Abstract

Traditional approaches to resource sharing make the assumption that the cooperating entities are able to exchange performance related information. This is not a valid assumption if the entities belong to different operators. In such a scenario, cooperation decisions have to be taken under uncertainty. We present our dynamic inter operator resource sharing framework, in which we formulate the decision problem as a Partially Observable Markov Decision Problem(PO-MDP). We provide an exact solution PO-MD and find the optimal borrowing threshold. We also propose a user-centric load measure that can be used independent of the underlying networking technology. We demonstrate our approach via a load balancing scenario for TCP traffic over 802.11g network.