Broadband Traac Modeling: Simple Solutions to Hard Problems

A much clearer picture of the progress towards an integrated high speed multi-service network is now emerging. Such networks were anticipated over twenty years ago, at a time when packet switching was just another way to transmit data. Now the technology is so mature that media barons are organising their investments in order to take advantage of the proot. Many of the technical problems are now solved and the fundamental protocols required for these networks are suuciently well-deened to support a rapidly expanding industry. However, consensus on how to talk about the statistics of the data which is transmitted around these networks has not been readily forthcoming. Nevertheless, there now exists a family of models with suucient richness to describe real traac fairly well, which can be parametrised conveniently, and which degenerates to a readily analysable Gaussian model in the situation of very large networks. This analysis leads to important architectural conclusions which accord with common sense, in particular the conclusion that integrated networks of the future should be able to provide better and better service with eeciency tending toward perfect. This is a rather happy conclusion which thoroughly connrms the current rapid drive towards an all-encompassing integrated multiservice network. Rather than the increase in traac and the diversity of service types leading to greater and greater complexity, it seems that the ow of traac in our networks may become steadily more manageable.