You could easily throw the components into an old tower case.

Getting the PSU to fit could be a bit tricky due to the rather short cables.

Workstations, like real workstations, are another beast and not what’s typically referred to as “office PCs”, those are indeed rather sff builds.

Again, optiplex sff 3060 as an example, it has two SATA ports, one x16 and one x1 (I think) PCIe, and looking at the PCB, apparently there’s a version with m.2 slots. Sure, not exactly server grade storage, but if you manage to find some version with m.2 slots or invest 10€ for a cheap SATA card, you can get enough storage attached.

GPU wise, absolutely no idea. My optiplex has a wx3100 that I got for cheap and its self reported power draw never goes under 5W, but since this machine is a desktop, it doesn’t run all day.

Sorry, but you’re either pulling those numbers out of your ass or haven’t kept up with the real world for 25 years.

The numbers I’ve posted above are measured using an external meter. I’m German, I have a vested interest in knowing how much power my devices pull.

And you don’t think, office PCs pay attention to power consumption, given they are intended to run 8h a day?

My optiplex sff runs at about 10-15W in idle, and it’s an i5 6500. The t variant in my elitedesk runs at 5W.

If you don’t need actually public DNS, something like Tailscale might be an option.

“Even deader internet theory”.

Yeah, but that’s the thing: it’s not fast. And that’s actually even more baffling.

All those rewrites cost money. All the shoddy software somewhere in the background of every corporation was written using these shitty libraries, frameworks, APIs. And that means cost.

I’m explicitly not talking about the newest crap Amazon or Google push, they operate under different principles, I’m talking about the data plumbing stuff. The eight billion Spring Boot apps, the PHP sites, the Python pipelines.

Writing and deploying a simple CRUD app, that just takes data from a request and saves it into a DB should be done in a few minutes. The actual “payload” is just input schema, DB schema, transformation rules, maybe auth. That’s it. However, if you want to do that in Spring Boot and K8s, it takes often hours or days to get everything right. Not because Spring Boot is hard, but because you forget a network policy in Helm or the dateformat between JSON/Java/DB is different or that library is outdated or there’s CVE somewhere or you have to look up that shitty mappedBy thingy from Hibernate again (that’s maybe just my problem, I can’t remember that crap) etc. etc. etc.

The basic problem behind is, that we have to take care of so much. Network, business logic, framework detail. You have hardly any layer or platform that is really solid.

Honestly, I would advocate the exact opposite.

Yes, programs became bloated and fragile, but the solution cannot be to return to the stone age, but be professionals for once.

Our entire industry is shit at actually engineering. There’s leaky abstractions everywhere, and that’s exactly why everything is so complicated and brittle. There’s no platform to build upon, only a scaffolding made of twigs, duct tape and three bananas for some reason. Every minor change in some minor library percolates through the entire stack.

You’re a simple developer, so am I. And we both probably wrote hundreds of apps that essentially do the same crud crap again and again and again. The same basic functionality gets implemented thousands of times, because we can’t get our shit together to build actually reusable components. Instead we rewrite the 12th iteration of “make stuff move in browser” and “make Java do business”.

We’re not engineers, we’re children with hot glue guns.

It’s just a question of time. Every platform will devolve into either obscurity or cesspool.

None. There is no model that can output anything even remotely usable on that tiny amount of RAM and certainly not using the few CPU cycles your vps has to offer.

The US default, that I never left Europe. What an achievement for the USA!

Oh come on, are you really that boneheaded not to understand that you’re not the norm?

I literally had not a single power surge in my entire life. The only power outages I had were for a few minutes maybe three times in the last 15 years.

The larping refers to you. Either you are truly an outlier who actually runs a small DC, or you just like the feeling you can get pretending to do so.

Your attitude is roughly the “only gold plated cables made from solid silver” equivalent in audiophiles. Technically maybe correct, practically a self-important waste of money.

But not for us.

That’s what I meant by larping. The vast vast majority of us here would probably not even notice if their systems went down for an hour. Yes, battery backup has its purpose. In a datacenter.

I mean, what’s on the line here in the worst case? 15min without jellyfin and home assistant? Does that warrant taking risks with old batteries or investing in new ones?

That equation might change if you’re in a place with truly unreliable electricity, but I guess those places have solutions in place already.

That’s typically a feature for servers or business desktops. Maybe your laptop has it, just look into the BIOS.

As I wrote in my other comment: try to be realistic about your needs. Chances are, pressing the power button every few months (if at all) is perfectly fine for your use case (and most others here).

And how much need is there for a UPS in this scenario - realistically.

Some of the people here take their admin-LARPing a tad too seriously. Most households have reliable enough electricity, and even if there’s an outage once every quarter, would a dead battery even help?

I advocate for being realistic with one’s own needs. Don’t build a five-nines datacenter for a glorified weather station or VCR.

In case you didn’t already do that: remove the battery. It’s probably dead anyway, you don’t need it and it poses a potential (albeit low) risk.