Western Digital offers datacenters 12 TB of Gold

Western Digital (WD) has a wide range of hard drive variations, sometimes a little tricky to keep track of. Regular consumer drivers come are labeled and colored Blue and Black, NAS drives are the Reds, surveillance storage is colored Purple, and the datacenter, endurance-oriented HDDs come in a Gold outfit. The Gold family is now getting a new member with a whopping 12 TB capacity.

This drive is the first with a 12 TB capacity in WD's entire lineup, and has a 7200 RPM rotation speed and a big 256 MB chunk of cache. According to WD, transfer speeds to and from the drive can hit 255 MB/s, likely thanks to its massive areal density. Since the HDD is part of the Gold family, reliability and endurance are two key characteristics. WD says the drive is capable of dealing with around 550 TB of I/O per year, a figure it claims is "among the highest of any 3.5-inch hard drive." The 12 TB Gold HDD's mean time between failures (MTBF) ought to be around 2.5 million hours, too.

The drive's heavy-duty characteristics don't stop there. As befits an enterprise hard drive, the WD 12 TB Gold has vibration protection, time-limited error recovery (TLER) for RAID setups, and a dual-stage head actuator. Interested system administators can check out the drive's detailed datasheet.

The Western Digital online store already has the drive at $522. Those who prefer to shop from Newegg can obtain it for a little more, at $540. WD offers a five-year warranty coverage across its entire Gold HDD lineup.

Comments closed
    • ludi
    • 3 years ago

    Uh-oh, nobody tell AMDisDEC about this.

    [/inside_joke]

    • ronch
    • 3 years ago

    Gold? How ordinary. I want a tesseract.

      • just brew it!
      • 3 years ago

      They should’ve at least gone for Platinum. OTOH, I guess Gold leaves them some room to introduce a higher product tier, without resorting to Unobtainium.

      Maybe this is a start of a broader rebranding effort? Red Pro could become Silver, and Red could become Bronze? Enterprise/NAS drives = metals, consumer drives = colors?

        • Brainsan
        • 3 years ago

        Double rainbow! Full double rainbow! What does it mean?

    • SlappedSilly
    • 3 years ago

    The WD link is teh borken. [url<]http://https//...[/url<] Newegg: "In stock. Limit 1 per customer." Uhhh... just one?

      • travbrad
      • 3 years ago

      1 copy of your 12TB of data. What could go wrong?

    • jts888
    • 3 years ago

    What tricks does this need to pull off this capacity?
    Helium is fine with me for now (no idea if leaking actually limits lifespans), but shingled recording in HDDs is probably hated by anybody not using them in place of archival tape cartridges.

      • Krogoth
      • 3 years ago

      The larger SKUs use helium in order fit more than four platters in the chassis.

      Helium will eventually leak out in span of years which cause the headers to crash. This is the worst way for a HDD to die on you. Modern HDDs usually die from spindle motor or actuator seizing.

        • jts888
        • 3 years ago

        If a drive’s helium bled out, wouldn’t air that bled back in keep the heads from crashing? I’d have guessed the failure mode would be more along the lines of too many platters spinning in more viscous gas just causing elevated heating that would kill thinks more gradually.

          • magila
          • 3 years ago

          One of the reasons they use Helium is because it produces less turbulence which allows the use of tightly packed, ultra-low-profile heads. If the drive were filled with air the heads would become unstable and potentially crash into the platters.

            • GTVic
            • 3 years ago

            No way, the cushion of air/helium is what stops the heads from crashing into the platter in the first place. Also, the height savings from an air gap of a few nano-meters to a smaller helium gap is not even measurable relative to overall drive height or the internal distance between platters.

            The helium allows the disks to be thinner and the gap between disks to be reduced because of reduced drag. So if air got in then the whole thing would start to fail but not due to head crashing in particular.

            To suggest that the helium will leak and air will get in and make the drives less reliable is speculation since I’m pretty sure no one has shown a drive where the helium has leaked out and the whole reason for the switch to helium in the first place is to increase reliability.

          • just brew it!
          • 3 years ago

          In addition to what [s<]maglia[/s<] magila said, no, air does not necessarily get back in. Helium atoms are very small, and much better than oxygen, nitrogen, etc. at slipping past (or diffusing through) seals. You also can't view it as a simple equalization of total pressures; it does not work that way when you have different mixtures of gasses on either side of the barrier (look up "partial pressure" sometime).

            • just brew it!
            • 3 years ago

            BTW, remaining helium level is apparently a SMART attribute on at least some helium drives.

            I personally would not trust them for long-term data archiving. They are being marketed to enterprise customers, who are typically going to replace their servers (along with the drives) on a ~5 year cycle. I would not be surprised if the helium tends to fall to unacceptable levels not too long after the warranty (typically 5 years for enterprise drives) runs out.

            • Brainsan
            • 3 years ago

            Yep, helium is extremely good at leaking. The only thing that’s better is hydrogen. But hydrogen does have certain drawbacks for doing things like leak checking vacuum systems.

      • ronch
      • 3 years ago

      Gold platters. It’s got gold platters.

    • Duct Tape Dude
    • 3 years ago

    Stupid question: Why are HDD caches still in MB in 2017? Can’t they just stick a gig or two of DDR3 in there, if only for marketing purposes?

      • just brew it!
      • 3 years ago

      Problem with large internal caches is that you vastly increase the potential for data loss if power gets cut unexpectedly during writes. Best case it’ll take you several seconds to flush all the data, and that’s if it is all sequential. If the data to be written is scattered around the disk it will take even longer due to seek and rotational latency.

      Denser caches also increase your susceptibility to random bit flips unless it is ECC protected.

        • DragonDaddyBear
        • 3 years ago

        OK, stupid question 2: Why not use non-volatile RAM? I assume it’s too expensive or slow for the benefit but maybe it’s more than that?

          • UberGerbil
          • 3 years ago

          What makes you think it would have much of any benefit? We hit diminishing returns on HD cache long before SSDs were even a thing. Doubling the cache would likely yield single-digit percentage improvement on benchmarks, maybe less, and would be unnoticeable in real-world usage. Besides, for anything on the HD that is getting hit often enough for cache to matter, it is probably [i<]already[/i<] being cached in system RAM by the OS. And if you're so low on system RAM that the OS file cache is getting squeezed, you have bigger problems. At this point HDs are warm (or semi-cold) storage; if performance matters, you're using alternative tech. But if larger caches really matter to you, Seagate and WD still sell Hybrid "SSHD"s...

          • Ryu Connor
          • 3 years ago

          Seagate does do this with their SSHD line. They call it Multi-Tier Cache or MTC.

          RAM to NAND to Platter and vice versa

          Their original SSHD models were designed for read only. The new FireCuda line finally does read and write cache in the NAND.

          [url<]http://www.seagate.com/internal-hard-drives/hdd/firecuda/[/url<] [url<]http://www.seagate.com/files/www-content/product-content/barracuda-fam/barracuda-new/files/multi-tier-caching-technology-white-paper-2017.pdf[/url<] The problem here is parallelism. You have a single NAND chip, not multiple chips interleaving like in a real SSD. It's faster yes, but nowhere near as fast as many enthusiasts would hope.

          • just brew it!
          • 3 years ago

          Large NVRAMs are expensive because they use SRAM cells to reduce power consumption (and SRAM cells take up a lot more die space than DRAM cells). Typical NVRAMs also require a small battery (like the one your motherboard uses to power the NVRAM that holds your BIOS settings).

          As UberGerbil notes, you can already get SSHD drives that have mechanical platters and use NAND flash as cache — maybe this is what you’re looking for?

          Edit: But and as Ryu notes, NAND flash cache is slower than RAM cache…

        • Waco
        • 3 years ago

        HDDs, at least recently, dedicate a section of the disk to dump the cache in the case of an unsafe power down. The motor backfeed is enough to keep the drive active long enough to dump the contents of cache for recovery on next power up.

      • Ryu Connor
      • 3 years ago

      An educated guess is cost.

      The logic board on an HDD would have increasing complexity with more memory. Traditionally SSDs and HDD limit their RAM cache to a single memory chip.

      More memory means more PCB space.
      Additional memory will require more traces.
      The above two items together increase the need for an expensive well built multi-layer PCB.
      Higher chip counts will increase the odds for cache memory errors (bit flips) and defective memory chips.
      Power loss woes. Nobody wants integrated batteries.
      As memory chip counts increase so too does the product, support, and QA costs.

      To create an 8GB RAM cache that matches the size of the NAND on Seagate SSHDs, you’d need to place 32 256MB chips or 16 512MB chips. That’s a great deal of PCB space and cost in a market, that is generally speaking, razer thin on margins.

      The more clever solution would be to integrate the memory into the ARM CPU on the PCB. That isn’t free either and doesn’t even fix all the issues listed above.

      If you really want a big memory cache, there are some SAS/SATA controllers with RAM.

      [url<]https://www.newegg.com/Product/Product.aspx?Item=N82E16816117224&ignorebbr=1[/url<] You'll want one of these as well. [url<]https://www.cdw.com/shop/products/Intel-RAID-Smart-Battery-memory-backup-battery-Li-Ion-1500-mAh/2872051.aspx[/url<]

      • ew
      • 3 years ago

      It isn’t really a cache. It is a buffer. Its job is to let the drive read/write from the disk and send/receive data from the host at different times. So it only needs to be big enough to hold data between these operations.

      Edit: The kind of caching you are thinking of is typically done by the host computer in system memory.

Pin It on Pinterest

Share This