I’m designing a voice network of an investment firm between Athens-N.York-Vienna-London. Traffic can be interoffice and long distance. My main problem at the moment is that the outgoing traffic can not be characterised by the Erlang distribution since the number of sources is finite (less than 60 per site).