Can someone explain....

Any off topic discussions should go in this forum. Post count is not increased by posting here.
FTP Access status is required to post in this forum. Find out how to get it
Forum rules
Any off topic discussions should go in this forum. Post count is not increased by posting here.
FTP Access status is required to post in this forum. Find out how to get it
Post Reply
Gnome
User avatar
Staff
Posts: 2671
Joined: Thu Nov 30, 2006 6:10 pm
Location: Land of The Tea.
Contact:

Can someone explain....

Post by Gnome »

....what raid actually does and its benefits as im thinking about putting my 2 sata drives in my server to raid.
I just don't get it lol.

Andy
User avatar
Administrator
Posts: 12629
Joined: Fri Aug 18, 2006 11:47 am
Location: United Kingdom
Contact:

Post by Andy »

A quick google will have found you plenty of information but here goes...


RAID is a redundancy solution in most cases.

For example:

RAID1 - Mirroring
Two identical sized drives (or two sizes but max usable space is that of the smallest) are used in a RAID1 configuration. You see the space of just one drive and everything is mirrored to the second. If one disk fails, the other can take over until a new drive is put in to replace the failed drive, and the raid is rebuilt.




RAID is also used for speeding up the read/write speed of the drives.

For example:

RAID0 - Striping
Two or more drives of any size are used in a RAID0 configuration. You see the space of all the drives in series (EG Drive1+Drive2's space. 80gb+80gb=160gb). The data is stored on these drives equally. If one drive fails, you lose the data on both drives.




RAID can also provide redundancy in that if one drive fails, you still have access to all your data.

For example:

RAID5 - Striping with parity
Any number of drives are used in a RAID5 configuration. You see the space of N-1 drives. (EG, 5x1TB drives will show as 4TB). The data is stored the same as RAID0 but the difference is parity data is added (hence the N-1). This means if one drive fails, you still have all of your data but your redundancy is now gone. Replace the failed drive with a new one, rebuild the array and your redundancy is back. If more than one drive fails before the redundancy is restored, you lose everything.

RAID6 is as above but is N-2, so up to two drives can fail before you lose everything.




You can also nest levels of RAID together (see http://en.wikipedia.org/wiki/Redundant_ ... ted_levels ).

Hope this explains it enough.

mrpijey
User avatar
Administrator
Posts: 8239
Joined: Tue Feb 12, 2008 5:28 pm
Contact:

Post by mrpijey »

Don't forget tho:

RAID is not a replacement for backup!

Now once that has been said, lemme explain... RAID doesn't protect you from data corruption other than such caused by a failure of the mechanical drive. So if you get a virus, accidently wipe data, or get a controller failure that writes bad data, then you're toast anyway.

RAID is good if you want to have very large volumes (spanning), if you want to have the same data to two drives (in case you want that "protection") (mirroring), or if you want to add parity data to any of that (like RAID5, 6 or any combination).

Drives in spanning mode (x*y, x being storage space per drive, y being amount of drives) tends to write and read faster (since both drives are used at the same time), but access time is slower (two drives or more has to be accessed and synced). This doesn't have any redundancy level, so if you lose one drive you lose ALL data on all drives. There is also a mode of this called "JBOD" (Just a Bunch Of Drives) where you can use drives of different sizes and chain them together into one volume. Regular RAID needs the drives to be of the same size (or you lose any drive space above the smallest drive size).

Mirroring (x/2) doesn't increase speed anywhere, it can actually decrease speed since same data has to be written to both drives, and system has to wait for that. Here you got redundancy (same data on two drives)., so if one drive fails the system continues as usual with the other drive. You can couple this with spanning to get larger volumes (RAID1+0 etc), but the drive count must be in pair of two. 2, 4, 6, 8 and so on.

You got RAID levels with parity information, most commonly known as RAID 5 and 6 (RAID5: x-1, 3 drive minimum, RAID6, x-2, 4 drive minimum) This is a good mode if you want to pick the best from RAID0 and 1, you get both larger volumes and (some) redundancy. The drawback is that you either need a very expensive controller to handle the parity calculation (also called XOR calculation), or a fast CPU that can handle it as well as the system requirements. The RAID5/6 simply works by adding parity data to each drive. So if one drive fails you replace it, and the system rebuilds the missing data by using the parity information stored on the other drives. This however means also that you can also only lose one drive at a time. Lose enough drives and there will not be enough parity data to rebuild the damage, thus losing everything. RAID6 = RAID5 but with an extra layer of parity, thus needing "an extra drive", storage = x space - 2 drives.

There are also more exotic variants of RAID levels with different configurations, stripes etc, but RAID0, 1, 5, 6, JBOD are the most common ones. RAID 50 perhaps too, which is RAID 5+0.

I hope this didn't get too messy . I am personally using two RAID5 stripes (a stripe is a set of drives configured as a RAID array) across my drives. Each stripe has 16 drives, giving me effective 15 drives storage across those 16 drives (1 drive for parity info, but this info isn't stored on ONE drive, it just takes up the space of one). In reality I shouldn't use such large arrays without using mirrored RAID5 or RAID6, but RAID6 is extremely expensive, and mirrored RAID5... well... expensive too. And my system always got "spares" installed, so if one drive fails the system automatically turns on a spare drive and starts the rebuilding, without any downtime and without me needing to do anything.

So far I have run two RAID stripes for 6 years (with different harddrives), have yet not seen one failed drive...

I hope it answers some of your questions lol.
Image
Official guidelines: The Definitive Guide to BetaArchive :: Abandonware
Channels: Discord :: Twitter

defect17
FTP Access
Posts: 41
Joined: Mon Feb 05, 2007 11:11 pm
Location: Birmingham, UK

Post by defect17 »

mrpijey wrote:Each stripe has 16 drives, giving me effective 15 drives storage across those 16 drives ... And my system always got "spares" installed, so if one drive fails the system automatically turns on a spare drive and starts the rebuilding, without any downtime and without me needing to do anything.
Oh my. Have you got pictures of the hardware behind that kind of installation?

It's also good to know that some vendors will omit the '+' indicating raid nesting, resulting in situations where RAID 01 = RAID 0+1 (a mirror of a striped set), different to RAID 1 (a mirror).

Post Reply