads, so in theory mirror mode ought to be able to service a 4MB request by reading 2MB off each drive simultaneously, thus doubling read speed.