Update: the link to the paper is now working. Duh. Sorry.
Sleep is another of those biological phenomena of which we lack a good theoretical understanding, despite there being lots and lots of (good) hypotheses and experiments. Van Savage and Geoff West have suggested a first step to building one, though. And (wouldn't you know it) it's based around metabolic rate.
The amount of time a mammal spends asleep, they find, is proportional to their relative, or cellular, metabolic rate. A fast-burner like a mouse sleeps for about 14 hours a day, whereas an elephant kips for only 3.5 hours. What's cool is that they find that sleep time is most closely matched to the brain's metabolic rate. Their suggestion is that sleep duration corresponds to the amount of time needed either repair brain cells, or reorganize them to process the day's input.
So unlike other organs, perhaps you need to shut down the (conscious) brain to maintain it. Fast-celled animals need more time for repairs (which, if this model holds, seems a more likely function of sleep - or at least of metabolically related sleep - than reorganization), because they do more damage while they're awake. The same argument applies to why small animals have shorter lifespans - fast cells, more free-radical damage, shorter lives.
What's that Lassie? A lot in life seems to depend on metabolic rate? Perhaps by understanding metabolic rate we can inch towards some kind of powerful, simple theory to explain a wide range of biology? You wish someone had written an accessible guide to this vibrant field and its long and strange history? Well, girl, it's your lucky day.