...
For fault tolerance, 2 servers are sufficient: one of them will be the entry point and serve all services, the second is only part of the RTP traffic. In case of problems with the 1st server, its address is automatically (VRRP protocol) picked up by the 2nd server and the entry point changes. Servers are preferably located in different data centers.
The platform is designed in such a way that there is a possibility of increasing productivity with an increase in the number of calls, by adding new servers. For example, add 2 media processing servers to the previous scheme.
In this case, "Server 1" and "Server 2" remain as interchangeable entry points for telephony / API, and "Media 1" and "Media 2" are used exclusively for voice / video processing.
...