The Erlang model is just as applicable to VoIP systems too, to argue it doesnt is to show a complete lack of knowledge of telecommunications principles (note I say telecommunications, not voice!)
In a voip network there will be a requirement for a certain amount of voice traffic, expressed as erlangs. CCS, call minutes or whatever. Erlang tables convert this to a quantity of ‘channels’ to carry that traffic. Each ‘channel’ is the equivalent of the voice compression standard (G.711, G.729A/B etc) plus IP overheads.
For instance, 10 Erlangs of traffic requires 18 lines/channel to carry that traffic. At 64kb (G.711) compression plus Ip overhead each ‘channel’ occupies 80kb, therefore the MINIMUN bandwidth for your network needs to be 18 x 80kb= 1.44Mb just to carry voice. This is easily achievable across a local 100Mb LAN, but must be considered if you are considering VoIP via say a 256kb DSL style connection.
Increase the number of phones, ERLANGS increase, bandwidth requirement increases. Once you hit b/w near 10Mb you may need to investigate the capability of hubs, switches and routers!
Throw in some IP video-conference and your LAN may just begin to sssllloooowwwwww down.