Note: the Ars Technica article links to a blogpost by Lydia Leong, a cloud computing analyst at Gartner, the URL seems to be broken. I believe they wanted to reference this blog post.

  • Alphane MoonOPM
    link
    English
    923 days ago

    I am curious to what extent IT repatriation is a discussion topic versus a real phenomenon. You hear a lot about this topic in the last few years, but very market-level numbers.

    On a purely back on the napkin analysis, it would make sense, since AWS/Azure/gcloud would want to raise prices once they have a foothold on the market (and some of the on-perm knowledge is lost).

      • HubertManne
        link
        fedilink
        323 days ago

        already happening and the company I was at reaction was different cloud provider which required so much work to move between they are never going to see the labor savings. the cloud should theoretically give them.

    • @[email protected]
      link
      fedilink
      4
      edit-2
      23 days ago

      The two things I see people forget when crunching the numbers:

      1. Cloud is a recurring cost. So you have to factor TCO over the life of the application which realistically is gonna be greater than 6 years. In those 6 years usually you’ve deprecated you’re on-premise hardware in 3-5, so those last few years are gravy, whereas with Cloud you’ll be paying the same amount year after year

      2. Application sizing and density. Quite often people will break out workloads on the cloud to offer better “segmentation”, but you’re paying for that division. On premise you can better utilize your resources through sharing at the hypervisor level. Couple this with people not sizing their workloads correctly (or at all) and they end up wasting a lot of money

      Cloud is good for rapid prototyping, or if you don’t have the resources to spend yourself, but if you have the size and means, on-premise will almost always win the cost argument.

      ETA: Another factor that businesses liked was that Cloud was a OpEx spend vs on-premise hardware being a CapEx spend and it props up your EBITDA, but that doesn’t really matter much unless you’re looking at selling your company or are a highly traded public company, but a lot of them got sucked in with that view too.

      (Source: I did these calculations and pointed it out to my leadership years ago and now am finally getting the “I told you so”)