Bunch of questions about RAID 0+1

[MxM]

Say I have 4 disks and I have RAID 0+1 (BTW is there any real difference for 4 disks to have 0+1 vs 1+0?)

Suppose that one disk fails during windows startup. What happens? Will windows be able to load and show the error message or what? And what kind of message will be shown? Like disk 2 is failed?

Then how do I repair this? I suppose I have to shutdown computer and replace the disk (hopefully they are marked according to the message, or is there a common practice to number them from top to bottom or from bottom to top?). Then I just restart computer and it fixes itself? Or I should run some kins of utility in windows to do that? So basically how does all this stuff work?

 

[CompTIA_Rep]

The raid controler has software that will warn you if any disk in the raid has failed. If any any time a disk fails, depending on the controler, you should get a Red screen of death with the raid controler telling you a drive has failed.


Now. Most built in controlers wont tell you which drive has failed... thats up to you to figure out. (some will... its in your manual).

 

[nobly]

Here is a pretty picture(s) that show the differences between 1+0 (aka 10) and 0+1.
http://www.acnc.com/04_01_10.html
http://www.acnc.com/04_01_0_1.html

I would say that 10 is better since you basically have 3 copies of the data, whereas 0+1 you only have 2 copies. Of course both are redundant, so it doesn't really matter.

So like Compia_Rep said, it'll say a disk failed, you replace it then rebuild the array from the other 3 drives.

 

[SomeJoe7777]

I would say that 10 is better since you basically have 3 copies of the data, whereas 0+1 you only have 2 copies. Of course both are redundant, so it doesn't really matter.

I don't know where that site you linked to got/created that diagram, but it's completely wrong.

RAID 10 and RAID 0+1 are similar but not the same thing. And neither has 3 copies of the data.

RAID 0+1 and RAID 10 are both nested RAID levels, where one level of RAID has been implemented multiple times on independent drives, and then a second level of RAID is implemented across the multiple first RAIDs.

RAID 0+1 takes physical drives and first creates 2 RAID 0 arrays, each with 1/2 the physical drives. Then, the first RAID 0 is mirrored to the second RAID 0. (i.e. RAID 0 is the first level implemented, RAID 1 is then implemented on top of that).

RAID 10 is the opposite. Physical drives are paired in RAID 1 mirroring arrangements. Then, a RAID 0 is created across the RAID 1s.

The difference between the two is how the arrays perform when a single drive goes down. In a RAID 0+1, when you lose a drive, the entire array now becomes non-redundant. The loss of any remaining operating drive results in loss of all data on the array. Further, if the failed drive is replaced, all drives in the RAID 0+1 array must participate in the rebuild operation.

In RAID 10, when you lose a drive, only the partnered drive in that mirror set is a single point of failure. Failure of other drives in other mirror sets is still allowable and the array will continue operating. When the failed drive is replaced, only it's partnered mirror drive participates in the rebuild - other mirror sets are not affected.

Thus you can see that RAID 10 is a more robust setup.

 

[justfcuk]

IMO a Raid 5 would be the better option if your M/B supports it.

 

[CompTIA_Rep]

AGREED. Dont bother with with complicated nested raid configurations.

 

[MxM]

The difference between the two is how the arrays perform when a single drive goes down. In a RAID 0+1, when you lose a drive, the entire array now becomes non-redundant. The loss of any remaining operating drive results in loss of all data on the array. Further, if the failed drive is replaced, all drives in the RAID 0+1 array must participate in the rebuild operation.

In RAID 10, when you lose a drive, only the partnered drive in that mirror set is a single point of failure. Failure of other drives in other mirror sets is still allowable and the array will continue operating. When the failed drive is replaced, only it's partnered mirror drive participates in the rebuild - other mirror sets are not affected.

Thus you can see that RAID 10 is a more robust setup.

Well for 4 drives with Raid 0+1 the data distribution across the drives is like this

1 2 1 2
3 4 3 4
5 6 5 6
7 8 7 8

Basically the first two drives are in raid 0 and the last two drives are in raid 0, and the first pair is in raid 1 with the second pair.

If you lose the first drive, then you still can lose the second drive and be fine. But you lose data if you lose first and third drive then the data are lost. The question is if you lose the first and last drive, then will Raid be able to recover that? In principle all information is stored on the second and third drives, so you should be able to reconstruct it in principle.

Now 4 drives in Raid 10

1 1 2 2
3 3 4 4
5 5 6 6
7 7 8 8

First two drives are in raid 1, last two drives are in raid 1 and the first pair is in raid 0 with the second pair.

If you lose first drive and second drive, you lose the data, but lost of (first and third) or (first and last) does not lose the data

So, fundamentally, if Raid 0+1 can recover the data if first and last drives lost, then I do not see the difference. The question is can it recover?

 

[CompTIA_Rep]

No. Do raid 5.

End of story.

 

[nobly]

I have to admit it did look funny when I saw it. I'm not a big RAID expert, so I defer to you guys.

 

[CompTIA_Rep]

I dont know why people use raid anyway.


Just load up 3 / 5 hard drives and every time you copy something to drive 1, copy it again to all the other drives. easy as pie, and you can lose all you drives but one and still be functional.

 

[ethel]

I dont know why people use raid anyway.


Just load up 3 / 5 hard drives and every time you copy something to drive 1, copy it again to all the other drives. easy as pie, and you can lose all you drives but one and still be functional.

Yeah, I agree really, unless you're running a multi-user server, in which case striping makes a ton of difference to performance. I suppose that if you're prepared to sacrifice the disk space for the speed and / or redundancy then I can see the point.

 

[CompTIA_Rep]

If thats the case... I would suggest tape backup too. Raid cards do fail, and in that case, you lose EVERYTHING.

 

[ethel]

Yeah, I've had it happen to me on a server, although I've also lost RAID cards and been lucky - I think it depends on whether there are pending writes in the cache when it dies.

Always scares the s**t out of me anyway.

 

[SomeJoe7777]

Well for 4 drives with Raid 0+1

...

If you lose the first drive, then you still can lose the second drive and be fine. But you lose data if you lose first and third drive then the data are lost. The question is if you lose the first and last drive, then will Raid be able to recover that? In principle all information is stored on the second and third drives, so you should be able to reconstruct it in principle.

Now 4 drives in Raid 10

...

If you lose first drive and second drive, you lose the data, but lost of (first and third) or (first and last) does not lose the data

So, fundamentally, if Raid 0+1 can recover the data if first and last drives lost, then I do not see the difference. The question is can it recover?

Ah, but let's look at this a bit closer: (Please excuse my crude text graphics. I didn't have time to build it to scale or to paint it. 8) )

[code:1:70cf002008]
Raid 0+1 Raid 10
======== =======
Drives Drives
A B C D A B C D
1 2 1 2 1 1 2 2
3 4 3 4 3 3 4 4
--- --- --- ---
| | | |
| | | \
| | | -- Mirrored (RAID 1) -
| | \ > Striped (RAID 0)
| | ------ Mirrored (RAID 1) -
| \
| -- Striped (RAID 0) -
\ > Mirrored (RAID 1)
------ Striped (RAID 0) -
[/code:1:70cf002008]

If drive A in the Raid 0+1 fails, the entire RAID 0 (drive A/B) set is offline. The only operational portion of the array is the drive C/D RAID 0 set. Yes, theoretically you could also lose drive B and not lose data, but that's not going to happen because drive B won't be getting accessed (the RAID 0 A/B set is offline). But, the loss of either drive C or drive D takes out the entire array.

If drive A in the RAID 10 fails, the A/B mirror set is no longer redundant. The subsequent loss of drive B loses all data on the array. But either drive C or drive D can subsequently fail and the array is still good.

Thus, in the 4-drive setup, the failed Raid 0+1 has no chance of surviving with another operating drive failure (the operating drives being C and D). But in the Raid 10, if another random operating drive fails, the array has a 2/3 chance of surviving.

And, there is the aforementioned recovery aspect. After the failed drive A is replaced in the Raid 0+1, the Raid 0 A/B set is now blank. The Raid 0 C/D set must be re-mirrored to A/B, thus all drives are participating in the rebuild. In the Raid 10, drive B re-mirrors to drive A. Drives C/D are not involved.

This is with 4 drives. If you increase the number of drives to 8, it gets worse:

[code:1:70cf002008]
Raid 0+1 Raid 10
======== =======
Drives Drives
A B C D E F G H A B C D E F G H
1 2 3 4 1 2 3 4 1 1 2 2 3 3 4 4
5 6 7 8 5 6 7 8 5 5 6 6 7 7 8 8
------- ------- --- --- --- ---
| | | | | |
| | | | | \
| | | | \ --- Mirrored (RAID 1) -\
| | | \ ------- Mirrored (RAID 1) \___ Striped (RAID 0)
| | \ ----------- Mirrored (RAID 1) /
| | --------------- Mirrored (RAID 1) -/
| \
\ -- Striped (RAID 0) \__ Mirrored (RAID 1)
---------- Striped (RAID 0) /
[/code:1:70cf002008]

Raid 0+1: Drive A fails, drives B,C,D now offline. Failure of any of E,F,G,H loses all data.

Raid 10: Drive A fails, only drive B is a single point of failure. Loss of any of drives C,D,E,F,G,H doesn't cause a problem.


Now, several people have made reference to RAID 5 being a better solution than either of these. That's true to an extent. You lose less drive space to the redundancy, and the setup is simpler. But the cost of a RAID 5 card that can read/write as fast as either the RAID 0+1 or RAID 10 is several hundred dollars, and will require a PCI-X or PCIe slot. If you have a bunch of old drives lying around and want a fast redundant array, then RAID 0+1 or 10 becomes attractive.

 

[I]

yes windows will load, and windows won't know squat about what has happened from a single drive failure.

That's why you run the RAID management software which monitors the array and creates an "event" which could be a log, a popup warning, or even an email to (whoever). A member being lost from an array is but one event possible but perhaps the most important to know, as well as the event that it was rebuilt.

Further, on some raid controllers you can hot-plug a drive and initiate the rebuild. You should install the software and see what it offers, if enough to suit your needs before investing in the raid card or creating arrays with it.

Keep in mind that if you are using a motherboard integral (particularly southbridge integrated) controller, you may find that if the board dies, you have it harder to source the same controller/board again to get that data off. Even though PCI card failure is less common, personally i have spares of any cards I use.

As for why you might do a raid 1 even if you make backups, the cumulative time, value of your work inbetween backup periods may exceed the cost of the drive(s) and controller over their lifetime. It's no substitute for offline backups, just another more timely layer of protection.

 

[pip_seeker]

I dont know why people use raid anyway.


Just load up 3 / 5 hard drives and every time you copy something to drive 1, copy it again to all the other drives. easy as pie, and you can lose all you drives but one and still be functional.

Yeah, I agree really, unless you're running a multi-user server, in which case striping makes a ton of difference to performance. I suppose that if you're prepared to sacrifice the disk space for the speed and / or redundancy then I can see the point.

probably because it sounds flashy... "hey look what I can do!" syndrome.

It's only worth it if you need it and you know what you are doing... if you don't then it will cost you $$, frustration and possibly even the stability of your pc.

Cons:

*Equipment cost
*complex setups cause complex problems
*HEAT!!! 4 drives spinning will create a toaster in your PC
*All harddrives should be identical... if you should have the luck of a good long life raid this will be a problem if you can't find your particular harddrive when the harddrive goes down. This will certainly happen it happened at my work and they had a 80GB Maxtor. Not Made anymore... seagate bought them out!


Pros:

*Performance
*Disk size: really doesn't matter they are making over half terabyte drives now with terabyte in the forseeable future.
*To say you can do it?


Raid is really only suitable these days for folks who can't live without it and can afford the cost. [IE. Servers with heavy load / Content producers capturing / editing HD Video. etc. ]

If the argument is for back ups you can buy a plug in USB backup drive with software that has simple one touch backups. That is fractions of what you'd spend for RAID with much less trouble.

 

[TeraMedia]

Here's a different viewpoint.

I'm running a 4n RAID 5 array, and a 4n RAID 0 array off the motherboard controller (D955XBK). I did fail a drive, and the BIOS supplemental software for the motherboard RAID controller notified me during bootup. I was able to boot with the non-redundant array. I was also able to put in another drive, rebuild the RAID 5, reset the RAID 0, and go about my business. No data lost. No manual (or automated) copying of files between drives.

RAID 5 is very good for write-once, read-many (e.g. your O/S, your photo collection, your music collection, and other media). It gives you the added benefit of read rates approaching RAID 0 (albeit with 1 fewer disks). RAID 10 (or 01 or whichever) will give you better write performance, but there is of course the larger storage penalty of 1/2 instead of (n-1)/n.

Make sure you put your temp directory, your page file, and other disposable files on a RAID 0 or a single disk. Windows will put these in the same place as the O/S by default, which will hurt your performance a lot if that's a RAID 5 array. There's quite a bit of config work to get this right, as each login has its own temp variables, etc.

 

[I]

Cons:
...

*HEAT!!! 4 drives spinning will create a toaster in your PC
*All harddrives should be identical... if you should have the luck of a good long life raid this will be a problem if you can't find your particular harddrive when the harddrive goes down. This will certainly happen it happened at my work and they had a 80GB Maxtor. Not Made anymore... seagate bought them out!

Heat is not much of an issue. You're talking about 3 addt'l drives at about 8W each IF all are actively used rather than separate volumes where one sleeps. 24W is not that big of a deal in today's systems.

To put it in perspective, one of my aging fileservers has only the 80mm PSU exhaust and one 80mm rear chassis exhaust, and stays pretty cool but it only has 6 drives in it ATM.

As for all drives having to be identical, NO!
This is usually an urban myth. While it is true a larger drive (than associated member(s) in same array) would have the addt'l space wasted, there is no reason to think this is a problem, on the contrary you are making the wrong choice to buy the same older technology drive if it were available as it'll only be slower (on average) than one more modern.

If the argument is for back ups you can buy a plug in USB backup drive with software that has simple one touch backups. That is fractions of what you'd spend for RAID with much less trouble.

Depends on what your time and data is worth. A USB drive itself is no protection between when you last made a backup and any given moment thereafter. Not everyone needs this kind of realtime protection but considering that many motherboards have it integral or a PCI card is about $20 USD, and another drive can often be had for about $50 AR, the fraction you allude to in price difference is a very small fraction, actually it's usually no more expensive than the external USB drive.

Even so, if someone could only have one or the other, removable backup (that they DO remove), OR internal raid, then I'd have to go with the removable (if removed) backup. It's just useful to remember that each has it's own purpose, neither is a complete substitute.

 

[MxM]

Well, I am not trying to run server. I am considering this for home computer. I just like when things are working fast, plus I do play games.

So I was considering Raid 0, but after I had bad incident with HD failure which resulted in loss of information (my backup was not very recent, when nothing is happening you tend to make backups less and less often in time), I decided that for such lazy person like me I have to have some redundancy in data storage.

However just RAID 1 is may be not good enough, because it has worse performance than a single drive for writes, and I am not sure if it gets much better performance for reads. (According to Dell representative, it is not much performance gain over single HD for their DataSafe thing). So I am considering to go with 0+1 or 1+0 and have it all: fast and reliable.

 

[Madwand]

However just RAID 0 is may be not good enough, because it has worse performance than a single drive for writes, and I am not sure if it gets much better performance for reads. (According to Dell representative, it is not much performance gain over single HD for their DataSafe thing). So I am considering to go with 0+1 or 1+0 and have it all: fast and reliable.

I'm not sure where this comes from.. RAID 0 has the best performance of the common RAID levels (not considering some of the rarer modes which are not supported by typical hardware), which is exactly why so many nested RAID levels incorporate a 0. As always, the raid performance depends somewhat on the implementation, parameters (esp. stripe size), and access pattern. But if you have a RAID 0 implementation/configuration that can't significantly outperform a single drive for both straight-ahead sequential reads and writes, you simply have a bad RAID 0 implementation/configuration.

In addition, thinking that RAID 10 or RAID 01 will magically perform faster than RAID 0, esp. for writes when RAID 0 has write performance problems is illogical -- the RAID 1 part never increases write performance, so how would RAID 10 be faster for writes than RAID 0? But there's no write performance problem with RAID 0 in a good implementation, so there's no problem in the first place, just some bad information, logic, or some bad implementations/configurations.

/em looks sceptical
/em wonders if salesmen are trying to sell more, overpriced, drives

Finally, your problem was laziness in backups, and you still haven't learned that lesson. RAID, regardless of levels, is not a backup. Every RAID level introduces new, significant points of failure. Even failure recovery with RAID can be error prone, esp. for newbies and cheap implementations. Doing regular, automated if possible, backups better is the more appropriate solution here, and would also help in the maintenance of any RAID level.

 

[CompTIA_Rep]

Well for my problem... Hoping to get any suggestions if there are any.


I have 350gb worth of music and another 100 worth of programs.
I just picked up 2 500gb sata drives and plan to run Raid 1. The mobo is common so if it dies I shouldnt have any problems replacing it and thus, the raid controller.

This IS my backup solution... and yes I realize that any outside source could destroy, burn, or steal both drives. Aside from that, I do not see why its a bad way to backup that much data.


Any reason to change my setup? (i have not implemented it yet and i am using 3 hard drives to store my data, which is not ideal).

 

[ethel]

As Madwand says, backups are the only real solution to protecting your data.

With mirroring, you could lose your data thus:

1) Accidentally deleting the data
2) Losing both disks within quick succession - i.e. the second disk goes before you get a chance to replace the first one
3) Your raid controller dying and corrupting your RAID (the likelihood of this depends on your controller, but it does happen)

With backing up, all of the above risks are eliminated.

The way I back up is easy - I have a second disk that I back up to every week. It takes time, but it is safer.

Even better, get a USB2 disk caddy for your disk (incredibly cheap) and then you can store the disk somewhere else in case someone takes a fancy to your PC.

Or more elegantly, get a cheap NAS device (£50) and put it on your home network, hidden away, and back up to that.

But mirroring is obviously less hassle. It's a choice between convenience and risk.

 

[MxM]

However just RAID 0 is may be not good enough, because it has worse performance than a single drive for writes, and I am not sure if it gets much better performance for reads. (According to Dell representative, it is not much performance gain over single HD for their DataSafe thing). So I am considering to go with 0+1 or 1+0 and have it all: fast and reliable.

I'm not sure where this comes from.. RAID 0 has the best performance of the common RAID levels (not considering some of the rarer modes which are not supported by typical hardware), which is exactly why so many nested RAID levels incorporate a 0.

I am sorry, I have mistyped there. I meant RAID 1, not RAID 0. I have corrected my post.

 

[Madwand]

But mirroring is obviously less hassle. It's a choice between convenience and risk.

Mirroring + external backup is probably fairly common in businesses, which like active redundancy; understand the importance of doing real backups, and also have the equipment and software for doing so.

Some people also use mirroring as a backup tool in this way -- they use the create & break mirror processes to duplicate drives. The part that makes this a backup and not just a mirror is disconnecting the drive once it's mirrored. Cycling three or more drives could accomplish mirroring + backup.

But operations like this are not safe for novices to undertake without having some other full backup of the data, which brings us back to simple external backups (e.g. to an external USB / firewire / eSata / etc. drives (if not tape, etc.)) first in safety.

If user is able to start off with replaceable test data, then he/she can learn error failure and recovery in context by testing it, attaching and detaching drives to simulate failure and recovery. This is recommended for any redundant RAID configuration. Many people have gone ahead and assumed that RAID will protect them from drive failures, and found out that when the failure happened, it didn't -- because the implementation wasn't great and the failure expanded during recovery to multiple drives or that they didn't do the recovery steps properly and ended up with data loss, etc.

Which reminds me.. I better update my file server backup..

 

[still_life]

I was going to use 4 hdds in a raid 0+1 setup. I think it gives both redundancy and performance benifits...but really, how often has your hdd failed??? 0+1 is great in my book and really is all you need (if at all)