RAID 0. Risky?

[Brad]

The current article on the site about "HD Encoding Face-Off: WMV-HD vs. DivX-HD" contains the following statement in regards to RAID 0.

However, before you consider RAID 0, keep in mind that even though it's fast and efficient, it's also a bit risky. It does not allow for data redundancy; meaning that if one drive fails in a RAID 0 array, the data is lost on all of the drives within the array.

How many times have you heard this? Who actually believes it?

The simple fact is that the chance of loosing all your data with RAID 0, it not much more than using a single HD for your data.

Look at it this way.

Lets say that the chance of a hard drive failing to the point of data loss is 1 in 100,000. (this is a made up figure as the real value would be different for different HD makes and models).

It does not allow for data redundancy;

Neither does any machine with a single hard drive. What is your point?

But I digress.

So you have 1 HD that has a 1 in 100,000 chance of failure in your system.

Now you get another identical drive, and run the two in a RAID 0.

Quiz.. What now is the chance of a HD failing and you loosing all your data?

50%?

Nope..

The chance now of a single hard drive failing is 2 in 100,000.

Hardly what I would call 'risky'.

In a nutshell. RAID 0 is in practice NO MORE RISKY than using a single hard drive.

And if you have a proper backup system (as you should anyway) or use something like RAID 1, there is little risk at all.

 

[tmac]

put it a different way, with actual numbers:
MTBF for 150Gb raptor is 1.2M hours - so having them in raid 0,
will mean the MTBF is 600k hours: running 24/7/52: = 63 years.
They don't put them in servers for nothing.

Of course you may be one of the unlucky ones: Always backup your
important data. - virus - lightning

I'm still debating Raid 0 -

I know the cost doesn't justify the performance gain -

But - Arggg - I love speed -

 

[htpcgb]

You just proved that running a RAID 0 is riskier than a single hard drive. I'm not flaming you here; I'm just saying that "risky" is subjective. You may not find it risky, most people here do. One more thing that you forgot to factor in here: Data recovery from a failed RAID 0 setup is MUCH more difficult. (That is why it's considered a lot more risky).

Just my two cents,
HTPCGB

 

[mark8987]

Man, don't you just love it when a poster comes on criticizing what other people have said, and his criticism indicates he doesn't know what he is talking about himself??

In terms of the post, of course Raid 0 is more risky. Everyone knows that. As you yourself note, in a Raid 0 setup, the failure rate goes up from 1 in 100,000 to 2 in 100,000... That is DOUBLE!!

What is wrong with your reasoning is the 1 in 100,000 figure you pull out of the air. Where did that come from? For your information, current hard drives are atrociously unreliable. All you need to do is go to www.storagereview.com. For reliability, you would typically give it over a certain lifespan - or are you saying that there is only a 1 in 100,000 chance that a hard drive will fail in total. In other words, in one million years time, 99,999 hard drives would still be working?

In reality, the failure rate is more likely to be something like 25% over the course of 3 years. This has been my experience, even though I have always sought to buy hard drives that have the highest reliability reputations possible. You can see on storagereview.com that people report failures for even the most reliable drives.

Hence, with Raid 0, the failure rate doubles (as you said). The indicates the chance of a HD failure over 3 years would increase from 25% to 50%! That is a massive increase. It changes from 1 in 4 chance to 1 in 2 chance.

You then say "And if you have a proper backup system (as you should anyway) or use something like RAID 1, there is little risk at all." Firstly, if you have Raid 1, then you don't have Raid 0. If you want redundancy as well as speed, you then need to go for something like Raid 10, and this takes 4 drives rather than 2 (much more $$$'s).

Secondly, what is a "proper backup system"? Backup is extremely problematic, so you don't want to have to rely on it if you can avoid it. For instance, one of my backup methods was an external hard drive. I plugged it in a couple of weeks ago and discovered it is no longer working. Fortunately, my main drive is still operational. What about backing up to optical? CDRs are notoriously unreliable. Many people find that if they try and read disks they burnt a few years ago, then the disk will just show up as blank. DVDRs may be better. However, you still have the problem that a DVD only holds around 4.5 GB (single-sided), which is hopelessly inadequate for many people these days who store video and music on their computers. For instance, I have 2 drives with 500 GB combined, and about 40GB free space total. Gee, I'm really looking forward to backing up all of this data onto 100+ DVDs.

 

[gudodayn]

I've been using Raid 0 setups since my Epox KT33 days back in 2002 or so and I've only ever had one drive in a raid 0 config die on me!!

Almost every other machine I've built is in Raid 0....maybe I am just lucky but even with the dead raid 0 config I managed to recover what I needed..........

But like mark8987 said, there is no relaibility or guarantee that these storage methods will last forever or as long as they say it would. Manufacturers just hope the component dies after the warranty period!!!

I think people just back up to various formats (DVDs, tape drives, external HDs, etc) just so they have multiple backup copies and the chances of them failing at the same time is rare.......I mean rare!!

If one goes for a raid 0 configuration, he must understand the risks (complicated recovery method require to recover data) but what he/she gains is performance.

People show there are no performance gains in real world applications by going to raid 0 config.

Whoever did that test should try this scenario....... converting 5+ video files, play a FPS game while allowing 10 friends to connect to his PC (the game server)......run that in a single drive system and compare it a raid 0 system (hard drive config being the only difference)......I think he/she will have a different view on raid 0 config. from then on!!

Raid 0 is risky because of data lost??
Here's another adavantage.....I notice my raid 0 config drives are not as hot as my single drive (Raid 0 - work load is spread onto to 2 drives; single drives - work load handled by one drive thus works harder).......I take it the read / write heads arent worn out as much in raid 0 considering if the same amount of data are processed!!

Its not like single drives dont break down!!

 

[Brad]

Hence, with Raid 0, the failure rate doubles (as you said). The indicates the chance of a HD failure over 3 years would increase from 25% to 50%! That is a massive increase. It changes from 1 in 4 chance to 1 in 2 chance.

By your fuzzy maths.. if you have 4 hard drives, you are at 100% chance of one failing in a 3 year period.. (4 x 25%).

Yeah, right. It doesn't work like that. See below.

Secondly, what is a "proper backup system"?

A 'proper backup system' is one that caters for YOUR need to backup what YOU find important. For some that might be burning a new back up DVD every time they add new images from their digital camera. For others it might be nightly backup to tape of development code that is then stored off site. For me it is and ethernet solution to a seperate computer. Of course, if my house burns down, I loose the lot. But my backup system is 'proper' to me.

And if I have a 'proper backup solution' the type of RAID I run is irrelevant ,making the idea that RAID 0 is 'risky' a moot point.

Firstly, if you have Raid 1, then you don't have Raid 0.

I never said you did. I was mearly point out that RAID 1 will offer you a degree of data backup. Note it was in the same sentence as the 'proper backup system'. It was clear what i was getting at (to some anyway)

Everyone knows that. As you yourself note, in a Raid 0 setup, the failure rate goes up from 1 in 100,000 to 2 in 100,000... That is DOUBLE!!

Wrong again.. Not 2 in 100,000 (My mistake to type it like that in my first post).

In other words

Drive 1 has a 1 in 100,000 chance of failure
Drive 2 has a 1 in 100,000 chance of failure

So you have 2 chances of a 1 in 100,000 failure.

Or in your example.

Drive 1 has a 25% failure chance over 3 years
Drive 2 has a 25% failure chance over 3 years

Again, you have a 2 chances of a 25% failure in 3 years

You do not add the failure chances. Each drive is mutually exclusive.

For reliability, you would typically give it over a certain lifespan - or are you saying that there is only a 1 in 100,000 chance that a hard drive will fail in total. In other words, in one million years time, 99,999 hard drives would still be working?

ummm. Dude? 100,000 is One Hundred Thousand.. Not a million.

Man, don't you just love it when a poster comes on criticizing what other people have said, and his criticism indicates he doesn't know what he is talking about himself??

Sigh.. Play nice..

 

[Slowburn]

Again, you have a 2 chances of a 25% failure in 3 years

You do not add the failure chances. Each drive is mutually exclusive.

Except, each drive is dependent on the other for not failing.

 

[Brad]

One more thing that you forgot to factor in here: Data recovery from a failed RAID 0 setup is MUCH more difficult. (That is why it's considered a lot more risky).

This is a good point. But remember, data you don't backup, is data you don't want.

 

[Brad]

Except, each drive is dependent on the other for not failing.

This is a good point too. But having 2 drives, that each have 25% chance of a failure in 3 years, does not give you 50% chance of a failure 3 years. Or like I said, that would be like saying that 4 drives gives you 100% chance of a drive failing.. Clearly nonsence.

So if you have 2 drives, and one fails with a RAID 0 you loose your data.

So what happens if you have one drive, and one fails? Bingo. Lost data.

 

[Slowburn]

Except, each drive is dependent on the other for not failing.

This is a good point too. But having 2 drives, that each have 25% chance of a failure in 3 years, does not give you 50% chance of a failure 3 years. Or like I said, that would be like saying that 4 drives gives you 100% chance of a drive failing.. Clearly nonsence.

So if you have 2 drives, and one fails with a RAID 0 you loose your data.

So what happens if you have one drive, and one fails? Bingo. Lost data.

It definately does not give a 50% chance of failure, but it does increase the risk of data loss as the chances of a drive failure is increased with every extra drive you use.

 

[Kholonar]

Ok, it's been a while since I did probability in Maths and I never liked probability anyway but here goes.

Taking the MTBF of a typical hard drive we get a 50% failure rate after 500000 hours.

Say you run 8 hours per day, that means you have a 50% chance of failure after 171 years.

Now, I have to just assume that the failure rate is linear over time but obviously this aint gonna be true as reliability decreases over usage. So we get a 0.405% chance of failure in 1 year ( 0.5^(1/171) ).

Put it another way, we have a 99.60% chance of getting through the year without failure.

If we now use RAID 1 (and assume that the fact two drives are not splitting the damage between them) then we have a 0.9960 x 0.9960 = 99.20% chance of both drives surviving. Not a massive decrease and it looks like an acceptable loss for ~30% decrease in load times.

However, I don't know about you, but us gamers tend to apply the whip to our drives and I tend to expect failure after about 4 years (which actually correlates quite well to the 25% figure the guy above said). If we use this rough figure then survival in RAID 1 for a year is 0.75 x 0.75 = 0.5625%.

Despite what the topic starter has said, RAID1 can have a significant impact on reliability although measuring it is extremely difficult. Manufacturers quote drive reliabilty to be extremely high compared to real world situations which is where I think this topic is having problems with. If drives really are extremely reliable then RAID1 is acceptable, if not, then suddenly the dangers of RAID1 become realistic.

The problem with a RAID1 array is that it is often used for encoding video and that is often far more precious to people than game saves. In fact, it is probably worthwhile for people encoding video to go for redundancy or parity data from the start. This leaves RAID1 best for loading up data that is easily replicable (such as windows installation files, game data and commonly used programs).

If you wait to use RAID1 for something more risky then try setting up a my briefcase file to backup the data to another system at regular intervals.

 

[taulin]

actually, firstly, an opinion i have that is shared by some other people i know is that a hdd is most likely to fail during its warranty period (or first year), i had a WD drive that died 4 months after purchase, so by that the best idea would be to buy say two drives, and run them in Raid 1 for a year, if either dies you can get it replaced without loss of data, then once theyve been broken in switch to the Raid 0

secondly, the 1 in 100,000 is a statistic, its an average, theres nothing to say that because X number hdd died then X+1 to X+99,999 hdd will function properly, its quite possible that a batch of hdds will be faulty, and that two you happen to buy may be from the same faulty batch or from seperate batches, in either case thinking "it will be alright, i only have a 1 in 50,000 chance of one of the drives dying and losing all my data" isnt going to save you when one of your drives dies, neither is it going to save you if the replacement drive you get happens to be from the same faulty batch and dies 2 or 3 months down the line

the main reason its seen as riskier than having the drives run seperately is because obviously if you have seperate drives and one dies all your data on the other drive is intact, with Raid0 all your data on the other drive is corrupt

 

[Corasik]

MTBF is a 'statisical' scam anyway. Hard drive manufacturers dont actually have hard drives turned on for years, just to find out how long they last.

Far more hard drives will be dead after 5-10 years of use than MTBF implies.

They run large numbers of drive tests in parallel for 'hours' (or days), and calculate MTBF based on how many broke. This is great for calculating short term reliability, but once your disk is a year old, or more, the MTBF really means very little, your drive could fail tomorrow, or it could run another 10 years.

As for drives lasting 171 years, your kidding yourself, after 10 years use the mechanicals of any drive will probably be getting pretty worn out, and so will the platter lubricant.

 

[allhell]

The point is,,, there many fools out there who will never back up anything be it on a single drive or an array. Even with any raid array you can loose all your data if you dont have a back up.
I haven been in the industry for more than 20 yrs & have seen arrays go bad from destrudtion of the entire array due to a PSU failure. If tie customer did not have a back up this would have been a total disaster.
Never kid yourself that a particular raid array is better than another.

THE IMPORTANT THING IS TO ALWAYS BACK UP YOUR DATA NO MATTER WHAT!!!!!!

 

[Kholonar]

Actually, their figures are even funnier than that. They say running 24/7/52 only half of their drives will fail after 57 years! Which is a joke of course, but that's a different topic.

 

[Corasik]

Absolutly, backup of important data is critical. Physical drive faults can be minimized, Raid 5+1 for instance, with each half of the mirrored raid on separate controllers, and separate PSU's.

But no matter how hard you protect the raid from hardware faults, there's always act's of god (lightening strike anyone), and even worse.... User error. Ooops, I just overwrote that file. Raid doesnt protect you from that.

However the chances of recovering data from Raid0 is the lowest of any type of raid storage.

 

[jamesgoddard]

Actually, their figures are even funnier than that. They say running 24/7/52 only half of their drives will fail after 57 years! Which is a joke of course, but that's a different topic.

You don’t understand MTBF fully, if the MTBF failure is quoted at 57 years, you have to read this as 1 drive in 57 will fail each year, it does not mean that some drives will live for 57 years. Also MTBF is normally only valid during the warranty period, after that failure becomes more common.

Thus if the MTBF is 57 years, this abounts to a 5.26% chance each drive will fail during the first three years, if you raid 0 that with 2 drives this becomes a 10.52% chance you will get data loss in three years.

 

[Pain]

I sort of agree with you. If you have one disk, or a RAID 6 array, you need a backup. If you have critical data and you lose it due to a disk or array failure, boo-fricken-hoo, I have no sympathy for you.

The thing about RAID0 is not only the failure rate of the disks, but the failure rate of the consumer grade controllers and drivers most people use, which adds another level of failure to the mix.

With all that said, I don't worry about it if I want to run a RAID0 array. I just don't rely on it and have a backup of critical data, just like I would anyway and everyone should.

 

[deweycd]

They say that its risky because the mean failure time is cut in half, but based on the math earlier, if its 63 years for a 2 drive setup, how long do you actually plan to have your computer for? I hope its not 63 years...

 

[clue69less]

You just proved that running a RAID 0 is riskier than a single hard drive. I'm not flaming you here; I'm just saying that "risky" is subjective. You may not find it risky, most people here do.

How do you know what most people here believe regarding RAID0?

 

[Pain]

hehe. I personally think those who avoid RAID0 simply because it is risky are fools, or at the very least uninformed or unrealistic. :lol:

It's one thing to decide that I'm not going to use it because of whatever reason, but to decide that my data might be at risk with RAID0, but then not have a backup plan while using a single disk is misguided.

 

[htpcgb]

You just proved that running a RAID 0 is riskier than a single hard drive. I'm not flaming you here; I'm just saying that "risky" is subjective. You may not find it risky, most people here do.

How do you know what most people here believe regarding RAID0?

Got me there lol. I don't "know" that for a fact, but that seems to be the general attitude towards RAID 0. (I only joined a while ago, but I've been reading these forums for more than 6 months.)

 

[qwertycopter]

Once you get the HDD out of the package and into your hands the "time before failure" decreases because of physical factors like handling, ESD, vibration, heat, impacts, etc. So pulling some arbitrary number like 1 in 100,000 from the manufacturer's notes is worthless. It doesn't account for user interaction.. only the intended lifespan of the mechanical parts under ideal conditions as reported by the manufacturer. In practice we see a lot more than 1 in 100,000 (or whatever arbitrary number you choose) of our HDDs fail. Surely you know someone who has owned fewer than 10 HDDs and has had multiple failures. Theory usually looks better than practice because of unknown factors and non-ideal environments. The risk is much greater than you think for a single drive when considering ALL the factors in play. I wouldn't want to double or quadruple the risk by using raid 0.

 

[apache_lives]

The current article on the site about "HD Encoding Face-Off: WMV-HD vs. DivX-HD" contains the following statement in regards to RAID 0.

However, before you consider RAID 0, keep in mind that even though it's fast and efficient, it's also a bit risky. It does not allow for data redundancy; meaning that if one drive fails in a RAID 0 array, the data is lost on all of the drives within the array.

How many times have you heard this? Who actually believes it?

The simple fact is that the chance of loosing all your data with RAID 0, it not much more than using a single HD for your data.

Look at it this way.

Lets say that the chance of a hard drive failing to the point of data loss is 1 in 100,000. (this is a made up figure as the real value would be different for different HD makes and models).

It does not allow for data redundancy;

Neither does any machine with a single hard drive. What is your point?

But I digress.

So you have 1 HD that has a 1 in 100,000 chance of failure in your system.

Now you get another identical drive, and run the two in a RAID 0.

Quiz.. What now is the chance of a HD failing and you loosing all your data?

50%?

Nope..

The chance now of a single hard drive failing is 2 in 100,000.

Hardly what I would call 'risky'.

In a nutshell. RAID 0 is in practice NO MORE RISKY than using a single hard drive.

And if you have a proper backup system (as you should anyway) or use something like RAID 1, there is little risk at all.

agreed

 

[Avadann_Kedeth]

The only reason I'm not running RAID 0 right now is because I can't afford to. I would stripe my data with a weekly backup if I had the resources. Gotta wait until I move into a bigger house so I have room to setup my secondary machine as a fileserver/NAS.