Problems and progress in switching networks for terabit networks

Summary form only given. The author reviews research on ATM (asynchronous transfer mode) switch architectures by exploring one architecture in some detail and exploring how that architecture can be modified to support a large number of gigabit-per-second transmission links. He then turns to how one can address the various challenges posed by the need for continuing improvement in performance, in the context of public networks with hundreds of millions of users. He believes that as technology improves to the point that gigabit-per-second user channel rates become economical, new uses will arise that consume the available bandwidth and spur even greater demands. The key challenge for today´s network architects is not simply how to design the biggest and fastest networks possible, but how to design them in such a way that they do not become obsolete overnight. New networks should be structured so that anticipated technology improvements can be incorporated in an evolutionary way, allowing incremental performance improvements with no disruption in network operation.<>