Elsevier

Atmospheric Environment

Volume 42, Issue 24, August 2008, Pages 5978-5990
Atmospheric Environment

Particle concentrations in data centers

https://doi.org/10.1016/j.atmosenv.2008.03.049Get rights and content

Abstract

Cooling buildings with large airflow rates of outside air when temperatures are favorable is an established energy-saving measure. In data centers, this strategy is not widely used, owing to concerns that it would cause increased indoor levels of particles of outdoor origin, which could damage electronic equipment. However, environmental conditions typical of data centers and the associated potential for equipment failure are not well characterized. This study presents the first published measurements of particle concentrations in operating data centers. Indoor and outdoor particle measurements were taken at eight different sites in northern California for particulate matter 0.3–5.0 μm in diameter. One of the data centers has an energy-efficient design that employs outside air for cooling, while the rest use conventional cooling methods. Ratios of measured particle concentrations in the conventional data centers to the corresponding outside concentrations were significantly lower than those typically found in office or residential buildings. Estimates using a material-balance model match well with empirical results, indicating that the dominant particle sources and losses have been identified. Measurements taken at the more energy-efficient site show nearly an order of magnitude increase in particle concentration when ventilation rates were high. The model indicates that this increase may be even higher when including particles smaller than the monitoring-equipment size limitation. Even with the increases, the measured particle concentrations are still below concentration limits recommended in industry standards.

Introduction

Data centers house the vast amounts of equipment that provide the computational power, data storage, and global networking integral to modern information-technology systems. The high concentration of densely packed computers in data centers leads to floor-area-weighted power densities 15–100 times higher than those of typical commercial buildings (Greenberg et al., 2006). Data-center energy use has doubled in the last 5 years. In the US alone, it currently accounts for about 45 TWh y−1 of electricity consumption, >1% of total demand (Koomey, 2007). A substantial portion of the energy use in data centers, perhaps as much as half, is dedicated to cooling the computer equipment (Tschudi et al., 2004). The data-center cooling load can be reduced by a substantial fraction when large amounts of outside air are used to cool internal loads during favorable weather conditions (Sloan, 2007). However, many owners and operators are reluctant to use this cooling technique owing to concerns about the risk of equipment failure posed by introducing outdoor particulate matter into data-center buildings.

Fine particulate matter can deposit on electronic circuit boards in the space between isolated conductors. When the humidity of the surrounding air rises above the deliquescence point, particles composed of water-soluble ionic salts can absorb moisture and dissociate to become electrically conductive (Weschler, 1991). Empirical results show that exposure to high sulfate concentrations at high humidity can cause electronic equipment failure (Litvak et al., 2000). However, the risk of failure under the environmental conditions typical of data centers is not well understood. Owing to the competitive nature and high economic value of businesses in this sector, failure data are not publicly shared. Furthermore, the effect of introducing greater flow rates of outside air (or any other design change) on equipment failure cannot be predicted with confidence, because little is known about the concentrations of particles in data centers, the sources of those particles, or their fate once introduced into the data-center environment. This paper addresses these unknowns by measuring and modeling particle concentrations at operating data centers. The results provide a partial basis for assessing the equipment failure risk posed by particles for current data-center designs.

In the present study, time- and size-resolved particle concentration data were gathered over weeklong periods at eight different northern California data centers. Building parameters for three of these data centers were documented and a material-balance model was employed to predict concentrations under various conditions so as to better understand the relative influence of potential sources and fates of airborne particles. Predicted indoor concentrations were compared against the measured results. The loss mechanisms of filtration, deposition, and ventilation removal were compared to assess particle fate. The model was also applied to estimate indoor concentration levels for sulfate particles, which are of particular concern because of their ambient abundance and hygroscopicity.

Section snippets

Study sites

Size-resolved particle concentrations were measured as a function of time at data centers in eight different northern California cities. With respect to ventilation and cooling, all the data centers are conventional except for the one at the Sunnyvale site, which was specifically designed to be energy-efficient and therefore has distinctive characteristics. This article presents detailed results from three of the monitored data centers—at Rocklin, Walnut Creek, and Sunnyvale. The Rocklin and

Measured and modeled particle concentrations

Table 2 presents time-averaged, size-resolved, measured indoor particle concentrations for all eight data centers monitored. Average indoor concentrations for particles of diameter 0.3–5 μm average <1 μg m−3 in all conventional data centers and are substantially higher at the data center with an energy-efficient design (Sunnyvale). A closer evaluation of the results from Rocklin, Walnut Creek, and Sunnyvale follows. Fig. 4 presents the cumulative distributions of outdoor measured, indoor measured,

Conclusions

Prudent implementation of energy-saving measures that would expose data-center equipment to more outside air requires two tiers of investigation: first, understanding how these design measures would change indoor particle concentrations, and second, understanding how such changes in concentration would influence equipment reliability. This study contributes to the former goal by presenting the first published measurements of particle concentrations in operating data centers. The data and their

Acknowledgments

We thank David Faulkner for assisting with the particle monitoring equipment and the staffs at the data-center sites for their generous cooperation. This project was funded by PG&E (Contract PGZ-0601) and by the University of California Energy Institute, California Studies Grant Program. Most of this work was performed at LBNL under the U.S. Department of Energy contract no. DE-AC02-05CH11231.

References (27)

  • M.M. Lunden et al.

    The transformation of outdoor ammonium nitrate aerosols in the indoor environment

    Atmospheric Environment

    (2003)
  • K.T. Whitby

    The physical characteristics of sulfur aerosols

    Atmospheric Environment

    (1978)
  • ASHRAE, 1992. Gravimetric and dust spot procedures for testing air-cleaning devices used in general ventilation for...
  • ASHRAE, 2005. Design considerations for data and communications equipment centers. American Society of Heating,...
  • J. Brusse et al.

    Zinc whiskers: hidden cause of equipment failure

    IT Professional

    (2004)
  • W.J. Fisk et al.

    Particle concentrations and sizes with normal and high efficiency air filtration in a sealed air-conditioned office building

    Aerosol Science and Technology

    (2000)
  • Greenberg, S., Mills, E., Tschudi, W., Rumsey, P., Myatt, B., 2006. Best practices for data centers: Results from...
  • J.T. Hanley et al.

    Fractional aerosol filtration efficiency of in-duct ventilation air cleaners

    Indoor Air

    (1994)
  • J. Koomey

    Estimating Total Power Consumption by Servers in the US and the World

    (2007)
  • A. Litvak et al.

    Hygroscopic fine mode particle deposition on electronic circuits and resulting degradation of circuit performance: an experimental study

    Indoor Air

    (2000)
  • C.M. Long et al.

    Using time- and size-resolved particulate data to quantify indoor penetration and deposition behavior

    Environmental Science and Technology

    (2001)
  • J.B. Milford et al.

    The sizes of particulate sulfate and nitrate in the atmosphere—a review

    JAPCA—The International Journal of the Air Pollution Control and Hazardous Waste Management

    (1987)
  • Cited by (27)

    • Model-based optimization of free cooling switchover temperature and cooling tower approach temperature for data center cooling system with water-side economizer

      2020, Energy and Buildings
      Citation Excerpt :

      Free cooling exhibits considerable energy-saving potential in different climate zones [5–7] and can be divided into air- and water-side free cooling [3]. Air-side free cooling directly brings cold outside air into the data center or utilizes an air-side economizer (ASE); however, it may increase the air risk of using poor-quality air, as well the costs for equipment replacement [8–11]. On the contrary, water-side free cooling is regarded as effective and economically viable, as it utilizes a water-side economizer (WSE) [12].

    • Overview of direct air free cooling and thermal energy storage potential energy savings in data centres

      2015, Applied Thermal Engineering
      Citation Excerpt :

      Large parts of the industry are still reluctant to use this technology due to risk of contamination. But Shehabi et al. [6,7] showed that staying below the ASHRAE recommended particulate contaminant concentration for data centres can be accomplished with proper filtration. Many researchers have examined the potential of free cooling for different infrastructures such as office buildings [8], commercial buildings [9] and more recently data centres [1,10].

    • Energy efficiency and renewable energy integration in data centres. Strategies and modelling review

      2015, Renewable and Sustainable Energy Reviews
      Citation Excerpt :

      In that sense, Dai et al. [26–28] studied different strategies to mitigate the risks for IT and telecommunication equipment under this cooling policy. But, even though Shehabi et al. [29,30] showed that with proper filtration, the ASHRAE [8] recommended particulate contaminant concentration for data centres is accomplished and important economic saving can be obtainable, the industry has avoided historically using free cooling technology when dealing with these infrastructures due to the potential contamination risk. In the pursuit of green practices and reduced energy costs, some data centre operators are leveraging to use seawater and geothermal energy to produce cooling for their infrastructure.

    View all citing articles on Scopus
    View full text