How to Select Optimal Gateway in Multi-Domain Wireless Networks: Alternative Solutions without Learning

Abstract

Gateways are a crucial part of wireless networks. In multi-domain wireless networks, the existing solution to the problem of optimal gateway selection is based on distributed learning. While such a solution is interesting and useful, it has a fundamental limitation: the learning algorithm may stay for long time in a Nash Equilibrium that does not correspond to the optimal gateway selection. This stay can be so long that people can hardly wait for the convergence of the learning algorithm to the optimal gateway selection. In this paper, we present a systematic study of the gateway selection problem. We distinguish three cases and present an alternative solution to the problem in each of these cases: for public link costs, an algorithm to compute the optimal gateway selection directly; for two domains with private link costs, a cryptographic protocol; for three or more domains with private link costs, a perturbation-based protocol. In all the three cases, our solutions accurately compute the optimal gateway selection within reasonable amounts of time, although we emphasize that our solutions are NOT improvements to the gateway selection solutions based on distributed learning (mainly because our solutions are centralized instead of distributed).