In wireless cellular networks and wireless local area networks nonlinear network utility maximization need to be conducted over both user rates and transmit powers. For each of the three cases considered in this paper, we present an algorithm that converges to the jointly optimal pair of rate vector and power vector. For the simple case when data rates are not limited by interferences, for example in single-cell downlink transmissions. Algorithm 1 we propose is an iterative bidding mechanism between the base station and mobile users, where knowledge about channel conditions and individual user utility functions is only needed locally at each user but not needed at the base station. In the case when data rates are limited by interferences, the utility maximization problem is complicated both by nonlinear coupling between powers and rates, and by interference among powers. Through centralized iterative steps. Algorithm 2 we propose converges to a joint and global optimum over the solution space of rates and powers. We then consider end-to-end transmissions in cellular networks, which traverse both wireless fading channels and many hops of wired links shared by other traffic. There is a tradeoff between attaining air-interface capacity in the wireless hop and controlling congestion in the wired backbone wide area network. We formulate this end-to-end resource allocation problem in such hybrid networks, and present a solution to obtain the Pareto optimal tradeoff between attaining wireless multi-access fading channel capacity and maximizing global network utility.