Imagine if someone told you they had submerged your PC into a tub of boiling liquid while the system was still running. You'd probably feel a mix of emotions, including anger, shock, and bewilderment. Well, nobody's going to do that to you (hopefully, anyway), but Microsoft did do that very thing to itself, plunging densely packed server racks into a steel tank filled with a special liquid developed by 3M.
Unlike naturally occurring water (read: not pure H2O (opens in new tab)), this specially engineered liquid does not conduct electricity and is actually an effective insulator. The general concept is not new, as PC users have been dabbling with mineral oil for over a decade. But Microsoft is embarking on uncharted territory as a cloud provider.
"We are the first cloud provider that is running two-phase immersion cooling in a production environment," said Husam Alissa (opens in new tab), a principal hardware engineer on Microsoft’s team for datacenter advanced development in Redmond, Washington.
Interestingly enough, Microsoft says it is applying lessons learned from cryptocurrency miners, who have been employing a similar method of cooling to help mine Bitcoin and other digital currencies.
How it works is cooling coils snake through a tank shaped like a couch filled with 3M's liquid. From there these coils connect to an external dry cooler. The heat exuded from the submerged server racks is then transferred to the liquid, and 3M engineered the liquid to boil at just 122F (50C), compared to water, which has a boiling point of 212F (100C).
Due to the low boiling point, fluid in the coils is never hotter than the surrounding air, negating the need to douse them in water to assist with evaporation. It's essentially a closed-looped cooling system.
"The boiling effect, which is generated by the work the servers are doing, carries heat away from laboring computer processors. The low-temperature boil enables the servers to operate continuously at full power without risk of failure due to overheating," Microsoft says.
It also facilitates a bit of CPU overclocking. Not for constant use, but to handle increased workloads that occur at certain times. For example, in mid-afternoon, there is a burst in computing power from Microsoft Teams, as more people join at the same time. When something like that happens, the servers can run faster while still staying cool.
"If done right, two-phase immersion cooling will attain all our cost, reliability, and performance requirements simultaneously with essentially a fraction of the energy spend compared to air cooling," added Ioannis Manousakis, a principal software engineer with Azure.
Microsoft's hope is that its servers dunked in 3M's liquid experience low failure rates on par with the underwater datacenters it is testing as part of Project Natick—servers in specially designed tubes, sitting on the seabed.