Research On Computation Offloading And Code Configuration Selection In Mobile Edge Computing

The emergence of 5G and computationally intensive applications has driven the development of mobile edge computing(MEC)and computing offloading,among which the offloading decision-making and coding configuration selection of parallel computing tasks have attracted extensive attention from the academic community.In this thesis,we use evolutionary game theory(EGT)to study computing offloading and code configuration selection for partitionable tasks and coded tasks respectively.The main work includes:(1)An offloading scheme for partitionable tasks is proposed.There are few works considering both ultra-dense networks(UDN)and the partitionable tasks.However,in a multi-user competitive environment,different task offloading schemes will lead to different loads on edge servers,which will affect the completion delay of tasks.In addition,due to different computing capacity of edge servers,appropriate offloading schemes are needed to minimize the overall delay of MEC system.First,we formulate a computing offloading model for partitionable tasks with the goal of minimizing the task completion time,under the constraints of the size and arrival rate of tasks,the coverage of base stations and the computing and communication capacity of edge servers.Then,we construct a task offloading game using EGT,and prove the existence and stability of evolutionary equilibrium within populations,as well as the convergence among populations.We propose an offloading scheme based on EGT to solve the formulated problem.Extensive evaluation demonstrates that the proposed scheme can effectively reduce the overall system delay and converge to the evolutionary equilibrium.(2)A code configuration selection scheme for coded tasks is proposed.Computationally intensive tasks can be parallelized to reduce execution time,but suffer from stragglers.Coded distributed computing(CDC)methods can provide robustness against stragglers for offloading tasks by introducing redundant computations,thereby ensuring the efficiency of task execution.However,the computing capacity and straggler effects between edge servers are different and unknown to users,it requires users to explore edge server and code configuration to reduce the execution time of task.First,we construct a code configuration selection problem for coded tasks according to the M/M/1 queue and the shifted exponential distribution.Considering the different computing capacity and straggler parameter of edge servers,the goal of the problem is to minimize the overall delay of the system.Then,we construct a code configuration selection game using EGT,and the existence and uniqueness of the evolutionary equilibrium is proved using replicator dynamics.To solve the proposed problem,a code configuration selection scheme based on EGT is proposed.Extensive evaluation shows that the proposed scheme can effectively reduce the overall system delay and converge to the evolutionary equilibrium.