While CFD is the current best practice for datacenter layout/design; the FLIR images provided a granularity that you can not get from a theoretical model – for example, this illustration is the front of a rack where the client has stacked storage servers on top of one another – the servers in this image did not have sufficient air flow pull (a design flaw of this particular server chassis) and thus have to either be repositioned and/or directed pro-active cooling to be funneled directly to the server positions in the rack. We would not have known this otherwise…

Furthermore, this rack’s position in a sealed “cool” row is the anomaly as the rest of the racks all read 68 degree F at the top of rack’s input and 60 degrees at base of racks…

Lastly, we also were able to see exactly where each “open” air space in the rack actually created “hot leaks” back into cool row that were unforeseen both by the rack designer and actual layout.

Whilst CFD is a great tool for the drawing board, having a strict regimen of monitoring and spot checks is the only way to ensure proper air flow as per our SLA’s…all that said, we are always open to suggestions and collaboration as we certainly can learn from everyone else’s work and effort, conversely, we are always open to sharing our own lessons (we are an open book) – :-)

The stuff this will show would be a missing “blank”, a fan malfunction, that sort of thing. Limited usefulness. CFD is nice because it shows the impact of changes on the entire room airflow. Still, its tough to beat the coolness and marketing factor of doing handheld FLIR – throwing a few of these pix in your sales powerpoint will garner serious geek points.

Still, you can get a lot of this value with a cheap infrared thermometer. Not as cool, however :)