Performance of resource allocation in wireless networks is strongly tied to
the amount and the accuracy of relevant network state information available to
the network controller. When the desired information is perfectly known
without any cost to the controller, then it is known that decisions can be
performed to guarantee optimal long-term throughput performance. However, in
practice, there exists a cost associated with obtaining -possibly imperfect-
network state information. We show that, ignoring this cost leads to a
substantial degradation in performance in a number of wireless networking
scenarios. In the mere point-to-point communication, there has been a better
understanding on the mechanisms and the cost of obtaining information via
channel estimation and the fundanmental limitations of channels with limited
side information have studied extensively. However, in a network setting, we
illustrate that the ideas for the single link scenario do not generalize due
to a complex interplay that arises between estimation and resource allocation.
We aim to move beyond such limitations and enable practical resource allocation
by accounting for the intricacies of obtaining and utilizing possibly imperfect
network state information.