Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
Data Storage

Storage Vendors Are Quietly Slipping SMR Disks Into Consumer Hard Drives (arstechnica.com) 221

"Storage vendors, including but reportedly not limited to Western Digital, have quietly begun shipping SMR (Shingled Magnetic Recording) disks in place of earlier CMR (Conventional Magnetic Recording) disks..." writes Ars Technica.

"In addition to higher capacities, SMR is associated with much lower random I/O performance than CMR disks offer."

Long-time Slashdot reader castrox shares their detailed report: Shingled Magnetic Recording is a technology that allows vendors to eke out higher storage densities, netting more TB capacity on the same number of platters — or fewer platters, for the same amount of TB. Until recently, the technology has only been seen in very large disks, which were typically clearly marked as "archival"...

Storage vendors appear to be getting much bolder about deploying the new technology into ever-smaller formats, presumably to save a bit on manufacturing costs... [S]everal users have reported that these disks cannot be successfully used in their NAS systems — despite the fact that the name of the actual product is WD Red NAS Hard Drive.

Citing a statement from Western Digital, the article concludes that "The writing on the wall here seems clear. Yes, Western Digital slid SMR drives into traditional, non-enterprise channels — and no, the company doesn't feel bad about it, and you shouldn't expect it to stop...

"Western Digital doesn't appear to be the only hard drive manufacturer doing this," they write, noting that the storage-news web site Blocksandfiles.com "has confirmed quiet, undocumented use of SMR in small retail drives from Seagate and Toshiba as well."
This discussion has been archived. No new comments can be posted.

Storage Vendors Are Quietly Slipping SMR Disks Into Consumer Hard Drives

Comments Filter:
  • by jolyonr ( 560227 ) on Sunday April 19, 2020 @05:09AM (#59964578) Homepage

    Another problem with SMR is that some RAID systems can kill them in a matter of days. Read the fine print of your raid system's supported disks section to see if SMR is supported.

    Drobo systems, for example, don't support SMR.

    So, for example, recently we purchased three new Seagate 8TB Barracuda drives to add to our Drobo system. Within three days two of the drives had completely failed and the third was flashing up warning lights leading to an urgent purchase of non-Seagate drives to replace it.

    Checking with Drobo and SMR drives aren't supported - are these SMR drives? I checked the online support sheet for the drive, no mention of SMR.

    I emailed Seagate and they claim they don't even know!

    "Thank you for contacting Seagate Support. "Typically"the new Barracuda over 8 TB uses SMR, we do not have access to this kind of information, so I can't confirm this."

    But more digging revealed they are SMR.

    https://www.reddit.com/r/DataH... [reddit.com]

    All three drives were returned to the dealer and a full refund received.

    Since then I've been using Toshiba disks without any problems.

    • Most people do not detect silent product substitutions such as sawdust in the sausages or hot dogs as it were, The solution to the above may be https://en.wikipedia.org/wiki/... [wikipedia.org] where you can set expectations precicely, and stress less disk holdings. I predict security calls to the SAM component on Registries may kill disks such as these, and at super-sale times of peak login activity.
    • by jolyonr ( 560227 ) on Sunday April 19, 2020 @05:48AM (#59964632) Homepage

      So, as a follow-up, it appears all major disk vendors are now using SMR on their higher capacity budget drives with various levels of openness about it.

      As far as I know WD are the only ones shipping NAS certified drives with SMR, which is REALLY shady.

    • by Kjella ( 173770 )

      That sounds more like a bad batch to me. That would mean that a little snippet of code that simulates RAID write patterns could kill your disks that fast too. What may be happening is that your work pattern is causing the buffers to fill up so it'll micro-stall and the RAID controller will drop it. Other than that I do see people say write performance will be very bad, including rebuild performance so not good for RAID5/6. If you do want to use them then RAID 0, 1 or 1+0.

    • recently we purchased three new Seagate

      That was the beginning of your mistakes. Don't you people read Backblaze's drive stats?? HGST, definitely, WD maybe, Seagate, fucking never.

    • So explain how continued writes can kill a platter drive? Worst case scenario seems that writes would just be incredibly slow.

  • Comment removed based on user account deletion
    • According to blackblaze, Toshiba is the brand to get right now.
      • According to backblaze (not blackblaze) whatever you buy right now is probably not representative of the reliability figures they have because their numbers seem to point to a new reliability champion every year ... but you need at least a year or so of data to make a relevant assessment.

    • The great thing about anecdotes is that everyone has one, ... about other products. I swore by WD because of multiple Seagate failures. Now I have a mix of WD and HGST (also WD) drives, and in my experience using my massive datatrove of a whole 15 drives I can say that WD and HGST are 100% reliable* whereas I've a 50% failure rate of Seagate drives in 3 years.

      But that's the problem with small datasets. My experience is not realistic and my failures do not reflect the reliability or quality control of any of

  • by burni2 ( 1643061 ) on Sunday April 19, 2020 @06:47AM (#59964694)

    The german wikipedia page for SMR has a section captioned "Verbreitung"(* -> spread of SMR disks in market) where links three links to "computerbase.de" which had done the research are cited.

    https://de.wikipedia.org/wiki/... [wikipedia.org]

    Daneben konnte durch Recherchen von Fachzeitschriften im April 2020 gezeigt werden, dass alle groÃYen Hersteller Festplatten mit SMR vertreibt, ohne dies zu kennzeichnen.

    Translation:

    Besides that research by computer magazines conducted in April 2020 have shown, that all big hard drive manufacturers ship disks with SMR without labeling this.

    There are also the following disks/manufacturers listed where research has shown that these are using SMR.

    Toshiba
                    DT02: 4 TB und 6 TB
                    P300: 4 TB und 6 TB
                    MQ04: 1 TB und 2 TB
                    L200: 1 TB und 2 TB
            Western Digital
                    WD*0EFAX: 2 TB, 3 TB, 4 TB und 6 TB
            Seagate
                    Barracuda ST2000DM008 (2 TB)
                    Barracuda ST4000DM004 (4 TB)
                    Barracuda ST8000DM004 (8 TB)
                    Desktop HDD ST5000DM000 (5 TB)

    • by Vario ( 120611 ) on Sunday April 19, 2020 @08:36AM (#59964926)

      Very frustrating for all users that bough these drives. With a builtin CMR-like cache they can and will fail when you need them most: in RAID usage with lots of written data. Let's hope the list does not grow and that these drives will be replaced by the manufacturer...

      • by AvitarX ( 172628 )
        It seems like if there's a rebuild there's about a 100% this will strike.
        • Unlikely. SMR drives are orders of magnitude slower at writing than reading. The healthy drives in the array won't time out while reading waiting for slower write drives. The likely read delays are more in the order of seek time, rather than the 7seconds or so many "RAID" drives default to reporting read errors to controllers (yes that's different but it points out that if your disk times out in 7 seconds your entire RAID would degrade with every unrecoverable I/O error, and that's just not the case).

      • Very frustrating for all users that bough these drives.

        Well that depends on why you bought them. :-)

  • by Dirk Becher ( 1061828 ) on Sunday April 19, 2020 @07:25AM (#59964774)

    it remembers nothing, its cache is empty and its plug hurts like hell.

  • I recently bought a 10TB external WD drive, but was told in advance that the internal drive was a WD Red drive specifically designed for NAS.

    Is there some kind of test I can run or a product number I can check to see if I have one of these drives?
    • Open it? Or, if you look at it in gparted or some other tool (hwinfo on windows) it should show you the drive serial number which you should be able to look up online (or in manufacturer's database usually accessible via RMA'ing). There's probably a serial number format by device type.

      I'm guessing the label will say what it is inside, though.

    • by Entrope ( 68843 )

      WD says their current 8-14 TB "WD RED" drives are CMR, not SMR. So you personally should be safe.

      The 10 TB hard drive you buy next month may be different. It looks like WD is aggressively trying to hide that parameter from consumers. They are more forthcoming with identifying "host-managed" SMR drives that are aimed at sequential-write workloads in data centers (such as log-structured files and databases).

    • Use a drive information tool (e.g. hdparm) to query the drive capabilities. If the drive supports TRIM, it is SMR.

      You cannot use the drive model number to check whether it is SMR or not - WD use the same model numbers for SMR and CMR models.
  • Western Digital has been shameless lately in their marketing. The 10TB drive I purchased is in reality a 9.1TB drive.

  • by Bengie ( 1121981 ) on Sunday April 19, 2020 @11:07AM (#59965168)
    What are the failure modes of these drives? When I originally read about these, the issue that stood out is they can't do random writes without read-modify-write. If you're writing out data in a forward only streaming fashion, all is well. But if you attempt to write anywhere except the end of the shingle group, the entire group of data has to be read off the drive, modified, and re-written. This could be something like 64MiB of data. In order to make this faster, some companies would re-map the data logically. This made smr drives like an SSD, in that if a modification was in flight during a power loss event, the drive could be left in a corrupted state.

    This was back when SSDs were sensitive to power loss, making SMR even worse.
  • by WaffleMonster ( 969671 ) on Sunday April 19, 2020 @12:07PM (#59965304)

    The only reason left to use spinning platters on less than gigantic drives is certain types of reliability. Lower read error rates, offline persistence, avoidance of power fail during remap lotteries.

    If vendors won't even say what you are buying so that you can't even know what the risks are what's the point of spinning disks at all?

  • Somehwat misleading (Score:5, Informative)

    by TheRealMindChild ( 743925 ) on Sunday April 19, 2020 @12:25PM (#59965350) Homepage Journal
    Every reporting outlet and post about this point at the WD Reds, then follow up with "Seagate and Toshiba doing the same!"... No they aren't. Western Digital is the **only one** to pull this on specifically labeled NAS/RAID drives. And none from Seagate or Toshiba had this "feature" nefariously hidden away like WD. It was right on their spec sheet
  • by parker9 ( 60593 ) on Sunday April 19, 2020 @06:15PM (#59966352) Homepage

    Being a former employee of Western Digital (originally with Hitachi before it was sold to WD), I'm not surprised. Western Digital has been deceitful for a while.

    First, their 'official' response to submarining SMR drives in clients:

    "In a typical small business/home NAS environment, workloads tend to be bursty in nature, leaving sufficient idle time for garbage collection and other maintenance operations."

    and

    "All our WD Red drives are designed to meet or exceed the performance requirements and specifications for common and intended small business/home NAS workloads."

    So, in other words, Western Digital claims they know better than you what you want a HDD to do. Perhaps that's true for some cases, but clearly not all.

    Their real deceit is their next generation HDD. Microwave assisted magnetic recording (MAMR):
    https://hardware.slashdot.org/... [slashdot.org]
    https://www.youtube.com/watch?... [youtube.com]

    If you watch their announcement, you see the only 'proof' that MAMR works is modeling results from a professor at CMU. Turns out this model didn't include all the relevant physics. If you include it, the 'resonant' condition MAMR requires is lost. Their 'prototype' MAMR drive was not MAMR but ePMR.

    ePMR drives an electrical current through the write pole to help sweep out magnetic domains for faster write field switching that improves jitter (allowing for more bits down the track). Driving an electrical current through the write head pole tip should heat the head which may reduce the reliability. They have chosen not to mention this publicly.

    Finally, I will point out their CTO and CEO have both left the company along with their VP of recording heads. I expect their COO will be next.

Love makes the world go 'round, with a little help from intrinsic angular momentum.

Working...