I given to understand that Google invests a lot in energy solutions for its data centers. The problem is that there is a time/space tradeoff for most computational problems. The less time you want to take to do a search, the larger the index you need. Effectively, Google’s main work is anticipating the searches that MIGHT be made. The more searches they anticipate, the faster the processing time.

The more searches they anticipate, however, the more data they have to store. That means running more hard drives. And more energy.

Of course, they also invest heavily in efficient algorithms, but the physical limits on this are much more easily approached at Google-scale computations than simply hosting a larger index. Hence Google must necessarily be an energy-intensive operation, whether that claim in the link is true or not. (And I’m often skeptical about those kinds of Fun Fact calculations even when done in a sophisticated way.)

I’m surprised that the Googleplex takes only 3300 homes worth of energy. Considering the number of people who work there and the fanciness of the digs, that actually sounds pretty efficient.

The power load is there whether anyone searches or not. “Running a search” doesn’t add anything measurable. All big resources, like Google, require power/energy to build and maintain.

Guestimates are that Google is running about 500,000 servers now, scattered across maybe ten sites. The big complex in Mountain View CA runs on power from the Hetch Hetchy hydro plant in the Sierras, so no emissions there. The newest one will be in Council Bluffs Iowa and is designed to run on power from a wind farm. They dumped a lot of money for energy efficiency upgrades into the old SunMicro buildings out here, and they’ve contracted with a architectural design firm whose name escapes me now that specializes in energy efficient construction. All in all they’re a pretty green-oriented bunch.

Dunno about that number as even a loose average of total power consumed divided by searches run. No substantiation in the linked article, and seat of my pants it seems way, way too high. I call bullshit; can’t be even 1/100th of that. This is a Fake Fact.

Calculation = computation = computing power = electricity = heat …

A baseline load is there whether anyone searches or not, but that is not the same as a system serving up search results.

[x] Any (D) in the general. [ ] ?????. [ ] Any mullah-sucking billionaire-teabagging torture-loving pus-encrusted spawn of Cthulhu, bless his (R) heart.

I assumed that they were taking into account the baseline load. The whole point I was making is that a large continuous baseline load is REQUIRED to run a fast search system.

I have more than a little bit of professional experience in this area.

Will it save 11Wh if you don’t do a search on Google today? In general, no it won’t.

If many people stop making searches so total searches drops enough, they will probably start turning machines off in their grid. (Probably an automated process.) That will make real power consumption go down some.

Loading the individual servers up and down will change power consumption some depending on what kind of CPUs and OSes they’ve chosen to implement on. But this is even less savings. A very good CPU can scale its power consumption up and down an order of magnitude or so, but this doesn’t affect the power draw of the whole system as much.

The big differences come in in planning data center expansions, so that means if many people stop using google for several months, it could make a large impact on planned increases of power consumption.