Jérôme Totel, group strategy and innovation director at Paris-headquartered European data-centre provider Data4, talks of AI’s potential to boost sustainability in the data centre and how his company is harnessing these possibilities.
How important is sustainability in today’s data centre?
Unfortunately, I would say that climate change is really here, with this summer showing some alarming trends from which we cannot hide. This doesn’t only concern the data-centre sector, but every sector and all of us individually.
At Data4 itself, sustainability is one of our key pillars of growth, alongside quality and scalability. We focus on this during every phase of the data centre’s lifecycle, meaning the construction as well as the operational phase. If we forget this construction part, we lose many benefits because the majority of greenhouse gas emissions happen then. Once the data centre is built, its environmental DNA is decided.
We also offer an easy-to-use Green Dashboard calculator embedded into our customer portal to help customers understand all their key environmental KPIs, as this type of transparency is crucial across the industry.
What potential does AI have for aiding sustainability?
Artificial intelligence offers new ways to optimise the operations of our data centres – and we hope to use it to reduce our power-usage effectiveness, cutting our electricity consumption.
We’re already operating our data centres well, but they can have between 5,000 and 10,000 sensors inside. The human mind cannot process all that information in real time, particularly when you have 32 data centres in operation like Data4 across five countries.
The capabilities of AI, however, allow us to manage and study data in a very large quantity, meaning it can support human decisions in a way that optimises operations.
What stage has Data4 reached in deploying AI in data centres?
After previously working on AI as a proof-of-concept, we’ve been in a rollout stage over the past 18 months. We have more than 10 data centres up and running with AI, with plans to increase that to 15 by the end of this year and then continue rolling it out to all our data centres. In the early stage, the technology will make some recommendations for optimisation of cooling systems.
But much of the activity in our AI-installed data centres so far has been about collecting data and training the AI model on it, which can take between nine and 18 months.
How much human input will be retained in AI data centres?
They won’t be autonomous data centres. This is because the AI doesn’t know everything, especially when it comes to customer constraints, or having all the knowledge if there’s a big event like the Olympic Games – which are happening in Paris next year.
In such circumstances, the AI may, for example, make a system change that’s not a good change, causing an outage. We therefore see AI as providing significant added value to the human in helping them handle huge quantities of data and make decisions, but not as a replacement.
Are there any challenges for Data4 in scaling up AI as it adds more data centres?
That’s why we took some time to think about how to design the long-term architecture, to ensure we have a system that’s completely scalable and won’t hit bottlenecks. We knew we had many data centres in which to incorporate AI, with more to be added in the coming years. And it’s a long journey to roll out such projects, so we don’t want to be redoing it every three years.
The only potential issue I can see is that one day we might have too much data coming into our data lake, which comprises a centralised database for collecting information – in which case we may need to upgrade the available bandwidth or add another data lake.
How big a long-term contribution can AI make to data-centre sustainability?
It’s hard to say at the moment. Like the algorithms for these systems, we’re still in the learning phase and I think we’ll discover new use cases using AI that we haven’t thought of yet.
In addition, we’ll have new changes to implement to the system on an ongoing basis: it’s a continuous approach using these tools rather than thinking it’s all done once AI is up and running in our data centres. Ultimately, as I say, it’s important to keep implementing activities in every phase of the data-centre lifecycle to ensure sustainability in the long term.
