How do you estimate On-Premises Data Center Infrastructure Requirements? [closed]

I need advice on estimating the compute, storage and bandwidth resource requirements for an on-premises web data center given a number of users, frequency of requests per user, and expected response latency. Some servers will be web servers and some machines will be database servers. No big data or heavy data analytics or machine learning. Also, how do you estimate the size of the HTTP request and response message given typical page display content? I am sure there is no simple cookie-cutter formula for these estimates, and elastic cloud would be much easier to deal with, but is there any advice on doing this off-cloud? I know the number of users will be in the range of 1 to 5 million users who each use the app about 10 times a day.