"you need a few servers with UPSes" imagines a world where companies can responsibly maintain a reasonable computing environment independently. oh such optimism! refreshing!!
we are near to having a world where that might be reasonably possible & likely, thanks to the availability of off-the-shelf open-source broadscale diy cloud software like Kubernetes. but until today, few companies have been able to create & more challengingly maintain healthy points of presence that they can manage & run competently. there have long been offerings like Puppet and Chef & countless others to help a company try to maintain systems, but it's always been a picking of assorted pieces, a cobbling together of various concerns. rarely has running a couple boxes been a reasonably strong competency for most companies, as simple & elegant sounding as the idea may come off as.
and if you do have people with less than stellar, 100% self-hosting experience, they will have no where to go to learn & skill up. all the articles, all the focus today is on using cloud systems. when you're on your own you're on your own.
i do think we are nearing a new era, a new place, where running systems may be possible. but it's only because we have started to upscale & compete with what the cloud offers. we've begun to reconsider how most software is ran, & have advanced our collective skill, knowledge & systems to align towards more managable, coherent, modern practices. we've learned so much. but we're only just beginning to figure out how to deploy & run & keep running these private clouds. and anything less seems like an act certain to become a tragedy.
> It will when these companies need to start competing on margin.
i see the future as supply side driven, not, as posited here, demand side driven. running your own has usually been a barrier to competing, to keeping your talent focused on product & development. we've had no general access to good, strong, competent, robust, well used alternatives for running software, versus using the cloud & it's array of easy to latch onto, well maintained, affordable, pay per use, easy to scale cloud service. we have been gaining that rapidly with Kubernetes, with open source & even proprietary, assistive Cloud Native Computing Foundation & other software. we're just starting to work together, to have known quantities we can put into place.
the option of trying to cut cost, avoid the colossal Hyperscaler surcharges has been out there. but it has absolutely, per the parent, been radically in favor of not running shit yourselves. i strongly agree, "for most devs, for most workloads, the cost of running a data center isn't worth it," because there's been few good understandable deployable patterns for running software. Kubernetes marks an early beginning where we emerge from this tar pit. nothing else has come close to providing a solid base for computing alike this. i'd seen some decent Rancher Fleets in action, which came semi-close, but like many of the possibilities from >5 years ago, it was esoteric, known to few, hard to skill up in, fraught with unknown perils, and few could help you. the situation has just started changing, we have just started finding common shared practices for computing, where skills & expertise are exchangeable across systems, where newcomers to your business have some chance of following along, understanding, on their own, with the out of the box tools. without having to deeply dive into the particulars that you have used to cobble together your systems.
cobbling shit together is how things have gone, & it's a long term nightmare, as easy as it is to start. we're finally becoming better. but i'd still say, for most entities, it's too soon.