©2014 This excerpt taken from the article of the same name which appeared in ASHRAE Journal, vol. 56, no. 10, October 2014.
By Mukesh K. Khattar, Ph.D., Fellow ASHRAE
About the Author
Mukesh K. Khattar, Ph.D, is energy director at Oracle in Redwood Shores, Calif. He is a past member of ASHRAE TC 9.9.
Six years ago, using outdoor air in cold climates for data centers was limited. Although free cooling from outdoor cold air, or air-side economization, is as old as civilization, data centers have been reluctant to adopt this technology. This may be because outdoor cold air is naturally dry, and when it picks up heat in data center spaces, its relative humidity drops to levels where humidification is needed to protect the IT equipment from static electricity.
Though some advocate maintaining a much lower level of indoor humidity in data centers, which would reduce need for humidification (ASHRAE TC 9.9, Mission Critical Facilities, Data Centers, Technology Spaces and Electronic Equipment is revising its recommendations to lower recommended and acceptable humidity levels to increase use of air-side economizers), users have yet to embrace it.
The conventional approach to humidifying large quantities of air requires injection of steam into air, typically from boilers or pan heaters. The generation of steam requires energy. The industry was quick to learn from its experience that it required more energy to humidify dry outdoor air than the energy saved from the use of its free cooling, and air-side economization did not make much headway in data centers. The concern with airborne contaminants and the need for increased filtration also were added reasons.
These problems were addressed efficiently in an innovative system installed in a new state-of-the-art data center in West Jordan, Utah. The first of the four 7.2 MW master planned supercells was built, leaving room for future expansion. The 25,000 ft2 (2323 m2) data center is supported with a 95,000 ft2 (8826 m2) structure to house infrastructure equipment and a 44,000 ft2 (4088 m2) office space. The system has operated for more than a year, and its operational fine-tuning is ongoing. Field-monitored data for energy use by cooling equipment (air circulation fans and the cooling plant) for the main data center hall as well as the supporting UPS hall as a fraction of the IT equipment energy use from August 2012 to July 2013 are summarized in Table 1.
The cooling energy use as a ratio of the total IT equipment energy use varies from a low of 1.05 in cold weather to a maximum of 1.25 in summer, depending on the ambient wet-bulb temperature. Its partial cooling-only power utilization effectiveness (PUE:1 measured as a ratio of all cooling energy consumption divided by the IT equipment energy consumption) was measured at less than 1.05 for February and less than 1.10 for the entire year as summarized in Table 1 and shown in Figure 1.
Figure 2 shows a plot of the operating partial cooling-only PUE versus daily average ambient wet-bulb temperature. The cooling energy use consisted of three components: fans for air distribution to the data hall; fans for air distribution to the UPS hall; and chiller plant to distribute trim cooling to the air-handling units. Total cooling energy required to remove each kWh of the IT equipment generated heat was quite low during periods of low ambient wet-bulb temperatures when free cooling and humidification were available but increased when ambient wet-bulb temperatures increased as evaporating cooling needed to be supplemented with trim cooling of ambient air.
In our design and operation, we elected to operate our system with free cooling when ambient wet-bulb temperatures were generally below 52°F (11°C) (74% of hours annually) with supply air of less than 59°F (15°C). Our design rationale is discussed later. It is thought possible to obtain free cooling even at higher ambient wet-bulb temperatures approaching 65°F (18°C) (>99% of the time in this climate) since cold air can still be supplied at 72°F (22°C) with a 7°F (4°C) approach with evaporative cooling effect.