Stories
Slash Boxes
Comments

SoylentNews is people

posted by martyb on Sunday May 10 2015, @10:55AM   Printer-friendly [Skip to comment(s)]
from the my-first-hard-disk-had-40-MB dept.

Japanese manufacturer Fixstars is releasing a 6 terabyte 2.5" solid state drive in July. The drive uses 15nm MLC NAND. 1 TB and 3 TB models are also available, but only the pricing for the 1 TB model is known: $820. The drive is not particularly fast; it uses the 6 Gbps SATA 3 interface to achieve 540 and 520 MB/s sustained read and write speeds.

For comparison, the highest capacity 2.5" hard disk drive is currently Toshiba's 3 terabyte MQ03ABB300, which uses four 750 GB platters. The Fixstars SSD is 9.5 mm thick, while the Toshiba HDD is 15 mm thick.

It's about time to bring the HAMR down.

Related Stories

World's Largest Capacity Storage Drive Announced: It's a 15.36 Terabyte SSD 47 comments

We have previously run stories about 2 TB, 4 TB, and 6 TB Solid State Drives (SSDs) and their seemingly inevitable but gradual increase in capacity over time. Samsung just announced a HUGE increase in drive capacity, leap-frogging all other storage devices out there — including spinning hard disk storage [takyon: a 6 TB 2.5" drive already leapfrogs spinning disk]!

Ars Technica is reporting that Samsung unveils 2.5-inch 16TB SSD: The world's largest hard drive. The third-generation 3D V-NAND is now up to 48 TLC layers and 256Gbit per die. From the article:

At the Flash Memory Summit in California, Samsung has unveiled what appears to be the world's largest hard drive—and somewhat surprisingly, it uses NAND flash chips rather than spinning platters. The rather boringly named PM1633a, which is being targeted at the enterprise market, manages to cram almost 16 terabytes into a 2.5-inch SSD package. By comparison, the largest conventional hard drives made by Seagate and Western Digital currently max out at 8 or 10TB.

The secret sauce behind Samsung's 16TB SSD is the company's new 256Gbit (32GB) NAND flash die; twice the capacity of 128Gbit NAND dies that were commercialised by various chip makers last year. To reach such an astonishing density, Samsung has managed to cram 48 layers of 3-bits-per-cell (TLC) 3D V-NAND into a single die. This is up from 24 layers in 2013, and then 36 layers in 2014.

Though claimed capacity is 16 TB, actual available storage is 15.36 TB (providing 640 GB of over provisioning.) The drive is 15mm high so it is geared to the enterprise market; it probably won't fit in your laptop where 9.5mm is an unofficial standard.

In case you were wondering, by some estimates this capacity is enough to store 1.5 copies of the uncompressed textual data in the print collection of the US Library of Congress (LoC).

It boggles my mind to consider such large storage capacities. Given the global population is about 8.3 billion, just one of these drives would be sufficient to store 1.8 KiB on every human being on the planet, never mind an entire rack of these drives.

What practical use is there for such capacities? What would you do with one (or more) of these? How would this fit into your "Big Data" application?


Original Submission

This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 2) by MichaelDavidCrawford on Sunday May 10 2015, @11:03AM

    ... that I had landed a job that left me with enough cash, that I was able to purchase a full-height, 5 1/4" inch Fujitsu SCSI drive, used, off of the USENET news.

    135 megabytes.

    Seven Hundred Dollars.

    Kids.
    These.
    Days.

    HOWEVER!

    That drive still works, and to this very day is in regular use. It belongs to a friend now; he uses it to play ancient Mac OS games.

    --
    Yes I Have No Bananas. [gofundme.com]
    • (Score: 2) by kaszz on Sunday May 10 2015, @11:35AM

      by kaszz (4211) on Sunday May 10 2015, @11:35AM (#181057) Journal

      And the ATA interface were an electrical and a protocol design horror.. ;)

      • (Score: 3, Interesting) by MichaelDavidCrawford on Sunday May 10 2015, @11:41AM

        the cable connector was unkeyed. The only way to tell that you plugged it in upside down, was that you couldn't read the drive. I actually did that once.

        The classic mac os required that one write a "scsi driver" to the initial sector of each drive. It was a little bit of 68k machine code that, in my understanding, adjusted for the timing characteristics of that individual drive.

        --
        Yes I Have No Bananas. [gofundme.com]
        • (Score: 2) by Dunbal on Sunday May 10 2015, @12:17PM

          by Dunbal (3515) on Sunday May 10 2015, @12:17PM (#181071)

          The cable had a red border down one side and if you looked carefully (and didn't have a drive made by a crappy manufacturer), there was usually a tiny notch on one side of the connector on the drive or a white dot painted onto the board where the connector was soldered on- that's where you put the "red" side of the ribbon cable... Sheesh you guys, it's a wonder we made it to this era of enlightenment at all if you never even noticed these things :)

    • (Score: 0) by Anonymous Coward on Monday May 11 2015, @01:16PM

      by Anonymous Coward on Monday May 11 2015, @01:16PM (#181453)

      5 1/4" inch

      5 1/4 square inch?

  • (Score: 0) by Anonymous Coward on Sunday May 10 2015, @11:15AM

    by Anonymous Coward on Sunday May 10 2015, @11:15AM (#181053)

    Cause I just finished my bewulf cluster of 1541 C64 Drives.
    Sadly, the project will have to be abandoned cause the power requirements just to reach 1 Tera is outweighing the cost to by the upcoming 6 Tera.

    Thanks to all who helped me in trying to attain my dream.

  • (Score: 1) by nitehawk214 on Sunday May 10 2015, @11:33AM

    by nitehawk214 (1304) on Sunday May 10 2015, @11:33AM (#181056)

    Doesn't HAMR have moving parts and a spinning platter? That would be a huge leap backwards. I am entirely done with drives that have moving parts. Because margins have become razor thin on the old technology hard drives, quality seems to be slipping badly. (I'm looking at you, Western Digital.)

    --
    "Don't you ever miss the days when you used to be nostalgic?" -Loiosh
    • (Score: 2) by MichaelDavidCrawford on Sunday May 10 2015, @11:38AM

      attaching flash to SATA doesn't make a whole lot of sense to me.

      Why not plug it into the memory module sockets?

      --
      Yes I Have No Bananas. [gofundme.com]
      • (Score: 2) by kaszz on Sunday May 10 2015, @11:54AM

        by kaszz (4211) on Sunday May 10 2015, @11:54AM (#181065) Journal

        Because memory sockets vary way more than S-ATA. And if you pick up random computer. Chances you find a S-ATA connection is way higher. And if you don't, there's a lot of adapters.

        • (Score: 3, Interesting) by MichaelDavidCrawford on Sunday May 10 2015, @12:05PM

          Granted that just about every computer has SATA. But I would expect that by now, the storage industry would have come up with a connectivity specification that was optimized for flash.

          Consider that rotating media can only read the bits that are passing under the head. If you want to read a whole sector, you have to wait for the disk to rotate a little bit.

          Modern memory controllers can read and write data in much wider chunks then the CPU registers, for example my Core Quad Xeon e5420 access four FB-DIMMs in parallel, to load a cache line all in one go.

          There should be a connection standard that works like that for flash.

          --
          Yes I Have No Bananas. [gofundme.com]
          • (Score: 2) by kaszz on Sunday May 10 2015, @12:10PM

            by kaszz (4211) on Sunday May 10 2015, @12:10PM (#181068) Journal

            Industry and coming up with a smart solution. No that doesn't happen particularly fast. So until that happens, expect to wait and suffer.

            (USB is an industry creation.. and it really suck when it comes to smart design)

          • (Score: 1) by nitehawk214 on Sunday May 10 2015, @06:23PM

            by nitehawk214 (1304) on Sunday May 10 2015, @06:23PM (#181125)

            There were PCI-e based SSD drives, but from what I heard these were not terribly reliable. But really I don't think SATA is the problem here. I believe the current SATA spec is still faster than the fastest SSD drives.

            USB certainly isn't the solution. It is no where near reliable enough for permanent storage on a computer.

            RAM sockets are not really it either. The way the computer interacts with RAM is much more different than permanent storage.

            --
            "Don't you ever miss the days when you used to be nostalgic?" -Loiosh
          • (Score: 3, Informative) by takyon on Sunday May 10 2015, @07:36PM

            by takyon (881) <{takyon} {at} {soylentnews.org}> on Sunday May 10 2015, @07:36PM (#181143) Journal

            It's called NVMe. It can allow SSDs to read/write more than 2 GB/s.

            http://www.anandtech.com/show/9090/intel-ssd-750-pcie-ssd-review-nvme-for-the-client [anandtech.com]

            There's a reason I wrote that this 6 TB drive is "not particularly fast", and it's called NVMe.

            --
            [SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
            • (Score: 2) by kaszz on Sunday May 10 2015, @11:19PM

              by kaszz (4211) on Sunday May 10 2015, @11:19PM (#181233) Journal

              Yeah, PCI-e is probably fast enough to make a serious difference. And it's a standard to. Otoh.. perhaps not much faster than S-ATA. BUT it can be utilized with many parallel channels.

              • (Score: 0) by Anonymous Coward on Monday May 11 2015, @10:16AM

                by Anonymous Coward on Monday May 11 2015, @10:16AM (#181423)

                Yeah, PCI-e is probably fast enough to make a serious difference. And it's a standard to. Otoh.. perhaps not much faster than S-ATA. BUT it can be utilized with many parallel channels.

                The fastest SATA is 6 Gbps. PCIe3 is 8 Gbps per lane. An x16 slot is 128 Gbps. PCIe4 will be twice as fast as PCIe3.

      • (Score: 1) by tftp on Sunday May 10 2015, @07:22PM

        by tftp (806) on Sunday May 10 2015, @07:22PM (#181139) Homepage

        attaching flash to SATA doesn't make a whole lot of sense to me. Why not plug it into the memory module sockets?

        The DDR sockets are optimized for super-fast, synchronous, random access to DRAM cells, with minimum unit of storage being one byte. SATA is accessing an asynchronous block device - one that consists of a bunch of storage sectors that are read and written as a whole, and the access time is not guaranteed. The DRAM controller handles refresh, which Flash has no need for. The ATA controller implements DMA from/to the RAM, and many can do RAID. Those are quite different technologies.

        • (Score: 1) by EETech1 on Monday May 11 2015, @03:22AM

          by EETech1 (957) on Monday May 11 2015, @03:22AM (#181326)

          www.zdnet.com/article/flash-dimms-hit-server-market/
          I believe IBM has them, if you can afford it!

    • (Score: 2) by kaszz on Sunday May 10 2015, @11:39AM

      by kaszz (4211) on Sunday May 10 2015, @11:39AM (#181061) Journal

      Which manufacturer is the king of quality right now?

      • (Score: 2) by Dunbal on Sunday May 10 2015, @12:21PM

        by Dunbal (3515) on Sunday May 10 2015, @12:21PM (#181072)

        I hear Samsung is getting good reviews and I myself have Crucial (Micron Technology Inc) 500GB-980GB SSD's that haven't failed in 4 years' of home use. But all I can offer is personal anecdote.

        • (Score: 3, Informative) by tynin on Sunday May 10 2015, @04:35PM

          by tynin (2013) on Sunday May 10 2015, @04:35PM (#181106) Journal

          We have several thousand 1TB Crucial M550 drives at work (and are getting ready to buy ~8k more in the near term, we've been very happy with them). Without having any numbers to work with, the failure rate I've noticed has been extremely low (lower than the HDDs). Before that we experimented with OCZ, but they were garbage and were failing out so often we couldn't keep the environment stable.

          For home use, I picked one up a 1TB M550 and I've continued to be pleased with it (used camelcamelcamel.com to watch the price, when it fell to $300 I picked it up). Prior to that for home use I was only buying Intel SSDs (X-25M, and two 520's), of which I've had zero failures.

        • (Score: 2) by kaszz on Sunday May 10 2015, @11:25PM

          by kaszz (4211) on Sunday May 10 2015, @11:25PM (#181236) Journal

          I really referred to mechanical disks ;)

      • (Score: 1) by nitehawk214 on Sunday May 10 2015, @06:45PM

        by nitehawk214 (1304) on Sunday May 10 2015, @06:45PM (#181132)

        For spinning disks? I don't know anymore. Certainly not Western Digital, though. I have had 3 out of 4 2T drives fail in a year. and two of those were DOA.

        But I have 2 SanDisk SSDs for one for over a year and the other for a few months. Both have been operating wonderfully. My work has several more Sandisk and Samsung drives in our database servers, and I have never actually seen a failed SSD yet. The ones at work get hammered pretty hard and have been for a couple of years, so I can say those work.

        --
        "Don't you ever miss the days when you used to be nostalgic?" -Loiosh
        • (Score: 2) by kaszz on Sunday May 10 2015, @11:37PM

          by kaszz (4211) on Sunday May 10 2015, @11:37PM (#181242) Journal

          Yeah, spinning. are you sure it's WD thats bad and not just the cutting edge harddisks?

          And I read that SSD doesn't degrade gracefully. It's a complete mess to recover if even possible in any meaningful way.

          • (Score: 1) by nitehawk214 on Tuesday May 12 2015, @03:12AM

            by nitehawk214 (1304) on Tuesday May 12 2015, @03:12AM (#181771)

            My backup strategies do not cover recovering a disk. I always have mirrors of everything I care about, and if a disk fails it gets pulled and replaced. (Though with my current WD Raid0 failure, I am using the ghetto rsync solution to a different drive.) If the mirror itself fails, that is what offline backup is for.

            I had been using WD drives for quite a long time now, but it has only been this recent batch that have had failures. My guess is that the rise of SSDs and consolidation has really put a pinch on profits, and WD has let quality slip. Things with moving parts are bound to fail eventually, but this has been quite absurd.

            --
            "Don't you ever miss the days when you used to be nostalgic?" -Loiosh
    • (Score: 2) by takyon on Sunday May 10 2015, @04:21PM

      by takyon (881) <{takyon} {at} {soylentnews.org}> on Sunday May 10 2015, @04:21PM (#181105) Journal

      As long as Google, Facebook, Amazon, etc. continue to use HDDs, you still are, and there's no better consumer option for offline bulk storage. No SSDs, tapes, BD-XLs.

      --
      [SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
  • (Score: -1, Troll) by Anonymous Coward on Sunday May 10 2015, @12:16PM

    by Anonymous Coward on Sunday May 10 2015, @12:16PM (#181070)

    Hey Ethanol-fueled, just imagine how many of your Japanese comics and animations you could store on a 6 TB SSD hard drive, and how quickly you could access them. You know which comics and animations I'm talking about: the ones with the tentacles.

    • (Score: -1, Offtopic) by Anonymous Coward on Sunday May 10 2015, @08:53PM

      by Anonymous Coward on Sunday May 10 2015, @08:53PM (#181181)

      Vanilla scum.

  • (Score: 4, Funny) by Techwolf on Sunday May 10 2015, @02:21PM

    by Techwolf (87) on Sunday May 10 2015, @02:21PM (#181085)

    SSD is nice for loosing 6 TB of data in a flash. ;-)

    • (Score: 3, Funny) by rts008 on Sunday May 10 2015, @02:58PM

      by rts008 (3001) on Sunday May 10 2015, @02:58PM (#181088)

      You are not using enough duct tape if it is getting loose and escaping. With enough duct tape, it will stay where you put it, nice and tight. ;-)

    • (Score: 3, Interesting) by takyon on Sunday May 10 2015, @04:18PM

      by takyon (881) <{takyon} {at} {soylentnews.org}> on Sunday May 10 2015, @04:18PM (#181104) Journal

      The controllers have gotten better over the years. I expect NAND reliability to outpace HDD reliability, but I haven't seen any definitive proof of that.

      V-NAND will make a huge difference. Bigger nodes (30-40nm) in the first V-NAND drives, lots of layers to increase over-provisioning. It buys manufacturers more time to come up with a better cell [phys.org], otherwise they will get replaced by memristors or something else.

      --
      [SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
      • (Score: 1, Informative) by Anonymous Coward on Sunday May 10 2015, @04:38PM

        by Anonymous Coward on Sunday May 10 2015, @04:38PM (#181108)

        http://techreport.com/review/27909/the-ssd-endurance-experiment-theyre-all-dead [techreport.com]

        This was pretty good. But no comparison to the life of a HD.

        I would be a bit leery to put them in my production env (we do 100-200 GB per day and plans for 500 by end of year). But that is not my decision.

        However, we have a robust backup system in place.

        For my home computers though. As much SSD as I can lay my hands on and whole disk system backups to external spinny drives. My laptop now lasts well over 6 hours on battery with little usage. 2-3 with my normal usage. That is a 3 year old battery to boot. The speed increase is pretty nice too.