Can somebody explain "R.A.I.D" to me in a language I will understand?

kingaaa

I just bought a n.a.s. 4 bay server and am about to populate it with some drives, but I'm getting a bit confused about r.a.i.d. as in, whats the difference between r.a.i.d 0, 1, 5, 6 ? I want to use it to store all my media files and distribute throughout the home network so I want to get the biggest capacity possible - anybody give me a "dummy's guide to R.A.I.D. " please?

Fysh

kingaaa wrote: »

I just bought a n.a.s. 4 bay server and am about to populate it with some drives, but I'm getting a bit confused about r.a.i.d. as in, whats the difference between r.a.i.d 0, 1, 5, 6 ? I want to use it to store all my media files and distribute throughout the home network so I want to get the biggest capacity possible - anybody give me a "dummy's guide to R.A.I.D. " please?

RAID is short for Redundant Array of Independent Disks.

RAID 0 ("Striping") - basically takes 2 or more drives and treats them like one big drive, giving some performance increase in terms of read and write speeds. Doesn't provide any security whatsoever, and if one disk goes, you lose everything across the whole array.

RAID 1 ("Mirroring") - takes 2 (or in some cases more) drives and sets them up in parallel. Basically the second drive is a copy of the first, so that if one drive dies you have an identical working copy.

RAID 5 ("striping with distributed parity") - takes 4 or more drives and creates one big drive volume, with parity data distributed across all drives such that if one drive dies you can replace it and rebuild the array (with no data loss). The tradeoff is that you lose 1 drive's worth of space to the parity data.

RAID 6 ("striping with double distributed parity") - like RAID 5, only it doubles the parity data and therefore the fault tolerance ie you can lose 2 disks from the array and still rebuild with no data loss - but you lose 2 drive's worth of space to do so.

It is important to note that most NAS units will also support a configuration known as JBOD ("Just a Bunch Of Disks") - each drive is set up independently of the rest.

In terms of how to choose what option's best for you...you need to answer some questions:

Is your NAS going to be the master copy of all data stored on it?
Which is most important: total available storage, read/write performance of storage, or reliability/recoverability of storage?

RAID 0 will give you the full capacity of your 4 drives (with some read/write performance increases) , but if one of them dies, all your stuff is gone.
RAID 1 will give you the capacity of 2 of your drives. You can survive up to 2 disk failures (though it depends which disks fail - if both disks in a single array fail, you're stuffed ).
RAID 5 gives you the capacity of 3 drives along with the ability to recover from a single drive failure without data loss. (It should be noted that with higher-capacity disks the possibility of a disk failure happening during an array rebuild becomes more significant).
JBOD will give you the full capacity of your 4 drives at standard read/write performance. If one of them dies, you lose the contents of that drive but the others are unaffected.

It should be noted that if your NAS is going to be a canonical master copy of All Your Stuff, you should also have some backup strategy in place - because if the NAS box dies, for example, you don't want to find out that the RAID controller used some Weird Ass-Backward Implementation of RAID5 such that having 4 working drives still doesn't let you get at your data. (For example, my home setup for the time being is to use a 4-bay HP Microserver with 4x1TB drives in JBOD as canonical master, with a backup on 2x2TB drives in my desktop.)

Have a read of the Wikipedia page for a few more details, and this article about the limitations of RAID 5.

Edited to add:

Should've included JBOD in the summary at the end. Also details as to why using RAID doesn't mean you can ignore backups.

kingaaa

cheers mate, very informative

kc66

kingaaa wrote: »

I want to get the biggest capacity possible

To get the highest capacity from your 4 disks and some fault tolerance, raid 5 is your best option. You will have the full capacity of 3 disks.

Edit: sorry didnt read to the end of Fysh's post. He effectively said what I said.

kingaaa

I think I'll just go for R.A.I.D. "0", want to get the most I can out of this. Can I put pre-loaded disks in or do they have to be formatted in the bay?

Torqay

It should be noted that all the safe-guard features (fault tolerance, redundancy, etc.) may not prevent the loss of data in worst-case scenarios.

To be on the safe side, I recommend another storage facility of similar capacity (preferably at another location) and scheduled incremental backups.

vibe666

kingaaa wrote: »

I think I'll just go for R.A.I.D. "0", want to get the most I can out of this. Can I put pre-loaded disks in or do they have to be formatted in the bay?

you REALLY REALLY don't want to do that if you value your data at all.

RAID0 should only ever be used where you need the maximum throughput with zero regard for reliability, such as a temporary drive for video editing and should NEVER be used for storing any data ou don't have reliable backups of.

the problem you have with RAID0 is that if WHEN you lose a disk (and you will), you will lose EVERYTHING, totally and with almost zero chance of any possibility of recovery.

as a rule of thumb, with 2 disks, you'd use RAID1, with 3 or 4 disks you'll tend to go for RAID5 and any more than that, you'll want to look into RAID6.

vibe666

as for formatting the disks, you just buy them, throw them in the NAS and then format them using the browser based GUI in the format of your choice.

Torqay

kingaaa wrote: »

I think I'll just go for R.A.I.D. "0", want to get the most I can out of this. Can I put pre-loaded disks in or do they have to be formatted in the bay?

RAID 0 for a NAS??? RAID 0 is the most error-prone option, why would you do that? Read/write speed is hardly of relevance for a file server. Remember, 1 drive goes down and all is lost.

bonzodog2

Strange, then, that RAID 0 is even part of the RAID spec, since there is no redundancy at all ! I suppose you could say there's a clue in the ZERO though.

BostonB

RAID isn't a backup either

http://www.2brightsparks.com/resources/articles/RAID-is-not-a-backup-solution.html
http://randomdrake.com/2009/01/02/why-raid-is-not-backup/

vibe666

bonzodog2 wrote: »

Strange, then, that RAID 0 is even part of the RAID spec, since there is no redundancy at all ! I suppose you could say there's a clue in the ZERO though.

whilst it doesn't actually offer any redundancy, it does have it's uses and uses exactly the same principles (and more specifically disk controllers) to do it's thing.

but yup, not very redundant at all.

Fysh

Heh, I'm reminded of the BOFH article in which RAID (-1) was introduced for secure storage ("Write once, read never").

Back on topic, I've edited my earlier post to try and make clear that RAID0 is really not what the OP should be using and is basically a more problematic alternative to JBOD or RAID5.

kingaaa

Fysh wrote: »

Heh, I'm reminded of the BOFH article in which RAID (-1) was introduced for secure storage ("Write once, read never").

Back on topic, I've edited my earlier post to try and make clear that RAID0 is really not what the OP should be using and is basically a more problematic alternative to JBOD or RAID5.

Sorry, thats the option I meant, "JBOD". I'm basically using it to distribute video files around the home network, I mainly wanted the option of not having to have the pc running and the idea of getting rid of a few of my usb drives is reason enough alone.

Torqay

Well, with RAID 0 they're not going to be "distributed around the home network" any faster, even at the slowest RAID option, the transfer rate of the drives is still light years beyond the speed of a gigabit network.

BostonB

If it fails, you'd want a backup to save all the work in setting it up again.

FruitLover

If your NAS supports RAID 1+0, this is your best option with 4 disks. RAID 5 is only ever a good idea if you have exactly 3 disks (or your RAID controller/software can't support RAID 1+0).

As has already been mentioned, RAID is not in itself a backup system. If the files on a NAS get deleted or infected, they're gone for good. Make sure any critical files are backed up elsewhere too.

bonzodog2 wrote: »

Strange, then, that RAID 0 is even part of the RAID spec, since there is no redundancy at all ! I suppose you could say there's a clue in the ZERO though.

RAID 0 hints at the 'I' part of the acronym - 'Inexpensive'. RAID 0 was particularly popular with admins who had a load of old, small SCSI drives that still worked perfectly well, but were too small to be very useful on their own. So rather than use (e.g.) six old 10GB disks as five individual drives, you could use RAID 0 to make the OS see them as a single 60GB drive. You balanced the relatively high risk of failure against the ability to get use out of old drives that you would otherwise just chuck out.

bonzodog2

FruitLover wrote: »

RAID 0 hints at the 'I' part of the acronym - 'Inexpensive'.

I thought it was I for Independent.

RAID 0 was popular at one point among people who wanted higher performance and didn't really care about data security (especially gamers). SSDs have pretty much made that redundant now... unless you wanted to RAID 0 some SSDs.

I've had to repair some Dell Dimension 9200s over the years which were shipped with RAID 0 and suffered a drive failure. In many cases the customer didn't even know it was a RAID 0 array.

Fysh

FruitLover wrote: »

If your NAS supports RAID 1+0, this is your best option with 4 disks. RAID 5 is only ever a good idea if you have exactly 3 disks (or your RAID controller/software can't support RAID 1+0).

As has already been mentioned, RAID is not in itself a backup system. If the files on a NAS get deleted or infected, they're gone for good. Make sure any critical files are backed up elsewhere too.

Why would RAID 5 not make sense in a 4-drive configuration? You've got whatever probability of a single drive failure, from which you can recover in either 0+1 or 5.

I mean, yes, 0+1 means that if you hit a second drive loss you've got a 2/3 chance of still having all your data intact - but if we're already talking about a system that should have a backup system in place, why would it be worth sacrificing half your storage capacity to introduce redundancy that can cope(ish) with two drive failures? Assuming that the usage plan doesn't require or particularly benefit from the read/write benefits that 0/1/0+1 provide, it seems like overkill to me.

vibe666

FruitLover wrote: »

If your NAS supports RAID 1+0, this is your best option with 4 disks. RAID 5 is only ever a good idea if you have exactly 3 disks (or your RAID controller/software can't support RAID 1+0).

i'm pretty confident that someone who was contemplating RAID0 with 4 disks is much more concerned with maximising capacity than redundancy.

and whilst there's definitely a business case to be made for RAID0+1, for the most part, it doesn't make much sense (at least to me) in a home NAS.

Karsini wrote: »

SSDs have pretty much made that redundant now... unless you wanted to RAID 0 some SSDs.

that would be me. :cool:

overkill? probably.

totally awesome? definitely!

Torqay

Yea, SSDs really shine in a RAID 0 array!

kingaaa

thanks for the info lads, now I have one last question. I bought the Verbatim powerbay off ibood, and even though it says maximum 8tb - there seems to be lots of sites selling the same model with 12tb (I've ordered 1 x 3tb disk to test it) but my question is, if I put a 3tb disk in and it works - will all the subsequent disks have to be 3tb or can I put any mixture of capacity into the powerbay?

Torqay

Although I'd recommend to use only disks of the same make, model and indeed firmware, you can use any mixture of different drives and volume sizes in a RAID array (at your peril).

vibe666

also, if you wanted to forgo RAID entirely (again, at your peril ) most (if not all) NAS's will allow you to create a storage volume using a single disk and also manage more than one volume, so you can use whatever disks you have.

most will still require reformatting the disks either way, so you'll want to do one, copy everything from your old drive(s) one at a time, then add them and reformat them.

i can't say for certain that the verbatim NAS will allow you to do this as i've never used one, but most NAS's would, so as long as you have at least one disk with more available space on it than your next biggest full disk, you can probably do it that way.

having said that, everyone here (inc. myself) would advise against it, as you WILL have a disk fail on you sooner or later and you'll lose whatever is on it if there's no redundancy in place.

Fysh

Torqay wrote: »

Although I'd recommend to use only disks of the same make, model and indeed firmware, you can use any mixture of different drives and volume sizes in a RAID array (at your peril).

I've always wondered about this - I've seen suggestions of standardising capacity but using more than one vendor and drive model to reduce chances of more than one disk failing at the same time, but it's difficult to know which way to jump...

vibe666

i'd imagine that differences in read/write performance due to different drive models having different specifications would potentially cause more problems than the minute chance a whole bunch of disks crapping themselves at exactly the same time.

i've seen plenty of individual failed disks in a RAID set and even buggy firmware with the potential to cause issues in a bunch of disks in my own NAS (you know who you are samsung! :P), but i don't think i've ever heard of a whole load of disks biting the dust at once like that, so i know where i'd rather put my money. (he says, instantly cursing himself!)

Torqay

Fysh wrote: »

I've always wondered about this - I've seen suggestions of standardising capacity but using more than one vendor and drive model to reduce chances of more than one disk failing at the same time, but it's difficult to know which way to jump...

Now matter how you spin it, even the most sophisticated RAID setup doesn't make a proper backup redundant, it's still the best safeguard against the loss of data.

vibe666

Torqay wrote: »

Now matter how you spin it, even the most sophisticated RAID setup doesn't make a proper backup redundant, it's still the best safeguard against the loss of data.

very true. i'm running RAID6 on my current NAS, but that hasn't stopped me mirroring all my valuable data elsewhere.

there was a pretty scary article written a couple of years ago that i thought was just scaremongering until it actually happened to me.

http://www.zdnet.com/blog/storage/why-raid-5-stops-working-in-2009/162

that was with 5x 1.5tb disks in a RAID5 set, so maybe it was just really bad luck, but i keep thinking about that article and since then, i always hope for the best, but plan for the worst when it comes to my NAS.

Capt'n Midnight

bonzodog2 wrote: »

Strange, then, that RAID 0 is even part of the RAID spec, since there is no redundancy at all ! I suppose you could say there's a clue in the ZERO though.

It's called Zero because that's the exact number of files you will recover if any of the drives fail.

an alternative is to use the 4 drives separately and use symbolic links on the first one so that from the OS's point of view the other three drives can be seen as subfolders on the first drive

it's not as fast as RAID0 , you still have to move files since you aren't sharing free space

But you can search all the drives at once and if any drive dies you only loose what was on that drive.


In general apart from Mirroring (RAID 1) RAID means moving your one point of failure from a hard drive to a controller card.
If the raid controller dies you may not be able to recover you data without the self-same card, and in extreme cases a card with the same firmware revision

Not an issue if you have plenty of spare hardware or a same day response contract, otherwise you'll looking at your Plan B.

Capt'n Midnight

FruitLover wrote: »

RAID 5 is only ever a good idea if you have exactly 3 disks (or your RAID controller/software can't support RAID 1+0).

You also need IT people who know that a red flashing light is top priority and will drop everything until the raid array is rebuilt.

As has already been mentioned, RAID is not in itself a backup system.

it is not a backup system
three reasons to use raid
increased space - but 2TB drives are far cheaper than RAID
speed - but only with expensive controllers
redundancy - but only from single hard drive failures

RAID 0 hints at the 'I' part of the acronym - 'Inexpensive'. RAID 0 was particularly popular with admins who had a load of old, small SCSI drives that still worked perfectly well, but were too small to be very useful on their own. So rather than use (e.g.) six old 10GB disks as five individual drives, you could use RAID 0 to make the OS see them as a single 60GB drive. You balanced the relatively high risk of failure against the ability to get use out of old drives that you would otherwise just chuck out.

?
you never saved money by using SCSI RAID instead of software mirrored IDE/SATA - the cost per GB was many multiples in price and the bathtub curve means that old hdd's are more likely to die , the argument is similar to airplanes should no more than two engines.

where possible I left new servers for a month before going live, and more than one had drive failures :mad:

Capt'n Midnight

Torqay wrote: »

Although I'd recommend to use only disks of the same make, model and indeed firmware, you can use any mixture of different drives and volume sizes in a RAID array (at your peril).

However, if possible don't use drives from the same batch. If there is a manufacturing problem then all drives could fail around the same time.

In theory RAID 5 means it's unlikely for another drive to fail while you are rebuilding from the hot spare. In practice this happens and it's not happy bunny time.

Having your OS on a separate pair of mirrored drives means you can recover quickly to a point where you can restore data drives from backup. It's handy to get the users to tell you which data / folders / systems they want restored first.