Research On Offloading Strategy Of Mobile Edge Computing Based On Multi-dimensional Attributes

Cloud computing enables smart devices to run more computing-intensive applications,but it incurs the problems in privacy protection and real-time application as well.Therefore,mobile edge computing technology emerges as the times require.By sinking computing resources,network-control functions,and caching functions to micro base stations and surrounding edge servers,mobile edge computing greatly reduces the time and system energy consumption for data round-trip processing.During the investigation,we found that most of current researches aim to reduce the system energy consumption or delay.Research on the optimization of centralized mobile edge computing services and communication architectures is relatively mature,while research on the optimization of distributed mobile edge computing services and communication architectures is ongoing.At the same time,most of current researches on computation offloading focus on one single factor.For example,they optimized the computation offloading process only considering Qo S(Quality of Service)with the assumption that these mobile users are stationary,which is non-meticulous in the optimization of global strategy.The above investigation inspired us to study the computation offloading strategy in combination with the multi-dimensional attributes of users.This paper designed a computation offloading strategy by considering the mobility,sociality and Quality of Service factors of users in mobile edge computing scenarios.Firstly,Kalman filtering algorithm was applied to predict trajectories of mobile users,then the social relationship parameters between mobile users and edge servers are calculated according to the historical connection relationship,and next the predicted results of trajectories and the social relationship parameters were applied in an Affinity Propagation clustering algorithm as input parameters to cluster edge servers.In this paper,two Qo S parameters were defined by considering the efficiency of the service provided by edge servers and the quality of channel links.Through the proposed AP based Sociality-Associate and Mobility-Aware Clustering Algorithm,a set of candidate edge servers were screened out,and then the target edge server with the highest Qo S benefits would be selected from the candidate edge servers to offload tasks.The results of the simulation experiments show that the proposed offloading strategy can enhance the data processing capability and reduce the computational delay in a powerconstrained network.In addition,this paper studied the computational offloading problem of taskdependent scenarios in mobile edge computing,where each application had its completion time limit and different tasks belonging to the same application had dependencies among them.With the applying of Graph Theory,the task scheduling decision problem was formulated as a DAG(Directed Acyclic Graph)optimization problem.To solve this problem,this paper divided it into two sub-problems.Firstly,we proposed an algorithm that can rank multiple applications and multiple tasks to obtain an efficient solution.Secondly,the problem was formulated as a potential game by jointly considering dependencies between tasks and the competition for resources among multiple devices,and a distributed computing offloading algorithm was designed to achieve a Nash equilibrium.The results of simulation experiments show that the proposed offloading strategy can effectively reduce the overall completion time of applications,and the computational cost of multiple applications,under their respective completion time limits.