Subventions et des contributions :

Subvention ou bourse octroyée s'appliquant à plus d'un exercice financier. (2017-2018 à 2022-2023)

Cloud computing for mobile devices offers the opportunity of performing complex computational tasks on a mobile device by offloading them to more powerful servers (the cloud) through wireless network connections. The objectives of cloud computing system designers include energy efficiency (at all systems components: the wireless device, the wireless channel and infrastructure, and the cloud servers), latencies and execution times, user utility and/or satisfaction, fairness in resources utilization, etc. In this proposal, the main concern is energy efficiency at all system levels, and its effects on other objectives (such as latency minimization or fairness).

Although energy use has probably been the primary objective in mobile cloud computing, due to the short life­time of a mobile device battery charge, its study has usually been done under certain idealized assumptions that facilitate analysis. We propose to study mobile cloud offloading (MCO) under realistic conditions that include (a) a competitive environment with selfish agents (mobile users, server providers, infrastructure operators, etc.) trying to optimize their utility, (b) the existence of hard constraints such as task deadlines, and (c) the placement and provisioning of mobile cloud computing infrastructure, such as cloudlets and Road Side Units for vehicular wireless communications. The offloading of interdependent tasks by several mobile users leads to competitive behaviour and policy decisions at several system bottlenecks, such as the wireless channels, the Access Points, or at cloud servers. Getting rid of central coordination and moving the decision responsibility from the infrastructure designer/operator to the individual user under the realistic constraints mentioned above is a central theme of this proposal. In MCO, a mobile user may decide not to offload its task to the cloud (e.g., due to task deadline violations), while another user who can meet all hard constraints may decide to offload, since that would increase its utility by using system energy to execute its task. On the side of cloud service providers, selfish behaviour is displayed in an effort to maximize profits and minimize energy consumption, but so does cooperation (in the game-theoretic sense), if it can lead to better services and larger mobile users participation, and, therefore, better provider utility. The delegation of decision responsibility to individual agents leads to suboptimal utilization of system resources. This proposal is focused on the study of system inefficiencies due to selfish behaviour, the alleviation of their effect through the design of scheduling and pricing policies, as well as the design and analysis of efficient algorithms for the building and provisioning of the infrastructure needed for mobile cloud computing.