The data centre industry’s rapid expansion in recent years is well documented and shows no sign of abating. However, with great power comes responsibility. A recent report suggests that by 2025, data centers will consume one-fifth of all the electricity in the world. Stats like this have served to bring the industry under the spotlight. Data centers have come under increasing pressure in recent years to reduce their carbon footprints and operate more efficiently.
The primary sources of energy consumption in most data centers are server operations and cooling. Much of the energy usage comes from large industrial equipment, such as pumps, chillers, and cooling towers, which are used to regulate data center temperature.
Metrics, like PUE, have helped steer the industry down a greener path, but it doesn’t tell the whole story. The use of water in data centers is increasingly coming under the microscope.
Global Water Crisis
It is said that we live on a blue planet, yet only 1% of the water it holds is actually useable. The U.N. estimates that $50 billion to $60 billion needs to be spent annually between now and 2030 to avoid future water shortages. That’s not to mention that, in 2015, the World Economic Forum in Davos listed “water crises” for the first time as the world’s leading threat.
Water is essential to achieving climate targets because climate change affects the availability, quality, and quantity of water for basic needs, such as industry, agriculture, and domestic use.
Water Usage in Data Centers
In recent years, optimization of supply air temperatures in data centers has been built on with the introduction of adiabatic cooling systems. This technique incorporates both evaporation and air cooling into a single system. The evaporation of water, usually in the form of a mist or spray, is used to pre-cool the ambient air to within a few degrees of the wet bulb.
In all, data centers throughout the U.S. were projected to consume about 174 billion gallons of water in 2020 (U.S. Department of Energy report). In the 2018 fiscal year, Google used about 4.2 billion gallons of water. Some went to offices, but the majority was consumed by its global fleet of data centers. During that same time period, Microsoft used about 1 billion gallons, also with the majority going to its data centers.
A 1-MW data center using adiabatic cooling can use around 6.5 million gallons of water per year, yet less than a third of data center operators track any water metrics, and water conservation is ranked as a low priority.
The water usage itself is not even the whole story. This water still has to be stored and treated, which increases capital costs and, as with any mechanical equipment exposed to continuous water contact, cooling plants can suffer from increased degradation, putting strain on opex costs too.
Answering the need for cooling systems that provide similar efficiencies that can be achieved with adiabatic cooling but with a more sensitive approach to water conservation, water-side optimization (WSO) has been presented as an alternative method for data centers.
The philosophy of WSO is based on taking an optimized air environment and looking at what other variables can be adjusted in order to deliver more free cooling. Assuming the air within the white space stays at the same temperature, the next step would be to reduce the approach temperature while opening the difference between water supply and water return. Implementing innovations within the plant equipment means the supply and return air remain as before, but supply and return water temperatures are higher, thus the approach temperature is reduced. There is a fixed temperature difference on the air side with the fluid side being opened out to 21°F and the approach temperature closing from 7°, all delivering energy efficiency benefits.
To achieve this, free cooling chillers are matched to chilled water coils in indoor CRAC units or fan walls. The air path is simplified using hot-aisle containment, creating a pressure differential that draws cool air through the servers, out of the white space via ducts, and back to the air conditioning plant via a common plenum. The air is introduced directly to the space via side wall diffusion, minimizing air-side pressure drops.
The benefits of this include less mechanical cooling, meaning more efficient chiller operation; lower fan speeds, which results in more efficient indoor unit operation; lower pump power that increases water transfer efficiencies; and large coil surface to increase cooling capacity in a reduced footprint.
This is all managed with an intelligent controls platform that monitors fluctuating demand within the white space and dynamically operates the system at its most efficient operating point.
Based on temperate climates, 14% more free cooling (59% in total) can be achieved with (WSO) with all but 1% of the rest of the year being covered by concurrent cooling (a combination of free cooling and mechanical), giving huge benefits in terms of chiller efficiency.
We have discussed data center HVAC architecture in terms of heat exchange, but temperature is not the only issue to consider for technology providers. IT equipment is sensitive to pollutants and particulate matter, so managing the purity of the air supplied into the data hall is an essential task.
Two global megatrends, industrialization and urbanization, are impacting a third trend: digitization. The rapid expansion of digital infrastructure is being compromised by the by-products. Air quality across the world, but particularly in urban areas, is compromised by pollutants, such as sulphur dioxide and hydrogen sulphide, both of which contribute to the corrosion of electronic components. The rise of data centers at the edge of load centers, i.e., towns and cities, means that it is more critical than ever to effectively manage air quality.
HVAC systems tend not to draw in outside air, especially in larger facilities, relying instead on constant recirculation. Many data centers rely maintaining positive pressure in the building to keep clean air inside. When a door is opened to admit a technician, air is forced out due to the air pressure in the hall being greater than atmospheric pressure in the rest of the building, preventing dirty air from entering.
Positive pressure is usually maintained via an air-handling unit fitted with high MERV filters, supplying clean, fresh air into the space from outside. Pressure within the hall is monitored using sensors, which integrate with a building management system to control the operation of the air handler. Care must be taken to replace or clean filters regularly, as the build-up of particulates on filters can block them or, at the very least, decrease the energy efficiency of the plant. The life span of a filter is influenced by factors, such as weather and the outside environment, so it is not enough to simply replace them as part of a periodical maintenance plan. Sensors can be used to detect blocked filters by measuring pressure drops across them.
Filters are also deployed on HVAC equipment, like CRAH units and fan walls. Care must be taken when designing such equipment to ensure that filters are positioned to reduce the pressure drop as much as possible. This type of equipment tends to run 365 days per year, so any extra work the fans have to perform to draw air across the filters can cost thousands of dollars in increased energy bills.
Optimally positioning the filters in units could save 12 kW of fan energy in some cases. Industry professionals have calculated that this alone could deliver a PUE improvement of 0.05, which, although seemingly small at face value, can result in significant cost savings over the life cycle of a facility.