Been using Seagate for years. They've all outlasted the Western Digital drives I've had. Anecdotal of course, but I'll use what continue to works for me in my 24TB NAS box
I had a batch of WD Green 500GB drives, and another batch of WD Green 1.5TB drives (roughly 6 drives per batch). Half of each batch died before warranty, and a few more died shortly after warranty (including some of the warranty replacements). The couple that did survive were discarded due to a loss of confidence. However, I've had 4 WD Red drives that are getting close to warranty expiration, and they haven't skipped a beat. I've also had a bunch of Seagate failures. All manufacturers have their failures, and some real shocker models.
WD green are compromised by default - the head parking issue. If you disable that they last a long time, I had about 10% failure rate of 1 and 2 TB greens two years after their 2 year warranty has expired, they are still running.
My Seagates were great, but I do feel quality has gone down since they bought Maxtor. Now you will randomly have that 1-2 drive models that every couple years that just suck. So I now tend to wait a few months before purchasing any new type of Seagate. Just to see if there's some issue with it.
I'd still rather deal with that than with WD drives (for storage). I find WD drives are better performance, but not as reliable. That's just me though. Others have no issues.
Then failure rates get any big at over 40%...which is nearly 1 in every couple drives sold. Once the drive falls, here comes Shitgate to the rescue... saving your data in theirs labs for... you guessed it... only well over $1200. Sea who?
Corrected: Then failure rates get big at over 40%...which is nearly 1 in every couple drives sold. Once the drive fails, here comes Shitgate to the rescue... saving your data in their labs for... you guessed it... only well over $1200. Sea who?
That temperature gradient spec is interesting. Is it possible to go outside that range if I shutdown a desktop while the HDD has a workload? It looks like these are just suitable for NAS use.
Everyone says these Seagate NAS drives have better reliability than "desktop class" drives, but the warranty is just 3 years. Same goes for WD Red models. Yet my WD Black drives have 5 year warranty. What's up with that?
I think it just has to do with pricing. The WD Red Pro gets 5 yr warranties. I don't think Seagate offers 5 yrs on any of their NAS drives, but they do on their pricier Constellation and Enterprise lines. It's a shame that literally anectodotal crapping on Seagate is more or less all based on their entry level Barracudas. Well, what do you expect for paying minimum dollar? same can probably be said about WD Blues. And the exclusion of Seagate NAS drives from the Backblaze data most certainly does not help the uninformed and fanboyism.
Like a lot of our pipeline content, it's a dissection of a product announcement based on a press release. The aim is to have lots of information and our analysis therein based on what we understand about the market, not just a copy/paste like others. We've been doing it for over a decade, and it's good to stay on top of announcements to let our readers know if what's announced immediately stands out as having potential or a crock.
Great explanation Ian, and these articles are one of the reasons I read AT as my primary Tech News source. Its not just a product announcement with a spec sheet.
Note the URE rate for the 6TB and 8TB drives has gone up to 1 in 10E15. That's a good sign, standard consumer 1 in 10E14 would mean a read failure after 100TB, or 13 drive reads. Or put another way, nearly a 10% chance of a read error per drive when rebuilding an array (for anyone foolish enough to still use RAID5, that would be a 1/3 failure chance for a 5-drive array!).
It's about time *SOMEONE* updated their rated numbers for prosumer HDDs. The non-occurence of the RAID apocalypse over the last few years means that real world number have been much higher than the number on the specsheet; but the clearly fantasy numbers - presumably picked by bean counters to avoid losing money from warranty claims based on higher URE rates in marginal drives - have made realistic risk assessment impossible.
Actually, 10E14 means one error per 12TB I think ( since it is 1 in 1e14 BITs, so need to divide by 8 for BYTEs). And current 10E15 means one error per 125TB, or about 15 drive reads, which still is not exactly great ;p
Ah, and on their 4TB HDD they have "only" 10e14, which means error in about 3 drive reads - definitely not great. Luckily, modern NASes like Synology support btrfs (beta) or QNAP support ZFS, so that could reduce issues with bit-errors ;p
The problem with perception of failure rates on consumer drives is vibration. Most drives have official failure rates and warranties when they're individual drives in a system. The minute you put two near each other, or say four/eight in a NAS, depending on the chassis in most situations the drives will continually vibrate due to spinning. Professional grade drives are built to withstand a multi-HDD environment, whereas if you design a drive to be on its own, it can be made a lot, lot cheaper. So when a major storage firm says '20% failure rate' and you realize they're putting 48 consumer drives in a 4U chassis, I want to say 'no duh'. The fact that other drives can withstand it better is perhaps a nod to base level quality, but ultimately if you want a multi-drive or NAS environment, make sure what you purchased is actually built for it.
Thank you for this post, Ian. Most people are referring to the Backblaze reliability report when they make claims about reliability, but they fail to recognize that Backblaze has not taken the steps necessary to make their metrics scientific or meaningful beyond their own environment. That includes aspects like what you mention, but also no real effort to randomly sample their sources, or make certain that they have statistically significant sample sizes of the drives they test, nor make certain that the environment is actually the same for each drive being included in the results.
At best, their results may help them make decisions for their own storage, but honestly given the lack of control for variables I don't know how useful it is even in that context.
And all of that is a completely moot point, since other drives, like the consumer HGST, WD Reds and even Seagate's own later 4TB drives all have much better (lower) failure rates than a certain ST3000DM001. Sometimes you have a dud design, and from how the other drives behave, it's pretty clear the ST3000DM001 is one such case.
for individual desktop drives, not so much. For those of us building larger arrays using the new high-density chassis options (30+ drives in 4U), the costs of replacing failed lower-end drives is often cheaper than buying higher-end enterprise drives, and for that, backblaze' data is extremely valuable.
Apart from physical construction differences (Enterprise/NAS drives being able to handle more vibration), the other main difference is the behaviour when the drive encounters a read error.
On a desktop drive, when the drive detects a read-error, it'll attempt to read the sector again, and again, and again and...until after some firmware-defined number of retries it gives up and reports a read-error back to the O/S (or succeeds at which point the data is successfully read). This process can take several minutes, possibly 10's of minutes.
In enterprise and NAS HDDs, the assumption is that these drives are most likely going to be put into some sort of redundant array, whether hardware or software RAID or some other software that can handle errors (ZFS etc). In this case, if there is a read-error, report the error immediately to host system (RAID controller, RAID software, etc.) and do not waste time re-reading the sector. The data can be rebuilt from the RAID array, therefore there is no need to waste time trying to re-read the data, doesn't matter if the drive can rebuild the data or not, it can be retrieved from another member of the array. And, since the chances are that if it's being put into a RAID array, it'd PROBABLY a multi-user system, with 10's, 100's, 1000's of disparate simultaneous request for data, such as in, say, a 100TB SAN array. You don't want to hold up those other accesses for minutes trying to re-read data that can be recovered at a tiny overhead from the RAID array.
Therefore in a desktop or other non-RAID-like configuration, the data reliability may in fact be worse, as it won't retry a failed read.
However, one other significant difference with enterprise/NAS drives, is that there are often software tools to access the drives firmware and configure such parameters (read re-tries etc) which in consumer/desktop drives are often locked-out, or the tools just aren't available from the manufacturer.
Everytime a HDD article... "Avoid WD" this and "Seagate, never again" that. Take it from a system builder: just avoid their budget lines. Both WD Greens and the standard Seagate Barracudas suck. Ever since I switched to WD red/se and Seagate es, not a single failure.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
38 Comments
Back to Article
trparky - Tuesday, January 12, 2016 - link
I wouldn't hit a dog in the ass with a Seagate drive.nandnandnand - Tuesday, January 12, 2016 - link
I wouldn't either, too busy using my Seagate drive.ddriver - Tuesday, January 12, 2016 - link
consumer grade product == compromised reliabilityclose - Wednesday, January 13, 2016 - link
== lower price. Not everyone can afford to put in enterprise grade components in their browsing machine.There's a compromise to be made with every choice. Except for the 5.25" drive. That's a total "no compromise" invention.
vnangia - Tuesday, January 12, 2016 - link
No, because that would be animal cruelty.Endda - Tuesday, January 12, 2016 - link
Been using Seagate for years. They've all outlasted the Western Digital drives I've had. Anecdotal of course, but I'll use what continue to works for me in my 24TB NAS boxddriver - Tuesday, January 12, 2016 - link
Reliability stats don't support your unsubstantiated claim.SlyNine - Tuesday, January 12, 2016 - link
That's an appeal to authority fallacy. He would be the authority on how his drives performed. And that's the only claim he made.Gigaplex - Wednesday, January 13, 2016 - link
I had a batch of WD Green 500GB drives, and another batch of WD Green 1.5TB drives (roughly 6 drives per batch). Half of each batch died before warranty, and a few more died shortly after warranty (including some of the warranty replacements). The couple that did survive were discarded due to a loss of confidence. However, I've had 4 WD Red drives that are getting close to warranty expiration, and they haven't skipped a beat. I've also had a bunch of Seagate failures. All manufacturers have their failures, and some real shocker models.ddriver - Wednesday, January 13, 2016 - link
WD green are compromised by default - the head parking issue. If you disable that they last a long time, I had about 10% failure rate of 1 and 2 TB greens two years after their 2 year warranty has expired, they are still running.khanikun - Wednesday, January 13, 2016 - link
My Seagates were great, but I do feel quality has gone down since they bought Maxtor. Now you will randomly have that 1-2 drive models that every couple years that just suck. So I now tend to wait a few months before purchasing any new type of Seagate. Just to see if there's some issue with it.I'd still rather deal with that than with WD drives (for storage). I find WD drives are better performance, but not as reliable. That's just me though. Others have no issues.
3ogdy - Tuesday, January 12, 2016 - link
Then failure rates get any big at over 40%...which is nearly 1 in every couple drives sold. Once the drive falls, here comes Shitgate to the rescue... saving your data in theirs labs for... you guessed it... only well over $1200. Sea who?3ogdy - Tuesday, January 12, 2016 - link
Corrected: Then failure rates get big at over 40%...which is nearly 1 in every couple drives sold. Once the drive fails, here comes Shitgate to the rescue... saving your data in their labs for... you guessed it... only well over $1200. Sea who?zodiacfml - Tuesday, January 12, 2016 - link
That temperature gradient spec is interesting. Is it possible to go outside that range if I shutdown a desktop while the HDD has a workload? It looks like these are just suitable for NAS use.p1esk - Tuesday, January 12, 2016 - link
Everyone says these Seagate NAS drives have better reliability than "desktop class" drives, but the warranty is just 3 years. Same goes for WD Red models. Yet my WD Black drives have 5 year warranty. What's up with that?etamin - Tuesday, January 12, 2016 - link
I think it just has to do with pricing. The WD Red Pro gets 5 yr warranties. I don't think Seagate offers 5 yrs on any of their NAS drives, but they do on their pricier Constellation and Enterprise lines.It's a shame that literally anectodotal crapping on Seagate is more or less all based on their entry level Barracudas. Well, what do you expect for paying minimum dollar? same can probably be said about WD Blues. And the exclusion of Seagate NAS drives from the Backblaze data most certainly does not help the uninformed and fanboyism.
ABR - Tuesday, January 12, 2016 - link
Is this an article or a press release? And what's "network area storage"?Ian Cutress - Wednesday, January 13, 2016 - link
Like a lot of our pipeline content, it's a dissection of a product announcement based on a press release. The aim is to have lots of information and our analysis therein based on what we understand about the market, not just a copy/paste like others. We've been doing it for over a decade, and it's good to stay on top of announcements to let our readers know if what's announced immediately stands out as having potential or a crock.creed3020 - Wednesday, January 13, 2016 - link
Great explanation Ian, and these articles are one of the reasons I read AT as my primary Tech News source. Its not just a product announcement with a spec sheet.edzieba - Wednesday, January 13, 2016 - link
Note the URE rate for the 6TB and 8TB drives has gone up to 1 in 10E15. That's a good sign, standard consumer 1 in 10E14 would mean a read failure after 100TB, or 13 drive reads. Or put another way, nearly a 10% chance of a read error per drive when rebuilding an array (for anyone foolish enough to still use RAID5, that would be a 1/3 failure chance for a 5-drive array!).DanNeely - Wednesday, January 13, 2016 - link
It's about time *SOMEONE* updated their rated numbers for prosumer HDDs. The non-occurence of the RAID apocalypse over the last few years means that real world number have been much higher than the number on the specsheet; but the clearly fantasy numbers - presumably picked by bean counters to avoid losing money from warranty claims based on higher URE rates in marginal drives - have made realistic risk assessment impossible.Nenad - Wednesday, January 13, 2016 - link
Actually, 10E14 means one error per 12TB I think ( since it is 1 in 1e14 BITs, so need to divide by 8 for BYTEs).And current 10E15 means one error per 125TB, or about 15 drive reads, which still is not exactly great ;p
Nenad - Wednesday, January 13, 2016 - link
Ah, and on their 4TB HDD they have "only" 10e14, which means error in about 3 drive reads - definitely not great.Luckily, modern NASes like Synology support btrfs (beta) or QNAP support ZFS, so that could reduce issues with bit-errors ;p
edzieba - Wednesday, January 13, 2016 - link
You're right, I derped and used the bit error rat as a byte error rate. So everything is 8x worse. Stop using RAID5 already!extide - Wednesday, January 13, 2016 - link
I noticed that as well. Good to see. I'll still stick with ZFS and 2 or 3 drive parity, though.Ian Cutress - Wednesday, January 13, 2016 - link
The problem with perception of failure rates on consumer drives is vibration. Most drives have official failure rates and warranties when they're individual drives in a system. The minute you put two near each other, or say four/eight in a NAS, depending on the chassis in most situations the drives will continually vibrate due to spinning. Professional grade drives are built to withstand a multi-HDD environment, whereas if you design a drive to be on its own, it can be made a lot, lot cheaper. So when a major storage firm says '20% failure rate' and you realize they're putting 48 consumer drives in a 4U chassis, I want to say 'no duh'. The fact that other drives can withstand it better is perhaps a nod to base level quality, but ultimately if you want a multi-drive or NAS environment, make sure what you purchased is actually built for it.LordanSS - Wednesday, January 13, 2016 - link
Many consumer cases come with anti-vibration HDD caddies, by either using rubber dampers or similar.I know my old (and now retired) CM690-II had their removable drive caddies set up in that way.
Wouldn't something of the sort lessen vibration issues on a multi-HDD setup?
Reflex - Wednesday, January 13, 2016 - link
Thank you for this post, Ian. Most people are referring to the Backblaze reliability report when they make claims about reliability, but they fail to recognize that Backblaze has not taken the steps necessary to make their metrics scientific or meaningful beyond their own environment. That includes aspects like what you mention, but also no real effort to randomly sample their sources, or make certain that they have statistically significant sample sizes of the drives they test, nor make certain that the environment is actually the same for each drive being included in the results.At best, their results may help them make decisions for their own storage, but honestly given the lack of control for variables I don't know how useful it is even in that context.
ZeDestructor - Wednesday, January 13, 2016 - link
And all of that is a completely moot point, since other drives, like the consumer HGST, WD Reds and even Seagate's own later 4TB drives all have much better (lower) failure rates than a certain ST3000DM001. Sometimes you have a dud design, and from how the other drives behave, it's pretty clear the ST3000DM001 is one such case.Reflex - Thursday, January 14, 2016 - link
True or not, citing Backblaze is not making a case.ZeDestructor - Tuesday, January 19, 2016 - link
for individual desktop drives, not so much. For those of us building larger arrays using the new high-density chassis options (30+ drives in 4U), the costs of replacing failed lower-end drives is often cheaper than buying higher-end enterprise drives, and for that, backblaze' data is extremely valuable.eddman - Wednesday, January 13, 2016 - link
Is it ok to use an NAS drive as a main drive on a PC, or should I use a desktop specific drive? Are there any special firmware differences?extide - Wednesday, January 13, 2016 - link
You would be fine, I'm sure.eldakka - Wednesday, January 13, 2016 - link
Apart from physical construction differences (Enterprise/NAS drives being able to handle more vibration), the other main difference is the behaviour when the drive encounters a read error.On a desktop drive, when the drive detects a read-error, it'll attempt to read the sector again, and again, and again and...until after some firmware-defined number of retries it gives up and reports a read-error back to the O/S (or succeeds at which point the data is successfully read). This process can take several minutes, possibly 10's of minutes.
In enterprise and NAS HDDs, the assumption is that these drives are most likely going to be put into some sort of redundant array, whether hardware or software RAID or some other software that can handle errors (ZFS etc). In this case, if there is a read-error, report the error immediately to host system (RAID controller, RAID software, etc.) and do not waste time re-reading the sector. The data can be rebuilt from the RAID array, therefore there is no need to waste time trying to re-read the data, doesn't matter if the drive can rebuild the data or not, it can be retrieved from another member of the array. And, since the chances are that if it's being put into a RAID array, it'd PROBABLY a multi-user system, with 10's, 100's, 1000's of disparate simultaneous request for data, such as in, say, a 100TB SAN array. You don't want to hold up those other accesses for minutes trying to re-read data that can be recovered at a tiny overhead from the RAID array.
Therefore in a desktop or other non-RAID-like configuration, the data reliability may in fact be worse, as it won't retry a failed read.
However, one other significant difference with enterprise/NAS drives, is that there are often software tools to access the drives firmware and configure such parameters (read re-tries etc) which in consumer/desktop drives are often locked-out, or the tools just aren't available from the manufacturer.
willis936 - Wednesday, January 13, 2016 - link
The world's most timevefficient way to destroy data.nwarawa - Thursday, January 14, 2016 - link
Everytime a HDD article... "Avoid WD" this and "Seagate, never again" that. Take it from a system builder: just avoid their budget lines. Both WD Greens and the standard Seagate Barracudas suck. Ever since I switched to WD red/se and Seagate es, not a single failure.I hate fanboys - Thursday, February 18, 2016 - link
Hey all fanboys and other ignorant corn holes (if you can read):http://arstechnica.com/information-technology/2016...
xicaque - Sunday, October 23, 2016 - link
I've had mixed feelings about Seagates.