Raising server room temperatures to improve data center energy performance

Cooling systems, essential to keep servers operating under optimal conditions, account on average for 30% to 50% of data centers’ energy spend. For a long time, the temperature commonly accepted by IT room operators was around 22°C ambient, resulting in 15 to 16°C supply air temperatures from the air-conditioning units. This required significant electrical power for cooling and generated high energy costs.

Since 2015, ASHRAE (American Society of Heating, Refrigerating and Air Conditioning Engineers) has recommended server inlet temperature ranges of 18 to 27°C, with relative humidity between 8% and 60%. For most recent equipment, temperature recommendations even extend from 15 to 32°C, with relative humidity between 8% and 80%. In other words: for the past few years, data center temperatures can be raised without compromising IT service continuity.

This increase in operating temperatures opens up new possibilities for data center operators. It makes it possible to increase IT room density without replacing air-conditioning equipment, and therefore without having to make additional investment. It also makes it possible to significantly reduce energy consumption with the same IT footprint.

Higher server inlet temperatures also make the use of free cooling or free chilling (systems that use outside air to blow cool air into the room or to cool water instead of using the chilled-water plant) far more attractive, especially in temperate regions such as France. Indeed, with a room temperature setpoint of 25°C instead of 15°C, the periods of the year during which free cooling can be used without turning on air conditioning increase considerably. This generates significant energy savings and improves PUE (Power Usage Effectiveness). The same applies to free chilling, which can be used more frequently throughout the year to cool water loops, as temperature setpoints are now set at 15°C instead of 7°C for water.

While direct or indirect air-side free cooling has become more widespread in recent years, particularly in large data centers, it remains constraining from a building standpoint because it requires space to install the ducts that bring air into the rooms. It also requires a certain level of expertise to operate properly. Implementing a free-chilling cooling solution often proves more suitable, particularly for data center renovation projects equipped with a chilled-water plant. For example, for a data center located in Lyon, by setting a room supply air temperature of 27°C and chilled-water temperatures of 15–20°C, it was possible to achieve a return on investment in under two years.

Raising the overall data center temperature requires a thorough understanding of the specificities of the cooling equipment and the IT hardware hosted in the rooms, and control of all the airflow circulating in the data center. Temperature must be controlled at the server inlet, not the room ambient temperature, nor even the return air temperature at the air-conditioning units. Poorly controlled airflow, even on a single rack, can cause “thermal contamination” across the entire room. Hence the need to structure the data center into cold aisles and hot aisles to ensure that cool air is properly delivered to the servers and that hot air, generated by the heat dissipation of IT equipment, does not mix with the air intended for cooling. This involves implementing containment and sealing systems that effectively control airflow circulation: blanking panels to cover unused spaces in a rack or various openings, cable grommets to limit airflow around cables, etc.

An energy optimization project for a data center is often complex because, most of the time, it must be carried out on live sites subject to service continuity requirements. It requires an in-depth audit and study phase to take into account all data center parameters: floor area, equipment power, load rate, physical infrastructure, cooling systems, airflow, environmental constraints (dust, humidity level, etc.).

In all cases, implementing such a project, as well as the resulting energy savings, depends on the characteristics and infrastructure of each site, the level of upstream control (regularity of monitoring, performance indicators calculated, proximity of project teams, etc.) and the operating policy adopted downstream (process definition, training, resources mobilized, etc.).

By Nicolas Miceli, Head of Energy Performance and Innovation, APL