Optimizing Resource Scheduling for Multi-Scenario Mixed Service Groups under Edge Cloud-Native Environments Using Simulation Learning

Abstract

The evolution of cloud and edge computing technologies has brought about resource management challenges. Traditional resource scheduling strategies fall short in dynamic cloud-edge environments, one of the challenges is identifying system state changes in multi-scenario edge cloud-native environments. The dynamic orchestration and deployment of container resources are crucial. To address this issue, we introduce a virtual environment, which generates interactions of multi-scenario mixed service groups. Furthermore, we proposed a multi-agent adversarial imitation learning approach, which is trained in the virtual environment. Experiments reveal that our approach, which is fully trained in the virtual mixed-service environment, results in no physical sampling costs and significantly outperforms traditional supervised approaches.