RAID 0 with 3 HDD's

PDH-NicFury

Distinguished
Mar 4, 2006
91
0
18,630
Time for more fun and games...

Anyone know if mobo's support RAID 0 with 3 discs? I ask this because I recall an issue of MaxxxxxPX testing 2, 3, and 4 disc RAID setups with 3 discs having the best performance. Putting together a lan party PC, and want to know if I need an add on card (which I won't buy, just go with two drives) or if the mobos chipset drivers for RAID will work with 3 HDD's.

Feel free to reply now...

PDH-NicFury
 

Dr_House

Distinguished
May 23, 2006
40
0
18,530
Time for more fun and games...

Anyone know if mobo's support RAID 0 with 3 discs? I ask this because I recall an issue of MaxxxxxPX testing 2, 3, and 4 disc RAID setups with 3 discs having the best performance. Putting together a lan party PC, and want to know if I need an add on card (which I won't buy, just go with two drives) or if the mobos chipset drivers for RAID will work with 3 HDD's.

Feel free to reply now...

PDH-NicFury

All recent and almost all (should be all) past RAID Controllers (thus, motherboards as well) support 1 to infinite drives in RAID 0 (okay, it isn't infinite, but it's a sizable number).
 

IVAces

Distinguished
Feb 9, 2006
84
0
18,630
I realize that you are shooting for the best performance but when you run raid 0, you triple your chance of a system failure. As I'm sure you know, if you loose a drive in raid 0, you loose everything.

If I were you, I'd run raid 5 with 3 drives. You get that same read speed as raid 0 and you have some data security to boot.
 

Dr_House

Distinguished
May 23, 2006
40
0
18,530
I realize that you are shooting for the best performance but when you run raid 0, you triple your chance of a system failure. As I'm sure you know, if you loose a drive in raid 0, you loose everything.

If I were you, I'd run raid 5 with 3 drives. You get that same read speed as raid 0 and you have some data security to boot.

You don't get exactly the same, because the way RAID 5 is setup, it can cause a little slower access times, but people talking about RAID 0 being "exponentially increased chance of failure" is getting annoying, and anybody with years of experience in RAID's (such as myself) won't read a website that says "RAID 0 increases system failure" because all drives are the same, if you have 5 independent drives being accessed at once vs. a RAID 0 of 5 HDD's, there's the same chance of losing data, though you won't lose everything with the former, the chance is the same, not increased with the RAID.
 

FITCamaro

Distinguished
Feb 28, 2006
700
0
18,990
It doesn't increase the chances of the system failing since if the drive was going to fail, it would fail regardless of whether its alone or in a RAID. The issue is that if its in RAID 0, you loose all your data instead of just some or none of it. So yes it does increase your chances of loosing all your data.
 
well it is a fact.... i have been running raid 0 for years....but if one drive goes i will NEVER get my data back...

Raid 0 splits your data on both(or 3 or 4...ect) drives so if u do not loose half(1/3 1/4 ...so on) a file....not like half a song.... u loose every second(3rd, 4th) "stripe".

With a 5 meg mp3 and a 128 k stripe size your would loose about 20(come fix my math someone) file parts so that song would be gone (kind of like the multi-source downloads on BT and other P2P's)

On raid 5 (correct me if i am wrong)
Data is stored accross all disks in the same way as raid 0 but with parody data to re-build in the case of a drive failure?
I have looked it up... but it was along time ago.

i just googled this to make the point
Raid 5

Point is Raid 5 IS safer. and i will be kicking my self the day one of my drives goes and i loose 500gigs of data, but thats what i have manual back-up drives for :)

i will be running raid 0 for the os(games) only in the near future. i don't need 100+ megs a sec for mp3's and PVR
 

IVAces

Distinguished
Feb 9, 2006
84
0
18,630
I realize that you are shooting for the best performance but when you run raid 0, you triple your chance of a system failure. As I'm sure you know, if you loose a drive in raid 0, you loose everything.

If I were you, I'd run raid 5 with 3 drives. You get that same read speed as raid 0 and you have some data security to boot.

You don't get exactly the same, because the way RAID 5 is setup, it can cause a little slower access times, but people talking about RAID 0 being "exponentially increased chance of failure" is getting annoying, and anybody with years of experience in RAID's (such as myself) won't read a website that says "RAID 0 increases system failure" because all drives are the same, if you have 5 independent drives being accessed at once vs. a RAID 0 of 5 HDD's, there's the same chance of losing data, though you won't lose everything with the former, the chance is the same, not increased with the RAID.

I'm sorry, but I have to respectfully disagree with you. The more points of failure in a system, the higher the chance that the system will fail. The chances of failure with 1 out of 3 hard drives are greater than the chances of a single drive. Lets say that "Joe's Hard Drive Company" has a failure ration of 1 our of 1,000 produced drives. If you purchase 1 of Joes drives your chance of failure is 1 in a thousand. If you purchase 3 hard drives from Joe, it goes from 1/1000 to 3/1000 or 1/333. Every manufacturing company on earth has a ratio of product failure to product produced. The work hard to keep the ratio low but it's impossible to eliminate it as long as humans are involved in the process. Therefore, if you purchase 3 hard drives from a company, the odds are better that you'll have one that fails prematurely than if you had only purchased one. I'm not trying to talk down to you or anyone, I'm just trying to show you the flaw in your logic.

As far as your statement goes with raid 0 vs. raid 5 read speed, the differences are negligible at best. Where you loose on performance with raid 5 over raid 0 is in the write performance because it has to write to the disks and it has to write the parody bits also.

I also hapen to be a huge subscriber and victim of Murphy's law. I've lost 3 raid 0 partitions of my own when the ones I built identical to my own for friends still work to this day. Go figure. If it were me, I'd run raid 5 with a hot spare or even raid 6. I hate reloading my system.
 
The chances of failure with 1 out of 3 hard drives are greater than the chances of a single drive. Lets say that "Joe's Hard Drive Company" has a failure ration of 1 our of 1,000 produced drives. If you purchase 1 of Joes drives your chance of failure is 1 in a thousand. If you purchase 3 hard drives from Joe, it goes from 1/1000 to 3/1000 or 1/333.

I perform reliability and SPC calculations for laboratory equipment, and if the same rules apply to computers, then the logic that I constantly see about RAID setups being twice or three times more likely to fail are not valid.

To start off, all parts will eventually fail. What really matters is that they don't fail before they are replaced. In the example that 1 in 1000 units will fail before being replaced means that 99.9% of the units will last before being replaced. For the total RAID system, the overall reliability will be the product of the reliablility of all parts. If we simplify things and only consider the hard drives (the RAID controller could go out too you know) for three hard drives with a reliability of 99.9%, the overall reliability would be .999x.999x.999=.997 or 99.7%. It drops, but not as much as most think.

The real issue is that I believe hard drives are the most unreliable component next to power supplies. I bet 1 in 10 hard drives fail before being replaced... it seems everyone has had it happen at least once. That would mean the reliability for a triple drive RAID is more like .90x.90x.90=0.729 or 72.9% (about 1 in 4 chance of failure)

Then again, maybe none of this applies to computer parts like it does in the manufacturing industry.
 

IVAces

Distinguished
Feb 9, 2006
84
0
18,630
HD's are manufactured parts. I never said that you increase your chances of having a problem are multiplied by how many hard drives you have in your stripe set. I simply said that you increase the chances. That is a true statement. We both agree in that fact. I also agree with you that 3 hd's doesn't triple your failure point. It simply increases it by the percentage of mean failure time and faulty drive ratio.
 

Dr_House

Distinguished
May 23, 2006
40
0
18,530
What I mean by chances is the likely hood of a drive dying, and that does not increase with more drives in a RAID 0. The only thing that increases is the ability to lose data in a RAID 0, because if 1 drive dies, all the data is lost whereas only 1 drive of data would have been lost. The chances of a drive failing is not increased, but the chances of losing more data does increase. I think we, or at least I, were talking about different things.
 

IVAces

Distinguished
Feb 9, 2006
84
0
18,630
What I mean by chances is the likely hood of a drive dying, and that does not increase with more drives in a RAID 0. The only thing that increases is the ability to lose data in a RAID 0, because if 1 drive dies, all the data is lost whereas only 1 drive of data would have been lost. The chances of a drive failing is not increased, but the chances of losing more data does increase. I think we, or at least I, were talking about different things.

I Think we are agreeing on the principle just not the finer points.

I can live with that.
 

raider_

Distinguished
Jan 12, 2006
6
0
18,510
Standard Statistics.. guys... come on now.. rwpritchett you are right when dealing with things that work Independently.... finding the failure would at that particular time be .999x.999x.999 but we are dealing with an item that works in conjunction with and directly depends on 2 others so the forumula would change to .999+.999+.999. Thats the difference. It is true that 2 drives in RAID0 double your chance for failure 1/1000 changes to 2/1000. Althought i take that risk everyday... backup backup backup!


The chances of failure with 1 out of 3 hard drives are greater than the chances of a single drive. Lets say that "Joe's Hard Drive Company" has a failure ration of 1 our of 1,000 produced drives. If you purchase 1 of Joes drives your chance of failure is 1 in a thousand. If you purchase 3 hard drives from Joe, it goes from 1/1000 to 3/1000 or 1/333.

I perform reliability and SPC calculations for laboratory equipment, and if the same rules apply to computers, then the logic that I constantly see about RAID setups being twice or three times more likely to fail are not valid.

To start off, all parts will eventually fail. What really matters is that they don't fail before they are replaced. In the example that 1 in 1000 units will fail before being replaced means that 99.9% of the units will last before being replaced. For the total RAID system, the overall reliability will be the product of the reliablility of all parts. If we simplify things and only consider the hard drives (the RAID controller could go out too you know) for three hard drives with a reliability of 99.9%, the overall reliability would be .999x.999x.999=.997 or 99.7%. It drops, but not as much as most think.

The real issue is that I believe hard drives are the most unreliable component next to power supplies. I bet 1 in 10 hard drives fail before being replaced... it seems everyone has had it happen at least once. That would mean the reliability for a triple drive RAID is more like .90x.90x.90=0.729 or 72.9% (about 1 in 4 chance of failure)

Then again, maybe none of this applies to computer parts like it does in the manufacturing industry.
 
Standard Statistics.. guys... come on now.. rwpritchett you are right when dealing with things that work Independently.... finding the failure would at that particular time be .999x.999x.999 but we are dealing with an item that works in conjunction with and directly depends on 2 others so the forumula would change to .999+.999+.999. Thats the difference. It is true that 2 drives in RAID0 double your chance for failure 1/1000 changes to 2/1000. Althought i take that risk everyday... backup backup backup!

I'm sorry, I think you misunderstood me raider. You are quoting probability statistics and you are absolutely right about the probability of a failure correlates directly with the number of items. Three items will have three times the probability of failure.

I was using reliability statistics, which is a different chapter. You made me actually dig up my old CQE statistics book, and sure enough it's in there:) For a system that is in series (if one part doesn't work, the whole thing doesn't work), you multiply the reliability of each part to get the total reliability.

The point I was orignially trying to make is that people need to look at the reliability of the RAID, not the probability of failure. After all, if I told someone that their computer as a whole has about 10,000 parts and therefore it is 10,000 times more likely to fail than a single part, it would scare the bejesus out of them. Another good example would be to compare a Honda with a Chevy. They have about the same number of parts, but the Honda is more reliable as a whole because the individual reliability of each part is better.

Here's the bottom line: Forget the statistics people, you should be backing up your important data if you have one hard drive or a hundred. If you're not backing up your data then you are an idiot and deserve to lose everything. Raider is right, "backup backup backup".

RP