A growing number of data center operators which I have been speaking with seem to have awoken to an alternative data center cooling approach based upon Pressure rather than Temperature. I guess it makes sense that Pressure would not necessarily be the first idea that pops out in someone’s head if everything else they know about cooling is based upon Temperature. But there really is a second approach to cooling and BOTH strategies are exciting, when done well….
The cooling concept is simple. Cooling is about optimizing heat transfer and essentially tries to match heat removal (using air or water) with heat generation (cpu and memory chips primarily). In a data center, each piece of equipment generates a known quantity of heat (actually a range) and each data center has a specific number of these sources. Simple math to figure out how much heat needs to be removed.
The trick is deploying a technology that most effectively (re)moves the heat. No more, no less. One strategy is to provide cool air at all of the INLETS (at a known temperature value, typically 68 to 74-degrees). Getting this cool air everywhere it is needed is a challenge and over-cooling is very common. The main metric: As long as ALL of the air aisles are constantly measured to be in that range, everything appears to be working well. Deploy a few temperature sensors and many data centers get their systems up and running. Is this optimal? Not really. Like eating lobster with a hammer, it gets the job done, but not very elegantly. Adding more sensors helps ALOT and when coupled with active management discipline/processes, the temperature values can identify cooling inefficiencies. Remember, too much cooling costs alot of REAL money. Temperature is a really good approach to cooling management when there are enough sensor points. (And many IT devices even have built-in sensors on the inlets to help with this goal).
A second approach is to manage by Pressure. Simply calculate how much air flows through each piece of IT gear based on manufacturer’s spec, add it all up, and then inject that amount of cool air into the front or ‘cold’ aisles. Heat will be transferred. If done properly, the two figures match perfectly. Too much air and the pressure rises and cooling resources are wasted. Too little and temperatures rise IT gear is at risk. Pressure management is about balance. (Think of this like a household bucket filled with water, a garden hose adding more water and a bunch of holes in the bucket leaking water out. As the flow from the hose is raised, there will be a point where the leaking water EQUALS the new water and the system will be working perfectly. The pressure of the water remains constant). This equilibrium value is the ideal Pressure goal used as a ‘normal’ for each part of the data center.
The best news: Most modern sensor vendors now offer BOTH types of devices, Temperature and Pressure, in easy to deploy systems. These systems WILL allow either approach to be used to optimize cooling in the data center. While it is clear that there are at least two camps of thought here, the technologies required to support both are now mature and commonly available. The only downside of sensors for the data center, is NOT deploying them! Sensors are STRATEGIC when deployed in enough quantity and along with a fully formed operational ‘plan’ on what to do with the information provided. Choose pressure or temperature, but choose something!