The development of simplified rack boundary conditions for numerical data center models

As cloud computing and computational needs grow, data centers will continue to become a larger part of our energy load. Proper design and layout are crucial to efficient energy use in data centers. Modeling the rack is one of the critical pieces in the design process. Often this is done as a black box rather than modeling the rack in detail. Modeling a computer rack as a black box has been done in numerous data center studies, but rarely has it been validated against experimental temperature and velocity data. This study looks at two simplified rack models and compares them against a rack composed of four 10U server simulators. The first model is an open box model that has a heating and fan plate and allows air to flow through the rack. The second model is a black box model that allows no flow through the rack and imposes a constant pressure boundary across the inlet and exhaust. The black box model adds the rack load from the upwind cells at the rack inlet plate to generate the exhaust temperature profile. Both models were tested across a range of airflows and rack loads. Agreements were found to be within 5.4°F (3°C) and 40 fpm (0.2 m/s), on average, over all experiments. An interesting finding of this study was the importance of correctly capturing the boundary conditions at the perforated floor tile. Modeling the perforated floor tile as a nozzle diffuser using the momentum method described in ASHRAE RP-1009 was found to produce acceptable results for airflow from the perforated floor tile.