MOBILE EDGE COMPUTING NETWORK OPTIMIZATION

Smart mobile devices are becoming more and more important in every aspects of human life, and mobile application are becoming more and more resources demanding with a widening gap between the required resources and those available on mobile devices. To bridge this gap, Mobile Edge Computing paradigm has been introduced to bring IT applications, computational and storage resources to the periphery, or edges, of the cellular mobile network. Several complementary technologies have been presented to implement Mobile Edge Computing, all of them considering the deployment of virtualization facilities within mobile access and backhaul network. In this thesis we face several optimization problems related to the planning of a Mobile Edge Computing Network. In the first part of the thesis we present the Mobile Edge Computing Network Design Problem (MNDP), that considers the design of a full mobile access and backhaul network together with the location of the virtualization facilities. Two variants of the problem are considered: either assuming a static condition of the network or dynamic variations of traffic demands and the human mobility that causes these variations. Matheuristics are proposed to solve MNDP and best practices are drawn on real world data. In the second part of the thesis, we face a tactical side of the optimal Mobile Edge Computing network planning, that is the routing in time of access point traffic to specific Mobile Edge Computing facilities on a fixed network structure. We present exact Branch-and-Price algorithm to solve the problem, experimenting on real-world dataset. Finally, driven by the fact that the knowledge of mobile user mobility represents a key data for the MNDP, in the third part of the thesis we face the problem of estimating human mobility given very aggregated data, that is the network traffic demand variations in time. We propose mathematical programming formulation and column generation algorithm to solve this problem, experimenting on both real-world and synthetic datasets.