Intelligent Virtual Machine Scheduling Based on CPU Temperature-Involved Server Load Model

作者：Huan Zhou, Jiebei Zhu, Binbin Chen, Lujie Yu, Heyu Luo · 发表于：Energies · 年份：2025 · DOI：10.3390/en18143611 · 研究领域：Cloud Computing and Resource Management、Advanced Technology in Applications、Distributed and Parallel Computing Systems

To reduce the significant energy consumption in data centers, virtual machine scheduling optimization and server consolidation are deployed. However, existing server power load (SPL) models typically adopt linear approximations for model developments, which results in inaccuracy with actual SPL characteristics, hindering the optimal solution of virtual machine scheduling. Therefore, intelligent virtual machine scheduling (IVMS) is proposed based on a CPU temperature-involved server load model for data center energy conservation. The IVMS establishes a novel server power load model considering the influence of CPU temperature to capture the actual server load characteristics. Based on the model, the Q-learning method is utilized to solve the problem with the advantage of global optimization to obtain the scheduling solution that further improves calculation accuracy. The performance of the proposed IVMS is evaluated and compared to existing methods by both simulation and experiments in data centers, proving that the IVMS can better predict SPL characteristics and further reduce server energy consumption.