0
Research Papers

Computational and Experimental Validation of a Vortex-Superposition-Based Buoyancy Approximation for the COMPACT Code in Data Centers

[+] Author and Article Information
Michael M. Toulouse

Mechanical Engineering Department,
University of California,
Berkeley, CA 94709

Amip J. Shah

Hewlett Packard Laboratories,
Palo Alto, CA 94304

Van P. Carey

Mechanical Engineering Department,
University of California,
Berkeley, CA 94709

1Corresponsing author.

Contributed by the Electronic and Photonic Packaging Division of ASME for publication in the Journal of Electronic Packaging. Manuscript received October 1, 2012; final manuscript received April 15, 2013; published online July 24, 2013. Assoc. Editor: Yogendra Joshi.

J. Electron. Packag 135(3), 030903 (Jul 24, 2013) (8 pages) Paper No: EP-12-1090; doi: 10.1115/1.4024302 History: Received October 01, 2012; Revised April 15, 2013

A simplified model and GUI package, named COMPACT (Compact Model of Potential Flow and Convective Transport), has been developed in previous years to provide a fast alternative to full computational fluid dynamic (CFD) thermofluidic models for a variety of data center applications. These include, but are not limited to, use as a first-order design tool, a potential improvement to plant-based controllers, and an initial guess for complex CFD solvers. COMPACT applies convective energy transport equations to a computed potential flow field to approximate a flow and temperature field, taking 30 s or less for a commercially available laptop to characterize a 7700 cell room. Previously, the results from this model were compared to experimental measurements taken from a data center at Hewlett-Packard Laboratories (HP Labs) in Palo Alto, CA. High localized temperatures in the model led to the conclusion that recirculation and buoyancy were contributing excessively to error in the model. This paper proposes a method of vortex superposition to account for these effects, in which locations with high temperature in the original model are analyzed and a corrective flow field consisting of Rankine vortices is superimposed on the solution. This approach is tested with further experimental measurements taken from the data center at HP Labs, as well as conventional commercially available CFD. These newer results show a marked decrease in mean deviation of the model from measured temperatures, as well as elimination of the highly localized temperatures which afflicted the original COMPACT results. The vortex superposition model is also “tuned,” with vortex strength optimized for multiple test cases at varying levels of recirculation.

FIGURES IN THIS ARTICLE
<>
Copyright © 2013 by ASME
Your Session has timed out. Please sign back in to continue.

References

Figures

Grahic Jump Location
Fig. 1

Simplified visual representation of experimental data center, (a) isometric view with annotation explaining air flow through the server room, (b) top view and (c) side view of room with locations of temperature measurements (gold circles)

Grahic Jump Location
Fig. 2

Mean deviation (at in-flowing server rack faces) of modeled temperatures from their measured values, versus the vortex strength multiplier used in vortex superposition optimization

Grahic Jump Location
Fig. 3

Streamline plot for optimized COMPACT for measured test case. Streamlines originate from floor inlets and out-flowing server rack faces.

Grahic Jump Location
Fig. 4

Temperature slice plot for optimized COMPACT for measured test case. Slices are located at x = 18 ft, y = 10.5 ft, and z = 6 ft, respectively.

Grahic Jump Location
Fig. 5

Velocity slice plot for fluent for measured test case. Horizontal slices are flow through floor inlets and ceiling outlets; the vertical slice is a vector plot at x = 22 ft.

Grahic Jump Location
Fig. 6

Temperature slice plot for fluent for measured test case. Slices are located at x = 18 ft, y = 10.5 ft, and z = 6 ft, respectively.

Grahic Jump Location
Fig. 7

Maximum and mean deviation of the basic COMPACT, optimized COMPACT, and ansys fluent models from measurements (85 points) taken at the in-flowing face of server racks

Grahic Jump Location
Fig. 8

Maximum and mean deviation of the basic COMPACT, optimized COMPACT, and ansys fluent models from all measurements (286 points) taken in the data center

Tables

Errata

Discussions

Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related eBook Content
Topic Collections

Sorry! You do not have access to this content. For assistance or to subscribe, please contact us:

  • TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
  • EMAIL: asmedigitalcollection@asme.org
Sign In