|
|
|
|
|
|
|
|
|
|
|
Validation Studies
|
|
|
|
A Well-Validated Tool for Airflow/Thermal Management in Raised-Floor Data Centers
TileFlowTM has been extensively validated using measurements in real-life raised floor data centers as well as in laboratory-scale setups. The airflow rates and rack inlet temperatures calculated by TileFlow are typically within 10% of the measured values. Details of four validation studies are provided here; for other validation studies, refer to various technical publications.
Laboratory-Scale Data Center at IBM
To validate the technical approach in TileFlow, detailed tests were performed in a data center located at IBM in Poughkeepsie, N.Y. Since the focus
of the study was on performing measurements of flow rates from the perforated tiles, a small section of the floor (6.06 m x 20.0 m), comprising two CRAC units was isolated. This arrangement provided a
controlled area where all the air exiting the CRAC units would exhaust from the perforated floor tiles. The CRAC units were manufactured by Liebert (Model No. FD411C). The perforated tiles had an open area of
19%. Five arrangements of perforated tiles, shown in the figure below, were investigated with either one or both CRAC units operating. Complete details of this study can be found in the publication by Schmidt
et al. (2001).
|
|
|
|
For brevity, comparisons are shown here for only two tests. In the first test, both CRAC units are operational. A
comparison of the predicted and measured tile flow rates for the four rows of tiles is shown in the following figure. The peak in the airflow velocity distribution is located closer to unit A. The results
given by TileFlow correctly reproduce this behavior.
|
|
|
|
In the second test, CRAC Unit A is off. The following figure shows the predicted and measured flow rates through the
various tiles. The data shows reverse flow near unit B. TileFlow also reproduces this bahavior.
|
|
|
|
Schmidt R.R., Karki K.C., Kelkar K.M., Radmehr A., and Patankar S.V., “Measurements and Predictions of the Flow
Distribution Through Perforated Tiles in Raised-Floor Data Centers,” Proceedings of IPACK01, The Pacific Rim/ASME International Electronic Packaging Technical Conference and Exhibition, July 8-13 2001,
Kauai, Hawaii, USA.
Data Center at Oakridge national Laboratories
To further validate TileFlow, measurement of airflow rates was undertaken in a large data center at the Oakridge National Laboratories. The floor
size is 88 ft x 40 ft. A 26 ft x 18 ft space in one of the corners is not available for placing electronic equipment. Six CRAC units are placed at different locations along the walls.
The following pictures show variation of the airflow rates over the floor, and the contours of pressure and velocity vectors under
the floor, calculated using TileFlow.
|
|
|
|
|
|
|
|
The next two figures show the comparison of airflow rates given by TileFlow with measurements for two rows of perforated
tiles. TileFlow accurately predicts the variation of the flow rates.
|
|
|
|
|
|
|
|
Data Center Designed by Engineering Design Group (EDG)
Engineering Design Group (EDG, a GE Digital Energy Company) recently designed and commissioned a new data center in Washington, D.C. The data center
has approximately 10,000 sq.ft. floor space, with 11 Liebert CRAC units and over 200 perforated tiles. The novelty of the design lies in locating CRAC units in a distributed manner. This arrangement ensures
that each perforated tile has a CRAC unit in its proximity, thus allowing better control over the flow distribution through the perforated tiles. Analysis of the flow distribution using TileFlow verified the
design concept.
|
|
|
|
Innovative Research also performed extensive measurements of the flow rates to validate the accuracy of TileFlow. The
following figure shows the predicted distribution of the flow rates through the perforated tiles. The lower flow rates are for the tiles with cable cutouts, which have small open areas. If the tiles with
cutouts are discounted, the flow rate through the perforated tiles is fairly uniform due to the distributed arrangement of the CRAC units. Comparison of the measured and the predicted flow rates is also shown
for selected rows of tiles. The predicted flow rates agree well with the measured values.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
NCEP Data Center in Bethesda, Maryland
This data center is the subject of a recent publication (Thermal Profile of a
High-Density Data Center--Methodology to Thermally Characterize a Data Center, presented at 2004 ASHRAE Meeting in Nashville, TN; Publication NA-04-4-2) by Dr.
Roger Schmidt, IBM, Poughkeepsie. Dr. Schmidt have reported details of the data center and measurements of airflow rates through perforated tiles and cutouts, heat
loads in the racks, and rack inlet temperatures.
TileFlow was use to calculate the airflow patterns and temperature distribution in this
data center. The calculated rack inlet temperatures were compared with the measurements of Dr. Schmidt. Representative results from this study are included here.
NCEP Data Center: Floor Plan
Rack inlet temperatures for racks in Row 9
Rack inlet temperatures for racks in Row 15
|
|
|
|
|
|
|
|
|
|
|
|
|
