Characterization of Transient Air Temperatures in a Data Center Following Chiller Failure
Vikneshan Sundaralingam, Yogendra Joshi, Steven Isaacs and Pramod Kumar
The task of minimizing the downtime of a data center is becoming increasingly important due to the necessity of availability, and maintaining the integrity of the data being handled. However, the prediction of the dynamic response of air temperatures during a failure, which can be used for mitigation options such as equipment shutdown, is not an easy task for most facilities. The data center section considered consists of 10 racks of servers, with maximum capacity of 28 kW for each rack. The racks are cooled using cold air supplied through computer room air-conditioning (CRAC) units. Prior to discharge from each rack, the heated air is cooled using an air-to-water rear door heat exchanger (RDHX) utilizing chilled water circulated from the building chiller. Measured data are presented following two chiller pump failure scenarios in the data center zone of interest. The data include CRAC supply and return air temperatures, chiller supply and return water temperature, chilled water flow rate, total server power draw, and CRAC fan speeds. The first failure event was unregulated, during which the data center can be viewed as a control mass. The second failure event was regulated manually using various methods, during which the data center can be perceived as a control volume. As a secondary objective, the paper presents a simple analysis based on overall energy balance to provide estimate of an average temperature rise based on an estimated thermal mass of the system following the first cooling system failure.
Keywords: Thermal Mass, Data Center, Lumped System Analysis, Chiller Pump Failure, Cooling Failure.