back to article Dell (Michael, that is): EMC's DSSD a 'game changer'

Walking around Dell World 15 with former Dell hyperconvergence guru Jimmy Pike was a bit like being with a rock star. As we strolled the corridors Pike, now a consultant and commentator at tech analyst firm Moor Insights & Strategy, fell into old colleagues who greeted him like a long lost friend. Even founder and chief …

  1. Anonymous Coward
    Anonymous Coward

    DSSD

    At which point does DSSD becomes a meaningful product rather than a series of blog posts and hype?

  2. TANKtr0n
    Terminator

    Not Vaporware

    Check out the TACC Wrangler system that's been deployed at the University of Texas (Austin). Not surprising, but this appears to have actually been designed/built as a joint Dell-EMC HPC behemoth.

    Link to their announcement back in April here: https://www.tacc.utexas.edu/-/wrangler-data-intensive-system-opens-to-scientists

    See the DSSD component (0.5PB - Wow!) listed in the Architecture section of the TACC Wrangler user guide here: https://portal.tacc.utexas.edu/user-guides/wrangler#architecture

  3. TeeCee Gold badge
    Meh

    The trouble with flash....

    .....is everyone knows that, in a minute, one of the "flash-killers"[1] will come to market.

    At that point, anyone having just invested a shitload in new flash product or manufacturing capacity will see their ROI vanish and go titsup.com.

    [1] i.e. One of the several RAMish RW speed, long lived[2] types well into development and already sampling

    [2] Speed and longevity removes the need for complex and (relatively) expensive controllers providing parallel IO and "wear levelling" to make it work.

    1. Nate Amsden

      Re: The trouble with flash....

      There was an article recently here I think on el reg that said Intel designed the NVMe interface specifically for those flash killers. So when they arrive products like DSSD won't really be impacted I think they can just swap their flash chips for whatever is the new thing, the software and hardware built around the super fast storage should stay the same.

      1. HonestAbe

        Re: The trouble with flash....

        "Intel designed the NVMe interface specifically for those flash killers." Exactly. DSSD sounds cool, but it doesn't appear to compete with anything at the component level. Rather it appears to be leveraging the rise of NAND (and presumably Intel's new NAND-killer) to fully realize their potential in storage.

        Sounds very cool, but if it's as good as Dell thinks then Oracle, HP, and IBM should be worried... and Intel, Micron, and Sanshiba should be rejoicing.

  4. captain_solo

    The implied subtext

    "We don't have anything today that can compete with Oracle Engineered Systems"

    Their problem is going to be that Oracle can already put very high speed low-latency flash in the servers of their engineered line, as well as closely connected all flash array or high speed disk fronted by flash, which is really all a purely agnostic storage hardware solution could realistically do.

    They won't have columnar compression, they won't have intelligent query offload, they won't have software in silicon (although I suppose you could argue that one will improve Oracle DB and Java performance with Sparc connected to any storage solution)

    Depending on workload and characteristics, its unlikely they will be able to overcome the software/integration advantages just by adding faster i/o

  5. Anonymous Coward
    Anonymous Coward

    Doesn't Intel Crosspoint do this better

    It seems like DSSD is a flash version of Intel Crosspoint, which means it will be much slower. Looks like Michael is missing something!

  6. Terafirma-NZ

    Where they will face a challenge is 3DXPOINT and RDMA this tech will see servers with TB of NVRAM replacing normal RAM and removing SATA/SAS/SCSI storage all together, mix that with RDMA and its ability to scale out to a whole DC and DSSD looks like yesterdays tech. Yes they have a jump but it is so highly engineered to a custom spec that by the time it hits market it could only be 12 months before we see 3DXPOINT arrive followed by a whole wave of new products and startups.

    Watch the Intel parts of SF8

    The network really is about to become the bottleneck of speed.

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

Other stories you might like