基于DQDG与DA-DDPG的容器分组调度算法研究OA

We address the container scheduling challenge in edge computing,where low-latency online tasks and delay-tolerant offline tasks with complex inter-container dependencies require mixed deployment.To overcome the lack of quantitative dependency assessment and low fragmented resource utilization,we propose an optimized strategy based on the Dependency Quantification and Grouping Algorithm(DQDG)and the Dependency-Aware Deep Deterministic Policy Gradient(DA-DDPG).the proposed method begins by constructing a multi-metric model to precisely quantify dependency strength.The DQDG then groups strongly interdependent containers to minimize cross-node communication.Subsequently,the DA-DDPG enhances the experience replay and policy network by incorporating dependency information for long-term optimal scheduling.Experimental results confirm that the proposed method ensures online service quality while significantly improving cluster load balancing and fragmented resource utilization,offering an end-to-end solution for deploying heterogeneous tasks with complex dependencies in edge environments.