Everybody knows the blocking on typical telephony, but on voice over IP are many differences. Suppose Poisson arrivals (wrong estrictly ), the strategy queueing no always is FIFO (could be FQ, WFQ, others).
Anyway if you find the process prob. distribution and get a blocking on that, how can you define the delays???
You’re right that there are many differences between IP and traditional telephony transport networks. However, the nature of voice demands immediacy, and a queuing strategy is not appropriate for voice in the same way as it is applied to data. Our own approach (and there may be others), is to calculate bandwidth requirements by working out the number of voice paths required during a busy hour, and then translating this to bandwidth requirements. In other words, we make it a two step operation. Blocking could probably be calculated in a similar way.
Any attempt to trade off bandwidth against bandwidth, as might be the case with data applications which are not so delay sensitive, must be doomed to failure.
Systems need to be engineered to avoid jitter (the variance of delay), but the delay experienced by a call seems to be very difficult to quantify. It depends upon many factors, including the voice coding algorithm chosen and the characteristics of the transmission network.