Lancaster University hails potential of AI software to cut datacentre power consumption

Lancaster University is developing artificial intelligence (AI) software for servers that could potentially drive down the amount of energy consumed by datacentres. Researchers from the university’s data science institute claim to have developed a software system that can alter its own behaviour and performance according to the tasks the server is doing.

The system, dubbed REx, consists of a set of software components that can rapidly reassemble themselves into the most efficient form without the need for any human intervention. The technique is known as micro-variation. These software components, which can be memory caches and various search and sort algorithms, can then automatically assemble themselves depending on the type of workloads the server is being used to run.

“With the help of machine learning and a highly adaptive virtual material with which to write code, for the first time we have computer programmes that are able to gain a deep understanding of their behaviour and performance while they do their job. This has the potential to completely remove the need for humans to understand these complex systems, pushing that responsibility to machines instead.” said Barry Porter, a lecturer at Lancaster University’s School of Computing and Communications, in a YouTube video to discuss the work.

Porter said the university has pinpointed the datacentre as an ideal use case for the technology, adding that it could help operators run their facilities more efficiently.

“Datacentres are highly dynamic environments, which experience major fluctuations in workloads over the course of a day, as popular content changes. By having the software that powers datacentres built from this self-assembling software that learns its own behaviour as it’s running, we have the ability to deal with these dynamics much more efficiently and therefore save energy in the long term.” Porter.

REx could also help enterprises cut software maintenance costs because its self-governing capabilities mean fewer developers will be needed to manage systems built on it.

“Self-assembling software models could also have significant advantages by improving our ability to develop and maintain increasingly complex software systems for a wide range of domains, including operating systems and internet infrastructure,” the university said, in a statement.

The team at Lancaster have been supported in their efforts to develop REx by the Engineering and Physical Sciences Research Council (EPSRC) and a PhD scholarship in Brazil.

Details of their work in this area can be found in the REx: A development platform and online learning approach for runtime emergent software systems research paper.

Credits: Caroline Donnelly from ComputerWeekly.com (Source: here)