Meet FlashText – the Python library alternative to Regular Expression25 January 2018
What types of conversions should you optimize first?1 February 2018
Google has achieved success in the machine learning process and claims that last year they managed to increase efficiency in all their data centres and reduce energy consumption by 15%.
The amount of energy used by large data centres has always been a problem for technology companies. Maintaining server temperatures is such a challenge that Facebook even built one of the devices on the edge of the Arctic Circle. Google is trying a different solution to this problem. It uses the DeepMind artificial intelligence unit and utilizes AI to manage energy consumption in its data centres parts.
Google wonders about the data centre performance as long as we know about them. Such centres require considerable energy for cooling as well as constant adjustments of temperature, pressure and air humidity to operate as efficiently as possible. In the beginning, they decided to design and build their own devices from the scratch to allow continuous piloting of new cooling technologies and operational strategies. Their data centres use advanced cooling techniques. They have reduced energy consumption by installing intelligent temperature and lighting controls and redesigning energy distribution to minimize energy losses. High-performance servers are specifically designed to consume as little power as possible and they are free of unnecessary components such as graphics cards.
Last year, Google looked for further ways to reduce used energy and introduced artificial intelligence to this process. They claim that they can reduce the energy consumption in their data centres by 15% by implementing machine learning from DeepMind, a British AI company that they bought in 2014 for around £ 400 million. The reduction in energy consumption was achieved by combining DeepMind with a more accurate prediction of upcoming computational load – i.e. when people most often demanded energy-consuming YouTube videos – and very quickly matched this forecast to the required cooling load. In a dynamic environment, such as a data centre, it can be difficult for people to see how all variables – IT load, outside air temperature, etc. – interact with each other. Computers are irreplaceable when it comes to historical data analysis.
Jim Bao, the chief data centre engineer at Berkeley, took the information they collect during daily operations and carried it through the model to help understand the complex interactions his team struggled with.
Google does not disclose exactly how much energy is used by data centres but says that as a company it is responsible for 0.01% of global electricity consumption, and most of them are data centres. DeepMind has reduced Google’s cooling energy consumption by 40% and total energy consumption by 15%. “The implications are significant for Google’s data centres, given their potential for significant improvements in energy efficiency and overall emission reduction,” wrote Evans and Gao. “It will also help other companies that use Google’s cloud to improve their own energy efficiency.”
Google also declared that it plans to use DeepMind artificial intelligence for other parts of the data centre infrastructure, not just for cooling.