One of the main pillars in a Data Center is its cooling capacity for the servers in the IT rooms. The market offers several methods, being the main ones the implementation of cooled air or water. Without an effective cooling system, the electrical components inside the servers hosted in a Data Center can suffer of thermal throttling, causing a slowdown in its processing capacity, and – in the worst-case scenario – causing breakdowns or even generating fires.
The evolution of the cooling techniques answers to the new needs derived from the increasing demand in network capacity, in the volume of data processed globally and in the computing capacity. This translates into an increment of the capacity of the installed servers and, therefore, and increment of the heat generated.
Nowadays, the trend in Data Center cooling combines both techniques, always supported by free-cooling systems, to increase the efficiency.
Air cooling
Air cooling was one of the first techniques implemented in a Data Center. It is based on lowering the temperature of the servers through the impulsion of cooled air. Generally, it works in two different ways:
Cooling through external air
Taking air from the environment, it is cooled and filtered to, later, impulse it to the servers, expelling the heated air in the process. Logically, this technique is more common in colder environments.
Cooling through recirculated air
This technique takes advantage of the hot air expelled by the servers, directing it to Air Treatment Units, where it is conditioned and finally redirected to the servers.
Water cooling
In this case, the element used for cooling the servers is water. In this method, we can find two techniques:
Direct water cooling
The cooled water is directly administered to the servers or racks of servers.
Immersion cooling
The servers are submerged in cooled water depots.
Experimental methods
There are other methods, less common or in experimental phase, which are being tested to improve efficiency and sustainability of scale cooling systems. The most outstanding being:
Immersion cooling
This method requires the immersion of the electronic components directly in a mineral oil or a special refrigerant liquid. These types of liquids are dielectric non-conductive, which guarantees the security of the systems. The liquid absorbs the heat generated by the components and it is cooled through dissipation, recirculating it again. It is a very water-efficient systems, but with a high cost.
Evaporative cooling
In this method, water is evaporated to cool the air. First, water is sprayed in panels through which air comes out, achieving the absorption of the heat. Finally, this air is redistributed in the Data Center.
Which is the best system?
The selection of the best server cooling system in Data Centers will always be the subject of an engineering project that factors in all the parameters involved: IT room design temperatures, server temperatures, IT room interior conditions, exterior conditions, system availability at regional level, maintenance availability, etc.
In all cases, the Net Zero trend is the common practice. This trend promotes minimum energy consumption (typically electrical energy), and its practical application in Data Centers means that, ideally, the only electrical consumption is that of the servers themselves.
The greatest energy consumption in data centers – excluding servers – is in the cooling systems, so the goal is to have these systems consume as little as possible. Since servers must have stable temperature conditions 24 hours a day, seven days a week, 365 days a year, systems must be selected to minimize their power consumption throughout the year, so often the combination of several systems is the best solution. Given the current trend of servers working at higher temperatures, systems that allow greater use of free cooling (free cooling), either direct or indirect, are currently the most widely used. Adiabatic cooling systems (evaporative type) are the most advantageous in this regard, although the water consumption of these systems may be an obstacle depending on the region where they are implemented.