Enabling Energy Efficient Data Center in Smart Power Distribution Systems

The goal of this project was to improve energy efficiency and enable demand response for data centers in smart power distribution systems. This project achieved two objectives: 1) it developed pre-commercial server, data center, and data center cluster energy efficiency technologies and strategies, and 2) it provided easily accessible software solutions to facilitate the adoption of energy efficient data center technologies. 

Technological and scientific advancements were achieved at three levels. At the server level, an innovative low power management system was developed that coordinates deep sleep states and dynamic voltage-frequency scaling and selects the optimal power state configuration for a given workload and traffic pattern. At the rack/data center level, a new workload scheduling algorithm was developed to improve the data center level energy efficiency. This new algorithm collects system statistics of worker servers to predict power levels and trigger load migration to require all servers to run at peak energy efficiency. At the data center level, the project team developed a solution to enable data centers to provide ancillary services to the electricity market by adjusting their energy consumption. 

If all data centers in California adopt the three technologies developed in this project, it could result in estimated annual electricity savings of 1,342 gigawatt hours, a corresponding cost reduction of $163 million, and a greenhouse gas emission reduction of 596,114 metric tons. The lessons learned from this project are being incorporated into IEEE Standards P1924.1.