↝ OWC / MacSales.com... ↜
↝ diglloyd Deal Finder... ↜
Buy other stuff at Amazon.com...
Choosing RAID-0, RAID-5, RAID-6
This discussion assumes eight drives as with the 8-bay iStoragePro used for the testing for this article.
With eight drives involved, the chances of a drive failure increases greatly over systems with 1-4 drives. So if one large volume is created, use of some level of fault tolerance is all but mandatory, except for the foolhardy.
RAID-0 = striping, failure of any drive means data loss
RAID-5 = striping + parity, tolerates loss of one drive, at risk during 10-12 hour rebuild
RAID-6 = striping + dual parity, tolerates loss of two drives
With any of these RAID setups, keep a spare drive (“cold spare”) on hand at all times, even if there is a hot spare in use, so that a failed drive can be swapped out immediately with the cold spare.
In other words, purchase 9 drives for an 8 drive enclosure (see this handy shopping cart at OWC for all parts needed to put together a working storage setup).
Making a 5/6/7/8 drive RAID-0 carries a much higher risk of a drive failing than a 1/2/3/4 drive setup. So use RAID-0 only for data you can afford to restore from a backup. RAID-0 does offer maximum capacity (and in theory speed), but you must maintain up-to-date backups to preclude total data loss, or perhaps spend $10K - $20K to get your data back.
RAID-5 tolerates one drive failure
For a blend of high performance and reliability, RAID-5 makes sense, but it really should be used with a hot spare, so that when a drive fails the RAID can be restored to its RAID-5 state automatically by the RAID card (it’s degraded into a RAID-0 until that process occurs, no more failures can be tolerated).
In an 8-drive enclosure, use seven drives for the RAID, designating one drive as a hot spare.
When a drive failure occurs with RAID-5, data is not lost. Assuming a hot spare, a recovery process begins that can take about half a day with 2TB drives, longer if the array is being utilized. Within that 8 hour time span, a second drive failure means loss of the RAID (total data loss).
In the context of an 8-drive enclosure, it makes little sense to opt for RAID-5 with a hot spare, because RAID-6 with dual parity drives is more robust, and the performance is the same.
RAID-6 tolerates two drive failures
Because RAID-5 with a hot spare is still at risk during failure recovery (rebuild), using RAID-6 with dual parity is the smart move for sensitive data. RAID-6 uses two drives for double distributed parity. If one drive fails, rebuilding the bad drive (after it is physically replaced) can still tolerate loss of one more drive.
Consider a hot spare for RAID-6— there is modest difference between 10TB and 12TB of storage, and so if you really need 12TB, you really should be getting a second enclosure, rather than trying to cheat on reliability. Appropriate mini-SAS units like the iStoragePro can be daisy-chained.
Suggested setup for an 8-drive enclosure:
- RAID-6 using 7 drives yields usable capacity of 5 drives (2 drives used for parity);
- + 1 hot spare.
With eight 2TB drives, you get 10TB of usable capacity and a hot spare, and you can be out of town when a drive fails, with the system automatically rebuilding using the hot spare. Also have that 9th cold spare on hand, so that a new hot spare can be added to replace the failed drive.
Choosing a hard drive
I tested the iStoragePro using the fast 2TB Western Digital RE4 enterprise grade hard drives. The RE4 is my top choice for this type of use scenario, but the 2TB Hitachi A7K2000 could probably be substituted and it draws much less power when idle (about 5.1 watts vs 8.1 watts for the RE4). But the Hitachi A7K2000 performs about 10-15% slower.
The cost-effective Hitachi 7K2000 could also be used, but it’s not an enterprise-grade drive and it uses more power than the A7K2000, and I am unsure that this is appropriate for an 8-bay RAID system.
For fast and reliable external storage, skimping on drives is a contradiction in theory; a qualified enterprise-grade drive designed for RAID is the smart move. Or so it seems.
If you really need more than 12TB of storage, it’s appropriate to get a second enclosure rather than sacrifice reliability. Remember, restoring 10TB of data is going to take a long time, so keeping the RAID intact is a major priority. Enclosures can be daisy-chained; another RAID card is not required, so long as the enclosure is an “expander” model.
For the figures below, subtract 2TB for each hot spare you’d like to keep online (but RAID-0 does not benefit from a hot spare in terms of preserving data).
|Capacity using RAID 0/5/6|
|Number of 2TB drives||RAID 0 striping,
tolerates no failures
striping + parity,
allows 1 drive failure
striping + dual parity,
allows 2 drive failures
|RAID 6 + hot spare
striping + dual parity,
allows 2 drive failures