Question : Problem: Stripe RAIDs, data recovery and imaging

Hello,

My question relates to using RAID 0 on a boot volume. The performance benefits are attractive, but I am confused about how the data is arranged between the separate drives. If I understand correctly, data is distributed among the drives which allows faster access because more than one drive is responsible for retrieving the data. I also understand that if one of these drives goes bad that you lose all your data. Does this mean that file recovery tools such as OnTrack EasyRecovery (http://www.ontrack.com/easyrecoveryprofessional/) would not be able to retrieve files from the non failed drive because there would be no complete files to recover? What if the MBR had become corrupt but the two drives were still functional?

I realize that regular file backup is a guarantee that file recovery would not be an issue, I am mostly just curious how this would work. Likewise, I understand that some Imaging tools (Ghost, Acronis) etc make various claims to support dynamic disks, but does anyone have any real world success or lack thereof regarding imaging RAIDs?  

Answer : Problem: Stripe RAIDs, data recovery and imaging

It sounds like you are talking about RAID 0 across two different drives.  The two comments already posted tell you about how it works with striping.  If you are on two drives, and want a performance boost as well as redundancy, take a look at RAID 1 which is MIRRORING.  Essentially you are keeping two copies of the data.  If one goes bad, it switches to the good copy and you are still up and running.  The two most used RAID types (for redundancy and performance) are RAID 1 and RAID 5.  They are the most used for a reason, so if you are looking into redundancy, look at these two.  Keep in mind that when a disk goes bad, you may be still up and running, but performance suffers.  In the case of the two drives in a RAID 1 MIRROR, you can read from both drives at the same time (essentially doubling your READ ability), but when one dies, you are at half that speed because you can only read from one drive now.  When you replace the bad drive with a good one, you lose even more speed, as that second drive has to rebuild itself.  While it is rebuilding, it uses up system and I/O resources.  The idea is that you would rather be up and running a bit slow, rather than be down all together.

If you want to look at very high redundancy and performance, look to RAID 1 + 0.  Essentially this is a mirror that is stripped across disks.  You get the speed benefits of mirroring AND striping.  You could potentially lose multiple disks (as long as they are not the primary and its mirror).  For example, you have drives 1A, 1B, 2A, 2B, 3A, 3B.  Drive 1A and 1B are mirrors of each other, and the data is striped across drives 1, 2, and 3.  You could lose up to three drives before failing, as long as two of the drives aren't in a combination of 1A and 1B.  There are also systems that have something called a HOT SPARE.  This HOT SPARE will automatically replace one of the failed drives.  So if you lose drive 1A in the above example, the HOT SPARE rebuilds the data on itself, and removes drive 1A from the RAID.  Eventually you will need to replace the broken drive 1A, but this is another redundancy option that larger datacenters tend to use.  As a matter of fact, in my datacenter, we use RAID 1+0 across sets of 16 disks with 2 HOT SPARES.  Upper management is paranoid about redundancy, so I gave them what they wanted, albeit for a very high price tag.

Good Luck!