Brendan Leonard, Managing Director at STULZ UK, explains the benefits of free cooling and how it can proactively create efficiency and have a positive impact upon data centre optimisation.
Data centre cooling is a continually evolving science due to the amount of equipment variations possible and the number of options in terms of facility design and operation. There is now a plethora of cooling methodologies, from hot aisle to cold aisle, row containment, rack containment and water cooling but one of the most effective – both in terms of cost and application – is free cooling.
Data centres consume almost as much energy for non-computing resources as they do in powering their servers. That is why Power Usage Effectiveness (PUE) has become such a prevalent industry metric – the closer it is to 1.0 then the better the facility is doing in managing its use of non-computing energy – and free cooling can help lower this figure.
So what do we mean by the term free cooling? Gartner defines it as ‘any technique used to reduce the energy consumed by cooling systems or the time that the cooling units run by using the outside temperature of air or water to cool the data centre or other facilities’. Free cooling helps save energy and for most of the time UK ambient air temperature is colder than that required in the data centre – so a simple ventilation system can maintain compliant conditions.
Before looking at free cooling in greater depth, it is worth noting that there are various measurable temperatures present in a data centre including supply air, return air, room air, cold aisle, and server inlet and exit temperatures. Server inlet temperature and server exit temperature are well known to have a significant effect on the efficiency of any cooling system and the higher the former, the more efficient a cooling system can become.
Just as importantly, attitudes towards data centre temperatures have changed drastically over the last 20 years or so. Previously, the computer room air conditioner (CRAC) return air temperature was the main control parameter and was set to 22-24°C. This resulted in unnecessarily low supply air temperatures in a range of 12°C. Today, the best practice is very different. In a modern data centre, supply air temperature is the main control parameter and it is typically set between 18°C and 27°C, as recommended by ASHRAE TC9.9. This has resulted in return air temperatures to the cooling equipment of 25-40°C.
Although the concept behind free cooling is relatively straightforward, there are a number of ways to go about achieving it.
However, in urban areas air quality might not be good enough and is often full of particles, which can be very hostile to modern IT equipment. Therefore, direct free cooling should only be used where filtering can control the quality of outdoor air. Also, if the outdoor air is too cold, a certain proportion of warm air from the data centre must be mixed in with it to supply controlled, tempered air to the IT equipment.
Humidification also has to be controlled and methods to humidify or dehumidify air can also be very expensive and complex, with large ductwork systems often required. In addition, large openings in the building fabric are needed to bring in the outdoor air, plus an equally large ductwork system and fabric opening for the exhaust air. These openings and ductwork bring a security risk to the data centre that has to be considered and in the event of an external situation such as a fire, the system must have the ability to operate independently of the air inlet.
Single stage indirect free cooling is based upon an air/water heat exchanger placed on the hot water return of the chiller. Outside air is blown across the heat exchange and helps removes the heat before it gets to the chiller. Therefore, with a return water temperature of 20°C, if the outside temperature is 19°C or less, it starts reducing the activity of the chiller and reduces running costs.
Two stage indirect free cooling systems have a considerably smaller footprint than their single stage counterparts, but are less efficient, as they function with two heat transfers. In the first heat exchanger, the heat from the air in the data centre is transferred to a liquid, normally glycol. This is then pumped to the outside in relatively small pipes, where heat from the data centre is rejected to the outdoor air via a second heat exchanger.
Air contains a certain amount of water and the total amount depends on the air temperature and barometric pressure. As water absorbs into the air, an adiabatic process takes place – this means the temperature of the air decreases, while the energy content of the air remains unchanged.
By using adiabatic assistance in direct free cooling systems the air that enters the data centre is cooled before entering. Indirect free cooling and adiabatic assistance is more conducive however, as the adiabatically assisted outdoor air never enters the data centre. In single stage indirect free cooling systems, adiabatic assistance reduces the temperature of the outdoor air entering the air/air heat exchanger and reduces the amount of mechanical cooling required to satisfy the supply air temperature to the IT equipment. In two-stage indirect free cooling systems, the warm outdoor air can be cooled by adiabatic assistance before entering the dry cooler to either increase the hours of usable free cooling, or extend the mixed mode at the upper limit to reduce the use of mechanical cooling.
As the density of installed equipment in the data centre has risen, so too has the amount of heat generated. While being able to fit more kit into a smaller space is generally considered a good thing, the need to control temperature has led to the growing use of free cooling. It should be remembered that no free cooling system is perfect though, and all have limitations. It is therefore important to evaluate what kind of free cooling is best suited to the application on a case-by-case basis.