• stoy@lemmy.zip
    link
    fedilink
    English
    arrow-up
    9
    arrow-down
    3
    ·
    2 days ago

    Humm, this will probably mean that the EU will need to look into if we need to setup a European mainframe manufacturer.

    I am talking AS400/iSeries type stuff.

    MasterCard and VISA process a huge number of transactions per second, and there can’t be any risk of loosing a transaction in progress, so you need an extremely stable central processing node with very high redundancy.

    At the moment I believe that only IBM and Fujistu makes mainframes these days, IBM is American which has now shown to not be an ideal long term trading partner, Fujistu is Japanese, with a strong presence in Europe, but they made the UK Post Office computer system, which makes me want to stay, far, far, far away from them.

    Either one, whoever we pick will make it easy to get the system going, but to migrate away will be a nightmare.

    I wonder if we could build something on open hardware like Risc-V, this make me wonder is Risc-V would even be suitable for this application

    • Sizing2673@lemmy.world
      link
      fedilink
      English
      arrow-up
      2
      arrow-down
      1
      ·
      4 hours ago

      Lmao someone would be very incompetent to actually propose the idea to create a mainframe system to do this

      It would be so stupid , it would be ancient slow and hard to maintain

      Everything this century that’s new is cloud, distributed, HA, real time, event driven, and fast low latency

      Mainframe only has some of those features, plus really ancient legacy and other stuff that makes it not perform as well in certain areas

        • Sizing2673@lemmy.world
          link
          fedilink
          English
          arrow-up
          1
          ·
          edit-2
          10 minutes ago

          I do

          I mean slow in terms of innovation, which they stagnate on

          But also performance actually

          Their TCP\IP stack is one such consequence. It doesn’t have any of the massive changes that happened in the last few decades that have optimized performance

          Open source stacks picked those up immediately. Windows, and other older platforms still use a much slower and more poorly designed stack

          That’s one such example. Plenty of others

          It’s not that they can’t solve problems. They can.

          Steam engines can solve everything too. But they are not the best at every task and these days it’s hard to find anything that couldn’t be beaten otherwise

          Mostly these systems ONLY exist because of legacy

          It is why none of the big compute players have touched any of that in decades. Because it is dead technology and a dead end

    • overcooked_sap@lemmy.ca
      link
      fedilink
      English
      arrow-up
      5
      ·
      24 hours ago

      Tech guy here. There’s no way in hell a new system would be mainframe-based. A distributed queue with delivery receipt and many nodes to process messages along with many distributed read-only DBs is the way to scale this thing. And you can be isolated form local power and connectivity issues. Tech isn’t the problem in this situation, market penetration is.

      • Sizing2673@lemmy.world
        link
        fedilink
        English
        arrow-up
        1
        ·
        9 minutes ago

        I know right!

        I think this is another case where even the engineers got sold into the marketing

        Because they heard three decades ago that this was the fastest best technology, and IBM sold it to them

        … The reality is, these technologies nobody uses for anything new and there’s a reason why. They are just too ancient and stagnating

        Plus, other technologies are open and you can see how much more innovation happens when you allow that. Mainframe never had that and that’s why it sat around

    • jenesaisquoi@feddit.org
      link
      fedilink
      English
      arrow-up
      7
      ·
      edit-2
      1 day ago

      There has been massive progress in the last 40 years in distributed computing and consensus algorithms (which is what you need for consistency in a distributed system). We don’t need 1970 style centralised systems anymore.

      • stoy@lemmy.zip
        link
        fedilink
        English
        arrow-up
        2
        ·
        1 day ago

        For normal tasks, absolutely, and if we can do it without a mainframe while maintaining the stability and redundancy of a mainframe system, then we should look into alternatives.

        However, mainframes have been in continous development since they were created, there are absolutely tasks they still do better these days.

        • Sizing2673@lemmy.world
          link
          fedilink
          English
          arrow-up
          1
          arrow-down
          1
          ·
          edit-2
          4 hours ago

          I think this is just old style thinking

          Cloud and modern technologies are far superior. There is a reason why the fastest most demanding data centers do NOT use any of that old technology

          Also, “continuous development since they were created” is pretty much nonsense . It’s technology, it’s all always been in continuous development

          Mainframe and other technologies have stagnated for the last several decades. They haven’t developed much. IBM is the main one in the game and their main strategy is vendor lock in. Not innovation, and definitely not them updating and keeping up with the times

          Even their TCP\IP network stack is decades behind other technologies.

          It’s all proprietary so it’s less efficient, less innovative, less secure, too

          • stoy@lemmy.zip
            link
            fedilink
            English
            arrow-up
            1
            ·
            2 hours ago

            Oh no, not the fucking cloud.

            Didn’t we just talk about taking the data back?

            Let’s run our own servers and not a needless third party…

            • Sizing2673@lemmy.world
              link
              fedilink
              English
              arrow-up
              1
              ·
              13 minutes ago

              Dude. They are technologies

              You can make your own damn cloud if you want to. I think you missed the point entirely

        • jenesaisquoi@feddit.org
          link
          fedilink
          English
          arrow-up
          4
          ·
          1 day ago

          Like what? I worked with some of them - they are nothing but the software that runs on them. Companies buy them to keep the ancient software they depend on running, not because they’re good or necessary. Software can be rewritten. The companies I worked with got rid of around 50% of their old propritary systems like SystemZ, POWER, Superdome, and NonStop in favour of plain Linux on amd64 in the last 20 years, and the trend is continuing.

    • barsoap@lemm.ee
      link
      fedilink
      English
      arrow-up
      5
      ·
      1 day ago

      CPUs don’t make something a mainframe, the whole system design does. They’re transaction-based throughput monsters with all kinds of bells and whistles when it comes to reliability, seamless fallover, etc. The European CPU initiative currently focuses on supercomputing (weather simulations and such) which is a completely different beast when it comes to dataflow but certainly a good foundation for general compute.

      Looking into my crystal ball, at some point SAP is going to enter the hardware business.

    • Illecors@lemmy.cafe
      link
      fedilink
      English
      arrow-up
      5
      ·
      2 days ago

      Mainframes have nothing to do with this.

      RISCV is still just a computer - would work just fine on a logical level. Raw compute would be an issue with today’s hardware.