Abstract
Mobile Edge Computing is an emerging technology which enables low latency applications by providing computation at the network edge. As resources at the edge are limited, optimizing resource allocation in MEC is a key research question. The goal of this dissertation is to investigate how the sharing of unutilized computing resources amongst entities can lead to higher efficiency. In the first work we investigate virtual machine (VM) sharing across base stations, in light of the network demand pattern. We consider the joint VM placement and pricing problem across base stations to match demand and supply and maximize revenue. To make this coupled and combinatorial problem tractable, we decompose it. Following which, we propose a Markov approximation based placement algorithm, and propose optimal and incentive compatible auctions. We prove optimality and truthfulness guarantees for our algorithms. In the second work we introduce a novel sharing economy-inspired model, where a platform facilitates the sharing of computing resource quota among users. This helps tackle the partial wastage scenario arising from coarse-grained plans. Based on our model, we design and analyse two dynamic pricing mechanisms, which maximize the social welfare and profit respectively. We prove the optimality and convergence of these proposed mechanisms. In the third work we propose a priority pricing scheme with different service classes, to enable users of heterogeneous delay sensitivities to share a resource node. Specifically, delay sensitive users get served first, for a higher price. We derive in semi-closed form the optimal prices and propose a partial-knowledge based pricing mechanism, where no knowledge of individual users’ functions is required. At equilibrium, users have chosen the social-welfare optimal priority class and offloading frequency.