Ben Hatton Direct Liquid Cooling: The Future for Data Centers?

December 15, 2014 by · Leave a Comment 

Direct liquid cooling

An application of direct liquid cooling in a data center. Photo courtesy of Green Revolution Cooling

For data centers, cooling is everything; not only is it what keeps colocated equipment running efficiently, it also makes up the majority of a data center’s power usage. Most data centers, including Data Cave, rely on cooling systems that use air for transferring heat away from equipment, while simultaneously supplying chilled air. However, a cooling method that was developed back in the 1960’s and is now seeing somewhat of a resurgence, is direct liquid cooling. Rather than using air, this method involves physically submerging servers in a dielectric fluid to keep them cool (NOT regular water). While this method of cooling has been anything but mainstream in the industry, there appears to be a new interest in it, according to a July report by 451 Research.

The report indicates that there has been some growth in the number of liquid cooling providers, as well as new interest in liquid cooling by larger companies that require very high density computing in their data centers.

Why liquid cooling hasn’t become mainstream

While the direct liquid cooling method saw a fair level of practical application cooling mainframes during the 60’s and 70’s, it eventually lost out to cheaper but equally efficient air cooling methods in the 1980’s. It was during this time period that the modern data center as we know it began to take shape, and air cooling quickly became the norm within the industry.

There are a number of obstacles that currently prevent most data center operators from considering liquid cooling as a viable option, and here are a few of them:

  1. Little to no vertical scaling: With this method, servers are submerged in large vats of fluid that are only accessible from above. This inhibits the ability to scale servers vertically (or stack them), since you have much less vertical space to work with. If a data center is already equipped with very tall cabinets stacked with equipment, then moving to a liquid cooling system would require a great deal of re-configuration.
  2. Not all server types are supported: Since liquid cooling involves the physical equipment being fully submerged in liquid, not all types of hard disk drives can be cooled this way. Drives must be designed to not only tolerate the liquid, but they must be fully operational while submerged as well. While this type of server equipment exists, any incompatible servers would have to be upgraded as well.
  3. Costs are too high for most data centers to justify: In addition to the costs involved with updating the server equipment itself, the numerous other costs of implementing a liquid cooling system are quite high. These include the physical equipment needed (the large vats as well as supplies of dielectric fluid), maintaining that equipment, and modifying the existing infrastructure and data center layout to accommodate for this equipment. In all likelihood, it would be more cost effective to start from scratch with liquid cooling, rather than trying to retrofit it into an existing data center.
  4. Extra precautionary measures are required: Data centers have always been environments where any type of liquid has been strictly forbidden(*), so the idea of introducing a liquid into that environment is something that would definitely raise a few eyebrows and cause a bit of fear on the data center operator’s part. It would also necessitate additional measures and physical changes within the data center to prevent other pieces of critical equipment from coming into contact with the cooling fluid.

Ultimately, it is the steep costs and long-term infrastructure changes that are the major barriers keeping most data centers from adapting liquid cooling anytime soon.

What it would take for it to become mainstream

While there has been some renewed interest in liquid cooling by some larger data centers with very high capacity computing needs, I think it will still take considerable time before it begins to gain traction throughout the industry. Here are a couple of things that could help it to take off though:

  1. Widespread adoption of high density computing: There has been a growing shift in the industry towards high density computing, a scenario that would be well served by liquid cooling. However, this isn’t a shift that has happened overnight, and until high density data centers start to really become the norm in the industry, liquid cooling won’t see a great deal of growth.
  2. Adoption by the industry trend setters: Many of the key trends and directions that the data center industry moves in are set first by some of the world’s largest tech companies (ie. Google, Facebook, Amazon, etc.). The computing and capacity requirements for these companies are incredibly above average, and they are often the pioneers in developing and working with new types of data center technologies to meet those needs. If many of these larger companies begin adapting liquid cooling as part of their data center operations, the rest of the industry would definitely take notice and likely follow suit over time.

Liquid cooling for data centers definitely has the potential to become a legitimate standard for cooling within our industry. It’s not there yet, but as the industry continues to evolve, there is a good chance we’ll be hearing more about it in the future.

Share this with your friends!

Share on Facebook Share on Twitter Share on LinkedIn Share on Google+

If you liked this post, sign up for our monthly newsletter!

Newsletter Signup



Speak Your Mind

Tell us what you're thinking...
and oh, if you want a pic to show with your comment, go get a gravatar!