James W. VanGilder

AIRFLOW UNIFORMITY THROUGH PERFORATED TILES IN A RAISED-FLOOR DATA CENTER

The maximum equipment power density (e.g. in power/rack or power/area) that may be deployed in a typical raised-floor data center is limited by perforated tile airflow. In the design of a data center cooling system, a simple estimate of mean airflow per perforated tile is typically made based on the number of CRACs and number of perforated tiles (and possibly a leakage airflow estimate). However, in practice, many perforated tiles may deliver substantially more or less than the mean, resulting in, at best, inefficiencies and, at worst, equipment failure due to inadequate cooling. Consequently, the data center designer needs to estimate the magnitude of variations in perforated tile airflow prior to construction or renovation. In this paper, over 240 CFD models are analyzed to determine the impact of data-center design parameters on perforated tile airflow uniformity. The CFD models are based on actual data center floor plans and the CFD model is verified by comparison to experimental test data. Perforated tile type and the presence of plenum obstructions have the greatest potential influence on airflow uniformity. Floor plan, plenum depth, and airflow leakage rate have modest effect on uniformity and total airflow rate (or average plenum pressure) has virtually no effect. Good uniformity may be realized by using more restrictive (e.g. 25%- open) perforated tiles, minimizing obstructions and leakage airflow, using deeper plenums, and using rectangular floor plans with standard hot aisle/cold aisle arrangements.

Experimental-Numerical Comparison for a High-Density Data Center: Hot Spot Heat Fluxes in Excess of 500 W/FT2

The current trend of using denser server environments is continuously increasing to satisfy the growing needs of e-commerce and other emerging technologies. The resulting high room-level heat fluxes present significant challenges with respect to maintaining acceptable computer rack inlet temperatures and minimizing total data center energy consumption. Numerical methods are widely used to model existing and new facilities. Validation of existing numerical techniques is an important step in facilitating good thermal design of data centers. This paper uses previously published experimental data to present a comparison between test results and numerical simulations. The example considered is a large 7400 ft2 data canter that houses over 130 heat-producing racks (1.2 MW) and 12 air conditioning units. Localized hot spot heat fluxes were measured to be as high as 512 W/ft2 (5.5 kW/m2) for a 400 ft (37 m) region. A numerical model based on computational fluid dynamics (CFD) was constructed using inputs from the measurements. The rack inlet air temperature was considered to be the basis for experimental vs. numerical comparison. The overall mean rack inlet temperature predicted numerically at a height of 1.75 m is within 4degC of the test data with a rack-by-rack standard deviation of 3.3 degC

Potential-Flow Modeling for Data Center Applications

Potential-flow-based airflow and heat transfer models have been proposed as a computationally efficient alternative to the Navier-Stokes Equations for predicting the three-dimensional flow field in data center applications. These models are simple, solve quickly, and capture much of the fluid flow physics, but ignore buoyancy and frictional effects, i.e., rotationality, turbulence, and wall friction. However, a comprehensive comparison of the efficiency and accuracy of these methods versus more sophisticated tools, like CFD, is lacking. The main contribution of this paper is a study of the performance of potential-flow methods compared to CFD in eight layouts inspired by actual data center configurations. We demonstrate that potential-flow methods can be helpful in data center design and management applications.Copyright

Prediction of Cold Aisle End Airflow Boundary Conditions Using Regression Modeling

This paper is a continuation of the development of software tools that estimate, in real time or in near-real time, the cooling performance of a cluster of racks bounding a common cold aisle in a raised-floor data center environment. A fundamental assumption within the algorithm of these tools is that the computation of airflow patterns inside the cold aisle can be decoupled from the room environment. The effect of the room environment impacts the solution in the estimation of the cooling performance primarily through the airflow boundary conditions prescribed at the ends of the cold aisle. Consequently, the accuracy of the cooling-performance tool is directly linked to the accuracy of the end airflow prediction for any room environment. The end airflow is a complex function of many factors including the location and airflow rate of each rack, the perforated-tile airflow rate, and room environment conditions. As shown here, the dominant room-environment parameter is the difference between ambient and supply air temperatures. This paper describes the model developed to estimate the end airflow rate. End airflow values are calculated from several hundred computational fluid dynamics (CFD) scenarios covering a broad range of rack airflow (and power) distributions, tile flow rates, and room environments. An end airflow model is developed based on a regression analysis from the CFD data, which facilitates the real-time prediction of the end airflow for any practical cluster layout and room environment. The difference between accepted and predicted end airflow values is typically less than 25% of the accepted value or per-tile airflow rate.

Partially Decoupled Aisle Method for Estimating Rack-Cooling Performance in Near-Real Time

The Partially Decoupled Aisle (PDA) method facilitates a near-real-time cooling-performance analysis of a single cluster of racks and, potentially, coolers, bounding a common hot or cold aisle in a data center. With the PDA method, the airflow patterns and related variables need be computed only within an isolated cold or hot aisle “on the fly” through CFD analysis or other means. The analysis is fast because the much larger surrounding room environment is not directly modeled; its effect enters the model through the boundary conditions applied to the top and ends of the isolated aisle. The proper boundary conditions in turn may be estimated from an empirical model determined in advance (“offline”) from the study of a large number of CFD simulations of varying equipment layouts and room environments. A software tool based on the PDA method, which uses a full CFD engine to solve the aisle airflow within the isolated aisle, can analyze a typical cluster of racks and coolers in 10–30 seconds and requires no special user skills. This paper formally introduces the general PDA method and shows several examples of its application with comparisons to corresponding whole-room CFD analyses.Copyright

Data Center Cooling Prediction Using Artificial Neural Network

An analytical approach using artificial intelligence has been developed for assessing the cooling performance of data centers. This paper discusses the use of a Neural Network (NN) model in the real-time prediction of the cooling performance of a cluster of equipment in a data center environment. The NN model is used to predict the Capture Index (CI) [1] as a function of rack power, cooler airflow and physical/geometric arrangement for a cluster located in a simple room environment. The Neural Network is “trained” on thousands of hypothetical but realistic cluster variations for which CI values have been computed using either PDA [2] or full Computational Fluid Dynamics (CFD). The great value of the NN approach lies in its ability to capture the non-linear relationships between input parameters and corresponding capture indices. The accuracy of the NN approach is 3.8% (Root Mean Square Error) for a set of example scenarios discussed here. Because of the real-time nature of the calculations, the NN approach readily facilitates optimization studies. Example cases are discussed which show the integration of the NN approach and a genetic algorithm used for optimization.© 2007 ASME

A statistical prediction of cold aisle end airflow boundary conditions

A statistical prediction of cold aisle end airflow boundary conditions. This paper is a continuation of the development of a software tool (VanGilder and Shrivastava, 2006) that estimates, in real time, the cooling performance of a cluster of racks bounding a common cold aisle in a raised floor data center environment. A fundamental assumption within the algorithm of the tool is that the computation of airflow patterns inside the cold aisle can be decoupled from the room environment. The effect of the room environment impacts the solution in the estimation of the cooling performance primarily through the airflow boundary conditions prescribed at the ends of the cold aisle. Consequently, the accuracy of the cooling-performance tool is directly linked to the accuracy of the end airflow prediction for any room environment. The end airflow is a complex function of many factors such as rack power and rack position, tile flow rate, and room environment conditions including the location of and airflow rate through the return vents. This paper describes the statistical model developed to estimate the end airflow rate. End airflow values are calculated from several hundred CFD scenarios covering a broad range of rack power distributions, tile flow rates and room environments. An end airflow model is developed based on a regression analysis from the CFD data, which facilitates the real-time prediction of the end airflow for any practical cluster layout and room environment. The difference between accepted and predicted values is typically less than 25% of the accepted value or per-tile airflow rate

Experimental Measurement of Server Thermal Effectiveness for Compact Transient Data Center Models

Server thermal mass can significantly affect the rate at which a data center heats up following a loss of cooling and moderate transient temperature fluctuations due to changing CPU utilization. Recently, a compact server model has been introduced which captures the effects of thermal mass while avoiding the impractical level of detail that would be required by an explicit representation of all relevant server components. Inputs to that model include server mass, overall effective specific heat, and a parameter called the “server thermal effectiveness”. The latter characterizes the server’s ability to transfer heat to/from the airstream passing through it and can take values between zero (no heat exchange) and one (maximum possible heat exchange). Server thermal mass is a physical property of a server and is not influenced by external factors.In order to use the compact model for practical applications, we must experimentally measure the thermal effectiveness of actual servers. The present study reviews the compact model and describes the development of an experimental technique for measuring thermal effectiveness. The technique is validated using simple plate fin heat sinks in place of an actual server. This “server proxy” is sufficiently simple so that it can be modeled accurately in detail in CFD, providing well-controlled benchmark data. CFD and experimental measurements both yield a value of server thermal effectiveness of approximately 0.6, providing confidence in the model and measurement technique for the future characterization of actual servers.Copyright

Cooling performance of ceiling-plenum-ducted containment systems in data centers

In an effort to improve the reliability and efficiency of data centers, racks and sometimes entire hot aisles are ducted to a dropped ceiling. The cooling performance of such systems strongly depends on IT and cooler airflow, the number and configuration of ducted objects and perforated ceiling tiles, the leakiness of the ceiling system, ceiling plenum depth, and other factors. Recently, a compact model has been proposed in which a Flow Network Model (FNM) representing the ducted equipment is embedded into a parent CFD model. By eliminating the need to explicitly model difficult-to-characterize leakage paths in CFD, this approach allows for realistic solutions while greatly improving the solutions speed and robustness of the CFD simulation. This paper employs the FNM (without CFD) to characterize and compare the cooling effectiveness of individually-ducted racks and ducted hot aisles subject to a given ceiling plenum pressure. Example resistance values needed in the FNM are provided. Additionally, an example data center layout is studied with the coupled FNM-CFD model to explore cooling performance as a function of ceiling leakiness, plenum depth, ratio of cooling to IT airflow, and rack density (IT airflow). Best-practice type recommendations for ducted equipment are provided.

A method for predicting key data center temperatures by known airflow streams

We present a simplified model for predicting key data-center temperatures, such as those of rack inlets and cooler returns. If the primary airflow streams into and out of all racks and coolers are known, these airflow values can be combined with the assumption of a well-mixed room ambient volume to create a simplified, but energy-balanced, set of temperature equations. Since the temperature estimates are restricted to only a small number of data center objects, solutions can be found quickly. This temperature model can be adapted to quickly address data center design or management issues without multiple Computational Fluid Dynamics (CFD) simulations. Optimized cooler set points can be quickly found with only slight adjustments to the model. In another application, the effects of cooler capacity models are easily understood when incorporated within the temperature model. The object-averaged inlet temperature estimates generated by this temperature model compare favorably to those found by CFD.