Are they just putting an 'Enterprise Grade' sticker on it and going for a different market segment?
March 19, 2010 1:23 AM Subscribe
Why does the Seagate Barracuda ES.2 cost almost twice as much as the 7200.12 for the same capacity?
Am I just paying for more reliability and quality control or is it the extra 16MB of cache?
I'm setting up a small test and development server for a team of 3-5 people working on a web-based application, I'll probably buy two disks for a RAID-1 setup. Do you think I would see any tangible benefit from the more expensive model?
Am I just paying for more reliability and quality control or is it the extra 16MB of cache?
I'm setting up a small test and development server for a team of 3-5 people working on a web-based application, I'll probably buy two disks for a RAID-1 setup. Do you think I would see any tangible benefit from the more expensive model?
Best answer: What he said above about the MTBF and 5 year warranty, also:
The firmware on the "ES" drives is also optimized to handle use in a RAID5 or RAID6 array, particularly as it relates to the NCQ ability of the drives. If you build a large RAID6 array with regular consumer SATA drives, when one of the drives decides to perform thermal recalibration or "pause" on a write, the RAID controller may see the delay and decide the drive's lag means it has failed and will mark the whole array as requiring a rebuild.
posted by thewalrus at 1:58 AM on March 19, 2010 [1 favorite]
The firmware on the "ES" drives is also optimized to handle use in a RAID5 or RAID6 array, particularly as it relates to the NCQ ability of the drives. If you build a large RAID6 array with regular consumer SATA drives, when one of the drives decides to perform thermal recalibration or "pause" on a write, the RAID controller may see the delay and decide the drive's lag means it has failed and will mark the whole array as requiring a rebuild.
posted by thewalrus at 1:58 AM on March 19, 2010 [1 favorite]
Ditto.
ES.2 is designed to operate 24/7, whereas 7200.12 is designed to be used in a desktop-like environment (8h per day, 5 days per week). link So, if you wanted to use 7200.12 in 24/7-environment you'd be running risk of its MTBF being actually lower.
Also, nonrecoverable read errors per bits read is an order of magnitude lower for ES.2 (1 per 10^15) than for 7200.12 (1 per 10^14).
posted by noztran at 3:09 AM on March 19, 2010
ES.2 is designed to operate 24/7, whereas 7200.12 is designed to be used in a desktop-like environment (8h per day, 5 days per week). link So, if you wanted to use 7200.12 in 24/7-environment you'd be running risk of its MTBF being actually lower.
Also, nonrecoverable read errors per bits read is an order of magnitude lower for ES.2 (1 per 10^15) than for 7200.12 (1 per 10^14).
posted by noztran at 3:09 AM on March 19, 2010
thewalrus, I think you've just cleared up some hitherto totally baffling behavior in our (now superseded) curriculum network server - thank you! It was RAID 1, not 5 or 6, but even so it would sporadically declare that one of the drives was bad and drop it from the array even though neither drive's SMART log ever showed any error events.
posted by flabdablet at 3:10 AM on March 19, 2010
posted by flabdablet at 3:10 AM on March 19, 2010
Keep in mind, ES drives are a generation behind consumer drives in terms of preformance. The ES.2's are based on the 7200.10 I believe, so they will be a little bit slower compared to a 7200.12 if you just compare the two side by side. That being said, I only deploy enterprise drives for servers, since it's not worth your headache for a drive to die sooner than it has to, and that little bit of extra preformance isn't very important anyway.
posted by tracert at 5:05 AM on March 19, 2010
posted by tracert at 5:05 AM on March 19, 2010
So how does one identify these "enterprise" vs "desktop" drives when browsing through the selection?
I've bought quite a few HDDs over the years, and I don't recall any dead giveaway or standardized naming convention across manufacturers.
Do I just look at the MTBF and read errors per bits read as mentioned or am I missing something else?
TIA
posted by InsertNiftyNameHere at 6:39 AM on March 19, 2010
I've bought quite a few HDDs over the years, and I don't recall any dead giveaway or standardized naming convention across manufacturers.
Do I just look at the MTBF and read errors per bits read as mentioned or am I missing something else?
TIA
posted by InsertNiftyNameHere at 6:39 AM on March 19, 2010
So how does one identify these "enterprise" vs "desktop" drives when browsing through the selection?
One of the easier ways is to only look at drives that carry a 5-year warranty. Each manufacturer has their own way of designating enterprise-grade models though. Seagate prefixes them with "ES", usually. But you really have to look at the specs to get the MTBF and bit error rates.
I don't think there's an easy way to filter exclusively for them on Newegg, if that's what you're asking.
(Also, if you're building an array one thing I wouldn't do is stock it with identical drives all purchased at the same time. I've always preferred to diversify a bit, just to avoid a potential Deathstar situation.)
posted by Kadin2048 at 6:57 AM on March 19, 2010 [1 favorite]
One of the easier ways is to only look at drives that carry a 5-year warranty. Each manufacturer has their own way of designating enterprise-grade models though. Seagate prefixes them with "ES", usually. But you really have to look at the specs to get the MTBF and bit error rates.
I don't think there's an easy way to filter exclusively for them on Newegg, if that's what you're asking.
(Also, if you're building an array one thing I wouldn't do is stock it with identical drives all purchased at the same time. I've always preferred to diversify a bit, just to avoid a potential Deathstar situation.)
posted by Kadin2048 at 6:57 AM on March 19, 2010 [1 favorite]
Best answer: How to identify? Enterprise drives are more expensive, generally. It used to be that enterprise drives were SCSI and always in specific capacities like 4.3, 9.1, 18.2, 36.4, 72.8, etc., where desktop drives were everything else.
These days, as other posters mentioned, a lot of the difference is in the firmware and what sort of data tasks the drive is "good at". Since enterprise drives are almost always in a RAID configuration and serving multiple users and processes, performance is improved by improving the random read/write ability. Different cacheing algorithms. Faster read/write head, more robust bearings, etc., since its entire life is going to be spent doing random seeks. You'd also see spindle motors designed for reliable sustained operation, versus the opposite end of the spectrum of the laptop drive, where it would be designed for tons of spin up and spin down cycles.
One thing enterprise drives were doing to improve that type of performance was making drives with 2.5" platters in the standard 3.5" form factor. Performance is improved because the RW head just physically has less far to go. And performance is more predictable because the relative speeds of the inner and outer edges are closer. The downside is that doing this costs more- more platters per drive.
Enterprise drives also can be designed with fewer compromises. Perfect example is the SATA drive in the 2.5" form factor. Enterprise drives require 12v power, because it is easier to make a 12v motor than a 5v one. Laptop drives have to get along using 5v, because laptops don't really have a 12v rail like a traditional machine does.
So, like anything, analyze what you are going to be doing with the machine and choose the hardware that will do that most cost-effectively. Also, there is a big difference in performance requirements for raid1 versus raid5. Raid 1 is just a mirror, so its performance requirements are more desktop like.
Don't discount the controller itself as being part of the enterprise versus desktop problem- a cheap controller is more likely to not handle momentary glitches as well as a better one. You'll either get premature failures or data corruption.
posted by gjc at 7:19 AM on March 19, 2010 [2 favorites]
These days, as other posters mentioned, a lot of the difference is in the firmware and what sort of data tasks the drive is "good at". Since enterprise drives are almost always in a RAID configuration and serving multiple users and processes, performance is improved by improving the random read/write ability. Different cacheing algorithms. Faster read/write head, more robust bearings, etc., since its entire life is going to be spent doing random seeks. You'd also see spindle motors designed for reliable sustained operation, versus the opposite end of the spectrum of the laptop drive, where it would be designed for tons of spin up and spin down cycles.
One thing enterprise drives were doing to improve that type of performance was making drives with 2.5" platters in the standard 3.5" form factor. Performance is improved because the RW head just physically has less far to go. And performance is more predictable because the relative speeds of the inner and outer edges are closer. The downside is that doing this costs more- more platters per drive.
Enterprise drives also can be designed with fewer compromises. Perfect example is the SATA drive in the 2.5" form factor. Enterprise drives require 12v power, because it is easier to make a 12v motor than a 5v one. Laptop drives have to get along using 5v, because laptops don't really have a 12v rail like a traditional machine does.
So, like anything, analyze what you are going to be doing with the machine and choose the hardware that will do that most cost-effectively. Also, there is a big difference in performance requirements for raid1 versus raid5. Raid 1 is just a mirror, so its performance requirements are more desktop like.
Don't discount the controller itself as being part of the enterprise versus desktop problem- a cheap controller is more likely to not handle momentary glitches as well as a better one. You'll either get premature failures or data corruption.
posted by gjc at 7:19 AM on March 19, 2010 [2 favorites]
If you're doing any sort of RAID then go with the ES drives. Or an RE drive if you're thinking of buying Western Digital. Mechanical drives are unreliable to begin with. Purposely buying the home/non-RAID version to save a few dollars isn't a wise move.
posted by damn dirty ape at 8:27 AM on March 19, 2010 [1 favorite]
posted by damn dirty ape at 8:27 AM on March 19, 2010 [1 favorite]
Response by poster: Yeah, I'm going to go with a pair of ES drives, even though it's more like 100€ vs 50€ for each around here, so it's not a trivial amount. This is supposed to be Serious Business, hopefully they will pay for themselves over time.
posted by Dr Dracator at 9:16 AM on March 19, 2010
posted by Dr Dracator at 9:16 AM on March 19, 2010
Kadin2048: "
(Also, if you're building an array one thing I wouldn't do is stock it with identical drives all purchased at the same time. I've always preferred to diversify a bit, just to avoid a potential Deathstar situation.)"
Great point. I bought, I think, four of those original IBM Deathstars (20GB, IIRC). Even under very light use in glorified terminals, three of the four failed. The fourth was never used much, and I eventually updated its firmware. The thing is actually still working last I checked. I may have the last working IBM Deathstar on the planet.
Thanks!
posted by InsertNiftyNameHere at 7:45 PM on March 19, 2010
(Also, if you're building an array one thing I wouldn't do is stock it with identical drives all purchased at the same time. I've always preferred to diversify a bit, just to avoid a potential Deathstar situation.)"
Great point. I bought, I think, four of those original IBM Deathstars (20GB, IIRC). Even under very light use in glorified terminals, three of the four failed. The fourth was never used much, and I eventually updated its firmware. The thing is actually still working last I checked. I may have the last working IBM Deathstar on the planet.
Thanks!
posted by InsertNiftyNameHere at 7:45 PM on March 19, 2010
If you're going to stock a RAID 1 array with diverse drives, some controllers will require you to fiddle with the logical sizes of all but the smallest one to make them all the same. I used MHDD to do this for my old array.
posted by flabdablet at 10:10 PM on March 19, 2010
posted by flabdablet at 10:10 PM on March 19, 2010
This thread is closed to new comments.
posted by wongcorgi at 1:50 AM on March 19, 2010