RAID 0 - is it worth it?

[cranbers]

I remember reading an in depth article about raid and the benefits. I have used raid in action, both raid 0 and 5.

The only time there is an increase in speed is when you dish out some cash for the very expensive pci cards. Typically from what i've seen on board raid, really is pretty worthless. You will see such a marginal performance increase and add in the risk for data loss, with no way to recover I personally don't see a benefit.

One hard drive goes in raid 0, thats it, audios data. Rebuild the array and start over.

Raid 5 is your best bet if you want best of both worlds, but that requires at least 3 hard drives, 1 for parity and 2 for the striping. Once again, most likely no noticeable differences. perhaps in programs that do very heavy read/writing like video editing, compression etc.

Never hurts to try it out though. In the end its up to you.

 

[groo]

Raid 5 is your best bet if you want best of both worlds, but that requires at least 3 hard drives, 1 for parity and 2 for the striping. Once again, most likely no noticeable differences. perhaps in programs that do very heavy read/writing like video editing, compression etc.

one of us is confused as to how a raid5 works

 

[Maximus_Delta]

I have used RAID for the last 5 years on several home PC's and it is a must for me - can't say I have ever had any problems with it. I have 3 x 320GB drives currently, 2 operate in RAID 0 while the other acts as a reserve and backup drive. In my opinion, performance increase is very noticeably and I wouldn't consider going back to a single drive set-up. The benefits of a more responsive system with reduced load times outweigh the extra complexity and slightly increased chances of a failure.

As a note, I'd avoid RAID 5 unless you have the appropriate hardware controller. The Intel controllers found on the vast majority of desktop motherboards will implement a RAID 5 array in software... you now have that processing overhead and the risks associated with relying on the OS to manage the array.

 

[JackNaylorPE]

Look at it this way. In all but DTR and access time benchmarks, a Seagate 7200.11 is faster than two 7200.10's in RAID 0. Exceptions will be in boot time and in access time say for example in accessing lots of small files with a database app. SO do ya buy two 7200.10's and go RAID or replace ya 7200.10 with a 7200. 11?

The downside is that you double your chance of data loss and spending lotsa time on recovery. But if you find yourself booting ya machine every morning and sipping coffee while you wait for it to finish, then you will notice the reduced boot times. Of course if ya keep machine son 24/7 then it ain't gonna do much for ya.

As an AutoCAD user, I am very conscious of disk usage and I just bought my 1st four non SCSI desktop drives about 18 months ago. I tries RAID 0 but frankly I didn't see a real impact with what I was doing.

Gaming sees the least impact. This is a good read to see if you are likely to benefit:

http://faq.storagereview.com/tiki-index.php?page=SingleDriveVsRaid0

Here's some benchmarks:
Benchmark / One Drive / Two Drives in RAID 0 / % performance increase
SR Office DriveMark 2002 / 395 IO/sec / 426 IO/sec / 7.8 %
SR High-End DriveMark 2002 /373 IO/sec / 408 IO/sec / 9.4%
SR Gaming DriveMark 2002 / 519 IO/sec / 529 IO/sec / 1.9 %
SR Bootup DriveMark 2002 / 288 IO/sec / 474 IO/sec / 64.6 %

So if you are expecting to double performance you are going to be disappointed. If you find the 64% increase in boot time performance exciting, then that may be enough for you alone. If you keep machine on 24/7, maybe not so much. In the end, that is once booted, if you find the T & E worth a performance increase from 2 to 10%, then by all means go for it.

 

[totakeke]

I've never personally used a setup with RAID 0, but it doesn't seem like it would be worth it. From reading lots of reviews across many different websites, hard drive failures seem to be quite common. (That's why I'm using RAID 1) Aside from just a small performance boost, remember that if one drive fails, that's it.

 

[croc]

I have two 150Gb Raptor hard disks on my new setup.

I intend to put Vista on one 150GB Hard disk and XP MCE 2005 on the other.

If I was to combine these two hard disks into a single RAID 0 array, despite the fact that I will be splitting the array in half again for the two OS installs would I gain any performance boost?

Also, what pitfalls will I need to watch out for when building/using this RAID array?

I'd use raid0 if I had a dedicaed, quality raid controller. For home use, Adaptec makes good controllers, but not very cheap. At least they're portable, so will move easily from one MB to another with a minimum of hassle.

Given the amount of work that the typical MB controller off-loads to your CPU, I just don't feel its worth it. Plus, if your MB dies, you'll have issues rebuilding the raid without data loss.

I do have an external raid storage unit, (an early model readynas unit) and it performs nearly as well as some of the high-end dedicated controllers we use at work.

my 2p's...

 

[apache_lives]

I have two 150Gb Raptor hard disks on my new setup.

I intend to put Vista on one 150GB Hard disk and XP MCE 2005 on the other.

If I was to combine these two hard disks into a single RAID 0 array, despite the fact that I will be splitting the array in half again for the two OS installs would I gain any performance boost?

Also, what pitfalls will I need to watch out for when building/using this RAID array?

1) RAID 0 setups make the system feel more robust and alive
2) Data Loss from failure isn't like what people here are saying, its easier to get a virus or accidentally delete your data
3)External caddys are your best friend when they have a sync setup
4) Intel MATRIX arrays are the best setups

 

[JackNaylorPE]

1) RAID 0 setups make the system feel more robust and alive
2) Data Loss from failure isn't like what people here are saying, its easier to get a virus or accidentally delete your data
3)External caddys are your best friend when they have a sync setup
4) Intel MATRIX arrays are the best setups

1. "Feel" being the operative word. They do boot a lot faster which ya kinda lose the benefit of when you leave machines on 24/7. With RAID 0 Raptors gaining you 7.8% in apps, 9.4 % in hi end apps and 1.9% in gaming, it seems awful paltry compared to the increases offered by say the Samsung F1.
2. It's not complicated, whatever it is with one drive, it's twice that what two drives.
3. Most caddies are not rated for "everyday" usage". If one chooses this route, make sure you clarify this.
4. Like croc, I'm an Adaptec man myself.....and I too made the switch to NAS's.

 

[jackieboy]

Rapter are are only a tiny bit faster in raid 0....as they are smoking fast already....I would stick with reliability and put an operating system on each drive....

 

[russki]

Never thought I'd say this, but I agree with Mr. Naylor, PI.

Any "it feels faster" comment gets dismissed as placebo, because it flies in the face of BENCHMARKS. Storage review article is a good one to read, or try Anandtech's writeups. And by all means, avoid Tom's benchmarks which show throughput increases that relate very little to the real world usage.

qwertycopter - just because it is the #1 source of support calls it doesn't mean the technology is not up to par; it mostly means that it is grossly misused by uninformed masses. It really has no place in the desktop. Think about this; the highest margin of improvement is up to 30% [edited; I recalled 20% but it's sometimes higher] in media encoding scenarios. That's not a really wide margin, and again, that's best-case usage scenario on the desktop. But some people might consider the extra investment (ordinarily) involved in setting up RAID0 worth it for 20%. Your typical margin of improvement, though, is well below that. RAID was never meant to be on desktops.

 

[B-Unit]

I had a couple of heart stopping moments when my array went missing at startup for a couple of different reasons. Even though my data is backed up to two different locations (LAN and USB HDD) unless I had performed a backup minutes before an array went down, something would be missing.

Heheheh, I know that feeling. Tryed pushing my overclock a few times after I set up my RAID, would have to reset the BIOS after it would no longer POST, but forgot that the RAID setup was part of the BIOS. Had given up hope and was ready to reinstall before I remembered, lol.

Overall, I like having the RAID setup, I have no benchmarks, but the system just feels faster with it, although I will say I did see a bigger speed increase going from PATA to SATA than from single drive to RAID. I use to keep my system on 24/7, but with the price of electricity, I've started to shut it down while I'm away at work. Its nice to be able to fire that puppy up in <2min and be ready to go, longest part is connecting to my wireless, lol.

 

[fergie]

i have 2 150 gb raptors in raid 0 and they are definately faster than 1 at loading maps on online games and i have never had a problem.i had 2 74gb raptors in raid 0 before in my last rig for 3 years and didn't have a problem.go for it

 

[JackNaylorPE]

The 74 GB Raptors did very well in the storagereview.com reliability survey. It was the newer 150 GB models that dropped to the 12th percentile.

Raptor WD360GD - 86% (released 2nd quarter 2003)
Raptor WD740GD - 94% (released 4th quarter 2003)
Raptor WD1500 - 12% (released 1st quarter 2006)

Don't assume it's the size that makes the difference. The Raptor is now in version 4 and I'd expect all version 4's regardless of size to perform similarly. They must have swapped a part or two in the latest versions that accounts for the difference.

The last Raptor design came out 9 quarters ago which accounts for the performance loss as it's rpm advantage is eclipsed by competitor's advances in areal density over the past 2+ years. If WD plops it's new hi density 320 GB platters into a 10k design, the Raptor should retake it's performance edge.

 

[bliq]

Yup, Just create the array and make it 2 partitions when you go to install the OS. You will still see a performance gain.

Only real pitfall with RAID 0 is the increased risk of complete data loss in a drive failure. If I remember the numbers correct, your chances of complete drive failure and or data loss on a bad drive in a single drive setup is less than 1% but in a RAID 0 setup the risk is ~9%. So, back up your data!

IMO, RAID 0 is an easy way to speed up load times and increased overall 'snappyness' of the OS.
Do some searches and you can find some pretty good benchmarks.

Actually the chances on a single drive used by a consumer is probably more like >0.01% and in a two drive RAID0 array it's just double that or technically a slight bit more than double as there's a RAID controller involved that could go wonky. On the plus side, Raptors are essentially enterprise class SCSI drives refitted with SATA controllers. So they're more robust than your average hard drive.

9% failure rates would mean no one would ever run R0

 

[JackNaylorPE]

9% failure rates would mean no one would ever run R0

Maybe they shouldn't ..... considering the current Raptor's failure rate:

9% was a good number for the old 2003 version of the Raptor. But the new 2006 versions has fared far worse.

http://www.storagereview.com/WD1500ADFD.sr?page=0%2C9
"According to filtered and analyzed data collected from participating StorageReview.com readers, the Western Digital Raptor WD1500 is more reliable than 12% of the other drives in the survey that meet a certain minimum floor of participation."

Out of 168 WD 1500 Raptors in storagereview.com's reliability survey, they published the following results amounting to 25 dead drives:

Dead on arrival
Dead on arrival
Dead on arrival
Dead on arrival
Dead on arrival
Failed before 1 month
Failed before 1 month
Failed before 1 month
Failed before 3 months
Failed before 3 months
Failed before 3 months
Failed before 3 months
Failed before 1 year
Failed before 1 year
Failed before 1 year
Failed before 1 year
Failed before 1 year
Failed before 1 year
Failed before 1 year
Failed before 2 years
Failed before 2 years
Failed before 2 years
Failed before 2 years
Failed before 2 years
Failed before 2 years

That equates to:

a) a 15% chance of dealing with a dead drive within 2 years
b) a 12% chance of lost data (excludes DoA's) within 2 years
c) a 24% chance of lost data if used in RAID 0 within 2 years

And there are still 3 years to go in it's warranty period however if something gonna fail, it usually goes in 1st 2 years.


By comparison the 2003 version (698 in survey) of the Raptor had:

9 DoA's
8 Failures within 1 month
3 Failures within 3 months
12 Failures within 1 year
8 Failures within 2 years
2 Failures within 3 years
1 Failures within 5 years
13 Peeps stopped using before it died.

That equates to 43 failures out of 685 drives:

a) a 6% chance of dealing with a dead drive within 2 years
b) a 5% chance of lost data (excludes DoA's) within 2 years
c) a 9% chance of lost data if used in RAID 0 within 2 years

 

[qwertycopter]

qwertycopter - just because it is the #1 source of support calls it doesn't mean the technology is not up to par; it mostly means that it is grossly misused by uninformed masses.

I was referring to their quality and reliability, not performance.

 

[Maximus_Delta]

To give some tangible figures:

Benchmarked WD 320GB on it's own with HDDTune and avg around 80-90 MB/sec.
Benchmarked 2 x WD 320GB in RAID 0 and avg 130-140 MB/sec.

It's synthetic yes but that is a significant increase and I certainly do notice my system is faster with RAID 0. Used RAID for years and year, never would build a personal machine without it now, and never had any major problems.

 

[russki]

Maximus, try SiSoft Sandra, that would show you a great improvement, also. I could do a search for synthetic benchmarks that show you the throughput benefit of RAID0, I'm sure there are many...

People like you kill me.

 

[JackNaylorPE]

Yes it will boot faster and it will do a DTR synthetic faster. But running gaming benches don't show any significant improvements. As a comparison, buying a 500 GB 7200.10 last year and a 7200.11 this year would have netted a user a 34% increase.

And it does double your failure rate.

Looking at the last Caviar in the storagereview.com reliability survey the 500 GB RE2 shows that it finished in the 4th percentile. IOW, 96% of the drives out there proved to be more reliable. With 187 drives in the 500GB RE2 reliability survey:

13 drive still running after 2 years
139 drives still running after 1 year (includes the 13 above)
7 DOA's
22 failures within one year
3 failures between 1 and 2 years

So that means:

17+% failure rate so far with most of the drives just one year old.
13+% of the people using them "lost data"

About 1 in 4 people using RAID 0 would experience a failed array within the 1st year alone.

 

[Maximus_Delta]

Maximus, try SiSoft Sandra, that would show you a great improvement, also. I could do a search for synthetic benchmarks that show you the throughput benefit of RAID0, I'm sure there are many...

People like you kill me.

The point I am making is that it is 'faster', even if not quite what the benchmarks show. Windows, games etc load quicker... big file transfers happen faster and it copes better multi-tasking with heavy disk work. I notice it to point of thinking that non modern RAID HDD setups are slow and bog down easily when I use them.

The thread was, is RAID 0 worth it? My answer is yes for justifiable reason - sorry if that's got your back up.

As a note I currently use the server versions of the WD discs which have 5 year warranty - I am sure they are more reliable than a 25% chance of my array failing this year. Anybody with RAID 0 should back up frequently - I have a few sync routines that do this automatically...

 

[apache_lives]

To give some tangible figures:

Benchmarked WD 320GB on it's own with HDDTune and avg around 80-90 MB/sec.
Benchmarked 2 x WD 320GB in RAID 0 and avg 130-140 MB/sec.

It's synthetic yes but that is a significant increase and I certainly do notice my system is faster with RAID 0. Used RAID for years and year, never would build a personal machine without it now, and never had any major problems.

Agreed mate

As for benchmarks showing little/no improvement, the benchmarks are synthetic and dont show real world performance, its the usability, same as Intel's HT, NCQ, SATA, 32mb cache (HDD), Ready Boost and many other technologies and specs - its not just high benchmark figures that make a computer "fast" and "robust". Why dont we all buy IDE hdd's and disable AHCI/NCQ and every other minor feature that "doesnt perform any better on synthetic benchmarks" - EVERY LITTLE BIT HELPS.

Failure rates? Most people are stupid and don't handle hardware correctly, short it out, drop it, misuse it or blame it for faults etc - thats something your figures dont show.

I trust my data on a Raid 0 hdd setup as much as i trust a single hdd setup - that is, i do not trust either. My data is backed up onto external drives as well as my server with raid1, all my pcs have raid 0 setups and no important data is stored on any of them, not that iv ever lost a Raid0 array.

Let the single drive users live with there little computers, we can use our pcs knowing that the bottlekneck is nearly half of everyone elses - the minimum spec is doubled.

 

[JackNaylorPE]

As a note I currently use the server versions of the WD discs which have 5 year warranty - I am sure they are more reliable than a 25% chance of my array failing this year. Anybody with RAID 0 should back up frequently - I have a few sync routines that do this automatically...

There is no real difference between the drives. Take for example the "enthusiast Raptor and the server Raptor". They are the identical drive except for that:

a) One has a metal cover and one a clear plastic cover

b) The server version has a 1,200,000 MBTF, the other a 600,000 MBTF. Now logic would dictate that this meant that the server version was better somehow. Not at all, the only difference between the two drives is how the MTBF is calculated. By changing the number of hours that the drive is expected to run, they essentially double the MTBF with the argument being, I suppose, that the more frequent start ups of a home drive puts more shock loading on moving parts than one that runs smoothly 24/7.

If you buy a "server" hard drive and use it in a home machine, the rated MTBF drops in half. Take a home use HD and put it in a server and its rated MTBF doubles.

Warranty to a certain extent is determined by price. If you buy something in Best Buy that has a warranty of one year, and you buy a 2 year 3rd party warranty, does that in and of itself make the product better. Same thing with manufacturer warranty. Very often the only difference between the server version and the home version is the:

a) Label
b) MTBF rating (and again, you can double any drive's rating simply by calling it a "server" drive.
c) Price

By charging the higher price, they pay for what is in effect an insurance policy against future warranty claims.

 

[groo]

waranties on something like this are like rebates. most people dont go through the hassle.

 

[JackNaylorPE]

I have obtained replacements for every failed part under warranty and every rebate I have ever been offered, with one exception. My last rebate was a $2,000 straight rebate and $2,000 trade in rebate on a $11,995 plotter which was "on sale" for $9,995.....Wound up saving me $6k over retail.

The exception was a failed HD that WD refused to cross ship 3 years ago. I used keep spare HD's handy. The spare goes in and the warranty replacement becomes my new spare. But w/ RAID 1 on the NAS, a failed HD doesn't mean that much anymore.

 

[Maximus_Delta]

JackNaylorPE, I can see how that would work with the higher prices just to cover the higher cost to them of providing a 5 year warranty. I won't worry too much about the home vs server editions in future then...