Artificial Intelligence is increasing computational demands, and the capabilities of data centres are being reframed. Ozgur Duzgunoglu, Design and Engineering Director, Telehouse Europe, urges operators to consider the impact of this transformation and how liquid cooling can be an indispensable tool to prepare data centres for the future.
Data centre operators are in a continual race to meet the changing demands of clients. As AI evolves at a staggering pace, many infrastructures now necessitate high-density compute capabilities for AI-centric applications. New Telehouse research has also found that the majority (89%) of IT decision-makers will demand high-density, high-performance computer systems by 2030. Yet, with this surge in power comes another challenge: the indispensable need for more robust cooling.
From their very inception, data centres have had integrated cooling mechanisms. Historically, air cooling has been the preferred method to maintain optimal temperatures. Using the simple principle of circulating cold air around active hardware, it dissipates the heat generated. However, the contemporary workloads we see today, especially those governed by AI, challenge the limits of air cooling.
Air cooling systems are fundamentally limited in their ability to be heat transfer solutions. This places strain on data centre infrastructure and also creates frustration among customers who are unable to deploy new resource-heavy services. While this is less of a widespread concern now, denser workloads and computing power requirements will grow over the next few years, and air cooling will need to make way for new solutions in these instances. Enter: liquid cooling.
While the appetite for liquid cooling currently remains stable, primarily due to supply chain disruptions, it’s gearing up to play a pivotal role in the future of digital infrastructure. As the urgency for enhanced power and cooling escalates, the focus on eco-friendly solutions will also intensify. The inflection point for liquid cooling is near and providers must prepare now ahead of this trend.
The rise of liquid cooling
The future landscape of data centre cooling is likely to be dominated by two distinct types of liquid cooling. Conductive liquid cooling taps into the principle of using liquid to directly siphon off heat from processor components. It employs heat sinks affixed directly to heat-generating elements like the central processor. These are then linked to tubes that facilitate liquid movement, ensuring heat is efficiently transferred.
Immersive liquid cooling, necessitates the submersion of servers in a specially engineered, non-conductive fluid, allowing heat to disperse into the liquid medium. This method, while efficient, requires specific alterations to servers to ensure their safe immersion.
The adoption of liquid cooling offers a myriad of advantages. Notably, it allows data centre operators to increase rack densities, sometimes reaching up to 100kW per rack. This, in turn, empowers clients focused on innovation to roll out power-intensive workloads essential for their expansion. Moreover, these cooling systems typically consume less energy, alleviating concerns for operators about escalating energy costs.
The lower energy needed also reduces the operator’s carbon footprint and leads to a marked improvement in its PUE number. As a bonus, the absence of CRAH/CRAC units liberates space in data halls, and the omission of fans results in quieter operations.
However, the transition isn’t devoid of challenges. Implementing liquid cooling introduces complexities, particularly in the initial design and setup phases. These complexities might ripple through to maintenance and troubleshooting stages. Operators must be vigilant against potential leaks or spills, which can spell disaster, leading to hardware damage or data loss. The cooling and heat rejection system’s water quality must be well-managed by operators. Additionally, the financial implications of repairs, replacements and ensuring optimal water quality cannot be overlooked. Specialty lifting and cleaning equipment should be considered by the building design team.
Planning for deployment
So, how should operators navigate the impending rise in liquid cooling demands? Establishing a resilient supply chain, achieved through diversifying suppliers, emerges as a foundational step. This diversification can act as a bulwark against potential component shortages. Furthermore, forging collaborative relationships with customers is paramount. Through mutual communication, operators can glean insights into projected workload growth, while customers can share their technology blueprints, ensuring both parties remain aligned.
In the foreseeable future, operators will increasingly rely on cutting-edge technologies to oversee data centre operations and energy consumption. AI will play a pivotal role, analysing building temperatures and highlighting optimisation avenues, and as demand swells, data centre layouts will evolve, seamlessly integrating liquid cooling without disrupting core operations.
Anticipating future needs
Currently, operators might be content with their existing infrastructure, especially if their customers do not demand heightened workloads. However, this status quo is going to change. A growing number of organisations will soon lean on high-density services, aiming to roll out new solutions to their users. With AI applications relentlessly forging ahead, and given the hefty computational power they demand, operators must respond proactively. Liquid cooling stands poised as the solution. Companies must also ensure they have open and honest conversations with their digital infrastructure providers on their future demands and requirements, as this will ensure supply meets demand in this critical coming decade.
Liquid cooling’s allure isn’t just limited to its efficiency. Its reduced energy needs ensure that operators don’t compromise their sustainability goals. By working in tandem with suppliers and fostering a cohesive relationship with the broader supply chain, operators can ensure the timely delivery of vital components. Maintaining open channels with clients, understanding their future needs and aligning strategies will be pivotal. As we look ahead, AI-powered tools will be indispensable, enabling operators to closely monitor in-house temperatures and pre-emptively deploy liquid cooling where needed. The time to prepare is now.
