People are perennially interested in the reliability of hard drives. Here is useful hard data. Summary:
At Backblaze, as of December 31, 2014, we had 41,213 disk drives spinning in our data center, storing all of the data for our unlimited backup service. That is up from 27,134 at the end of 2013. … The table below shows the annual failure rate through the year 2014.
I spend time in hardware enthusiast communities and not so impressed with Backblaze. Even here, the Seagate failure rates seem suspiciously anomalous.
Also, SSDs, which are probably a better match for most people here (my rig has run a 256 GB SSD for the past 2.5 years and I’m yet to want for more storage). Especially for laptops; they use less power (= your battery lasts longer) and can stand up to shock (so your laptop doesn’t break if you drop it).
I did not mean to endorse any particular service or give recommendations as to which storage devices should people buy. I found hard data which is rare to come by, I shared it. If you think the data is wrong or misleading, do tell.
Consensus is that modern HDD’s from reputable manufacturers have approximately equal low failure rates, especially after the first year. You should still back up important data (low != 0), but the differences failure rates in consumer space is small enough to not really sway purchasing decisions.
Their methodology probably doesn’t extrapolate well because they’re testing the drives in what amounts to a NAS and the WD reds (which did well) are NAS drives, and therefore designed to operate 24⁄7 with vibration and nongreat cooling, whereas the Seagate Barracudas are just absolutely not NAS drives (unlike, say, the Seagate NAS drives). So, it’s not really surprising they had a much higher failure rate, but it’d also be incorrect to conclude that you should avoid them. If I’m building a rig for work, internet use, or gaming {1}, then my HDD’s going to be in a well-cooled, non-vibrating environment, and not used in use 24⁄7, so I’m essentially throwing away 15% price premium for the WD Red’s (or 60% for the HGST Deskstar’s). OTOH, if you’re backing up your data locally on a NAS, pay the gorram premium.
{1} Again, though, SSD’s are increasingly likely the way to go. You can get a sufficiently good 256 GB SSD for about the price of a 3 GB HDD and if you’re never going to use more than 250 GB (which, I’m guessing is at least 80% of people reading this who don’t already know whether an SSD or HDD better meets their needs), you’re essentially getting substantially better performance (up to an order of magnitude), more reliability, and less noise for free. I harp on this because SSD’s come in a 2.5-inch form factor and the more the standard storage option is SSD, the more cases won’t have a whole bunch of room taken up with 3.5-inch bays I don’t use. More importantly, there’ll finally be budget laptops that I don’t have to immediately take apart, clone the OS onto an SSD, reassemble, and figure out what to do with the HDD it came with just to get a decent experience. Gah! SSD’s are the right choice for most people and there’s externalities when they get HDD’s instead because “more gigabytes”.
Consensus is that modern HDD’s from reputable manufacturers have approximately equal low failure rates, especially after the first year.
I am sorry, the link shows hard data which disproves that statement and not in a gentle way, either.
So, it’s not really surprising they had a much higher failure rate
Didn’t your first sentence state that all failure rates are “approximately equal”? Make up your mind.
my HDD’s going to be in a well-cooled, non-vibrating environment
Assumption not in evidence. I’ve seen a LOT of computers totally taken over by dust bunnies :-) The reason you go look at that grey disk where the fan vent used to be is that your bios starts screaming at you that the machine is overheating :-D
SSD’s are the right choice for most people
Yes, but that’s irrelevant to the original post which looks at reliability of rotating-platter hard drives. If you think you don’t care about the issue, well, what are you doing in this subthread?
Consumer-grade HDD’s, used properly, all have about same, low failure rate. If you treat your desktop like a NAS or server, they will drop like flies (as evidenced). If you treat your desktop like a desktop, then a lot of the price-raising enterprise-grade features (vibration resistance, 24⁄7 operation) count for zilch. They’re still higher-end drives, and will last longer, but assuming you give your desktop a fraction of the maintenance you give your car (like, take 5 minutes to blow it out every other year), not a lot.
tl;dr Looking at this data and concluding “avoid Seagate Barracuda drives” is a bit like noticing that bikers survive accidents more often when they’re wearing a helmet and then issuing a blanket recommendation to a population primarily of car-drivers to wear bike helmets. Sure, it’ll reduce your expecting mortality when you go out for a drive, but not nearly as much as you’d expect from the biking numbers.
Consumer-grade HDD’s, used properly, all have about same, low failure rate. If you treat your desktop like a NAS or server, they will drop like flies (as evidenced).
Sigh. No. Really, go look at the data. I am not going to take the “consensus” of the anand crowd over it.
Hitachi Deskstar 7K2000 is a consumer-grade non-enterprise hard drive. In the sample of ~4,600 drives it has 1.1% annual failure rate in the NAS environment.
Seagate Barracuda 7200.14 is a consumer-grade non-enterprise hard drive. In the sample of ~1,200 drives it has 43.1% annual failure rate in the NAS environment.
Those are VERY VERY DIFFERENT failure rates.
I, for example, have five-drive zfs array at home which is on 24⁄7. I am very much interested in what kind of drives will give me a 1% failure rates and which kind of drives will give me 43% failure rates. I am not average, but I hardly think I’m unique in that respect in the LW crowd.
We certainly agree that the Barracuda’s are crap in NAS’s. I believe that WD Red’s are a major improvement and Hitachi Deskstars a further improvement, which is just reading the Backblaze data (which is eminently applicable to NAS environments), so I’m we’re in complete agreement that, for NAS’s, Barracuda << Red < 7K2000.
However, I also contend that, in a desktop PC, a lot of what makes the Reds and 7K2000 more reliable (e.g. superior vibration resistance) will count for very little, so they’ll still fail less often, just not 1/40th as much. Even if they’re four times as reliable, moving from, say, a 4% annual failure rate vs a 1% annual failure rate may not be worth the price premium (using Newegg pricing, the Hitachi drive costs 72.5% more, but on Amazon, the Hitachi drive is cheaper. Yay Hitachi?), especially since RAID 1 is a thing (which would give us a 0.16% annual failure rate at a 100% price premium). Obviously, if you can find higher-quality drives for less than lower-quality drives, use those. But, in what we’d naively expect to be the normal case, if you’re paying for features that drastically reduce failure rates in NAS environments, but using your drives in a desktop environment where these features are doing little to extend your drive life, then you’re probably better off using RAID 1.
(Why do I use low single-digit annual failure rates? Because I remember Linus of Linus Tech Tips, who worked as a product manager at NCIX and therefore is privy to RMA and warranty rates, implied that’s about right. He produces a metric shit-ton of content, though, so there’s no way I’m going to dig it up.)
I’m also interested why you’re dismissive of AnandTech. I currently believe they’re gold standard of tech reviews, but if they’re not as reputable as I believe they are, I would very much like to stop believing they are.
Yes. You keep saying that there are no significant differences in reliability between hard drives of similar class (consumer or enterprise, basically) in similar conditions. I keep saying there are.
I’m also interested why you’re dismissive of AnandTech. I currently believe they’re gold standard of tech reviews, but if they’re not as reputable as I believe they are, I would very much like to stop believing they are.
I don’t follow the hardware scene much nowadays, but I don’t think AnandTech was ever considered the “gold standard” except maybe by AnandTech itself. It’s a commercial website, not horrible, but not outstanding either. Garden-variety hardware reviews, more or less. In any case, I trust discussion on the forums much more than I trust official reviews (recall the Sturgeon’s Law).
I’ve found that modern hard drives tend to be quite reliable for consumer purposes; we’ve come a long way since the bad old days of the Click of Doom.
Their enclosures, not so much. I’ve had three backplanes for external hard drives, from three different manufacturers, fail in as many years. And one cable. But that table won’t give you any information on how common this sort of thing is or how to mitigate your risk.
modern hard drives tend to be quite reliable for consumer purposes
Heh. I’d say the reverse: modern hard drive are not reliable enough for consumer purposes since consumers typically don’t make backups and a failed hard drive is a disaster. They are sufficiently reliable for professional purposes where when a drive fails you just swap in another one and continue as before.
Their enclosures, not so much
Yeah, these are usually cheaply made. But then if an enclosure fails you just get another one and no data is lost or needs to be recovered from backups.
People are perennially interested in the reliability of hard drives. Here is useful hard data. Summary:
tl;dr Avoid 3Tb Seagate Barracuda drives.
I spend time in hardware enthusiast communities and not so impressed with Backblaze. Even here, the Seagate failure rates seem suspiciously anomalous.
Also, SSDs, which are probably a better match for most people here (my rig has run a 256 GB SSD for the past 2.5 years and I’m yet to want for more storage). Especially for laptops; they use less power (= your battery lasts longer) and can stand up to shock (so your laptop doesn’t break if you drop it).
I did not mean to endorse any particular service or give recommendations as to which storage devices should people buy. I found hard data which is rare to come by, I shared it. If you think the data is wrong or misleading, do tell.
Consensus is that modern HDD’s from reputable manufacturers have approximately equal low failure rates, especially after the first year. You should still back up important data (low != 0), but the differences failure rates in consumer space is small enough to not really sway purchasing decisions.
Their methodology probably doesn’t extrapolate well because they’re testing the drives in what amounts to a NAS and the WD reds (which did well) are NAS drives, and therefore designed to operate 24⁄7 with vibration and nongreat cooling, whereas the Seagate Barracudas are just absolutely not NAS drives (unlike, say, the Seagate NAS drives). So, it’s not really surprising they had a much higher failure rate, but it’d also be incorrect to conclude that you should avoid them. If I’m building a rig for work, internet use, or gaming {1}, then my HDD’s going to be in a well-cooled, non-vibrating environment, and not used in use 24⁄7, so I’m essentially throwing away 15% price premium for the WD Red’s (or 60% for the HGST Deskstar’s). OTOH, if you’re backing up your data locally on a NAS, pay the gorram premium.
{1} Again, though, SSD’s are increasingly likely the way to go. You can get a sufficiently good 256 GB SSD for about the price of a 3 GB HDD and if you’re never going to use more than 250 GB (which, I’m guessing is at least 80% of people reading this who don’t already know whether an SSD or HDD better meets their needs), you’re essentially getting substantially better performance (up to an order of magnitude), more reliability, and less noise for free. I harp on this because SSD’s come in a 2.5-inch form factor and the more the standard storage option is SSD, the more cases won’t have a whole bunch of room taken up with 3.5-inch bays I don’t use. More importantly, there’ll finally be budget laptops that I don’t have to immediately take apart, clone the OS onto an SSD, reassemble, and figure out what to do with the HDD it came with just to get a decent experience. Gah! SSD’s are the right choice for most people and there’s externalities when they get HDD’s instead because “more gigabytes”.
I am sorry, the link shows hard data which disproves that statement and not in a gentle way, either.
Didn’t your first sentence state that all failure rates are “approximately equal”? Make up your mind.
Assumption not in evidence. I’ve seen a LOT of computers totally taken over by dust bunnies :-) The reason you go look at that grey disk where the fan vent used to be is that your bios starts screaming at you that the machine is overheating :-D
Yes, but that’s irrelevant to the original post which looks at reliability of rotating-platter hard drives. If you think you don’t care about the issue, well, what are you doing in this subthread?
My above comment was poorly written. Sorry. Hem.
Consumer-grade HDD’s, used properly, all have about same, low failure rate. If you treat your desktop like a NAS or server, they will drop like flies (as evidenced). If you treat your desktop like a desktop, then a lot of the price-raising enterprise-grade features (vibration resistance, 24⁄7 operation) count for zilch. They’re still higher-end drives, and will last longer, but assuming you give your desktop a fraction of the maintenance you give your car (like, take 5 minutes to blow it out every other year), not a lot.
Mea culpa. I’ll give you heat, but vibration tolerance and 24⁄7 operation are enterprise-grade features with minimal relevance to desktop hard drives. Evidence. Evidence. Why I’m inclined to distrust anything Backblaze publishes + evidence.
tl;dr Looking at this data and concluding “avoid Seagate Barracuda drives” is a bit like noticing that bikers survive accidents more often when they’re wearing a helmet and then issuing a blanket recommendation to a population primarily of car-drivers to wear bike helmets. Sure, it’ll reduce your expecting mortality when you go out for a drive, but not nearly as much as you’d expect from the biking numbers.
Sigh. No. Really, go look at the data. I am not going to take the “consensus” of the anand crowd over it.
Hitachi Deskstar 7K2000 is a consumer-grade non-enterprise hard drive. In the sample of ~4,600 drives it has 1.1% annual failure rate in the NAS environment.
Seagate Barracuda 7200.14 is a consumer-grade non-enterprise hard drive. In the sample of ~1,200 drives it has 43.1% annual failure rate in the NAS environment.
Those are VERY VERY DIFFERENT failure rates.
I, for example, have five-drive zfs array at home which is on 24⁄7. I am very much interested in what kind of drives will give me a 1% failure rates and which kind of drives will give me 43% failure rates. I am not average, but I hardly think I’m unique in that respect in the LW crowd.
Do we actually disagree about anything?
We certainly agree that the Barracuda’s are crap in NAS’s. I believe that WD Red’s are a major improvement and Hitachi Deskstars a further improvement, which is just reading the Backblaze data (which is eminently applicable to NAS environments), so I’m we’re in complete agreement that, for NAS’s, Barracuda << Red < 7K2000.
However, I also contend that, in a desktop PC, a lot of what makes the Reds and 7K2000 more reliable (e.g. superior vibration resistance) will count for very little, so they’ll still fail less often, just not 1/40th as much. Even if they’re four times as reliable, moving from, say, a 4% annual failure rate vs a 1% annual failure rate may not be worth the price premium (using Newegg pricing, the Hitachi drive costs 72.5% more, but on Amazon, the Hitachi drive is cheaper. Yay Hitachi?), especially since RAID 1 is a thing (which would give us a 0.16% annual failure rate at a 100% price premium). Obviously, if you can find higher-quality drives for less than lower-quality drives, use those. But, in what we’d naively expect to be the normal case, if you’re paying for features that drastically reduce failure rates in NAS environments, but using your drives in a desktop environment where these features are doing little to extend your drive life, then you’re probably better off using RAID 1.
(Why do I use low single-digit annual failure rates? Because I remember Linus of Linus Tech Tips, who worked as a product manager at NCIX and therefore is privy to RMA and warranty rates, implied that’s about right. He produces a metric shit-ton of content, though, so there’s no way I’m going to dig it up.)
I’m also interested why you’re dismissive of AnandTech. I currently believe they’re gold standard of tech reviews, but if they’re not as reputable as I believe they are, I would very much like to stop believing they are.
Yes. You keep saying that there are no significant differences in reliability between hard drives of similar class (consumer or enterprise, basically) in similar conditions. I keep saying there are.
I don’t follow the hardware scene much nowadays, but I don’t think AnandTech was ever considered the “gold standard” except maybe by AnandTech itself. It’s a commercial website, not horrible, but not outstanding either. Garden-variety hardware reviews, more or less. In any case, I trust discussion on the forums much more than I trust official reviews (recall the Sturgeon’s Law).
I’ve found that modern hard drives tend to be quite reliable for consumer purposes; we’ve come a long way since the bad old days of the Click of Doom.
Their enclosures, not so much. I’ve had three backplanes for external hard drives, from three different manufacturers, fail in as many years. And one cable. But that table won’t give you any information on how common this sort of thing is or how to mitigate your risk.
Heh. I’d say the reverse: modern hard drive are not reliable enough for consumer purposes since consumers typically don’t make backups and a failed hard drive is a disaster. They are sufficiently reliable for professional purposes where when a drive fails you just swap in another one and continue as before.
Yeah, these are usually cheaply made. But then if an enclosure fails you just get another one and no data is lost or needs to be recovered from backups.
Unless the manufacturer in their infinite wisdom has enabled hardware encryption with the keys stored in the backplane.
Ah. Well...
-- Doctor, it hurts when I do this.
-- Don’t do this, then.
The trouble is that it’s the manufacturer that does it, and the user who gets hurt.
It’s up to the user not to buy broken hardware :-P