The explosive growth of smartphones and other portable devices, along with new traffic types generated by M2M applications, are creating huge volumes of mobile data traffic and signaling overhead, therefore requiring a radical change to the current mobile network architecture. This has promoted new virtualization paradigms, which combine diverse packet core services, and provide network functions implemented in software, rather than in dedicated hardware appliances, in order to scale capacity and introduce new services in a fast and cost-effective way. In this paper1we study the optimization and resource allocation problems taking into account the deployment of virtualization structures. Our aim is to develop a theoretical framework of resource orchestration for mobile access networks, deriving the fundamental performance limits as well as the tradeoffs among the key system parameters. We therefore study optimal, time-varying placement and chaining of network functions. With respect to existing works, our optimization framework provides a much more precise system modeling, with, among others, a separation between control and data plane functions. We perform an extensive numerical analysis using both real traffic traces provided by a mobile operator (Vodafone UK) and real positions for radio access points for the UK area, and discuss the impact of network parameters on the system performance. Numerical results show that our proposed optimization framework permits to carefully model key aspects of network virtualization and service deployment/chaining in such scenarios, thus representing a very promising framework for the design of efficient and cost-effective mobile networks.