Current advances in wireless communication are driven by an increased demand for more data and bandwidth, mainly due to the development of new mobile platforms and applications. Ever since then the network operators are overwhelmed by the rapid increase in mobile data traffic, which is primarily fueled by the viewing of data-intensive content. In addition, according to the statistics, the ratio of downlink and uplink data traffic demands have changed drastically over the past decade and they are increasingly asymmetric even over small time periods. In recent years, different solutions, based on topological and architectural innovations of the conventional cellular networks, have been proposed to address the issues related to the increasing data requirements and uplink/downlink traffic asymmetries. The most trivial solution is to scale the network capacity through network densification, i.e., by bringing the network nodes closer to each other through efficient spectrum sharing techniques. The resulting dense networks, also known as heterogeneous networks, can address the growing need for capacity, coverage, and uplink/downlink traffic flexibility in wireless networks by deploying numerous low power base stations overlaying the existing macro cellular coverage. However, there is a need to analyze the interplay of different network processes in this context, since, it has not been studied in detail due to complex user dynamics and interference patterns, which are known to present difficulties in their design and performance evaluation under conventional heterogeneous networks. It is expected that by centralizing some of the network processes common to different network nodes in a heterogeneous network, such as coordination between multiple nodes, it will be easier to achieve significant performance gains. In this thesis, we aim at centralizing the control of the underlying network processes through Centralized Radio Access Networks (C-RAN), to deal with the high data requirements along with the asymmetric traffic demands. We analyze both large-scale centralized solutions and the light-weight distributed variants to obtain practical insights on how to design and operate future heterogeneous networks.