Originally posted by Tom M
Nothing left to say on topic.
Well there is, since no one has adequately explained the different levels of RAID storage...
Raid 0 (Span Set): 2 - n disks. Combines multiple disks into one large volume, 'spanning' all the disks. Volume size is the total size of all disks. The volume can span disks of different sizes. Has the advantage of increased speed, but there's no fault tolerance. If a single disk dies, you lose ALL the data on the whole array. This is good for people who need simply massive volumes, although given that you can now buy 160GB single disks, I would say you're reasoning for installing this kind of array was shaky.
Raid 1 (Mirror Set): 2 Disks. Combines two disks into one volume which has the size of the smaller of the two disks, so is usually half the total disk size. Each disk has an exact copy of the data, hence the data is 'mirrored' across the disks. Provides elementary fault tolerance, in that if one drive fails, you have a complete copy of the data, although there is no tolerance to the second disk failure. When you replace the failed disk, the mirror rebuilds.
Raid 2 (Mirror Span Set): 4 - n disks. Combines Raid 0 and 1, using two mirrored span sets. Highly inefficient.
Raid 3 (Stripe Set): 3 - n Disks. Data is not written to a single disk, as with lower-numbered RAID types, but is instead split across multiple disks. Hence if you wish to store the file "ABC" across a three-disk array, you may well get "A" written to disk 1, "B" written to disk 2, and "C" written to disk 3. Volume size is the total size of all disks. Provides the speed advantage of RAID 0, but adds a small amount of fault tolerance. In RAID 0, if ANY of the disks dies, ALL the data is lost. In Raid 3, only the data on the dead disk is lost.
Raid 4 (Stripe Set with Parity): 3 - n Disks. Data is striped as in RAID 3, but a second piece of information is written to a separate disk. So, if you wish to write the numbers 2, 3 and 4 to a 4-disk array, you would find that "2" was written to disk A, "3" to disk B, and "4" to disk C. Additionally, "9" (2 + 3 + 4) would be written to disk D. Therefore, volume space is 36n-36, where 'n' is the number of disks, and 36 is the capacity in GB of the smallest disk. The beauty of this is that if you lose Disk A, you can rebuild it from the contents of B, C and D. Also, you can expand the volume, by adding disks. The parity information doesn't take up much space, so you can add more volumes to the Data set without having to expand the Parity set. However, you CAN add to the parity set too, thus your theoretical volume size is infinite. The downside is that you can't lose more than one of the disks. This is the technology that the NetApp Network-Attached Storage system uses, on which you can create single volume of up to 6 terabytes.
Raid 5 (Stripe Set with Striped Parity): As with Raid 4, only the parity information is also striped across all the disks. In this way, you can, incredibly, lose up to 50% of the disks in an array, and still rebuild it without losing data (although it will run like a dog!).
Raid 10 (Mirrored Stripe Set with Striped Parity): A Raid 10 array is simply two mirrored Raid 5 arrays (so it's Raid 5 + 1). This is really only for people who have loads of money AND who ABSOLUTELY CANNOT lose any data. Obviously, since at least 3 disks will be completely redundant, it's not an economic solution unless data integrity is paramount. I've only ever seen one solution (from EMC2) which offers Raid 10, and that was in the region of £300,000 for a 2TB array!