Proposal for standard compact server model for transient data center simulations

• Published in: ASHRAE transactions Vol. 121; p. 413
• Main Authors: Pardey, Zachary M, Demetriou, Dustin W, Erden, Hamza Salih, VanGilder, James W, Khalifa, H Ezzat, Schmidt, Roger R
• Format: Journal Article
• Language: English
• Published: Atlanta American Society of Heating, Refrigerating, and Air-Conditioning Engineers, Inc. (ASHRAE) 01.01.2015; American Society of Heating, Refrigeration and Air Conditioning Engineers, Inc
• Subjects: Chilled water systems; Communities; Computational fluid dynamics; Computer simulation; Cooling; Data centers; Engineering models; Exhaust; File servers; Heat transfer; Mathematical models; Servers; Software; Temperature; Thermal properties
• ISSN: 0001-2505

Accurate prediction of data center temperatures and airflow following a transient event, such as the loss of cooling or variation in server load, requires that all sources of thermal mass in the data center he modeled with sufficient accuracy. One of the most important and simultaneously difficult-to-model types of thermal mass in the data center is that of the servers that populate the facility. Although compact models of servers for transient applications have recently been proposed and much experimental measurement of servers undertaken, no one model or technique for estimating required server properties has emerged as the de facto standard for computational fluid dynamics (CFD) users and vendors. This paper, a collaborative effort by industry and university researchers active in the subject field, reviews the progress in transient server modeling to date and recommends a single, standard compact server model to be adopted by the data center community. Techniques for determining the transient server properties thermal capacitance and thermal effectiveness from detailed measurements as well as simple correlations with server mass and size are discussed. Example server properties are provided based on measurements from two independent laboratories. Finally, the sensitivity of server exhaust temperature predictions to uncertainties in server properties is discussed.