Hi Heath,
The thing with darks is that you want to get *exactly* the same amount of of noise and warm/hot pixel level readouts as for your light frames; that is, you want to replicate the exact same circumstances that you're taking the light frames in.
Then afterwards, by comparing the two, you can perfectly correct the light frames as you know what part of the signal was 'false' (e.g. it was not part of the actual image, but part of the noise) and which part of the signal was 'real'.
The longer your exposure, the more noise you gather, so the 'lighter' your dark frame.
The higher your ISO the quicker you gather noise, so the 'lighter' your dark frame.
The more frames you take, the smaller the random noise fluctuations, but the more consistent the read out noise and thermal noise becomes.
Strictly, you don't really need to take the same amount of dark frames as your light frames, but you should take at least enough to generate a smooth noise profile. But, just as with light frames, the more the better (but light frames have priority obviously

).
EDIT: Bartman beat me to it