Research On Task Offloading Decision In Multi-access Edge Computing

The use of cell phones and other terminals for face recognition or other complex computing tasks has been integrated into people's daily life.The frequent offloading of computing tasks on terminals often cannot be offloaded or takes a long time due to insufficient resources,which seriously affects user experience.The offloading of computing tasks based on cloud servers needs to be forwarded through the core network,and the offloading task delay and energy consumption are seriously affected by the network,and data security cannot be guaranteed.The offloading task method based on edge computing server studied in this paper,the computing task is not forwarded through the core network,and the offloading is completed on the edge server near the user side after permission confirmation and resource allocation,which strongly compensates for the shortcomings of cloud computing,and the specific work is as follows.For the demand of multi-task continuous offloading,an edge computing simulation system optimization framework is proposed to retain the minimal system in edge computing,discard the modules of billing and third-party compatibility,and optimize the framework according to the application scenario to obtain a low-cost and lightweight implementation of the simulation system centered on the decision algorithm.The paper focuses on computational task offloading algorithms.In the modeling and subsequent simulation experiments,instead of computational offloading for tasks in a particular scenario,the tasks in the actual scenario are abstracted as a set of specific information,such as the Qo S index of the task,the memory size occupied,the number of CPU clock cycles,and the maximum latency.For the problem of finding the optimal multi-task continuous offloading scheme,Shannon's formula is studied,a mathematical model of single-task offloading is established,the constraints of single-task offloading are analyzed,and an optimization model with the delay and offloading rate of multi-task offloading scheme as the target is established.For the multi-task single offload decision problem,the ant colony algorithm is optimized from three perspectives of taboo table,pheromone and heuristic factor,and the genetic algorithm is fused to speed up the algorithm iteration.Using the same task and computational resources for single offload,the simulation is compared with different decision algorithms.The experimental results show that the joint offload algorithm based on ant colony algorithm and genetic algorithm has better performance,and the offload time is less and the offload rate is higher compared with pure ant colony algorithm and genetic algorithm.The simulation system is built using Docker containers and other technologies,and the system is visualized using node.js.We conducted experiments with different decision algorithms in the simulation system to compare the offloading performance and benefits of the two algorithms,and concluded that with 10 times of decision data analysis,the optimized ant colony algorithm with "superiority and inferiority" behavior has better performance in the continuous offloading scenario,the waiting time is lower than the genetic algorithm by 0.08 seconds,and the offloading rate is 6% higher,and The higher the number of decisions,the higher the gain of the algorithm.