An air-to-air cooling architecture operates much like the central air system in your home, except that in a data center, cool air circulates through compute racks to absorb generated heat. Hot air then passes over coils containing cold refrigerant that absorbs the heat and cools the air for recirculation. The refrigerant transfers the absorbed heat outdoors, where it's released to the environment and re-cooled through compression, condensing and evaporation. This process can be summarized as a cycle of heat absorption, heat transfer, and heat release and cooling. To control airflow, racks are typically mounted on raised floors and arranged in back-to-back rows to form cold aisles and hot aisles.

As power levels rise, hot aisle containment prevents hot air from bleeding into and raising cold-aisle temperatures controlled by computer room AC units (CRACs). Computer room air handlers (CRAHs) manage airflow and the transfer of heat to refrigerant in the evaporator coils. Increased heat load requires larger blower fans and evaporator coils, along with increased server fan airflow. In the 20kW to 50kW rack power range (assuming that all racks in the computer room are similar), efficiency and operational demands require a different heat transfer method.