• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Ok, is raid 0 really that unsafe? Is it worth it?

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Originally posted by: iboomalot
by your thinking we should all be using 5400 rpm drives since you could make an arguement that 7200 and 10K drives are not worth it.

raidO is faster raptor is faster and 7200 rpm drives are faster than 5400 rpm Hard drives

the 10K speeds help transfer rates and does load faster and has better access speeds.

Raid0 gives about the same improvement a 10K vs 7.2K drive.

The hard drives are the slowest thing in the system and has only gained in capacity not speed. 5% gain is pretty good when dealing with the slowest part of the overall system.

if SSD didn't cost 2,600.00 for 4GB I would have one.

every little bit of speed helps

5% is 5% little speed gains are good but don't forget the price
 
Originally posted by: SuPrEIVIE
Originally posted by: iboomalot
by your thinking we should all be using 5400 rpm drives since you could make an arguement that 7200 and 10K drives are not worth it.

raidO is faster raptor is faster and 7200 rpm drives are faster than 5400 rpm Hard drives

the 10K speeds help transfer rates and does load faster and has better access speeds.

Raid0 gives about the same improvement a 10K vs 7.2K drive.

The hard drives are the slowest thing in the system and has only gained in capacity not speed. 5% gain is pretty good when dealing with the slowest part of the overall system.

if SSD didn't cost 2,600.00 for 4GB I would have one.

every little bit of speed helps

5% is 5% little speed gains are good but don't forget the price

Also, it's not like hard drives are, say, half the speed of the rest of your system. Going to disk (say, 10 milliseconds access time for a random read that's not cached, 20-30MB/sec. transfer) is about 1000 times slower than RAM (perhaps 5-10 nanoseconds access time, transfer rates up in GB/sec). A 5% improvement (heck, even a 50% improvement) is hardly going to make a difference unless you have an application that is totally disk-bound (such as a very, very large database that is read and written essentially randomly, or a file server hosting enormous files on a gigabit LAN).
 
if price is all people cared about you could build an entire system for what my CPU costs. I got a good deal before the price drop last week at 469.00 for my 3500+ People can build entire systems for that.

point is speed and Raid0 is faster hell a single 5400 rpm hard drive IDE 80GB is only 53.00 vs 180.00 for a raptor. The raptor might only be 15-20% faster but I sure wouldn't down grade.

hell SATA is a joke my HHD is one of, if not the fastest you can buy and only tops out at 74 MB/s transfer rate and SATA is 150 MB/s

does anyone really need a FX-53 or 3.4EE ?? no but I've had to use my old athlon 850 and its dirt slow.

IMO his money is better spend on 1 10K raptor + large drive vs two 7200 in raid0

or do this get a Raptor and backup or store your MP3s to DVD.

speed or save money its your call
 
State what you will about RAID0 but the truth of the matter is:

It feels (response) faster
It frees up IDE Ports
When you buy the latest n greatest HD and then need more room you can buy 1 nother then RAID0...
{Ex}I bought my 1st Raptor and SATA'd it then needed more space and the Drive went down in price so I RAID'd it...

Backup is necessary on any config if you wanna keep the Data assured!
 
Originally posted by: LED
State what you will about RAID0 but the truth of the matter is:

It feels (response) faster
It frees up IDE Ports
When you buy the latest n greatest HD and then need more room you can buy 1 nother then RAID0...
{Ex}I bought my 1st Raptor and SATA'd it then needed more space and the Drive went down in price so I RAID'd it...

Backup is necessary on any config if you wanna keep the Data assured!

How does it free up ide ports? I would understand it if it was because you were moving to sata ports, but you would also be occupying these once free sata ports right?
 
Raptors don't make that much sense because a PAIR of 200GB drives in RAID 1 is actually both cheaper and faster for real-world use in addition to providing redundancy.

Oh, and BTW, the difference in access time between disk and RAM is close to 250,000x, not just 1000x.
Burst transfer rate difference is more like 100x. This is why most of the time waiting on disk is from slow access time, not transfer rate, and the transfer rate benefit you get from RAID0 (with no access time benefit) is not that great. RAID 1 effectively improves the average access time.
 
Originally posted by: RockGuitarDude
Will I see a benefit from Raid 0 when encoding large mp3 files? If so, how much are we talking?

None. You'll only see a second or two difference when *loading* massive files (like, say, a new level in somegame - rather than taking 30 seconds, it might take 28.)
 
Originally posted by: glugglug
Raptors don't make that much sense because a PAIR of 200GB drives in RAID 1 is actually both cheaper and faster for real-world use in addition to providing redundancy.

How is a RAID1 mirror of older drives faster than a RAID0 Raptor setup?
 
Originally posted by: dclive
Originally posted by: glugglug
Raptors don't make that much sense because a PAIR of 200GB drives in RAID 1 is actually both cheaper and faster for real-world use in addition to providing redundancy.

How is a RAID1 mirror of older drives faster than a RAID0 Raptor setup?

Raid 1 improves effective access time, which is the most real-world performance-critical aspect of the disk subsystem, because on each random read, you only have to wait for whichever drive gets there FIRST instead of having to wait for BOTH drives.

Raid 0 only improves burst transfer rate, which has little effect on real-world applications, with no positive effect on access time. (and even some potential for a negative effect on access time).
 
Originally posted by: glugglug

How is a RAID1 mirror of older drives faster than a RAID0 Raptor setup?

Raid 1 improves effective access time, which is the most real-world performance-critical aspect of the disk subsystem, because on each random read, you only have to wait for whichever drive gets there FIRST instead of having to wait for BOTH drives.

Raid 0 only improves burst transfer rate, which has little effect on real-world applications, with no positive effect on access time. (and even some potential for a negative effect on access time).[/quote]

http://www.storagereview.com/articles/200102/20010214ST1000_vs_Escalade_5.html

And good benchmarks out there?
 
Originally posted by: CVSiN
wrong again ronald..
mirroring = no performance gain at all..
sorry this is basic computing anyone that has passed an A+ (retard cert which I have for god knows what reason other than job required it) knows this is not true..

Proving once again, that A+ certs mean nothing, and lack the depth of real-world experience.

The correct answer to that question is, of course, "it depends". It is possible, for a good implementation, to actually nearly double the read speeds of a single drive, with a mirrored RAID array. The reason why is, you have two independent spindles, both containing identical copies of the same data blocks. So if you get two read requests, instead of servicing them one after another, the RAID controller can send one of the requests to one drive, and one to the other drive, and together, they can complete in half the time. Basically, it can issue parallel read/seek requests, up to the number of drives in the mirror-set, as opposed to issuing them sequentially.

Of course, cheap and/or brain-dead RAID implementations may not do this. Those include many low-cost "software RAID" controllers for PCs. So if those are all that you are working with, then you're right, they generally offer no performance gain.

Originally posted by: CVSiN
RAID0 = striping =FAST

Potentially-higher STR, at the expense of lowered IOs/sec as compared to a different configuration of spindles.

Originally posted by: CVSiN
RAID 1= mirroring no gain other than redundancy to bit slower pretty worthless in most computing environments as most peeps that need redundancy will run RAID5 or a regular single drive solution with a good backup plan
RAID5 =striping with parity= FAST with redundancy= uber but costly 3 drives minimum required

Actually, write speeds to a RAID5 are usually lower than that of a single drive, and/or a mirror set.
Reads can be better than a single drive, but lower than a stripe set. (But can handle more IOs/sec, usually.)
 
Originally posted by: drag
Raid 0 is worthless. A waste of money for what your getting.

see storage review and their test of RAID 0 with 2 maxtor 200gig drives and a PATA Promise controller.


Is more then doubling the cost for less then 10% performance increase worth it? Not in any galaxy I am from. (although more storage space is nice, but you don't need RAID 0 for that.)

I remember anandtech had an article posting about the cons of Raid 0.
Talk very little about the pros of it. Enough said.
 
OK, lets shoot another myth. "RAID 0 is twice as likely to fail". Only with simple statistics. The real way of calculating what the failure rate does not yield half the time. Real MTBF, Mean Time Before Failure, is an observed number. MTBF with HDs is mostly a guess. The IBM Deathstar's are a good example. They sure never made it to their mean.

In calculating simple MTBF, adding a second like device into the mix makes the drive part of the equation half. But including all of the other components that could fail smooths it. Wish I could remember where in the basement those books were, but not worth the time to hunt for them.

But, in really calculating a failure, you would have the mean, the standard deviation from the mean. Say that two standard deviations from the mean is 2 years (that is some drive) and the MTBF is 8 years. That would infer that 95% of the machines would experience some failure from 12 to 8 years. It would be less than 5% of that population that experiences a failure under 8 or over 12 years. But it is stats, which just try to explain observed patterns.

Even with a single drive, you are prone to failure. It is more sensitive to failure with two drives is a vaild characterization. But twice as likely is not accurate.

Why do I use RAID0? Because 2 200GB drives are cheaper than 1 400GB drive. Two 37GB Raptors were cheaper than 1 74GB Raptor. I write big files (.5 to 15GB each). I never really like Just a Big Old Drive. 😉 I just sounds slow (queue bango music).

BTW, Statistics majors are wierd. Math majors don't like them. A math major will say 2 + 2 = 4. A statictic major will say that it is improbable that 2 + 2 is something other than 4, but won't say it is 4. Don't ask a physics major for the answer!
 
Larry the personal attack wasn't called for.. 16 years real world experience working with Companies such as US Navy, NASA JSC, Molex, Getronics, as well as many other smaller jobs building systems and doing Desktop support.. I do not pretend to be a know it all be end all of Techs but I am far from a freaking paper cert.. I have been working with Computers and Electronics with the Military and other government firms since the Commodore 64 was considered state of the art.

I'm tired of arguing with peeps on this site.
if you say you have seen a perfomance increase fine you have..
Im not gonna argue with you..
in my experience with 2 of the same exact drives it is about the same or a tad slower depending on the box.

BTW what kinda crappy RAID5 array have you built that a single drive can beat ?
scary...
oh well no more arguing here.. even though its in my nature to debate.
 
Yeah, Raid 0 is pretty stupid, unless you religiously back up your data. I've set up a LOT of RAID arrays for customers, and I swear that almost every RAID 0 or JBOD (just a bunch of disks) array that I had to set up has failed within 3 years. These are were good Seagate and IBM SCSI drives, too, not those cheesy Serial ATA consumer drives with only a 1 year warranty.

Honestly, It's gotten to the point where I'm turning away business from people who are too cheap to buy a RAID controller and at least 3 hard drives for RAID 5. It's just not worth the hassle of having to restore from backups, assuming that the customer even bothered to pay for that as well.
 
eh, i ran a RAID 0 array for a while. it was fast, for sure. but the array broke twice in a year. the hard drives were fine, just one day the computer would crash and my array would be toast. make a new array and it would work fine again for another 6 months... i lost a lot of important data that way. the second time i got sick of it and ran RAID 1 from then on out. i recommend against it unless you NEED the extra speed and you don't mind backing up all the time.
 
Originally posted by: drag
Raid 0 is worthless. A waste of money for what your getting.
see storage review and their test of RAID 0 with 2 maxtor 200gig drives and a PATA Promise controller.
Is more then doubling the cost for less then 10% performance increase worth it? Not in any galaxy I am from. (although more storage space is nice, but you don't need RAID 0 for that.)

Two 160GB HD's are cheaper than One 320gb HD, so it is actually cheaper with approx 10% performance increase. Like all data if it is on a single drive or RAID0 it still needs to be backed up if you dont want to risk losing it.
 
Ok ultra long time lurker here that wants to speak out on the RAID 0 issue, might seem a bit subjective from my own experience. Furthermore I'm away from home atm and I'm unable to show some raw data.

I've been using my current RAID array for ~ 2 years without problems. But from personal experience, the RAID drive only really shines when it isnt performing mixed Read/Write operations. When extracting huge rars from the WD and dumping that data onto the RAID array, it is significantly faster than from one drive to another. The problem I see is that foremostly, most RAIDs are set up as boot drives.

My own personal conclusion find it better to have a boot drive as a small fast HD with low seek times and use the RAID array as a program HD/storage. Another comment for those that say game loading times aren't significantly reduced, from personal experience RAID helped loading times ALOT in Planetside.

In my new comp that I'm still putting together the parts for, I'm gonna test out using 2 RAID 0 arrays. I picked up 2 Segate 80 Gb 8 Mb's the past weekend for $30 each (haha for new 5 year warranty) and am probably gonna RAID em along with the Maxtors. Just giving stuff a try.... never know what might happen heh. Personally I'm afraid of oversaturating the PCI bus. 4 Drives might be able to do it heh.
 
RAID0 is the devil.

on a home PC, unless you're doing something hard-drive intensive, I don't think that even the above-average user would notice the speed increase (I'm applying this to RAIDs across the board).

with 10K raptors in a RAID1, the only time I notice a substantial speed increase is when I'm installing software, which I don't do often enough to warrent the risk of going with a RAID0 or 5.
 
Originally posted by: ultimatebob
Yeah, Raid 0 is pretty stupid, unless you religiously back up your data. I've set up a LOT of RAID arrays for customers, and I swear that almost every RAID 0 or JBOD (just a bunch of disks) array that I had to set up has failed within 3 years. These are were good Seagate and IBM SCSI drives, too, not those cheesy Serial ATA consumer drives with only a 1 year warranty.

Honestly, It's gotten to the point where I'm turning away business from people who are too cheap to buy a RAID controller and at least 3 hard drives for RAID 5. It's just not worth the hassle of having to restore from backups, assuming that the customer even bothered to pay for that as well.

I'll beat the dead horse... but someone who goes 3 years without backing up anything deserves to lose their data. I refuse to warranty any RAID system if they have a proper backup system, if it's a business. For most home users I won't even bother with RAID unless they specifically state they understand the risks...
 
I don't know about you guys nor do I know the numbers look like there's no difference b/w RAID and non RAID. However, I have tried both and I can say that I do feel that the RAID is significantly faster. Unless you have used it you would not know other than looking at the test numbers. I have been using RAID 0 for almost 4 years now and it failed once. And that's it. Not too bad. While at that, I had HDs failed more in non RAID than in RAID.
 
Back
Top