Sustainable Data Center Energy Management Through Server Workload Allocation Optimization and HVAC System

The paper presents a new approach to reducing power consumption in data centers by optimizing server workload allocation and considering the impact on cooling impact from air conditioning (HVAC) systems, such as server static pressure difference and ambient temperature. We built highly accurate generic server power prediction models by investing in more than 20 algorithms. We proposed a workload allocation optimization (WAO) algorithm that extends Kubernetes API and uses it to evaluate energy efficiency in a distributed computing environment. Unlike the simulation studies, this research was conducted in a test-bed data center with around 200 on-premises servers featuring four different CPUs to validate the proposed solutions. The results of the experiments indicate that the optimized workload allocation with HVAC systems can lead to potential energy savings of approximately 51.5% and 25.7% on high-performance CPUs, such as Intel Xeon Gold and AMD's ROME, respectively. The study highlights the importance of optimizing workload allocation with HVAC systems to reduce data center energy consumption and promote sustainable computing practices.