> Amazon really does not have a history of throwing huge margins on raw compute resources
What? My $2000 Titan V GPU and $10 raspberry pi both payed for themselves vs EC2 inside of a month.
Many of AWS's managed services burn egregious amounts of EC2, either by mandating an excessively large central control instance or by mandating one-instance-per-(small organizational unit). The SFTP example you list is completely typical. I've long assumed AWS had an incentive structure set up to make this happen.
"We're practically selling it at cost, honest!" sounds like sales talk.
Yes. Look at the requirements for the EKS control plane for another example. It has to be HA and able to manage a massive cluster, no matter how many worker boxes you plan to use.*
*Unless things have changed in the last year or so since I looked
It is currently 10 cents an hour flat rate for the control plane. That actually saved us money. Even if you weren't going to run in HA, that is still the cost of a smallish machine to run a single master. I am not sure who running K8s in production would consider that too high. If you are running at the scale where $72 a month is expensive or don't want to run HA, you might not want to be running managed Kubernetes. I'd just bootstrap a single node then myself.
You said it yourself: production at scale is the only place where the current pricing makes sense. That's fine, but it means I'm not going to be using Amazon k8s for most of my workloads, both k8s and non-k8s.
What? My $2000 Titan V GPU and $10 raspberry pi both payed for themselves vs EC2 inside of a month.
Many of AWS's managed services burn egregious amounts of EC2, either by mandating an excessively large central control instance or by mandating one-instance-per-(small organizational unit). The SFTP example you list is completely typical. I've long assumed AWS had an incentive structure set up to make this happen.
"We're practically selling it at cost, honest!" sounds like sales talk.