How often does RAID 0 really fail?

[Gunslinger08]

So you always see people talking about how RAID 0 is unstable and will fail. This doesn't make sense to me. I've used probably 50 hard drives over the past 10 years and never had a single one fail. Being in a RAID 0 array shouldn't affect the reliability of individual disks at all. Why do people think it's so unstable?

 

[chusteczka]

I have seen a few people over the past year in the Technical Support category asking for help in recovering files from drives that were in a RAID 0 array.

It is not so much a failure of the RAID array than it is a failure of the hard drive. More drives fail in the summer due to the heat than any other time of the year. Over the hottest point of the summer, the end of July through the beginning of August, there are usually 5 - 7 threads per day asking for help with a failed hard drive. The rest of the year, the cause of a failed drive is typically a cheap or failing PSU.

 

[Gunslinger08]

Originally posted by: chusteczka
I have seen a few people over the past year in the Technical Support category asking for help in recovering files from drives that were in a RAID 0 array.

It is not so much a failure of the RAID array than it is a failure of the hard drive. More drives fail in the summer due to the heat than any other time of the year. Over the hottest point of the summer, the end of July through the beginning of August, there are usually 5 - 7 threads per day asking for help with a failed hard drive. The rest of the year, the cause of a failed drive is typically a cheap or failing PSU.

So the controller introduces no new threats to the drives. All I needed to know. Thanks.

 

[soydios]

The controller does not make the drives fail any faster, no.

But, a RAID-0 array will fail twice as often as a single hard drive.

 

[DaveSimmons]

Originally posted by: soydios
The controller does not make the drives fail any faster, no.

But, a RAID-0 array will fail twice as often as a single hard drive.

Exactly, if either one of the 2 drives fail, you lose 100% of your data.

Let's say the odds of either drive failing within one year is just 10%.

Then with RAID-0 the odds of losing your data from either or both dying is equivalent to (not)(both drives surviving) which is 1 - (0.9 * 0.9) = 1 - 0.81 = 19% chance of complete data loss, slightly less than double the odds.

 

[dirtrat]

I learned my lesson the hard way. Even if the data gets corrupted on one drive then you have lost it all. The minimal speed increase for the casual user/gamer isn't worth the risk but by all means take your chances if you don't believe what other users have posted.

 

[RebateMonger]

Originally posted by: joshsquall
So the controller introduces no new threats to the drives. All I needed to know. Thanks.

Well, that's not completely true. First, many RAID0 arrays are built using built-in RAID chipsets, some of which aren't quite as reliable as high-end separate RAID controllers. RAID chipsets seem to be somewhat more prone to failure than "standard" IDE chipsets, which are pretty well developed and tested.

Second, ANY RAID array is more complex than a simple IDE interface, and people make mistakes managing their RAID arrays. I've never seen an estimate of this risk, but it's obvious that the human factor is there and causes array failures. A search through AnandTech's Forums will show many people losing their RAID data even though no drive failed.

 

[JEDIYoda]

Originally posted by: RebateMonger

Originally posted by: joshsquall
So the controller introduces no new threats to the drives. All I needed to know. Thanks.

Well, that's not completely true. First, many RAID0 arrays are built using built-in RAID chipsets, some of which aren't quite as reliable as high-end separate RAID controllers. RAID chipsets seem to be somewhat more prone to failure than "standard" IDE chipsets, which are pretty well developed and tested.

Second, ANY RAID array is more complex than a simple IDE interface, and people make mistakes managing their RAID arrays. I've never seen an estimate of this risk, but it's obvious that the human factor is there and causes array failures. A search through AnandTech's Forums will show many people losing their RAID data even though no drive failed.

I use raid 0 on 2 of my riggs and I will tell you that without a doubt 99% of all raid0 failures are due to human error.....

Both of mine were my fault not the systems fault...yet
Rule #1 to anybody using raid0 always back your important stuff to another harddrive!!

 

[mindless1]

Do you always see people talking about how it's "unsafe and will fail" or are you making it up?

Do you have evidence, references, citations? I think you are embellishing and it "might" be a sign of an illogical user which is the greatest risk of all to an array - user error.

The risk of a (assuming RAID0) is 200% failure vs. non-raid, single drive, but once an illogical user is involved that doesn't show due diligence in making sure they start out with a good stable raid bios and refrain from fiddling with the array (user error is one of the greatest sources of loss, not drive failure) then the risk is up a lot.

The better question is why ask? What do you hope to gain? It is better to find tech solutions to a problem rather than the other way around, to take some bit of tech and try to force an implementation of it.

 

[RebateMonger]

Originally posted by: joshsquall
I've used probably 50 hard drives over the past 10 years and never had a single one fail.

You've been a bit lucky. But that's how probabilities work. Your fortunate experience helps balance those who post on AnandTech, "Why have I had three hard drives fail in three months? What am I doing wrong?"

Hard drives DO fail. ALL of them fail, eventually. Google's recent study of a zillion hard drives gave failure rates very close to my own experience. I see about one-in-twenty hard drives failing each year. Google said about one-in-sixteen.

Can you run five years with a RAID0 array and not have a failure? Sure. But the odds are something like 50-50.

 

[Sunner]

I'd say by far the most common misconception with RAID0 is that it'll give your average user a big performance boost in day to day tasks.

 

[Roguestar]

Originally posted by: soydios
The controller does not make the drives fail any faster, no.

But, a RAID-0 array will fail twice as often as a single hard drive.

Came here to post this.

 

[LazyGit]

Originally posted by: Sunner
I'd say by far the most common misconception with RAID0 is that it'll give your average user a big performance boost in day to day tasks.

You're not the first person on here to say that. What is the basis for this? I'm looking to build a RAID0 myself in the hope that it will speed up load times for my OS and programs as well as speed up file transfers and unRARing. Is that what you would call 'average use'? Is a RAID0 not going to help in those situations?

 

[Tlkki]

The load level of the drives is also a factor when calculating the odds in failure. Especially if it is a low grade desktop HD.

My RAID-0 failed in 2 months

EDIT: unrarring is a good example where raid-0 is faster. i could say maybe 60% faster in most cases. Of course it speeds up even more if you unrar to another disk.

 

[Sunner]

Originally posted by: LazyGit

Originally posted by: Sunner
I'd say by far the most common misconception with RAID0 is that it'll give your average user a big performance boost in day to day tasks.

You're not the first person on here to say that. What is the basis for this? I'm looking to build a RAID0 myself in the hope that it will speed up load times for my OS and programs as well as speed up file transfers and unRARing. Is that what you would call 'average use'? Is a RAID0 not going to help in those situations?

If you have a task that's bound by sustained transfer rates, RAID-0 will help substantially.
Transfers of large files(a DVD image for example), un-rar'ing large archives(presuming you don't get CPU bound there), etc are good examples of this.
Load times for most things aren't very dependent on STR, access times will be more important here.

Average use would be, booting your computer, opening your browser, office app, whatever, play a game or two, maybe watch a movie, etc etc.
If you spend much of your time waiting for WinRAR to finish decompressing DVD images and such, yeah, you'll benefit from it.

 

[robmurphy]

Have a look at the following article:

http://www.anandtech.com/storage/showdoc.aspx?i=2969

If you look there is very little performance advantage for a desktop machine in using the 2 drives in raid 0. The fact that raid 0 means that you are twice as likley to lose your data is a basic fact. The loss of 1 drive looses all the data, and as you have 2 drives the chance of 1 of the 2 drives failing is double that of a single drive. This makes raid 0 a high risk for little reward, and as the article points out a backup nightmare.

Just my 2p worth.

 

[ch33zw1z]

Never fails, you're all set!!

/me waits for imminent "WTF MY DATA" thread.

 

[Roguestar]

You mean YART?
Yet Another RAID Thread :laugh:

Edit: no, I'm serious. There are too many goddamn RAID threads because people either screw it up or refuse to use wikipedia/google to find out the basic differences between RAID 1 and RAID 0 and crap like that.

 

[ch33zw1z]

hehe, exactly.

 

[gsellis]

R0 is NOT twice as likely to fail as a single drive. The statistics are more complicated than that. And I have had more problems with controllers than drives failing. Having that problem right now with my ICH5R.

 

[Hyperlite]

mine lasted about 2 years. 2xWD Caviar 80gb on an NF7-S

 

[Hyperlite]

Originally posted by: ch33zw1z
Never fails, you're all set!!

/me waits for imminent "WTF MY DATA" thread.

OMGWTF MY DATAZ IZ GONE!!

:laugh:

 

[Riverhound777]

If you want the "increase in speed" from RAID-0, use RAID-5. It isn't like hard drives are that expensive these days.

 

[Idontcare]

I can't wait for people to realize that having 2 cores instead of 1 core means your computer is twice as likely to start corrupting data and killing itself.

Same for those 2 platter hard-drives, everyone knows you must go with single-platter to avoid data loss.

Oh and sucks for you if you do dual-channel RAM...everyone knows your likelihood of loosing all your data (corrupted on its way to your single-platter hard-drive after being processed on your single-core processor) is twice that over the smarties who use single-chip DDR memory (no 8 chip or 16 chip sticks in their boxes, that would be dumb) in single-channel mode.

If you are worried about Raid-0 dying (and you should be) then you should be worried about your single hard-disk dying, in either case you lost your data whether it be in 4 months (Raid-0 example) or 8 months (single-drive example). Either case you are a moron for not having a proper backup strategy in place.

Get used to losing reliability. And take off your tinfoil hats because you will be using quad-core chips next year or soon thereafter...OMG 4 times the probability of any one of those cores causing data corruption!!!! SkY IS FALLING!!!!

 

[RebateMonger]

The original question was:
"Is it true that RAID 0 is unreliable and will fail?"

RAID 0's probability of failure is (approximately) two times higher than a single drive. But ALL drives will fail eventually. If you don't want your data at risk, keep independent backups. Then, drive failure becomes a matter of inconvenience, rather than tragedy.