Research On Distributed Task Offloading Algorithm In Mobile Edge Computing

With the major breakthroughs in Io T technology and 5G technology,more and more emerging applications have put forward higher requirements on the computing power and energy consumption of user equipment.Compared with traditional cloud computing,mobile edge computing(MEC)has unique advantages in ensuring the service quality of mobile devices and reducing energy consumption.Task offloading is an important content that affects the effect of MEC.In recent years,researchers have carried out a lot of research works on task offloading of MEC.User mobility and computational task randomness are inherent features of many MEC applications;however,many research efforts devoted to task offloading tend to ignore these two highly dynamic behaviors,which pose significant challenges for achieving reliable computation.In addition,compared with centralized methods,which cause a lot of decision-making overhead,distributed methods do not require global information and are more suitable for dynamically changing network topologies.To fill these gaps,the specific research work of this paper is as follows:1)For the user mobility problem,this paper proposes a distributed task offloading algorithm in MEC.When users move,in order to ensure the continuity of service,computing tasks need to be migrated from one edge server to another edge server.In this paper,with the goal of minimizing energy consumption,the task offloading problem is modeled as a combinatorial optimization problem under the constraints of limited computing power and delay.The complexity of solving this problem will increase exponentially with the number of edge servers and users.For this reason,this paper uses the Markov approximation technique to design a distributed algorithm to solve it.First use the Log-Sum-Exp function to approximate the objective function,and then,the user's mobility problem is transformed into the state transition problem by constructing a Markov chain that can be realized in a distributed manner with a steady-state distribution.Thus,when the Markov chain converges,an approximate solution to the problem is obtained.The simulation results show that the distributed task offloading algorithm proposed in this paper can quickly converge and obtain performance close to the optimal energy consumption.2)For the task randomness problem,this paper further extends the static Markov chain to a new dynamic scenario.In this scenario,the number of active users in the MEC system changes over time as new computing task arriving or old task completing.The problem of task randomness brings great challenges to the design of Markov chain.To this end,this paper assumes that the MEC system allows jumping from one state to another state only when a new computing task arrives or an old computing task completes.Based on this,this paper also designs a dynamic Markov chain that can be implemented in a distributed manner and has a steady-state distribution to study the problem of task randomness.Simulation results show that the dynamic distributed task offloading algorithm can quickly converge,and only has a small loss bound.