€40 gives you a dedicated server, not just a VPS.

Getting metrics on that is not a hard problem, there are various projects that are relatively simple to set up.

If you want to make it easier manage resources, metrics out of the box, and avoid (hoster) lock-in then I'd use a hyper visor distro like Proxmox VE (disclaimer, am a dev there) or the like, and you can migrate (or backup/restore) VMs or Containers easily to other providers. That gives you a (relatively) simple web-interface to manage most things and also opens the possibility to just add a second or third dedicated host down the line to scale out, if those new hosts are in the same DC or have a good interconnect (latency wise) you could even cluster the nodes.

To make a fair comparison you need to consider the time cost for setting all of that other stuff up compared to having it out of the box. I'd an engineer on 100k takes a week to get it up and running then your vps cost 2k to set up and 40/month going forwards.

If an engineer needs that much time you have a serious technical debt for setup of your software or a inexperienced or inept engineer. The developer or at least operators need to be able to setup your software for more frequently for testing anyway, if they cannot do that rather quick you got other problems..

I can set up Proxmox VE as hyper visor, some container for each DBs, load balancer in front and some app in about an hour max from scratch, with good testing and some bells and whistles, and here I really do not want to brag or the like, as such operations are not my job to do at all, I only know because I do that occasionally for some tests and for some private infra I just maintain out of interest - so I really want to say, if some operation-dork can do that, the engineer you hired should be able to do it at least as quick.

But yes you're right in the general point, upfront setup and frequent maintenance is naturally something you need to price in. I just think that if you have that many different parts with complex coupling to induce such a huge maintenance effort required to keep your product running, the cloud offer may not really be your salvation and just delay the fall while costing all the more.

> If an engineer needs that much time you have a serious technical debt for setup of your software or a inexperienced or inept engineer.

Everyone does something for the first time once. Just because someone has not set up a hypervisor before doesn't mean they're inexperienced.

> I can set up Proxmox VE as hyper visor, some container for each DBs, load balancer in front and some app in about an hour max from scratch,

And I can spin up containers + load balancer on AWS in less than five minutes. That doesn't mean that it's just an easy thing to do. (although, this specific example is).

> upfront setup and frequent maintenance is naturally something you need to price in. <...> the cloud offer may not really be your salvation and just delay the fall while costing all the more.

Agreed 100% on both counts.

>Just because someone has not set up a hypervisor before doesn't mean they're inexperienced.

Wait, what? If you never did something, then you're unexperienced, ain't you?

You can be an experienced $job without having ever done $one-particular-thing-related-to-job.

Experienced != Knows 100% of things.

The difference with a PaaS/serverless system is that I don't need to hire you, or have someone on my team learn to be you.

I'm sorry, but all that stuff you describe doesn't bring any business value. My customers don't care what hypervisor I'm running, so I don't care either. PaaS means someone else deals with it, forever. The last time I had to employ an ops (or devops) person was 2007.

True, but setting things up on AWS isn’t free either.

That's true - if you want to make a fair comparison between the two you need to consider the costs of the setup on AWS vs the cost of setting it up on whatever your platform of choice is. For a small team with no/development only loads, then a $5 digital ocean droplet would likely work for them, maybe even 10 of them. It's not worth managing a VPS for deploying 5 containers when you can have DO do it for $25 behind a load balancer. For a small team with moderate load, the question is "is it worth spending X on setup to save Y but potentially spend Z on maintenance of the systems on Hetzner/whoever, vs spending A on setup, B on compute and C on maintenance". If the difference is < 6 months salary, you go with whatever your current team is comfortable with and reevaluate in a year.

For a large company, it's not about $ cost, it's about risk management and avoiding cost centers.

Good point. If you are using Linux on a daily basis, it's easier to set up a server than configuring AWS.

You'd have to figure out the setting up stuff just once though and then automate it. It's not like you have to go through this for every additional server you will add in the future or when you have to rebuild it.

Also, it doesn't take a week.

> You'd have to figure out the setting up stuff just once though and then automate it.

You're assuming that this is for recurringly set up infrastructure. Sometimes infra is set up once and maintained, othertimes it's set up and spun down. It's also not always automated. The time spent automating something like that might not be worth it in the medium to even long term.

> Also, it doesn't take a week.

The actual amount of time it takes doesn't matter; if it's a day or a month. what matters is costing the time spent on setting it up and maintaining it, and pricing that against AWS costs.

You can use Hetzner's cloud. You get metrics and still have a lot of free traffic with very low cost above that.