7x24 2015 Fall Magazine

71 / 88

71
underfloor issues result in increases
in hot air re-circulation over the tops
of cabinets or around the ends of
cabinet rows. a common response to
address high it equipment intake
temperatures is to increase cooling
airflow volume or reduce
temperature setpoints.
often ignored ΔT:
exhaust back to
Cooling unit
(#2 in Figure 1)
in contemporary data centers, the Δt
from servers back to the cooling
source is usually negative; the
temperature of the return air tends
to decrease after it is exhausted from
the it equipment as it makes its way
back to be re-cooled. this makes
sense, as the excess cooling air is
bypassing the data center heat load
and returning to the cooling units,
reducing the temperature of the
return air along the way. there are
several causes of such bypass
airflow:
• improperly located perforated floor
tiles that can cause cool air to
bypass the heat load.
• large underfloor pressure
differentials which may result in
high pressure zones that pump
significant quantities of air into a
cold aisle in excess of the demand
indicated by the intake fans of the
associated it equipment.
• unsealed cable access holes in tiles
located toward the rear of server
racks that will result in cold air
joining the exhaust air inside the
back of the server rack. this source
of bypass airflow can be especially
troublesome when managing
cooling efficiency issues via Δt’s, as
it can mask the true it equipment
Δt. before monitoring or taking
corrective action in response to this
Δt, plug any holes in the unsealed
floor cut-outs in the rear of server
racks to help eliminate false
measurements.
RemeDial
ConsiDeRaTions
when server inlet temperature is
more than 5˚f above the supply
temperature being produced by
cooling units, there is hot air re-
circulation occurring from open
pathways between a hot and cold
aisle, inside or around the server
racks, or an inadequate flow zone
that may require Cfd analysis to
determine the source of pressure
variations under the floor. if the
return air intake of the cooling units
are more than 5˚f lower than the
exhaust temperature from the it
equipment, there is a bypass airflow
problem. areas to check include
improperly placed floor tiles,
unsealed floor openings, or simply
excess airflow being delivered into
the room.
these 5˚f guidelines for calibrating
Δt’s are only suggestions to help
increase the efficiency of the data
center. the overall goal (particularly
if free cooling is available) is to get
the differentials between the server
exhaust and the cooling unit return
intake and between the cooling unit
supply and the server intake as close
to zero as possible. that alignment
will result in a better harmonization
between expenses made for cooling
and the true cooling work required.
it will also increase opportunities for
more free cooling hours if free
cooling is part of the data center’s
design.
in most circumstances, optimum
efficiency in the data center will be
achieved when:
• there is minimal difference
between the supply air
temperature and the server inlet
temperature (#4 in figure 1).
• there is minimal difference
between the it equipment Δt and
the cooling coil Δt (#1 and #3 in
figure 1).
• the supply temperature can be
elevated to a temperature
approximating the maximum
specified upper threshold for the
space (once the above two
conditions are met).
Monitoring all four Δt’s can provide
valuable information for calibrating
the data center to meet its ideal
performance level, regain stranded
capacity and reduce operating cost.
Lars Strong is Senior Engineer of Upsite Technologies. He can he reached at
Ian Seaton is Technical Advisor to Upsite Technologies. He can he reached at
¹ There will be some conditions in which the ΔT will not remain constant, especially with the proliferation of variable speed fans in IT equipment.
When inlet temperatures are allowed to exceed the maximum recommended threshold and move toward the upper allowable levels, many servers
will increase fan speeds to protect equipment, thereby reducing the ΔT through the IT equipment. Additionally, with the increased adoption of
cloud-based data centers, large caches of work can be transferred between data centers with the resultant increased work load producing higher
chip temperatures and, therefore, higher ΔT’s. Nevertheless, in normal conditions, the equipment ΔT remains a constant.