back to article Don't want to fork out for NAND flash? You're not alone. Disk still rules

NAND flash shipments are not catching up with disk, casting doubt on the idea that disks will disappear from data centres, according to the latest research. Stifel MD Aaron Rakers used data from Samsung, Gartner and others to tabulate capacity-based shipment data for consumer and enterprise disk and flash storage devices. …

  1. Naselus

    Not a big surprise

    Chris has pointed this out a few times in the past couple of years. HDD isn't going anywhere, as it's the only thing which is both fast enough for mid-performance storage and can be produced in large enough quantities to keep up with the accelerating rate of data creation.

  2. John 104

    Give it time

    Consumer awareness of SSDs is growing. As more peeps become aware of the performance increase, demand will go up and so will manufacturing.

    1. jabuzz

      Re: Give it time

      Consumer drives account for a small fraction of the shipped used capacity. My educated guess is most capacity is going in the data centre or at the very least consumer level NAS drives, for which SSD makes no economic sense. So unless there is a huge drop in the cost per GB of SSD's disk is not going anywhere. It's just like tape, need to back up 500TB, well doing that to D2D is just not economic even with dedupe.

      So while spinning disks (there has been no spinning iron in disks for decades) may well disappear from end user equipment, and 10k/15k RPM hard drives are on the chopping block, the bulk storage 4TB+ 7200/5400 drives are going nowhere fast. I have a pair of 4TB 7200 RPM drives in a RAID1 and it happily could do six different simultaneous 720p streams from my Plex server. At that point I ran out of playback devices, in these sorts of applications SSD is not replacing HDD any time soon.

      1. Sandpit

        Re: Give it time

        Need to factor in the cost of powering the disks and the cost of cooling them and the cost of the larger racks in a data centre before claiming that SSDs make no economic sense there. Then there is MTBF.

        Look at TCO before deciding how far SSDs are behind HDD

        1. Trevor_Pott Gold badge

          Re: Give it time

          I have. I depth. SSDs still make no sense outside of Tier 0 and some Tier 1 apps. They make no sense at all below Tier 2. Even at Tier 2, hybrid is the way to go.

    2. Paul Shirley

      Re: Give it time

      Even in the consumer space it's hard to see massive scope for growth. You don't need large SSD's to gain most of the speed benefits and given a decent amount of RAM the advantage mostly vanishes once drive caching kicks in in Windows.

      Adding more flash won't make boot any faster, it won't even make much difference on PCs short on RAM. If you've gone cheap on RAM you're probably going to cheap out on the drives as well and buy spinning rust. And if you've gone big on RAM there's even less reason to avoid big, cheap hard drives!

  3. JaitcH
    Meh

    Just ordered a bunch of ASUS UN62V Core-7 units ...

    and even though they are customised for us, ASUS insists they come with NAND drives. Unlike some competitions vertical stack computers there is not even physical provision for regular drives.

    And the added cost of NAND, even at manufacturers quantity discounts, sure would buy very large spinners.

    Let's see if Lenovo comes up with some answers.

  4. PNGuinn
    Go

    Hmmmm...

    In the early days there were all sorts of anecdotal stories about the unreliability of flash. History has shown that some of that was undeniably true. Manufactures haven't helped themselves by being, shall we say, somewhat obscure with their specification and application data. I certainly get the impression that at least some of the manufacturers either didn't fully appreciate or care about data write / rewrite implications of some applications.

    The result was that there was - and still is - deep suspicion over these devices.

    Spinning rust on the other hand is well known, trusted, and more important, the various failure modes well known.

    IT is a somewhat conservative profession, particularly where data security is concerned. (I wonder why?)

    Early devices were also well overpriced and of very low capacity compared with spinning rust. That is changing very quickly.

    There is going to be a tipping point, and I think it's quite near - unless flash hits a major reliability problem as capacity increases. I'm assuming that capacity will continue to rise and relative prices fall.

    Or if some of the "real soon now" technologies really do come on stream.

    Personally I think that the tip will be from the other side. I'm getting increasingly concerned about disk reliability as capacity increases. I've just got a gut feeling that we are quickly reaching a limit with the physics and chemistry of what is after all a mechanical device.

    We live in interesting times.

  5. Gordan
    Stop

    Wait just a minute!

    He has 3 data points, NOT including this year (this being the first year that SSD capacities in 2.5" form factor have not just caught up but exceeded the capacities available as spinning rust) and from that he is trying to project the next 5 data points? I call bullshit.

    1. Anonymous Coward
      Anonymous Coward

      Re: Wait just a minute!

      He's got 3 data points indeed versus your zero.

      I guess he wins. At least for now.

      1. Gordan

        Re: Wait just a minute!

        Basic understanding of statistics and quantitative analysis methods indicates that he may as well have rolled dice to project 5 future data points from 3.

        1. John Tserkezis

          Re: Wait just a minute!

          "Basic understanding of statistics and quantitative analysis methods indicates that he may as well have rolled dice to project 5 future data points from 3"

          Sure you COULD use stats and other analysis, but it's easiest to draw a line in Excel. You can draw some pretty nice graphs you know. And the results are the same in this case anyway.

  6. ntevanza

    Throughput

    Let's talk about throughput. SATA RAID rebuild times are now unacceptable, and flash isn't the answer.

    The obvious answer was once to put more arms and heads in the drive. No-one did that because RAID came along and and you could stripe IO across many heads for cheap.

    That does not work when you're rebuilding a bad disk.

    The Winchester disk guys (and RAID controller guys) can build a dual-arm SATA drive, doubling throughput, that's competitive with flash on unit price. Discuss.

    1. Gordan

      Re: Throughput

      Flash is a _part_ of the answer. The rest of the answer is moving to smaller form factor drives (use 2.5" disks instead of 3.5" ones), and the biggest part of the answer is to switch to a post-RAID technology like ZFS, with appropriately sized vdevs (i.e. don't have more than, say, a 6-disk RAID6 (RAIDZ2), or 11 disk RAIDZ3 (n+3)). if you need a bigger pool, have multiple such vdevs (equivalent of 12 disk RAID60).

      1. ntevanza

        Re: Throughput

        Have a vote, since RAID is part of the problem, and small drives are nicer.

        So why isn't ZFS everywhere?

        RIP Sun.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Biting the hand that feeds IT © 1998–2019