AI may help create more sustainable data centres – but work still to do
Enterprise data centre provider Aegis Data argues in its latest note that utilising artificial intelligence (AI) could be key in winning the battle for sustainable data centres.
“There’s no escaping the reality that as more connected devices and technology trends sweep the market, more demands will be placed on the data centre to provide the high-powered servers and cooling systems required,” said Greg McCulloch, CEO of Aegis. “But in the pursuit of guaranteeing performance, it is having an accumulative effect on the global share of data centre emissions… it’s not an understatement that the industry needs to take immediate action and AI may just be the solution in order to help achieve more sustainable results.”
For data centre providers, PUE (power usage effectiveness) figures have traditionally been the name of the game for sustainability, dividing energy for the whole facility by the energy for the facility’s IT equipment and aiming for an ideal score of 1.0.
Back in April 2015 this reporter attended the opening of a new Rackspace data centre (below) in West Crawley, which had an impressive 1.15 PUE rating; the average was nearer 1.7 with Rackspace admitting themselves their target was 1.25. The data centre utilised other natural benefits, including cooling using natural air – a feature that was tailor made for the UK – while the design was donated to the Open Compute Project.
Picture credit: Rackspace
It was pretty impressive stuff all round. Yet not everyone agrees that PUE is the way forward. Professor Ian Bitterlin, chair of the British Computer Society’s data centre group on server lifecycle, told E&T last year that “improving server effectiveness is the only way to improve data centre effectiveness”, and that people are erroneously putting PUE as a data centre ‘goodness’ metric.
As far as Aegis is concerned, the work Google is undertaking by putting its DeepMind technology in its data centres – and experiencing a 15% energy efficiency improvement in the process – is on the right lines, but not without its flaws. “[AI is] a technology that’s very much in its infancy, and if it is to overtake human interaction in the data centre, then it must be rigorously researched and tested to guarantee performance,” said McCulloch. “But once this hurdle is overcome, AI has the ability to provide a comprehensive visualisation, automation and monitoring process that can envision the necessary power, cooling and energy requirements needed.”
Google is not the only company to be experimenting with cognitive intelligence in its data centre facilities; earlier this week IBM announced the launch of four new centres in the UK, and revealed how one customer, travel operator Thomson, was trialling new search functionality based on supercomputer Watson.
“With the wide range of operations that occur in a data centre, the ability to eliminate human error and have an ‘always on’ approach will go a long way in helping reduce energy consumption,” added McCulloch.