Computational Study Of Data Center Hotspot Mitigation With Snorkels As An Alternative Rack-level Containment Strategy

Abstract

One of the major concerns of companies around the world is business continuity and with that comes the need to prevent operational dysfunction by keeping their information systems available and functional. IT professionals therefore, are constantly striving towards network sustainability by boosting the cooling cycle and infrastructure in data centers, while also reducing overall energy consumption. The combination of increasing IT workloads and packaging densities place stringent requirements on data center cooling systems and so, various traditional airflow management technologies have been implemented that propose significant energy savings and potential reduction in net annual cost for running data centers.In this work, a parametric study of a 1008 ft2 baseline data center with 161kW of IT load is used as an example for CFD analysis and the concept of using rack-level containment with snorkels has been explored as an alternative airflow management strategy. Snorkels minimize rack inlet temperatures by directing and aiming cool air from the sub-floor plenum to individual racks/cabinets, hence removing hotspots. Over 40 CFD cases were analyzed to compare different such strategies. Two different designs of snorkels, the flat-top and the angle-top were specifically examined to determine the design with the optimal performance. Test models include the baseline case, separate cases with hot-aisle and cold-aisle containment as well as separate cases with the two snorkel designs. Other models were built for different simultaneous combinations of cold aisle containment with the two snorkel designs for the purpose of comparison. A comparative metric based on industrial thermal guidelines as specified by ASHRAE was used to measure the efficacy of rack cooling at increasing supply temperatures for all the cases compared, and the snorkels especially ones with the flat-top design showed a significant inlet temperature modulation and improvement along the height of the racks, maintaining a rack cooling index-high of 100%. Other performance metrics like the return temperature index and supply heat index were also used to analyze and compare the models.In the later part of this work, it was needful to see the effectiveness of snorkels at elevated rack density. Heat load was increased from 4kW/rack to 10kW/rack for a selected row of cabinets, and cooling improvements starting from over 13% were achieved at the different set-points monitored by temperature sensors.These results present the feasibility of this strategy as a preliminary review and give an initial guideline on where this concept can be implemented, depending on the respective data center layout and design.