TR Forums

Krogoth wrote:

From what I read over the years. Large-scale HDD failures aren't really vendor specific and they are just a bad batch that managed to get through QA/QC and get out onto the market.

Backblaze would disagree with you.

Deanjo wrote:

Backblaze would disagree with you.

http://blog.backblaze.com/2014/01/21/wh ... uld-i-buy/

Source for those interested.

Their blogs are really worth reading, do consider that their experience can be somewhat backblaze specific since they had to take care to watch things like vibrations and interactions between various vibrations with large drive installations, so it's not industry wide stats, but they do go into how they account for their statistics which anybody can learn to apply.

But I've seen tons of people that tried to justify their purchase for a home HD based on backblaze, and that is just pure lunazy.

Aphasia wrote:

Their blogs are really worth reading, do consider that their experience can be somewhat backblaze specific since they had to take care to watch things like vibrations and interactions between various vibrations with large drive installations, so it's not industry wide stats, but they do go into how they account for their statistics which anybody can learn to apply.

But I've seen tons of people that tried to justify their purchase for a home HD based on backblaze, and that is just pure lunazy.

I agree that using it as sole justification is silly, but completely ignoring it doesn't make sense either.

I realize that all hard drives have a statistical chance that they will fail. But there's also a lot of variation, so it comes down to the luck of the draw, regardless of what brand or model you buy. All else being equal, I'll choose Hitachi/Toshiba/WD drives over Seagate. But if the price is right (and especially if I plan to use RAID-1, which mitigates the impact and inconvenience of random failures), I'm still willing to use Seagate too. This has been my stance since before the Backblaze report; it was based mostly on personal experience, and that of friends and co-workers.

The 7200.11s did indeed seem to have a much higher failure rate overall; but I think they've gotten significantly better since then. Are they as good as the competition *today*? I really don't know for certain, but my gut says "maybe not quite".

True that, of course it's an indicator, but you have to look what is behind the failures as well, like their incredible failure of green drivers that were because spin up/spin down interferences for instance.

As far as I can remember all the hard drive makers have gone through difficulties, when I worked in sore, we had a shipment of Wester Digital Caviars that we had to send back 30% from... not to mention there were others like Connor that also had troubles, and of course, the infamous IBM Deathstar's.... that worked fine for me, but not many else So it's not to say only one brand is bad and another is good, some models might have troubles, then others are just fine while it's the opposite for another manufacturer.

I have 8 2TB Seagate Barracuda LP drivers running, 4 in Raid 5 in my Server, and 4 in my NAS, then as spare cold drive I bought just to have as a replacement if either of the raid 5 sets has a drive die. Funnily enough, I had my first drive go in a week or so after I had just built the server. Had to rebuild the array outside on my workstation hotplugging each driver on a started workstation because I use a highpoint el cheapo raid. Cant start windows with broken array(despite another drive as boot drive) and cant rebuild in raid bios, needs to go from within windows.

Except that first drive going bad, they have been rock solid for a couple of years now. That's when I got the NAS to run mirror backups to because when the failure occurred, I only had USB backups that of course, don't get done as often since it was a manual backup.

Aphasia wrote:

True that, of course it's an indicator, but you have to look what is behind the failures as well, like their incredible failure of green drivers that were because spin up/spin down interferences for instance.

Yup, I've used a few, and I monitored the load/unload count closely at first, to make sure I wasn't getting "had" by this issue.

Wow, that is crazy. Recently though, I had a 2TB WD Green drive kick the bucket on me after 2 months of use. Smart data revealed 10 days of actual powered on use. Since I travel for work quite often I'm never home and I just keep the computer turned off. So most of the time, even while on these drives were idle as they were basically storage drives. The Samsung 840 Evo 250GB SSD & WD Green 500GB drives do all/most of the work from day to day.

Just goes to show, no one is safe! lol

Aphasia wrote:

not to mention there were others like Connor that also had troubles

Seagate purchased Connor in 1996. Then they purchased Quantum in 2005, and Maxtor in 2006. Quantum was ok, but Maxtor was another drive manufacturer I didn't trust because of my experience with high failure rates.

I just bought a seagate 1tb for backup. It is the 7200.12. I am scared. Like Blair Witch Project when the girl tell the camera she is scared scared.

Quantum. MMmmm Bigfoot. Big drive, big storage, big noise, big vibration. I have one somewhere, I believe it works to this day.

I'm not too worried about my Seagate drives, and I don't think anyone else here should be, either. Important data goes on fault-tolerant arrays, and gets backed up sometimes, so the risks are mitigated as much as feasible.

toki wrote:

I just bought a seagate 1tb for backup. It is the 7200.12. I am scared. Like Blair Witch Project when the girl tell the camera she is scared scared.

Like I said back on page 1, I've got four 7200.12s which have logged thousands of hours of uptime, problem-free. So it's not like they are cursed.

The problem I have with Seagate is that they seem to have more "bad batches" than the others. You may buy a couple that are perfectly fine and will last ages. I have some 7200.10's that are about 8 years or so old and still going strong in a RAID10 array on a 3ware controller. Most of my mechanical SATA drives are WD though.

Well, and the 7200.10 was before they had the big run of problematic drives (7200.11).

i do WD for my main drive as I have always used them. I did the seagate for backup because it was a little cheaper and I won't have it plugged in except to move stuff onto and off.

The thing is that Seagate has tons of technology that should be good. The barracuda series was long used in business on scsi and their 7200RPM barracudas and 10K RPM cheetah drives. I remember using the old 3.5" half height(Yes, that is the heigh of a 5.25" slot) scsi drives with 10K RPM. Those were monsters back then. The spin up had the sound of a turbine revving up until a point.

During the writing of a long post on a Swedish forum as a reply to the backblaze blog, I perused the Wikipedia articles on HD manufactuers... interesting read actually. Especially look at the chart on the Second link.

http://en.wikipedia.org/wiki/History_of ... isk_drives
http://en.wikipedia.org/wiki/List_of_de ... ufacturers

Deanjo wrote:

Krogoth wrote:

From what I read over the years. Large-scale HDD failures aren't really vendor specific and they are just a bad batch that managed to get through QA/QC and get out onto the market.

Backblaze would disagree with you.

Taking the word of a single study rather a list of many studies? If you look around and go back far enough. You will realize that almost every extant manufacturer had periods of bad units hitting the market. It is usually a single line or two not their whole inventory. It creates a massive flamewar among the datacenter and geek crowd who get burn by it and swear to never try the vendor again.

You realize that Seagate is currently ships out the most HDDs on the market? Statistically speaking, you would normally expect them to a more HDDs failure on average than their competitors.

To be honest though, HDDs have been getting less reliable over the last decade. They have become victims of commoditization and the laws of physics. The manufacturers are in deadlock to grab the most marketshare and they already used practically every trick you can pulled with HDD technology to increase performance and capacity. Yet there's still massive demand to keep increasing both areas. The performance-size (10K-15K RPM units) already ceded their niche to the emergence of SSD market. It is becoming more and more difficult to increase the density of the platters. Modern HDDs are engineered with rather fine tolerances while dealing with razor-thin profit margins. It is a practically a miracle that failure rates aren't more commonplace.

Krogoth wrote:

Hard disk drives have ceded their niche to the emergence of SSDs.

Buggy whips.

JustAnEngineer wrote:

Krogoth wrote:

Hard disk drives have ceded their niche to the emergence of SSDs.

Buggy whips.

SSDs have their own set of issues and limitations. It is unlikely that they will ever reach the same GB/$$$ and density ratio as magnetic media.

It is not cost-effective nor practical to replace PiB of capacity in a large datacenter with solid-state media.

JustAnEngineer wrote:

Krogoth wrote:

Hard disk drives have ceded their niche to the emergence of SSDs.

Buggy whips.

A more accurate comparison would be the introduction of automobiles.

Trains lost some of their usage, but they still have their niches. A single train is going to haul the same amount of goods as a fleet of semi trucks and is cheaper than air travel.

It would be more likely for a new memory system to replace BOTH hard drives and SSDs.

I'm old enough to remember when Seagate made excellent drives. I remember Connor drives, although I never had any. Had a couple of Quantums over the years. All failed. I had fairly good luck with the Maxtors. They had a 20GB model that was around forever. The IBM drives were awesome right up until the moment they bought the farm, which was usually fairly quickly. Didn't WD go though several years when they were garbage? It's really weird how unreliable hard drives have become in the past 7 or so years.

The Swamp wrote:

I'm old enough to remember when Seagate made excellent drives. I remember Connor drives, although I never had any. Had a couple of Quantums over the years. All failed. I had fairly good luck with the Maxtors. They had a 20GB model that was around forever. The IBM drives were awesome right up until the moment they bought the farm, which was usually fairly quickly. Didn't WD go though several years when they were garbage? It's really weird how unreliable hard drives have become in the past 7 or so years.

You're always going to have a percentage of bad drives, but it seems that drive makers go through "bad periods", or have certain product lines which are more prone to trouble. I had a rash of defective Western Digital drives in the 120-250GB range when they first released SATA models about 10-11 years ago. It was enough to make me switch to Seagate almost exclusively for a period of time, because they were giving 5-year warranties. A handful of years later, Seagate shortened their warranties and drive quality seemed to go down some, so now I'm mostly back to WD.

What concerns me is how few drive makers are still around. I can easily name a half-dozen brands I've owned which are now defunct. Today, you've got Seagate, Toshiba, Western Digital, and that's it. 3 manufacturers. Everyone else has either been bought out, sold their HDD division, or went bankrupt. It's a very unhealthy situation.

Krogoth wrote:

Deanjo wrote:

Krogoth wrote:

From what I read over the years. Large-scale HDD failures aren't really vendor specific and they are just a bad batch that managed to get through QA/QC and get out onto the market.

Backblaze would disagree with you.

Taking the word of a single study rather a list of many studies? If you look around and go back far enough.

You will realize that almost every extant manufacturer had periods of bad units hitting the market. It is usually a single line or two not their whole inventory. It creates a massive flamewar among the datacenter and geek crowd who get burn by it and swear to never try the vendor again.

What i relevant is what is their recent history like. Going way back to the years of lore has very little relevance to how their products are now. AMD made competitive chips at one time that could go head to head with intel in terms of performance. It doesn't mean that they still are.

I am not only going by backblazes findings. I also saw similar percentages of failures with seagates when I was at Apple in their products. Again, there the situation was the same with the Seagates having the largest percentage of failure over several years. Then there is also my personal experience dealing with thousands of systems over the year as well as the observations of several IT professionals that I know personally that also fall in line. Backblaze isn't the only source, it's just yet another that has observed the same and put their data out, with a good sample size, out in public.

You realize that Seagate is currently ships out the most HDDs on the market? Statistically speaking, you would normally expect them to a more HDDs failure on average than their competitors.

In total numbers of failures sure, percentage of failure wise however it should fall in line with the other manufacturers. It doesn't and is much higher than the others. They also have shown a trend of sharply dropping in reliability as time goes on during the life of a drive. That too does not fall in line with other manufacturers. In fact backblaze's data show that drop off during what would be the drives warranty period. The drives in question in their data largely consisted of post 7200.11 drives (only 539 in use out of the 12,765 Seagates in used in the study) and newer drives were still experiencing much higher failure rates then their competitors and the sample size with the Hitachi's were about the same but experienced far lower failure rates, despite them having a longer length of service.

On the other hand, it wasn't 12K drives of the same model though.

And they also went on to see that temperature can be a factor for some drives, while it doesn't matter for other, which muddle the waters a bit. I've always had some cooling on any mechanical drive, because usually it doesn't hurt, but it seems that for some drives it can make a difference, you just don't know which.

Then you have the option of recent history, considering IBM/Hitachi's track record, backblaze still finds them very reliable, so when are that recent history getting a downgrade to worth another look, because things like that can easily change with a technology shift, engineering or leadership shift, etc, etc.

The most telling thing about the Backblaze study I find is that over a 36 month period, they find that the survival rate plummets for Seagates with 1 in 4 drives failing vs the others that are 1 in 20 failing.

And yet they are continuing to use those drives despite the failure rate because it's still economically feasible.

Think you missed the point there though, despite being a hint, you can extrapolate a pure number outside of their environment, for instance, 2 drives wouldn't run at all basically because of vibration in their environment, does that mean's the drives are junk, probably not, just that you can't load 48 of them into a chassi and expect it to work without taking care of other factors. Other drives they found had issues with temperature adding another dimension, that is 3 specific models of drives, etc, although you do need to read about the other posts to those numbers.

Meanwhile, if you use the 4TB model, they have the same annual rate as the wD's, The 3.0TB model bit more, while the 1.5TB model has a vastly higher annual failure rate accourding to data. You can probably draw more hint's of that, but you do need to factor in the fact that since 2 specific models make up 75% of their Seagate data, and another one for another 15% or so. The numbers on those two models will skew the results if one of those models have a much higher rate of failure, so is that representative for the whole of Seagate's lineup.

Perhaps, but it's not clearcut since you would need an environment akin to backblaze to be able to directly compare numbers. So I expect any numbers would be vastly different in an home environment, both better and worse in many instances. Just look, one of the "worst" ones, avg age of just short of 4 years in a 24/7 48 drive chassi environment, now translate that into another environment, bet you would have a hard time to extrapolate which factors would affect the drive live more, spin up/down cycles, running time, temperature, usage pattern.

Note that they don't have any spec's on WHAT failed in the drives, only that they failed. So if the bearings failed, they the drives would be toast either way, now of something in the actuator failed, in a different environment you might have a medium span of 6-8 years depending on usage patterns, etc.

Aphasia wrote:

And yet they are continuing to use those drives despite the failure rate because it's still economically feasible.

Think you missed the point there though, despite being a hint, you can extrapolate a pure number outside of their environment, for instance, 2 drives wouldn't run at all basically because of vibration in their environment, does that mean's the drives are junk, probably not, just that you can't load 48 of them into a chassi and expect it to work without taking care of other factors. Other drives they found had issues with temperature adding another dimension, that is 3 specific models of drives, etc, although you do need to read about the other posts to those numbers.

Granted they are not in an ideal environment, they are however in the same environment as the others. Level playing field.

Meanwhile, if you use the 4TB model, they have the same annual rate as the wD's, The 3.0TB model bit more, while the 1.5TB model has a vastly higher annual failure rate accourding to data. You can probably draw more hint's of that, but you do need to factor in the fact that since 2 specific models make up 75% of their Seagate data, and another one for another 15% or so. The numbers on those two models will skew the results if one of those models have a much higher rate of failure, so is that representative for the whole of Seagate's lineup.

You also have to remember that the 4 TB drive also has the shortest length of time for that data. 0.3 of a year. As the graph clearly shows, the seagates traditionally start off well and trend steadily down as time goes on.

Perhaps, but it's not clearcut since you would need an environment akin to backblaze to be able to directly compare numbers. So I expect any numbers would be vastly different in an home environment, both better and worse in many instances. Just look, one of the "worst" ones, avg age of just short of 4 years in a 24/7 48 drive chassi environment, now translate that into another environment, bet you would have a hard time to extrapolate which factors would affect the drive live more, spin up/down cycles, running time, temperature, usage pattern.

It's not as different as you think. Once you start adding drives to a system the vibration also increases. Things get even worse in consumer grade multi-drive storage products as they usually have next to nothing for vibration and usually cooling and ventilation is poor.

Note that they don't have any spec's on WHAT failed in the drives, only that they failed. So if the bearings failed, they the drives would be toast either way, now of something in the actuator failed, in a different environment you might have a medium span of 6-8 years depending on usage patterns, etc.

That is true, they do however notice a trend with "green" drives that spin down frequently. They observe that is likely due to their spinning down. Spinning down is fairly common in a consumer system where as in a mass storage array like Backblaze uses they are kept in a fairly steady state and do not see the amount of spindown cycles a consumer would put their drive through. Consumers have also notice this correlation over the years with the various laptop drives they have had over the years. That is why many people that are concerned about the longevity of their drives use utilities or set the drive feature to keep spinning no matter what even though it comes at the expense of battery life.

While we don't have particularly concrete evidence on reliability in typical consumer applications, we have a whole lot of anecdotal evidence and the Backblaze study showing Seagate doesn't do very well. Sure, it's not perfect data, but that doesn't mean it's not helpful. I am prepared to say, based on that imperfect data, that Seagate is not more reliable than WD.

Given that I'll consider it at best equal in reliability, and potentially much worse, what I'm lacking is any reason Seagate is a good option. I haven't checked in on the HDD market in a while (when I last did the baseline capacity was ~320 GB, not 1 TB), but last I knew it was a very undifferentiated market, with little reason to choose one option over another besides price, capacity, and (perceptions of) reliability. Browsing Newegg just now, Seagate and WD still seem to be fairly even on $/GB. Are Seagates faster? Do they consume less power? Are they quieter? The two companies have roughly equal revenue despite the reliability perception, so obviously I'm missing something here.

I was very sleepy the first time I saw this picture and the first thing that came to mind was "nice hubcaps."