Liquid Assets: The significance of Liquid Cooling for Data Centre Sustainability

Amongst the escalating demand for more energy from data centre (DC) customers, maintaining a low Power Usage Efficiency (PUE) rating by DC providers is paramount. With low PUE comes the ability to identify cost saving measures and recognise areas for improvement, whilst also reducing carbon footprints and impacts on the environment. 

Yet, as more organisations deploy high performance computing (HPC) and greater IT workloads to support their business requirements, the by-product of heat from power usage is also on the rise. In order for DC providers to ensure that their equipment and systems remain in top condition to serve demand, suitable cooling and heat management is a must. 

One effective strategy is building DCs in cold climates and leverage free cooling methods. When the external temperature of a DC is considerably lower than the internal temperature of a facility (as little at 2°C), natural air can be used to directly cool the site and therefore, reduce – if not completely eliminate – the need of mechanical cooling processes. As a result, energy consumption is decreased significantly, and equipment life can be extended reducing embedded carbon. However, geographical constraints such as proximity to key markets and infrastructure mean that not every organisation can benefit from this advantage. For those unable to build in cooler regions, advanced cooling techniques offer a promising alternative. 

Among these, liquid cooling has emerged as a particularly effective solution, offering significant advantages over traditional air-based methods. Let’s delve into the options and benefits of liquid cooling in modern DCs. 

What is liquid cooling?

Rather than utilising traditional air-cooling technology, liquid cooling is a method that uses liquids such as water to conduct heat away from IT equipment. 

Compared to air, liquid has a much higher heat capacity (1.004 vs 4.18). Therefore, by using liquid cooling methods, the absorption and transfer rate of heat released from IT equipment will be much higher and more effective when dealing with increasing levels of power. 

The different types of liquid cooling 

Several liquid cooling methods are available for DC providers. Rack-based liquid cooling circulates coolant through racks to absorb and remove heat directly from servers. Direct-to-chip cooling, including liquid-to-chip methods, targets heat sources by circulating cool liquid through a 'cold plate' in direct contact with components like GPUs. In contrast, immersion cooling submerges entire servers in a thermally conductive liquid, efficiently dissipating heat across all components. Let’s delve into this in more detail. 

Rack-based liquid cooling: 

Rack-based liquid cooling is a method where coolant is circulated through the racks that house servers, absorbing and removing heat directly from the equipment. This approach is highly efficient, allowing for better heat management and enabling higher server densities within DCs. By reducing reliance on traditional air-based cooling, it can significantly lower energy consumption and operating costs. However, implementing rack-based liquid cooling requires significant upfront investment and infrastructure modifications. Additionally, managing potential leaks and maintaining the cooling system can be complex, posing challenges for DCs not initially designed with this technology in mind.

Immersion Cooling:

Immersion cooling submerges servers in a thermally conductive, dielectric fluid that absorbs excess heat and stabilises IT equipment temperatures. The process can be single-phase, where the liquid remains in a constant state, or two-phase, where the liquid evaporates and re-condenses to remove heat. Immersion cooling offers significant benefits, such as drastically lowering PUE levels – sometimes as low as 1.10 – by eliminating the need for traditional air cooling systems like CRAC units, which also reduces noise and saves space. However, implementing immersion cooling requires specialized infrastructure and careful management of the dielectric fluid.

Direct-to-Chip Cooling:

Direct-to-chip cooling involves circulating cool liquid through a system that directly contacts the chips and other heat-generating components. This method is more efficient than immersion cooling in dissipating heat at the chip level because the circulating fluid is typically cooler. Direct-to-chip cooling allows for higher compute densities without needing additional space, making it an excellent option for upgrading existing air-cooled DCs. However, it still requires supplementary cooling for other IT equipment, typically using chilled air, which adds some complexity to the overall cooling strategy.

Hybrid approaches for a balanced PUE  

While liquid cooling is an innovative solution, this technology is not yet in a position to completely replace air cooling in DCs. Even if equipment is cooled by liquid, heat will be transferred to it and some of this will be dissipated into a room or surrounding space where air will be required to remove this. Therefore, a hybrid approach is the best option where liquid and air techniques are used together to offer the best balance of PUE performance. 

Furthermore, it is important for organisations to know what options are available to them by working with a DC partner that fully understands their needs. DC providers must be able to review and manage their operations in real-time to ensure efficiency and to meet corporate sustainability goals. 

By using modern data analytics tools, businesses can monitor power usage, internal and external temperatures, and electricity usage for cooling to optimise processing loads for cost-effectiveness and proactively monitor equipment maintenance. Hybrid solutions are increasingly being adopted to combine the strengths of different cooling methods, improve PUE, and pave the way for a more energy-efficient future.