If you’re looking to use windows, then take a wide birth on AWS, and head to Azure, then understand the pricing models as you can save up to 80% off the rack rates by choosing the right machines and other discounts based on how many machines you have (aka available via a partner), and should be around 20% cheaper than AWS.
The advice on how memory hungry windows is is horrific and completely wrong and embarrassing to be honest that linux fanbois are so out of touch with tech.
We run windows 2016 container scale sets with lucee, and spin up windows instances (about 6 seconds to spin up 5 instances to serving traffic and scalesets are automatically under a load balancer so no configuration needed)… and for pricing… dirt cheap and way cheaper than windows on AWS #fail
We find no performance difference between windows and linux for lucee on windows containers vs ubuntu.
Personally we chose MSSQL has more advanced features than RDS or mysql so we went that option as the features meant we had performance gains on the db side. Also having R as part of MSSQL means we can use relational and nosql in a single query with performance, as well as having json output from queries.
If you need to know what VM types and sizes to use, hit me up, but as a general rule B-series are great for burstable instances and F series are great for heavy processing sites (remember vm types on Azure have different cpu, ram and network speeds)
btw you could run nanoserver on any of those instance sizes (t2.micro (1GB) 250 USD/3 years,
t2.small (2GB) 488 USD/3 years, t2.medium (4GB) 974 USD/3 years) with no issues. In the end the deciding factor is more lucee/tomcat when it comes to performance in my experience on smaller machines.