Dependency-aware Task Offloading In Mobile Edge Computing

The growing popularity of mobile devices and deep learning technique have been creating explosive growth of realtime applications such as face recognition,on-line gaming,and AR/VR services.However,the resource-limited mobile devices usually cannot afford to execute the computation intensive tasks efficiently without degrading users' real-time experience.The centralized Mobile Cloud Computing(MCC)paradigm is not an appropriate solution as its centrality execution requires a long-distance data transmission from users to the remote cloud,which will lead to severe back-haul congestion.Pushing computation workload to the network edge,Mobile Edge Computing(MEC)has been proposed as a new paradigm to meet the latency requirement of applications and save energy for mobile devices.Mobile users can offload the computing tasks to the nearby edge servers(deployed at the base station or access point).After executing the offloaded tasks,MEC servers will feed the results back to the mobile users.In this work,we consider a multi-user and multi-server MEC system with taskdependent applications.We study the global optimization problem and multi-user gaming problem under both continuous time and discrete time situations.First,we study the global optimization problem,aiming to improve the QOE(quality of experience)of all users by task offloading and scheduling.After the transformation and reconstruction of the problem,the global optimal solution is obtained and evaluated in the simulation.Based on the global optimization problem,considering the selfishness of users in real network,we study the multi-user gaming problem in which each user makes the offloading decision independently.In the continuous and discrete-time scenarios,through the modeling and simulation of the game problem,we study the behavior of each user and the evolution process of the network.The performance of the gaming problem is compared with centralized optimization problem.To sum up,this paper mainly studies the global optimization problem and multi-user gaming problem in the multi-user multi-server scenario,and obtains the optimal solution of the global optimization problem and the behavior of the multi-user game problem,which provides theoretical support for the follow-up research of task offloading in mobile edge computing.