Latest Backblaze reliability data shows carnage for 3TB Seagate drives

Online backup provider Backblaze made headlines last year with a reliability study based on over 25,000 mechanical drives. Unlike previous publications in this vein, the report listed failure rates for specific makes and models. The data confirmed a lot of the anecdotal evidence that suggested Seagate drives were less reliable than the competition. Now, there's a new dispatch with updated stats through the end of 2014.

The most interesting trend pertains to 3TB units. Drives with that capacity suffered higher failure rates regardless of the manufacturer, and there's a familiar face in the spotlight.

Yikes.

The vast majority of the 3TB Seagate failures are tied to a single model: the Barracuda 7200.14. That drive's annual failure rate jumped from under 10% at the end of 2013 to over 40% a year later. There's no indication of why those 'cudas are failing at a dramatically higher rate, but the sample size is pretty big. Backblaze has over 1,100 examples with an average age of 2.2 years.

Although the 3TB contenders from HGST and WD also suffered higher failure rates over the past year, the overall percentages are much lower—especially for HGST. Drives from that manufacturer, which is owned by Western Digital, continue to be the most reliable in Backblaze's storage pods.

The numbers for Seagate's newer Desktop HDD.15 4TB provide some salvation for the company's battered reliability rep. Across over 12,000 units, the HDD.15's failure rate is now only 2.5%. The average age for those drives is less than a year, but they're failing less frequently than the 3TB 'cudas did at the same point in their lives. Perhaps the HDD.15's slower 5,900-RPM spinde speed is a factor.

Updated reliability stats on 17 different drives are available in the full report, which is worth perusing. Do Backblaze's experiences continue to jibe with what TR readers are seeing in their own systems?

Comments closed
    • hunkeler1961
    • 5 years ago

    I have had a 3 TB Seagate drive failure within the past six months and replaced it with the same drive. If you are good at keeping track of warranty receipts, purchase your drive with American Express as they double the manufacturer’s warranty period. Make sure you make good backups. I was able to restore all the data from backup (it was a data only drive) and file a claim with American Express Insurance for a refund of the original purchase price. The new drive cost less so I ended up having to spend less money for my trouble.

    Registering the drive with the manufacturer makes it easier to file a warranty claim with them if you need to in the future. I did not do that originally so to investigate if the drive was still under warranty, I had to register the drive retroactively. Surprisingly, Seagate’s date of when the warranty expired seemed to be based on the manufacture date and not the date I purchased it. Luckily the AmEx extended warranty kicked in and solved the issue for me. Had Seagate honored the purchase date to determine the warranty expiration, AmEx Insurance would have been off the hook.

    • DarkMikaru
    • 5 years ago

    “Do Backblaze’s experiences continue to jibe with what TR readers are seeing in their own systems?” In a word? Yup. Without question! Newegg reviews pretty much corroborate Backblaze’s findings.

    [url<]http://www.newegg.com/Product/Product.aspx?Item=N82E16822148844[/url<] If you look at the reviews for this drive it lines up almost exactly. 42% is horrendous! When looking at Newegg reviews I add the bottom two as negatives and the top 3 as positives. I've found over the years that gives the best overall view. However, it is worth mentioning that Seagate isn't alone. I've said before I've been eyeballing replacing some of my older drives with newer 3TB & 4TB models and they all seem to suffer from high failure regardless of manufacture. But it does make you wonder... what the hell happened in a year?

    • flip-mode
    • 5 years ago

    By the two dead Seagate 3TB drives sitting in the drawer next to me, I am completely not surprised.

    Edit: I am surprised the 4TB model is looking so good.

    • Prototyped
    • 5 years ago

    As part of the horse-trading deal between WD and Toshiba, when WD bought HGST, Toshiba got plants and tooling for 3.5″ drives from HGST. Toshiba’s current 3.5″ drives, like the DT01ACA300, are just rebadged old HGST drives, so unless Toshiba has chnged something at the plants, the reliability numbers should apply to the 3.5″ Toshiba drives. (I’ve seen some comments whining about the 2.5″ drives — those have a different pedigree. Turns out, in the 21st century, branding is near worthless, in general. Everyone that has good products also happens to have a larger number of crap products.)

    [url<]http://goughlui.com/2013/02/26/toshiba-dt01aca300-aka-hitachi-hds723030ble640-drives/[/url<] I have one of these and one Barracuda 3 TB 7200.14 drive (model ST3000DM001) in a RAID 1 mirror. Wish me luck!

      • ronch
      • 5 years ago

      As I understand it, WD gets HGST’s enterprise and mobile drive lineups. The standard desktop lineup went to Toshiba. I have a Hitachi 1TB unit that looks exactly like the Toshiba desktop drives you can buy today.

      I’ve been sold on Hitachi since I got mine, buying Hitachi for family members as well. I will have to go with Toshiba the next time I get an HDD.

    • Juba
    • 5 years ago

    got several Seagate 3TB HDD running for years, no issue.

    i remember when i was a WD only and had all the problems with there “blu” and “black” HDD, then made the switch to seagate and never looked back.

    • ZGradt
    • 5 years ago

    I have an 8 drive raidz2 array that started out as mostly 3TB Barracudas. I only bought them when they were on sale, and they went on sale a lot. It started out as a 5 drive raidz1 I believe. Over the two years or so I’ve had it, I’ve had to return at least 3. Another failed out of warranty. After the 2nd or so return, I swore off Seagate and switched to Toshiba. The array is 5/3 Seagate/Toshiba right now. One of the surviving Seagates has a disturbing amount of allocated sectors, and another has 128. The Toshibas have none. It’s a very good thing that I can lose two drives in that array without data loss.

    I also have a 7 drive raidz1 array made up of a mix of different NAS rated 3TB drives. WD Reds, Seagate NASWorks, and a Hitachi. I haven’t had any problems with those. It made me wonder if my problem was due to using the non-NAS drives 24/7, but articles like these lead me to believe that they’re just crap.

      • Waco
      • 5 years ago

      It’s a stupid notion, but I feel like the 1.5 TB and 3 TB drives are rejects from the 2 TB and 4 TB drive pools, hence the higher failure rates.

    • Krogoth
    • 5 years ago

    I rather have data that includes firmware version, model number, when and where it was made.

    It will give us a better idea what to look for.

    • NovusBogus
    • 5 years ago

    Damn, I think I bought one of those not too long ago and actually paid extra because it had gotten better reviews than similarly priced competitors. Well, it’s a cold-storage backup drive so it shouldn’t be too big of a thing.

    • MadManOriginal
    • 5 years ago

    Bad hard drive models are bad.

    Thanks, Backblaze.

    • joyzbuzz
    • 5 years ago

    Haven’t bought a Seagate for years solely because they always ran way hotter than WD.

    • dogchainx
    • 5 years ago

    I have had problems with all makes and models. My 4x3TB array of Seagates is still going strong after running 24/7 for two years. I have a 4x4TB seagates too, and not a problem.

    However, I have had one *external* 3TB seagate that bit the dust, though I’m betting its because of a bad USB controller and not the drive.

      • shaq_mobile
      • 5 years ago

      why didn’t you pull the drive and test it?

    • moose17145
    • 5 years ago

    I understand that these numbers are not exactly totally scientific, as many have stated. Yes there are holes in the data and stuff… But I don’t consider the data provided any less relevant or interesting.

    These drives are in working environments. So no. Not every drive is going to be the exact same temp as every other drive, and not every drive will see the same work load as some of it’s counterparts.

    I would be particularly interested to see google’s statistics for drive failures (I know they have published many documents on the topic, but I have never seen them publish the brands and their failure rates).

    • geekl33tgamer
    • 5 years ago

    Should I start to feel less confident about the 2 x 1TB 7200.14 in RAID0 and the 3 TB 7200.14 backing that up as well as the system SSD???

      • Visigoth
      • 5 years ago

      Hell yes! SSD all the way, man. And for archiving, I believe most manufacturers have dedicated hard disks for these purposes.

    • odizzido
    • 5 years ago

    My personal experience can be summed up by the chart provided.

    • egon
    • 5 years ago

    Hard not to see similarities in the way techies approach the question of hard disk reliability and the way adherents to homeopathy and the like come to believe what they do. For hard disks, there’s no really high-quality research, and in the vacuum, we trust the perceptions formed by our personal experiences, and are credulous about anything (such as the Backblaze report) that seems to confirm our beliefs.

    One major difference I suspect – if there were a gold standard study into hard disk reliability which contradicted us, many would accept the science, whereas for something like homeopathy, the science is clear but people tend to be ‘true believers’. Still, it’s surprising the extent to which, in the absence of that, we accept without much doubt our personal experiences as highly reliable and relevant.

      • A_Pickle
      • 5 years ago

      I think the uncontrolled reports of hard drive reliability are still preferable to the bunk science of homeopathy…

        • egon
        • 5 years ago

        But it’s only robust science that confirms it’s bunk. Take that away, and you’d have for homeopathy much the same as you have here – strong beliefs based on personal experience, backed by research which produces compelling results, but is low in quality.

        The two main, current sources of data on HDD reliability are the Backblaze and hardware.fr reports, and they give starkly different pictures. There’s little reason to be confident one is more robust and relevant than the other, but it’s the unfortunate default for the human mind to accept that which confirms its existing beliefs, while being dismisive toward contradictory information.

          • Deanjo
          • 5 years ago

          [quote<]There's little reason to be confident one is more robust and relevant than the other,[/quote<] As I pointed out in another post, they are measuring two very different things. hardware.fr only has data from the products it receives back for exchange, it does not include devices that are past their return period and have been sent back to the manufacturer for RMA or devices that simply were deemed not worth going through the hassle of exchange. Backblazes data comes from their own personal experiences where every defective item is accounted for. If you were to pit one againt another, Backblaze will have a more complete data set thus offering more true to real life first hand data, not just what gets reported back to them. Also hardware.fr is more susceptible to "bad batches" to skew their data. Back blaze acquires their drives from hundreds of different sources which is less susceptible to "bad batches" skewing the results. Also with Seagates lower DOA rate with the 4 TB drives that could also simply be a case of better packaging for shipping. Long term numbers will bring a better overall picture if the improvement is from changes made to their drives or simply a case of better packing.

    • Eggrenade
    • 5 years ago

    I had a ‘cuda that went down in a blaze of glory.

    It was a 500 GB 7200.10 (I believe; I can look it up when I get home) I bought with my first build in 2008. By the summer of 2011, it was a storage drive.

    I had my case open because I was working on some components. I turned my computer on and everything was working fine. About a minute after I got to my desktop, I started smelling smoke.

    I looked down at my computer and saw flames coming from my hard drive carriage. I flipped the switch on my surge protector and blew out the flame. The SATA power connector must have shorted or something—that’s where the fire was coming from. The power cable was fused to the drive.

    The whole experience was a little surreal, and it still seems like it had to be a dream. I really never thought my computer would catch on fire.

    There were a few files I wanted off of that drive (game saves and ripped media, most files were backed up) so I chipped away at the melted plastic power connector enough to connect a new cable from my power supply. To my great surprise, it worked. After that I quickly shut down my computer and removed the drive for good.

    Seagate honored their 5-year warranty (they don’t offer those anymore) and the replacement has worked ever since.

    • Takeshi7
    • 5 years ago

    It’s pretty obvious Backblaze purposely looks for statistics that make Seagate look bad. In this graph, the 6TB Seagate drives have a 0% failure rate, but they weren’t included in the graph because that would make Seagate look better than WD’s 6TB 3.1% failure rate.

    In their previous update they used graphs showing that their servers with WD drives filled up faster than Seagate, to make it seem that the Seagate drives were slower, but then they admitted the drives really had nothing to do with how fast the servers fill up, and had more to do with how much traffic they received the day the servers were installed.

    I want another source for HDD failure rates, because it really seems BackBlaze is biased.

    Edit: Hater’s gonna hate, but I make valid points that all of the downvoters know are true.

      • exilon
      • 5 years ago

      It’s pretty obvious you didn’t RTFA.

      Their 6 TB Seagate drives have been in service for less than half a year. They also didn’t put their 4 TB WD drives with 0% failure rate because… they were in service for less than a year. Coincidence? Or CONSPIRACY?

        • Takeshi7
        • 5 years ago

        The WD 6TB drives have been in service for 0.1 years. That didn’t stop them from including them in the graph… So yes it looks like a conspiracy to me.

          • exilon
          • 5 years ago

          You know that makes WD look bad instead of Seagate, right?

            • Takeshi7
            • 5 years ago

            The way they present it, by not including Seagate 6TB drives it’s less obvious by just looking at the graph. They purposely try to cover up WD’s shortcomings while emphasizing Seagate’s shortcomings.

        • Takeshi7
        • 5 years ago

        You still haven’t explained to me why they didn’t include the 6TB Seagate drives and the 4TB WD drives because they were in service for a year, but decided to include the WD 6TB drives that have been in service even less time.

          • DarkMikaru
          • 5 years ago

          Hate to say it Takeshi, but go look at Newegg, Amazon reviews for these drives. You know what they day about smoke right?

    • Bensam123
    • 5 years ago

    Definitely wish there were more studies like this. It’s impossible to do without a LOT of hardware and time, not something most people have access to.

    Based on where they’re getting hardware, it’s entirely possible that the pool could be tampered with (trash drives bought wholesale, cherry picked samples being sent to them because they know they report stats on them).

    • Bananaman
    • 5 years ago

    My heart just pooped its pants and my wallet will have to clean it up: I built a six-drive FreeNAS server with these drives. I had one fail within the first six months which had been replaced with a WD red. Now I’m really regretting my decision to not shell out the extra money for a whole fleet of Red’s from the beginning.

    • TEAMSWITCHER
    • 5 years ago

    Academically interesting, but irrelevant. Their application is massively different from my own and since no drive is 100% infallible – I still need to buy additional drives for backup regardless of which drive I buy. So far, going cheap has worked well for me and the drives do offer good performance. If it fails…it fails…and will get replaced. But I won’t loose sleep (or data) from reading this article.

    • swaaye
    • 5 years ago

    Working with drives of any sort means you’ll eventually deal with a failure. The statistics don’t seem to really matter. I mean, I had 75GXP Deathstars for years, I have had 3 of these 3TB Seagates (one did die after >1year), I’ve had head-park-happy Greens for years, etc. You just want to have a warranty to fall back on for a free replacement, and have backups of your data.

    If these people are using consumer drives in server roles, I wonder if the warranty is voided for them…

      • Deanjo
      • 5 years ago

      You can only RMA and ship back a drive for replacement so many times before it becomes a losing proposition. Courier shipping isn’t exactly cheap.

        • Ninjitsu
        • 5 years ago

        WD has free pick up and delivery, in my country.

          • Deanjo
          • 5 years ago

          [url<]http://support.wdc.com/warranty/rmapacking.asp?custtype=end[/url<] [quote<] Use a traceable carrier. Use a traceable carrier for transport or a signed for service (i.e., Airborne, Fed Ex, UPS, U.S. Priority Mail, etc.) with the post office (European/Canadian customers only) as they provide you with a tracking number and proof of delivery. [b<]The cost of sending a drive to the returns service center for warranty service (also known as the return shipping cost), is paid by the customer.[/b<] [/quote<]

      • Ninjitsu
      • 5 years ago

      [quote<]I had 75GXP [b<]Deathstars[/b<][/quote<] You [i<]what[/i<]?

    • Krogoth
    • 5 years ago

    Looks like the 3GB Seagate results are skewed by a bad batch. Nonetheless, it is rather problematic. Bad batches come from every HDD manufacturer, so there are no safe bets. Just hope that you don’t pick-up a unit from a bad batch.

    In my personal experience, HDDs typically die out after several years of mainstream usage patterns. I’ve never had much experience with units that undergo 24/7 usage patterns.

      • LoneWolf15
      • 5 years ago

      Remember the Seagate 7200.11 drives? Seagate has had entire model lines that were problematic –there’s no guarantee this is just a quality control issue. I had two of three 7200.11 1.5TB drives (the worst model of the line) fail in ten months. Fortunately, it wasn’t simultaneous, I was using RAID-5 and also had backups, but their bad attempts to fix the situation (which included firmware that bricked a number of users’ drives) and the shortened lifespan of those drives (it was rumored there was a hardware design flaw) caused me to dump them and go Western Digital (whose RMA policy has been much simpler for me in the event of the rare failure).

      I could see Seagate having more than a bad batch of 3TB drives. Even if it isn’t, quality control shouldn’t be that level of abominable on something that people regularly trust to hold their data. (Note: WD doesn’t get a free pass either; some of their Green drives have had issues in the past too, and I won’t buy them either).

        • Krogoth
        • 5 years ago

        The problem is that HDDs have become commodity goods and there’s constant pressure to keep $$$$/GB ratio low as possible especially now that SSD media presents a creditable threat. On top of that, HDD guys are getting pretty close what is theoretically and economical viable with spinner-based tech. Modern HDDs are built with very fine tolerances and firmware tailored to each unit (the reason why the old controller switch trick doesn’t work anymore).

        It is more of a surprise that failures aren’t more commonplace.

      • VincentHanna
      • 5 years ago

      indeed, and having the serial number of that line is good solid information.

      hard to call the data skewed, however, seeing as the data is just the data. “The 48 month failure rate of the Seagate Barracuda 7200.14 (ST3000DM001) 3TB has been skewed by over 1000 drives failing in their first 48 months of operation.”

      And recidivism statistics in the inner city are skewed by former felons committing crimes after being incarcerated.

    • anotherengineer
    • 5 years ago

    What “Grinds my Gears” the most about these, is the fact that most of these Hitachi drives are several years old (which is fine), but are no longer available at retailers, which grinds my gears.

    It would be nice if they had a “reliability chart” for drives that are readily available and in stock at most e-tailers.

    Don’t see car reliability charts comparing a model T ford with a 2013 ford edge.

    While all nice to know, it does not really help me if I want to go out and buy something that’s not available.

    edit – spelling & grammar

      • Deanjo
      • 5 years ago

      When it comes to long term reliability studies that is pretty much always going to be the case, nothing much you can do about that (unless you know how to time travel). All you can do is observe historical trends. One could argue as well that drives less than a year old shouldn’t even be included in the study (unless you are looking for infancy death).

      • TheMonkeyKing
      • 5 years ago

      While it is interesting what they discovered, they are not a testing shop. They are a datacenter and they are reporting on all their hard drives. So while it is interesting to note, their information is still anecdotal.
      – no standardized test plan or environment (unless you count the datacenter as a single unifying environment – which it ain’t)

        • Deanjo
        • 5 years ago

        [quote<]no standardized test plan or environment[/quote<] The flipside is most testing in a "no standardized test plan or environment" doesn't relate to real world use. There are typically too many static variables in a standardized test to reflect real world use. For example, automotive manufacturers use a very limited set of variables in their MPG rating which nearly nobody is able to achieve because they are not driving at 80 kph, in a wind tunnel, without elevation or driving surface changes.

      • VincentHanna
      • 5 years ago

      So, you want the 6 year failure rate on the 2015 models sold by WD and Seagate respectively?

    • HisDivineOrder
    • 5 years ago

    This just reminds me why I think I’ll sit out the 5TB and 6TB (and future 8TB) drive beta tests for a few years.

    • Arvald
    • 5 years ago

    Well it jives with my home server. I have had 4 Seagate 3TB and 1 WD 3TB.
    I’m down 3 of the Seagate’s (replacing with Red 4TBs now)
    All are just reporting excessive SMART errors.
    All 3 were consumer green variants and lasted over 1 year but under 3, so the warrantee was gone.

      • 3SR3010R
      • 5 years ago

      A sure sign that drives are unreliable is when the vendors went with 1 year warranties.

    • 3SR3010R
    • 5 years ago

    I hate drives that do the head-unload/load cycles aggressively.

    The WD drives I own that had that feature have failed for me way too early. Since failures occurred in the warranty period I have gotten replacements but then I sell those replacements on eBay as I don’t want to keep going through that install/fail/install cycle over and over again.

    [url<]https://koitsu.wordpress.com/2012/05/30/wd30ezrx-and-aggressive-head-parking[/url<]

      • stdRaichu
      • 5 years ago

      Yeah, I had read about this as well (but it doesn’t appear to be present on any of my WD30EZRX drives) and that the WD DOS tool wdidle3.exe didn’t work on it.

      Have you tried using [url=http://idle3-tools.sourceforge.net/<]idle3ctl[/url<]? It's what I used on my green drives and lots of people saying it works for the EZRX drives too.

    • Shambles
    • 5 years ago

    [quote<]A drive is recorded as failed when we remove it from a Storage Pod for one or more of the following reasons: The drive will not spin up or connect to the OS. The drive will not sync, or stay synced, in a RAID Array. The Smart Stats we use show values above our thresholds. [/quote<] This is probably the disparity between what consumers experience and what Blackblaze reports. A consumer will usually only replace a drive when scenario 1 happens. Hardly any consumer runs RAID and less even log SMART stats regularly. Also most consumer drives aren't rated for 24/7 use as they are used by Back Blaze. That being said a drive that can't stand up to a bit of workload and isn't reliable for more complex setups isn't a drive I want storing my media on my PC either. I have a couple 3TB Seagates and they've been fine for the past couple years but I'll be trying to stay away from them in the future. (It's also funny how HGST drives are more reliable than WD drives when HGST [b<]is[/b<] WD.

      • ddarko
      • 5 years ago

      Western Digital may now be the parent company of HGST but HGST drives are not re-badges of WD drives, at least not yet. HGST and WD drives are still mechanically different and anyway, many of the drives used Blackblaze were bought when the two companies were separate.

        • Shambles
        • 5 years ago

        Very true. You’d think that the WD engineers would go visit the HGST engineers and learn a few things but being from the mother company they probably just assume that there way is the best way.

    • DPete27
    • 5 years ago

    Probably why Seagate hdds are consistently the cheapest cost/GB. Cause/effect, hard to say.

      • stdRaichu
      • 5 years ago

      Someone on stronger medication than me should probably accuse backblaze of artificially collapsing the price of seagates so they can pick them up cheaper 😉

    • c1arity
    • 5 years ago

    It’s a running joke where I work when it comes to seagate failure rates. The mass majority of drives that fail within 1-2 years are seagate 7200.xx with the WD green drives being next in line. After that I’d say it’s probably WD blue but the blue drives are a distant 3rd. As such, every seagate drive I replace gets replaced with Hitachi (HGST), WD Black, or Toshiba.

    Best thing about seagate is the amount of money we make doing data recovery! Their drives are absolute crap.

      • chriscasimir
      • 5 years ago

      Not sure I’d recommend Toshiba drives. I have 5 of their 2.5″ drives here (admittedly all 4-7 years old) that have serious issues powering up correctly.

    • ThatStupidCat
    • 5 years ago

    HGST is a subsidiary of WD. [url<]http://en.wikipedia.org/wiki/HGST[/url<] As for myself I've never really had problems with WD drives and have no reason to complain here.

      • Voldenuit
      • 5 years ago

      Didn’t WD only keep their 2.5″ and 1.8″ drive business sold the 3.5″ business to Toshiba?

        • Deanjo
        • 5 years ago

        Yes that is the case. WD kept the < 2.5 drive business and then sold the 3.5 side to Toshiba.

          • Chrispy_
          • 5 years ago

          So it’s Toshiba’s fault that we have too many models?
          WD Red, Green, Black, Blue, Purple?

            • Deanjo
            • 5 years ago

            Different models with different specification and criteria (there are differences). WD had those models before they even acquired the HGST.

            • ronch
            • 5 years ago

            We have too many WD lineups because their execs just love Johnny Walker.

    • ClickClick5
    • 5 years ago

    My last Seagate was in 1997.

    The reason still stands above. Every. Single. One. Died prematurely. Longest one of the four I owned lasted a year and a half. Then pow. No longer booting Windows 98 SE.

    Got pissed at having to reinstall everything, so I went to WD. Been there ever since.

      • l33t-g4m3r
      • 5 years ago

      I’ve had similar experience with WD. However, I haven’t bought any of their newer drives, being WD is not that price competitive. I have had one of my 1.5 seagates die, but the 2gb drives seem to be really reliable. Probably not going to buy any more HDD’s since SSD’s are taking over. HDD’s are just too much of a pita to manage.

        • ClickClick5
        • 5 years ago

        I have only one complaint about SSDs: When they fail, they fail. I have tried many times to pull data off of failing SSDs and the success ratio to a failing hard drive is laughably bad. I tell the clients now, “If you experience a single hard lock or BSOD, turn it off, being it to me.” It might be nothing, or the drive is failing. Granted they controllers have improved leaps and bounds in the past two years, but when you have to tell a manager the super awesome and fast 1TB SSD lost all their data, they are not happy.

        Backup, backup, backup.

          • 3SR3010R
          • 5 years ago

          [quote<]I have only one complaint about SSDs: When they fail, they fail.[/quote<] You also need to be concerned with SSD's losing their data is left unpowered too long. [url<]http://www.virtium.com/blog/z-ssd-data-retention[/url<] So be sure to disable the "power off disk" in the OS and leave systems in sleep instead of powering them completely off. It is also a good idea to have a UPS on the system with a SSD if said system doesn't already run on a battery (laptop/notebook) as this report shows massive data loss is possible on SSD's when power is removed unexpectedly. [url<]http://www.infoworld.com/article/2613584/flash-storage/test-your-ssds-or-risk-massive-data-loss--researchers-warn.html[/url<]

          • stdRaichu
          • 5 years ago

          Sadly it’s the same with almost any solid-state technology; when it stops working, it normally does so catastrophically. In my experience HDD failures are about 40% mechanical (and therefore some chance of fixing it with some gentle percussive maintenance) and about 60% electronic (with about a 50/50 chance of fixing it with a replacement controller board and almost zero chance of repairing it if the actuator is FUBAR).

          A friend of mine does data forensics as a profession. They offer an SSD recovery service that can go as far as de-soldering the NAND chips and reading them individually and then re-assembling the drive image in software (they tried soldering the chips into a known-working controller but had a very low success rate). Even he doesn’t know how much this costs because it’s so expensive that not a single one of their clients has been able to justify the cost.

            • cygnus1
            • 5 years ago

            I bet they also provide no guarantee of success on that type of SSD nand recovery either. Between that and the price, that’s what steers most people toward just taking the loss of data and spending that money on setting up proper backups for the future.

      • albundy
      • 5 years ago

      Same experience with Seagate and WD. I also had other bad experiences with seagate drives. Had 2x 320gb drives where i found out that i was the unfortunate sucker that bought 7200.10 drives with the AAK performance dropping firmware. Apparently the AAK firmware was created for a specific OEM-server customer and should never have reached the retail channels. Thats what i get from buying from dell, cheap useless POS junk. I was pissed for a long while until I started buying Hitachi’s from newegg and never looked back!

    • auxy
    • 5 years ago

    Why are we still paying attention to these shills? The last study was thoroughly discredited, and I’m sure this one will be too. (´・ω・`)

      • Welch
      • 5 years ago

      I’m not sure I’d call their stuff discredited, it sure had some holes in it. I think that at times they make casual links by jumping to conclusions at times.

      Every brand has its POS model. I can name quite a few Seagate ones, there are WD junkers too.
      The 2 I can think of for Seagate were there first 2TB drives (the freaking worst) and then 160gb barracuda’s back in the day. WD and Seagate both had some issues with 750gb drives at some point.

      Otherwise I’ve been running Seagate 1TB Constellations (ES.3) without any issues at all in client file servers. I think I was using the earlier ES and ES.2 versions maybe, but in 500gb variants. All of which were the SATA variants, not SAS.

      • puppetworx
      • 5 years ago

      It irks me that the TR is still reporting this without any comment about there being no controls at all. The fact that temperature, vibration, reads/writes weren’t controlled means that this information is useless. Maybe the Seagate drives failed because they had by far the most use or were exposed to the highest temperatures or most vibration. We simply don’t know. They don’t even attempt to correlate, they just dump the failure rates and sites repost it as if it’s meaningful. It’s not.

        • nanoflower
        • 5 years ago

        I would expect that they attempt to keep all of their systems at a reasonable temp. Controlling for read/write and vibration isn’t going to happen since this is the real world results where usage varies. Over time I would expect that the usage for all drives would equal out but there’s no guarantee of that.

        • stdRaichu
        • 5 years ago

        In their past publishings, they’ve repeatedly stated that these results are only a release of their failure rate data in their environment and aren’t necessarily indicative of anything else.

        That said, running any sizeable number of drives under a production workload in a controlled environment where individual temperature and especially vibration are tallied with the results is next to impossible.

          • puppetworx
          • 5 years ago

          Yet this article doesn’t mention the lack of controls at all, all I’m saying is that it should.

        • Scrotos
        • 5 years ago

        Temp isn’t necessarily the issue you think it is:

        [url<]http://storagemojo.com/2007/02/19/googles-disk-failure-experience/[/url<] "The Google team took temperature readings from SMART records every few minutes for the nine-month period. As the figure here shows, failure rates do not increase when the average temperature increases. At very high temperatures there is a negative effect, but even that is slight." Certainly at the data center I have some stuff hosted at, they mentioned that they don't keep the temps super cold since there's no reliability difference and it saves on running cost. Plus, there's a ton of different hardware when I walk down the aisles of the data center, some all pro, some kinda slapped together. I don't see the Backblaze stuff being anything out of the ordinary. Also, it doesn't make them money to purposely try and fail their own hardware. They run a business. People think it's just a platform to slur Seagate, but hey, if I had that info I'd share it too, no matter the manufacturers involved, because I don't have any binding contracts with them. 😀

      • maxxcool
      • 5 years ago

      Ill back them up… Seagate SUCKS.. Went Wd and never have gone back. to may failures..

      • bwcbiz
      • 5 years ago

      Considering that we’re dealing with huge sample sizes, and the fact that they’re very happy with the Seagate 4 TB models, I’m not sure where you are seeing a bias. While they can’t give full control measurements for other environmental factors like power quality, temperature and vibration, it’s not likely that they would have different build standards for different data centers, or that drives from any particular brand would be concentrated in any particular data center.

      • ronch
      • 5 years ago

      Hmmm…. Do you happen to be connected in any way to Seagate?

        • auxy
        • 5 years ago

        It’s a shame that most people upvoting you on this are probably not aware of the clever and subtle reference you’re (hopefully/I assume) making to my vocal support of #gamergate.

        Of course, if I had any such affiliation, I would disclose it! (*´∀`*)

        I will say I buy a lot of Seagate and Toshiba drives; have had very few issues. Did have a Toshiba 3TB DOA a few weeks ago, not a big deal. DOAs happen.

        Mainly I have a hate-on for WD and their dishonest (or at least disingenuous) business practices; like selling the WD Blue and Black drives as separate products when the only difference is firmware. (For the record, I also sharply frown upon Intel, Nvidia, and AMD’s artificial market segmentation. Quadros are a lie!)

          • ronch
          • 5 years ago

          I never expected people to upvote that post. Was just bit by the sarcasm/witty/whatever bug.

          I’d really like to think that Seagate drives are as reliable and long lasting as WD or HGST but horror accounts about people owning Seagate drives seem to abound. I can honestly say I personally haven’t had any issues with Seagate but then I have never had a Seagate since 2004 except for a mobile 320GB unit I got in 2007. I’ll have to reconsider them on my next drive purchase.

      • Waco
      • 5 years ago

      Trolling as usual, eh?

        • auxy
        • 5 years ago

        Whom? Surely not me.

          • Waco
          • 5 years ago

          A bad study does not make them shills. It’s uncontrolled, therefore anecdotal at best, but there’s no evil intent here like you imply.

            • auxy
            • 5 years ago

            Aww, shucks, I was just being facetious! (⌒▽⌒) In my circles we find the relevance of the Backblaze data so laughable it honestly didn’t occur to me that someone might take [s<]it[/s<] my comment seriously.

            • Waco
            • 5 years ago

            In your circles?

            What circles would those be? How many thousands of hard drives do you buy per year?

            • auxy
            • 5 years ago

            Well, we comprise the largest hardware-oriented community on IRC, with over 500 concurrent members at any given time. Naturally as with any time you get >500 people in a room not everyone agrees but we have a lot of guys in there who do hardware purchasing for a wide variety of companies both large and small. Σ(´∀`;)

            We also have a rule that says you can’t require someone else to have personal experience with something to remark on it… (*´艸`*)

            • Waco
            • 5 years ago

            I see you’re using your Gold powers to downvote me. Cute.

            500 concurrent members, times maybe 2-3 drives each…that’s still nothing in the grand scheme of things.

            As for your rule? I manage petabyte scale filesystems. My personal server at home can survive 2/3 of the drives failing and 4/5 drive controllers failing without a hiccup. I do storage, it’s my job.

            • auxy
            • 5 years ago

            Is this dong-waving supposed to impress or intimidate me? (‘◇’)ゞ

            Either way you’re doing a rather poor job of it! ( *´艸`) You really think someone in charge of hardware purchases for a “large company” buys “2 or 3” drives? Geez. Even assuming I was using a relative or regional scale for “large company”, that’s still a foolish notion.

            • Waco
            • 5 years ago

            You complain about “dong-waving” when rule #1 of your IRC is the same damn thing?

            It makes no difference, your insinuation of knowledge is but anecdotes of others that you don’t really know.

            All hard drives fail. All batches have problems. All firmwares have bugs. No company, at scale, is substantially worse than any other in the HDD arena. By scale, I’m talking tens of thousands of drives at the minimum, not a specific run. No company worth it’s salt would buy a few thousand of the same batch and rely on them.

            • auxy
            • 5 years ago

            What? Rule #1 is “no idiots allowed”. And the aforementioned rule to which you refer is specifically the opposite of what you seem to think it is…

            As usual you make a bunch of presumptions based on things I didn’t actually say and then use those presumptions to justify name-calling and finger-pointing. (・へ・)

            For what little it is worth, my viewpoint has always been one of brand neutrality, so we actually agree on that (unless you’ve somehow misspoken entirely?), which makes this whole hate campaign you’re waging on me here even more juvenile and pointless than usual. Hooray for insecure little boys! \(^o^)/

            By the way, a guy with a job like yours must make pretty good money, yah? I do retail computer services and support. I make bunk; $10/hr. But even I was able to throw TR $45. Instead of whining about me downvoting you (totally deservedly), why not step up to the plate and give as good as you get? Unless you’re actually fine with being upstaged by a little girl all the time… (゜レ゜)

            • Waco
            • 5 years ago

            Hate campaign? Pointing out incorrect statements is not a hate campaign. I don’t care if you downvote me, but you do make it pretty obvious that you care.

            Keep on trolling, it doesn’t bother me one bit.

            • VincentHanna
            • 5 years ago

            [quote<]You really think someone in charge of hardware purchases for a "large company" buys "2 or 3" drives?[/quote<] You really expect anyone to take seriously your "friend of a friend on IRC" as being a reliable source to stand up to something like the backblaze internal study?

            • auxy
            • 5 years ago

            When did I say anything of the sort?

            • Waco
            • 5 years ago

            You implied it, no matter what you choose to say.

            • VincentHanna
            • 5 years ago

            39,696 drives would be a lot of anecdotes.

            • Waco
            • 5 years ago

            It is, which is why it’s still extremely interesting. 🙂

      • divide_by_zero
      • 5 years ago

      As someone who has literally replaced *thousands* of Seagate drives due to premature failure (mostly .11, but plenty of other series as well), the Backblaze results usually don’t surprise me. But wow, even I was shocked at the failure rate of the 3TB .14 drives.

      Every time Backblaze publishes more data, there’s always a few people who claim that their data is worthless and “discredited”.

      This argument is ridiculous.

      Yes, if you were devising a long-term HDD reliability study you would certainly design things differently. I’ll grant you that. It does not change the fact that this is the only data of this scale available to anyone in the general public. It also does align with what myself and my tech-industry friends have observed as well in more anecdotal settings.

        • JustAnEngineer
        • 5 years ago

        I believe the data. Anecdotally, my four examples of the ST3000DM001 are still working fine (fingers crossed and backups up-to-date). At less than 1/3 the cost of equivalent Western Digital Caviar Blue or Black drives, it would have been hard not to buy the Barracudas. Now that Toshiba and Hitachi drives are available for reasonable prices, I’ll give them a try next.

      • Pholostan
      • 5 years ago

      Reading their stuff makes me think they don’t know what they are talking about. Also see:
      [url<]http://www.enterprisestorageforum.com/storage-hardware/selecting-a-disk-drive-how-not-to-do-research-1.html[/url<] [url<]http://www.tweaktown.com/articles/6028/dispelling-backblaze-s-hdd-reliability-myth-the-real-story-covered/index.html[/url<]

        • auxy
        • 5 years ago

        There are a lot of articles like this. Mostly I think Backblaze likes the free publicity they get from posting controversial not-data. (*´﹃`*)

    • ronch
    • 5 years ago

    I am sure Seagate is aware of Blackbkaze’s findings, but what are they doing about it? You can only last so long until people start to realize you make crap and will avoid buying your stuff next time they buy.

    Also, it’s just too bad Hitachi spun off HGST. Been using a 1TB Hitachi since May of 2011 and it’s still going strong, and I’d choose Hitachi-branded drives again if they will still be on the market the next time I buy an HDD. Funny thing is, a hard disc monitoring app I’m using called ActiveSMART says its performance and reliability is low.

      • Deanjo
      • 5 years ago

      You have to take SMART readings with a grain of salt. Powered on hours effect the reliability rating for an otherwise perfectly fine drive. Consumer drives will often read as low reliability by SMART if they are powered on 24/7 in a fairly short time.

        • ronch
        • 5 years ago

        What I mean to say was, it’s ironic how SMART seems to tell ActiveSMART that the drive is not very reliable given how Hitachi drives seem to top the reliability list. I guess the proof is in the pudding.

          • 3SR3010R
          • 5 years ago

          [quote<]it's ironic how SMART seems to tell ActiveSMART that the drive is not very reliable[/quote<] SMART is not telling ActiveSMART about how reliable a drive is it is ActiveSMART interperting SMART data and then coming to the conclusion on reliability. If ActiveSMART is using Power-On-Hours to determine reliability then many of my current drives would be showing reliability concerns when there is none to be found. As others here have stated Relocated-Sector-Count, Pending-Sector-Count are a much better indicators of future drive failures. You can use the FreeWare program CrystalDiskInfo to monitor your drives and it doesn't care about how long a drive is powered on. [url<]http://crystalmark.info/software/CrystalDiskInfo/index-e.html[/url<]

        • nanoflower
        • 5 years ago

        Is that for some programs reading SMART info? I ask because I’m using Acronis Drive Monitor and it doesn’t seem to care about the Power-On Hours. It’s when the error counts on Reallocated Sectors and Read Errors start to rise that it starts to send warnings.

        For instance I’ve got an old Hitatchi 1TB drive that it is giving me warnings about (for good reason.) That drive has 41352 Power On Hours and is listed as ‘OK’. However the Reallocated Sectors Count is listed as 9 which is a ‘failure’. The Reallocation Event Count is listed as 9 which is ‘degradation’. So even though the drive has over 4 years of constant use that isn’t considered an issue but the reallocation events are a real problem.

        (I only have the drive connected now for junk use since you can’t trust it to store and return valid data all the time.)

          • Deanjo
          • 5 years ago

          [quote<]Is that for some programs reading SMART info?[/quote<] It's all about what your programs are reading. If that program reads the powered on hours status for a drive it will show the reliability as low if those hours are high. Personally I just use smartctl for gathering the hard drive smart status (I've have 250 Gig 10 year old maxtors that have tripped their power on reliability hours 9 years ago but they still going strong despite r/w operations that make TR's SSD reliability tests look like a light load).

      • Takeshi7
      • 5 years ago

      It seems Seagate is responding to Backblaze’s finding. Their new 6TB drives have a lower failure rate than the WD 6TB drives. Backblaze is just biased and didn’t include the Seagate 6TB drive on the graph because they only use stats that make Seagate look bad.

        • Deanjo
        • 5 years ago

        [quote<]Their new 6TB drives have a lower failure rate than the WD 6TB drives.[/quote<] That just says they have a lower infancy death rate than usual. Says absolutely nothing about their long term reliability is. Their graph clearly shows that as time goes on the seagates fail more often than the others and that has been the trend for quite a while.

          • Takeshi7
          • 5 years ago

          We shall see.

    • Anovoca
    • 5 years ago

    And yet, one year later, they are still servicing those drives to store client data.

      • emredjan
      • 5 years ago

      They told in their blog. It’s still cheaper to replace failed drives than using more reliable (but expensive) enterprise drives. And as they consume so many hdds, they just buy whatever’s available at that time.

      The client data are stored in multiple RAID arrays. So unless they forget to check the S.M.A.R.T. status, temperatures and multiple drives start failing without replacing, the data should be fairly safe.

        • Deanjo
        • 5 years ago

        [quote<]So unless they forget to check the S.M.A.R.T. status, temperatures and multiple drives start failing without replacing, the data should be fairly safe.[/quote<] One thing to note however, they do not use ECC ram on their pods so corruption still can occur, just that HD failure would not be the cause.

          • emredjan
          • 5 years ago

          Well, they backup all your stuff for 5 dollars a month, cutting corners here and there is probably expected.

            • Deanjo
            • 5 years ago

            Backing up corrupted data doesn’t do much good.

    • SuperSpy
    • 5 years ago

    The backblaze reports have mirrored my experience fairly well: all my WD failures have been either 5+ year old models showing sector errors, or drives that are DOA or dead in under a month. Seagate drives have been much more erratic in their failures, and have happened much more often.

    I don’t have much information on HGST, having just recently started using their 4-6TB 7200 rpm NAS drives, but I haven’t had any early failures yet.

      • w76
      • 5 years ago

      I’ve been extremely lucky apparently with Seagate, I’ve had the total opposite experience. But this is only proof of why big data sets trump personal data sets of 1 every time. Just because I’ve had luck personally and I noticed one of our component saws at work had an 8yr old Seagate still going strong (running Win98SE!) doesn’t mean anything. I definitely won’t be buying any more for a couple years, to give the data time to roll in and prove they’ve resolved their issues.

    • stdRaichu
    • 5 years ago

    [quote<]Do Backblaze's experiences continue to jibe with what TR readers are seeing in their own systems?[/quote<] Pretty much, yes, but as with almost everyone else I don't have a large enough sample size and the last bunch of seagates I bought were the die-happy 1.5TB models about six or so years back. Yoiks I'm old. My ancient 1TB Hitachi drives are still going strong. I'd be all over their drives if they a) came in 5400rpm versions and b) weren't routinely 10-20% more expensive than the WD reds. I'm very interested to see the big discrepancy with the WD 3TB 2013 vs. 2014 reliability figures. I bought about twelve 3TB reds in 2013 and all are still working perfectly (but then I have a 0% failure rate on my four-year old 3TB greens that are still running in my file servers as well), but soon after that the rumour mill said that the newer 3TB and 4TB models were becoming about as trustworthy as a Disney villain on crack and that seems to be borne out by the data. Happy to see the new 6TB lines seem to be a bit more reliable, at least at present, I would have expected higher failure rates . Looks like I'll be waiting for backblaze's next publication before I commit to a file server capacity upgrade...

    • dragosmp
    • 5 years ago

    I just wanted to corroborate with another study published twice a year of the biggest reseller of PC hardware in France. In hardware.fr’s study there’s no particular problem with the 7200.14 3TB. They publish the return rate of various bits, this is for the HDDs, numbers are self-explanatory:
    [url<]http://www.hardware.fr/articles/927-6/disques-durs.html[/url<] *EDIT: Seems like backblaze use them in enterprise environment, while LDLC/hardware.fr sell to home users. Is it ok to say that reliability is highly impacted by the workload and backblaze study isn't really relevant to home users?

      • Deanjo
      • 5 years ago

      The problem with PC hardware’s stats is that unless it is an infancy failure within the return period failures do not count towards those stats. After that return period, failures are processed by the drive manufacturer and don’t count towards their numbers.

      [quote<] Is it ok to say that reliability is highly impacted by the workload and backblaze study isn't really relevant to home users?[/quote<] Considering the data you are referring to with hardware.fr, no, it is not OK to say that.

        • dragosmp
        • 5 years ago

        Good point, they have said all manufacturers except Toshiba cover the RMA directly, while Toshiba does it thru the resellers. It so happens Toshiba had the highest failure rate in hardware.fr’s stats

      • keltor
      • 5 years ago

      Without a full set of data, it’d actually be hard to compare the hardware.fr data. Also the high failure rates are drives that are 7200.14@2.2 years and 7200.11@4.7.

      Backblaze also only uses the drives heavily initially and then doesn’t use them much at all.

      • Pholostan
      • 5 years ago

      The Backblaze data is pretty much worthless. They pack the cheapest consumer desktop drive into self made “storage pods”. 40+ drives a piece, very vibration prone and hot. They claim high temperature does not impact the life of the drives. Ridiculous. Many of the drives have been extracted from external drive enclosures. Etc.

      Backblaze is cool company that do cool stuff, but their sensationalist blog posts and their unreliable data is not part of that.

        • VincentHanna
        • 5 years ago

        Backblaze isn’t playing with house money, You make it sound like they [i<]want[/i<] the drives to fail for sensationalist reasons, but A) the data doesn't show that, the majority of their drives(ignoring the seagate data which is all over the map) have a failure rate of less than 1% and B) they are conducting the study to save themselves money long term.

    • Deanjo
    • 5 years ago

    Of course, as seagate drives get older they become far less reliable than their competition. Once again, real life data proves what I have said for years.

      • adisor19
      • 5 years ago

      Already replaced 1 3TB in my FreeNAS after 2 years and now a second one is showing unreadable sectors. Thanks Seagate.

      Adi

Pin It on Pinterest

Share This