Edge-edge Collaboration Based Micro-service Deployment in Edge Computing Networks

With the sixth generation (6G) proposal, collaboration at the edge of the Internet of Things (IoT) has been widely studied to coordinate limited edge resources. Kubernetes has emerged as a promising solution for flexible and efficient resource scheduling. However, the default scheduler of Kubernetes only allocates pods separately according to the resource utilization condition of the cluster, which ignores the effect of the correlation between micro-services on latency. Under this circumstance, we propose a micro-service deployment strategy based on edge-edge collaboration, which takes the correlation between micro-services into account and models it as Service Function Chain (SFC), aiming to reduce the delay and balance the utilization rate in the edge cluster. Furthermore, we propose a model-free Distributed Deep Reinforcement Learning Deployment (DDRLD) algorithm to solve the multi-objective optimization problem. The master node trains the Q network and updates the parameters to the other nodes in the cluster, where each node can determine the deploying decision separately. Simulation results show that the proposed scheduling strategy can reduce user delay while ensuring the balance of the utilization rate.