Home » Opinions » The Rules of RAID

Introduction - Part One

by | Go to comments


Most features about RAID tell you what RAID is and then help you to decide whether or not you need RAID in your PC or workstation. We could easily follow the same path and do no more than quickly run through the different types of RAID. But then we're going to throw out the rulebook and throw a spanner in the works as, to paraphrase our Prime Minister, "The rules have changed", though he admittedly wasn't talking about PCs. Yes, we've going to talk about each type of RAID but we're going to follow this with some exhaustive test results to back things up.

RAID (Redundant Array of Inexpensive Disks) is a way of ganging up a number of hard drives to either increase performance, increase reliability, or preferably, both. Referring to RAID 0 as RAID, is something of a misnomer as though it stripes data across two or more drives to increase performance it doesn't actually provide any redundancy. If a drive or the RAID controller fails then the array is broken beyond repair and your data is lost. Assuming this doesn't happen though, performance will be good as the controller is usually able to repeatedly read and write to the cache on each drive in turn.

RAID 1 mirrors data from one drive to another so you get full redundancy but there is a trade-off. Write time increases compared to a regular PC with a single drive as the controller has to write to two drives. Read times often improve however, as Windows has two drives with the same data to choose from.


RAID 10, more commonly known as RAID 1+0, uses four drives with one pair of striped drives mirrored onto another pair of striped drives. This retains the speed of striping yet also gives you redundancy as it's tolerant of one drive failing. If you're lucky, it will even tolerate the failure of two drives. On paper RAID 10 looks good until you consider that despite having to accommodate four drives, you only get the capacity of two.

The last form of RAID that interests us is RAID 5. This uses at least three drives with data striped between two of them and parity data written to the third drive. In RAID 3 the relationships are fixed but RAID 5 rotates the parity information. This means that though you lose the capacity of one of the drives the array can tolerate the failure of any one drive. In addition, having all drives share the burden of dealing with parity information increases overall reliability as no single drive has to work twice as hard to deal with parity from the other drives in the array. In the event of a failure you can replace the faulty drive and the controller will rebuild the array over a period of some hours without any loss of data, although there is a risk of data loss while the array is degraded as it has no redundancy during this time.

RAID 6 takes this one stage further by designating two drives as parity drives, so you lose the capacity of two drives, but if one drive fails then the array degrades to RAID 5 until you rebuild it so your safety net is in place at all times.

Go to comments
comments powered by Disqus