scope:

ABSTRACT

Data center energy conservation is becoming increasingly challenging due to the increasing number of servers, their increasing power consumption, and the reliance on IT services. To achieve sustainability goals of efficient data center operation, digital twin models - a real-time digital representation of the physical data center - are being leveraged for the optimization of server arrangement and the operation of the data centers. As the industry looks to make the next gains in data center energy efficiency, digital twin models will become integral to how control systems and smart building technologies understand and predict the data center performance. A pre-cursor to the development of a data center digital twin model, are accurate server-level models. These are critical to accomplish the well-represented data center digital twin. In this paper, the development of state-of-the-art server digital twin models for computational fluid dynamics are discussed to highlight the critical features that are needed to assure an accurate representation of commercial IT equipment. These features include the introduction of back-pressure curves of various servers, which have been experimentally obtained and applied into the detailed simulation models. Physical aspects of servers with two cooling domains have been tested individually and combined to understand the level of modeling fidelity needed. The impact of how the airflow exhausts from the server and the impact of these outlets to the simulation model are described. Lastly, accurate representations of the server fan operating points under various inlet temperatures and external impedance are shown to be of critical importance to enable detailed server digital twin models to be more efficient and accurate within a data center digital twin. The improved digital twin design can contribute to more accurate data center control and more efficient data center operation.