RAID explained ?!

[tomos]

RAID storage explained

by George Ou at TechRepublic. Link

RAID was originally defined as Redundant Array of Inexpensive Drives, but RAID setups were traditionally very expensive so the definition of “I” became Independent.  The costs have recently come down significantly because of commoditization and RAID features are now embedded on to most higher-end motherboards.  Storage RAIDs were primarily designed to improve fault tolerance, offer better performance, and easier storage management because it presents multiple hard drives as a single storage volume which simplifies storage management.  Before we start talking about the different RAID types, I’m going to define some basic concepts first.

-
-
I've been trying to get my head around RAID for a while now
I think this is a good article - in spite of the fact my eyes start glazing over after the first half page -
that has more to do with my capacity for technological info than anything else ...

[tomos]

RAID Level 6: RAID Level 6 is a cluster-level implementation of data striping with DUAL distributed parity for enhanced fault tolerance.  It’s very similar to RAID Level 5 but it uses the equivalent capacity of two hard drives to store parity.  RAID Level 6 is used in high-end RAID systems but it’s slowly becoming more common as technology becomes more commoditized.  Dual parity allows ANY two hard drives in the array to fail without data loss which is unique in all the basic RAID types.  If a drive fails in a RAID Level 5 array, you better hope there is a hot spare that will quickly restore the array to a healthy state in a few hours and you don’t get a second failure during that recovery time.  RAID Level 6 allows that second drive failure during recovery and is considered the ultimate RAID Level for fault tolerance.  Out of an array with “N” number of drives, the total capacity is equal to the sum of “N-2″ hard drives.  For example, an array with 8 equal sized hard drives will have the combined capacity of 6 hard drives.

Sounds good

[Hirudin]

I love that there is also "JABOD" (or is it "JBOD"?)! Just a Bunch of Drives! Haven't used it, but I guess it's like RAID except that it doesn't require the same capacity drives.

[f0dder]

Hirudin: yup, it just gives you "one big disk", linearly used, without the speed advantages of striping, and slightly less chance of losing all your data (although you will need a decent filesystem recovery app if a single disk dies).

RAID-6 is imho a good idea, since disks purchased at the same time tend to die around the same time. Especially if the problem is wear&tear + heat in a crowded server room, disk #2 might die from the added stress when the system tries to reconstruct the array with help of a spare drive. And then you're screwed.

[lanux128]

tomos, add this article to your list: http://en.wikipedia.org/wiki/RAID

[justice]

If you are thinking of setting up a RAID for your desktop think again:
see these two excellent articles from CodingHorror:
You Want a 10000 RPM Boot Drive
Desktop RAID: Oversold?

[MrCrispy]

The only situation in which I would ever use or recommend is RAID 5 with a hardware controller with parity calculation. Even then if your controller goes bad you have to get the exact same one because RAID arrays are not generally transferable. So software RAID might actually be preferable because its more fault tolerant, if cpu load is not an issue.

RAID was never meant for performance and I see custom pc builders using it for that all the time. ITs meant for data redundancy and IMO those needs are much better served with a judicious backup strategy, for which RAID is no substitute! I've seen many horror stories where people think RAID 1 is backup, then try to recover something they deleted on one drive

[tomos]

What most articles forget to say is (WTF) RAID is for and HOW IT WORKS (in the important sense! - how it works when things go wrong) -
or maybe the glaze factor caused me to miss that bit 

I think the idea is -
if your main HD dies or whatever you can continue working with one of the others?

Or,
is it simply that, again, if your main HD dies or whatever, you can replace it with one of the others, then buy a new one to replace that.

Or, do you have to buy a new drive & then copy "back" your contents?
[EDIT: all much of a muchness I suppose - if I've got the gist of it correct]

Anyways the general opinion here seems to be not too favourable so I wont loose any sleep over it

[tomos]

frantically reading 

If you are thinking of setting up a RAID for your desktop think again:
see these two excellent articles from CodingHorror:
You Want a 10000 RPM Boot Drive
Desktop RAID: Oversold?

Both very interesting - find myself more interested in what people use to build their machines though -
the glaze factor for reading bout RAID got too much 

as often the comments make for a very good - if sometimes contradictory read.
I reckon there's a few hundred there so I'll read at my leisure,
thanks
tom

[f0dder]

If you juggle around a lot with big files, trust me, a raid stripe can be invaluable. Try compressing/decompressing a big file from one harddrive to the same harddrive - then try the same with two separate harddrive, or a raid stripe.

Fortunately single-drive performance has been getting a lot better the last few years, but if you need high sustained write rates, it's hard to beat stripes. And while they don't reduce seek time in my experience, they do help against disk thrashing when doing multiple things on the same disk.

Mirroring (and raid-5 and raid-6 etc) is not a substitute for backups, and only retards use it that way. They're intended to protect against disk crashes before the nightly backup, so you don't lose a day's worth of work. And they're intended to be able to keep your data available while hotplugging a spare disk and replicating your array, as opposed to shutting down the server, replacing disks, and restoring from (last night's) backups.

I've got a 10k rpm raptor drive as my system disk, where I keep windows and my apps, et cetera. Low seek time = love. And the disk has very good sustained read and write speeds as well. Then I have two 160gig so-so performing maxtor disks, which are supposed to be in mirror for all my data (which I don't back up near often enough >_<), but currently they're forming a (filled >_<) 320gig stripe, because I wanted to play around a bit.

Btw, software raid isn't all that bad - for consumer on-board raid, lots of stuff is being done in software by drivers, and performance is still decent. It's only if you want big and really fast arrays that hardware solutions are really necessary. And if you want big arrays, you're probably doing it for a file server, that probably won't be spending CPU cycles on much else... so the hardware solution is really only if you build a weak system, or if you go all the way with battery-backed ram on the controller for safe caching

[tomos]

f0dder,
in spite of the fact you're speaking a very different language to me -
and,
I'm not talking Danish here 

You have, I believe summarised the reason for raid very nicely
.

Mirroring (and raid-5 and raid-6 etc) is not a substitute for backups, and only retards use it that way. They're intended to protect against disk crashes before the nightly backup, so you don't lose a day's worth of work. And they're intended to be able to keep your data available while hotplugging a spare disk and replicating your array, as opposed to shutting down the server, replacing disks, and restoring from (last night's) backups.

.
and also helped me realise I don't really need that kind of thing -
and I don't think I will either ..
(although I may have to "juggle around a lot with (very) big files" in the near future & the prospect has been bothering me, but more in the sense of doing incremental backups ... but I think I can always burn older backups to DVD if necessary)

But, again,
it's interesting to hear about the hardware setups people are using ...

[Lashiec]

Btw, software raid isn't all that bad - for consumer on-board raid, lots of stuff is being done in software by drivers, and performance is still decent. It's only if you want big and really fast arrays that hardware solutions are really necessary. And if you want big arrays, you're probably doing it for a file server, that probably won't be spending CPU cycles on much else... so the hardware solution is really only if you build a weak system, or if you go all the way with battery-backed ram on the controller for safe caching

The problem is not performance, but security and reliability. Adaptec is not selling RAID controllers at 200 € just to make us poorer, you know.