New 15,000 rpm disk drives. Why?

New 15,000 rpm disk drives. Why?

Summary: Seagate and HGST have announced 600GB 15k SAS drives. Weren't SSDs supposed to kill fast hard drives? Here's why they haven't.

SHARE:
TOPICS: Storage, Hardware
20

Five years ago pundits predicted that SSDs would kill fast hard drives. And yet, new 15k drives are being announced and shipped.

Why?

Unlike the consumer SSD, the enterprise SSD (eSSD) is very expensive. Figure $2-$3/GB vs a quarter that for consumer SSDs. 

15k SAS drives are closer in price to the consumer SSD. If you're buying hundreds of drives that adds up fast.

But why is the enterprise SSD so costly? Several reasons, perhaps not all of them good.

  • Over-provisioning. Consumer SSDs may have 7 percent more capacity than advertised to support reasonable garbage collection performance and allow for flash die failures. While there's no general spec for eSSDs, a good one will have much more capacity, say 50 percent, to ensure performance and capacity for write-intensive workloads.
  • Dual-port SAS interface. Less common and more expensive, especially in the latest 12Gb/sec flavor.
  • More robust - and expensive - architecture to guard against data-scrambling faults.
  • Better data integrity with a maximum unrecoverable read error rate of 1 per 10E16 vs 1 per 10E14 for consumer SSDs.
  • Profits. Enterprise products carry higher margins because buyers aren't as price-sensitive.

The Storage Bits take
It may be that Seagate and other disk drive vendors are also willing to cut enterprise disk margins to keep them competitive with SSDs. Nor is it clear that the added functionality of eSSDs helps arrays work better or simply helps IT sleep better.

But whatever the case, drive vendors wouldn't be announcing new 15k drives unless some customers were lined up to buy them. Those early predictions of high-end disk drive extinction were premature.

Comments welcome, as always. If your data center is still buying 15k drives, why?

Topics: Storage, Hardware

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.

Talkback

20 comments
Log in or register to join the discussion
  • Couple of other reasons

    1) SSDs are often marketed as more reliable but conventional drives, even the notebook and other portable drives have become so reliable in recent years that the difference is immaterial. In fact, they conventional drives actually outlast cheaper SSDs due to limited erase/write cycles of cheaper (consumer friendly) SSDs.

    2) Another area of contention is performance and yes, the SSDs do perform better in read and write but for most real life uses, the performance difference is not very noticeable. Boot times of operating systems have gotten better ( though SSDs do boot noticeably faster) but app loading times are different only by a few secs at most and most people don't care about the performance gain compared to price/GB difference.

    3) Most OEMs are still installing conventional drives due to cost advantages and most consumers do not replace the OEM peripheral.
    GrabBoyd
    • You hit the nail on the head

      "the conventional drives actually outlast cheaper SSDs due to limited erase/write cycles of cheaper ... SSDs."
      Dr_Zinj
  • 15,000 rpm drives?

    If they are reliable and the price is competitive with the slower drives, consumers may also be interested. 500 Gig SSD are in the $500 range now, and if you go down to a 256 Gig, it will be in the low $200 range. Reliability is key to the attractiveness of these devices.
    rgeiken@...
  • Isn't this typical?

    Don't we usually see the old technology roll out some final tricks it had up its sleeve in an attempt to stave off the newcomer? "Hey, look at us! Our speeds are near what that new SSD kid can do!"
    ejhonda
  • new?

    15k rpm hdd are nothing new. Seagates launched their first 15k rpm drive in 2000.
    Jean-Pierre-
  • All data isn't the same...

    There's a place for both. The virtual SAN architecture takes advantage of this by looking "ahead" and placing the most likely needed data on the (small) eSSD while keeping the large majority of non-read-write intensive stuff on the spinning disks. A further refinement of this is to use gobs of memory instead of or with the eSSD, while keeping most data on disk.
    ClearCreek
  • Somebody said it....

    "Patience is overrated."
    trm1945
  • 15K RPM Drives vs SSD's

    This is a "6 of one, half dozen of another" argument.
    Paul on the Mesa
  • Disaster Recovery

    I don't know nearly enough about Disaster Recovery as it pertains to SSDs. However, it's my guess that "traditional" hard disk drives are much more likely to be "recoverable" in the event of component failure. (due to difficulty and/or cost)

    Maybe this isn't a mitigating factor in deciding on SSD over HDD, but I'd bet in some cases it is. Especially if a person/business has ever been "burned" by a failed SSD. Often times, we humans throw out the baby with the bath water.
    alboulley
  • Fastest drive won't matter much..

    ..if it's going through a network connection. If you're saturating a GigE (250MB/sec) or 10GigE 2.5GB/sec) connection, buying super-fast SSD's will not make much of a difference when you can RAID several 15K SAS HDDs together *AND* have redundancy.

    I'm sure a 8-disk RAID6 array can easily saturate a 10GigE link. Plus, HDDs can be recoverable (minus a major head crash). I'm not too sure how to recover data from a failed SSD.

    Also, 15K SAS HDDs have been around for awhile. Heck, 15K SCSI drives were out before that.
    Simba7
    • Minor nit

      How are you getting 250MB through a 1Gb connection? At most, 125MB and with overhead, drop that to the 100MB range. Add the extra zero for 10Gb.

      Assuming you did use b for bit and B for byte (8 bits to the byte).
      DNSB
  • What about energy use?

    Nobody has mentioned the presumed energy savings from using SSDs. Wouldn't a 15K hard drive use even more energy than a slower hard drive?
    Roger911
    • Energy use and reliability

      A 15K drive would probably use about 4 times the energy of a standard 7200rpm drive with the same weight. And you have to worry about wear and tear in the long run, since the stresses will also go up.
      wilback
      • Plus the heat

        Large numbers of 15k hard drives in the server room will add substantially to the energy bill compared to SSDs.
        mdsock@...
  • Slower drives with better response times - YES it's possible!

    One extremely simple way of improving the performance of a spinning hard drive is to have more than one read/write head assembly. Suppose that we have 2 assemblies on opposite ends of a diameter. This would halve access times because the disc would only have to rotate half a revolution at worst for any sector to be readable. Another major speed improvement would result where data is being copied from one part of a disc to another; in this case, one head assembly could do the reading while the other did the writing. This would eliminate the need for repeated (slow) repositioning of the head assembly from the read position to the write position.
    I am not aware of any disc drive company using this technique and I wonder why not. One obvious reason is cost but since increasing the speed of the spinning dic is costly and people/companies are prepared to pay for better performance. Another possible reason is form factor - having more than one head assembly would make the drive larger, but consider a 2.5 inch disc in a 3.5 inch form factor housing - it should be possible, or maybe a new form factor could be introduced? It's been done before! I think it would be worth doing. What about it Western Digital, Seagate, Hitachi, etc.?
    JohnOfStony
    • Slower drives with better response times - "new" form factor already here!

      The good old 5.25 inch form factor used for CD/DVD/BLU-RAY drives should be big enough for a double-headed "3.5 inch" disk. Mind you, the 3.5" drive has been around so long that technological improvements probably make it practical for a double-header drive to fit the same form factor.

      I, and probably a few thousand other people, had also thought of having 2 head assemblies, however I like the idea of using it for copying - with the right firmware and drivers the data could be copied straight across without leaving the drive, eliminating any IO bottlenecks, freeing up computer resources, and allowing data to be copied.at maximum speed.
      ausvirgo
  • I'll buy an SSD

    when they drop below 40 cents per gigabyte. Maybe.

    A super fast HDD can match the sequential read speed of an SSD, primarily because they're both bottlenecked by the SATA III connection. It just can't hit the random read speed, but for a lot of things that doesn't matter so much.

    Until they pass that bottleneck and SSD's get cheaper, I'll stick with HDD's.
    Jacob VanWagoner
  • ??? 15K drives a big deal ???

    15K rpm drives have been around in the SCSI space for many years. I still have some old IBM numbers that are probably 10 years old. This is nothing new. Why the big deal now?
    wjwood64
    • Not new tech,

      I believe Robin was just pointing out that there are new "enterprise" hard drives being introduced when all the prognosticators told us the spinning platter would be dead by now.
      WozNotWoz
  • HGST on 15K Drives

    Robin, great insights here on 15K drives. To reinforce your point, HGST announced these 15K drives because our customers are using them in complement with SSDs in tiered storage pools and servers dependent on their performance needs. We’re seeing that customers are still looking to balance IOPS performance with the more favorable price/GB metrics provided by 15K drives, as they’ve been telling us that only about 15% of workloads handle “hot data” that would require the raw IOPS needed by an SSD drive.
    sachinpiplani