Dante_Stella
Rex canum cattorumque
It seems that a lot of RAID 5 solutions have been pushed onto the market - but do these really help?
I was looking at RAID units in conjunction with outfitting an iMac Retina 5K to have some kind of realistic storage. In looking at using a very big, very fast RAID box to store user profiles and about 5tb of data, I ended up with the following observations after looking hard at a Thunderbolt RAID array:
Am I crazy? My conclusion was to use a striped set for working files, a RAID NAS (not on all the time) to deal with backups, and a couple of clones in external housings (FW800/USB3/eSATA) for the real disasters.
Dante
I was looking at RAID units in conjunction with outfitting an iMac Retina 5K to have some kind of realistic storage. In looking at using a very big, very fast RAID box to store user profiles and about 5tb of data, I ended up with the following observations after looking hard at a Thunderbolt RAID array:
- The only redundant systems that don't put a huge hit on read speed are striped;
- Striped sets are very sensitive to drive issues (as I have learned firsthand...), and the more drives, the higher the likelihood of a failure;
- To overcome stripe deficiencies, you have to resort to mirroring (RAID 1+0) or parity (as in RAID 5 or 6), but either puts a big hit on read and write speeds;
- In a RAID 5 or 6 system, which is supposed to overcome the risks of striping, the rebuild times for data in the terabytes is monumental. It can take 36 hours to sync 3 x 3Tb drives because even the initial sync hits every block on the drive.
- If resynching occurs on failure, all of the drives in the array are running full tilt to reconstruct the failed drive - and in the interim, if thermal issues caused the original failure, you're inducing them again. And in RAID 5 particularly, a second drive failure takes the whole system out.
- Resynching a RAID also puts a days-long hit on performance.
- If you are not watching for a failure, and it happens, then the degraded array is running unprotected for as long as it takes you to detect and mitigate the issue.
- The "hot spare" concept is a great idea - except that the hot spare costs money, cuts capacity and is being cooked in the enclosure until it is called upon. At best, it is sitting still and baking; at worst, it is spinning and incurring wear and tear like the other drives.
Am I crazy? My conclusion was to use a striped set for working files, a RAID NAS (not on all the time) to deal with backups, and a couple of clones in external housings (FW800/USB3/eSATA) for the real disasters.
Dante