Factoids

Infographic on Data Center Power Consumption

“…Data centers use more energy than the entire country of Russia…”

http://articles.economictimes.indiatimes.com/2012-11-24/news/35332573_1_internet-power-consumption-proxy

Internet World Stats
http://www.internetworldstats.com/

The Web Foundation’s estimates of the relative importance of the Internet, by country
http://www.webfoundation.org/projects/the-web-index/

Internet (infrastructure) doubles in size every 5 years|
http://www.labnol.org/internet/internet-size-to-double-every-5-years/6569/

Number of Internet users doubled between 2007 and 2012 (5 years)
http://royal.pingdom.com/2012/04/19/world-internet-population-has-doubled-in-the-last-5-years/

By 2017, there will be about 3.6 billion internet users – more than 45% of the world’s projected population.

Internet Energy use,  US and World
http://uclue.com/?xq=724

Average energy use of a desktop PC per year
The average PC/monitor combination consumes 588 kWh of electricity every year.

“…With more than 1.5 billion people online around the world, scientists estimate that the energy footprint of the net is growing by more than 10% each year.”
http://www.guardian.co.uk/technology/2009/may/03/internet-carbon-footprint

Overall energy use by the Internet worldwide is only about 2%
http://www.newscientist.com/blogs/onepercent/2011/10/307-gw-the-maximum-energy-the.html

but this doesn’t take into account that many countries with have low Internet access use lots of the word’s energy. If we look at the USA and Europe, where more energy is consumed, the value jumps much higher. This allows us to predict that the 2% figure will rise, and that the overall absolute carbon footprint will also rise as more countries get universal broadband service.

Current growth rates for the Internet, by region:

world internet population doubles from 2007 to 2012, with Africa, the Middle East, and Asia growing the fastest

  • Africa has gone from 34 million to 140 million, a 317% increase.
  • Asia has gone from 418 million to over 1 billion, a 143% increase.
  • Europe has gone from 322 million to 501 million, a 56% increase.
  • The Middle East has gone from 20 to 77 million, a 294% increase.
  • North America has gone from 233 to 273 million, a 17% increase.
  • Latin America (South & Central America) has gone from 110 to 236 million, a 114% increase.
  • Oceania (including Australia) has gone from 19 to 24 million, a 27% increase.

Note that the biggest growth is seen in Africa, the Middle East, and Asia. North America is growing very slowly now.

A hypothetical endpoint for Internet subscriber growth (if Internet Penetration reaches the levels in the US everywhere)

  • The Internet as a whole would have 2.6x as many users as today.
  • Asia would have 3.3x as many Internet users as today.
  • Europe would have 1.3x as many Internet users as today.
  • North America would pretty much be unchanged.
  • Latin America would have 2.2x as many Internet users as today.
  • The Middle East would have 2.4x as many Internet users as today.
  • Oceania/Australia would have 1.3x as many Internet users as today.
  • Africa would have 6.8x as many Internet users as today and would rise to become the second-largest region in the world after Asia.
  • Asia would go from representing 44% of the Internet to 56%.
  • Africa would go from 5.7% to 15%.
  • Europe would go from 22.7% to 11.8%.
  • Latin America would go from 10.3% to 8.6%.
  • North America would go from 13% to 5%.
  • The Middle East would go from 3.3% to 3.1%.
  • Oceania/Australia would go from 1% to 0.5%.

http://royal.pingdom.com/2011/09/30/thought-experiment-an-equal-internet-chart-numbers/

One hour of reading a print paper, approximately equals one hour browsing an online paper
“… newspapers have a lower total environmental impact than printed. However, when considering the time people actually spend reading the printed newspaper or the electronic version, the printed version has the advantage…”
http://www.almamedia.fi/carbon-footprint-per-reading-hour 

Confirmation of the negative effect for newsprint (meaning that it is less sustainable) and for other “broadband” activities.

Negatives numbers mean that we USE more oil to prusue online activities. In other words, wtih the exception of telecommuting, many online replacements for the physical have very small positive effects, or even negative ones.

Exhibit 32: Summary of Monte Carlo Simulations for the US
http://gesi.org/files/Reports/Measuring%20the%20Energy%20Reduction%20Impact%20of%20Selected%20Broadband-Enabled%20Activities%20within%20Households.pdf

Source: ACEEE and Yankee Group, 2012

Savings in Million
Barrels of Oil

Online
News

Music
Streaming

Online
Banking

Telecommuting

Online
Shopping

Online
Education

Digital
Photos

E-Mail

Total

Lower

-0.6

0.2

-2.2

130.8

1.5

-0.6

-3.6

-0.7

163.6

Upper

1.0

3.4

17.7

298.4

15.7

4.6

26.3

7.4

335.8

Average

0.2

1.8

7.8

214.6

8.6

2.0

11.3

3.4

249.7

Common electronic device energy use

  • 17″ LCD Monitor – 75 watts
  • Low-end destkop PC – 70 watts idle, 120 watts in use
  • Medium PC – 150 watts idle, 200 watts in use
  • X-treme Gamer PC -300 watts idle, 900 watts during full gameplay
  • Wi-Fi router – 8 watts
  • Smartphone – 20 watts

http://www.frequencycast.co.uk/howmanywatts.html

Network energy usage related to PC users
for every 100 kWh of demand from desktop PCs, another 25 kWh is needed for networking components): http://cache-www.intel.com/cd/00/00/10/27/102727_ar024103.pdf

Impact of Mobile networks
From the 2009 book, How Bad are Bananas. The calculations may seem high, but that is because the author included the costs of manufacture, transport, and maintaining and operating cellphone towers and networks.

Mobile Activity C02 Equivalents
Manufacture 16kg
1 minute mobile use 57g
3 minutes mobile use mailing a 2nd class letter
1 year at 2 minutes per day 47kg
1 year at 1 hour per day 1250kg

The weight of the Internet
In electrons – about 1 grain of sand (2007)

http://discovermagazine.com/2007/jun/how-much-does-the-internet-weigh

Another weight: 2 ounces, 50 million horsepower, 40 billion watts for the servers  (doesn’t count the client side) (2007)
http://adamant.typepad.com/seitz/2007/04/weighing_the_we.html

“…Each person alive today has six watts of computational power at the disposal of their twenty watt brain…”

http://adamant.typepad.com/seitz/2007/06/the_sincerest_f.html

Zuckerberg’s Law (a Moore’s Law for sharing)

http://bits.blogs.nytimes.com/2008/11/06/zuckerbergs-law-of-information-sharing/

“…Mr. Zuckerberg pinned his optimism on a change in behavior among Internet users: that they are ever more willing to tell others what they are doing, who their friends are, and even what they look like as they crawl home from the fraternity party.

‘I would expect that next year, people will share twice as much information as they share this year, and next year, they will be sharing twice as much as they did the year before,” he said. “That means that people are using Facebook, and the applications and the ecosystem, more and more.’

Call it Zuckerberg’s Law…”

Traffic rise, 2006-2015 estimated for WordPress.com
http://royal.pingdom.com/2012/08/10/traffic-wordpress-com-blog-platform-2006-2015/

Energy Use at a Design Firm for one Developer Seat (use to calculate embodied energy/person-hour)

  1. Client-side (5 LCD screens, Two Desktops, one Laptop)    ~0.4kWH
  2. Server-side (web server plus backup server at hosting company)   ~0.4kWH
  3. Amortization of equipment  ~$1.39/hour

from Ryan McIntyre, Lead Design Monkey at Seraph Group, Inc. (Chicago) – Linkedin

Google’s Electricity use and Carbon Footprint

Google: 15 x e9 kWh/year
America: 3.9 x e12 kWh/year

6.1e9kWH in 2006 – Rich Brown, an energy analyst at the Lawrence Berkeley National Lab in California

Google’s calculations:

“…a Google search uses just about the same amount of energy that your body burns in ten seconds…a typical individual’s Google use for an entire year would produce about the same amount of CO2 as just a single load of washing….”

http://googleblog.blogspot.com/2009/01/powering-google-search.html

Report setting the record straight – Alexander D. Wissner-Gross, Ph.D.  did the original calculations

http://www.technewsworld.com/story/Harvard-Prof-Sets-Record-Straight-on-Internet-Carbon-Study-65794.html

Another quote attributed to Wissner-Gross

 “…viewing a simple web page generates about 20 milligrams of CO2 per second. This rises to about 300mg of CO2 a second when viewing a website with complex images, animations or videos…”

Alexander D. Wissner-Gross, Ph.D. homepage – with links to his company, Enernetics, “…Focused on building advertising and analytics technologies that positively influence online visitor behavior…”

http://www.enernetics.com/

Home Page for Dr. Wissner-Gross:
http://www.alexwg.org/

Technical commentary from a blogger:

“…Google says that a typical search query uses an amount of energy equivalent to the release of 0.2 grams of CO2. If that’s true, and if Google handles a billion queries a day, that’s equivalent to a net release of 200 tons of CO2 per day. But remember, Google’s data centers are not powered by coal-driven generators (they use hydro or nuclear power instead), so in essence a Google search costs nothing, in terms of carbon dioxide…”

http://asserttrue.blogspot.com/2009/03/google-uses-more-electricity-than-most.html

An older post by Steve Souders
http://www.stevesouders.com/blog/2008/03/06/how-green-is-your-web-page/

A fawning report, assuming Google is part of the Conspiracy for Good:

“…Google uses more electricity every year than almost two-thirds of the world’s nations. No wonder it’s been pushing the Obama administration to invest in alternative energy sources and smart grid technology…”

http://www.networkworld.com/community/node/39445

Google Gmail is 100x more efficient that running your own mailserver
http://highscalability.com/blog/2012/8/23/economies-of-scale-in-the-datacenter-gmail-is-100x-cheaper-t.html

How Power used by IT and Hosting breaks down

APC estimates:

  • IT Equipment    44%
  • Cooling  38%
  • Lighting 3%
  • Power System itself   15%

Posted by: John Baschab at December 5, 2006 02:05 PM on http://www.roughtype.com/archives/2006/12/avatars_consume.php

How has data center energy use grown?

Between 2000-2007, data center power use doubled. Use by servers went up 10%, 90% of the expansion was the rise in the number of servers. In 2007, data centers consumed about 1% of the electrical power in the US.

http://www.informationweek.com/news/197006210

By 2020, data centers will produce more greenhouse gases than the airline industry, rising fourfold. The problem is

“In the old mainframe days, data centers were far more efficient but inflexible. In modern data centers, which use standardized technology from the personal computer industry, things are flexible but uncontrolled.”

http://bits.blogs.nytimes.com/2008/05/01/data-centers-are-becoming-big-polluters-study-finds/

How much energy per YouTube video?

Google gives the following estimates:

  • 1 year of Gmail = 1.2kg CO2
  • 3 days of YouTube = 3kg CO2 = 1 manufactured DVD

http://googleblog.blogspot.com/2011/09/gmail-its-cooler-in-cloud.html

This page estimates the number of YouTube videos viewed, and comes up with a big number – Jevons’ Paradox at work (a more efficient system may stimulate higher net consumption)

In 2010, the net carbon output of the 50 million or so daily video views on YouTube resulted was 600,000 DVDs.

http://www.energyrebels.com/blog/2011/09/10/youtube-carbon-footprin/

How much energy per tweet?

Video link: http://mehack.com/from-chirp-energy-tweet-100-j-something-tweet

According to Raffi Krikorian, a developer for Twitter’s Platform Team, each tweet sent consumes about 90 joules. That means each tweet emits about 0.02 grams of C02 into the atmosphere.

How much energy per spam email?

“…The average greenhouse gas emissions associated with an individual spam email are about 0.3 grams of CO2…”

http://gigaom.com/cleantech/carbon-footprint-of-spam-hello-the-internet-uses-energy/

The original study

http://img.en25.com/Web/McAfee/CarbonFootprint_12pg_web_REV_NA.pdf

Here’s a good counterpoint to the study, and questions the whole way that energy use is being measured

 “…The global annual energy used to transmit, process, and filter spam is the equivalent to powering 2.4 million homes, and spam filtering saves 135 terawatt hours–the equivalent of taking 13 million cars off the road. The study decides that the average greenhouse gas emissions associated with an individual spam email are about 0.3 grams of CO2. Fascinating, right?… but we can conclude that the real carbon footprint of a spam message lies in the energy wasted by PCs, notably the fixed amount of time users spend dealing with spam…In my eyes, you could really substitute “Bejeweled” or email in general or any other computing activity for spam and reach a similar computing, if all we’re doing is quantifying uptime. Bottom line: PCs waste energy, and humans leave them on too long, wasting energy. But spam? It’s kind of meaningless here…”

http://goodcleantech.pcmag.com/computing/279586-counterpoint-spam-eco-study-misses-the-big-picture

How much energy per causal web-based game?

The Tragic Cost of Google’s Pac-Man home page in 2010

  • 4.82 million man-hours
  • $120,483,800 paid to surf while at work

http://blog.rescuetime.com/2010/05/24/the-tragic-cost-of-google-pac-man-4-82-million-hours/

How much energy bloat per web page?

“…According to statistics, since 2003, the average web page size has quintupled. That is to say it has increased five-fold. “From 2003 to 2009 the average web page grew from 93.7K to over 507K”. While performance has improved, it seems we have forsaken basic compression logic for increased quality, hence larger files behind our web pages…”

“…a simple web page generates about 20 milligrams of CO2 per second. This rises to about 300mg of CO2 a second when viewing a website with complex images, animations or videos…”

http://www.smashingmagazine.com/2010/09/20/making-the-web-a-better-place-guidelines-for-green-web-design/

Average web page increased in size to 1.7 megabytes in 2013

http://www.sitepoint.com/average-page-weights-increase-32-2013/

Average web page sextuples since 2003

Chart showing average web page six times larger in 2011 than 2003

How much more rich media on web pages?

The average web video has tripled in size since 2003

Chart showing that web page video has tripled from 2003 to 2012

http://www.websiteoptimization.com/bw/1108/

http://wp.me/p1ZvVh-77

Which parts of the web page are bloating the fastest?

JavaScript, probably due to the introduction of HTML5

increase in web download per page, broken down by media type

How much more energy efficient is are LCD monitors compared to CRT monitors?

For 15″ LCD and 17″ CRTs, the average power usage of the CRT was 0.57 W/in2, and that of the LCD was 0.187 W/in2. The LCD used 68% less power that the CRT on average.

http://www.seanmichaelragan.com/html/[2009-03-26]_Power_use_of_white_vs_black_screens_in_LCDs_and_CRTs.shtm

For comparable LCD and CRT monitors, the LCD used 74% less power on average.

http://savingenergy.wordpress.com/2006/11/21/saving-energy-one-monitor-at-a-time/

How much energy does a CRT use on black and white screens?

A 17″ eView e17  CRT monitor consumed 0.48 W/in2 to display black pixels, and 0.65 W/in2 to display white pixels. Thus, as expected, the CRT requires more (+34%) power to display a white background than a black one.

http://www.seanmichaelragan.com/html/[2009-03-26]_Power_use_of_white_vs_black_screens_in_LCDs_and_CRTs.shtml

A CRT consumed 43% more energy displaying a white, versus a black screen

http://savingenergy.wordpress.com/2006/11/21/saving-energy-one-monitor-at-a-time/

How much energy does an LCD use on black and white screens?

A 15″ Planar systems model 996-0547-00 LCD monitor consumed 0.190 W/in2 to display black pixels and 0.183 W/in2 to display white pixels. Here, again as expected, the color preference is reversed: the LCD requires less (-5%) power to display a white background than a black one.

http://www.seanmichaelragan.com/html/[2009-03-26]_Power_use_of_white_vs_black_screens_in_LCDs_and_CRTs.shtml

How do brightness settings affect energy use on CRT and LCD screens?

Both use more energy at higher brightness settings.

http://savingenergy.wordpress.com/2006/11/21/saving-energy-one-monitor-at-a-time/

How do contrast settings affect energy use on CRT and LCD screens?

Not at all, for either kind of monitor. But high-contrast pages are more easily seen on dimmed monitors.

http://savingenergy.wordpress.com/2006/11/21/saving-energy-one-monitor-at-a-time/

How much energy does Adobe Flash use?

Flash is an energy hog
http://arstechnica.com/apple/reviews/2010/11/the-future-of-notebooks-ars-reviews-the-11-macbook-air.ars/3

Flash increases energy usage of a highly efficient laptop by 33%
http://jstsch.com/post/adobe_flash_is_bad_for_the_environment

Minimizing CPU cycles stolen by Adobe Flash
http://help.adobe.com/en_US/as3/mobile/WS4bebcd66a74275c36cfb8137124318eebc6-8000.html

Flash use cut by 50% in 2 years, only 7% of average page size
http://royal.pingdom.com/2012/07/06/flash-cut-in-half-in-2-years-now-only-7-of-average-web-page-size/

Growth in web servers, 2003-2012
http://news.netcraft.com/archives/category/web-server-survey/

10 Comments

  1. These factoids (and the other resources on this site) have been very useful! I’m writing a blog post on “Carbon impact of good/bad UX” blog post, and it will be well-peppered with resources taken from your collection. One thing I didn’t find (yet) is a good energy measure for the effectiveness of CDNs and other optimisation techniques. I’d love to show how applying some WPO to a large website (code optimised, responsive techniques, CDNs etc) x 800k / page x 1 million visits/yr x 2 minutes a visit (3-4 pages) = ??? MWh saved. I’m going with 1MWh = ~1 Ton CO2 (assuming coal).

  2. Yea, I’ve been sending requests out to WPO companies, searching for case studies. So far, no takers. Ditto for traditional LCA companies who compute the green-ness of buildings, no interest and even one hostile response!

    The fonts question is particularly well-taken, since using webfonts via a CDN (plus making part of your interface via a font “vector sprite” should have a big impact.

    I may try working on a theoretical analysis like this during the next couple of weeks – if you do one, let me know.

    Finally, I’m looking for someone to correctly normalize the UC Berkeley CS dept study on Internet energy. They reported that the Internet uses only 2% of world power, but neglected to adjust for the fact that many of those countries have very low Internet penetration. I’m thinking that the Web Index by the Internet foundation – http://www.webfoundation.org/projects/the-web-index/ could be used for normalization.

    Finally, keep in touch, and let me know when your blog post is out! -pete

  3. Hi, still haven’t finished my post. I’m still assembling facts and figures.
    Most recently I was looking for a crude measure of network electricity usage, as kwh/GB downloaded. This whitepaper on GHG footprint of internet adverts (http://evanmills.lbl.gov/commentary/docs/carbonemissions.pdf) claims 9-16 kWH / GB.
    If true, this gives a very handy way for amateurs like me to think about WPO – 250k saved per page on an average ecommerce site = 195 tons CO2e/year saved. (5 pages x 250kb per page x 12 months x 1million/visitors/month x 1ton CO2e per MWh (dirty coal)).

    But those are 2008 figures, and apparently the network is getting more efficient all the time.

  4. This is great stuff, even if it is conservative, with things getting more efficient.

    As for the “getting more efficient”, I’m guessing one could look at hardware to get long-term estimate. Current technology is reaching physical limits, so it seems likely we might get another 10-fold or more efficiency increase, but probably not 1,000 or 1,000,000 without something exotic like quantum computing. Combine this with the “law” that the Internet is doubling in size every 5 years – http://www.labnol.org/internet/internet-size-to-double-every-5-years/6569/ and we can see that things will add up.

    1. The impact of energy-intensive technology (often mislabeled as “green” releative to paper) remains to be seen. Web carbon footprints as a plank in theplatform?

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s