heat.
Making matters worse, the
efficiency of the power conver-
sion drops dramatically when
UPS systems are lightly loaded,
which almost always is the case
because of desires to maintain
redundancy and keep loads around 40 percent.
Figure 1 shows wide variations in UPS efficiency.
With a more-efficient UPS system, an immedi-
ate reduction in overall electrical-power demand
W
ith annual energy costs per square
foot that are 10 to 30 times those
of typical office buildings, data
centers are an important target in energy-saving
efforts. They operate continuously, which means
their electricity demand always
is contributing to peak utility-
system demand, an important
fact given that utility pricing
increasingly reflects time-de-
pendent tariffs. Energy-effi-
ciency best practices can hold
the key to significant savings,
while improving reliability and yielding other
non-energy benefits.
This article will summarize best practices
developed from an extensive study of energy use
in 22 data centers.
SPECIFY EFFICIENT UPS SYSTEMS AND
IT-EQUIPMENT POWER SUPPLIES
One of the best ways to improve
data-center energy efficiency is to
reduce heat loads attributed to power
conversion within both information-
technology (IT) equipment and
the data-center infrastructure. With
45
HPAC Engineering March 2006
William Tschudi, PE, is a principal inves-
tigator, Evan Mills, PhD, a staff scientist,
and Steve Greenberg, PE, a staff mechan-
ical engineer for Lawrence Berkeley
National Laboratory in Berkeley, Calif.,
while Peter Rumsey, PE, is principal of
Rumsey Engineers in Oakland, Calif.
Findingsand resulting best practicesfrom
a study of energy use in 22 data centers
Data-Center
M e a s u r i n g a n d M a n a g i n g
Energy Use
By
WILLIAM TSCHUDI, PE
,
EVAN MILLS, PHD
,
and
STEVE GREENBERG, PE
,
Lawrence Berkeley National Laboratory,
and
PETER RUMSEY, PE
,
Rumsey Engineers
Percent of rated active power load
Efficiency
, per
cent
100
95
90
85
80
75
70
20
0
40
60
80
100
Flywheel UPS
Double-conversion UPS
Delta-conversion UPS
FIGURE 1. Factory measurements of UPS efficiency using linear loads. of 20 to 30 percent can be achieved.
Additionally, downsizing HVAC and
upstream electrical systems (in new
construction) can result in capital-cost
savings and excess capacity.
A similar phenomenon occurs within
IT equipment, such as servers, where
multiple power conversions typically
occur. Conversion from alternating
current to direct current and then
multiple direct-current conversions
contributes to IT-equipment energy
loss. Lawrence Berkeley National Labo-
ratory (LBNL) found a wide range
of efficiencies in power supplies used
in servers. With more-efficient power
supplies, additional energy- and capital-
cost savings can be obtained.
OPTIMIZE AIR MANAGEMENT
As computing power skyrockets, data
centers are beginning to experience
higher concentrated heat loads. In facili-
ties of all sizesfrom small data centers
housed in office buildings to large data
centers essentially dedicated to IT equip-
menteffective air distribution has a
significant impact on energy efficiency
and equipment reliability. Energy bench-
marking using a metric that compares
energy used for IT equipment to energy
used for HVAC systems (Figure 2) re-
veals that some data centers perform bet-
ter than others. For this metric, a higher
number indicates that proportionately
more electrical power is being provided
for computational equipment than for
cooling. In other words, the HVAC
system is more effective at removing
heat from IT equipment. The variation
from worst to best is fivefold. This can be
attributed to a number of factors, includ-
ing how cooling is generated and distrib-
uted; however, air management is a key
part of effective and efficient cooling.
Improving air management, or opti-
mizing the delivery of cool air and the
collection of waste heat, can involve
many design and operational practices.
Air-cooling improvements often can
be made by addressing:
The short-circuiting of heated air
over the top of or around server racks.
The short-circuiting of cooled air
back to air-conditioning units through
openings in raised floors, such as cable
openings and misplaced floor tiles with
openings.
Misplaced raised-floor air-discharge
tiles.
Poorly located computer-room air-
conditioning (CRAC) units.
Inadequate ceiling height or an
undersized hot-air-return plenum.
Air blockages, which are common
with piping and large amounts of cabling
under raised floors.
Openings in racks that allow air
bypass (short-circuiting) from hot
areas to cold areas or vice versa.
Poor airflow through IT-equipment
racks caused by restrictions in rack
structure.
IT equipment with side or top air
discharge adjacent to front-to-rear-
discharge configurations.
Inappropriateeither too high or
too lowunderfloor pressurization.
Ones general goal should be to mini-
mize or eliminate inadvertent mixing
between cooling air supplied to IT
equipment and hot air rejected from the
equipment. Air distribution in a well-
designed system can reduce operating
costs, reduce investment in HVAC
equipment, allow increased utilization,
and improve reliability by reducing
processing interruptions and equipment
degradation attributed to overheating.
Solutions to common air-distribution
problems include:
The use of hot-aisle/cold-aisle
arrangements, by which racks of com-
puters are stacked with the hot discharge
sides facing each other and the cold inlet
sides facing each other (Figure 3).
46
March 2006 HPAC Engineering
D
A
T
A
-
C
E
N
T
E
R

E
N
E
R
G
Y

U
S
E
Ratio of IT equipment to HV
AC load
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Data-center number
1
2
3
4
5
6
7
8
9 10 11 12 14 16 17 18 19 20 21 22
FIGURE 2. Data-center HVAC effectiveness.
Return-air plenum
Hot
aisle
Hot
aisle
Cold
aisle
Physical
separation
Raised floor
FIGURE 3. Typical hot-aisle/cold-aisle arrangement. Cold air can be distributed from
above or below. 48
March 2006 HPAC Engineering
Sealing openings in underfloor
systems.
Blanking unused spaces in equip-
ment racks.
The careful placement of CRAC
units and floor-tile openings, often
through the use of computational-fluid-
dynamics modeling.
Collecting heated air through high
overhead plenums or ductwork and effi-
ciently returning it to the air handler(s).
Minimizing obstructions to proper
airflow.
CAPITALIZE ON FREE COOLING
Data-center IT-equipment cooling
loads are nearly constant throughout the
year. Water-side economizers utilizing
evaporative cooling (usually provided by
cooling towers) can be used to indirectly
produce chilled water to cool a data
center when outdoor conditions are mild
or at night. This free cooling is best
suited to climates with wet-bulb temper-
atures lower than 55 F for 3,000 or more
hours a year. Free cooling can improve
the efficiency of a chilled-water plant by
lowering chilled-water approach temper-
atures (i.e., precooling chilled water
before it enters a chiller) or eliminate the
need for compressor cooling, depending
on the outdoor conditions and overall
system design. With free cooling, chilled-
water-plant energy consumption can be
reduced by up to 75 percent, with related
improvements in reliability and mainte-
nance through reductions in chiller
operation. Because this solution does not
affect the quality of air entering IT equip-
ment, it can be an economical alternative
to air-side economizers in the retrofit of
a chilled-water-cooled data center.
Air-side economizers also can provide
free cooling; however, their use is some-
what controversial. While some IT-based
data centers routinely use outside air
without apparent complications, others
are concerned about contamination
and thermal control in their equipment
rooms. Having seen the use of outside
air result in energy-efficient operation in
several data centers, LBNL is planning
to examine the validity of contamination
concerns. The American Society of
Heating, Refrigerating and Air-Condi-
tioning Engineers data-center technical
committee, TC 9.9, is expected to
develop guidance. For now, simply using
a standard commercial-building econo-
mizer is not recommendednot with-
out an engineering evaluation of the
local climate and contamination condi-
tions.
Temperature and humidity fluctua-
tions, as well as particulate and gaseous
pollutants, must be considered. Mitiga-
tion may involve filtration or other
measures.
If outside air is to be used for cooling,
D
A
T
A
-
C
E
N
T
E
R

E
N
E
R
G
Y

U
S
E 50
March 2006 HPAC Engineering
adequate access to the outside must be
provided. Central air-handling units
with roof intakes or sidewall louvers
are used most often, although some
internally located CRAC units offer
economizer capabilities when provided
with appropriate intake ducting.
REPLACE OUTDATED RULES OF THUMB:
THE EXAMPLE OF HUMIDIFICATION
A remnant of the days of mainframe
computers and tape storage, the need
for tight humidity control generally
can be relaxed or eliminated, as there
is very little humidity load from within
data centers. In the study undertaken
by LBNL, many of the data centers
attempting tight humidity control were
found to be humidifying and dehumidi-
fying simultaneously. TC 9.9 addressed
that issue by developing guidance
concerning temperatures and humidity
supplied to the inlet of IT equipment.
1
That do