Datacenter design used to be a lot of trial and error, and "flood cooling" (throw ever more tonnage of air conditioning at the heat load until stuff stops overheating).
This is inefficient, hideously expensive, and frankly nobody worth their salt does it this way anymore: Either a "standard" datacenter design is used (known volume and layout, estimated heat load, and balanced cooling) or a room is actually designed and engineered for efficient cooling.
Engineering airflow in a room isn't an easy task (as Iain mentioned, it's really the job of an engineer - preferably one who has built datacenters before and knows what they're doing].
There's a lot to consider:
- Intended Airflow
- Containment (keeping Hot and Cold zones separate)
- Raw cooling required ("heat load")
- Distribution of the heat load (which is often overlooked)
- Failure modes (and how to handle them gracefully)
(and that's just the heat/airflow -- there are human factors that need to be considered too)
As Chopper3 noted, the tools used for this are CFD (Computational Fluid Dynamics) packages -- many of them are the same things people use for designing aircraft, as you noted (and yes, the really good ones often take some substantial training to use well).
There are also some that specifically cater to datacenter design - I've personally used one called CoolSim which I found to be "adequate" (most of my other CFD experience is from engineering classes, modeling wings and such, and I've not used any other datacenter-specific tools, so take that for what it's worth, which is "not much").
You should also never underestimate the value of walking through a room with a smoke stick (especially once it's live) to see what having machines in the room really did to the airflow.
(Please remember to cut out your room's smoke sensors first, and ensure that no sensors are tripped before you turn them back on.)
In addition to modeling software there's also monitoring techniques you should be investing in. Pretty much every modern server has a small army of temperature sensors inside, and good PDU/CDUs have the ability to support probes that report temperature and humidity as well.
This can data can be grabbed via SNMP and fed into monitoring software (or custom-written stuff) in order to generate a real-time plot of your datacenter's temperature, and allow you to pinpoint "hot spots" before they actually become problems.
Depending on how much time (or cash) you're willing to spend this can be anything from a simple
Page someone if the temperature is above X degrees alerting system to a
Here's a thermal map of the datacenter for the last 5 days, sampled every 30 seconds and time compressed. You see where the temperature dropped on Tuesday? That's the old SAN we decommissioned being powered off.