How to calculate the required terminal server performance?

Question

Performance evaluation

It is required to calculate the required performance of the CPU and disk subsystem.
Namely, _calculate_ in numbers, with links to the source, how many CPU resources each application needs, and how wide the disk subsystem channel is needed.
Comments like “I have such a server, everything flies, buy yourself the same one and don’t worry” will be studied with interest, but what is needed is a calculation with justification and links to authoritative sources.
With RAM, everything is clear (2-3 GB per person), but with the processor, everything is incomprehensible.
I used the search, but I didn’t find anything sensible, maybe I didn’t search well?
Required server configuration.
Number of employees: 50 - 100 people.
Operating system: MS Windows Server 2012 R2 (2019) either as a virtual machine or as a separate server on dedicated hardware.
Client applications that will run on it: MS Office, Adobe Acrobat, Skype (in text mode), 1C, Google Chrome, Mozilla Firefox, Thunderbird, file antivirus, several self-written applications on .NET and other office small things like notepad.
Currently using Dell R720 CPU 2xE5-2640 RAM 127 GB, HDD 8xSAS 10K RAID6 and it is no longer enough.
The new server will definitely be on a mixed-use SAS SSD, this is the only thing that is clear so far.

Answer 1

If all users work the same way, you can count.
If one opens 1C with reports, and the other solitaire, you won’t count.
Buying one server for this is not very good. Usually they put a cluster of smaller machines, and increase the volume as needed. At the same time, it is more convenient to maintain, and the fall of one node in the cluster will simply require some users to re-login.

Answer 2

Here is an interesting article.

Answer 3

IMHO you will not find such a rhyme.
Only empirically, removing the Performance counter from an existing server or workstation and extrapolating to the number of users on the terminal server.

Answer 4

No way. The solution is selected according to the task, if the task is described in numbers, then the solution can be obtained in numbers, if in estimates plus minus a kilometer, then the solution goes in the same format.
In your case, either edit the zoo (to be honest, I don’t really understand why you need rdp at all), or figure out a solution for the resource monitor.
Current cpu's have significantly more cores, this won't help you for long. Plus there will be two servers, so users can be divided.

Answer 5

About 2 years ago, I also tried to figure out the whole thing under an RDS farm - all garbage.
The biggest minus of RDS (when compared with VDI) is that there is no resource management: everything for everyone. And everyone suffers.
Creating an RDS farm improves the picture a little. different groups of users can be given different hosts/resources to use.
I have a couple of RDS farms from 4 to 16 hosts in each and everywhere the CPU is 90+ percent full during working hours and it can be "guzzled" by 2 users by opening 100500 chrome tabs and 20 boobs with their 1C and excel - just "comfort" of work in such cases is different.
Empirically, I came to the conclusion that it is optimal to keep + - 15 users per host (in my practical standard "load" regarding software).
So if you buy a new server - raise an RDS farm (there are a lot of manuals) - you will be able to increase fault tolerance and you can manage resources somehow. In this case, it is better to keep user profiles on UPD.
Take the CPU with a higher frequency if there is 1Ska.
Disk subsystem - SSD unambiguously!