Google aims for an autonomous data centre management system - FactsnFigs - Submit Your Guest Post, Articles for more Exposure, Credibility and Traffic.

AMAZON


Google aims for an autonomous data centre management system



Using Artificial Intelligence by Google as a means to optimise data centre efficiency has moved to a new level. Machine learning algorithms are autonomously tuning cooling plant settings in real-time tirelessly. The development of the system is based on the platform unveiled by Google previously; an AI-powered recommendation engine.

Most of the data centre personnel are unlikely to consider a typical hurricane or tornado season as the best time to deploy cooling-system settings for energy savings. Instead, it’s time when humans rely on mother nature and hope the power flow remains intact; well, humans do have their priorities. But then, there’s AI-powered algorithm capable to seek and shave-off every kilowatt hour irrespective of the weather anomalies.

The actual occurrence

During a tornado watch, one of Google’s data centres in the Midwest of which the cooling plant was managed by AI system changed the plant’s settings that were considered a counterintuitive by human facilitators. But after a detailed evaluation, it did what had to be done for energy saving under the given circumstances.

Weather conditions that complementto a severe thunderstorm system results in a significant atmospheric pressure drop and shift in temperature as well as humidity level! That said, weather plays a crucial role in the way a typical tier 3 data centre tune its cooling system whereas the Google operated software recalibrate the existing setting and manage changes accordingly, even if the benefits are few.

Back in 2014, it was revealed that Google is to deploy AI for improvingits data centre energy efficiency. That particular system developed by Jim Gao; data centre engineer at Google back then and was deployed as a recommendations engine.

With ever advancing AI as a means to achieve energy efficiency, Google data centres just moved up to a new level. Rather than just making recommendations, the system’s capable to perform all the tasks associated to plant cooling management independently.

The initial/first version of the system was developed by Gao titled as “20-Percent Project” which was later incorporated with the Google’s DeepMind AI algorithm which allows it to move up by 40% in saving energy whilst the facility’s temperature is regulated by the cooling system. Further upgrades add another 15% thus totalling it to 55% when concerned with energy savings.

The basics of the system were developed on the authentic work collaboration of Gao and Google’s DeepMind. The input variables are same that totals to approximately 21 including external air temperature, barometric pressure, wet and dry-bulb temperature, data centre’s power load, dew point, pressure of air generated by servers from the rear end during active process and more.

Various tweaks & modifications

The example of tornado-watch is perhaps the best illustration of how a Google-empowered machine algorithm can save energy in a data centre which is beyond the human capability. The combined benefits are that of the significant savings even from minor tweaks and modifications being performed continuously.

In case the outside temperature rises from 72 to 76 degree Fahrenheit while the wet-bulb temperature staying same, it’s highly unlikely for human operators to execute any change to the cooling plant setting to cope with that minor temperature change even if it can reduce energy consumption. But then there’s the Google algorithm which will make sure all changes are performed accordingly.

The system can prove exceptional when someone plans to initiate a new data centre or perhaps augments it to a tier 3 data centre facility. That said, a newly inaugurated facility operates at its least efficiency since it doesn’t utilise the true potential of the infrastructure.

It’s possible Google to deploy cluster of servers in the new facility from day one onwards. Irrespective of the number of rows, the network fabric is spread across the entire data centre to meet all the electricity requisites in a perfect way. We as humansneed to give machine-learning algorithm some control of the mission-critical infrastructure and it can improve the performance in an exceptional way.
Theme images by linearcurves. Powered by Blogger.