The Google Teakettle - Was The Times actually that wrong?

Latest twist in the tale of the Google datacentre tea boiling system (our original coverage here) is that the original author said that he didn't say what the Times said he said, but the Times said it instead. TechCrunch:

....according to Wissner-Gross he never said anything of the sort. For starters, he says he would never refer to any sort of measurement having to do with tea (he’d go with coffee). But his findings have nothing to do with Google as a company, either - they’re concerned with much more generalized stats, like your computer’s rate of CO2 production when you look at a webpage.

Wissner-Gross says that the widely circulated 7 gram/search figure came from some other source (he’s not sure where), and notes that if you read the article carefully it only makes it sound like it’s from his data. He has confirmed that he did make some vague statements regarding Google, including “A Google search has a definite environmental impact” and “Google operates huge data centers around the world that consume a great deal of power”. But the “tea kettle” statistic that has been repeated ad nauseum simply isn’t his.

As I wrote in the TechCrunch comments section, this still doesn't quite ring true.

While the original figures are possibly overblown as some have noted, Google’s own figures could do with a little hard probing rather than just taking them as gospel. I did these rough calcs below in my original post:

Google notes that their system consumes 0.0003 kWh per search. If you take c 200m searches / day as a 2008 mean, at that rate you get c 60,000 kWh per day, or 21.9m kWh per annum. Given that the average UK citizen consumes in the order of 8,000 kWh per annum all in, that is in theory about 2,700 UK citizens worth - ie according to Google’s data, they should consume about the equivalent amount of power annually as a typical village for all Search needs.

However, Google’s infrastructure is nothing like village sized - Google themselves have indicated before that their power requirements are something like half that of a city like San Francisco (pop 800,000).

So something is not quite adding up - that implies Google Search in totality is c 1% of Google's energy usage..

And funnily enough, if you take the original and much berated Times / Wissner-Gross estimates of c 1/2 a kettle boil (c 0.05 kWh) per search, and run the above numbers through it, you wind up with Google needing the energy of a city of about 450,000 people.

So, if Google ain’t burning that energy in search, what exactly are they burning 99% of their energy on?

We suspect that Google is also being a bit disingenuous, and the quoted energy usage for search is merely the direct component and neglects a lot of indirect energy usage.

There are two things this episode has brought up that need to be addressed by the Webnaughts:

Firstly, the IT and Web industry does need to stand up and understand clearly that its consumption of energy is increasingly becoming a global issue (as Nick Carr noted, a Second Life avatar consumes more energy than the average Brazilian). Big Datacentres (Google's included) have massive energy impacts - they are the "Satanic Mills" of the 21st century and increasingly have to be situated close to direct power transmission and natural cooling like the great industrial plants of old. Undoubtedly some of the IT industry energy usage is replacing more costly approaches, but there is also an argument (as the original article was pointing out) that ease of use - eg of Google Search - potentially creates far more energy use than was originally replaced.

Secondly, the reporting on the original article and the Google data response by many bloggers (and some mainstream media journalists, it must be said) was by and large woeful, with very little real analysis or in many cases understanding of the data being presented and what it meant. As I noted in an earlier post, the issue with the blogosphere taking over as Citizen Journalists is that the Noisy Know Nothings will always drown out those who do grasp the issues, and it was shown in spades here.

It also must be said that the original quoted physicist, Mr Wissner-Gross seems to be doing rather well out of the alleged Times misreporting (which, as we've shown above, is - perhaps accidentally, perhaps not - probably not that wildly inaccurate) which also makes one wonder if The Times did get their end of the stick so very wrong. The simple "black box" analysis of looking at Google's posted energy needs, dividing by the reported number of searches per day, and calculating the implied energy per search gives a number not that far off the original half a kettle boil as we show, so we suspect they are not two orders of magnitude out, as Google implies.

There is another shoe to drop here methinks.