There’s always a sense of irony when data centre cooling methods become a hot topic. The eruption of data we generate on a daily basis demands processing power of seismic proportions. Of course, with this comes the troublesome task of keeping the data centre within the correct environmental conditions. Most of the heat in a data centre is generated by computer chips, CPU’s, GPU’s, and memory modules that consume a fraction of the facility’s real estate. Processors alone run between 75°C and 100°C on average, while larger data centres can produce enough energy to heat a skyscraper. So often the case when building a data centre, an efficient cooling system not only represents the biggest challenge, but an enormous opportunity.
There have been many inventive methods that companies have used to cool their facilities, but one of the most innovative must be direct to components. To anybody unfamiliar with the technology, managing the waste heat energy of electrical components with liquids might sound unconventional, if not downright crazy. However, considering how small the electrical components responsible for the heat output are, direct liquid cooling method actually makes a lot of sense.
Usually, the hot components are kept cool by an electronic device called a heat sink. These heat sinks often utilise a fan to cool the component or aluminium-finned radiators that dissipate heat through convection. Direct Chip Cooling essentially replaces the heat sinks with cold plate units that allow cooling liquid to be routed directly to the component. It is able to capture the waste heat and recycle it via liquid-to-liquid heat exchangers. This means higher operating temperatures and the removal of compressor dependant cooling. This reduces overall energy consumption and improves energy efficiency, especially in higher ambient locations. Also these higher operating temperatures makes the reuse of the data centre waste heat more practical.
Suddenly, using liquids to cool electronic components makes a lot of sense. After all, it isn’t as if they are directly in contact with the liquid. Unless of course you decide to make a splash and submerge your entire component rack. Yes, it is something that sounds so impractical that it couldn’t possibly work, and yet it does sublimely. Engineered as an efficient way to reduce energy consumption whilst optimising performance, it is a solution called Immersion Cooling. But what makes it possible?
We all know that water and electricity generally aren’t the best of friends, but that’s mainly because water contains impurities that make it conductive. The difference here is that the fluid in which racks are submerged is completely dielectric, meaning that the components can operate normally. Through direct contact with them, heat is removed via the fluid.
A further advantage of immersion cooling is that it allows for a higher density of components. The computing power per square meter that conventional air cooling allows for, dielectric fluids can accommodate ten times over.
Big Data is on a meteoric rise, now more so than ever. 90% of the data in the world today has been created in the past two years alone. It estimated that by 2020, 40 trillion gigabytes of data will be created, and the demand for energy efficient solutions will grow with it. The question has always been how to make our facilities as efficient as possible and liquid-based solutions could very well be one answer.