Minimizing Electricity Bills for Geographically Distributed Data Centers with Renewable and Cooling Aware Load Balancing

Abstract

Large Internet service providers usually operate multiple geographically distributed data centers to guarantee the quality of service for their worldwide users. The enormous power consumption of these data centers may lead to both huge electricity bills and considerable carbon emissions. To mitigate these problems, we propose a novel renewable and cooling aware load balancing policy GGLB-ARMA in this paper. First, GGLB-ARMA models the runtime cooling efficiency for each data center with their ambient temperatures. Then based on the ARMA models, it predicts the fine-grained future on-site renewables at the beginning of each scheduling period. Through these methods, GGLB-ARMA is finally able to solve the challenging total electricity bills minimization problem for data centers with on-site renewable energy plants. Evaluation results based on reallife traces show that GGLB-ARMA is able to reduce the total costs by as much as 12.58% compared with an hourly executed cooling aware GGLB while avoiding the costly frequent power off/on of servers.