Advanced Computing in the Age of AI | Tuesday, April 23, 2024

How To Reduce Water Consumption in Datacenter Cooling Systems 

Saving water may not be as straightforward as you think ...

Is water a scarce resource? In some places it is. It can also be costly and a drain on the environment (California cities, farmers, businesses and environmentalists are in a constant battle over how scarce water drained from the mountains is allocated.)

The Green Grid has developed the Water Usage Effectiveness (WUE) metric to help datacenter operators measure their progress in cutting water use. Water usage (in addition to carbon emissions,) says the Green Grid in its paper introducing the WUE, “are emerging as extremely important considerations in the design, location, and operation of data centers in the future.” The WUE is another metric in the attempt to “help the data center community better manage energy, environmental, societal, and sustainability parameters associated with building, commissioning, operating, and de-commissioning data centers.” it adds.

So how does a datacenter manage go about reducing water consumption? One way is to examine the design and type of cooling systems used. In a guest column at Data Center Knowledge, Ron Vokoun, director of the Mission Critical Div. at Gray Construction, discusses the issue of cooling systems.

First, he stresses the use of ASHRAE TC 9.9 for both temperature and humidity guidelines, including the use of Outside Air (OSA) economization techniques.

He also warns that, while there is a general assumption that evaporative cooling uses more water than a water-cooled chiller system, that is not always the case. In Phoenix, Arizona, for example, OSA with direct evaporation can use just 11 percent of the water of a water-cooled chiller with no evaporation.

In the attempt to save water, you might also end up using more energy. Again in Arizona, air-cooled chillers with no evaporation use more energy than water-cooled chillers with no evaporation. The cooling techniques used depend on the location of the datacenter.

EnterpriseAI