Google Using Machine Learning To Optimize DCs


Google is using machine learning and neural networks to optimize its data centers.  The company has developed and tested a neural network framework that learns from actual operations data, to model plant performance and predict power usage effectiveness (PUE), a measure of how efficiently a data center uses energy,  within a range of 0.004+/-0.005 (mean absolute error +/- 1 standard deviation), or 0.4% error for a PUE of 1.1. The model has been extensively tested and validated at Google data centers, says the Mountain View giant.

The algorithms were conceptualized by Jim Gao, a Google data center team engineer.  The work started out as a 20% project; Google customarily lets employees work on independent projects in 20% of their time.  What Mr. Gao designed works a lot like other examples of machine learning, like speech recognition: a computer analyzes large amounts of data to recognize patterns and “learn” from them, according to Google.

In a dynamic environment like a data center, it can be difficult for humans to see how all of the variables—IT load, outside air temperature, etc.—interact with each other. Mr. Gao collected the information Google gathers in the course of its daily data center operations, and ran it through a model to help make sense of complex interactions that his team may not otherwise have noticed. After some trial and error, his models are now 99.6 percent accurate in predicting PUE, says Google.

A typical large scale data center generates millions of data points across thousands of sensors every day. Advances in processing power and monitoring capabilities can create a large opportunity for machine learning to increase efficiencies.

According to  industry estimates, data centers comprised 1.3% of the global energy usage in 2010.

[Image courtesy: Google]


Also see: