Traffic forecasts predict mobile video to be the most significant slice (66\%) of the world's mobile data by 2014. However, today's networks are designed primarily for pre-compressed and packetized data. For video, this assumption of data packets abstracts out one of the most relevant design problem - the lossy compression problem. Another key design aspect with video streaming over wireless is the need for interaction among layers of the protocol stack, including congestion control, routing, access and physical layer transmission. We generalize the traditional network utility maximization framework to include compression (rate-distortion) control, and show that this new framework provides a better understanding of the optimal operation of and the interaction among layers. 

Next, we focus on Gaussian interference networks to address the weighted-sum rate optimization (or capacity problem) that results from the utility maximization framework. Majority of existing information-theoretic results are specific to two-user networks. Three-or-more user networks are fundamentally different, and therefore, these results are inadequate. We establish one of the first non-trivial exact capacity results for Gaussian interference networks - ``the sum capacity of degraded networks in closed-form''. This exact capacity result is extremely important, but limited in scope to degraded networks. This motives us to consider the most practically relevant scenario of MIMO interference networks under channel uncertainty. Inspired by recent results on interference alignment, we consider linear transceivers and pose the resulting design as a robust weighted-sum rate optimization. We develop new provably convergent iterative algorithms for solving this optimization through ingenious convex sub-problem formulations.