Up-front and Lifetime Costs
A few years back my good friend Craig Meyer pointed out that in the days when the Chevy small-block engine was designed the primary criterion for engine design was how many horsepower the engine could produce for a given cost. Of course the packaging had to be reasonable, as did the ease of service and manufacturability, but both of these feed back into overall cost in the end. Efficiency and pollution were, if they were considered at all, at most afterthoughts. It took the oil shocks to change that mindset. I bring this up because until recently computers have been in the same boat. With the exception of laptops, the same criterion has been applied to computers in general and microprocessors in particular: how much power can we squeeze out of every dollar? But it seems like Google, among others, is running into an electricity shock: lifetime costs of servers will soon outstrip up-front costs. This, and a lot of other things suggest that maybe Wally Rippel, the resident mad scientist at my previous employer, was right. He said something along the lines of "[the last] century was when we learned the value and limitations of data; in [this] century we'll learn the value of energy."