Drought, Dystopia And Data Center Design
It’s August in America and angst is in the air.
Obviously, it’s August everywhere, but in America, it’s a time that creeps up like a sweaty surprise, even though it’s right there on the calendar.
While Europeans have the common sense to vacation during the month of August, we Americans muddle along, and for the most part shuttle from air-conditioned space to air-conditioned space like ants avoiding the bright beam from a massive magnifying glass in the heavens.
Weathermen across the nation breathlessly tell us how hot it is and, more importantly, how hot it actually “feels.” Global warming aficionados (both pro and con) point out the latest examples to prove or disprove the trend. And then there are the many stories on drought.
Never mind that drought conditions exist year-round. In August, it just makes for better TV to do stories on unhinged homeowners washing their cars, or to show water skiers traversing the lake surface well below the reservoir’s bathtub ring.
But even though I hate the overheated hype, I will use this month of despair to jump on the bad news bandwagon and offer some insights on water usage and data center design.
What’s the Deal with Water?
Most of us have heard some version of the following stanza from The Rime of the Ancient Mariner:
Water, water, everywhere,
And all the boards did shrink;
Water, water, everywhere,
Nor any drop to drink.
When 70% of the earth’s surface is covered with water and 97% of that water is problematic salt water, those last words are a cruel accounting of our situation. Irony joins pain when you consider that here in the Midwest, we have seen record rainfall and flooding this year while the West experiences the greatest shortfall in 40 years.
For states that depend heavily on hydro-generation like California and Nevada, low reservoir levels can lead to power shortages. Water shortages affect us at the municipal level when water is rationed between agricultural, domestic, and industrial users.
And it isn’t limited to the U.S. Dry conditions in Brazil in the upcoming dry season will likely affect data centers negatively, due to a lack of makeup water and impacts on the power grid thanks to hiccups at hydro plants.
And if Mother Nature weren’t tough enough, the philosophical progenies of Mother Jones can wreak havoc as well. In 2014 for example, lawmakers threatened to cut off the water supply to the NSA’s massive Data Center in Utah because they viewed NSA’s mission as unconstitutional.
We may take water as a given, but you can see that drought, politics, and civic priorities (agriculture v. domestic v. industry) all contribute to muddying the water. [insert rim shot here]
While electricity is often seen as the base utility, with modern data centers requiring thousands of gallons of make-up per day (40 to 80 L/min (10 to 20 gpm) per MW), water might in fact be the more integral necessity.
Taken one step further, when you adopt a plant-to-plug mindset, and consider the quantity of water required and consumed at the source of generation, water’s importance is elevated even further.
Waterless Urinals at the Thunderdome
What does the future hold? For my part, I don’t see the dire dystopia that some predict. If the imagined Mad Max Dustbowl is in fact a coming reality, it’s pretty likely that HVAC design will fall by the wayside. In turn, we will all have more pressing matters to attend to.
However, there likely will come a day when emerald golf greens in Phoenix and colossal dancing water fountains in Las Vegas will be memories from waters golden age. It’s probable that the silliness of bottled water will eventually dawn on us. And it’s a good bet that engineers will ultimately make the oceans water usable.
But as long as we are a society that believes using 1.1 gal of fresh water instead of 1.6 gal of fresh water to flush a toilet is significant progress, don’t expect to be sucking pebbles any time soon.
With that said, we still have an obligation as good designers and stewards of the environment to minimize our footprint and use our resources judiciously. And in the mission critical environment, we are further mandated to provide designs that reduce risk and assure uninterrupted operation.
So, just as we do with energy, as good mechanical engineers we should reduce the amount of water our systems require.
Measure to Manage
Luckily, for all of us, someone else has been working this issue. And more importantly, they have developed a metric that will allow the designer to assess both the risk and the value of water use and conservation. By them, we are referring to the Green Grid, and the metric provided is Water Usage Effectiveness (WUE).
Like Power Usage Effectiveness (PUE), the WUE metric uses the familiar value of IT Equipment Energy as its denominator. Once determined for PUE, the same value would be used as the denominator for WUE as well.
PUE is dimensionless and has an ideal value of 1.0, indicating all energy consumed is IT-related only. WUE on the other hand has dimensions (liters per kilowatt-hour [L/kWh]) and the ideal value is 0.0; indicating that no water use is associated with the data center’s operations.
There are two versions of the metric. WUE is site-based and includes water used for humidification and water evaporated on-site for energy production or cooling.
WUE = Annual Site Water Usage
IT Equipment Energy
WUEsource is source-based and includes water used on-site and water used off-site in the production of the energy used.
WUEsource= Annual Source Energy Water Usage + Annual Site Water Usage
IT Equipment Energy
While the site metric should be straightforward to determine, the source data is a bit trickier.
Power Production and Water Consumption
Intuitively, we know that water is in the power equation. But most conversations regarding the plants that serve the grid focus on the fuel used or the means of production be it coal-fired, nuclear, hydro, or renewable. Water use is somewhere in the fine print, if at all.
Thankfully, the National Renewable Energy Laboratory (NREL) developed a useful resource for determining the intensity of water use across the country: The Water Consumptive Use for US Power Production report. This report uses the Energy Water Intensity Factor (EWIF) to contrast and compare different technologies as well as regions in terms of water consumption (see Tables 1, 2 and 3, and Figure 1).
With EWIF available, the WUEsource can be calculated as follows:
WUEsource= [EWIF x pue]+ Annual Site Water Usage
IT Equipment Energy
But this information is shared with a few caveats. First, this is aggregated state data based on power generated throughout each state. So even though Massachusetts looks sweet with an EWIF of zero, if you are building in Boston, power must be generated somewhere. Last time I checked, Massachusetts was not off the grid (p.s. Boo, Patriots).
Second, this accounting does not factor in the risks associated with states that may be dominated by hydroelectric generation. So California may look good with an EWIF of 0.19, but if you consider hydro, the aggregate number would jump significantly.
And last, this data is over ten years old. So it’s great for high-level scoping decisions, but you would really want to get into the local nitty gritty when making final decisions on siting and strategies.
Also note that if you generate or collect your own power on-site via co-gen, tri-gen, or through renewable sources, it will render the regional data moot.
But what can we quickly glean from the basic info?
- Renewable energy sources are obvious candidates when water use is a consideration.
- Natural gas-fired plants are relatively efficient, water-wise.
- Projects located in the Western United States will likely have a lower EWIF.
- The Rust Belt is a particularly water intensive power corridor.
- And lest we forget, PUE still matters.
Wringing out the Water
So, we have metrics that help us incorporate water consumption into our design optimization algorithm. The Green Grid provides guidance on a third metric for Carbon Usage Effectiveness (CUE), but at this time, data is lacking and in this author’s opinion, it is too early to embrace that measure fully.
With that, what should or can we do to reduce our WUE?
Not surprisingly, most of the strategies we employ to reduce our PUE will typically cut our WUE. So all of the good design strategies that have been developed over the last few years still apply, including but certainly not limited to:
- Employing hot-aisle cold-aisle separation strategies (containment, curtains, etc.).
- Aggressively managing airflow within the data center through intelligent placement of perf tiles, attacking hot and cold spots, and limiting leakage.
- Expanding the environmental parameters within the data center, both temperature and relative humidity.
Regarding the environmental conditions in the data center, ASHRAE’s guidance has been out there since 2011. And now ASHRAE has even more data to support lower humidity levels. But even though the big guys are running hotter (Google, Yahoo, Amazon, et al), the typical enterprise operators are still reticent.
All we can say at this point is that lower humidity and higher temperatures are not the bleeding-edge anymore. As designers and operators, we should all commit to spending less time making excuses to avoid the new set points, and instead invest more energy into finding ways to embrace them.
When the data center feels more like a warehouse than a surgical sweet (a tough image to accept), cascading opportunities arise. Chilled water can become less so, meaning chillers may be eliminated or simply used for trim on the worst day. Free cooling via water-side or air-side economizer (both direct and indirect) is in play. And although it doesn’t solve the water equation completely, direct and indirect evaporative cooling strategies can cut water use by more than half when compared to legacy chilled water systems.
Cooling Tower Optimization
The biggest culprit in water usage for the typical data center is the cooling tower. Whether it’s a traditional chilled water system or a compressorless system using the towers to maintain a warmer cooling water temperature, evaporation, drift, and blowdown all conspire to turn the tap on.
Entire articles can and have been written on the topic of tower optimization. So we will touch on the highlights here, but provide direction to a greater resource. Specifically, Chapter 6.3 of the EPA’s WaterSense at Work e-booklet .
The basic steps include the following.
- Maintain system efficiency.
- Monitor the water chemistry and flow.
- Work with a vendor who focuses on conservation.
- Minimize blowdown and maximize the cycles of concentration (CoCs).
The ratio of evaporation to blowdown is called the CoC. And if possible, the towers should be operated at a CoC of six or more for maximum water-efficiency. The quantity of blowdown can be controlled and the CoC increased, using an automated conductivity controller and maintaining proper water chemistry.
Table 4 shows the percentage of make-up water savings that can be expected by increasing the CoC, while Figure 2 shows how increasing cycles of concentration can decrease water use in a 100-ton cooling tower. The actual CoC should be determined with the vendor, based on water available and the equipment employed. The opportunities are significant.
What Would Sarah Connor do?
In 2009, James Hamilton, VP and Distinguished Engineer at Amazon, stated, “Water is tomorrow’s big problem. The water consumption (in data centers) is super embarrassing. It just doesn’t feel responsible. We need designs that stop using water.”
Not only does James have the best hair in data center engineering, he was spot on six years ago. And the only caveat to the quote is that tomorrow may be today.
Do you know what all of those apocalypse movies have in common (beyond bad acting)? The good guy always wins.
Engineers are the good guy, and we have to find ways to continue to support the information age without draining the global bathtub. Just as it was unimaginable to have a compressorless data center 10 years ago, it may seem like a waterless data center is a daft dream today. But it really is a realistic goal … and not just in some odd one-off experiment.
I’m pretty sure we can do this. ES
Dickens is a Design Principal for Jacobs Global Buildings, focused on the Mission Critical and Government Sectors. He can be reached at firstname.lastname@example.org.
a. Modenesim, Sidney. 2015. “Running on Empty in Brazil.” Datacenter Dynamics, 04 (02): 16-17.
b. Carlisle, N. (2014, November 19). Shutting off NSA’s water gains support in Utah Legislature. The Salt Lake Tribune.
c. The Green Grid. 2011. “White Paper #35. Water Usage Effectiveness (WUE™): A Green Grid Data Center Sustainability Metric.”
d. Consumptive Water Use for U.S. Power Production, P. Torcellini, N. Long, and R. Judkoff, 2003, NREL/TP-550- 33905, http://www.nrel.gov/docs/fy04osti/33905.pdf
e. Beaty, Donald. 2015. “Are Data Centers Drying Up?” ASHRAE Journal, 57 (03): 72-77