Research And System Construction Of Task Offloading Algorithm For Cloud-edge Computing

With the continuous increase in the number of user terminal devices,the limitations of cloud computing and edge computing in processing delay-sensitive and computing-intensive tasks can no longer meet the needs of users.In order to fully exploit the respective advantages of both,cloud-edge computing has emerged.In the field of cloud-edge computing,how to determine the offload location of each task and allocate reasonable computing resources to it has always been a very popular research direction.On the one hand,when studying task offloading algorithms for coarse-grained applications,it is necessary to consider the computing time,transmission time,and waiting time of tasks to prevent task timeouts.On the other hand,the network topology between tasks in fine-grained applications constitutes a directed acyclic graph,and the completion time of a single task may have a significant impact on the generation time of other tasks.In addition,we need to allocate the right amount of computational resources to each running task so that all tasks are processed quickly or so that the computation time of the task is balanced against the energy consumption of the individual servers in the system.Aiming at the above two different types of applications,we study the task offloading problem in the cloud-edge computing environment.The main work is as follows:(1)A multivariate particle swarm optimization algorithm is proposed for the task offloading problem of multiple coarse-grained applications.This method regards all computing tasks in the environment as a particle,and builds a population containing multiple particles,then assigns each particle a different initial speed and position,and later determines their calculation according to the priority information of the tasks in the particle order.The offloading position of each task is finally determined through the interaction between particles in multiple rounds of iterations.The experimental results show that,compared with particle swarm optimization and simulated annealing algorithm,the algorithm not only has lower time complexity,but also has better effect in reducing the total weighted average computing time of the task.(2)Aiming at the task offloading problem in fine-grained applications,a task offloading algorithm for multi-decision energy efficiency optimization is proposed.First,tasks are divided into different levels according to the network topology relationship between tasks in fine-grained applications.The offloadable location and server deployment are constructed to build a Markov decision process model,the task offloading process is then divided into multiple time periods for processing,and appropriate computing resources are allocated to the running task,finally the value iteration method is used to solve the problem.Compared with online dispatching algorithm,first come first serve algorithm,short job first algorithm,this method not only enables the application to finish running in the shortest time,but also balances the computing time and energy consumption of the servers in the system.(3)A cloud-edge computing task offload monitoring system is constructed.The front-end interface is responsible for receiving various data transmitted by the user,and displaying the offloading of the task in real time according to the selected algorithm.The interactive middleware is responsible for the communication function between the front-end and the server-side.The serverside processes the tasks according to the received information.The system is not only suitable for various types of task offloading algorithms,which simplifies the user deployment time and ensures the smooth execution of the task offloading process,but also provides result feedback through a visual interface,which is convenient for users to monitor in real time,and has high application value.