There is no such thing as a typical data centre. Each differs in size, number of installed servers, security and availability. However, one thing most of them tend to have in common is a variety of basic design and configuration flaws which prevents them from achieving their full potential cooling capacity. These problems generally go unrecognised as traditionally data centres have operated at power densities well below their design values. But recent increases in the power density of new IT equipment is now pushing data centres to their design limits and many are now incapable of providing effective cooling.

 The raised floor method has been the preferred option for many years as it is not only cost effective, but the technology is simple and flexible. CRAC (Computer Room Air-Conditioning) systems nonetheless do have a number of disadvantages; the airflow around the rack is critical to cooling performance and if the height of the raised floor is reduced or swirl mixing occurs (perhaps as a result of cable laying), the cooling performance can be diminished. Similar problems can also occur if further enclosures or additional outlet tiles are installed as the cooling now has to be shared even further which can alter the equilibrium decisively. Servers which had been receiving sufficient cooled air could develop ‘hot spots’ leading to sensitive equipment becoming endangered.

It is essential to ensure that the air intake and outlet openings to all the servers blow into defined cold and warm aisles, and that a sufficient body of cold air is available which cannot mix with any warm exhaust air before it is consumed. Although the cabinet is often thought of as having a supporting role is actually provides a critical function as it prevents hot exhaust air from the installed equipment from circulating back into the apparatus air intake. Most people believe that as hot air rises it will rise away from equipment however the extent of this effect is much greater than the degree of buoyancy of hot exhaust air.

A recent development has been the cold aisle containment system which takes into account the above issues by installing the climate control components at the top of the enclosure, away from the raised floor. This method prevents the cold and warm air from mixing as well as improving the efficiency of the CRAC system and limiting the energy demand.

EC-Controlled fans are now used in the newest generation of CRAC systems, from which huge energy savings of up to 30% can be achieved in normal part-load operation. By placing the fans within the double floor, additional space is also provided for the obliquely fitted heat exchangers, thereby unnecessary deflections of the air are avoided and flow resistance is minimised.

Proper airflow is essential to effective cooling, but is not sufficient alone. Proper layout of the racks is also critical to ensure that the temperature and quantity of air is available. It is well documented that by placing racks in rows and reversing the direction that alternate rows of racks face, recirculation can be significantly reduced. The distribution of loads can also stress the data centre capabilities. Pockets of heat loads or ‘hot-spots’ typically occur when high density, high performance servers are packed into racks that may give rise to adding more CRAC units or decreasing the air temperature setting point, which could lead to negative consequences.

The concept of redundancy also plays a central role within the data centre. Nowadays, almost all business processes are considered critical, as generally workflows can be displayed in an IT-based way. Technically speaking, data centres can be built in accordance with Tier II or Tier IV standards, which assure a certain level of availability. However, in many industries, the expenditure involved is in no reasonable relation to the risk, which makes it even more essential that the infrastructure is as secure as possible, even if that means making a compromise.

To minimise system failures and data loss a monitoring system which continuously monitors specified factors ensures any potential risks are quickly identified. Wireless I/O units are connected to a processing unit that collects information from specified sensors such as hotspots, smoke, fire, airflow or perhaps a blocked filter mat. Using this type of monitoring allows any problems to be quickly rectified before they interfere with the data centre operation.

Avoidable mistakes that are routinely made when installing cooling systems and racks in data centres or network rooms compromise availability and increase costs. These unintentional flaws create hot-spots, decrease fault tolerance, decrease efficiency and reduce cooling capacity. Technologies such as blade servers are a huge investment and it is vital that the cooling and related technicalities are tailored to suit the application – only in this way can a efficient and cost effective solution be reached which meets the performance demands.

 by John Wilkins at Rittal Ltd.

 Contact details

electrical industry news featured articles