Royaume-Uni: Intelligence artificielle et refroidissement de centres de données (en anglais)

Data centre traffic is on the rise – projected to grow by 27% per year from now until 2020 – so many companies are looking at how they can use artificial intelligence (AI) to make their data centres more efficient. However, according to Dr Stu Redshaw, Chief Technology Officer at EkkoSense, many risk seeing AI as a magical solution. AI solutions cannot solve problems on their own without huge amounts of data and clever algorithms. A key part of ensuring that AI solutions are actually effective, therefore, is machine learning.In spite of recent developments, it seems that many UK data centres are performing poorly when it comes to cooling: according to a survey carried out by EkkoSense, 66% of installed cooling equipment does not deliver any active cooling benefits. Perhaps more worryingly, almost 8 out of 10 data centres do not comply with ASHRAE guidelines for data centre thermal testing.

According to ASHRAE, temperature data should be collected from at least one point every 3 m to 9 m of rack aisle in data centres. In reality, this is very difficult to achieve – individual racks would need to have their own dedicated thermal sensors if the data centre operators want to obey target limits. AI solutions will need to identify these performance exceptions and be able to manage them in order to improve data centre cooling. The continued, significant investment in cooling equipment will hopefully move UK data centres in the right direction – overall energy levels have already been reduced by 30%, but further progress would be possible with a truly AI-led approach.

For further information, please follow the link below.