Simon Brady, Product Manager, Liquid & High Density Cooling at Vertiv, outlines some of the challenges – and opportunities – of hybrid cooling.
Today, sustainable and resilient data centre design hinges on effective cooling, but thermal management is undergoing a period of change. Traditional heating, ventilation, and air conditioning (HVAC) systems, comprising air handlers, chillers, etc., are gradually maxing out on their ability to provide the cooling densities for next-generation compute. As data centres prepare to cope with the growth of artificial intelligence (AI), machine learning (ML), and high-performance computing (HPC), chip power densities are soaring – and so is the heat generated. These evolving technologies require new cooling methodologies for both optimal performance and minimised downtime.
Liquid cooling technologies will be essential alongside the current legacy thermal solutions that use air to dissipate heat. These traditional solutions simply do not have the thermal transfer capacity to cope with emerging chip thermal design power (TDP). Emerging technologies such as ‘direct to chip’ and immersion liquid cooling solve the increasing TDP issue by bringing a liquid cooling medium directly to the surface of the chips and servers.
These solutions are not as widely adopted as air cooling yet but are gaining traction, particularly with hyperscale operators. With nearly one in five data centres already leveraging liquid cooling and an additional 61% considering its adoption, it’s clear that this technology will become mainstream. It offers unparalleled efficiency in dissipating heat, particularly in high-density environments, where traditional air cooling methods fall short.
The benefits
Liquid cooling provides many benefits for data centres addressing rising heat levels and efficiency needs. From improved processor reliability and performance to space savings with higher rack densities.
It enhances thermal efficiency by directly cooling heat-generating components, providing optimal temperatures and reducing the risk of overheating. With higher heat transfer capacity, it supports denser workloads like HPC and AI applications. Additionally, it minimises energy consumption by efficiently removing heat, leading to cost savings and reduced environmental impact.
Liquid cooling is also operationally quieter because the fans in a liquid-cooled system tend to rotate more slowly and quietly than those in an air-cooling system. Space-saving and quieter operation make liquid cooling ideal for environments with limited space or noise restrictions. This can enhance the working environment for data centre personnel and mitigate noise pollution in surrounding areas. However, it’s important to consider that in a 100 kW rack, only 80% of that load will be liquid-cooled. The other 20 kW per rack still needs to be cooled with air.
Beyond immediate benefits, liquid cooling technology is a future-friendly data centre infrastructure, offering flexibility to adapt to evolving hardware requirements and thermal challenges.
The challenges
While the benefits of liquid cooling are compelling, its successful integration requires careful planning and coordination. IT, facility and power teams should work closely together because the IT stack, power usage and building chiller will all be impacted by AI deployment and the data centre infrastructure that enables it. A clear strategy and communication will enable a smooth process for infrastructure deployment and operational efficiency.
Checking that existing infrastructure can support liquid cooling systems, or identifying and addressing any necessary upgrades, is critical to the success of the deployment. Budget constraints could limit the scope of the project, requiring careful prioritisation and resource allocation to maximise the return on investment.
Sustainability goals are also playing an increasingly significant role in decision-making processes. Organisations must evaluate the environmental impact of cooling systems and look to align them with broader sustainability objectives. This may involve implementing energy-efficient technologies that use low global warming potential (GWP) refrigerants and/or adopting closed-loop water cooling systems to reduce the carbon footprint of data centre operations.
A hybrid approach
Although air cooling is commonplace in the industry and will still exist for years to come, a transition from 100% air cooling to hybrid liquid cooling will become the preferred and necessary solution if data centres are to cope with AI clusters effectively. This is due to traditional air-cooling systems being unable to effectively cool at higher TDP temperatures and densities.
The adoption of liquid cooling heralds a new era in data centre thermal management, marking a significant leap forward in addressing the escalating demands of modern IT infrastructure. However, a hybrid approach to cooling, which combines both air and liquid cooling technologies, is poised to be the way forward for most data centres in the future to effectively remove the heat from both the room and the racks. In most cases, ‘direct-to-chip’ liquid cooling technologies can work in conjunction with air cooling systems. In this way, the strengths of each cooling method can be leveraged to optimise thermal management in data centres, offering several key advantages.
Strategically employing liquid cooling to target specific hotspots and enhance temperature control optimises efficiency and reliability in cooling systems. By leveraging liquid cooling’s superior heat transfer capacity, the hybrid approach can support dense workloads without compromising performance whilst reducing overall energy consumption in comparison to conventional air-only cooling methods.
What’s more, its flexibility and scalability allow for tailored solutions that can adapt to evolving workload demands, facilitating incremental upgrades and expansions as necessary. The diversification of cooling methods in a hybrid approach enhances redundancy and reliability, diminishing the risk of system failures and enabling uninterrupted operation, even during maintenance or outages.
However, the journey towards successful deployment of liquid cooling necessitates meticulous planning, seamless collaboration across multidisciplinary teams, and unwavering commitment to implementing industry best practices. It is important to work with experienced providers that can provide valuable consultancy and customised designs.
