A Collaborative Migration Algorithm for Edge Services Based on Evolutionary Reinforcement Learning

Multi-access edge computing (MEC) enables users’ smart devices to execute computing-intensive and delay-sensitive applications by sinking computing power to edge servers, thereby meeting users’ quality of service requirements. However, due to the limited computing and storage resources on the edge server, it is impossible to migrate all user service requests to the edge server. At the same time, due to the heterogeneity of resources among edge servers and the uneven distribution of user service requirements, it can easily lead to unbalanced loads among edge servers in the edge system. Consequently, this results in low resource utilization and a decreased success rate of requests. Therefore, this paper builds a collaborative edge service migration model based on software-defined networking (SDN), which supports cloud computing centers and edge servers to collaboratively process user requests and service request migration between edge servers. Taking minimizing the response delay of user requests and the weighted sum of device energy consumption as the optimization goal and transforming the optimization problem into a Markov process. A collaborative edge service migration algorithm based on evolutionary reinforcement learning (ERL) is proposed to solve the service Migration strategies and resource allocation decisions. Experimental results show that the proposed algorithm (DEDRL) performs better than other algorithms in response delay, energy consumption and request success rate.