Data centers account for more than 2% of total electricity consumption in China. The average power usage effectiveness (PUE) rating of global large data centers is 1.67, showing that there is plenty of scope for improvement.

Further, traditional data center construction generates a large amount of dust and waste and consumes large amounts of water.

As governments step up action on the climate crisis and cloud computing needs increase, making data centers more efficient is a priority.

One challenge is that data centers are typically designed and built using a sub-contractor model, meaning no one takes overall responsibility for issues such as PUE.

Huawei developed an end-to-end integrated data center management service solution that improves energy usage efficiency based on PUE simulation and intelligent energy-saving optimization. The BestDC platform brings all the stakeholders together by linking information, from the design and construction phase through to ongoing optimization.

During the design process, the BestDC performs simulation using 3D models of the physical, control, power distribution, and cooling system of the data center. These simulations are based on experience from more than 1,000 projects as well as the integration of over 600 standard equipment models, 300 equipment working condition curves, 25 system architecture models, and 80 national climate databases.

The core algorithms are used to simulate system running modes under different climates and load rates. The PUE simulation model has been shown to be 97% accurate.

Smart design means data centers can be built in a prefabricated, modular fashion, reducing cost and waste.

During the construction phase, 3D modeling integrates information on the project delivery plan, tasks, and quality requirements in real-time, enabling global visibility of the project construction status.

Once the data center is operational, the BestDC platform provides core capabilities to reduce energy usage. The system collects and analyzes the running data of key devices in the data center and completes intelligent model training on the cloud. Based on over ten core algorithms, such as deep neural networks and random forests, an overall energy efficiency model is built for the data center and delivered to BestDC’s local energy-saving platform.

Automatic optimization provides inference results and checks against service level agreements to drive continuous improvements.

Finally, heating simulation ensures that equipment room temperature remains within limits.

One data center in southeast of China has reduced its PUE from 1.6 to 1.31, which saves 38 million KWh/year per 3,000 racks. It also reduces carbon emissions by 28,400 tons per year, which is calculated to be equivalent to planting almost 1.5 million trees.