Dynamic Power Simulator Utilizing Computational Fluid Dynamics and Machine Learning for Proposing Task Allocation in a Data Center

摘要

A dynamic power simulator for a data center was demonstrated by combining computational fluid dynamics (CFD) and machine learning. The total power consumption of the data center was simulated. The sensitivity of the temperature distribution along the virtual machine (VM) allocation was analyzed using a non-parametric process for the CFD. An allocation of server tasks was proposed for reducing the power consumption of the air conditioner installed in the data center. This simulation showed that the optimum operating temperature increases with the power usage effectiveness. These results indicate that the power simulator developed in this study is a powerful tool for dynamic power simulation and for estimation of better operation parameters, including VM allocation, from the aspect of power consumption.