page 1 public interest energy research (pier) california institute for energy & environment...

Page 1

Public Interest Energy Research (PIER)California Institute for Energy & Environment (CIEE)

Bill Tschudi Lawrence Berkeley National Laboratory

http://hightech.lbl.gov

Data Center Energy Research and Demonstrations

Critical Facilities Roundtable Meeting - March 25, 2005

• Benchmarking and Best Practices identification• Research-directed benchmarking• Self-benchmarking protocol

• Improve uninterruptible power supplies (UPS)

• Develop performance metrics (Computing vs. Watts)

• Improve power supplies in IT equipment

• Adding demonstration projects

2

Current LBNL data center related activities

Wide range of efficiencies:For example: data center HVAC

0102030405060708090

100

1 2 3 4 5 6 7 8 9 10 11 12 13 14

Facility Identifier

W/S

q.F

t.

Current Computer Load Projected Computer Load

Average 39.3

Average ~25

HVAC (as a % of total load)

0%

10%

20%

30%

40%

50%

60%

1 2 3 4 5 6 7 8 9 10 11 12

Data Center Identifier

% o

f to

tal l

oad

3

HVAC effectiveness varies

Load intensity is increasing in our supercomputing center

4

LBNL NERSC Facility Computer Power Density

0

10

20

30

40

50

60

70

80

90

100

25-Oct-2002 17-Aug-2004 14-Feb-2005

W/s

f

5

Factory Measurements of UPS Efficiency

80%

85%

90%

95%

100%

0% 20% 40% 60% 80% 100%

Percent of Rated Active Power Load

Eff

icie

nc

y

Flywheel UPS

Double Conversion UPS

(tested using resistive loads)

Efficiency ranged from 86% to 97% at peak load, but efficiency under typical operating loads ranges from 81% to 97%

Measured UPS Efficiency

Typical operating range

6

Data Centers – UPS measured vs mfg. efficiencyMeasured and Manufacturer-Reported UPS

Efficiency at Nominal Load

89%90%91%92%93%94%95%96%97%98%99%

1 2 3 4 5 6 7

UP

S E

ffic

ien

cy

Measured Efficiency

Manufacturer-reported efficiency

• Manufacturer’s reported efficiency in product literature consistently within 1% of measured performance

• Unfortunately, manufacturers often do not report efficiency in typical data center operating range (30% - 50%)

Power Supply Findings:

7

Measured Server Power Supply Efficiencies

45%

50%

55%

60%

65%

70%

75%

80%

85%

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

% of Nameplate Power Output

% E

ffic

ien

cy

80 Plus

Non-PFC Power Supplies

PFC Power Supplies

Typical operating range

Power Conversion Conclusions:

8

There is a wide range of efficiency for both UPS and power supplies and they have lower part load

efficiency (where most operate)

Server power supplies are less efficient than PC’s

Computer manufacturers caution that setting criteria for power supplies may lead to “gaming” of the rest of the power supply chain

Elimination of electric power conversion(s) could be a large impact on energy use

• Heat removal from servers without fans

• DC powering rack of computers• IT/facilities design process for new

supercomputer center–

9

Data center demonstration projects

• Start-up company – Nisvara has patented technology using carbon micro fiber to efficiently conduct heat in one direction to a point where it can be transferred to air or liquid.

• Seeking “industry partner(s)” to provide servers and rack systems to modify to

use this technology. Also seeking a demonstration site.

–

10

Heat removal without fans demonstration

• Discussions with server manufacturers, processor and other component manufacturers

• www.nisvara.com

–

11

Heat removal without fans …

• Demonstrate energy savings using commercially available DC power at rack level to eliminate or consolidate a power conversion

• Various potential configurations and manufacturers

• Seeking “industry partner(s)” to provide servers to modify to use this technology. Also may involve a demonstration site.

• Potential overlap with heat removal demo–

12

DC powering demonstration

13

Thank you

Contact me if you would like to participate

wftschudi@lbl.gov

510-495-2417