Preparing data centres for a much cooler, more fluid future as AI advances

By Ozgur Duzgunoglu, Design and Engineering Director, Telehouse Europe.

Data centre operators know they must constantly adapt to the changing needs of their clients. With the relentless advance of AI applications, the demand for high-capacity computing is set to grow significantly. The high processing speeds and energy demands of this technology present increased pressure on cooling systems, necessitating new, more robust approaches.

 

Most senior people within the IT world, regardless of industry, know compute requirements are set to increase substantially as AI and edge computing gain greater adoption. Recent Telehouse research revealed that 89% of IT decision-makers anticipate they will need high-performance, high-density computing systems by 2030. AI workloads are not currently large but could expand quickly, requiring low latency, high-bandwidth connectivity from data centres.

 

From the outset, data centres have always incorporated cooling mechanisms. Traditionally, the favoured technique has been air cooling, a method that relies on the simple physics of circulating cold air around operational hardware to mitigate the heat produced. But the modern workloads we witness, particularly those spearheaded by AI, are stretching the capacities of air cooling to their very limits.

 

With natural boundaries to heat transfer capabilities, air cooling technology could hinder the rollout of newer, more energy-intensive, heat-generating services to meet the new demand. This could cap the total volume of AI workloads a data centre can handle unless there is a change to a more efficient cooling technology. This is where liquid cooling technology comes into play.

 

The inherent limitations of air cooling systems in terms of heat transfer efficiency pose significant challenges. As we look to the near-future, ever-increasing computing power is likely to side-line air cooling. Operators will need to think very seriously about liquid cooling technology.

 

Although the current demand for liquid cooling remains low & steadyit is poised to assume a central role in the future of digital infrastructure. With intensifying pressure for cooling systems to be sustainable and eco-friendly, liquid cooling is on the brink of becoming the industry standard. Data centre providers need to ready themselves now for what is an impending shift.

 

Understanding liquid cooling

 

Two primary forms of liquid cooling are set to dominate the data centre sector. First, conductive liquid cooling harnesses the potential of liquids to directly extract heat from processor components. This system uses heat sinks attached straight to heat-generating units like central processors. These are then connected via tubes, facilitating fluid circulation and ensuring efficient heat removal.

 

Conversely, immersive liquid cooling necessitates the complete submersion of servers in a purpose-designed, non-conductive fluid. This enables the heat to be effectively dissipated into the liquid medium. However, this approach requires specific modifications to the servers to guarantee safe immersion.

 

Adopting liquid cooling brings a plethora of benefits to the table. It allows for the augmentation of rack densities, with some reaching an impressive 100KW per rack. This capability enables innovation-driven clients to implement power-hungry workloads vital for their growth. Furthermore, these cooling methods typically have a lower energy consumption, mitigating operators' worries about rising energy costs.

 

The reduced energy consumption translates into reduction in  an improved PUE (power usage effectiveness) rating and reduction in the overall carbon footprint. Additional advantages include freeing up data centre space due to the removal of CRAH/CRAC units and a reduction in noise levels with the elimination of fans.

 

Yet, the transition to liquid cooling is not without hurdles. The introduction brings its own set of complexities, especially during the design and installation phases. Potential leaks pose significant threats, leading to catastrophic hardware destruction or data loss. The quality of water used in the building's cooling system requires rigorous monitoring, and the financial implications of potential damage and maintenance should not be underestimated. A comprehensive approach, involving specialised equipment, must be considered during the planning phase by the design team.

 

Planning for implementation

 

Given this level of investment, how should operators position themselves in anticipation of the rising demand for liquid cooling? A robust and diverse supply chain is crucial. Diversifying suppliers can act as a safeguard against potential component shortages. Additionally, fostering close ties with customers is essential. Through transparent dialogue, operators can gather insights about expected workload trajectories, ensuring all parties are on the same wavelength.

 

AI is also set to become one of the cutting-edge technologies within this process. Operators will lean heavily on advanced systems to manage data centre functions and power consumption. AI will monitor building temperatures and recommend optimisation tactics. As the demand curve rises, the layouts of data centres will undergo a series of transformations, seamlessly incorporating liquid cooling solutions.

 

Anticipating the liquid future

 

While some operators may currently be satisfied with their well-established set-ups, especially if their customers are not crying out for demand-intensive workloads, such equilibrium is likely to be fleeting. An increasing number of organisations of all types will soon rely on high-density computing services, aiming to offer more advanced and competitive solutions to their customers.

 

Given the surging demand generated by AI applications for significant computational power, operators need to be one step ahead. Liquid cooling stands as the most effective means to meet these heavy heat-reduction demands. Companies need to start open dialogues with digital infrastructure providers about their anticipated needs so they ensure that the supply is in sync with demand. Everyone should be looking ahead to the next ten years.

 

Liquid cooling's appeal is multifaceted, not restricted to its efficiency alone. By closely collaborating with suppliers and establishing a harmonious relationship with the broader supply chain, operators can guarantee the on-time acquisition of essential components. Continued open communication with customers will be utterly essential, so operators have insight into their changing objectives and are able to plan with confidence, meeting sustainability goals in the process. Looking ahead, tools powered by AI will be invaluable, enabling operators to consistently monitor ambient temperatures and proactively implement liquid cooling solutions without compromising on emissions and energy efficiency targets. The clarion call is clear – the time to prepare for liquid cooling has arrived.

By Daniel Beers, Senior Vice President, Global Data Center Operations of Ardent Data Centers, a...
By Mark Nutt Senior Vice President, International Sales at Veritas Technologies.
By Fabrizio Garrone, Enterprise Solutions Director at Aruba S.p.A.
with Sasan Moaveni, Global Business Lead – AI & High Performance Data Platforms, Hitachi Vantara.
By Richard Eglon. CMO Nebula Global Services and Joanne Ballard, MD Mundus Consulting.
By Francesca Colenso, Director of Azure Business Group at Microsoft UK.