Separation of user (data) plane from control plane in networks helps scale resources independently, increase the quality of service and facilitate autonomy. In ad-hoc networks, plane-separation through clustering introduces a hierarchy where control functions can be carried out by some designated cluster heads and other nodes perform according to the decisions of such functions. Therefore, clustered topologies can be considered as a natural consequence of the control and user plane separation (CUPS). Moreover, hierarchical routing protocols, which are constructed upon those clustered topologies, enable the use of CUPS architecture for the end-to-end communication. However, there is no rule of thumb to apply clustering algorithms that are directly dependent to network characteristics, and the routing protocols designed for clustered topologies cannot effectively utilize CUPS since they neglect the role of the nodes in the data plane. This study investigates the application of CUPS architecture in ad-hoc networks by considering clustering and routing protocols holistically. First, the applicability of the clustering techniques for different purposes: stability, energy efficiency and reliability is discussed; and Dependability-based Clustering Algorithm (DCA) is proposed. DCA is a dynamic clustering algorithm which exploits a cross-layer architecture. Its parameters are analyzed and optimized using the sensitivity analysis technique, Moment-Independent Delta Analysis. Then, the hierarchical routing protocol CUPS-based Hierarchical Routing Algorithm (CHRA) is proposed. In CHRA, the separate functions of the control and data planes are explicitly defined to provide the quality of service and energy-efficiency. The overall CUPS-centric framework including DCA and CHRA is implemented in discrete event-based simulator, OMNeT++. The results show that DCA outperforms its rivals when it is optimized for different scenarios. Besides, the study reveals the significant points that need to be considered for designing similar clustering algorithms with the discussion of such optimization process. Finally, CHRA offers a better quality of service and a fair energy consumption thanks to its novel approach that considers the active use of the data plane as well as the control plane.