Container Scheduling Strategy Based on Container Clustering and Deep Reinforcement Learning

The microservice architecture has accelerated the development of containerization in cloud computing. In Kubernetes, frequent communication among a large number of microservices generates significant east-west traffic. Traditional scheduling strategies primarily focus on resource load balancing while neglecting the service latency caused by cross-node communication. To address this issue, this paper proposes a container group scheduling method that integrates spectral clustering with deep reinforcement learning. Specifically, spectral clustering is first used to group correlated containers, thereby reducing communication overhead. Then, a DQN-based algorithm is employed to optimize the scheduling process and achieve balanced resource utilization. Experimental results demonstrate that the proposed method effectively balances both load distribution and communication cost, outperforming traditional scheduling strategies.