Abstract: Starting from the key parameters of data center power utilization efficiency (PUE), this paper analyzes all aspects of data center energy efficiency, including infrastructure server, network and reliability, and points out that these key parameters can be used for the overall comprehensive evaluation of data center, so that the design and evaluation of data center infrastructure has a quantifiable basis. Aimed at parameter optimization, this paper studies the supervised learning algorithm, unsupervised learning algorithm and reinforcement learning method, and points out that the most distinctive feature of machine learning is that there are many algorithms, different ideas and different development paths. Then, based on various learning algorithms, this paper analyzes and designs the data center parameters automatically from the aspects of operation and maintenance parameters optimization, PUE optimization and so on. Finally, the combination of machine learning and edge computing in the future is analyzed and introduced. It is believed that machine learning algorithm will have better use scenarios in the new field of data center, and can bring along greater benefits.

Key words: PUE, key parameter, machine learning, optimization