Note, that presentation is from 9 years ago. Our Moodle is now hosted on AWS. In the AWS docs include a 'Moodle reference architecture' and what we do has a lot of similarity with that. However, as everyone here alwasy says - because it is true - the only way to know how much
server resources you need is to set something up and then test with simulated load which is similar to the type or real load you expect.
(With AWS, and other cloud providers, there is also the approach of using the flexibility to scale things up and down. Once your system is running, you can use the live monitoring to see where you are paying for larger capacity that you really need, and scale down to save money. Or, find a bottle-neck and scale it up to save money.)