Powering the future: The role of cabling in data centre modularity
Industry Voices

Powering the future: The role of cabling in data centre modularity

AdobeStock_ AI Automation.png

Niklas Lindqvist, Nordic General Manager at Onnec believes data centre operators on the race to deliver capacity

Microsoft plans to more than double its existing data centre capacity this year, while AWS invested $650 million in a data centre next to a 2.5 gigawatt nuclear power station to support its AI goals.

As all operators battle for a share of AI's projected $15.7 trillion boost to the global economy, it’s critical they also think about how to adapt to this burgeoning demand. As operators invest in expanding data centre capacity, they must follow a holistic approach to design. Static sites that cannot change or adapt are no longer sufficient; instead, operators must embrace flexibility and a forward-thinking approach. This means designing data centres with the ability to interchange hardware and infrastructure from the outset. A new type of modularity is needed. But – to enable it – cabling is key.

The Capacity Challenge

Predictions suggest that operators will soon need to install infrastructure capable of accommodating 50-100kw per rack to meet AI demands – a significant increase from the 8-10kw per rack in 2020. However, operators face three key challenges: escalating demand, construction slowdowns, and hardware longevity.

To meet AI and other workload needs, flexible data centres are essential to enable rapid hardware refreshes – particularly with demand for GPU-powered racks. Here, re-evaluating data centre design for adaptability offers benefits beyond scalability, such as aligning with sustainability goals, amid increasing pressure to improve ESG metrics.

During the design phase – where the focus is mainly on power, cooling, and what goes into the building – cabling issues tend to slip through the cracks. Not enough thought goes into potential future problems. This is why it's important to think ahead about future demand in order to be ready for changes if customers leave and the site needs to adapt to shifts in demand.

To proactively tackle issues hindering flexibility—such as bad cabling, subpar installation, and resource shortages—operators must adopt a fresh approach to modularity.

Cabling and ‘New' Modularity

Traditionally, modularity was about constructing data centres using prefabricated containers that include everything from power to cooling. But this approach – intended to streamline construction – often fell short in helping operators scale effectively. As a result, major hyperscalers like Google and Microsoft shifted away from using containers to expand data centre capacity.

Today, the concept of modularity must evolve beyond "Lego brick" containers. Instead, operators should embrace a new kind of modularity, designing data centres with built-in flexibility to accommodate growing workloads by making plug-and-play hardware a reality. Failure to do so will put new or existing data centres at increased risk of falling into disarray and requiring early refreshes.

While the containerised concept modularity may not always be enough to meet the demands and workloads facing operators, its underlying principle remains crucial for managing the challenges of more frequent hardware refreshes. It’s about building modular IT infrastructure where interchangeability and connectivity are key.

Cabling will become a critical foundation that facilitates flexibility in modular data centres. Bad cabling can cause poor connectivity, data corruption, and troubleshooting issues – all leading to downtime. Replacing it is costly compared to best-practice installation. What’s more, bad cabling could potentially shorten a data centres’ lifespan from 10–15 years to just 3. For modularity to become a reality, cabling must take a more pivotal role in the design phase.

Cabling Best Practice Steps

As AI workloads ramp up, operators should consider how modular data centres can help them scale to meet demand. Success hinges on adopting a holistic approach to planning and adhering to cabling best practices, including:

1. Shifting the mindset to modular: Operators should view sites as integrated engineering projects rather than conventional commercial developments. Identifying the services the site will provide ensures optimised designs that can evolve with changing workloads and development of cabling infrastructure that supports interchangeability.

2. Being more strategic about cabling: Benchmark and minimise the number of high-quality cabling suppliers chosen for consistent, standardised architecture. Carefully limiting suppliers facilitates support, leverages economies of scale, and ensures tight alignment for efficient inventory management.

3. Building a skills pipeline: Foster a diverse skillset pipeline, including technical, installation, and maintenance competencies. Develop local talent to enhance data centre sustainability and agility for efficient project delivery. Anticipate future needs through AI, analytics, and predictive skills, while collaborating with educational institutions for workforce development.

4. Developing a blueprint for cabling strategy: Create a detailed cabling project plan prior to installation. This must factor in the data centre’s unique requirements, cable lengths, cable trays, preparation time, and delivery schedules to ensure precise infrastructure design without delays.

5. Creating a meticulous installation guide: Create a clear installation guide detailing cable types, connections, compatibility, measurements, and patch cabling. This guide will be a working plan for modular data centres, facilitating seamless hardware interchange.

Unlocking the power of cabling in data centres

Cabling is crucial for modern modularity. To keep pace with AI workloads, data centres must adopt a modern approach to building modular IT infrastructure – emphasising interchangeability and connectivity. Following cabling best practices and taking a holistic approach to data centre construction and operation enables operators to seamlessly upgrade hardware, deliver AI-compute, and accommodate future workloads. This approach allows for offering a wider range of services to diverse customers, scaling, and adapting to new demands efficiently – avoiding costly refits and ensuring uptime to meet customer SLAs.

Gift this article