A Deep Reinforcement Learning Framework for Optimized Container Scheduling and Load Balancing

Abstract

Unlike VMs, containerization is a modern method for packaging and deploying software in distributed environments like the cloud. Containers are widely used due to their efficient software packaging and deployment. Efficient management of containers is crucial in dynamic cloud environments with heterogeneous infrastructure. Deep learning techniques are being applied to optimize resource utilization in cloud environments, including mapping containers to suitable nodes for energy conservation. However, the existing works on container scheduling have limitations like inability to cope with dynamic runtime scenarios. To overcome this problem, the aim of this paper is to design and implement a framework using deep reinforcement learning techniques to improve container scheduling and load balancing. The proposed algorithm, Reinforcement Learning based Container Scheduling (RLbCS), uses an action-reward iterative approach to optimize container scheduling. Experimental results showed that RLbCS outperformed existing methods, achieving a 92% success rate in placing containers and optimizing resource utilization. The proposed method can be integrated with cloud-based systems to automatically schedule containers for resource optimization and load balancing.