Here is the abstract you requested from the Thermal_2014 technical program page. This is the original abstract submitted by the author. Any changes to the technical content of the final manuscript published by IMAPS or the presentation that is given during the event is done by the author, not IMAPS.
|Dynamic Thermal Management of Data Centers with Hybrid Air-Liquid Cooling Systems|
|Keywords: Data Center, Hybrid Cooling System, Dynamic Cooling Control System|
|Server under-utilization and cooling over-provisioning are endemic in data center operations and account for many of the energy inefficiencies present in air-cooled data centers. Under-utilization in servers leads to inefficient use of power supplied to servers. Over-provisioning of cooling as a result of poorly controllable air cooling systems leads to excessive use of cooling that is manifested as electrical power consumption in chillers and fans. Ideally, efficient operation of data centers would start with smart allocation of IT load to maximize individual server utilization and proportioning of the IT load to specific spatial zones within the room that would allow localized zonal cooling. Furthermore, the use of nimble, dynamically responsive cooling systems would allow smart, on-demand cooling that would enable provisioning of cooling “when needed” and “where needed.” This presentation will describe research in the NSF ES2 center in the modeling and experimental characterization of dynamically responsive hybrid air-liquid cooling systems in rear door, in-row, and overhead configurations. We will discuss the development of mathematically detailed and mathematically compact dynamic models of heat exchangers, racks and servers and demonstrate newly developed techniques for embedding both types of models within a CFD simulation environment. Experiments designed to characterize the dynamic response of heat exchangers and servers for model validation will be described. Finally, the use of 2nd Law based Exergy destruction as the principal metric for identifying system inefficiencies and searching for optimal performance in dynamic cooling systems will be demonstrated.|
|Alfonso Ortega, Associate Vice President for Research and Graduate Programs