You could do that.

But let's say need 4x vCPU: 72, Memory (GiB): 144 for 4 hours. Or you need that 12 hours a day but for the rest of the time you need 2 cpu and 4 GiB of memory.

You need to handle traffic spikes such as TV traffic.

Yes you could self-host a cloud env but you can't scale your resources the way you can with cloud.

I’ve never worked anywhere where that was really necessary. Even when I worked at Microsoft, the services my team built needed to be big scale, high perf, etc… But they would have easily run on a fixed number of beefy machines even at our peak load for a fraction of the cost of Azure.

It really was necessary due to some silly people not listening to me 2-years ago and ignoring technical debt until it got to the point the service they sold with a 1.5 million penalty fee for failing to deliver was needed to be delivered and load tested.

And that is literally the largest AWS Elasticsearch cluster option. So clearly that will be deployed for multiple orginsations. Otherwise they wouldn't have created a default node that size.

> But let's say need 4x vCPU: 72, Memory (GiB): 144 for 4 hours.

I would probably send you back to where ever you came from and tell you to re-engineer that. Cloud or no cloud.

And you would be told that it needs to go live because of a customer contract and failure to deliver would be a 1.5 million euro penalty fee.

Sometimes you need to spend lots of money on tech debt. I think it's nuts that was required but it was.

The German companies are really nuts. The cost to value ratio is through the roof. I’m a happy Netcup customer, and I honestly don’t know how they do it and make any profit at all.

I wish they’d bring those same prices to some US data centers.

I'm sure there are some big political flamewars to be had in this area regarding per capita productivity.