RAID 1 is optimal for redundancy. For an array of n drives, you have n-1 copies of the data, and can lose up to n-1 disks before you lose any actual data. Read performance is amazing, write performance is the same as 1 disk. Capacity is the same as 1 disk.
RAID 3 is not particularly optimal for anything. At minimum, you want to use RAID 5 (RAID 4 is RAID 3 with improved performance, and RAID 5 is RAID 4 with improved performance). RAID 5 does a little bit of everything. For an array of n drives, you have n-1 capacity, 1 copy of data (you can lose any 1 drive and still have all your data). Read performance is roughly equivalent to n-1 disks, and write performance is between 1 disk and n/2 disks.
RAID 6 is RAID 5, but with 2 parity disks, so you have 2 copies, n-2 storage, n-2 read speed, and about write speed of 1 disk.
RAID 0 provides optimal performance and capacity at the cost of redundancy. n Capacity, n read, and n write, and 1/n redundancy (if any drive dies, the whole array is gone.
For a backup system, I would recommend RAID 1, which prioritizes redundancy. I use RAID 5 for data that I can replace, but it would be a pain to, for which I need large storage capacity (read, my media fileserver). I run RAID 0 for my games - If a drive dies, I can simply reinstall games from their disks or Steam, and I want maximum performance from this disk to reduce loading times.
RAID is completely transparent, and a RAID 1 system with 2 drives is similar to your setup now, except the system would keep them perfectly in sync at all times. Any RAID array can have any number of drives, though some require certain multiples (RAID 5 needs at least 3 drives; RAID 1+0 or 0+1 needs at least 4 and must be even, and everything else needs at least 2). The entire array will show up as a single, large drive. If any drive fails within the arrays redundancy limits, then the array will continue to function fully. As soon as you replace the failed drive, it will restore the redundancy as quickly as it can copy data to the new disk; you can use the array while it is rebuilding or in this degraded state. If you lose more drives than the array can handle, then the array will fail and disappear. This would be similar to if your main 500GB hard drive and your 1TB backup drive failed at the same time.
@Darth Android: +1. Very informative and logical. To summarize, for my clarity, you state that a RAID1 set up with n external drives will provide the same security (n-1, and perhaps more speed) than backing up to 2 drive with, for example, Retrospect (which I have). Will the RAID setup give me a "mirror image" which I can use to restore my operating system, etc.? – Xavierjazz – 2010-10-08T01:59:26.537
1RAID 1 will provide bit-for-bit exact mirrors. You can pull any drive out of the array and slap it in another computer, which will boot up just fine (Windows might complain of an invalid license key in this exact scenario). RAID protects against drive failure, but is not a replacement for manual backups - those protect against accidental deletions, and other scenarios that you need to roll back files. – Darth Android – 2010-10-08T03:15:42.917
I can't tell if you're confused about how RAID drives are attached; RAID devices should be internal, not external harddrives because all drives in the array must be attached to the computer at all times. If a drive is removed at any point while the system is online, the RAID controller will immediately mark the disk as failed, requiring a full rebuild as soon as it or another working disk is attached and added back into the array. – Darth Android – 2010-10-08T03:27:34.150
The requirement that RAID drives need to be continually attached does not preclude that they may be external. This is a home use system so I can keep anything attached as I want. So far all of your information has been helpful, thanks. Regards, – Xavierjazz – 2010-10-08T03:39:29.060
After reading, once again, your help, I think this is what I see: "Backup" systems back up a singular entity (for example, my C/ drive), whereas a RAID system basically IS the singular entity: there is no "backup", just redundancy. i hope this is clear. Am I correct? Regards, – Xavierjazz – 2010-10-08T03:43:31.783
@Darth Android: I am not concerned with "... accidental deletions, and other scenarios that you need to roll back files". Again, +1. Regards, – Xavierjazz – 2010-10-08T03:44:58.943
1@Xavierjazz You are correct in that a RAID array is a singular entity, not a place to where entities are backed up. The idea is whenever you write a byte to the virtual drive, the RAID controller actually writes it to both disks. When you read a byte, the controller reads it from whichever disk is least busy. When the RAID controller cannot read from a disk (i.e., the disk crashed), it will transparently read the byte from the other disk and notify you that the disk has failed. Windows will continue to operate just fine during this. – Darth Android – 2010-10-08T03:50:20.587
I would just point out, since it's gone unmentioned in this thread, that while there are ways to RAID external disks it's a huge pain and will almost certainly require additional hardware. You're much better off using internal ones. – Shinrai – 2010-10-08T14:53:32.747