I have seen in the past here some questions related to how to size a * server with regards to the number of phones, the number of trunks, the number of active calls, etc and I wanted to throw some comments and things out to the group here to try to make sure that I have things straight in my head.
First, with respect to the number of active calls that a server should be able to handle. I would expect this to be dependent upon the TYPE of call.
Scenario #1
User A calls User B. If a user with a SIP phone dials another extension and that call completes then the call is really between both the endpoints and that call should not be placing much load on the server at all. The server is not involved in the RTP transmission at all nor is it involved in the call control once the call has been completed.
Scenario #2
Conference call between three or more users. This is a scenario where I would expect the server performance to be impacted since it is up to the server to mix the RTP data and present it to all the endpoints.
Scenario #3
User leaving a voice mail. Another scenario where the server is completely involved in the call and I would expect its performance to be impacted by the number of active calls.
Scenario #4
Call arriving to a user endpoint via a hardware card. In this case I would expect the server to also be fully involved in the call, at least form the perspective of getting the voice data off of the card and into the RTP stream then off to the IP endpoint.
Scenario #5
Call arriving to a user endpoint via a external gateway device like a Sipura box or a some other external gateway device. This is NOT a scenario where I would expect much server involvement since the call will end up being between the IP side of the gateway and the destination IP endpoint.
I am sure there are more call scenarios that you can come up with but these are some quick ones that came to mind.
Anyone here care to comment on them?