High-end disks dead in 2 years

High-end disks dead in 2 years

Summary: Dave Donatelli, a senior exec at EMC - the largest data storage vendor - predicted that high-end flash drives will replace high-end hard drives in 2 years. Is this the beginning of the end for disk drives?


Dave Donatelli, a senior exec at EMC - the largest data storage vendor - predicted that high-end flash drives will replace high-end hard drives in 2 years. Is this the beginning of the end for disk drives?

What is a high-end hard drive? Unlike the SATA or PATA drives in your desktop or notebook system, high-end disks share some important features:

  • They're faster: 10k or 15k RPM and up to 3x the IOPS
  • They're more reliable: 10x better read specs; 3-5x MTBF
  • They use high-end interfaces like Fibre Channel (FC) or Serial Attached SCSI (SAS)
  • They're much lower capacity: 74 GB and 146 GB capacities are most common; there are no 1 TB high-end disks.
  • They're way more expensive: 10x per gigabyte

Vendors make huge volumes of SATA drives. But they make huge profits on their high-end drives. Most of them go into high-end arrays from EMC, IBM, Hitachi and HP whose prices start at $100,000 and soar into the millions.

The loss of this high-end, high-margin business has the drive vendor's attention. Seagate's suit against STEC, a maker of high-end flash drives, is part of an effort to claim this business for themselves.

Flash performance EMC has measured STEC’s flash drives at 30x the IOPS of a high-end disk. And they give sub-millisecond access times. That alone would justify a premium over existing drives.

Flash cost A 15k 74 GB Seagate SAS drive is about $175 or roughly $2/GB. A 2 GB Single Level Cell (SLC) flash chip is currently about $8/GB on the flash spot market. If flash keeps dropping at 50% a year they’ll be where the current disk price is in mid-2010.

Flash reliability High-end flash drives use a more costly and more reliable flash called Single Level Cell or SLC. These are spec'd at 1,000,000 read/write cycles. Write-leveling means that customers have no worries, even with database index files. Cheaper MLC flash, used in notebook drives, is now spec'd at 10k and will soon be even lower.

But that’s raw chip vs finished disk The remaining question is how much does the chip controller and other infrastructure. STEC isn’t selling its 74 GB flash drives for $8/GB - $80/GB is closer to the mark. Volume will cover their engineering costs. PC boards are cheap. Then it's the cost of the flash chips.

Flash drives have other advantages: lower power; less cooling; higher MTBF and lighter weight. These are important when you have a couple of thousand drives in an array.

The Storage Bits take Flash drives don’t need absolute price parity to win against high-end FC drives. Getting within 30% should do it for most people. Their performance advantages are worth at least that.

Of course the drive vendors aren’t going to sit still. Many have claimed disks are dead and they’re all gone. But this looks serious. It won't be easy to innovate against flash drives.

For us SATA drive users the death of the high-end disk, when it comes to pass, will mean that SATA drives will be more expensive as they shoulder the R&D costs that the high-end currently absorbs. Disk research will also be cut back when array vendors are no longer experiencing complex disk problems.

One thing it doesn't mean: flash notebook drives won't become any more attractive. Notebook disks are running about $0.40/GB and they'll be a quarter that by end of 2010. Not many people will pay 10x per GB for very little benefit.

Comments welcome.

Topics: Web development, EMC, Hardware

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


Log in or register to join the discussion
  • High end Flash

    "High end" Flash has been spec'ed at 1M write cycles for years now. And even with wear leveling they will still wear out a lot faster with heavy use. Finally, write times are still slower.
    Hemlock Stones
  • Hybrid systems using DRAM

    We've had very good results with SSD units based on cheap DRAM for ultra-high-volume, high-volatility storage. Previous units had integral hard disk and battery backups, but we're working with a couple of vendors to use flash as the backup going forward. It has many of the benefits of flash-only SSD unites (lower heat, no moving parts, smaller battery requirements) and gets us away from the wear-leveling issue, since writing to the flash only occurs during infrequently.

    But there is a perversity factor in using DRAM: the cheapest DDR memory now runs at a far HIGHER clock speed than is required for SSDs limited by FibreChannel or other storage interfaces. This causes heat and power issues. If we could find cheap DRAM modules that ran much slower, but with the same density and reliability, it would be great. The chipmakers know how to build them, but since there is no mass-market for such devices, they don't build them, or if they do, it is at a much higher selling price.
    terry flores
  • What about vertical recording?

    As people demand more capacity to store just about everything on their computers, won't vertical recording keep "high-end" drives competitive?
  • RE: High-end disks dead in 2 years

    What about holographic recording?
    My bet is that this technology will replace Blu-Ray in 2-3 years! Maybe even some combination of holographic and SLC will come to the forefront.
    Let's keep guessing... That's what keeps things interesting!

  • RE: High-end disks dead in 2 years

    It seems like you are saying SLC flash chips are rated for 1M write cycles, they are not, more like 100,000 write cycles.

    SLC-based flash drives as a whole might be rated around 1M write cycles but your comparison of SLC and MLC write endurance seems a little muddy...
  • RE: High-end disks dead in 2 years

    A few thoughts:

    1) High reliability is absolutely "table stakes" in the high-end. As time has proven, high reliability drives costs up. I don't think we've seem a valid cost structure for high reliability flash yet. What's good for a MacBook Air is not good enough for a storage system.

    2) I think a more valid comparison in this part of the market is $/IOP. Is there any published comparison from SPC I haven't seen yet?

    3) It seems to me the moving target of any $/GB crossover point will continue to move out for some time. Certainly two years is over aggressive. EMC's storage business must be longing for some drama these days - I think they are missing their spot in the limelight relative to their brethren at VMWare.
    • More fodder

      did you see Mike Workman, CEO of Pillar Data Systems, take a jab at EMC's premise?