Here is the abstract you requested from the thermal_2015 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.
|Experimental and Numerical Analysis for Dynamic Airflow Provisioning at the Rack|
|Keywords: Airflow management, Thermal management, Dynamic provisioning|
|Data center cooling design generally focuses on room, row and rack-based cooling to provide raw cooling capacity and effective airflow distribution. The air conditioning systems are utilized to provide the effective bulk cooling and distribute the air through the Information Technology (IT) equipment at the rack. The aforementioned cooling designs enable different ways in provisioning conditioned air at the rack inlet. In traditional and free air cooling there is an opportunity to make significant cost savings by streamlining the entire airflow path in a Data center. The key aspect missing in all the airflow management schemes being utilized at an operational data center, colocation or company-owned, is real-time monitoring of conditioned air being provisioned at the server or rack. The variabilities of IT loads, rack inlet temperatures, pressures and volume of conditioned air across the rack, spatially and in time, are considered. An ideal data center would breathe air through the rack while constantly provisioning the varying heat loads and monitoring its functional health. This presentation explores the impact of server designs, operating at randomized IT utilization patterns, and airflow management practices on cooling efficiencies at the rack level. The degree of increase in volumetric flow rates due to raising the rack inlet temperature is established experimentally. The interdependencies of aisle pressurization and fan systems are demonstrated through experiments. The impact of variations in airflow provisioning at the rack for a modular and a raised floor data center is also investigated. Finally, the degree of inefficiencies incurred due to the stated variabilities in provisioning the rack is presented. The results shall provide a better understanding of airflow provisioning at the rack and enable dynamic thermal management.|
|Ashwin Siddarth, Gratuate student
University of Texas at Arlington