Jump to content

  •  

  • Photo

    Another highly questionable post from BackBlaze


    • Please log in to reply
    47 replies to this topic

    #1 Drashna Jaelre

    Drashna Jaelre

      HSS Legend

    • Moderators
    • 3,065 posts
    • LocationSan Diego

    Posted 21 January 2015 - 04:07 PM

    https://www.backblaz...est-hard-drive/

     

    Take with a grain (or bag) of salt.

     

    What they fail to mention is that that a MAJORITY of the failed Seagate drives in question (ST3000DM001) are shelled external drives. 

     

    And that their stats are PURELY anecdotal until they decide to buy EVERY model of drive on the market and test them EQUALLY. 

     

     

    It's a commonly held belief that external drives are "binned". Eg, that drives that didn't perform as well are used for externals. 

    Agree or not, we CAN agree that external disks use different firmware than internal (retail) drives. This means that it exhibits different behavior. 

     

     

    My biggest issue here isn't that I'm a seagate fanboy, but that they tote these reports as unbiased and fair, when they DEFINITELY are not.

    Additionally, they fail to disclose this statistical bias at all, and present it as unbiased.

    Additionally, they are about the only (or THE only) company that releases stats, so people take their word as law, when it's anecdotal, AT BEST. 

     

    Additionally, there is no way to register or comment on their blog apparently, so there isn't a way to call them out on this directly.


    Christopher Courtney
    Microsoft MVP for Windows Home Server 2009-2012
    Lead Moderator for We Got Served
    Director of Customer Relations for CoveCube (StableBit)


    Windows Server 2012R2 Essentials:
    Supermicro SC847E26-R1K28LP, SuperMicro X10SAT, Intel E3 1245v3 (Haswell) 16GB RAM (ECC), ADATA Premier Pro SP900 240GB SSD OS disk, StableBit DrivePool: 117TB, IBM ServeRAID m1015 cross-flashed to "IR" (RAID) mode

     

    HyperV Server 2012R2:

    SuperMicro SYS-6016T-NTF, SuperMicro X8DTU motherboard, Dual Intel Xeon X5560, 16GB (ECC-Registered), 120GB HDD OS Disk, Crucial MX100 500GB SSD for HyperV storage.

     

    Dedicated Emby Server (Windows 10 Pro):

    Random 1U case, SuperMicro X8DTi motherboard, Dual Intel Xeon X5660, 20GB (ECC-Registered), 64GB ADATA SSD OS Disk, 120GB ADATA SSD for data storage

     
    Sophos UTM: 
    Antec ISK 110 VESA case, ASRock RACK J1900D2Y, 4GB of non-ECC RAM, Samsung 850 Pro 120GB SSD OS drive.

     

    When I say "linux", I'm being lazy and I do mean "*NIX/UNIX based operating systems". That includes Linux, BSD, UNIX, etc. 

     


    #2 ikon

    ikon

      HSS Elite Genius

    • Donating Member
    • 16,090 posts

    Posted 21 January 2015 - 04:41 PM

    OK, I get what you're saying, but I think you're being just a bit overly pessimistic.

     

    I don't think the BackBlaze guys are trying to claim they're providing a comprehensive, market-wide study. I think they're just saying, "Hey, these are the drives we're using, and these are the results we got". It doesn't mean there aren't other makes or models that are better. There well could be; they just haven't used any of them. That said, they really should not title the page as, "The Ultimate Hard Drive Test: What Hard Drive is Best?" because, as you said, they don't cover every make and model, not by a long shot.

     

    As for the recommended drives to buy, what they should have said is, "Of the drives we've tested......."

     

    I am curious how you know the high-failure Seagates are shelled external drives. Is that model only used in external enclosures? It does seem unlikely they would take the time to unshell over 1,100 drives, especially Seagates. I don't know if they're still the same, but I had to unshell 4 Seagate external drives around 8 years ago. It was a royal PITA. Maybe (hopefully) they've changed in the years since I had to do that.


    If at first you don't succeed, do it like your mother told you.


    #3 schoondoggy

    schoondoggy

      HSS Genius

    • Moderators
    • 5,432 posts

    Posted 21 January 2015 - 04:52 PM

    In one of their other posts they stated that the Seagate 3TB were pulls from externals. I still do not believe that drives in external enclosures are less reliable. I do believe that the extra handling from dis-assembling enclosures could very much reduce drive life. The data they provide is interesting, but it can not be used for comparison.
    • Drashna Jaelre likes this

    #4 ikon

    ikon

      HSS Elite Genius

    • Donating Member
    • 16,090 posts

    Posted 21 January 2015 - 05:11 PM

    That's interesting schoondoggy. I guess they figure that counts as 'disclosure'; probably should have been in this post as well, but there it is.

     

    I'm not convinced that shelling a drive would necessarily shorten its life, especially not compared to the travails they experience in normal shipment from overseas to their various markets.

     

    I do think the data can be used for comparison of the drives they actually use. They have been getting similar results for 3 or 4 years now, at least. The preponderance of evidence starts to add up. It may not be 100% fully scientific, but I don't think it's useless either.

     

    EDIT: correction; this is the 2nd year of reporting drive failure rates.


    If at first you don't succeed, do it like your mother told you.


    #5 schoondoggy

    schoondoggy

      HSS Genius

    • Moderators
    • 5,432 posts

    Posted 21 January 2015 - 05:16 PM

    In the late 90's I worked for a company that sold a lot of drives to local PC manufacturers. If a customer had a high fail rate we would investigate. The majority of these checks revealed drive handling issues.

    #6 ikon

    ikon

      HSS Elite Genius

    • Donating Member
    • 16,090 posts

    Posted 21 January 2015 - 06:53 PM

    I find this to be a very interesting discussion. So, I decided to spend a little time researching some of the BackBlaze blog posts.

     

    schoondoggy, you are correct, BB did use some drives that they harvested from external enclosures. However, they are no where near a majority of their drives. Here is a link to the relevant blog from November, 2013: https://www.backblaz...sk-drives-last/

     

    I've extracted a few lines from the blog post, but please do read it for yourselves:

     

    We are up to 75 petabytes of cloud storage now.

     

    By far the majority of these hard drives are “raw” or “internal” hard drives. However, because the Thailand Drive Crisis made it nearly impossible to find internal hard drives for sale at reasonable prices, Backblaze started to farm hard drives. Thus, approximately 6 petabytes of the drives in this analysis were originally “external” hard drives that were “shucked” out of their enclosures.

     

    So, a relatively small percentage (6 petabytes out of 75) of their drives were shucked from external enclosures. I couldn't see anywhere in the post where it says what makes/models of external drives were used. If someone can spot it, please post it. I did read the article that's linked to from the first link I posted: https://www.backblaz...rive_farming-2/. It does show some photos of external drives they bought. From those, it appears, to me at least, that they wound up with a real mix of brands and models. The only consistent thing I could find is that they were all 3TB drives. They had all kinds of people scrounging drives from wherever they could get them at the right price.

     

    Drashna, you will be happy to know that you can indeed post comments about BackBlaze blog posts. Just scroll down to the bottom of the article, just below Brian Beach's bio and you will find a spot where you can post your thoughts.

     


    If at first you don't succeed, do it like your mother told you.


    #7 Drashna Jaelre

    Drashna Jaelre

      HSS Legend

    • Moderators
    • 3,065 posts
    • LocationSan Diego

    Posted 21 January 2015 - 07:33 PM

    OK, I get what you're saying, but I think you're being just a bit overly pessimistic.

     

    I don't think the BackBlaze guys are trying to claim they're providing a comprehensive, market-wide study. I think they're just saying, "Hey, these are the drives we're using, and these are the results we got". It doesn't mean there aren't other makes or models that are better. There well could be; they just haven't used any of them. That said, they really should not title the page as, "The Ultimate Hard Drive Test: What Hard Drive is Best?" because, as you said, they don't cover every make and model, not by a long shot.

     

    As for the recommended drives to buy, what they should have said is, "Of the drives we've tested......."

     

    I am curious how you know the high-failure Seagates are shelled external drives. Is that model only used in external enclosures? It does seem unlikely they would take the time to unshell over 1,100 drives, especially Seagates. I don't know if they're still the same, but I had to unshell 4 Seagate external drives around 8 years ago. It was a royal PITA. Maybe (hopefully) they've changed in the years since I had to do that.

    yes, I am probably being pessimistic, but a LOT of people take these statistics as FACT, and don't bother looking at the actual data.

     

    And I think Schoon is partly right. Handling does make a big difference. However, I suspect that there is more than this at play... I've experienced the same failure rate with these drives (the shelled Seagate Backup Plus drives).

     

    As for how I KNOW these are shelled?

    https://www.backblaz...ng-hard-drives/

    Costco is also where I bought mine. And they're the same model (ST3000DM001).

     

    And in some of their blog posts, they do mention it off hand (way into the post, with no emphasis). 


    Sorry if I'm coming off nasty... really bad week....

     

    And no I dont' have any comment section. Though I am using chrome... so must be a bad theme (they're using wordpress)

    https://drashna.net/...6fa10f22d41561b


    Woops, my bad.

    Sophos was blocking Disqus.... :P


    Christopher Courtney
    Microsoft MVP for Windows Home Server 2009-2012
    Lead Moderator for We Got Served
    Director of Customer Relations for CoveCube (StableBit)


    Windows Server 2012R2 Essentials:
    Supermicro SC847E26-R1K28LP, SuperMicro X10SAT, Intel E3 1245v3 (Haswell) 16GB RAM (ECC), ADATA Premier Pro SP900 240GB SSD OS disk, StableBit DrivePool: 117TB, IBM ServeRAID m1015 cross-flashed to "IR" (RAID) mode

     

    HyperV Server 2012R2:

    SuperMicro SYS-6016T-NTF, SuperMicro X8DTU motherboard, Dual Intel Xeon X5560, 16GB (ECC-Registered), 120GB HDD OS Disk, Crucial MX100 500GB SSD for HyperV storage.

     

    Dedicated Emby Server (Windows 10 Pro):

    Random 1U case, SuperMicro X8DTi motherboard, Dual Intel Xeon X5660, 20GB (ECC-Registered), 64GB ADATA SSD OS Disk, 120GB ADATA SSD for data storage

     
    Sophos UTM: 
    Antec ISK 110 VESA case, ASRock RACK J1900D2Y, 4GB of non-ECC RAM, Samsung 850 Pro 120GB SSD OS drive.

     

    When I say "linux", I'm being lazy and I do mean "*NIX/UNIX based operating systems". That includes Linux, BSD, UNIX, etc. 

     


    #8 ikon

    ikon

      HSS Elite Genius

    • Donating Member
    • 16,090 posts

    Posted 21 January 2015 - 07:36 PM

    OK. I just wondered because I found the same model of drive for sale from Newegg in an internal desktop version: http://www.newegg.co...8-844-_-Product.

     

    I agree handling can make a difference, but I also agree there's more going on. For example, the photos of shelled drives I looked at showed primarily a mix of WD (Passports) and Seagates. If it was the handling, I would expect the WD drives to experience as many failures due to mishandling as the Seagates. And it still doesn't explain why there are so many Seagate failures when the shelled drives are such a relatively small % of the overall total.


    If at first you don't succeed, do it like your mother told you.


    #9 schoondoggy

    schoondoggy

      HSS Genius

    • Moderators
    • 5,432 posts

    Posted 21 January 2015 - 07:44 PM

    Do they talk at all about what the drives are used in? Storage array or server? White box or branded?

    #10 ikon

    ikon

      HSS Elite Genius

    • Donating Member
    • 16,090 posts

    Posted 21 January 2015 - 08:00 PM

    From what I read, pretty much all of them are used in one version of their proprietary Storage Pod or another. I don't recall any reference to white box or other, but they did say they put out RFQ's for drives (maybe 2,000 drives at a shot) and pick the lowest price. Each RFQ specifies the makes and models that are acceptable.

     

    I also read that they have worked to reduce the vibration in the Storage Pods. To me, this is the biggest difference between Data Center HDD usage and home usage. The high density drive cages used in Data Centers tend to have awful levels of vibration. They bolt the drives down firmly to the chassis, thinking this is actually a good thing, when they should really be working to suspend the drives as lightly as possible, to help isolate each drive from the others. That's one of the things I like about Lian Li cases  — they use rubber bushings on the screws to reduce vibration. I still don't think even Lian Li goes far enough, but it's better than most chassis.


    If at first you don't succeed, do it like your mother told you.


    #11 HSS-Dave

    HSS-Dave

      Founder

    • Administrators
    • 2,703 posts
    • LocationIndiana

    Posted 21 January 2015 - 08:37 PM

    You have more fires to fight Drashna.  Lifehacker picked the story up and published it.

     

    http://lifehacker.co...tb-o-1680887763


    HSS Forums - Read Me First Post (Updated)

     

    Listen to RESET Podcast. http://mccabe.io


    #12 ikon

    ikon

      HSS Elite Genius

    • Donating Member
    • 16,090 posts

    Posted 21 January 2015 - 08:41 PM

    I agree with Yev from BackBlaze: make backups.


    If at first you don't succeed, do it like your mother told you.


    #13 Drashna Jaelre

    Drashna Jaelre

      HSS Legend

    • Moderators
    • 3,065 posts
    • LocationSan Diego

    Posted 22 January 2015 - 02:09 AM

    Do they talk at all about what the drives are used in? Storage array or server? White box or branded?

    Nope, unless you confront them about it.

    Apparently only a small UNNAMED percent are shelled externals.

     

    Again, see my point about disclosure....


    I agree with Yev from BackBlaze: make backups.

    Yup. Best post from BackBlaze. :)


    Christopher Courtney
    Microsoft MVP for Windows Home Server 2009-2012
    Lead Moderator for We Got Served
    Director of Customer Relations for CoveCube (StableBit)


    Windows Server 2012R2 Essentials:
    Supermicro SC847E26-R1K28LP, SuperMicro X10SAT, Intel E3 1245v3 (Haswell) 16GB RAM (ECC), ADATA Premier Pro SP900 240GB SSD OS disk, StableBit DrivePool: 117TB, IBM ServeRAID m1015 cross-flashed to "IR" (RAID) mode

     

    HyperV Server 2012R2:

    SuperMicro SYS-6016T-NTF, SuperMicro X8DTU motherboard, Dual Intel Xeon X5560, 16GB (ECC-Registered), 120GB HDD OS Disk, Crucial MX100 500GB SSD for HyperV storage.

     

    Dedicated Emby Server (Windows 10 Pro):

    Random 1U case, SuperMicro X8DTi motherboard, Dual Intel Xeon X5660, 20GB (ECC-Registered), 64GB ADATA SSD OS Disk, 120GB ADATA SSD for data storage

     
    Sophos UTM: 
    Antec ISK 110 VESA case, ASRock RACK J1900D2Y, 4GB of non-ECC RAM, Samsung 850 Pro 120GB SSD OS drive.

     

    When I say "linux", I'm being lazy and I do mean "*NIX/UNIX based operating systems". That includes Linux, BSD, UNIX, etc. 

     


    #14 Drashna Jaelre

    Drashna Jaelre

      HSS Legend

    • Moderators
    • 3,065 posts
    • LocationSan Diego

    Posted 24 January 2015 - 05:16 PM

    Apparently, I'm not the only one that feels the way I do about the backblaze "statistics".

     

    http://youtu.be/e85aRCFH8gM?t=52m23s


    Christopher Courtney
    Microsoft MVP for Windows Home Server 2009-2012
    Lead Moderator for We Got Served
    Director of Customer Relations for CoveCube (StableBit)


    Windows Server 2012R2 Essentials:
    Supermicro SC847E26-R1K28LP, SuperMicro X10SAT, Intel E3 1245v3 (Haswell) 16GB RAM (ECC), ADATA Premier Pro SP900 240GB SSD OS disk, StableBit DrivePool: 117TB, IBM ServeRAID m1015 cross-flashed to "IR" (RAID) mode

     

    HyperV Server 2012R2:

    SuperMicro SYS-6016T-NTF, SuperMicro X8DTU motherboard, Dual Intel Xeon X5560, 16GB (ECC-Registered), 120GB HDD OS Disk, Crucial MX100 500GB SSD for HyperV storage.

     

    Dedicated Emby Server (Windows 10 Pro):

    Random 1U case, SuperMicro X8DTi motherboard, Dual Intel Xeon X5660, 20GB (ECC-Registered), 64GB ADATA SSD OS Disk, 120GB ADATA SSD for data storage

     
    Sophos UTM: 
    Antec ISK 110 VESA case, ASRock RACK J1900D2Y, 4GB of non-ECC RAM, Samsung 850 Pro 120GB SSD OS drive.

     

    When I say "linux", I'm being lazy and I do mean "*NIX/UNIX based operating systems". That includes Linux, BSD, UNIX, etc. 

     


    #15 Drashna Jaelre

    Drashna Jaelre

      HSS Legend

    • Moderators
    • 3,065 posts
    • LocationSan Diego

    Posted 26 January 2015 - 01:33 PM

    Totally not the only one: 
    http://www.pcper.com...-and-statistics

     

    :P


    Christopher Courtney
    Microsoft MVP for Windows Home Server 2009-2012
    Lead Moderator for We Got Served
    Director of Customer Relations for CoveCube (StableBit)


    Windows Server 2012R2 Essentials:
    Supermicro SC847E26-R1K28LP, SuperMicro X10SAT, Intel E3 1245v3 (Haswell) 16GB RAM (ECC), ADATA Premier Pro SP900 240GB SSD OS disk, StableBit DrivePool: 117TB, IBM ServeRAID m1015 cross-flashed to "IR" (RAID) mode

     

    HyperV Server 2012R2:

    SuperMicro SYS-6016T-NTF, SuperMicro X8DTU motherboard, Dual Intel Xeon X5560, 16GB (ECC-Registered), 120GB HDD OS Disk, Crucial MX100 500GB SSD for HyperV storage.

     

    Dedicated Emby Server (Windows 10 Pro):

    Random 1U case, SuperMicro X8DTi motherboard, Dual Intel Xeon X5660, 20GB (ECC-Registered), 64GB ADATA SSD OS Disk, 120GB ADATA SSD for data storage

     
    Sophos UTM: 
    Antec ISK 110 VESA case, ASRock RACK J1900D2Y, 4GB of non-ECC RAM, Samsung 850 Pro 120GB SSD OS drive.

     

    When I say "linux", I'm being lazy and I do mean "*NIX/UNIX based operating systems". That includes Linux, BSD, UNIX, etc. 

     


    #16 ikon

    ikon

      HSS Elite Genius

    • Donating Member
    • 16,090 posts

    Posted 26 January 2015 - 02:32 PM

    Well, Jeremy does say he's glad to have any data, since the big testing organizations won't release it to the public.

     

    As I said earlier, what I think BackBlaze's report can be used for is to compare the drives they use. All their drives are used under the same conditions, so any comparison between them should be valid in that respect. Does it mean that the drive(s) that work the best for them are the best in the world? Not necessarily. As you said, they haven't tested every drive, so that claim cannot be made. Their sample is limited, in variety and size. There can also be issues around lot numbers. Still, taking all those factors into account, their experience over 3 to 4 years is that the HGST drives they've used have proven somewhat more reliable than the Seagate.

     

    More than that though, what I saw was that I personally would avoid 2 particular models of Seagate: the ST31500341AS and ST3000DM001. After that, I would avoid the ST31500541AS, WD30EFRX, and ST33000651AS. Why? Well, I still have tons of choice, so why not? In particular, the HGST drives look good, particularly the 3 and 4 TB, which is somewhat odd since HGST is owned by WD. I can only assume WD is not interfering with HGST's operations.

     

     

     


    If at first you don't succeed, do it like your mother told you.


    #17 Drashna Jaelre

    Drashna Jaelre

      HSS Legend

    • Moderators
    • 3,065 posts
    • LocationSan Diego

    Posted 26 January 2015 - 11:59 PM

    Precisely. It's nice to have some data. However... it's not anywhere good enough to get a real good idea of failure rates, etc.

     

    I will agree that the "ST3000DM001" is a line of drives that should be avoided. You know those drives that failed on me? All of them. That model. ....

    So I think there is an issue with that particular model. and that is good information.

     

    However, it would be nice if Seagate and WD were more transparent about this...


    Christopher Courtney
    Microsoft MVP for Windows Home Server 2009-2012
    Lead Moderator for We Got Served
    Director of Customer Relations for CoveCube (StableBit)


    Windows Server 2012R2 Essentials:
    Supermicro SC847E26-R1K28LP, SuperMicro X10SAT, Intel E3 1245v3 (Haswell) 16GB RAM (ECC), ADATA Premier Pro SP900 240GB SSD OS disk, StableBit DrivePool: 117TB, IBM ServeRAID m1015 cross-flashed to "IR" (RAID) mode

     

    HyperV Server 2012R2:

    SuperMicro SYS-6016T-NTF, SuperMicro X8DTU motherboard, Dual Intel Xeon X5560, 16GB (ECC-Registered), 120GB HDD OS Disk, Crucial MX100 500GB SSD for HyperV storage.

     

    Dedicated Emby Server (Windows 10 Pro):

    Random 1U case, SuperMicro X8DTi motherboard, Dual Intel Xeon X5660, 20GB (ECC-Registered), 64GB ADATA SSD OS Disk, 120GB ADATA SSD for data storage

     
    Sophos UTM: 
    Antec ISK 110 VESA case, ASRock RACK J1900D2Y, 4GB of non-ECC RAM, Samsung 850 Pro 120GB SSD OS drive.

     

    When I say "linux", I'm being lazy and I do mean "*NIX/UNIX based operating systems". That includes Linux, BSD, UNIX, etc. 

     


    #18 ikon

    ikon

      HSS Elite Genius

    • Donating Member
    • 16,090 posts

    Posted 27 January 2015 - 01:04 PM

    However, it would be nice if Seagate and WD were more transparent about this...

     

    Heck, all of the drive makers.


    If at first you don't succeed, do it like your mother told you.


    #19 Drashna Jaelre

    Drashna Jaelre

      HSS Legend

    • Moderators
    • 3,065 posts
    • LocationSan Diego

    Posted 27 January 2015 - 01:26 PM

    Heck, all of the drive makers.

    Didn't I list all of them? :)

     

    not counting SSD of course.


    Christopher Courtney
    Microsoft MVP for Windows Home Server 2009-2012
    Lead Moderator for We Got Served
    Director of Customer Relations for CoveCube (StableBit)


    Windows Server 2012R2 Essentials:
    Supermicro SC847E26-R1K28LP, SuperMicro X10SAT, Intel E3 1245v3 (Haswell) 16GB RAM (ECC), ADATA Premier Pro SP900 240GB SSD OS disk, StableBit DrivePool: 117TB, IBM ServeRAID m1015 cross-flashed to "IR" (RAID) mode

     

    HyperV Server 2012R2:

    SuperMicro SYS-6016T-NTF, SuperMicro X8DTU motherboard, Dual Intel Xeon X5560, 16GB (ECC-Registered), 120GB HDD OS Disk, Crucial MX100 500GB SSD for HyperV storage.

     

    Dedicated Emby Server (Windows 10 Pro):

    Random 1U case, SuperMicro X8DTi motherboard, Dual Intel Xeon X5660, 20GB (ECC-Registered), 64GB ADATA SSD OS Disk, 120GB ADATA SSD for data storage

     
    Sophos UTM: 
    Antec ISK 110 VESA case, ASRock RACK J1900D2Y, 4GB of non-ECC RAM, Samsung 850 Pro 120GB SSD OS drive.

     

    When I say "linux", I'm being lazy and I do mean "*NIX/UNIX based operating systems". That includes Linux, BSD, UNIX, etc. 

     


    #20 ikon

    ikon

      HSS Elite Genius

    • Donating Member
    • 16,090 posts

    Posted 27 January 2015 - 02:03 PM

    Didn't I list all of them? :)

     

    not counting SSD of course.

     

    Toshiba?


    If at first you don't succeed, do it like your mother told you.





    Skins By Invisioneers