PERC H200 with SATA drives (Poweredge T310)

Thank you very much for your help.

The H200 looks the better card so I will buy that one.

SAS also supports SATA - it is part of the spec (SATA connectors to NOT support SAS drives), so yes, the H200 will support SATA drives.

The H200 controller is a hardware RAID card (H = Hardware), while the S300 is a software RAID card (S = Software - it is little more than a connector card for all the drives).  While neither is an exceptional card, the H200 will be hands-down the best one.  Besides being a software RAID solution, the S-series cards only work with Windows, if that is a consideration for you.

You can order it with an H200 with a single SATA drive, but you will need to select a non-RAID configuration, then configure RAID on the card when you put your drives in.

Just a note about drives ... don't use Desktop-class drives.  It is tempting because of the price, but they are not suitable for use in servers.  If you value your data or the services your server will host, you need to invest in Enterprise-class drives.  Certified drives are even better, as they are specifically tuned to work with the controller.  You don't have to pay the premium of buying them from Dell ... get the part numbers and shop elsewhere.  You can buy Dell/Certified drives from dozens of suppliers for cheaper than directly from Dell.

"Just a note about drives ... don't use Desktop-class drives.  It is tempting because of the price, but they are not suitable for use in servers. "

I'm sorry but that's simply not true. We use standard desktop drives in servers for many years, and they barely have a higher failure rate than Enterprise-class drives (which at the price difference isn't really a problem). Due to their lower performance compared to 15krpm SAS drives they are of course not suitable for high I/O workloads (i.e. busy database servers), but for other tasks like file serving they are perfectly fine.

Also, I'm sure that Google also would beg to differ, as they operate most of their server blades with bog standard desktop SATA hard drives.

At the end of the day, any drive (even the $500+ Enterprise server drive) can fail tomorrow, so no matter what, appropriate provisions for redundancy (i.e. RAID) and a proper working(!) backup strategy are always a must-have.

The real problem with 3rd party drives is compatibility. For example, we had problems with WD Caviar Green drives in some HP servers because of the drive firmware. Some few desktop drives (most of them seem to be WDC drives) tend to drop out on hardware RAID controllers, so it's very important to do some thorough tests before deploying them into production.  But then, we also had drives sporadically dropping out with some original HP Enterprise-class drives which had a buggy frmware, so even server manufacturer certified drives are in fact no guarantee for troublefree operation.

Maybe it is not apparent in your experience, but I'm not just talking about my experience (which does show it to be true):

download.intel.com/.../enterprise_class_versus_desktop_class_hard_drives_.pdf

Any drive can die at anytime - that is the nature of the beast - but not all drives are created equal.

You mention "compatability" between certified and non-certified drives ... the firmware is tweaked to match the controller's requirements.  OEM's release firmware updates to address bugs ... just because they have a tailor-made firmware doesn't mean they are less prone to buggy firmware than generic drives.  The exact same reason can be applied to desktop and enterprise drives.  TLER and many other drive settings are just as different between c/nc and d/e drives.  I also am not saying that going enterprise/certified would equate to trouble-free operation ... simply that it is much more likely that they will work without issue ... not to mention that you are on your own to figure out what problems exist when you have an unstable array, where using certified drives, you have the OEM's guarantee that it will work - they will (and can) work with you until it works as advertised.

I have gleaned what little knowledge I have from years and hundreds of drive's worth of experience, but also from several other drive and controller experts.  If you want the REAL deal ... head over to Experts-Exchange and post a question there.  There are at least two independent (not associated with any OEM, controller, or drive company) drive and controller programmers that can talk more in-depth about the hows and whys.

All I can say is that we operate over 50,000 drives in various scenarios, and every drive's lifecycle is electronically captured, and the picture we get is very clear.

The intel document doesn't really prove anything, it just states generalizations that may or may not be valid for specific drives and in were not valid for almost a decade any more. For example, it says that Enterprise disks have a higher media density which is nonsense: current 3.5" Enterprise SAS disks have less than 350GB per platter (the 600GB Cheetah NS.2 for example has 150GB per platter), consumer disks are nearing 1TB per platter). In terms of media density, Enterprise-class drives have long been left behind by desktop drives. The pdf also mentions that Enterprise disks have a more sturdy and precise actuators (another statement that was valid say 5 years ago but not any more, simply because the extreme data density of desktop drives requires very high precision and highly advanced reading techniques) or that only Enterprise drives have two processors (which is wrong, see for example WDC Caviar Black). This document hasn't kept up with the developments in hard drive technology and needs a lot of updating and refreshing (as does the outdated knowledge of many experts).

I already mentioned Google, which operates all its services with custom-designed blade servers running consumer-grade SATA drives. You would think that at the number of global services they operate they have a pretty good idea of what they do. They also did a study on hard drive reliability (the one with the by far biggest data basis btw) which you might find interesting:

labs.google.com/.../disk_failures.pdf

Don't get me wrong, Enterprise drives do certainly have their use. They still offer much higher IOPS and shorter access times than any desktop SATA drive, and for high workload they are the way to go. Desktop SATA drives certainly have their limitations (lower IOPS,  high average access times) which makes them a bad choice for certain scenarios of high workload. But saying that desktop SATA drives are generally unfit for server use is (sorry) just complete nonsense.

As to compatibility with Dell PERC controllers: if you guranteed compatibility and support through Dell service then yes, you need to go for Dell certified drives (no matter if SATA or SAS). However, if your specific scenario and requirements can be served with SATA drives, and you don't really need hard drive supportability through Dell and are fully aware that if something goes wrong it#s your problem, then 3rd party drives are perfectly fine.

While I would be interested to see what kind of data you have captured on drives and their lifecycle, what you have said cannot overturn my opinion based on my experiences and information received from others.  I don't discount what you say - I just don't know your credentials to trust that what you say comes from any more than electronically capturing drive info at work.  

People I have worked with in the past have written firmware for controllers and drives, created diagnostic utilities, and written storage-related applications for most of the major OEM's and other manufacturers ... I know their creds and what they have shared with me closely matches my experiences on real, live systems.  Your information does not.  I have had disks not initialize, continuously fail, fall offline, not rebuild, etc. in RAID arrays when the drives tested good and worked otherwise.  Granted, my experience is almost exclusively with Dell servers and controllers, but I have had very little problems with any certified products, and have seen reliability problems with desktop drives (in another small data center) on Dell servers.  

Google might be using desktop drives, but they are probably all running on a linux platform with linux-implemented RAID, on custom-built servers ... tweakable to THEIR specifications.  That, to me, minimizes the study's relevance here.  Even if they didn't fail as often, the fact that you are using them on non-user-customizable PERC controllers means we are then not talking about reliability from drive longevity, but about reliability from their compatibility with the controller.  In the end, the result is problems.  Unless you are a highly-trained storage expert (like maybe you are), why invite problems that you won't know how to solve, other than to pay an expert to sort it out for you or go back to a "good" configuration like you will wish you had done in the first place?

Just my two cents.

you did see WHO wrote the paper you linked?

Enterprise Platforms and Services Division - Marketing

Marketing.....you really believe it?  I wouldnt trust that data any further than i can throw a server class hard drive ;) (which i guess technically is fairly far, but still...you get my point)

I see everyones point of view here, including Dell's

I not greatly into Dell certified drives either, they are over priced to the the point of gouging. I could see a small premium, say 20%, maybe even 40%  over a normal retail price.

I just purchased a basic server with 2 drives, costing $249/each, which can be obtained via the web for $113/each , granted you get the drive chassis with the drives from Dell. Figuring the chassis is really worth $5.00/per , that is $131 over retail, that is over 200% even with shipping. Dell obviously is gettting the drives cheaper than retail......Greed

On Dell's side, I would not want to support raids with drives of different/old/buggy firmwares, but greed is the primary force behind the drive pricing. I do understand the Dell firmware's are slightly tweaked for stability and standardization of firmwares are important, and the drives Dell uses or replaces are tested, that is a factor which is important... so would a released list of compatible  drives and firmwares.

I just priced up a SATA based server for a client, clients unwilling to go SAS. For the client to go for it, I need to buy the minimum number of drives, then purchased the majority retail. The difference is $600 less on a $3500 server between Dell certified and retail drives.

I'm not into spending more than is necessary either:

$317 - 1TB Enterprise SATA drive from Dell.com, part number F811C

$195 - 1TB Enterprise SATA drive from Neworldit.com, part number F811C
http://www.neworldit.com/ProductDetails.asp?ProductCode=F881C-NB

$159 - 1TB Enterprise SATA drive from XByte.com - Dell Certified
http://www.xbyte.com/ProductDesc.aspx?code=CAT_113_50

$140 - 1TB Enterprise SATA drive from Newegg.com - not Certified (retail/generic)
http://www.newegg.com/Product/Product.aspx?Item=N82E16822136798

I never buy directly from Dell - I think it is ridiculous for them to charge the prices they do - I buy the same drive from resellers/suppliers, and the price difference between a supported/qualified/certified and generic/retail is small enough that it is well worth the small cost of certified drives to avoid the questions and the problems altogether.

Thanks Flash for the links. Yersterday I spend the morning sifting through web retails trying to locate an elusive  "new" retail  with 5 year warranty"  drive for a raid which is 7 years old...fun   . The number of recerts and refurbs advertised make it real hard to find a decent drive.

Agreed!

Simon, as to the H200 make sure you have the drive cache enabled (read caution as to loss of power), otherwise performance is remarkably dismal.

Flash,  Dell should send you a deluxe wks for all the support calls you saved them.... or at least give you a considerable discount on certified drives

lol :)

Thank you for this document.  In reading the .pdf, it is clear that just because you can, doesn't mean you should.