KAWASAKI, Japan — Fujitsu Laboratories Ltd. has announced the development of high-accuracy predictive technology for such metrics as temperature and humidity to enable energy-saving operation of air conditioning equipment in data centers.
Air conditioning in data centers can account for 30 percent to 50 percent of total electricity use, said Fujitsu Laboratories. To flexibly respond to the dynamic status changes that are unique to data centers, such as moving IT equipment in and out and changing rack arrangements, Fujitsu has developed a technology that sequentially builds a model that predicts air conditioning effects from collected data, enabling reductions in air conditioner energy use.
Existing data center air conditioning equipment operates on the basis of information from a variety of installed sensors so as to maintain the target temperature and humidity even when the operating rate of the IT equipment it contains, such as servers, increases. When sensor data exceeds a set value, the system carries out rapid cooling for safety reasons, which causes air conditioning to operate excessively.
In order to use model-based control in data centers, Fujitsu Laboratories applied just-in-time (JIT) modeling, which can build a model in response to a situation. With existing JIT modeling, useful information is selected from sensor data, including temperature and humidity, through mathematical processing, and this data is used to create a predictive model and calculate predicted values. With the sensor data alone, however, this method does not select enough useful information to immediately reflect the status of dynamic changes in the data center, making highly accurate predictions difficult.
The new technology now creates a database that incorporates the status of air conditioning equipment, including the utilization rate of machines and fan speed, and has a requirement to not only select useful information against prediction targets, but also to automatically select at least one variable from the air conditioning equipment status database. By creating a model using the selected variables, this technology succeeds in improving predictive accuracy. Fujitsu Laboratories has thus created a high-accuracy predictive technology that can respond to the frequent changes in the state of a data center by building sequential models.
With regard to this newly developed technology, Fujitsu Laboratories conducted a simulation using actual 100-rack scale data, and confirmed it was possible to predict the temperature of servers’ air supply with high accuracy, within an average margin of +/- 0.17°C and maximum +/- 2.1°C, even in a frequently changing environment. These results were applied to simulated conditions for a 1,000-rack data center with servers that would consume about 70 million kWh/year, and air conditioning that would use 22 million kWh/year, and calculated that, with the application of this technology, such a data center could expect to save 4.5 million kWh in electricity consumption, a reduction of about 20 percent.
Fujitsu Laboratories will be conducting field trials of this technology throughout fiscal 2016 at data centers operated by Fujitsu Limited, after which it is scheduled to be put into actual operation in fiscal 2017, and incorporated into Fujitsu’s operations management infrastructure software.
For more information, visit www.fujitsu.com/jp/group/labs/en/.
Publication date: 7/14/2016