Google Lowers Data Center Energy Use through Machine Learning
Google recently released a white paper on how it is optimizing its data center operations through the use of machine learning. The practice is helping the company drive down its energy use.
According to a May 28 Google blog post, the machine-learning model is the brainchild of Jim Gao, an engineer on the company’s data center team. Although Google said its data centers use half the energy of typical data centers, Gao wanted to develop a model that would predict and improve data center performance.
The model works similarly to speech recognition software. It runs variables such as IT load and outside air temperature through a computer, and the computer analyzes the data to find patterns and “learn” from them. Google engineers then use this data to see how these variables interact with one another so they can develop new ways to make data centers operate more efficiently.
Google said the models are 99.6 percent accurate in predicting power usage effectiveness (PUE).
Image credit: A simplified version illustrating how the model works. Image via Google.
- Advanced Rooftop-Unit Control (ARC) Retrofits: Field Demonstrations Validate Energy Savings
- Operationalizing EHS Management: Bridge the Gap from Strategy to Execution
- 2015 Insider Knowledge
- There’s Money in the Trash
- Top 10 Steps for a Successful EMIS Project
- Financing Environmental Resiliency and a Low-Carbon Future with Green Bonds
- 2016 Environmental Leader Product & Project Awards
- Approaches to Managing EHS&S Data
- 2016 Energy and Sustainability Predictions Findings from Facilities Professionals
- Practical Guide to Transforming Energy Data into Better Buildings