Investigation of induced CRAH bypass for air-cooled data centers using computational fluid dynamics

Abstract

Previous studies investigated energy implications of computer room air handling (CRAH) unit bypass configuration and indicated significant energy saving potential for enclosed aisle data centers. These results were based on the thermodynamic modeling of the data center both at the room and cooling infrastructure levels coupled with the air flow network modeling (FNM) at room level. The contained structure of the enclosed aisle data centers allows FNM tools to have decent performance in predicting key air flow rates and fan power use in these configurations. However, any thermodynamic modeling tool to assess cooling energy consumption by the air-cooled data centers with CRAH bypass configuration needs to have more detailed quantitative information about the non-uniform temperature distribution among servers to ensure that none of the servers exceed recommended temperature limits for continuous operation. Furthermore, FNM cannot resolve complex air flow patterns in open aisle data centers, where some of the hot air recirculates back into the cold aisle and some of the cold air in the cold aisle bypasses racks and leaks into the room. We need more sophisticated tools such as computational fluid dynamics (CFD) to capture temperature non-uniformities at the rack surface; so that energy simulation tools utilize this information to assess the feasibility of CRAH bypass configuration primarily as an energy saving retrofit for air-cooled data centers. Here, we conduct a series of CFD simulations to extract critical information about the flow pattern and temperature non-uniformities of a specific data center layout with 1 MW IT load operating within a range of CRAH bypass fractions.