Comment by londons_explore

Don't be scared of 'one big server' for reliability. I'd bet that if you hired a big server today in a datacenter, the hardware will have more uptime than something cloud-native with az-failover hosted on AWS.

Just make sure you have a tested 30 minute restoration plan in case of permanent hardware failure. You'll probably only use it once every 50 years on average, but it will be an expensive event when it happens.