|
|
|
The Consortium for Energy Efficient Thermal Management (CEETHERM)
was initiated in 2002 to address research challenges of the intermediate
and long term nature associated with thermal and energy management of electronics.
This activity brings together the expertise of many researchers at the Georgia Institute of Technology that have
been focusing on related problems for many years. The consortium offers to industry partners
a unique approach to leverage sponsorship to handle research challenges of a
pre-competitive nature. The consortium focuses on research topics of
medium and long-range interest as identified in discussion with the partners.
- Rack Flow Rates and Characterization: Manufacturers are designed
servers with variable airflow rates in order to maintain a
constant temperature rise.
This may lead to a large amount of by-pass air when the
servers have low power dissipation and result in low energy
efficiency for the data center.
A Particle Image Velocimetry (PIV) system has been constructed to investigate
the air flow vectors, as well as possible recirculation effects.
- Waste Heat Recovery: As data center power densities continue to rise, cooling costs
have become a serious concern . Also,
incorporating new high power racks into the same infrastructure as legacy racks presents
cooling design challenges. Waste heat recovery can potentially reduce or eliminate
changes to the cooling infrascture, as well as reduce the overall power used to cool
the electronics, effectively improving the PUE.
- Predictive Modeling for Adaptive Data Centers: Most data centers are designed for specific optimal operating conditions.
However, many adverse conditions arise during typical operation, causing the data center to perform
at below optimum efficiency. By employing a POD (proper orthogonal decomposition) algorithm, which reduces
computation time significantly, near real-time responsiveness can be achieved, thereby maintaining
design efficiencies. Also a new approach, using wavelets and wavelet transforms, hopes to further improve
the modern data center's ability to adapt to its volatile operating conditions.
- Raised Floor
Plenum Design: The geometry of the plenum, flow obstructions and
perforated tile characteristics are very important in delivery
cool supply air to the racks in a raised floor plenum data
center. Some suggest that
beyond 2 feet, the plenum depth is unimportant and that for smaller
depth plenums, cables and obstructions should be located near the
floor. Can this be
verified? The effect of
tile grate opening (10% - 56%) is important and best-practices for
tile selection as a function of floor heat loading should be
developed. Due to
non-uniformity in data center heat loads, one may wish to create
zones with different airflow rates. Is the best way to accomplish
this with baffles or some other technique?
- Experimental Plenum
Characterization: The
experimental data center lab provides us with a unique opportunity
to characterize the raised floor plenum of a data center while
exerting some control over the flow obstructions. The primary concern is the
effect of under-floor geometry on perforated tile flow rates. This can be investigated by
putting blockages (e.g. foam blocks and PVC pipes) in the plenum
at various locations and measuring the resultant tile flow
distribution. Another
concern in raised floor plenum data center is the leakage flow, or
supply air escaping the plenum through seams between the tiles and
cable cut-outs. This can be
documented in the experimental facility to determine the
difference between the CRAC flow rate and the net perforated tile
flow rate.
- Data Center Modeling: Many data center airflow and
heat transfer investigations are computational in nature and the
physical models as well as optima numerical methods need to be
verified. Detailed full
field measurements using particle image velocity (PIV) are
possible in the lab and can be used to validate the quality of
approximations resulting from standard turbulence models
implemented in commercially available CFD codes. The measurements can also be
used to identify the limitations of CFD modeling and quantify the
error incurred with common simplifications. Such issues included using a
lumped resistance to model perforated tiles and the appropriate
boundary conditions to accurately model the CRAC units.
- Transient Scenarios: Even with a high
level of redundancy in power and cooling equipment, system
failures at the CRAC and rack level are still a concern. The CEETHERM lab allows us to
perform these studies without consequence to actual computing
hardware. The objective
would be to determine the time before the data center is unable to
adequately cool the computing equipment for various types of
failures
- Liquid Cooling: With ever-increasing power dissipation
levels, forced air cooling is becoming impractical for high power
density racks. Developing a
framework for liquid cooling of ultra-high power racks can greatly
increase the amount of power dissipated from a rack and lead to
increased energy efficiency in the data center. Liquid cooling can be integrated
with facility level cooling and possibly eliminate air as an
intermediate heat transfer media.
|
|
|
|
