ASHRAE Journal - August 2009 - 49

exit air temperature from the servers. The extent of such a reduction in the fan performance depends on the individual design of the racks. Figure 1 illustrates the hypothetical interaction between the server fan and the system resistance offered by the racks. The intersection between the system performance curve and the server fan performance curve determines the operating airflow rate for the server rack system. The dotted curve shown in this figure represents the system curve for a single server during the bench test analysis. This figure further indicates, due to increased resistance to airflow, that the operating airflows through the server rack systems can be lower than the desired airflow rate of the bench test of a single server. As the restrictions in the rack enclosures increase, the operating airflow through the servers can decrease. More restrictive rack enclosure designs can reduce the operating airflow rates through the servers. Rack Enclosure Designs Play Important Role Rack Design 2 Rack Design 1 Server Fan Curve Bench Test Pressure Drop Delivered Airflow Rates Airflow Rate Design Airflow Rate As mentioned previously, the design of a rack enclosure plays an important role in determining the airflow rate through the servers. Some of the factors that can affect the performance of server fans due to the design of racks are as follows. A typical rack enclosure consists of front and back perforated doors, railings to hold the servers, and several cables that occupy the space between the back perforated door or blind panel. All of these components create obstructions to the airflow through the racks. In addition, the airflow paths within rack enclosures are crucial. Most racks are designed for a front entry and back exit flow path for the cooling air. Other variations of the flow paths are front entry and top exit or bottom entry and top exit. As air takes more turns before exiting the rack, it induces pressure loss in the system for which a server fan must compensate. A few manufacturers offer a total containment of hot air within the rack enclosures by using ducted exhaust. In these racks the hot air is exhausted from the top either into the room or into a ceiling return. This type of system avoids short-circuiting and mixing of hot air with the room air. Since the air must pass through narrow spaces, the server fans are required to work harder to overcome these additional resistances. If this type of system is not designed properly, it can result in reduced fan performance. Another design offers a rear door heat exchanger to provide instant cooling of exiting hot air. The heat exchanger can create additional resistance to the airflow, which can deteriorate the server fan performance. Other designs offer additional fans in the cabinets to assist in exhausting the hot air. All fans in the data center, including the fans of the air-handling units, should work in harmony with other fans to deliver desired airflow rates and cooling performance. In the ducted exhaust August 2009 Figure 1: The intersection between the server fan curve and the system curve determines the operating airflow rates through the system. This illustration shows operating airflow rates through the server rack system are lower than the desired airflow rates determined for a single server during the bench test. More restrictive rack enclosure designs allow less airflow through the system. with ceiling plenum return (closed loop) system, all the fans (including fans of air-handling units, server fans, and any additional fans in the racks) operate in series, fans in the cabinets can become a bottleneck in delivering desired airflow from the air-handling units. Rack heat loads and how the loads are distributed within the racks can be another factor that can affect airflow through the rack enclosures. Several combinations of these design concepts are available in the market. Consider all of these factors and evaluate their effect on server fan performance before selecting rack enclosures for data centers. Pressure Buildup and Air Recirculation Pressure can build up in the gap between the back perforated door and back of servers from where hot air exits from the servers. Such a pressure buildup can cause internal recirculation of hot air within the cabinet that can be entrained into the servers. Computational fluid dynamics (CFD) analysis was performed for a typical rack enclosure with a front-to-back airflow path to illustrate how such a pressure buildup can affect airflow within the racks and the cooling performance of servers. This model considered a 42U rack with 63% open area ratio of front and back perforated doors. The heat load in the model varied from 5 kW to 20 kW. Accordingly, the airflow rates through servers and the total supply airflow rates were proportionately adjusted to maintain a 20°F (11°C) rise (DT) across the servers. CFD analysis predicts the pressure distribution, temperature distribution, and airflow patterns within the rack enclosure. As shown in Figure 2, CFD analysis predicts higher pressure ASHRAE Journal 49

ASHRAE Journal - August 2009

Table of Contents for the Digital Edition of ASHRAE Journal - August 2009