How about a RAID guide?

[Soul Rider]

With more and more people venturing into the RAID arena, how about we get a guide written providing as much information as possible about RAID and how it works?

If people want to submit questions that would be useful in a guide, then either people can supply the answers, or they can be researched on the net, to provide a comprehensive list of information.

Put you Q&A below and we will try and get something up here for people to refer to..

 

[semi-pro waster]

Thing is it isn't exactly hard to either set up a Raid system now or to find information about it given a quick search but anyway this might work as a starting point.

At a consumer level there are commonly two types of Raid to be discussed, Raid0(striping data across two or more drives) and Raid1(mirroring data across two or more drives). Raid has a couple of meanings but one of the most common is Redundant Array of Inexpensive Disks. The system is setup in a form of second 'bios' where you build the array, how exactly this is done differs from controller to controller but as a general rule you will have to press a combination of keys at a prompt(on a Promise controller it is normally ctrl + f) to enter the Raid 'bios'. Ideally the drives used should be exactly the same, ordered at the same time, from the same manufacturer and matching in every way to minimise the chance of problems.

Raid0 is where you connect two or more physical drives through a Raid controller to create one logical drive that will appear to your OS. You get the total size of all hard drives used in this version of Raid, strictly however it isn't Raid as it doesn't have the redundancy of other forms. Raid0 works by striping the data across multiple drives to increase both the read and write speed so it can be very useful for working with large files(e.g. video, graphics or audio processing) however it does increase the risk of data loss due to the reliance on two drives continuining to work as one.

Raid1 again works with two drives and the second mirrors the contents of the first, you will have to go into the Raid 'bios' again and choose the Raid1 option instead. You get to use the size of the smallest disk so as you will normally use drives of exactly the same size that means you get half the total size if you used the disks separately and again you will only see one logical disk. Raid0 gives you an exact copy of what is on the first drive, it does improve read speeds but does nothing for write speeds(and depending on the overheads may prove marginally slower) but is useful in situations where data integrity is important.

This is all from memory so I'm happy to make any corrections and I haven't covered Raid0+1 or Raid5(as the most common others but of course there are loads more) as I know less about them and frankly I can't be bothered right now

 

[Soul Rider]

That covers the basic, what is RAID question.

I think we might also need to look at the benefits of each type, For example

RAID1 - Provides redundancy in the event of disk failure but no performance enhancements. Works out expensive as you have to pay for the disk space required twice (1 for storage, 1 for mirror image).

RAID0 - Provides no redundancy (therefore isn't really RAID), offers improvements to read and write speed (Maximum increase of 100% - 2 drives, 200% - 3 drives, 300% - 4 drives.

Any other ideas?

 

[semi-pro waster]

RAID1 - no performance enhancements.

I did sort of cover the benefits and drawbacks briefly but Raid1 does improve the read speed because data can be taken from either drive or both together.

If you can get someone to cover the other types of Raid then it may become more useful and/or suggest improvements to what I wrote. Part of the problem is the same as always in that most people don't read stickies for whatever reason.

 

[Jimbo Mahoney]

Q: What size should my stripe / cluster be? (RAID 0)

A: There's no right answer, but here's a link to some freeware which will help give a breakdown of the file sizes on the drive you wish to RAID:

http://www.jgoodies.com/freeware/jdiskreport/

A good starting point could be:

stripe size = median file size / stripe width

where stripe width is the number of drives across which data will be striped. (e.g. two for a two drive RAID 0 setup).

If your stripe size is large (like 128k), HDTach will like you, but your OS might not. Anything smaller than the stripe size will NOT benefit from the RAID and will be just like a single drive.

Cluster size should perhaps be between 1/2 and 1/4 the stripe size, but remember a large cluster size could waste drive space if you have lots of files smaller than the cluster size.