[ownagefool]

The consequence of running ingress and DNS poorly is downtime.

The consequence of running a database poorly is lost data.

At the end of the day they're all just processes on a machine somewhere, none of it is particularly difficult, but storing, protecting, and traversing state is pretty much _the_ job and I can't really see how you'd think ingress and DNS would be more work than the datastores done right.

Now with AWS, I have a SaaS that makes 6 figures and the AWS bill is <$1000 a month. I'm entirely capable of doing this on-prem, but the vast majority of the bill is s3 state, so what we're actually talking about is me being on-call for an object store and a database, and the potential consequences of doing so.

With all that said, there's definitely a price point and staffing point where I will consider doing that, and I'm pretty down for the whole on-prem movement generally.

[vidarh]

I'm generally strongly in favour of bare metal (not so much actually on prem) but your case is one of the rare cases wher AWS makes sense. Even for cheap setups like that, bare metal could likely be cheaper even factoring in someone on call to handle issues for you, but the amounts are so small it's a perfectly reasonable choice to just pick whatever you're comfortable with.

That's the sweet spot for AWS customers. Not so much for AWS.

The key thing for AWS is trying to get you locked in by "helping you" depend on services that are hard to replicate elsewhere, so that if your costs grow to a point where moving elsewhere is worth it, it's hard for you to do so.