OWC Mercury RAID

I think we're getting a little esoteric here, but the term RAID is introduced and defined in the Berkeley paper "A Case for Redundant Arrays of Inexpensive Disks (RAID)" (1988)[http://www-2.cs.cmu.edu/~garth/RAIDpaper/Patterson88.pdf]. In it, Patterson, Gibson, and Katz define RAID as redundant (hence the acronym) and propose it as an antidote to the greatly reduced MTBF caused by non-parity stiped set and JBOD setups implemented to increase throughput. In other words, RAID was created specifically to solve the problems caused by "RAID 0" striping. Calling striped sets "RAID" is like calling the flu cough syrup. If anything confuses people--and leads to inexperienced users losing data they thought was safe--it's calling striped sets and JBODs "RAID."

Here's a damn good explanation of RAID http://en.wikipedia.org/wiki/Redundant_array_of_independent_disks#RAID_0

Jay, I'm not denying that. But RAID 0 /is/ called RAID, even though it is indeed not fault-tolerant and the likelyhood of failure only increases. It isn't meant for data protection-- it's meant to maximize storage space and performance. /You/ may call it 'pseudo-RAID', but that doesn't make it so. Please, go with what the rest of the world calls it instead of making up names that just confuse people.

CyBeR, I respectfully disagree, and I think you'll find most references agree with me. RAID stands for Redundant Array of Inexpensive Disks. That is, all RAIDs have parity with a greater or lesser degree of fault tolerance, the tolerance increasing with the parity for each stripe. "RAID" 0 is simply AID: and array of inexpensive disks with stiping, but no redundancy. We call it RAID 0 because it's convenient to lump all striping schemes together, but it isn't properly a RAID at all. Personally, I object to the term because many people believe, rightly so, that RAIDs protect their data, and discover too late to their dismay that "RAID" 0 doesn't. RAID MTBF should scale with the number of parity stripes, or in simple setups like mirroring with the number of disks. In "RAID" 0, however, MTBF actually increases the likelyhood of failure, with MTBF decreasing with the number of physical volumes, and destroys all of your data if a single disk fails. So I stand by my charaterization of "RAID" 0 as a pseudo-RAID, and my statement that RAID 1, mirroring, is the only RAID than can be implemented with less than three physical disks.

"a true RAID 1 (the only RAID you can implement on two disks)" Wrong, though you half-correct yourself. You can run both RAID 0 (Striping) and RAID 1 (mirroring) on two disks. RAID isn't necessarily about fault-tolerance, so there's nothing 'pseudo' about running RAID 0 on two disks. Not that I recommend doing RAID 0 at all..