An Adaptive Predictive Threshold (APT) Algorithm for Load Balancing in Cloud Environments

Cloud computing environments necessitate effective load balancing strategies to manage the growing volume of high-demand applications. This paper introduces an Adaptive Predictive Threshold (APT) Load Balancing Algorithm aimed at improving performance and energy efficiency through smart decision-making. The APT algorithm utilizes two thresholds that are updated dynamically-an adaptive overload threshold and a dynamic underload threshold-to precisely identify overloaded and underloaded Virtual Machines (VMs). Additionally, a predictive assessment enhances the mechanism by anticipating future VM loads, thereby averting sudden overloading and ensuring consistent task assignments. In contrast to traditional static algorithms, APT implements a consolidation strategy where an overloaded VM must finish a designated number of tasks before shifting to an underloaded state. Experimental findings indicate that the APT algorithm significantly surpasses commonly used methods such as Round Robin, First Come First Serve, and Equally Spread Load Balancing, especially under substantial task loads. Its adaptive, prediction-based methodology results in decreased instability, enhanced resource utilization, and improved overall system performance in practical cloud environments.