back to article Huawei, DataCore fire torpedoes into SPC-1 storage box benchmarks

Huawei has broken the 3 million IOPS barrier in the SPC-1 storage performance benchmark. Meanwhile, DataCore has shot past 1.5 million IOPS with staggeringly low latency and $/IOPS numbers. These two suppliers are virtually rewriting the future of SPC-1 performance. The Storage Performance Council's SPC-1 benchmark measures, …

  1. Anonymous Coward
    Anonymous Coward

    I hear Volkswagen are very good at benchmarks.

  2. Anonymous Coward
    Anonymous Coward

    Volkswagen makes cars... you are on the wrong forum.

  3. Anonymous Coward
    Anonymous Coward

    The fishing's good today I see...

  4. ecarlseen

    Muh real-world configurations

    After about two decades of watching storage benchmarks it's still hard to imagine how people don't take them with a New Panamax freighter full of salt. These configurations are built specifically to win benchmarks, and don't even remotely resemble anything you would find in the real world. Yay, somebody built a box that would sell for a few million dollars with 200 TB of storage. That's not a PO that's going to cross many (if any) purchasing desks.

    I understand why the vendors spend the time and money on it - they make for fun press releases; I fail to understand how any respectable IT journalist would use it for anything other than giggles and mockery.

    1. GlockandRoll

      Re: Muh real-world configurations

      OK, so then tell me why (since the car analogy was used above) do we have an SAE net HP rating?

      DO you think Dodge would sell as many Hellcat's if it wasnt certified by a vendor agnostic 3rd party that is industry accepted? Becauase that is what you are doing.. you are saying that Dodge cheated and the car doesn't really make 707hp.

      1. Dan Wilkie

        Re: Muh real-world configurations

        Well yes they probably would. I don't know how you do things in America, but generally over her we tend to test drive a car before we buy it, rather than just buy the one with the highest horsepower or the most torque or the best fuel economy.

        Although having driven a Prius as a pool car a few times, it seems clear that not everybody believes in actually test driving something before you buy it :\

        1. GlockandRoll

          Re: Muh real-world configurations

          Ok, you've lost me here. What is the point you are trying to make. Are you saying that DataCore doesn't drive faster IO at lower latency by using parallel IO technology?

          Please explain.

  5. Anonymous Coward
    Anonymous Coward

    The great thing about benchmarks, is that they clearly highlight what vendors are capable of. What always makes me smile, is when a vendor gets criticized for putting their best foot forward in a benchmark. Let's be very clear, every single vendor who goes through the SPC-1 benchmark, puts their best foot forward in hopes of showcasing what they are capable of - I don't think anybody is trying to hide that. What is also very clear, is that DataCore was 9x faster than their closest competitor at a fraction of the price on a standard x86 server!!!! Anyone who is critical of that is missing the point entirely. This is an absolute game changer

    1. Anonymous Coward
      Anonymous Coward

      This isn't about the usual putting a best foot forward and gaming the benchmark in your favor, this is making a mockery of the whole process.

      The DataCore numbers look great on paper but what isn't obvious is that it included single points of failure throughout, the only pretense made to redundancy was to provide mirroring of the data disks on that single node (not other disks) since SPC mandates some form of raid protection for those.

      It's effectively a single controller direct attached result with no semblance of high availability and therefore not even close to a real world configuration. No one else is submitting such configurations to SPC so the only fair comparison DataCore can make is with themselves.

      Add all the missing bits and suddenly the cost doubles, latency rises and the IOps tank, I'm very surprised it was even accepted in its current form. Run the test with the missing bits and we can then compare, but I suspect we'll have a long wait.

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