Project 2016 – five percent energy savings in the data center
For four years now, we have been working in accordance with the ISO 50001 standard. This standard obligates certified companies to set up energy management systems and to use the results to develop measures to reduce energy consumption. In our 2016 optimization project, we set ourselves the objective of reducing the energy consumption of cooling by 5 percent (measured based on the total energy consumption of the data center). Three measures enabled us to successfully complete the project.
Measuring to improve
Two calibrated sensors constantly monitor the temperature and humidity at the hottest point in our cold aisles. These measurements provide a rich pool of data for accurately analyzing temperature fluctuations. 11,200 measurements per month are stored and evaluated for each cold aisle. This data was supplemented by the performance data of the air cooling units. Combining these two sets of data was the prerequisite for recognizing potential action areas.
We came to the conclusion that:
1. The temperature settings in the cold aisle, even at high loads, were too low.
2. The temperatures in the room outside the cold aisle varied greatly. Isolated hot spots had formed.
3. The workload of the air cooling unit’s ventilators varied since they had to work very hard in some parts of the room.
Step 1: Higher temperatures in the cold aisle
For a long time, hardware manufacturers in particular demanded room temperatures of 18 to 20 degrees Celsius for the safe operation of IT equipment. But these reference values are outdated. The Swiss Federal Office of Energy (SFOE), for instance, recommends a working temperature of 26 degrees Celsius for servers. As confirmed in several independent studies, this temperature increase has no impact on the lifetime of the equipment.
As a data center operator, we are obliged to maintain certain parameters in the cold aisle. We also have to ensure that older hardware can also be operated with no problems. That is why our values are based on the highest requirements.
Agreed values in the Green cold aisle:
18 to 26 °C temperature
40 to 60% relative humidity
Since the temperature in our cold aisles was 18 to 20 degrees Celsius, we gradually increased the temperature by 4 degrees Celsius within a one-year period. This resulted in a greater temperature difference between the uncooled room and the cold aisle, a thoroughly welcome result. When the cooling is optimal, the difference is 6 to 8 degrees Celsius – a value that we now achieve.
Step 2: Networking of air cooling units
The air cooling units offered us further optimization potential. In a 600 m2 data room, we place 14 cooling units. The room temperature is different everywhere in the room and the cooling requirement in the cold aisles also varies. This means the workload of the air cooling units is uneven. In the past, some air cooling units had a very heavy workload and consumed a disproportionate amount of energy.
It therefore made sense to network the air cooling units. Now each unit in the room also has a master function. The results are positive. Cooling can be better controlled at specific points, energy consumption has gone down, and the workload is better distributed.
Step 3: Improved customer installations
Our deployment team not only helps with setting up racks and cabling, its members are also trained to optimize the energy efficiency of customer installations. This cooperation pays off. It is easy to forget a rack panel when setting up hardware, or to place hardware in an unfavorable position. Both are common reasons for increased cooling requirements in the cold aisle.
After one year, we have a demonstrable energy savings of between 4.5 and 5 percent. Increasing the temperatures in the cold aisle, networking the air cooling units, and optimizing the cold aisle setup made this possible. But we don’t see our task as being complete. For data center operators, optimizing cooling is an ongoing task. Thanks to our experience and the improved cooling control, this task has become easier.