I'm a bit conflicted about this. Yes, AWS is expensive. And depending on your workload you can certainly save a fair wedge of cash by colo-ing things. For wor

I'm a bit conflicted about this. Yes, AWS is expensive. And depending on your workload you can certainly save a fair wedge of cash by colo-ing things. For wordpress, (the site that is running serverthehome) its perfectly possible to run it in a lambda, with aggressive caching and save a whole bunch of cash. depending on plugins, load and a number of other things, it could be a potential saving of 90%. (plus a massive reduction in attack surface) Before you ask, yes I do know from bitter experience, parts of the financial times had WP wedged into them. making them fast and secure was an interesting experience. AWS is only cost effective when you are using ec2 with a duty cycle of less than 50%. This means that you are not using AWS for hosting 24/7 compute. In terms of storage, There are two compelling offerings, S3 and EFS/lustre. however if you're a large scale EFS user, its better to run your own GPFS system. The hardest part of "scaling" is orchestration. If you're using K8s, there really isn't much difference between running it on virtual vs real steel, barring bringup scripts(don't get me started on networking in K8s, its totally warped.) TLDR: IF you need 100+ machines on 24/7, AWS is going to be more expensive. If you have transient loads, and the average time on for a machine is 4 hours or less, then AWS is for you. However, you'd better use all the other bits that come with AWS to make it cheaper, like fronting it with fastly/otherCDN