Edge–cloud collaborative predictive auto-scaling for industrial IoT: A multi-objective optimization approach considering equipment health status

Abstract

This paper presents an innovative edge–cloud collaborative predictive auto-scaling framework for Industrial Internet of Things (IIoT) environments, specifically addressing resource management challenges in equipment health monitoring and predictive maintenance scenarios. Traditional autoscaling approaches often fail to consider the equipment’s health status and its impact on resource demands, leading to suboptimal resource allocation and potential equipment risks. We propose a three-tier framework that integrates equipment health monitoring, workload prediction, and multi-objective optimization. First, we develop a novel deep learning-based workload prediction model incorporating equipment degradation indicators to accurately forecast resource demands. Second, we formulate a multi-objective optimization problem that simultaneously considers resource utilization, energy consumption, and equipment health risk. Third, we design an adaptive edge–cloud collaboration mechanism that dynamically adjusts resource allocation based on immediate equipment health status and predicted maintenance requirements. Through extensive experiments using real-world data from multiple manufacturing facilities, our approach demonstrates significant improvements over the baseline methods: 25% reduction in energy consumption, 30% increase in resource utilization, and 20% decrease in equipment health risk ($p<0.05$). Furthermore, the framework shows robust performance under various industrial scenarios, including sudden equipment degradation and maintenance events. These results validate the effectiveness of our approach in managing IIoT resources while maintaining equipment reliability.