RAID Explained

Written by IT Manager.

What the hell is Raid…and do i need it?

Raid arrays allow you to use multiple hard drives to boost performance or redundancy. Here are the flavors:

RAID 0 is all about performance, employing what’s called striping, where data is broken up into fragments and written across multiple drives, sort of treating them as one giant drive. Let’s assume we’ve got a setup with four hard disks. The performance edge comes from the fact you’re getting massive throughput—it’s like going from one lane to four, since you’re writing and accessing all four drives in parallel. It’s for pros and crazies handling massive files, like HD video editing. The downside is that if even one of the hard drives fails, you lose everything. Every file is now incomplete. It’s not technically RAID since there’s no redundancy going on—possibly hence the zero. You need at least two disks for this.

RAID 1 is the main configuration most novices should learn about. It writes, or mirrors, data to multiple disks, so you’ve got multiple hard drives that are exactly the same. Obviously, this is good for data reliability, since if one fails, you’ve got another. If you don’t have an independent disk controller or host adapter for disk, however, performance can be kind of crummy trying to write to the disks simultaneously, and performance isn’t going to be as good as a striped RAID configuration, obviously. On top of that, you have to buy two 500GB disks just to get 500GB of storage, so it gets a little costly, too. Safety first! You need at least two disks here.

RAID 2 stripes data like RAID 0, but at even smaller level (bits instead of blocks) and uses additional hard drives and what’s called Hamming code for error protection and parity which allows it to recover corrupt data. Guess what? No one uses it anymore, because it requires a ridiculous number of disks.

RAID 3 stripes data across multiple drives as well, but at the byte level, and it has a single disk dedicated to data parity and error correction. Because of the byte level split, all the drives work together simultaneously as one unit, which means it can only do one one read or write operation at a time. Pretty rare to see, and nothing you, Joe Q. Consumer have to worry about. It’s good for high transfer rates (again, HD video editing comes to mind) with a measure of security that you don’t get with RAID 0, since you can lose a disk and still be okay. You need at least three disks for this party.

RAID 4 is a striping+parity disk setup too, but at the larger block level, so disks can be more independent, and you can have multiple read operations in different places going on. Since you’re only using a single disk for parity, which has to be written to every time you write data, you can still only have one write operation going on at a time. Three’s the magic number of disks here too.

RAID 5 is where much of the NAS in a box action is today if you’re not rolling with RAID 1, and tries to offer the best of all RAID worlds—performance and redundancy—by combining the various configurations. It stripes data across multiple hard drives, but instead of just dumping parity data onto a single drive, it spreads it across all of the hard drives too, meaning there’s no bottleneck from writing parity data. (Though writing parity data is still kind of a drag.) In this configuration, you can lose one hard drive and be okay, since that drive’s parity data is on a different disk. Sadly, there’s some concern about its probability for failure over the next year as hard drives increase in size and the system expands. Three disks gets you in the door.

RAID 6 is like RAID 5, but it uses two disks for parity and correction which are setup so that if one of the drives bombs out during data recovery, the system keeps on chugging. Obviously, you need one more disk than RAID 5, making the minimum four.

• Beyond the standard RAID levels, you’ve also got the multiples: RAID 1+0 and 0+1. In RAID 0+1, you take a set of drives that striped in RAID 0 configuration (so they’re acting like one hard drive) and then you mirror them in a RAID 1 setup, so you get some redundancy. So if you’ve got a pair of RAID 0 drives, you’d be mirroring them, for a total of four drives. RAID 1+0 is the reverse: It stripes across a set of mirrored drives. With hard drives getting cheap, the army of hard drives you need to make this happen is easier to get going, so you might see more of it.

Just to be clear: RAID’s not a total substitute for backup. For average internet geek people, the best way to think about RAID is as a way to more reliably store a large bank of data (which we all have now with photos, videos, music and more) with extra protection against physical hard drive failure. You can still totally lose an entire RAID array via virus or accidentally overwriting it.