Covering Disruptive Technology Powering Business in The Digital Age

Written by: Bailey Martin, Tech Journalist, AOPG.

Last week, AOPG was invited to Schneider Electric’s Prefabricated Data Centre Module Factory in Barcelona, for a press event to touch on the world’s issues regarding the need for further development in data centres. I was curious to know how one of the globe’s most notoriously sustainable and environmentally responsible power supplier would present their ‘take’ on tackling the current issues at hand for data centres. Namely, the need for more sustainable and cost-effective computing power supply and cooling architecture to meet the energy consumption demands of  Large Language Models (LLMs) and generative Artificial Intelligence (AI).

The event featured a distinguished panel of experts, including Natalya Makarochkina, Senior VP of Secure Power Division & International Operations, Maurizio Frizziero, Director of Cooling Innovation and Strategy, and Shaheen Meeran, VP of Business Development and Secure Power. With such a lineup, my curiosity quickly turned into eagerness, and I was off to Barcelona!

In the photo: Maurizio Frizziero (first on the left), Natalya Makarochkina (in the centre), and Shaheen Meeran (second from the right).

Upon listening to and eventually conversing with the speakers at the event, I was pleasantly surprised by their ability to present and simplify the complex issues facing data centres. They clearly shared and demonstrated their solutions and approach to tackling these challenges.

Addressing Modern Data Centre Challenges

If I could sum up the key themes of the event into three words, they would be Efficiency, Sustainability, and Cost-Effectiveness. From this event’s takeaways, we can get a better look at how Schneider Electric’s prefab data centre modules, cooling distribution units, AI purpose-built servers and EcoStructure IT software all play massive roles in addressing and resolving the discussed issues.

The concept of a ‘data centre’ has evolved over the decades since its origins in mainframe computing rooms in the 1950s. Now, data centres have become much more crucial to modern digital life, providing the infrastructure to accommodate cloud computing, data processing, web hosting and more. At the beginning of the 21^st century, we have seen data centres begin to evolve further to overcome new hurdles.

Tackling the issue of AI workloads first, Schneider has partnered with NVIDIA and placed its focus on fitting their racks with AI purpose-built servers equipped with GPUs (Graphic Processing Units). As stated by Natalya Makarochkina “GPUs run parallel to AI.” These GPUs can manage AI ‘training’ times in development stages more efficiently by using algorithms to break down large learning tasks into smaller tasks that run simultaneously.

This is why GPUs are now being favoured in AI servers in place of CPUs. In Shaheen Meeran’s keynote presentation, she further explained that while these AI servers will solve the issues of ‘development’ (training) and ‘deployment’ (inference) of AI, this leads to the next issues to be solved – power consumption and cooling solutions.

Tailoring Cooling to Specific Needs

These AI servers are better equipped to handle the development and deployment of AI software. However, in doing so, the power utilisation is at a constant 100% (comparatively, traditional servers operate at 10-40% power utilisation), and energy consumption triples. With energy consumption up and rack space remaining the same, this brings in the second issue of a higher increase in temperature output thanks to GPUs having higher thermal densities. It gets real hot, real fast!

Maurizzo Frizziero, Director of Cooling Innovation and Strategy, was available to provide insight and solutions in this area. He made it abundantly clear that air cooling has its limitations. It has been known for a while now that air cooling is approaching its limit in capabilities, with the method’s maximum ability averaging under 20 kilowatts (kW) of power. With these AI servers and GPUs, we can expect the kW to be much higher than this.

Liquid cooling is the necessary step to be taken here. One of his key points was, “It’s not about a product but an architecture…,” emphasising the unique cooling requirements of each data centre or specially customised data centre modules for Schneider’s clients. Schneider Electric’s Coolant Distribution Unit is a further addition to their prefabrication factories, aimed at addressing these cooling issues. While the unit itself comes with a manual for assembly, proper implementation requires skilled distribution design, of which Schneider has no shortage.

The Advantages of Prefabricated Data Centres

Lastly, we get down to the Prefabricated Data Centre Modules that have been mentioned throughout this article. Schneider Electric’s data centre modules bring us back to those three key themes mentioned above: Efficiency, Sustainability, and Cost-Effectiveness.

Think about a typical modern data centre now… You build it on location, fill it with servers, and power them up. Once you’re up and running, clients and hyperscalers come in and rent out digital space there. With this approach, the product or unit comes first before the clients, meaning there is ‘white space’ with limited customisation for each client. With Schneider Electric’s Prefabricated Modular Data Centres from their factories, including the one I visited in Barcelona; these data centre modules can be deployed much quicker to each client as a ‘standard model’ or as ‘engineer to order’ (customised) exactly to their data and server requirements, allowing for upgrades or alterations as needed. Moreover, they are portable and can be reused in new locations, ensuring no motion, resource, or rack space is wasted. This approach cuts costs, saves time, and maintains sustainability.

At this point, Schneider Electric’s EcoStructure IT platform (DCIM 3.0 vision) should receive its own additional mention as another extension for client’s CIOs to monitor and significantly reduce energy costs as well as carbon footprint from energy consumption. This platform itself uses AI for analytics.

Alison Matte, Sustainability Lead of EcoStructure IT made the interesting point of the cycle of data centres accommodating the development of AI to in turn be optimised by AI themselves. With safeguards and regulations around ESG and sustainability only expanding from Western nations to Asia, it’s a no-brainer for data centres to use a competent DCIM such as this – it’s not only about saving costs now but also maintaining a clean and habitable planet.

Towards Sustainable Data Centres of Tomorrow

All in all, I found the Schneider Electrics press event to be most insightful. It was refreshing to see a well-diversified company clearly explain to the public its roadmaps for adapting to changes and solutions to challenges that come with them. The experts who gave their time to the press and presented these issues and solutions reflected the ethics and quality that Schneider Electric strives for. It just goes to show that Schneider Electric is one of the few companies that prioritise the talent they recruit just as much as the product and service they provide.

(0)(0)