Humans might not seem naturally energy efficient — after all, your teenage son’s only got three energy states: sleep, eat and Grand Theft Auto. But we humans have actually evolved to use energy only when we need it. When we’re running around, we burn a lot; when we’re kicking back, we burn very little. And, it turns out, that efficient use of energy is something that’s actually pretty difficult to replicate in the technology world.
But Google thinks we should try when it comes to servers. This week, the search engine giant published an online tome explaining its strategies and thought processes surrounding really large data centers, or “warehouse-scale computers” (WSC). In one section of the 120-page “mini book,” Google drills down into “Energy and Power Efficiency,” saying that servers should be redesigned and enhanced with software so that they can be as energy efficient when used lightly (say, when web services are pinging them less frequently) as they are at maximum use (during the livestreaming of Obama’s inauguration, e.g.).
At the heart of the idea is the notion of “energy proportionality.” A system that’s “energy proportional” is equally energy-efficient across its range of activity levels — basically, the system consumes little energy when it is being used only a little bit, and a lot more energy when it is doing a lot. Biological systems (like you and me) have naturally evolved to be this way. Think about it: after millions of years, an organism whose very existence depends on how effectively it can consume and use resources will reach a more energy proportional (and efficient) state.