Multi-server Cooperative Offloading Strategy for Dependent Tasks Based on Improved Genetic Algorithm

Mobile edge computing (MEC) is an effective means to solve the problem of insufficient processor computation capability and limited battery power, when edge devices are processing computationally intensive and time-sensitive application tasks. Offloading strategy is one of the key technologies of MEC. Current task offloading strategies often lead to problems of low server utilization rate and high delay in the offloading process, because they overlook internal dependency of tasks and time-varying wireless channels. For this problem, a multi-server cooperative offloading strategy based on improved genetic algorithm is designed and implemented for multi-user with multi-server small cell network system scenario. This scenario jointly considers the internal dependency of tasks, time-varying wireless channels and combines with the heterogeneous computation capability of edge devices and servers. The goal of our paper is to find the best task offloading scheme with low latency and high server utilization. The improvements to the genetic algorithm mainly include the crossover and mutation process. Our paper uses the normal distribution crossover operator and the polynomial mutation operator to improve the genetic algorithm, so that it will not fall into the local optimum. The convergence speed is fast and the search space is wider. The simulation results verify that the proposed method can greatly reduce the user’s execution delay and increase the utilization rate of the server.