raid 0

[djnes]

Even their own "followers" are finding flaws in their methodology and findings. Sorry to say, but you don't easily say Anandtech and StoreReview are flat out wrong, and expect people to just buy into it. Those are also only a small sample of the reputable sites who have done their own testings. MaximumPC was yet another who had a multi-page article debunking RAID0.

Tweakers.net's article even mentions that RAID0 can make some functions slower, but they don't go into detail. This is a very large detail because the function in question is seek times....finding small files scattered around a hard drive, which is possibly the most common task. The very same people who used to spout that throughtput meant nothing on a drive, seek time is all that matters, are the same hypocritical people extolling the virtues of RAID0.

Furthermore, is as basic math concept as any to see why RAID0 is much more risky in terms of drive failure than a single drive. The amount of people who can't grasp that is staggering and shocking. All I'm saying is, do the research. Many MANY sites have proven RAID0 to be hype. Most of the boutique system builders also have gone away from it. As Fox Mulder used to say, the truth is out there.

 

[PaHick]

Even their own "followers" are finding flaws in their methodology and findings. Sorry to say, but you don't easily say Anandtech and StoreReview are flat out wrong, and expect people to just buy into it. Those are also only a small sample of the reputable sites who have done their own testings. MaximumPC was yet another who had a multi-page article debunking RAID0.

Tweakers.net's article even mentions that RAID0 can make some functions slower, but they don't go into detail. This is a very large detail because the function in question is seek times....finding small files scattered around a hard drive, which is possibly the most common task. The very same people who used to spout that throughtput meant nothing on a drive, seek time is all that matters, are the same hypocritical people extolling the virtues of RAID0.

Furthermore, is as basic math concept as any to see why RAID0 is much more risky in terms of drive failure than a single drive. The amount of people who can't grasp that is staggering and shocking. All I'm saying is, do the research. Many MANY sites have proven RAID0 to be hype. Most of the boutique system builders also have gone away from it. As Fox Mulder used to say, the truth is out there.

Ive done research, and im happy with the results.And as far as you saying"you don't easily say Anandtech and StoreReview are flat out wrong", I can just as easily as you say they are right,correct? This is a highly debated subject,and im not trying to prove that Raid0 is best for everyone.What I am trying to do is get everyone on the same page here. And by that I mean everyone needs to weigh the pros and cons,which articles like Anandtechs and Storage Review CLEARLY state cons,but rarely pros. They are very one sided articles and I think everyone needs to hear the other side of the coin, just as you among others feel the need try and shut Raid0 down further.

"Tweakers.net's article even mentions that RAID0 can make some functions slower", just as AT and SR mention some instances where Raid0 is faster, correct again?Seek times arent the be all end all of a system. Disk density and where the data is placed on the platter have just as much bearing on speed.And your "math" equation does not make sense,so I guess im one who cannot grasp and will continue to shock you.A dead drive is a dead drive.Oh and I undertstand with Raid1 I willl still continue to function. But im unwilling to spend money and another drive to mirror another. Complete waste of money unless your someone unwilling to backup your data.What happens if both drives in Raid1 go bad? Dead system right? And what kills a drive, more than likely a lightning strike or power surge IMO. Which could theoretically take out both drives in your array, correct? And as far as "Many MANY sites have proven RAID0 to be hype", there is an equal number of sites that can be posted in favor of Raid0,besides the equal number of users who are happy with Raid0 that have ACTUALLY tried it.AND IMO most "boutique" system builders that have quit using have done so because their customer base has grown in such a way that they are catering to a broader scope of users, users who may not have a use for Raid0. Besides the point of saving money by offering one large drive over several smaller ones.Remember they get better deals than we do, so price/Gig is lower as the drive gets bigger.Oh! almost forgot where you stated "Even their own "followers" are finding flaws in their methodology and findings"...Maybe I missed something, but it looked like and even debate to me.And heres something from SR FAQ....

Test 1x 200 GB DM+9 2x 200 GB DM+9 RAID 0
IPEAK Average Read Service Time 13.2 ms 13.3 ms
WinBench 99 Outer Zone Transfer Rate 55.3 MB/sec 109.4 MB/sec
Inner Zone Transfer Rate 31.6 MB/sec 63.1 MB/sec
SR Office DriveMark 2002 395 IO/sec 426 IO/sec
SR High-End DriveMark 2002 373 IO/sec 408 IO/sec
SR Bootup DriveMark 2002 288 IO/sec 474 IO/sec
SR Gaming DriveMark 2002 519 IO/sec 529 IO/sec

JMO but it looks as though there was a benefit there.Maybe you could explain why there was an increase everywhere but the IPEAK Average Read Service Time.Im not trying to step on your toes,I understand the pros and cons. And I think a user needs to go over the options when he(or she) decides. But IMO theres too many one sided debates on this subject,Thank you.

 

[TeeJayHoward]

I once read someone describing a RAID-0 drive failure as being like this:

You have two drives. Both are 99.99999% good. If you RAID-0 them, the array is only 99.99998% good. The .00001% failure rate doubled. Do you care? Not really

And a lightning strike is not the only thing that will take out a drive. Ever overclocked too far and had your system flip out and corrupt your Windows install? How about a simple file system corruption issue? It doesn't take much to corrupt a drive.

I don't RAID-0, but that's because I don't need it. From what I've seen, it will offer me no performance increase in my Maya render times, it won't help me load Maya or Firefox faster, and honestly, why would I need RAID-0 to watch DVDs?

 

[PaHick]

I once read someone describing a RAID-0 drive failure as being like this:

You have two drives. Both are 99.99999% good. If you RAID-0 them, the array is only 99.99998% good. The .00001% failure rate doubled. Do you care? Not really

And a lightning strike is not the only thing that will take out a drive. Ever overclocked too far and had your system flip out and corrupt your Windows install? How about a simple file system corruption issue? It doesn't take much to corrupt a drive.

I don't RAID-0, but that's because I don't need it. From what I've seen, it will offer me no performance increase in my Maya render times, it won't help me load Maya or Firefox faster, and honestly, why would I need RAID-0 to watch DVDs?

If you overclocking and your hard drive somehow gets corrupt files, what would be the difference? Raid0 would need reloaded,but so would Raid1 correct?Its a mirror of the first hard drive,which is corrupt? The mirroring just didnt majically stop before the file corruption.And Maya is CPU,GPU intensive,not hard drive so yes,Raid0 is not for you.DVD watching??????Hmmm, isnt that what DVD players are for???..lol. Seriously, if thats all you do with your system then definately arent going to benefit, besides maybe boot time . I didnt say its for everyone.

 

[TeeJayHoward]

If you overclocking and your hard drive somehow gets corrupt files, what would be the difference? Raid0 would need reloaded,but so would Raid1 correct?Its a mirror of the first hard drive,which is corrupt? The mirroring just didnt majically stop before the file corruption.

I never said RAID-1 would be better, or even mentioned it. RAID in general is not suited to desktop users (Now workstation users are a different matter...)

DVD watching??????Hmmm, isnt that what DVD players are for???..lol.

First off, one question mark, not 6, please. And my monitor is bigger than my TV. Why would I pay for a DVD player when I have a perfectly good DVD drive in my computer?

 

[DougLite]

In the Tweaker's article, they increase the multi-tasking load by having EMule and WinAmp running in the background. Why would anyone in their right mind have these tasks running on Raptors? It makes much more sense to have a second independent spindle service these requests, especially when you consider that a higher capacity 7200RPM drive can fill the role, and get you MORE gigs for the SAME dollars. For the same amount of money as the second 73GB Raptor, you could have a 300GB DM10. Your MP3 playback and EMule work will certainly not be disk limited, and a 7200RPM drive is more than up to the task. You could use the DM10 to keep an image (or two!) of the Raptor, still be 200GB ahead of having the second Raptor, AND not have RAID-0's reduced reliability (indeed, reliability would actually be increased with the availability of the image). In addition, this would allow you to continue to leverage locality on your Raptor while still enjoying the multi tasking boost of having two drives.

ONCE AGAIN, RAID-0 is not the answer on the desktop. Even the multi tasking argument is now dead. For the same amount of money, you get the multi tasking boost, 150% more capacity, and an increase in reliability, by going to the trouble of planning your spindle allocation.

 

[djnes]

1) How is it possible, that on a forum of advanced computer users, so many have trouble understanding a basic mathematical concept? When your data is dependent on 2 drives to exist, instead of just one, your twice as likely to experience data loss? If one drive dies, you lose ALL your data....from both drives. Someone PLEASE tell me why this is so hard to grasp?

 

[TeeJayHoward]

1) How is it possible, that on a forum of advanced computer users, so many have trouble understanding a basic mathematical concept? When your data is dependent on 2 drives to exist, instead of just one, your twice as likely to experience data loss? If one drive dies, you lose ALL your data....from both drives. Someone PLEASE tell me why this is so hard to grasp?

Because computer guys are notorious for having bad math skills. Calc on up is cake for us, but addition, subtraction, and, heaven forbid, division are all death in small bundles.

 

[PaHick]

Why would anyone in their right mind have these tasks running on Raptors? .

Because that might have been a users choice to run 2 Raptors, would it have made a difference running 2 7200rpm drives? No. Still would have been the same outcome.

AND not have RAID-0's reduced reliability

Do you have any data showing a Raid0 array going bad more than a single drive? I have yet to see a study on this.

multi tasking boost of having two drives

Do you have a link to a test showing the benefit of 2 singles against Raid0 in this environment? I would like to read it if so.Like I said before, different uses,different needs. But the numbers dont lie,and neither do the users who have seen a benefit,even if its just "feels snappier". Im learning and in the process ive made my decision, but I would like to continue to learn and if you could so kindly point me to the articles or tests I would be obliged.Thank you.

1) How is it possible, that on a forum of advanced computer users, so many have trouble understanding a basic mathematical concept? When your data is dependent on 2 drives to exist, instead of just one, your twice as likely to experience data loss? If one drive dies, you lose ALL your data....from both drives. Someone PLEASE tell me why this is so hard to grasp?

I understand where your coming from, but like I said a dead drive is a dead drive. In a Raid1 array you would continue to operate, right? But when the other drive is replaced it would need "mirrored again right? So basically if your "reinstalling the mirror, it wouldnt be to much harder to recreate the Raid0 array from a backup correct? Thats the key, from my point of view. Backups are necessary nomatter what array your using.DVD's,external hard drives....either way you still have your data if you have a good backup plan.

 

[DougLite]

Because that might have been a users choice to run 2 Raptors, would it have made a difference running 2 7200rpm drives? No. Still would have been the same outcome.

Yes, it would have. You leverage the Raptor's better seek performance in your disk limited tasks, then use a less expensive 7200 drive for tasks that access the disk regularly, but are limited by other factors. A 7200RPM drive will carry out disk limited tasks (such as loading programs, bootup, etc) slower than a 73GB Raptor, but the Raptor is not going to playback your MP3s any differently, nor will it make your Internet connection faster than a 7200RPM drive.

Do you have any data showing a Raid0 array going bad more than a single drive? I have yet to see a study on this.

It doesn't require a study. Running RAID-0 does not reduce the reliability of the drives themselves. It only reduces the reliability of the array. Simple mathematics. Given tires of equal design on a car and motorcycle, the car is more likely to get a flat tire (and incapacitate the whole car), as it travels over two axes instead of just one with the motorcycle. The tires may be equal, but the risk is still higher.

Do you have a link to a test showing the benefit of 2 singles against Raid0 in this environment? I would like to read it if so.Like I said before, different uses,different needs. But the numbers dont lie,and neither do the users who have seen a benefit,even if its just "feels snappier". Im learning and in the process ive made my decision, but I would like to continue to learn and if you could so kindly point me to the articles or tests I would be obliged.Thank you.

Once again, you don't need to test to figure this. If your Raptor is only performing your disk intensive tasks, and is not performing more 'mundane' tasks limited by other system factors, then it will be faster than the same drive performing all of the tasks. A scientific comparison would be required to compare the performance of a RAID-0 setup to a independent spindle hybrid setup. However, I am confident that you will realize most of the gains of having the second spindle, will have more capacity, and enjoy additional flexibility, as you will not have the drives stand or fail together.

I understand where your coming from, but like I said a dead drive is a dead drive. In a Raid1 array you would continue to operate, right? But when the other drive is replaced it would need "mirrored again right? So basically if your "reinstalling the mirror, it wouldnt be to much harder to recreate the Raid0 array from a backup correct? Thats the key, from my point of view. Backups are necessary nomatter what array your using.DVD's,external hard drives....either way you still have your data if you have a good backup plan.

Your comparison to RAID-1 is flawed. When rebuilding a RAID-1, the system remains available (admittedly at reduced storage performance while the rebuild proceeds). When your RAID-0 fails, the new drive must be installed and a restore from backup performed before the system becomes available again. That's a significant amount of downtime. Even after half of a RAID-1 dies, I can still hit the power button and boot the system. If a RAID-0 fails, I'll get no farther than the RAID BIOS puking out. You are correct that no form of RAID is a replacement for backups, which I also address in my "Truth About RAID" write-up.

My overall point with this is this. RAID-0 may increase performance in some applications/access patterns. However, it comes with several costs. Reduced reliability, doubled cost, and less flexibility. With some creative thought, a solution that is at least close, and possibly better, on performance and brings the additional bonuses of 150% more capacity, more reliability, and more flexibility for the same monetary cost becomes apparent. Why not go that way?

 

[djnes]

Shorty, you tried to hi-jack our thread, so I hi-jacked your post. I've moved it [thread=924003]here[/thread] and gave it a thread all its own. - DL

Do you have any data showing a Raid0 array going bad more than a single drive? I have yet to see a study on this.

Would you do a study to prove that 50 cents is equal to half of one dollar? Then why do a study to prove something almost as basic?

If you've ever worked in a datacenter, you'd know that the more hard drives that you employ, the more likely you will to, at some point, experience a drive fail. If you have one single drive, you still always have a risk of a drive failure. If you put two drives in your system, your risk doesn't change. One drive fails, you lose the data on that drive. Now when you use both of those drives in RAID0, if either of your drives die, you lose ALL your data. Simple enough? Should be, but here's more.

Imagine if your car exploded if only one of the four tires went flat. How safe would you feel driving it? Now imagine an 18-wheeler, same rules. One tire goes flat and the whole truck explodes. See the point? Your risk increases because your more reliant on each drive. There's no redundancy, so if one dies, your screwed.

 

[PaHick]

There's no redundancy, so if one dies, your screwed

Agreed, but again thats where the backup comes in.Same way with a Raid1 setup. You still have to rebuilt the mirror on that system. And yes DougLite, I understand you will be able to power on that system,but it still takes time and CPU usage to rebuild that array. We are still talking desktops here right? I dont see downtime being a huge problem in a desktop. And how much time are we talking? Throw a drive in there to replace the bad, copy your image.....shouldnt be a huge difference over rebuilding the mirror right? And since your rebuilding your mirror what else would you be doing with this system while it was rebuilding?I dont think anyone would be using a system that was rebuilding the mirror.

Yes, it would have. You leverage the Raptor's better seek performance in your disk limited tasks, then use a less expensive 7200 drive for tasks that access the disk regularly, but are limited by other factors. A 7200RPM drive will carry out disk limited tasks (such as loading programs, bootup, etc) slower than a 73GB Raptor, but the Raptor is not going to playback your MP3s any differently, nor will it make your Internet connection faster than a 7200RPM drive.

If I use your analogy, why would I be using a 7200rpm drive if I was accessing the disk regularly? Wouldnt I be better off with the Raptor with its seek time, cause as you say seek time is everything in a desktop."A 7200RPM drive will carry out disk limited tasks (such as loading programs, bootup, etc) slower than a 73GB Raptor".....but 2 7200rpm in Raid0 will be as fast or faster.Seek time isnt everything.Data density and placement on the platter are just as important as you can see by the performance of newer,larger drives...correct?

BTW..sorry bout the quotes..lol..in too much of a hurry...fixed

 

[DougLite]

Agreed, but again thats where the backup comes in.Same way with a Raid1 setup. You still have to rebuilt the mirror on that system. And yes DougLite, I understand you will be able to power on that system,but it still takes time and CPU usage to rebuild that array. We are still talking desktops here right? I dont see downtime being a huge problem in a desktop. And how much time are we talking? Throw a drive in there to replace the bad, copy your image.....shouldnt be a huge difference over rebuilding the mirror right? And since your rebuilding your mirror what else would you be doing with this system while it was rebuilding?I dont think anyone would be using a system that was rebuilding the mirror.

If I use your analogy, why would I be using a 7200rpm drive if I was accessing the disk regularly? Wouldnt I be better off with the Raptor with its seek time, cause as you say seek time is everything in a desktop."A 7200RPM drive will carry out disk limited tasks (such as loading programs, bootup, etc) slower than a 73GB Raptor".....but 2 7200rpm in Raid0 will be as fast or faster.Seek time isnt everything.Data density and placement on the platter are just as important as you can see by the performance of newer,larger drives...correct?

BTW..sorry bout the quotes..lol..in too much of a hurry...fixed

There's no way that two 7200RPM drives in RAID-0 will match the localized seek performance of a 73GB Raptor. Localized seek performance is by far the most important facet of hard drive performance on the desktop. Yes, increasing areal density does increase performance, as it makes locality more acute. However, no amount of density will overcome a deficit of more than 5ms in seek performance given roughly equal buffer strategy, for the frequent localized seeks that dominate desktop disk access. Look over the basics of RAID-0, and remember it is designed to increase transfer performance, not seek performance.

Also, how does a Raptor do a better job of playing MP3s or pulling in downloads over your extremely slow Internet connection? Why pay $180 for the 73GB Raptor when the same $180 gets you 300GB in 7200RPM and the exact same performance in media storage tasks? Admittedly, not all users care about capacity, but I can wipe out a Raptor with music alone, let alone movies or anything else. Awfully high price to pay when the Raptor's advantage, performance, is irrelevant in this particular model.

Also, your arguments do nothing to refute the advantages of my suggested purchase of the Raptor and 300GB 7200RPM drive.

 

[djnes]

Your right about how back ups make this a moot argument. I've been working in corporate IT for a long long time, and if I said 10% of the people, even computer saavy people, actually kept regular back ups, I would be awfully generous.

 

[PaHick]

There's no way that two 7200RPM drives in RAID-0 will match the localized seek performance of a 73GB Raptor. Localized seek performance is by far the most important facet of hard drive performance on the desktop. Yes, increasing areal density does increase performance, as it makes locality more acute. However, no amount of density will overcome a deficit of more than 5ms in seek performance given roughly equal buffer strategy, for the frequent localized seeks that dominate desktop disk access. Look over the basics of RAID-0, and remember it is designed to increase transfer performance, not seek performance.

Also, how does a Raptor do a better job of playing MP3s or pulling in downloads over your extremely slow Internet connection? Why pay $180 for the 73GB Raptor when the same $180 gets you 300GB in 7200RPM and the exact same performance in media storage tasks? Admittedly, not all users care about capacity, but I can wipe out a Raptor with music alone, let alone movies or anything else. Awfully high price to pay when the Raptor's advantage, performance, is irrelevant in this particular model.

Also, your arguments do nothing to refute the advantages of my suggested purchase of the Raptor and 300GB 7200RPM drive.

First,I didnt say anything about a Raid0 array having better seek times than a Raptor did I?Second, I never said a Raptor was better at MP3's.What I did say is density and placement on the platter are just as important as seek time. Wouldnt that be the reason a 300G Maxline III is close to the performance of a Raptor?And as far as my arguments doing nothing to refute the advantages of your suggested Raptor and 300GB 7200 rpm drive setup, thats acceptable,but....where is your test I asked for(Raid0 vs. 2 single drives)? I will gladly read what you offer. And I agree, a 5ms deficit cant be overcome. But your telling me (just for example) that 2 newer drives,such as Maxline III in Raid0 cannot reach or excede the Raptors speed?

BTW...Why did you say something about an internet connection?What does that have to do with this topic?

 

[xonik]

First,I didnt say anything about a Raid0 array having better seek times than a Raptor did I?

You're right, you didn't, but you also didn't refute the well-accepted claim that seek performance is more important in a desktop usage model than throughput performance. Seek performance becomes very pertinent when you consider this observation.

Second, I never said a Raptor was better at MP3's.What I did say is density and placement on the platter are just as important as seek time. Wouldnt that be the reason a 300G Maxline III is close to the performance of a Raptor?

If by performance you mean read/write, then yes, the MaxLine III is comparable to the Raptor. However, the Raptor stands in its own league in terms of seek performance. And if you yourself considers seek performance to be (only) just as important, that puts the MaxLine III down a notch even in your own book.

But your telling me (just for example) that 2 newer drives,such as Maxline III in Raid0 cannot reach or excede the Raptors speed?

Not in seek. In read/write, sure.

 

[PaHick]

You're right, you didn't, but you also didn't refute the well-accepted claim that seek performance is more important in a desktop usage model than throughput performance. Seek performance becomes very pertinent when you consider this observation.

Again,I ask for a study or test that proves that, im learning,prove it. So instead of quoting everything else you stated, because it all pertains to seek time, please show me a test or study that shows seek time as the most important factor in these situations.


Im going to state this comparison from SR again....

Test 1x 200 GB DM+9 2x 200 GB DM+9 RAID 0
IPEAK Average Read Service Time 13.2 ms 13.3 ms
WinBench 99 Outer Zone Transfer Rate 55.3 MB/sec 109.4 MB/sec
Inner Zone Transfer Rate 31.6 MB/sec 63.1 MB/sec
SR Office DriveMark 2002 395 IO/sec 426 IO/sec
SR High-End DriveMark 2002 373 IO/sec 408 IO/sec
SR Bootup DriveMark 2002 288 IO/sec 474 IO/sec
SR Gaming DriveMark 2002 519 IO/sec 529 IO/sec

Now, both hard drives have the same seek time. When setup in Raid0, they lost .1ms in seek time. Yet, there is in some cases a dramatic increase. What did seek time have to do with that?

 

[AtomicFire]

Can we just answer the poster's question without fighting over it?

Here is my own conclusions:

1. It is NOT significantly faster in web browsing, word proccessing...etc...
2. It is SLIGHTLY faster in gaming - Loading times of course, not FPS-wise.
3. It IS more likely to fail. In a RAID0 array, if one drive dies, all your data is gone.

In other words, I have it and I love it. Windows seems to load faster, games seem to load faster, and the computer just feels quicker in general. Those thinking its a myth, that its all in my head, fine. It probably is all in my head, but it feel faster and I like it.

The one thing that we can all agree on is that failure would be bad, lose 1 drive and its all gone. I was well aware of this when I built my computer, so I do weekly ghost backups to a USB 400gb drive.

If you have two matching drives, and you do backups often, then I don't see why not.

-AtomicFire

 

[krizzle]

I had a weird week once... I run a little tech-support service in my neighborhood. My neighbor called me up and said his drives are dead, on both his and his son's comp.

Two western Digitals. Both dead, within one day of each other. Next day, my own 80gig Caviar dies! Out of nowhere! Cyclic redundancies, and a WHOLE lot of shit for me to go through before I could get some fresh corporate emails off it.

Two days later, my good friend's 80gig HD dies. And a month after, another neigbor had a caviar fly. You guessed it, WD. It's as if they were all timed to fail...

So I say that to say this: Don't use Caviars for sensitive data in RAID-0 arrays.

We all know that consumer hard drive lines are of roughly equal reliability, and anyone can come up with equally valid data and experiences to refute any claim that "brand x is junk." Shipping/handling/installation damage and environmental problems, such as heat, poor power supply, etc are much more likely to cause a drive to fail than a manufacturing defect. As such, I expect that brand recommendations will be kept positive. - DL

About the math... hmm, lesse here. a bit of differentiation, integration, aproximation... related rates... conics and fractals... aha! Yes! You have twice as much chance of data loss due to HD failure on RAID-0 as you do when running seperate. But given that the chance of hd failure is 1 in 10^5, that chance is now 2 in 10^5. Not too bad.

 

[djnes]

So I say that to say this: Don't use Caviars for sensitive data in RAID-0 arrays.

And the industry studies always show Seagate and WD as the most reliable drives. I have 11 computers between my home and my office, and I ONLY use WD drives, with the exception of one Seagate I got at a really good price. You've just shown an example of a personal opinion versus a large sampling. That applies greatly here. Personal opinion of a small few says that RAID0 is useful and performance boosting. A large sample across the enthusiast community, where these tests are repeated over and over, show that it has no real place in a desktop environment.

 

[PaHick]

A large sample across the enthusiast community, where these tests are repeated over and over, show that it has no real place in a desktop environment.

I beg to differ.Where else do you find these arguements? Only in the enthusiast community do you find such a debate.I live real close to Adelphia, and 99% of my friends and family work there.This topic isnt even a thought to most of those people.Why, because in their line of work it isnt in there best interest to run Raid0. They are more worried about security and servers than anything.I think you truely underestimate the shear number of Raid0 users and supporters in the enthusiast community. We can put links to all kind of articles showing both ways,to support and to not support the idea. And one things for certain, as its been shown before, is that before Anandtech and Storage Reviews articles came out Raid0 was very accepted in this community. And since their articles, so many debates have popped up and the main arguement of those who oppose Raid0 is touting those articles as the holy grail.Sure you have other articles that have popped up, but most are riding the coat tails of the Anandtech and Storage Review articles.For every so called scientific tests these other articles show, I can link one showing the opposite.

The article I linked for tweakers.net really showed it isnt all one sided as you would believe. Maybe I didnt prove my point any farther than that, but I asked for a test proving that the 2 single drives would be better.If your going to argue the point,show me.Bring the tests I asked for. I said before im learning just as we all are, but im not going to take what you say as truth until you show proof. I am not the only supporter of Raid0, there are many more not just in this forum, but all over this community. There are plenty of users who run Raid0 and are happy with it, despite those articles and what has been said here. Other articles and tests need to be shown , because with just the one article I linked to, it puts a shadow of doubt on many opponents arguement.

 

[m3th0dm4n]

IMO...

I run a RAID 0 with 2 x Raptors 36GB. Personally, I think for most users, the best benefit would just be a single FAST drive (ie 74GB raptor, new large cap NCQ drives). But for me, 2 highly reliable drives, with no data that I am worried about losing, I think it's worth it. It's like a JBOD but with nice read rates

Oh and yes, I love to flaunt my 100% seq. reads


Also, onboard RAID 'controllers' are a joke for the most part... best performance 'controller' I've found is the Intel SATA RAID controller on the 865/875 chipsets. I really don't like the Sil Image ones...

 

[djnes]

The article I linked for tweakers.net really showed it isnt all one sided as you would believe. Maybe I didnt prove my point any farther than that, but I asked for a test proving that the 2 single drives would be better.If your going to argue the point,show me.Bring the tests I asked for. I said before im learning just as we all are, but im not going to take what you say as truth until you show proof. I am not the only supporter of Raid0, there are many more not just in this forum, but all over this community. There are plenty of users who run Raid0 and are happy with it, despite those articles and what has been said here. Other articles and tests need to be shown , because with just the one article I linked to, it puts a shadow of doubt on many opponents arguement.

You missed the boat, honestly. I had my own test results on my web site, but they were taken down a few months ago, simply because the debate on this topic was over. I'm not trying to start a flame war here, but the tests were done by many of us on these boards, with all of our results posted. We all came to the general conclusion that it offers no performance enhancements. The myth was debunked, and the debate died. Your about 6 months to a year too late to catch them. If you venture out of this subforum, you'll only find people laughing and flaming those trying to still maintain that RAID0 is useful. I'm not sure what else you want, but facts are facts. It's a basic prinicple of the scientific method, that a test must be able to be repeated with similar results. There was upwards of 30 people doing their own testing on various systems with various drives. The only thing in common was the results. Some of the graphic design folks did some wonderful testing using Photoshop and Premiere. The fact is, it has no real place in a desktop environment, That's a fact. Accept it or not, but that's how it is.

 

[DougLite]

Once again, Tweakers used a very poor multi tasking model. The WinAmp and EMule work are much better served by an independent spindle. For such media task only one hard drive attribute matters, and that's capacity. We all know that you don't buy Raptors for capacity. I know that I can't afford a Raptor to store MP3s. 7200RPM drives serve such tasks much better. The Tweaker's article is a classic example of the benchmarks not being reflective of real world application.

This thread is tetering on the brink of earning the dubious distinction of being the first one I have to lock. Before getting passionate on this issue, I highly recommend that you review the [thread=760666]Rules[/thread] and verify that your writings stay within them. The next time I put red ink in this thread, it will be the reason why I closed it. - DL

 

[djnes]

Tweaker's article is a classic example of the benchmarks not being reflective of real world application.

Which is exactly why it took so long to debunk anyway. I had benchmarks up on my site showing results that would lead one to believe my dual Raptors in RAID0 were smoking a single one. However, when I did some real-world testing, like file transfers across my network, some scripted media files, etc, the "real" results came out.

The bottom line is, I had no performance loss, by breaking my array. I kept onhe Raptor for my system drive, and put the other one in my fiancee's computer. I am getting much MUCH more benefit out of both drives than in a RAID array. We both have 120 GB Seagate SATA drives as storage/data drives. I lost absolutely no performance by going down to a single Raptor. My only regret is that instead of buying 2 36 GB Raptors, I should have spent the money on a single 74 GB one.

 

[s_s256]

Really, I recently had to break my Raid into separate drives. Albit I lost no performace in games such as Doom 3 i noticed a slow down in the loading performace.

 

[Cyraxx]

And let the FUD-spreading continue! Someone please e-mail Anand at Anandtech, and all of the other reputable sites, along with all the print mags, like MaxPC that did the actual testing and found it DOES NOT give you any performance increase. This is unbelievable how many people still buy into a myth, despite the fact it's been debunked repeatedly. Some of you still believe in Santa Claus too, right? Tooth Fairy?

Not to mention many showed performance LOSS

 

[djnes]

Not to mention many showed performance LOSS

Exactly! Too many times, these articles ignored the fact that seek times actually went up.

 

[PaHick]

And as I showed seek times arent everything...look im not saying its for everyone. Im not saying im right. But what I am saying again is this. If you are going to argue the point,show tests. I asked how many times for a test to show that 2 separate drives are better than 2 in Raid0. Did you or DougLite show them? No, and I will not take that as truth until I see proof. And it is uncalled for to even think someone should take that as truth until proven.If you would have proven your arguement for 2 separate drives I would completely agree with you then.But you havent.

As far as going outside this forum to find Raid0 users.......how many forums do you want me to list? Believe me,djnes, outside of this forum plenty of users can be found,and this debate continues alot more than you think.Ill keep checking back, and when I see the tests I asked you to show ill comment,thank you.

 

[DougLite]

And as I showed seek times arent everything...look im not saying its for everyone. Im not saying im right. But what I am saying again is this. If you are going to argue the point,show tests. I asked how many times for a test to show that 2 separate drives are better than 2 in Raid0. Did you or DougLite show them? No, and I will not take that as truth until I see proof. And it is uncalled for to even think someone should take that as truth until proven.If you would have proven your arguement for 2 separate drives I would completely agree with you then.But you havent.

As far as going outside this forum to find Raid0 users.......how many forums do you want me to list? Believe me,djnes, outside of this forum plenty of users can be found,and this debate continues alot more than you think.Ill keep checking back, and when I see the tests I asked you to show ill comment,thank you.

I don't need to run tests to show that the hybrid setup of a Raptor and the largest 7200RPM drive you can afford blows away the 2x Raptor RAID-0 setup, even if it is somewhat slower. Why? Because you totally dismiss the obvious advantages of the hybrid setup - 150% more capacity and more flexibility in allocation/recovery. In a performance only equation, the RAID-0 setup may be 'better.' However, it is extremely foolish to ignore flexibility and capacity. I will gladly trade whatever gains from RAID-0 in desktop access patterns for the capacity and flexibility of the hybrid setup.

Of course RAID-0 gives the absolute fastest linear transfer performance possible. As such, it can increase performance in linear access. Such tasks include content creation and certain games that use large compressed archives in lieu of a myriad of small files, depending on how those programs access the disk(s). However, in no situation does RAID-0 double performance (despite double the cost), it ties both disks into one volume of reduced reliability, and it also reduces flexibility in space allocation. Furthermore, when a second 73GB Raptor for RAID-0 is compared to an independent 7200RPM drive of comparable price, one writes off over 200GB to stay in the same budget.

With a hybrid setup, one could allocate the Tweakers' multitasking burden of WinAmp playback and EMule traffic to the 7200RPM drive, reducing the Raptor's load to a typical single task desktop access pattern. This is just like AT's and SR's RAID-0 tests, which have proved there is little to no benefit in using RAID-0 on the desktop. This is precisely the goal behind having the 7200RPM drive - to make background disk access irrelevant to primary disk intensive application performance. With the 7200RPM drive servicing background tasks, the Raptor is free to service the disk intensive applications as swiftly as possible. Once again, little to no performance gain, depsite still being burdened with all of RAID-0's drawbacks. There's your test results. In a performance only world, RAID-0 is justifiable. When you consider cost effectiveness, capacity, and flexibility, its luster rapidly fades.

 

[PaHick]

DougLite, Im guessing your assuming I have always meant 2 Raptors in Raid0,which is not the case. I merely stated Raid0. I cannot see the benefit of having a Raptor and a 300G drive over 2 250's or 300's in Raid0.And once again saying something isnt proving it.Please show me a study or test that shows that 2 independant disks are better than 2 in Raid0. Ive yet to see such a test and you continually state it as being truth and that users do not need to see a test. Why not? It has to be some sort of test to show proof or it is just your opinion. I am sending you a message shortly. Youve debated this topic before which IMO was never conclusive. There remains to be doubt in my mind. I may try to message mikeblas to see how he is doing with his findings. thank you.

 

[Cyraxx]

Not that im saying much.....but how often do you see raid with non-raptor drives anymore heh.

 

[PaHick]

Believe it or not...alot! I know alot of people who have a small OS drives with 2 in Raid split into PS, game, and storage.

 

[djnes]

Believe it or not...alot! I know alot of people who have a small OS drives with 2 in Raid split into PS, game, and storage.

Just because people are doing it, doesn't mean it's a better option. You need to realize, computers are much like cars....there are ricers in computers as well. Think of RAID0 as a big-ass wing on a Honda Civic. Does it look cool (arguable) and does it give them bragging rights? Sure, I guess. Does it add performance...certainly not. In fact, performance can be degraded due to the extra weight. RAID0 is the same way. Sure it sounds cool, and makes your e-wang bigger, but does it add performance? No. Couple that with the higher seek times, and for the usual desktop tasks, performance decreases.

The biggest problem with this issue is that very few are willing to admit they were wrong. Also, one of the most common psychiatric prinicples apply, in the form of the placebo effect. Users spend all this money on two drives, and they think it "feels" faster. In reality it's not. Look, the debate is over and dead. The facts were revealed, and the proof was stated. The tests were repeated over and over by many people, but the results stayed the same. If you want to run RAID0, go for it, it's your system. That doesn't change what is fact. RAID0 offers no performance increase. Most of us long-timers have moved on to debating other things still in the air, like pagefile usage, etc. This one, however, is dead and over.

 

[PaHick]

The biggest problem with this issue is that very few are willing to admit they were wrong. Also, one of the most common psychiatric prinicples apply, in the form of the placebo effect. Users spend all this money on two drives, and they think it "feels" faster. In reality it's not.

That goes for both sides. And I like how you doubt what others say....lol. When someone says their system "feels faster" it is ok for you to say they are full of it.But on the same token they cannot, in your mind,tell you the same. In the end your going to believe what someone wrote,and im going to believe what someone else wrote. For me it has worked great and I wouldnt have it any other way. There is no placebo effect, it works. Backups are alot faster as I stated in my message to you.Placebo....lol. If it walks like a duck,quacks like a duck,must be a duck....lol. Ill continue to believe in it and use it.Maybe one day ill change my mind but for now, im happy with what ive got.Im done with this thread ,but im sure it will pop up again somewhere down the line. Always does. Take care.

 

[latief75]

I didn't expect my simple question would generate all this debate ....

good job guys ....

maybe this should be a sticky !!!!

 

[Blitzrommel]

Not sure it'd do any good if this were a sticky -- people would still probably be RAID-0 advocates.

My new Fujitsu drive that DL suggested works for me!

 

[DougLite]

^ Another satisfied customer. Fujitsu's MAU series has moved to the top of the "Absolute Fastest" category in the Buyer's Guide, owing to its higher desktop performance than the Atlas 15K II. The 15K II still makes an entry, as it now occupies the honorable mention spot formerly held by the MAU. The Atlas is still the way to go for your server work, if you have an unlimited budget

As for the sticky, I prefer to have good news at the top, such as the scuttlebutt surrounding the redesign of the core logic on the Raptor which currently occupies the top spot on this board. I am considering adding a RAID-0 hall of shame, where some of the good info on its drawbacks and myths can be displayed for all to see. However, given that ANSWERS has only managed to amass 8,000 views in the time it's been up, it won't do much good.