In this Q&A, Robert Bunger, liquid cooling program director, CTO Office at Schneider Electric, discusses all things liquid cooling and how an innovative combination of immersive and direct to chip methods can provide facilities managers with the best of both worlds, whilst doing away with the disadvantages.

Firstly, what is currently driving the adoption of liquid cooling?

Changes in IT technology have always been a key driver in the development of new cooling solutions. Although liquid cooling isn’t new, having been deployed for years in high performance applications, modern day demands such as the cloud, IoT, AI and edge are all forcing a renewed look at liquid technology.

Right now, traditional air-cooling methods are reaching their limits, as IT chip densities continue to increase, which is currently what’s driving the demand for liquid cooling. The onus placed upon energy efficiency and sustainability is also more prevalent than ever, as well as new IT use cases that we weren’t seeing before, such as harsh and industrial edge computing environments. This culmination of challenges is pushing data centre operators to look to new and improved cooling solutions.

What are the main methods of liquid cooling currently?

There are two main methods at present, the first, is ‘direct to chip’, which comes with its own advantages and disadvantages. Starting with the positives, direct to chip liquid cooling can be easily retrofit into an existing chassis or standard rack form factor and is maintained similarly to an air cooled set up.

That said, this method does carry a risk of leaks, meaning having to invest in leak prevention measures, which of course carry an additional cost and energy consumption that most operators could do without. This method also isn’t ideal for harsh environments, nor does it capture all heat.

The second method is oil-based immersive cooling. This method is not only simple and cost effective, but eliminates all heat as well as the need for fans. Pitfalls here include messy maintenance (an oil bath is never going to be too pleasant), an unusual form factor and limited flexibility.

This got us thinking, wouldn’t it be great to have the best of both worlds where some of these disadvantages were mitigated? And that’s where chassis-based liquid cooling comes in.

What are the advantages of chassis-based liquid cooling over conventional air cooling?

Well of course if you want your facility to be able to cope with ever-rising chip and rack densities, then this type of liquid cooling is going to be advantageous. Chassis-based liquid cooling works with dielectric fluid contained in the chassis itself. This means liquid doesn’t come all the way into the server, nullifying the potential risk of leaks.

This technology also reduces the need for as much water, achieving effective cooling using very little. Space restrains are also no issue here, with liquid cooling taking up three times less data centre space than conventional air cooling equipment.

Additionally, there is minimal heat added to the room, whilst still being able to do away with fans altogether. The absence of fans not only eliminates noise, but can save anywhere from 4 – 15% on energy consumption and therefore overall costs.

Waste heat can also be recovered, which if put to good use, i.e used to heat a building, will help reduce the operational expense of a facility, not to mention its carbon footprint, adding to that all important sustainability factor.

Isn’t liquid cooling more expensive than traditional methods?

Many people assume liquid cooling costs more, when in fact this improved technology ends up costing just the same if not less than air cooling methods. Liquid cooling provides far greater savings in the long run when you factor in the compaction it enables. You are getting rid of a lot of costly facility equipment i.e. chillers, fans, piping changes etc.

Then there is the simple fact that the smaller a facility, the smaller your outgoings. There will of course be a shift in costs, but when you start space saving/compacting you in turn need less lighting, less money spent fitting out a building, the list goes on. You’d be surprised how quickly these savings add up.

If I wanted to buy a liquid cooling solution right now, what factors should I take into account to make the best choice for my business?

 There are three key things to consider:

1. Overall timeline: Every new data centre deployment needs clear and concise planning, so the first step is to understand your project timeline. Schneider has worked closely with Iceotope and Avnet to create the industry’s first commercially available liquid cooled IT solution, but by comparison, this is not as readily available as air-cooled products, meaning that there is still a lead-time. There are also different types of liquid cooling to consider, so I recommend the end-user begin investigating the type they wish to deploy, even in the initial discussion stages. 

 2. Facilities readiness: The next step is to ask if the system is going into a new build facility or being retrofitted into an existing data centre? Can you, for example, easily install a warm water loop to remove the heat from the IT rack? If not, it will need some work before it is ready for liquid cooling. Today there are many skilled engineering firms who have the expertise to convert air-cooled equipment into liquid cooled solutions quickly, which is great if you are planning to upgrade an existing facility.

3. Type of liquid cooling: Finally, there are many factors involved in the decision of which type to deploy and it is best to seek consulting advice that’s specific to your situation, needs and business objectives. Key factors will include; solution availability, chip Thermal Design Power (TDP), chip performance requirements, the type of environment (whether in a traditional facility, a distributed IT application or harsh and remote edge computing location), the noise level, the potential increment of growth (server by server, or rack by rack) and the frequency of IT upgrades or maintenance. For those looking at deploying the technology, Schneider Electric White Paper 265, ‘Liquid cooling technologies for data centres and edge applications’ gives a detailed account of types of liquid-cooled systems available today.

Is liquid cooling a sustainable solution, or could it be in the future?

The answers are yes and yes! Liquid cooled IT can be cooled by warm water, so it is very easy to make an energy efficient data centre of any size, in any climate. 

Additionally, less water will be used at the facility level for cooling, saving a precious resource, which is scarce in many locations. For chassis-based immersion-based cooling, there are a variety of different dielectric fluids available, each with its own trade-offs.  

Containment, reclamation programmes and research into new fluids are already underway, giving promise to even better options in the near future. Another side effect of immersive cooling is greater reliability of IT equipment. Early indications show that the stable temperature of immersion cooling leads to fewer component failures, improving the circular economy metric of durability.

Finally, how long do you predict it will take liquid cooling to become the go to method of thermal management?

I think it’s definitely on the way and I’d predict we are about two years out from this becoming the case. However, in order to secure widespread adoption, the industry needed a commercialised solution, which would require some collaboration from vendors.

With that in mind, here at Schneider we wanted to merge the positive aspects of both direct-to-chip and immersive strategies, while at the same time eliminating the disadvantages and we figured three heads would be better than one.

So, taking care of the physical infrastructure, as mentioned above, Schneider Electric teamed up with liquid cooling vendor Iceotope, alongside systems integrator Avnet. Together, we combined our knowledge, experience and technology to come up with an integrated rack with chassis based liquid cooling.

With this solution, our aim was to take the convenience that direct-to-chip methods bring and combine it with some of the performance that you can get out of immersive solutions. No one has ever offered a rack-based solution in the past and we are proud to be able to say we are the first.