Artificial Intelligence Used to Improve Google’s Data Center Cooling Efficiency

artificial-intelligence-energy-efficient-semiconductor
Valerie Eacret for Zondits, July 9, 2014

Reducing the amount of energy used for cooling Google’s data centers is now being achieved by artificial intelligence with the help of one naturally intelligent data center engineer.

Google has been collecting power usage efficiency (PUE) information, such as such as cooling tower speed, processing water temperature, pump speed, OAT, and humidity at 5-minute intervals for the past 5 years. As you may already know, Google allows employees to use a portion of their time at their own discretion. After taking an online class on machine learning, one of Google’s engineers started using this time to create a computer model that analyzes all of these data points and can learn to improve the energy efficiency of the data center’s operations as more data becomes available over time.

With the 99.6% accuracy of the predictions that the model produces, Google is able to set up alerts if the data center isn’t performing as expected, allowing them to nip problems in the bud. This model is also useful for improving the overall efficiency of the data center by determining the optimal time and frequency for maintenance, such as cleaning the data center’s heat exchangers. Google is also able to play with the PUE using simulated tests, which are risky enough that they would not be conducted in the actual physical environment.

While there aren’t any plans at this point to market this model, a white paper authored by the engineer encourages others to create their own models.

______________________________________________________________________________________________________________________________________________

Google is harnessing machine learning to cut data center energy

Gigaom, May 26, 2014

Leave it to Google to have an engineer so brainy he hacks out machine learning models in his 20 percent time. Google says that recently it’s been using machine learning — developed by data center engineer Jim Gao (his Googler nickname is “Boy Genius”) — to predict the energy efficiency of their global data centers down to 99.6 percent accuracy, and then to optimize the data centers in incremental ways if they become less efficient for whatever reason.

Part of Gao’s day-to-day job at Google is to track its data centers’ power usage efficiency, or PUE, which demonstrates how efficiently data center computing equipment is using energy. Traditionally many data center operators were seeing about half of their energy consumed by cooling equipment, but in recent years data center leaders like Google, Facebook and others have focused on tools like using the outside air for cooling, or running the server rooms at warmer temperatures, to dramatically cut energy use.

Read More