Sign in with
Sign up | Sign in
Your question

The Best in Enterprise Hard Drives

Last response: in Storage
Share
April 2, 2007 11:09:11 AM

We put 13 SAS and SCSI hard drives to the ultimate test. And we also launch our new Enterprise HDD Charts!

More about : enterprise hard drives

April 2, 2007 1:58:29 PM

Not a bad article for those in the market for a SCSI/SAS drive .... I wish that you would have included a Raptor in the graphs and charts for reference.
April 2, 2007 2:13:40 PM

I would have liked to seen how well the FC drives are doing in comparison to the new SAS segment.
Related resources
April 2, 2007 2:42:15 PM

That IBM drive looked almost to good to be true, it had a Windows XP startup time of just 3.6 seconds.
April 2, 2007 3:16:28 PM

Oh my thats a problem.

Thats a performance benchmark. I believe higher is better for that graph.

If you check the bottom of the graph it's measuring MB/s not startup time in seconds.
April 2, 2007 3:31:39 PM

Don't burst his bubble, we all need to dream. XP loading in 3.6 sec. ah, nirvana.
April 2, 2007 5:29:07 PM

Raptor isnt there because this is an Enterprise drive comparison.

It focuses on SAS and SCSI, NOT SATA.
April 2, 2007 5:53:10 PM

Quote:
Raptor isnt there because this is an Enterprise drive comparison.

It focuses on SAS and SCSI, NOT SATA.


SATA makes it into Enterprise servers.... in fact HP pushes those just as much as the SAS based drives.
April 2, 2007 6:31:32 PM

Quote:
Raptor isnt there because this is an Enterprise drive comparison.

It focuses on SAS and SCSI, NOT SATA.


SATA makes it into Enterprise servers.... in fact HP pushes those just as much as the SAS based drives.

Not very true at all. Yes, they do have servers with SATA drives, but those are systems that don't require high disk throughput. For example, take the DL380, a good general purpose server be it a small database server, file serving, or web services, and it has the option for SATA drives, but nobody would choose that if they cared about throughput.

When you look at either blades or the DL100 series, those are very entry level servers, which are good for lightweight services for disk or clustered services, where you are looking for processor power.
a b G Storage
April 2, 2007 7:13:45 PM

Lol spend 8k on a server and put 50$ sata drives in it.
April 2, 2007 7:52:20 PM

True, for any high-throughput application (SQL server, large file server, etc.) the SAS/SCSI drives benefit you greatly.

But sometimes, SATA is an overwhelming favorite.

Recently I needed to put together 6TB of space. I just couldn't do it with SCSI/SAS - I would have spent $25K+ just on the drives. With Seagate Barracuda ES 750GB SATA drives, I only spent about $4K.

Since they're all going into a decent RAID-5 SAN unit on iSCSI, throughput isn't an issue (the SAN can max out the iSCSI bandwidth with room to spare). And since they're Barracuda ES series (enterprise series), they should be in the reliability range of higher-end drives.
April 2, 2007 8:47:52 PM

Quote:
Raptor isnt there because this is an Enterprise drive comparison.

It focuses on SAS and SCSI, NOT SATA.


Funny how westerndigital.com has Raptors listed under ENTERPRISE DRIVES...they should have been in this ENTERPRISE DRIVE comparison regardless of interface. Not everyone requires more bandwidth, only a fast drive.
April 2, 2007 8:51:42 PM

Quote:
Raptor isnt there because this is an Enterprise drive comparison.

It focuses on SAS and SCSI, NOT SATA.

WRONG :!: A quote from the link below
Quote:
Some hard drive companies design desktop-class drives with the SATA interface. To meet the demands of enterprise storage, WD is going one better. WD is the only company combining a 10,000 RPM enterprise-class mechanical platform with the SATA interface to meet all the demands of the enterprise environment—reliability, performance, and reduced cost.

Applications
Servers, network attached storage, scientific computing, video surveillance, enterprise backup, document and image management, digital video.

http://www.westerndigital.com/en/products/products.asp?...
April 2, 2007 9:15:25 PM

Quote:
WRONG :!: A quote from the link below
Some hard drive companies design desktop-class drives with the SATA interface. To meet the demands of enterprise storage, WD is going one better. WD is the only company combining a 10,000 RPM enterprise-class mechanical platform with the SATA interface to meet all the demands of the enterprise environment—reliability, performance, and reduced cost.

Applications
Servers, network attached storage, scientific computing, video surveillance, enterprise backup, document and image management, digital video.

http://www.westerndigital.com/en/products/products.asp?...

WD may market it that way, but it's just marketing. If we believe that, then we should also accept every other manufacturer's claim for their products.

As a replacement for our current database server, there was no way I was going to throw 112 SATA (called FATA on our SAN) drives into an enclosure. They simply do not perform as well as their opponents in the same configuration under the same testing. What's with the huge font anyway?

Quote:
Not everyone requires more bandwidth, only a fast drive.


Not sure what you mean there. Load up a hundred or so Raptor drives and you still cannot fill the bandwidth you have. That's a drive limitation, and that's why SATA/FATA drives are out of the picture.
April 2, 2007 9:19:40 PM

Quote:
True, for any high-throughput application (SQL server, large file server, etc.) the SAS/SCSI drives benefit you greatly.

But sometimes, SATA is an overwhelming favorite.

Recently I needed to put together 6TB of space. I just couldn't do it with SCSI/SAS - I would have spent $25K+ just on the drives. With Seagate Barracuda ES 750GB SATA drives, I only spent about $4K.

Since they're all going into a decent RAID-5 SAN unit on iSCSI, throughput isn't an issue (the SAN can max out the iSCSI bandwidth with room to spare). And since they're Barracuda ES series (enterprise series), they should be in the reliability range of higher-end drives.


Very true. If your host interface is limited to a theoretical maximum (125MB/sec for GigE) less than what the drives can deliver, there's no point in increasing drive throughput. I'm referring to a SAN where you are running 2 and 4Gb connections.
April 2, 2007 9:23:10 PM

Interesting conversation.

I recently rolled out a db server that back ends for an e-commerce application. We had an internal discussion about SCSI vs. SATA and based on an article we found about some research Microsoft did for their mapping project we chose SATA figuring if the in-depth analysis MS did and wrote a white paper about made it good enough for that type of project then SATA was probably appropriate for us to consider.

The server that was being replaced was an Intel server with 2.8Ghz Hyperthreaded Xeons and 4Gb RAM with RAID 10 on Seagate 10K Cheetahs. The new one uses 2 X Dual Core Xeons with the same amount of RAM (probably faster, but I'd have to check) and RAID 10 on Raptors.

The new server is clearly faster than the older one but under heavy I/O load it is slower. We hang a USB drive off the server to do hourly db snapshots and this even takes slightly longer.

My vote is for SCSI or SAS, not Raptor in any server I really care about I/O on. Anything I don't really care about I/O on, I probably wouldn't use a Raptor on anymore because of the cost difference even though I am generally a fan of Raptors.
April 2, 2007 9:35:00 PM

I was just jerking sandmanwn's chain, because he bolded the "NOT SATA". WD gives a five year warranty and and insists that the Raptor contains enterprise class components. What do you know, that I don't, that allows you to make the judgment that the Raptor is not an enterprise class drive and that it is just marketing on WDs part?
April 2, 2007 9:37:28 PM

The fact they claim enterprise performance is what I'm against. An enclosure of SATA drives just doesn't match up performance-wise. For some applications, it fits just fine, even in the enterprise. But for high I/O operations, it doesn't cut it.

What do you know, that I don't, besides WD's claims, that makes it an enterprise drive?
April 2, 2007 9:47:17 PM

Quote:
The fact they claim enterprise performance is what I'm against. An enclosure of SATA drives just doesn't match up performance-wise. For some applications, it fits just fine, even in the enterprise. But for high I/O operations, it doesn't cut it.


Perhaps if the Raptor had been included in the review, this would be proven? :roll:
April 2, 2007 9:50:57 PM

Quote:
Perhaps if the Raptor had been included in the review, this would be proven? :roll:


Yeah, but what can we do? :D  Keep in mind, their tests may not reflect your usage. So, what may be true in one scenario is entirely false in another.

For example, comparing single drive's performance will yield you very little difference. However, using multiple drives in a different scenario will show you the differences in performance.
April 2, 2007 9:55:44 PM

Some of the drives tested fall into the performance area of the Raptor. I believe that enterprise class relates more to reliability. The Raptors should have been in the article anyway.
April 2, 2007 10:04:23 PM

Quote:
Some of the drives tested fall into the performance area of the Raptor.


Agreed.

Quote:
I believe that enterprise class relates more to reliability.


What would you do if tasked to spec out a high I/O server? You can't just eliminate the performance factor. If reliability is the main point behind enterprise class, then why test the performance? The reason is simple: when someone is given a budget, they go with the highest performance for the price first and foremost. Nobody that I have seen goes to the reliability of a drive first, then performance second.

Quote:
The Raptors should have been in the article anyway.


Given how strongly people feel about the Raptor line, I agree. It would have made a nice point.
April 2, 2007 10:29:03 PM

I'm not in IT so I don't know what the parameters are for selecting drives, I'll just take your word for it. The real reason that Raptors should have been in there is because this is a mostly enthusiast site and it would have been nice to see. However, in THGs defense there are plenty of articles that address the Raptor. This includes a recent one on RAID0 vs. the Raptor. Anyone that wants to compare can,so it's not really that big of a deal. I only jumped in when the Raptor was called desktop class. I have one and I got robbed for it :cry:  so nobody gets to piss on my Raptor :lol: 
April 2, 2007 10:45:49 PM

Quote:

Funny how westerndigital.com has Raptors listed under ENTERPRISE DRIVES...they should have been in this ENTERPRISE DRIVE comparison regardless of interface. Not everyone requires more bandwidth, only a fast drive.


You can 'think' whatever you like, but just because an OEM calls a part an Enterprise part, does not make it so, just because they state it. The reason WD calls their Raptors 'Enterprise drives' is because they have a 10K RPM spindle rate, have a longer MTBF, and a 2.5 platter,that is basically all. They would not even compare to a true 10K RPM drive, because they do not handle vibrations the same way a 10k RPM SCSI Enterprise would, and thus degrade in performance. Also, Enterprise SCSI drives tend to live much, much longer, and have a long proven track record.

As for your comparrisons, the Raptors get around 74MB/s throughput, and there are claims, in this article that some of these SAS drives are getting around 128MB/s. The speed differences should be clear, if Toms hardware data is accurate.

Thinking a HDD is going to boot Windows, or any other OS faster than a SSD is plain . . . uh . . . silly. Solid state drives can not boot Windows at 3.6 seconds, what makes you think ANY HDD, with access times much greater, is going to boot any faster ?! The plain fact of the matter is: it wont.

SATA HDDs have their place, even in enterprise, but calling any SATA drive 'Enterprise' is borderline criminal. There is a reason why enterprise machines, that NEED performance, and a lot of it, use SCSI (and more and more in the future SAS) exclusively. Matter of a fact, there have been studies done, that prove that the ES Seagates, and other 'enterprise' SATA drives offer no real benifit over regular HDDs in this situation. I am all for paying extra for better parts, but the proof is not there.
April 2, 2007 10:46:02 PM

I'm pleased to see the inclusion of Enterprise Hard Drives in the THG reviews. Well done, and with great information as usual.

I would also encourage THG to glance at Solid State Storage solutions for enterprise in a future article. Although pricey, these solutions are something my company are looking at for the sheer throughput capability they have. As we consolidate to centralized facilities, throughput will be a big consideration. The price is far out of our reach at the moment, but as the prices fall this may be a reality within the foreseeable future.

Keep up the great work.
April 2, 2007 11:12:20 PM

Quote:
I'm not in IT so I don't know what the parameters are for selecting drives, I'll just take your word for it. The real reason that Raptors should have been in there is because this is a mostly enthusiast site and it would have been nice to see. However, in THGs defense there are plenty of articles that address the Raptor. This includes a recent one on RAID0 vs. the Raptor. Anyone that wants to compare can,so it's not really that big of a deal. I only jumped in when the Raptor was called desktop class. I have one and I got robbed for it :cry:  so nobody gets to piss on my Raptor :lol: 


I agree - the Raptor should have been in there from a PC enthusiast angle.

If it makes you feel any better, from the reviews I have seen the Raptor actually does extremely well against 15K SCSI drives for desktop usage, but gets pissed on from a great height when it comes to server-oriented benchmarks.
April 2, 2007 11:18:54 PM

How does a SCSI enterprise class drive handle vibrations? Do they have special mechanisms in place that the Raptor doesn't? The drive in the link below is going to be my next desktop drive. Unfortunately I'm not sure it's fast enough for my needs e.g., browsing, word etc.

Texas Memory Systems - Products - RamSan-400 Solid State Disk
http://www.superssd.com/products/ramsan-400/?gclid=CPDv...
April 2, 2007 11:35:30 PM

Quote:
If it makes you feel any better, from the reviews I have seen the Raptor actually does extremely well against 15K SCSI drives for desktop usage, but gets pissed on from a great height when it comes to server-oriented benchmarks.


I really can't complain about my Raptor. However I was a little irritated when I found out that automatic acoustic management was turned on. The link below I showed how to check it and turn it off. This technique works for other drives as well.
http://forumz.tomshardware.com/hardware/modules.php?nam...
April 3, 2007 12:27:34 AM

Quote:
How does a SCSI enterprise class drive handle vibrations? Do they have special mechanisms in place that the Raptor doesn't?


Better components + better build quality + mo money = SCSI enterprise drive :p 
April 3, 2007 1:48:04 AM

So there are no special mechanisms in place? I thought not.
April 3, 2007 3:35:33 AM

Quote:
So there are no special mechanisms in place? I thought not.


google, read, understand. Many people out ther have written articles on the matter, ask any datacenter monkey why he would never mix SAS and SATA in the same enclosure. They use the same interface, they both work on SAS controllers.

Dont expect everyone, or even ANYONE to do your homework for you. Google. Read. Understand.

By the way, to the guy who CLAIMED the Raptors stand up fairly well to a 15k RPM SAS. Send me some of that good stuff you have been smoking . . . I too can use a break from reality....
April 3, 2007 10:58:55 AM

Quote:
By the way, to the guy who CLAIMED the Raptors stand up fairly well to a 15k RPM SAS. Send me some of that good stuff you have been smoking . . . I too can use a break from reality....


Read carefully. You're twisting things around. That "guy" stated:

Quote:
If it makes you feel any better, from the reviews I have seen the Raptor actually does extremely well against 15K SCSI drives for desktop usage, but gets pissed on from a great height when it comes to server-oriented benchmarks.


And it is totally accurate. If you place a Raptor, or Ultra320 SCSI, or even a single SAS drive in a workstation, you will notice very little difference. The major differences come into play when you have them in an array and many, many users are using the system at once. A single game or app is not going to be affected terribly between a Raptor or SAS drive.
April 3, 2007 1:05:27 PM

Quote:
By the way, to the guy who CLAIMED the Raptors stand up fairly well to a 15k RPM SAS. Send me some of that good stuff you have been smoking . . . I too can use a break from reality....


As belvdr says, I merely iterated that Raptors do very well against 15K SCSI in desktop benchmarks. A lot of people think that just because 15K SCSI drives have vastly superior seek and STR that they will be vastly superior on the desktop. This is not the case.

Check this article out:

http://www.storagereview.com/articles/200601/WD1500ADFD...
April 3, 2007 1:15:10 PM

Quote:
So there are no special mechanisms in place? I thought not.


Zorg - you “thought” wrong – what part of “Better components + better build quality” do you not understand. That is the special mechanism. Honestly, there is a difference in the design and build quality of an enterprise drive. This is why these devices cost substantially more than their consumer SATA counterparts, and why people buy them for very specific applications.

:roll: Do you think a girl is pretty based solely on her bust-waist-hip measurements? :roll:

You can buy a hundred bolts with a certain diameter, thread pitch, length etc for $2.00, and Pratt and Whitney comes along and buys the “same bolt” for $86.00 each – yes $86.00 each. You think you got a deal. Pratt and Whitney made an investment. Why?

The bolts that Pratt and Whitney bought are going into the gas turbine of a $160 million dollar aircraft that will carry hundreds of passengers at a time, and each bolt will be subject to critical stress, shock, corrosion and extreme heat-cool cycles. So Pratt and Whitney pays a premium for special materials like inconel or nonel and the special machining to make the bolt, and then they pay a guy with a lot of training and experience to physically verify and inspect every step in the manufacture and installation of that bolt, as well as test every single bolt.

To you, it is just a bolt, and maybe you got a deal or maybe you did not. You have no idea if somebody sold you counterfeit crap, or if the machining tolerances slipped - weakening the bolt threads by 15% or if any of a hundred other factors made the bolts you bought worthless. To Pratt and Whitney, an $86.00 bolt is cheap and they pay to know that their bolts will perform as specified, because if one fails, it could kill hundreds of people, expose the company to lawsuits, ruin their reputation and drive them out of business. People used to buy the cheap bolts, many still do – then their bridges fall down, buildings collapse and airplanes fall out of the sky. :roll:

As Yyrkoon said “Google. Read. Understand.”
April 3, 2007 2:01:47 PM

Quote:
I'm not in IT so I don't know what the parameters are for selecting drives, I'll just take your word for it. The real reason that Raptors should have been in there is because this is a mostly enthusiast site and it would have been nice to see.

WRONG
First off you arent even in the industry and you have the galls to come on here and tell people with mounds more experience than you that they dont know what they are talking about. What an idiot!

Who said this was an enthusiast site. I dont see any signs anywhere saying enthusiasts only, do you???? From what I can tell it seems to be pretty much a hardware review site.

No one, and Ill repeat that again for slow people like yourself, NO ONE in the real "Enterprise" sector uses Raptors for mission critical servers. SATA drives are for the extremely LOW end of the enterprise market. They go into those little rinky dink 1U and 2U webservers that get little to no traffic.

If you dont mind, pardon the rest of us while we have some decent discussions about real ENTERPRISE class drives.
April 3, 2007 5:18:18 PM

Quote:
I'm not in IT so I don't know what the parameters are for selecting drives, I'll just take your word for it. The real reason that Raptors should have been in there is because this is a mostly enthusiast site and it would have been nice to see.

WRONG
First off you arent even in the industry and you have the galls to come on here and tell people with mounds more experience than you that they dont know what they are talking about. What an idiot!

Who said this was an enthusiast site. I dont see any signs anywhere saying enthusiasts only, do you???? From what I can tell it seems to be pretty much a hardware review site.

No one, and Ill repeat that again for slow people like yourself, NO ONE in the real "Enterprise" sector uses Raptors for mission critical servers. SATA drives are for the extremely LOW end of the enterprise market. They go into those little rinky dink 1U and 2U webservers that get little to no traffic.

If you dont mind, pardon the rest of us while we have some decent discussions about real ENTERPRISE class drives.
Based on what I have seen over several years in my opinion this site is primarily an enthusiast site. There are people from all walks of life here. It is not primarily an IT professionals site. So, get down off your high horse.
Quote from pro SAS article:
Quote:
AB: Best reliability: SAS is designed for enterprise-class drives and thus offers extremely high reliability in spite of its lowest cost of ownership. Complete fault tolerance with dual porting and full SCSI failure prediction features add to this reliability.
HS: Right you are; SAS is designed for enterprise applications. SAS servers and SAS storage systems are designed to accept SAS drives AND SATA drives. So the industry consensus is SATA is viable in the enterprise (if SATA weren't viable for enterprise systems, then the industry would not have collaborated to create the SAS-SATA compatibility as an industry standard). Western Digital has an established track record of producing enterprise-class disk drives (that happen to have a SATA interface). And these drives have been widely accepted by the market. So in the future we expect SAS systems to deploy; they will accommodate both SAS drives and enterprise SATA drives. And the market will determine how many of each are manufactured.
Quote:
There is a very different market that requires high capacity with reliability and affordability. Characterising SATA drives as non-mission critical is an un-informed statement. WD Enterprise SATA drives have a 1.2 million hours MTTF rating, exactly the same rating as Fujitsu’s latest SAS enterprise drive, the MAV2073.


Link
Techworld.com - SAS or enterprise SATA drives?
http://www.techworld.com/storage/features/index.cfm?fea...

So it appears that the point is arguable. All I am saying is that it would have been nice to see the disparity between The Raptors and the SCSI drives. I believe all this talk of the Raptors not being enterprise class in reliability is total BS as was stated in the quote above. Also you don't hit anyone with bold and I won't hit you with bold. Have a nice day.
April 3, 2007 5:23:12 PM

Quote:


Not everyone requires more bandwidth, only a fast drive.


Not sure what you mean there. Load up a hundred or so Raptor drives and you still cannot fill the bandwidth you have. That's a drive limitation, and that's why SATA/FATA drives are out of the picture.

Mearly that Raptors have their place in this comparison as WD claims they are up to the task. Also, not every "Enterprise" situation will require SAS/SCSI drives. A small office can be served by a SATA RAID just fine.

I do suppose a less broad definition of "Enterprise" could clear up some of the arguments here. ie.. It could be argued that RAID 0 is NOT RAID.

Quote:

You can 'think' whatever you like, but just because an OEM calls a part an Enterprise part, does not make it so, just because they state it.



Are the Raptors the best drives? No, this STILL does not change the fact that they should have been in this comparison..which was my argument all along.

WDs claims.. "With 1.2 million hours MTBF, 3 Gb/s SATA technology and best-in-class vibration tolerance, these drives offer the best combination of reliability, high capacity and performance for enterprise applications."

"Built for speed! WD Raptor is a unique class of hard drives that matches SCSI reliability and performance while providing simplified connectivity – all at less cost than Parallel SCSI drives."

WD used to make SCSI drives, I own and use some of them still. Their claims about Raptors should have been proven or disproved in this article. Since vibrations can cause so much trouble and being able to handle them is what sets "Enterprise" drives apart from "desktop" drives, then Toms SHOULD have set up a vibration test.
April 3, 2007 5:55:39 PM

Quote:
So there are no special mechanisms in place? I thought not.


google, read, understand. Many people out ther have written articles on the matter, ask any datacenter monkey why he would never mix SAS and SATA in the same enclosure. They use the same interface, they both work on SAS controllers.

Dont expect everyone, or even ANYONE to do your homework for you. Google. Read. Understand.

By the way, to the guy who CLAIMED the Raptors stand up fairly well to a 15k RPM SAS. Send me some of that good stuff you have been smoking . . . I too can use a break from reality....
I like that argument
Quote:
Better components + better build quality + mo money = SCSI enterprise drive :p 
It is not my responsibility to Google anything. It is the responsibility of the individual that made the claim to back it up with some proof. I'm surprised that you were not aware of that.
April 3, 2007 6:16:40 PM

Quote:
So there are no special mechanisms in place? I thought not.


Zorg - you “thought” wrong – what part of “Better components + better build quality” do you not understand. That is the special mechanism. Honestly, there is a difference in the design and build quality of an enterprise drive. This is why these devices cost substantially more than their consumer SATA counterparts, and why people buy them for very specific applications.

:roll: Do you think a girl is pretty based solely on her bust-waist-hip measurements? :roll:

You can buy a hundred bolts with a certain diameter, thread pitch, length etc for $2.00, and Pratt and Whitney comes along and buys the “same bolt” for $86.00 each – yes $86.00 each. You think you got a deal. Pratt and Whitney made an investment. Why?

The bolts that Pratt and Whitney bought are going into the gas turbine of a $160 million dollar aircraft that will carry hundreds of passengers at a time, and each bolt will be subject to critical stress, shock, corrosion and extreme heat-cool cycles. So Pratt and Whitney pays a premium for special materials like inconel or nonel and the special machining to make the bolt, and then they pay a guy with a lot of training and experience to physically verify and inspect every step in the manufacture and installation of that bolt, as well as test every single bolt.

To you, it is just a bolt, and maybe you got a deal or maybe you did not. You have no idea if somebody sold you counterfeit crap, or if the machining tolerances slipped - weakening the bolt threads by 15% or if any of a hundred other factors made the bolts you bought worthless. To Pratt and Whitney, an $86.00 bolt is cheap and they pay to know that their bolts will perform as specified, because if one fails, it could kill hundreds of people, expose the company to lawsuits, ruin their reputation and drive them out of business. People used to buy the cheap bolts, many still do – then their bridges fall down, buildings collapse and airplanes fall out of the sky. :roll:

As Yyrkoon said “Google. Read. Understand.”

The bolts, that Pratt & Whitney paid additional money for, have different characteristics. Pratt and Whitney put specs out for the bolts to have a minimum strength, corrosion resistance etc. These things are all quantifiable. Pratt and Whitney didn't just say well if we pay a small fortune then the bolts will be better. The same goes for hard drives. It is possible that the drives cost more due to the SCSI design given that you get additional features. That is fine but the old "it costs more so it is made with better parts" is rubbish. As you can see from the quote in my post above the Raptor has the same MTBF as the Fujitsu enterprise class SCSI drive. MTBF is the only way, of which I am aware, to rate the reliability of a drive. Don't get me wrong I am not trying to say that the Raptor or SATA in general is preferable or even as good in many respects as SCSI/SAS. I am saying that the ridiculous blanket statement that SCSI drives are more reliable because they cost more does not hold water. Given that Raptors can and are used in small networks, I just thought it would be nice to see a comparison.
April 3, 2007 6:36:05 PM

Quote:
Link
Techworld.com - SAS or enterprise SATA drives?
http://www.techworld.com/storage/features/index.cfm?fea...

Is this a joke?

The guy tries desperately to make a comparison, shows no benchmarks whatsoever, and cant figure out access times so scews the results by using seek times and you want to show this as some sort of proof??????????

Hell I had some SAS vs SATA benchmarks waiting for some sort of slick response from you but this post isnt even worth the effort of showing you the differences. I guess youll just have to go google and learn like others are imploring you to do.
April 3, 2007 8:11:57 PM

18 years IT: DEC VMS, Sun OS 3+, SGI IRIS, UNIX, and even Win3.11 up to current Linux, OSX, VISTA and Solaris.

But still don't trust me and do the research yourself.

Enterprise class HDD is a marketing ploy; MTBF is an educated guess; don't believe either.

Go with hard measurable facts:Replacement policy (cost and speed)
Warranty
Performance

Replacement Policy first because this has the most effect on downtime and lifetime price.
Warranty second because this affects the lifetime price.
Performance is last because the top drives from all of the top manufacturers perform similarly and because different roles require different performance factors.


Check out a recent Carnagie Mellon University study:
Official Peer reviewed research paper: http://www.usenix.org/events/fast07/tech/schroeder/schroeder_html/index.html
Very nice summary at Daily Tech: http://www.dailytech.com/article.aspx?newsid=6404

To quote that article and the paper:
Enterprise Class:In our data sets, the replacement rates of SATA disks are not worse than the replacement rates of SCSI or FC disks. This may indicate that disk-independent factors, such as operating conditions, usage and environmental factors affect replacement rates more than component specific factors.
MTBF:Garth Gibson, associate professor of computer science at Carnegie indicated that the study was proof that MTBFs are not a reliable way of measuring drive quality.
also:It is interesting to note that even Google's own data centers use mainly SATA and PATA drives.

Also, this has been my personal experience for the past 18 years as well.

To re-iterate:
Go with hard measurable facts:Replacement policy (cost and speed)
Warranty
Performance (based on actual need)
April 3, 2007 10:32:23 PM

Quote:
Link
Techworld.com - SAS or enterprise SATA drives?
http://www.techworld.com/storage/features/index.cfm?fea...

Is this a joke?

The guy tries desperately to make a comparison, shows no benchmarks whatsoever, and cant figure out access times so scews the results by using seek times and you want to show this as some sort of proof??????????

Hell I had some SAS vs SATA benchmarks waiting for some sort of slick response from you but this post isnt even worth the effort of showing you the differences. I guess youll just have to go google and learn like others are imploring you to do. The only proof that I was using that article for is that Raptors are being used in place of SCSI/SAS and more importantly that the MTBF is the same as the Fujitsu SCSI enterprise class drive. Take MTBF or leave it, I have not seen any figures showing that Raptors are any more prone to failure than SCSI enterprise class drives. I asked what mechanism in the drive itself makes SCSI drives less prone to failure and the answer that I got was "They cost more therefore they are made with better parts."Give me a break, that is the weakest argument that I have ever heard in my entire life. That train of thought is moronic (Definition 1. a person who is notably stupid or lacking in good judgment.) I asked again if there was any proof to back up the claim that the SCSI drives were less prone to failure and what mechanisms did the SCSI drives employ that made them less prone to failure. I got the same old "if it costs more then they use better parts therefore it is obviously better" and "Why don't you Google it, do your own research". Well that doesn't fly in any arena and it doesn't fly here. I don't care about the difference in throughput I have never claimed that the Raptor was a contender. So spare me your benchmarks. Why don't you instead concentrate on the mechanics of an Enterprise class SCSI drive and how they differ from a Raptor that make it less prone to failure. You Google it,you do the work, because you made the claim. By the way I have a bridge for you, It will cost more but it is made with the best steel.
Quote:
No one, and Ill repeat that again for slow people like yourself, NO ONE in the real "Enterprise" sector uses Raptors for mission critical servers. SATA drives are for the extremely LOW end of the enterprise market. They go into those little rinky dink 1U and 2U webservers that get little to no traffic.


A quote from a post following yours:

Quote:
To quote that article and the paper:
Enterprise Class:In our data sets, the replacement rates of SATA disks are not worse than the replacement rates of SCSI or FC disks. This may indicate that disk-independent factors, such as operating conditions, usage and environmental factors affect replacement rates more than component specific factors.

and again

Quote:
also:It is interesting to note that even Google's own data centers use mainly SATA and PATA drives.

So I guess that Google is not a in the real Enterprise sector. Gee I wonder how they store all of their data? Both claims of yours refuted and backed up with source documents. You may not like the source, but it's better that what you provided.

Quote:
WRONG
First off you arent even in the industry and you have the galls to come on here and tell people with mounds more experience than you that they dont know what they are talking about. What an idiot!


By the way it is either balls or gall not both. You called me an idiot, I guess maybe you are the one that is an idiot.
April 3, 2007 11:02:21 PM

Quote:
No one, and Ill repeat that again for slow people like yourself, NO ONE in the real "Enterprise" sector uses Raptors for mission critical servers. SATA drives are for the extremely LOW end of the enterprise market. They go into those little rinky dink 1U and 2U webservers that get little to no traffic.


A quote from a post following yours:

Quote:
To quote that article and the paper:
Enterprise Class:In our data sets, the replacement rates of SATA disks are not worse than the replacement rates of SCSI or FC disks. This may indicate that disk-independent factors, such as operating conditions, usage and environmental factors affect replacement rates more than component specific factors.

and again

Quote:
also:It is interesting to note that even Google's own data centers use mainly SATA and PATA drives.

So I guess that Google is not a in the real Enterprise sector. Gee I wonder how they store all of their data? Both claims of yours refuted and backed up with source documents. You may not like the source, but it's better that what you provided.

Without touching the rest of the conversation ( :)  ), I believe Sandmanwm was referring to SATA drives being in the lower performing systems of any business, enterprise or small business. At least that's the way I read it. I have many systems with SATA drives, but the performance just doesn't compare to the other systems I have. It could very well be the bus and so forth, but they were cheap servers, so I cannot expect a top notch performer. At least they weren't bought to be that. :) 

Now my other question is this (and we'll probably never know the answer): If Google mainly uses SATA and PATA drives, what are they used in? Are they one of the web servers? I think that is a key point in this discussion, as the "mainly" term means that it is not in use in every system. I recall reading before that some of their offline storage were SATA drives, but for the life of me, I cannot find the article.

I cannot believe that Google would house any database, including warehousing, on SATA or PATA drives. But, like I said, we'll probably never know.

As for reliability, you cannot trust one source. Otherwise, we'd all be driving Fords or Chevys. It all depends on how the systems are cooled and used.
April 3, 2007 11:35:25 PM

I have no problem with any of what you said. I just don't like people making blanket statements and then when asked to support their argument come back with a ridiculous statement about how they cost more so they are better and then say Google it. If they are going to make claims then they are obligated to back them up. Also it is clear that "they" were not responding to my posts. I clearly stated that the Raptor fell into the bottom of the throughput benches. My argument all along is that the Raptor was enterprise class in the category of quality of build and rate of failure. All I got was a bunch of hot air about nuts and bolts etc. Also, that since this was not an IT professional only website but more of an enthusiast website that it would have been nice to see the benchmarks just for fun.
April 4, 2007 12:41:30 AM

Oddly enough most of our EMC's and SHARK's have Seagate drives. They seem to perform just fine on our SAN.
April 4, 2007 1:27:47 AM

Quote:
The bolts, that Pratt & Whitney paid additional money for, have different characteristics. Pratt and Whitney put specs out for the bolts to have a minimum strength, corrosion resistance etc. These things are all quantifiable.


Do not try to BS your way out of this argument with your bogus theories, I was a licensed engineer and have an MS in Operations Research (analysis). I retired from the US Navy after working on aviation and marine propulsion gas turbine maintenance programs. For any given size bolt, made a the same specified alloy and the same thread pitch, that you can buy from a wholesaler like McMaster Carr; Boeing, Pratt & Whitney, Rolls Royce etc. pay at least one to two orders of magnitude more. Why? Each bolt is inspected at each stage of manufacture and tested individually. You can buy the bolt cheaper, but the aviation industry pays a hefty premium to guarantee they get the bolt they paid for. A lot of manufacturers and suppliers cheat and supply counterfit crap. There is a heavy financial incentive to do so. That does not cut it in critical industries (aerospace, defense, medicine, nuclear etc) so they pay to have a QA engineer verify the parts. They are paying top money for a highly specialized machinist to build the part and a much higher paid QA engineer to look over his shoulder. That is where the cost goes – 100% verification.

MILSPEC electronics are a similar, and more relevant example. Two chips do the identical function, and probably come out of the same factory, one is guaranteed to function from –25 to 60 deg C, the MILSPEC chip is guaranteed to work from –75 to 120 deg C. The first chip costs $0.50, the MILSPEC chip costs $60.00. The MILSPEC chip is a bargan if it is part of a fuse assembly that keeps a 2,000 lb warhead in a torpedo from blowing up in the tubes of a $4 billion dollar submarine with a nuclear reactor in the back.

Quote:
Pratt and Whitney didn't just say well if we pay a small fortune then the bolts will be better. The same goes for hard drives. It is possible that the drives cost more due to the SCSI design given that you get additional features. That is fine but the old "it costs more so it is made with better parts" is rubbish.


I did not make the argument that simply paying more gets you a better drive, and as far as I can tell, you are the only one advancing that theory.

Quote:
As you can see from the quote in my post above the Raptor has the same MTBF as the Fujitsu enterprise class SCSI drive. MTBF is the only way, of which I am aware, to rate the reliability of a drive.


As a trained analyst I am laughing at you because you clearly have no idea what MTBF is. :roll: :lol:  :roll: Your average SATA hard drive now has an MBTF of one million hours. Do you think anyone actually ran a batch of them for one million hours to compile relavent statistics on failures? Hint, MBTF is an ESTIMATE based upon failure rate. It is not a measure of reliability. There are a lot of other, more relevant data needed to predict service life. Guess what happens when marketing meets estimates? Manufacturers lie.

And those of you saying that there is no difference in “build quality” are full of crap. As the article pointed out, some of these 2.5” SAS drives have no counterpart in the SATA world. And no corporation is paying $600.00 for a 147GB SAS drive if a $39.00 160GB SATA drive could do the same job. Does everyone need enterprise drives – of course not. But for those who do, there is a difference.
April 4, 2007 2:12:00 AM

Quote:
And it is totally accurate. If you place a Raptor, or Ultra320 SCSI, or even a single SAS drive in a workstation, you will notice very little difference. The major differences come into play when you have them in an array and many, many users are using the system at once. A single game or app is not going to be affected terribly between a Raptor or SAS drive.


I disagree. One key advantage of SCSI technology is that a lot of overhead is off loaded from the CPU and moved onto the host adaptor and drives. This is why a decade ago SCSI was the only way to go if you wanted to guarantee being able to burn CDs, even while doing other tasks. Most IDE HD equiped machines at the time were just not able to maintain an uninterupted data rate to the CD recorders.

I also remember getting a homework assignment in grauate school to solve an integer programing problem by brute force and record the time of completion. My machine (a 266mhz PII) beat everyone in the class with the next closest machine taking over an hour longer to solve the problem, even though several people had faster (300MHz if I recall) processors, but no one else had SCSI. The problem had so many variables, that each matrix manipulation had to be stored to the hard drive, and having SCSI drives was a demonstatable advantage, freeing up the CPU to work the problem, not tend to HD writes and reads.

CPUs and RAM have gotten so powerful compared to HD tech that maybe this is not much of an issue now except for the most demanding applications.
April 4, 2007 4:19:01 AM

Does Pratt & whitney only make milspec turbines? I think not. Nowhere in this thread is milspec mentioned, this is the first. Are the SCSI drives that we are discussing enterprise class and milspec as well?


Quote:
So Pratt and Whitney pays a premium for special materials like inconel or nonel and the special machining to make the bolt, and then they pay a guy with a lot of training and experience to physically verify and inspect every step in the manufacture and]installation of that bolt, as well as test every single bolt.


I seriously doubt that you would be able to get the same bolt for $2.00 with the Iconel and nonel and the special machining but maybe you could. Of course the Pentagon was buying toilet seats for $600 in 1983, so I don't know that I would use the military as my guide as to what things should cost.

Quote:
To you, it is just a bolt, and maybe you got a deal or maybe you did not. You have no idea if somebody sold you counterfeit crap, or if the machining tolerances slipped - weakening the bolt threads by 15% or if any of a hundred other factors made the bolts you bought worthless. To Pratt and Whitney, an $86.00 bolt is cheap and they pay to know that their bolts will perform as specified, because if one fails, it could kill hundreds of people, expose the company to lawsuits, ruin their reputation and drive them out of business. People used to buy the cheap bolts, many still do – then their bridges fall down, buildings collapse and airplanes fall out of the sky.


And exactly what does this have to do with the Raptor or anything for that matter? Is WD counterfeiting their own drives?

Quote:
How does a SCSI enterprise class drive handle vibrations? Do they have special mechanisms in place that the Raptor doesn't?


Better components + better build quality + mo money = SCSI enterprise drive :p 

So, you said better components, what is you source for that assertion or did you get an Epiphany? Same with build quality, do you have any proof that the build quality of the Raptor is inferior? Ah, and yes mo money, that I'm sure you understand, with all your buying bolts for the Navy and such. The Raptor really isn't that much cheaper than an equivalent size SCSI you know. Let's see if we can bring this back to hard drives and you give me proof, any proof, that the WD Raptor is inferior in any way that relates to components used, build quality, or testing to all the enterprise class SCSI drives. I'll say it one more time I don't dispute that SCSI drives are worth their weight in gold. I'm just saying that the Raptor drive is an enterprise class drive with regards to reliability nothing more nothing less. Now this whole bolt and milspec discussion is fun but it has absolutely no bearing on my single assertion. You can bring up your whole life history and you still will not have addressed my assertion. I know what MTBF means and how it is calculated and I am no fan of MTBF. Do you have another metric which might prove your point with regards to reliability? If so, please share it with us. Because in the absence of another metric that one will have to do.
April 4, 2007 9:34:07 AM

Quote:
So there are no special mechanisms in place? I thought not.


google, read, understand. Many people out ther have written articles on the matter, ask any datacenter monkey why he would never mix SAS and SATA in the same enclosure. They use the same interface, they both work on SAS controllers.

Dont expect everyone, or even ANYONE to do your homework for you. Google. Read. Understand.

By the way, to the guy who CLAIMED the Raptors stand up fairly well to a 15k RPM SAS. Send me some of that good stuff you have been smoking . . . I too can use a break from reality....
I like that argument
Quote:
Better components + better build quality + mo money = SCSI enterprise drive :p 
It is not my responsibility to Google anything. It is the responsibility of the individual that made the claim to back it up with some proof. I'm surprised that you were not aware of that.

It sounds like it is your job to be misinformed, and to misinform. I do not claim anything, I mearly state what I've read, and have heard from friends who actually WORK in the industry.

It is funny, you qouted the same guy who already stated you took his comment out of context, yet you did not seem to mention that.

Anyhow, to the Raptor fans, sorry, any 15k SAS, on a desktop, that is doing ANYTHING disk intensive, will stomp the bejesus out of the Raptor.

Also, Fujitsu SAS drives are CRAP. Period. It is possible they are semi fast, to fast, however when you buy 8 new drives, and 2 are bad within a month, something is up . . . especially if it is suppose to be a true enterprise class drive. Maybe it was a bad batch, or some other reason, but all I care about, is getting what I pay for (btw, I did not experience this, a friend did, I buy nothing but Seagate).

There are many reasons why SAS drives outclass SATA drives, but if you really want to spend less, and think you are getting more, sure go right ahead. Just do not say no one ever told you so . . .

Now, having said all that. Desktop systems really are not the place for a SAS drive. They can be, if you have loads of cash, and really want the best period. Most top tier OEMs will serve you just fine for reliability, and speed. Personally, I would rather spend $250 on a 750GB Seagate vs ~$200 for a 150GB Raptor any day of the week (yes, 750's are this low now, if you know where to look). The reasons are simple: MUCH more storage, same warranty, comparrible performance (obviously the Raptor will win, by about 10MB/s, but if you RAID, you get deminishing returns, and if not, then speed probably does not matter so much to you).

We can go round, and round about this for the rest of our lives, if you do not agree with what I have to say, fine, I really do not care. If you do care about being informed, and spreading real/ correct information, I suggest you google/read/understand.

Now, I have said this once already, I'll say it once more, then you guys are on your own concernign the subject:

1) Raptors get ~74MB/s reads, writes, I do not recall.
2) This article says that some of these SAS drives get ~128MB/s reads.

You still think Raptors can hold their own comparred to a 15k RPM SAS, on even a desktop, during drive intesive I/O ? I say you are dreaming .. .

[EDIT]

1) Spindle rotation
2) Vibration dampening
3) Deeper I/O queue depth
4) Controller quality
5) Controller interface type (ie PCI-X vs PCIe)

These are a few reasons why SAS is superior to SATA. #5 can be a little mis leading, since you CAN buy a PCI-X SATA controller, and very good ones, however the question is WHY. SAS controllers can be used to interface SATA drive also, and control a lot more drives, depending on what you want to do, SAS controllers will cost less ( Granted, a cutting edge RAID 5/6 controller, will run you into the $1000's, but these will most likely be able to control up to 255 drives from the same controller + a boat load of expanders).

Will anyone need this much horse power on a desktop ? Not I . . .
April 4, 2007 10:07:35 AM

Quote:
So there are no special mechanisms in place? I thought not.


google, read, understand. Many people out ther have written articles on the matter, ask any datacenter monkey why he would never mix SAS and SATA in the same enclosure. They use the same interface, they both work on SAS controllers.

Dont expect everyone, or even ANYONE to do your homework for you. Google. Read. Understand.

By the way, to the guy who CLAIMED the Raptors stand up fairly well to a 15k RPM SAS. Send me some of that good stuff you have been smoking . . . I too can use a break from reality....
I like that argument
Quote:
Better components + better build quality + mo money = SCSI enterprise drive :p 
It is not my responsibility to Google anything. It is the responsibility of the individual that made the claim to back it up with some proof. I'm surprised that you were not aware of that.

It sounds like it is your job to be misinformed, and to misinform. I do not claim anything, I mearly state what I've read, and have heard from friends who actually WORK in the industry.

It is funny, you qouted the same guy who already stated you took his comment out of context, yet you did not seem to mention that.

Anyhow, to the Raptor fans, sorry, any 15k SAS, on a desktop, that is doing ANYTHING disk intensive, will stomp the bejesus out of the Raptor.

Also, Fujitsu SAS drives are CRAP. Period. It is possible they are semi fast, to fast, however when you buy 8 new drives, and 2 are bad within a month, something is up . . . especially if it is suppose to be a true enterprise class drive. Maybe it was a bad batch, or some other reason, but all I care about, is getting what I pay for (btw, I did not experience this, a friend did, I buy nothing but Seagate).

There are many reasons why SAS drives outclass SATA drives, but if you really want to spend less, and think you are getting more, sure go right ahead. Just do not say no one ever told you so . . .

Now, having said all that. Desktop systems really are not the place for a SAS drive. They can be, if you have loads of cash, and really want the best period. Most top tier OEMs will serve you just fine for reliability, and speed. Personally, I would rather spend $250 on a 750GB Seagate vs ~$200 for a 150GB Raptor any day of the week (yes, 750's are this low now, if you know where to look). The reasons are simple: MUCH more storage, same warranty, comparrible performance (obviously the Raptor will win, by about 10MB/s, but if you RAID, you get deminishing returns, and if not, then speed probably does not matter so much to you).

We can go round, and round about this for the rest of our lives, if you do not agree with what I have to say, fine, I really do not care. If you do care about being informed, and spreading real/ correct information, I suggest you google/read/understand.

Now, I have said this once already, I'll say it once more, then you guys are on your own concernign the subject:

1) Raptors get ~74MB/s reads, writes, I do not recall.
2) This article says that some of these SAS drives get ~128MB/s reads.

You still think Raptors can hold their own comparred to a 15k RPM SAS, on even a desktop, during drive intesive I/O ? I say you are dreaming .. .
I can't believe my eyes. Did you even bother to read any of my posts? Do you have some serious problem with comprehension? I am only concerned with the Raptors reliability, not performance. There is no way that you could have read this whole thread and given a reply that was so off the mark. It's simply impossible. Reliability not performance, reliability not performance, reliability not performance. Can you understand that or would you like to disregard it as you have all previous posts and babble on about how much faster SCSI is, which every poster has already conceded.

Edit: Seriously you should read the rest of the thread. Then post that you did so and understand what my point was. You will look considerably less foolish for you last post.
!