Optimal layered architectures of wireless networks

A general wireless networking problem is formulated to jointly optimize end-to-end user rates, routes, link capacities, transmitted power, and power allocation across subcarriers and fading states. It has been recently shown that such general wireless networking problems in the presence of fading, while non-convex, have zero Lagrangian duality gap. In this paper the optimality of two different layered architectures is derived as a consequence of this result.