Thursday, December 3, 2009

Energy footprint of Google Searches

I like Google. Slowly, step by step, I have started to use more and more services from Google. It started with searches, then continued with Gmail, Google Earth and a number of other applications and services (including Blogger, which was bought by Google in 2003 and which provides the technical platform for this blog).

I have found almost all the services I have tried useful (for myself or at least for someone else), but it is now time to scrutinize Google through the lenses of my peak oil glasses. Even if I primarily write about Google in the text that follows, “Google” could mean any business competing with Google, or even the entire Internet with all the services it provides to us in our daily lives.

In the beginning of this year, the young Harvard physicist Alex Wissner-Gross claimed that Google searches contribute significantly to CO2 emissions. The angle that the newspaper chose was that two Google searches produce the same amount of CO2 (carbon dioxide emissions) as boiling a kettle of water for a cup of tea, and that Google searches thus have “a definite environmental impact”. More specifically, Wissner-Gross claimed that one Google search generates around 7 grams of CO2 emissions. The carbon dioxide originates from producing electricity to run your computer and to run Google’s data centers. A few grams of CO2 may not sound like much if not for the fact that the number of Google searches each day exceeds two hundred million searches, and may be closer to one billion according to some sources

Wissner-Gross further claimed that the architecture of Google’s search engines was an important factor as each Google search is distributed to several data centers across the world that then compete against each other to find and return the fastest answer. What you gain in speed thus comes at a cost of higher energy consumption caused by all the extra computer capacity (unused, misused or redundant) built into the larger system.

To no one's surprise, Google rejected this interpretation and stated that the company is ”among the most efficient of all internet search providers” because their data centers are relatively energy efficient compared to "average" data centers. Google furthermore claimed that the number 7 grams CO2 per search is ”many times too high” and that the true numbers are 0,0003 kWh of energy and 0,2 grams of CO2 emissions per search. This small amount of energy is in parity with the energy burned by the human body in 10 seconds, and the CO2 emissions are thousands of times lower than the CO2 emissions caused by the average car traveling only a few kilometers. Another powerful formulation from Google is that “In the time it takes to do a Google search, your personal computer will likely use more energy than we will use to answer your query”.

Another comparison is with the numbers from this report (pdf) which claims that each spam mail ending up in your mail box on average generates 0,3 grams of CO2 emissions (the same amount as if you drive your car 1 meter). Since the number of spam e-mails sent during 2008 was approximately 62 000 000 000 000 (62 trillions), the total amount of CO2 emissions caused by spam is not insignificant and more precisely corresponds with the amount of CO2 emitted by a car driving around the world 1,6 million times. If we assume that there are about 800 million cars on Earth, then all the spam sent during 2008 corresponds to the accumulated CO2 emissions from all the world’s cars driving 80 kilometers each. I am not sure whether this is much or little in a big-picture perspective, but I have no problems being judgmental and deeming spam e-mails 100% unnecessary, and now for yet another reason. Where are the technological and social solutions to stop them`

I saw a reference in January 2008 stating that a "bizarre" record was broken one day in October the preceeding year (2007). During that one day, more than 160 000 million spam e-mails - roughly two dozen per man, woman and child on Earth - were sent. Comparing this number with the total number of spam e-mails send during 2008 (see above), we find that the record from 2007 is actually lower than the daily average of spam e-mails sent during 2008...

Other experts who have made claims about the energy use of (Google) searches state that CO2 emissions are between 1 and 10 grams (depending on whether you have to turn on your computer first), or between 7 and 10 grams (if you use your computer for 15 minutes). The Times of London, which published the original article (above), informed its readers a few days later that the newspaper accepted Google's official claim that one (simple) search (taking less than a second) produces only 0.2 grams of CO2, and that the "search" refererred to in the article involved several attempts over a teme period of several minutes. In a clarification by the physicist Wissner- Gross, he states that he never mentioned Google specifically, that the example with the kettle of water was not of his origin, and, between the lines, that the newspaper made a hen out of a feather based on the interview with him.

Maybe the number 7 grams of CO2 emissions per search originally came from this blog (May 2007)? We should anyway probably take Google's numbers with a pinch of salt since the company probably counts only the marginal cost of performing one extra search, and not the nergy cost for temporarily inactive servers, support and maintenance, and the distributet and thus reduntand work being done in several data centers for each search. Also, the idea that badly written inefficient software code wastes electricity and thus has a bigger ecological footpring than neat code is thought-provoking (although I do not mean to imply that this is a problem for Google).

The "news" about Google searches and kettles of water quickly spread to several newspapers, but seem to have been a storm in a teacup when viewed in the read mirror. So let us step back and think about the larger issues that this text touches upon. Using computers has an environmental impact. What we read and look at when we use our computers is stored on many servers that are all connected by computer networks. All of these parts require electricity (personal computers the most, servers and data centers the least, and computer networks in-between).

The electricity that we use for these purposes is generated mainly from fossil fuels like coal and natural gas (85% of the energy consumed on Earth is produced from fossil fuels). when Google states that each search generates 0.2 gram of CO2 emissions, they surely count only the energy costs for their enormous-but-highly-efficient data centers, whereas the big energy thief is right in front of your nose - your own computer at home or at work. The use of a personal computer may cause CO2 emissions of somewhere between 40 and 80 grams of CO2 per hour, and if you include this energy consumption it is easy to reach 7 gram of CO2 for an advanced search that constitutes of several steps.

As I wrote in the beginning of this text, "Google" could in this context represent something much bigger than the company itself and it is somehow difficult to understand why exactly Google was singled out for its energy consumption. In general, Google should be acknowledged for their energy policies and for their lobbying in Washington for cleaner energy sources. A few searches on the Internet (there we go again :-) reveal several examples of interesting and good inititatives from Google, like their report "Clean energy 2030 and their work on RE<C (renewable energy less than coal), where the goal is to produce renewable energy cheaper than electricity generated by coal plants. According to Google, the company's data centers use only half as much energy as the average data centers.

The global IT sector is responsible for 2% of the global CO2 emissions (according to the firm Gartner Inc.). It may not sound much, but it is as much as the global airline industry emits, and in contrast to an airline industry in crisis, the IT sector is growing rapidly on a global basis. Many actors consider numbers about energy use business secrets, and for exampe Google does not want to tell how many or how big their data centers are, or how many servers they own.

Google clims that using their search engine on the larger whole saves money and natural resources, since a Google search replaces more energy demanding activities - we no longer have to use as many car trips, time, paper or ink to have our questions answered. This is a valid argument, but it assumes that we are doing more or less the same things (the same number of searches) as before, but now in a more resource-efficient way. But we obviously did not perform a hundred million searches per day before Google and other search engines existed. Furthermore, we burn energy by doing a lot of new things with our computers which could not be done easily - or at all - before:

We may be obsessive about turning off the lights when we leave a room, but at the same time we may happily spend hours clicking around online, oblivious of the electricity lighting up our screen, heating our chip, and powering ad cooling the data centers we're connected to. (It's true that in some cases Internet use may substitute for other activities, such as travel, that would consume more energy, but let's not kid ourselves: the vast majority of computer and Internet use represents additional energy consumption.) How many Twitterheads think about their electricity use before they tweet? Not many. How many blogger think about it before they blog? Not this one.

More interesting than to examine Google in particular is to think about the energy cost of computer use in general. Alex Wissner-Gross (again) has calculated that each second of watching a web page generates 0.02 grams of CO2 emissions. This applies to "static" website content - if you watch animations or video, that number quickly becomes ten times higher. The rule of thumb is of course that the more you use a computer, the more energy you consume, and some activities (playing computer games, watching movies) are more energy intensive than others (reading a document, working with a word processor). Regarding the energy consumption of avatars, I wrote the following almost one year ago:

It is difficult to determine the benefit (or damage) of using virtual worlds. On the one hand you use considerably less energy (and generate a lot less in terms of CO2 emissions) if you cancel a trip and set up a meeting in a virtual world. A computer on the other hand uses a lot of electricity compared to non-electricity-consuming activites (go for a walk, talk with a friend, help your children with their homework).

To play World of Warcraft several hours a day can hardly be described as an activity which "replaces traveling". It is more probable that playing such a game for a long time increases the chance that you will make new (faraway) friends whom you would later like to visit (sometimes by hopping on an intercontinental flight). I am here walking on a minefield of trying to differentiate between "good" and "bad" uses of computers and the Internet. I prefer to avoid this particular discussion at this particular point in time, but might return to the issue later. We can at least for sure state that computers and galloping use of electricity may be problematic in the long run - a characteristic shared by all types of exponential development: "If not addressed, unlimited, ever-increasing compute performance will ultimately consume all the energy on the planet".

I think it is definitely legitimate to critically investigate the energy consumption and CO2 footprint of using for example YouTube, Twitter and virtual worlds. Even without approaching the issue in a normative manner (making claims about "good" or "bad" use), one may thus find clues as to which activities could become painfully expensive if the electricity and energy prices will rise and keep on rising in the future. According to a vice president at Sun Microsystems, it is totally clear that "We need more data centes, we need more servers. Each server burns more watts than the previous generation and each watt costs more".

Something to further take into account are proportions. A person who uses a computer one hour per day (40 to 80 grams of CO2 emissions per day) generates emissions somewhere in the range of 15 to 30 kilos per year. A hardcore computer user who uses his/her computer 10 hours per day thus generates something between 150 to 300 kilos of CO2 per year. Is this a lot? Driving an average car 1000 km/620 miles (a single round trip between Stockholm and Gothenburg) generates approximately 200 kilos of CO2. This by no means absolves us from caring about the energy use of computers, but it hints at the fact that the potential of reducing CO2 emissions in the computer/IT sector is - for now - limited compared to the potential of reducing our energy use by changing our habits of travelling.

This text was originally published in Swedish on October 18, 2009.

No comments:

Post a Comment