A Distributed Microservice Scheduling Optimization Method

In the current Internet of Things environment, factors affecting scheduling are gradually increasing, and service performance is affected. There may be service delays, service non-response, and service feedback error messages. At the same time, service scheduling needs to be continued. Therefore, the scheduling The algorithm puts forward higher requirements. This paper proposes a scheduling algorithm based on depth-determined policy gradient (DDPG). By combining deep learning and reinforcement learning, it solves the problem of continuous action control and constructs a policy network and a value network. Double network structure to solve the problem of slow algorithm convergence. Secondly, this paper designs a microservice-based visual monitoring system to obtain the experimental data required by the algorithm and verify the effectiveness of the algorithm. Finally, 30 sets of data collected by the monitoring system are used to test and analyze the resource balance and service delay on the node in combination with the constraints of service and data dependency, and compare the delay and resource utilization with three typical algorithms. Experiments show that DDPG can show better performance under various constraints of the complex environment of the Internet of Things.