Research from 10 server rooms

Project purpose
The project ran from June 2003 to May 2004. The purpose was to obtain detailed evidence about the electricity consumption and savings possibilities in 10 selected server rooms.

Project – background
A considerable proportion of the electricity used in office buildings is used in server rooms. This consumption equates to approximately 200,000 MWh per year, of which around 1/3 is used for cooling purposes.

Despite the focus on electricity consumption in individual workplaces, very little attention is paid to electricity consumption in server rooms, where the primary emphasis is on security and uptime.

Project – conclusions
Between1/3 and 1/2 of the electricity used in server rooms can be saved if you:

• Choose IT equipment with low energy consumption.
• Save on cooling by removing, or switching off, superfluous IT equipment.
• Organise and layout the room in a suitable manner.
• Use free cooling.
• Use good-quality cooling systems.

Types of server rooms measured
A conscious decision was taken to select very different types of server rooms, which is why both large and small server rooms in the public and private sectors were included in the survey. In addition, there are server rooms which are very server ‘heavy’ and others which are relatively server ‘light’. The total annual power consumption in the server rooms ranged from 11.6 MWh to 5,500 MWh.

Project – measurement methods
Server rooms in the pilot project were equipped with supplementary meters to register the power used by servers, disk drives, networks, other powered equipment and the cooling system. Temperature sensors were installed to measure the temperature of the air to the servers and to the cooling systems’ evaporators and condensers.

(1) Air temperature to servers in the selected server rooms was only measured for 1 individual server. Because of temperature variations in server rooms, one should ideally place the incoming airflow temperature gauge near each server. At CFI and ‘Rigsrevisionen’ the temperature of the cold air was measured up through the floor and under the racks.  
(2) The server room’s cooling efficiency is the total thermal load in the server room (kWh) divided by the power consumption of the cooling system (kWh). The figure reflects the efficiency of the cooling system itself, combined with how efficiently the cool air is ducted to the components that require cooling.

Read the report (in Danish)