[Rmgill]

This is stuff you can estimate, but until you get systems in the data center and you're working with them on a daily basis, it's still quite nebulous at times.

Things that help:
Environmental monitoring from Netbotz or a similar company (at a minimum, walk account with a temperature probe or just go by feel if you're a really good scot's engineer...."I dunno captain, but the dilithium matrix just doesn't feel right, I need to run more tests"

Power-strips that display per-circuit amperage readings, this allows you to fine tune system installs in a rack and really know when you CAN'T add another system to a rack. I've had engineers sneak in and install more memory and hardware to a system such that it's power jumped, if I'd gone on the old power readings, I'd have been screwed if I'd installed more hardware on that circuit. Pulizzi makes good stuff. So does PDI and a few other companies.

Run the space colder if you can. I like 60°. If you have a CRAC/Air Handler Failure, the extra buffer gives you more time. In a room that's at capacity, you may have 10 minutes between a unit failure and 80° ambient if you run at 70°. Drives start throwing errors when my room reaches 85° ambient (it's usually 90° at the intake of the top of a rack) and then you start seeing sun's spontaneously shutting down to protect the processors.

If you can add environmental monitoring to the room, and use SNMP, you might be able to snag temperature data from systems themselves, either as SNMP traps or some other device crawler function.