Predictive container orchestration in the cloud using artificial intelligence techniques

Predictive container orchestration in the cloud using artificial intelligence techniques

연구 분야: Software Development

학회: Computing

As cloud computing continues to evolve, efficient container orchestration has become critical for optimizing resource utilization and maintaining scalability. This paper addresses the challenges of container placement and autoscaling in cloud environments by proposing a predictive Artificial Intelligence based solution. We utilize a Bidirectional Long Short-Term Memory model to predict CPU and memory utilization of containerized microservices, achieving a Mean Squared Error of 0.00172 and R score of 0.7055. We then employ a Random Forest Regressor to predict required container replicas, resulting in an R score of 0.9248. Our experimental validation using the Alibaba Cluster Trace dataset confirms that our Bidirectional Long Short-Term Memory approach significantly outperforms traditional techniques and alternative models like Convolutional Neural Network and Temporal Convolutional Network for resource prediction. This integrated approach enables proactive resource management, reducing operational costs and improving application performance through more accurate autoscaling decisions.