The Unexpected Impact of Raising Data Center Temperatures

Overview

The paper models a 1 MW packaged air-cooled chiller plant with a dry-cooler economizer in three cities to test whether raising IT inlet temperature saves money. The model is 1 MW fully loaded, 3 air-cooled chillers N+1 sized to the 20-year extreme, a VFD dry cooler, fixed-speed pumps, N CRAHs with hot aisle containment, 4 kW/rack, $0.10/kWh, 3% cost of capital and a 10-year TCO. Chiller energy always falls as economizer hours rise, but server and CRAH fans ramp up with higher server airflow demand and extra CRAHs may be needed, so total energy and TCO can get worse.

Floating the inlet between 15.6 and 26.7 C versus a fixed 20 C at full load changes total energy by a 1% saving in Chicago, a 14% saving in Seattle and a 12% increase in Miami; TCO is unchanged, 6% better and 6% worse respectively. Raising to a fixed 27 C gives a 13% increase in Chicago, a 2% saving in Seattle, a 12% increase in Miami and a 31% increase in failures (X-factor) regardless of location. The TCO-optimal maximum float temperature is 27 C in Seattle, 23 C in Chicago and 21 C in Miami. Floating with outdoor conditions beats a new fixed point on both energy and failure risk.

The paper's main warning is that cooling-only partial PUE improved in every city even where total energy rose, for example Miami from 1.327 to 1.314, so PUE alone is a misleading decision metric. Results are sensitive to the server airflow curve: Seattle ranges from a 36% saving with no CFM rise to a 10% increase with a high rise, and Miami from a 13% saving to a 26% increase. At 50% load, floating saves 13%/11%/1% while a fixed 27 C raises energy 5%/10%/1% with an X-factor of 1.309. Uptime Institute assessments found CRAH airflow oversized 2.6 times on average; that cuts fan energy by the cube law, but its capex typically exceeds 10-year energy savings.

Key points

1. Model: 1 MW, 3 air-cooled chillers N+1, VFD dry cooler, 4 kW/rack, $0.10/kWh, 10-year TCO.
2. Floating 15.6-26.7 C vs fixed 20 C: total energy 1% saving in Chicago, 14% saving in Seattle, 12% increase in Miami.
3. TCO unchanged, 6% better and 6% worse respectively.
4. Cooling-only pPUE improved in all three cities (Miami 1.327 to 1.314) even where total energy rose.
5. TCO-optimal max float temperature: Seattle 27 C, Chicago 23 C, Miami 21 C.
6. Fixed 27 C: Chicago +13%, Seattle -2%, Miami +12% energy; 31% more failures (X-factor) everywhere.
7. Server airflow curve sensitivity: Seattle 36% saving to 10% increase; Miami 13% saving to 26% increase.
8. CRAH airflow oversized 2.6x on average; its capex typically exceeds 10-year energy savings.

Using it in design

• When the user raises the inlet setpoint, include server fan power and airflow increase (flat/low/moderate/high CFM-vs-inlet curve) and CRAH fan and count changes; warn when PUE improves but total energy rises.
• Use site bin weather data; offer fixed and floating setpoint modes and show the TCO-optimal max float temperature.
• Compute the annual-average ASHRAE X-factor from the setpoint profile.
• Require containment or blanking before raised setpoints; flag chillers whose max chilled-water temperature is below the required reset.

Who should read it

Cooling designers and operations managers weighing a higher inlet-temperature setpoint.

Related standards

• ASHRAE TC 9.9 Thermal Guidelines for Data Processing Environments
• ISO/IEC 30134-2

Publication details

Authors

Wendy Torell; Kevin Brown; Victor Avelar

Publisher

Schneider Electric

Year

2016

Document number

WP 221

Document language

English

Source type

Manufacturer

Access

Free to read

Licence

Copyright retained; citation only — the publisher’s terms apply to the original document

Citation