Increasing governmental regulation is also affecting data center cooling operations. Some states, such as Washington, have energy codes that require use of air economizers for data center facilities. Furthermore, the U.S. Environmental Protection Agency, in its
Report to Congress on Server and Data Center Energy Efficiency, recommends the development of "systems and guidelines for reliable use of outside air economizers." The report also acknowledges the need to develop a "scientific understanding of the impact of environmental conditions (temperature, humidity, particulates and other pollutants) on IT equipment operation and reliability, to expand operating ranges without decreasing reliability." Inferred here is that applying air economizer design methodologies used in commercial office buildings does not address the specific needs of the IT equipment in data centers.
This two-part series addresses the recommendations in the EPA's report to "develop systems and guidelines for reliable use of outside air economizers." An intimate understanding of data center operational requirements and environmental conditions combined with an application of computation fluid dynamic (CFD) airflow modeling to the external environment provide a "scientific understanding," which can result in increased energy-efficiency.
Refrigerant-based air conditioning
Refrigerant-based air conditioning systems developed early in the 20th century were commonly used in commercial buildings in the United States through the 1960s. These design practices expended vast amounts of relatively cheap energy to create controlled microclimates inside the building, and focused on restricting the intrusion of the external environment. During the energy crisis of 1970s these practices were challenged. One response to rising energy costs was the development and use of HVAC systems that used outside air for "free" cooling. Fan-powered air economizers are now standard practice in office buildings in the United States.
Once again, increasing energy costs have focused attention on energy use. Concurrently there are concerns with the national electrical infrastructure generation and distribution capacity as well as a facility's carbon footprint. "Sustainable development" practices are gaining popularity in commercial construction in the United States, and cooling concepts, such as natural ventilation, are becoming more common. Despite these advances, most data centers' environmental criteria call for relatively tightly-controlled environments that rely on energy-intensive refrigerant-based air conditioning cooling solutions. So, if the concerns with the temperature and humidity levels within the data center are resolved, can the application of air economizers yield economic benefits for data center design?
Addressing risk, advancing engineering
Beyond temperature and humidity control, one objection to the use of air economizers for data center facilities is the potential exposure of the data center to airborne contaminates. This argument holds that with the introduction of outside air comes possible contamination (particulates, salts, and other pollution). This may result in detrimental affects on electronic equipment and the health of personnel (such as from entraining cooling tower or generator exhaust). Also, the energy savings anticipated may not be realized due to the recirculation of waste heat back into the facility. The methodologies to handle these concerns on typical commercial office are well established. But in data centers, these issues are magnified because of the relatively large quantity of airflow required to maintain the server floor environmental conditions. Whereas a typical commercial office building may require a mass flow rate of 1.0 to 2.0 cubic feet of airflow per minute per square foot, a typical data center may require airflow rates that are 10 times that amount. Increased mass flow presents an increased risk for entrainment of contaminates (particulate, waste heat, and pollution) from the outside environment into the data center facility.
To account for this, a data center design incorporating air economizers must address the potential of entrainment of waste heat and outside air contaminates. These elements may vary by location, time of day, and isolated events that are external to the facility's operation. The question remains: How can the external airflow patterns be analyzed quantitatively to allow the development of engineering and architectural designs to mitigate these situations?
Computational fluid dynamics applied to the external environment
Image 1:Airflow streamlines around the data center
(click image for larger version)
CFD modeling is used to anticipate the potential entrainment of contaminants and determine the effectiveness of applied design solutions. While CFD is often applied to a data center's internal environment, it may also be applied to external environments. Modeling for multiple wind directions and speeds will identify the most problematic conditions and provide a base level of performance to judge the potential for entrainment problems. Analysis of the wind, its transient nature, and complex turbulence is possible with CFD software. External modeling is accomplished through a quasi-static analysis approach. While this is a simplification of the actual situation (which is very dynamic), a transient analysis requires an increased order of magnitude in effort, which is often unnecessary to make good engineering decisions.
Application of CFD to the external environment provides a basic understanding of the airflow field patterns surrounding the data center. It provides guidance for development of engineering and architectural design and operational strategies to minimize waste heat and contaminate entrainment. This increases the facility's energy efficiency and reduces the carbon footprint and actual operating costs.
Applying CFD to external data center environment to improve operational efficiencyTypically, data center studies that address the energy efficiency using air economizers for cooling assume 100% removal of waste heat; that is, no recirculation occurs in the outside air intakes. However, unless external conditions are quantitatively addressed during the design effort, this assumption is unrealistic. Recirculation of waste heat results in an increase in energy consumption over that which is anticipated. The result is increased energy costs and a decreased return on investment. Furthermore, entrainment of generator radiator and engine exhaust, and contaminants such as cooling tower air discharge, can create internal air quality issues that may have an adverse effect on human health. These factors suggest the need for a reliable means of quantifying the external airflow patterns around the data center building.
Through the application of CFD modeling, the data center design team can quantify and qualify these external airflow patterns. This allows for the evaluation of the architectural features and development of mechanical system configurations to optimize the air intake and exhaust performance to minimize impacts to the facility's operations.
Application of exterior CFD airflow modelingUnderstanding wind direction and speed, wind profile, geographical terrain, sources and concentrations of airborne contaminants, and building aerodynamics are required to develop an external CFD model. The approaching wind is modeled based on the terrain surrounding the project site and consideration of the prevailing wind conditions. A wind profile, known as the atmospheric boundary layer (ASHRAE Fundamentals Handbook, Chapter 16, Atlanta, Ga., 2005), is an airspeed cross-section taken in parallel with the wind direction.
Image 2:Atmospheric boundary layer (click image for larger version)
Within the wind profile, the wind speed increases in magnitude as the height above the ground increases until reaching the free wind speed. In rural locations, the wind speed increases rapidly with height as there are few obstructions. In urban locations, the wind speed increases less with height due to the numerous obstructions. Although the nature of wind blowing over a building is essentially dynamic, (i.e. its airspeed and direction vary continually with time) the approach taken in typical CFD studies is a quasi-static approach. Hence, for a particular wind direction, it is assumed that the airflow around the building reaches steady-state conditions.
Sources of contamination may include cooling tower exhaust, generator combustion and radiator exhaust, loading dock and parking lot locations, and waste heat rejected from the air economizers. Furthermore, off-site sources of contamination may also need consideration. Nearby sources may include emergency generators and cooling towers located on an adjacent site, exhaust from nearby industrial facilities, and transportation corridors.
The key to analyzing the external building environment and excluding contaminants from entering the air intakes lies in understanding the building's aerodynamics. As wind approaches a building, the air is displaced up and around the structure, creating vortices and recirculation zones. A steady-state analysis identifies where the low pressure zones will typically occur around the building and rooftop obstructions and where the resultant vortices are formed. Exhaust streams and waste heat (that impinge on or are released within these vortices) will likely be entrained into the air intakes of air economizing air handlers. Through the application of external CFD modeling these conditions can be better understood.
Applying the results from CFD modeling
Image 3:A composite image that demonstrates three different types of physical placement/design of rooftop air-intake handlers. From left to right: Standard -- Rooftop AHU; Engineered -- AHU with vertical 8-foot relief air stacks; Integrated -- Roof above AHU with 10 foot relief air stacks. (Click image for larger version)
By applying basic knowledge of building aerodynamics, a holistic design approach can dramatically improve the indoor air quality, the energy savings, and the overall building performance.
Many challenges to air economizer applications for data centers may be solved through a holistic design approach. Basic design techniques, utilizing separation between the building air intakes and the exhaust locations by elevation separation and increased discharge velocities, have proven to be successful in data centers. As noted in Part 1, mass airflow rates required for data centers may be tenfold that of commercial office buildings. Therefore, the standard design techniques for applying air economizers may not adequately address the operational risks involved.
Image 4:Plume (click image for larger version)
Architectural features that may be considered include wind towers, air-stilling shafts, vertical-physical separation, and equipment screens that are pervious to airflow. These elements must be quantitatively evaluated. For instance, at increased airflow rates required for data centers, perforated equipment screening systems may effectively appear as a solid wall to the flow of air. However, through predictive CFD modeling techniques, the integration of architectural and mechanical design features and their coupled performance can be quantified to improve the overall building performance.
For example, a 100,000-square-foot data center located in the San Francisco Bay area was designed for 200 Watts/square foot. The original design concept used rooftop HVAC equipment with an architectural screen wall in a configuration typically applied on a low-rise commercial project. A CFD model of the facility shows that, because of recirculation of waste heat, the average air temperature entering the air handler outside air intake (under full economizer operation) increased by over eight degrees Fahrenheit above ambient temperature. This intake produces about a 50% recirculation rate, which resulted in 2,300 kW of unnecessary energy use, at a cost of almost $2,000,000 per year (based on an electrical cost of $0.10 per kWh). In contrast, the application of integrated design features yields a decrease in the recirculation rate to less than 1%, thereby recapturing nearly all the energy that would have otherwise been expended.
Contaminate control: CFD limitationsWhile most external contamination can be controlled through filtration technologies in the air handling system, there are some contaminants for which economical control may not be practical, but they must be considered. For example, providing adequate separation of generator engine exhaust from the air intakes on an air-economizer-equipped facility to preclude entrainment of exhaust (under all wind conditions) may not be practical or allowed by local building codes or site planning restrictions. The entrainment of diesel exhaust presents three areas of concern: the impact on IT equipment, on human health, and on air quality (i.e.: odors). Locating the contaminate source downstream in the prevailing wind reduces the potential for generator engine exhaust entrainment. However, while CFD analysis can predict operational issues, the wind direction is unpredictable. Therefore, there remains a need for mitigation to keep contaminates from entering the facility.
Providing filtration technology to remove diesel fumes from the incoming air increases the air handling system cost and drives up the operational and maintenance costs; it does not address the potential heat issues from entrainment of radiator exhaust. To overcome these issues, performing generator testing only under favorable wind conditions has been applied at many facilities. For emergency conditions, outside air intakes may be automatically closed. However, if outside air intakes are closed, an immediate source of cooling (e.g., a thermal storage system) must be available with adequate capacity to maintain the data center environment while the mechanical cooling system is started.
Air economizers are being used in data center projects in many locations. Some jurisdictions now require their use, and the applications of air economizers will likely increase in the future. Design solutions that use air economizers must not only address the spatial requirements for the equipment, but also the external configuration of the facility, associated equipment (that may produce undesirable conditions like generators and cooling towers), and external environmental conditions (e.g., temperature, humidity, site terrain, surrounding structures and adjacent contaminate sources). By applying CFD modeling techniques, the impact on data center operations from the external environment can be anticipated and minimized.
ABOUT THE AUTHORS: David Seger, P.E. is the principal mechanical engineer for IDC Architects (IDCA) Critical Environment group, and heads up the firm's critical environment mechanical engineering design staff. He has 25 years of experience in critical environments for manufacturing and data processing. David has overseen data center facility designs in the United States and internationally, including those that use air-side economizers, water-side economizers. He holds a B.S. degree in Mechanical Engineering from Oregon State University and is a registered professional engineer in the State of Oregon.
Andy Solberg, P.E. is a mechanical engineer within the IDC Architects (IDCA) Critical Environment group where he has led the engineered systems modeling group for the past 10 years. He applies CFD modeling to develop design solutions that optimize features of advanced technology facilities, mission critical facilities, and high performance buildings (green buildings). He holds a B.S. in Mechanical Engineering from the University of Nevada, Reno, and he is LEED accredited.
This was first published in April 2008