In Part I of this article I discussed the fact that many data centers use an enormous quantities of water as part of the cooling process. Paradoxically, it seems that all the “Green the Data Center” type organizations that are leading the charge to improve efficiency of the data center; most noticeably the “Roman High Council” (a.k.a. The Green Grid), and the “US Environmental Protection Agency” (a.k.a. The Roman Senate), have totally ignored water usage in their efforts to promote and measure efficiency.
The widely used PUE measure put forth by The Green Grid, has done a great deal of good by raising widespread awareness of energy efficiency in the data center. However, it has been out for almost 2 years and has been amended and enhanced several times, yet it still does not make any attempt to factor in water usage. Earlier this year I mentioned this to one of the members of The Green Grid who was speaking at Data Center Dynamics in New York City. I also posted this on the Data Center Dynamics Linked-In page, with no results:
I am merely proposing that water usage should be considered….. I decided to create my own metric (beta ver. 0.9)…
Existing…PUE= (Total Power / IT Load)
I propose….
PUE-W= PUE x WUF (Water Usage Factor)
WUF = 1+ (gallons per hour / KW)
I recognize that I am mixing time based factor “gallons per hour” within the other variables that are non time related values. Nonetheless, this is only a simple example something that could be more fully developed. The WUF (Water Usage Factor) may need some tweaking as to gallons per hour GPH vs KW, but it at least it reflects basic water consumption compensation.
Here an example of what this would look like:
PUE = (200KW Total Power) / (100KW - IT Load) = 2.0
PUE-W = PUE x WUF > [WUF = 1+ (GPH/KW)]
Example: the 200KW site has a “standard” PUE of 2.0 and it uses 20 gallons per hour, therefore:
PUE 2.0 x WUF (Water Usage Factor)
WUF = 1+ (20GPH / 200KW) = 1+ (20/200) = 1.1
PUE-W = PUE (2.0) x WUF (1.1) = 2.2
Of course for those systems that do not use water, the WUF factor would always be 1.0, thus the PUE-W would equal the basic PUE.
I recognize and acknowledge that this is not a perfect formula. My intention is to spur the engineers who design cooling systems to help create a better formula to more accurately reflect the “WUF”.

Moreover, in September the US Environmental Protection Agency (EPA), home of the Energy Star program, just completed their study of 108 large US data centers. They are now in the process of finalizing the first “Energy Star for Data Centers” certificate program (actually it will be an actual plaque that will be issued to the Data Centers that meet the pending requirements). This was an 18 month study that looked at Energy usage (not power – a story into unto itself - The EPA Energy Star Program has officially coined the term “EUE” Energy Usage Effectiveness.). Unfortunately, this new Energy Star program has also totally ignored the issue of water usage.
I understand that measuring “energy” efficiency is the easiest method to promote efficiency. However, it seems very shortsighted to not factor in water usage in calculating the efficiency of a data center, especially by the very organizations whose primary mission statement and names imply and invoke the mantle of “Green” and “Environment”. In an age where virtually all new buildings are judged by the Green Building Councils LEED standard, which looks at every bit of material used in construction and every building’s energy and resource usage, it seems that the “Green Data Center” does not want to consider water as a resource to be a factor in declaring their “Greenness”.
So why not stop using water based cooling and just use air cooled chillers or direct expansion (DX) systems? They use no water, but do use more energy and are larger, since they need to use larger air cooled heat exchangers, instead of the evaporative process. They are also more expensive (per ton of cooling) since the exterior heat exchanger is more expensive to manufacture than a comparable cooling tower. They are used in many smaller (i.e. under a MW) installations especially where it is impractical to install or operate a cooling tower. However, some forward thinking organizations are considering using a combination of air cooled chillers and some DX systems to minimize water usage, as well as air and water side economizers to save energy. I believe that just as energy costs and supply limits forced the energy efficiency to the forefront, water will become a more visible factor, as it gets more expensive and more constrained.
As we enter the second decade of the new millennium we need to re-evaluate the significance of water usage when gauging the efficiency of a data center. So at the risk of invoking the wrath of the “Roman High Council”, as we build more and larger data centers (sometimes just so anyone can watch a 15 minute HD video clip of someone’s cat sitting on a couch), think about excesses and the Roman Empire. Then please consider how much water your data center will use (and perhaps its source), when designing and building more “Aqueducts” to cool your next “Coliseum”.
The views expressed are those of the author and are not those of DatacenterDynamics.