There was an interesting post over at the Datacenter Knowledge blog a couple of days ago about the increasing use of Computational Fluid Dynamics within datacentres. I thought it would be useful to share some details on how we’re using this ourselves.
Powering the cooling infrastructure within a datacentre is one of the largest components of the operational expenditure related to running these facilities. As such, it’s been very important for us to fully understand the impact that various design options would have on our new Centro datacentre facility.

CFD Model of various rack layouts within Gyron hosting suites
The model above shows a horizontal temperature plane with a number of different rack layouts, cooling technologies, CRAC unit placements, and containment options. As you can see between the right and left hand sides, even very minor things like moving structural columns by a few centimeters can have big knock on effects… and in a lot of cases you can take some quite counter-intuitive steps to actually improve your cooling efficiciency (for example, installing baffles to slow down air flow through some sections of the hosting suites, removing some vented tiles, increasing the target set-point on the cooling management software etc).
In addition, now that we have a model of the facility in the CFD software, we can quickly and cost effectively plan out large customer deployments and verify that higher density rack loads will be cooled effectively before equipment is actually installed. We can even build these models in to our services on a contractual basis to give customers comfort that they will receive what we promise in the sales pitch, and when coupled with thermal imaging and enviromental monitoring built in to our infrastructure can form part of our Service Level Agreements.