PCNN-BCMO: A Novel Deep Learning Espoused Task Scheduling Approach for Enhancing Energy Efficiency of Cloud Data Centers

Abstract

The rapid development of cloud computing and data centers has intensified the need for efficient task scheduling to enhance energy efficiency and resource utilization. This study presents a novel task scheduling approach leveraging a part-based convolutional neural network (PCNN) optimized with balancing composite motion optimization (BCMO) to address these challenges. Historical data from cloud data centers is pre-processed utilizing guided box filtering (GBF) to eliminate noise, allowing the PCNN to accurately predict and schedule tasks. The BCMO optimization method fine-tunes the PCNN's weight parameters, ensuring effective scheduling while minimizing power consumption. Implemented in JAVA, the proposed method achieves remarkable accuracies of 99.67% and 99.78% on NASA and Saskatchewan HTTP traces benchmark datasets, respectively, outperforming existing models.