Research On Computation Offloading Based On Dynamic Resource Allocation In Mobile Edge Computing

With the rapid development of the mobile Internet of Things,mobile devices have become more and more popular,and at the same time a large number of computationintensive applications have also emerged.These applications generally require a large amount of computation and generate high energy consumption,but mobile devices have limited computing capability and battery capacity.The conflict between the two cases above has driven the development of mobile edge computing(MEC).Mobile edge computing deploys cloud servers at the edge of the wireless access network,which are physically located near mobile device users to provide users with cloud computing resources.Therefore,users can offload computation tasks to nearby MEC servers for execution,to reduce the response time and energy consumption of their computation tasks.However,on the problem of computation offloading,most of the researches have considered that the communication resources and computing resources allocated to each offloading user remain unchanged during a computation offloading period,that is,the static allocation of resources,leading to the waste of constrained resources and affecting the performance of MEC computation offloading.In order to solve the above problems,we propose a solution for the dynamic allocation of wireless bandwidth resources and computing resources during the task computation offloading process,that is,reallocating the bandwidth resources of transmission-finished tasks to transmission-unfinished tasks,and computing resources of computation-finished tasks to computation-unfinished tasks,thereby improving resource utilization.First,we analyzed the computation offloading problem based on the dynamic allocation of communication resources and computing resources in the MEC network scenario.We established the communication model,computation model and cost model,respectively,and proposed a game model for computation offloading.The goal is to minimize the computation cost of each user in the system.Then,by showing that the computation offloading game is a potential game,we have proved the existence of the Nash equilibrium of the computation offloading game model.Next,we propose a computation offloading algorithm(ECO-MEC algorithm)based on the game model.After a limited number of iterations,the algorithm finally reaches the Nash equilibrium of the computation offloading game.We analyze the convergence of the ECO-MEC algorithm and derive the upper bound of the number of slots when the ECO-MEC algorithm converges.Further,in terms of the number of beneficial cloud computing users and the system-wide computation cost,we use PoA to evaluate the performance of the ECO-MEC algorithm,and derive the upper and lower bounds of PoA.Finally,we set up simulation scenarios and conduct simulation experiments.We first perform experiments on the convergence of the ECO-MEC algorithm.The experimental results verify the convergence of the ECO-MEC algorithm,and show that the ECO-MEC algorithm can converge to the Nash equilibrium of the computation offloading game in a limited number of slots.Then,simulation experiments are performed for different lengths of the task data size range,different lengths of CPU cycles' range for computation,and different number of users.The ECO-MEC algorithm is compared with several other algorithms.The experimental results show that: ECOMEC algorithm has better performance in terms of the number of beneficial cloud computing users and the system-wide computation cost.For example,compared with the JPBR algorithm,in terms of the number of beneficial cloud computing users,the ECO-MEC algorithm can achieve 30.8% performance improvement.In terms of the system-wide computation cost,the ECO-MEC algorithm can achieve 34.1% cost reduction.