I would like to know what the average carbon footprint of an email is.

The answer should take into consideration the composition,
transmittal, and reading of the email. The carbon footprint of the
storage of the email is NOT a consideration.

Some hurdles to gaining an accurate answer may be:

-the fact that a computer will have differing power consumptions based
on the hardware configurations the user has.

-impossible to determine whats source of energy the comuter draws
upon: By which I mean that it is (I assume) not possible to determine
whether a given computer is drawing its electricity from a source
powered by solar, wind, oil, gas, coal etc

-email sizes will vary especially considering attachments

-email transmissions sizes will vary because of differing bandwidths,
email sizes and attachments

-email would most likely be only one of many applications running on a
given computer, as it is impossible to know for sure for my purposes
you can assume email is the only application running and consumes all
of computer resources while being used. (Far from accurate I know)

What I am looking for here is as factually based an estimate of the
average as possible complete with references

---

Clarification of Question by ashdezign-ga on 27 Jul 2006 22:04 PDT

I have noticed that noone has yet started answering this question.
Perhaps I was not clear enough regarding what exactly I am looking
for.

Let me start by explaining the premise I am working off of.

As you compose an email you are using a computer which is most likley
drawing power from a power station burning oil or coal. This produces
carbon dioxide. As the email is transmitted it travels through various
servers all drawing power in a similar fashion. When the recipient
downloads and reads the email, again ir falls under the same
parameters.

What I would like to know is on average how much carbon dioxide is
produced for the entirity of an average emails journey.

I hope this helps, please if anything is unclear, ask me for clarification.

Thanks!

---

Clarification of Question by ashdezign-ga on 29 Jul 2006 10:36 PDT

First let me thank everyone who has commented. Sometimes the hardest
part in getting an answer is in finding the right question and you all
have been tremendously helpful.

As pointed out the transmission time would seem to most likely be
instantaneous, or so close to instantaneous as to make no practical
difference. As such the carbon footprint of transmission would likely
be immeasurably small.

That leaves us with composition, storage and reading the email.

Storage can also be fairly easy to work out by simply measuring the
power consumption of the server(s) used to store the emails so we can
eliminate that from the question as well.

So it appears the real question is:

How much is the carbon footprint of the average email: specifically
how much electricity is consumed by the computers being used during
the times used to compose and read the email.

Emails will vary in size, but we are looking for averages and
attachments can be left out as they are composed separate from the
email. So we are considering the duration of actually writing the
email by the sender before transmittal and reading the email once it
has arrived with the recipient.

I guess what we need to know is how long the average content portion
of an email is and, on average, how long does it take to compose and
how long does it take to read.

We can, for simplicities sake, assume that the web mail service is the
only thing running on the computers in question and therefore we
simply need to translate the length of time spent in composing and the
length of time spent in reading into an amount of electricity consumed
by the computers in question.

The answer can be expressed as such: X number of Kilobytes of data on
average takes X amount of time to be written and X amount of
electricity is consumed by the computer being used for the
composition. That same amount of data on average will take X amount of
time to be read and X amount of electricity will be consumed by the
computer being used to read the message.

If we can arrive at an average amount of electricity used per kilobyte
of written data and read data it should be fairly easy to work out the
carbon footprint.

I will add a 100 dollar tip if you can also work out how much
electricity is used on average for X Kilobytes of data to be
transmitted.

Please provide as much supporting references for your conclusions as possible.

---

Clarification of Question by ashdezign-ga on 29 Jul 2006 10:39 PDT

To further clarify the following section in the previous calification:

The answer can be expressed as such: X number of Kilobytes of data on
average takes X amount of time to be written and X amount of
electricity is consumed by the computer being used for the
composition. That same amount of data on average will take X amount of
time to be read and X amount of electricity will be consumed by the
computer being used to read the message.

Could instead read:

The answer can be expressed as such: X number of Kilobytes of data on
average takes X amount of time to be written and X amount of
electricity in total is consumed by the computer being used for the
composition during that period of time. That same amount of data on
average will take X amount of
time to be read and X amount of electricity in total will be consumed by the
computer being used to read the message during that period of time.

---

Request for Question Clarification by hedgie-ga on 01 Aug 2006 20:56 PDT

Re:
"As pointed out the transmission time would seem to most likely be
instantaneous, or so close to instantaneous as to make no practical
difference. As such the carbon footprint of transmission would likely
be immeasurably small"

Email is using 'store and forward' algorithm, and automatically retransmits
the message several times before it gives up (in case of a problem).
It is not instantaneous and cpu cycles spent in transmission,
 depend on path (there are more'stops' when going between continents
then between users of same ISP)...

 The bulk of the power consumption is used  'to keep the servers switched on';
 Increase caused by simple processing (forwarding the mail) is small, but there
 are many servers dedicated to processing of mail.

 Expression 'email' is not well defined: Does it include spam? (there is almost
 no compose time per message) and (minimal time spent reading it) but there is
 lot of computation done when filtering mail.
 "Now, two-thirds of all e-mail is spam". Do you include address harvesting?
 
 Include : Intranet communication? (within company?)

  So, the question would have to be reformulated (or rather formualated)
 It would be more an essay, then calculation.
 I am willing to give it time and consideration, if you  understand
that you will not get accurate numbers, just very rough estimates.

 I would also need to know your purpose or goal and use of the numbers.

 You a power consumption (per kB of a message, or per period of time).
 It could be per-user, infrastructure differs in different areas of the world.
 The statistics on number of users (penetration) are available.

 What power sources are being used to generate that power, you would have to
 infer from power industry stats.

 Would that be acceptable?

Hedgie

---

Request for Question Clarification by hedgie-ga on 01 Aug 2006 20:59 PDT

"You a power consumption .."

should be

 " You would get a power consumption .."

Sorry about a typo.

---

Clarification of Question by ashdezign-ga on 04 Aug 2006 23:14 PDT

Sorry for the delay in response Hedgie, I had almost given up on
getting an answer and had not checked for a couple of days.

First the purpose of the question would be to calculate the estimated
carbon footprint of a commercial webmail service with a view to using
carbon offsets to make it carbon neutral. I am building one so I would
once the service is live have access to a variety of stats. For now I
need as much info and estimates as possible for planning.

You stated that the term email was not well defined and asked if it
included spam. Yes it would include any data sent/recieved via email,
spam and attachments as well as general emails. I would not include
address harvesting. Or Intranet considerations.

re:
"Email is using 'store and forward' algorithm, and automatically retransmits
the message several times before it gives up (in case of a problem).
It is not instantaneous and cpu cycles spent in transmission,
 depend on path (there are more'stops' when going between continents
then between users of same ISP)..."

If you could give an overview as well as an average approximation that
would be great.

Re:
" So, the question would have to be reformulated (or rather formualated)
 It would be more an essay, then calculation.
 I am willing to give it time and consideration, if you  understand
that you will not get accurate numbers, just very rough estimates.

 I would also need to know your purpose or goal and use of the numbers.

 You a power consumption (per kB of a message, or per period of time).
 It could be per-user, infrastructure differs in different areas of the world.
 The statistics on number of users (penetration) are available.

 What power sources are being used to generate that power, you would have to
 infer from power industry stats.

 Would that be acceptable?"

Absolutly acceptable, thanks!

Hello again ashdezign-ga

   The search came out a bit differently than I anticipated when I started.
   I got energy consumption figures, but once I had them, a more meaningful
   answer demanded I go further down the path.

   It is impossible to answer the broad question - of the 
   
   SEARCH TERM: energy consumption in telecommunication
   
   but your explanation of the purpose:
   "..commercial webmail service with a view to using
   carbon offsets to make it carbon neutral..." 
   
   was useful and allows narrowing the topic a bit. I will focus
   on the following scenario: imagine a company which is still doing
   business the old-fashioned way (snail mail, POTS , fax ..) (if such still
   exist) and imagine that this company would switch to the modern, Internet way
   of handling the necessary communication and use a commercial webmail service.
   
   How would such a decision affect energy consumption (as a measure of
   the impact on the C02 emission)?. The topic is still quite broad:
    a company in the business of global import-export has different needs than
    the local grocery store or a shipping firm  or a stock brokerage ..
     but we can guess (in advance of a search for quantitative facts)
that impact of the decision to switch will be ecologically positive in
all such cases.
   
   In general, the energy requirements of electronic communication are
much smaller than the conventional means ( such as paper based
(hardcopy) letters
  sent bu US post office - a method called snailmail ).
    
   In this narrow interpretation of the question, we are excluding the
impact of phenomena such as spam: the fact that someone in Asia is
flooding the world with billions of unwanted ads for 'fake Rollexes
and quack remedies' has really little to do with a decision of our
'typical' company to use email for their business communication.
   
   Let's start by  finding the Watts/cycle energy consumption rates.  
   
   That is the power used to make computations. Those numbers are known:  
   
   PART 1 OF THE ANSWER TO YOUR IMMEDIATE QUESTION IS HERE:
   ===========================================================
   
   http://news.taborcommunications.com/msgget.jsp?mid=710609&xsl=story.xsl
   
   Note how low the numbers are -- (1 nano-watt per Megaflop)
            -----                   --------------------------
   
   http://www.zettaflops.org/PES/P1-Sterling.pdf
 
   PART 2 OF THE ANSWER TO YOUR IMMEDIATE QUESTION IS HERE:
   =========================================================
   
   This subquestion is : how many Megaflops does it take to process one "hop" of 
   1 Mb (Megabyte) of email?
  
   SEARCH TERMS: CPU consumption per email message
   
   Number depends heavily on anti-spam filtering methods
   and is summarized in Fig 5. of this report
   
   [PDF] Resisting Spam Delivery by TCP Damping
File Format: PDF/Adobe Acrobat - View as HTML
Unsolicited email messages, also known as spam, have become a serious
problem ... Figure 5: CPU Usage per Email. 4.2. Damping TCP by
Resource Consumption ...
http://www.cs.uga.edu/~kangli/src/ceas2004_kangli.pdf

What it means is: It takes more CPU cycles to filter the mail (to
eliminate spam), than to store and forward it.
As a rough measure of each step of the necessary processing ( compose,
store and forward (per hop) and deliver (via POP server) to users mail
box, one can take 100 cycles per bit of volume and each step).

What is a "hop":
 There is a network tool called 'traceroute' which show how many hops
(intermediate steps) are needed to deliver email from your location to
a given e-mail address. The number of hops ranges from 1 to 20. For
long distance measages, the number  depends on 'distance' to the
Internet backbone.
  
 Here is a web-based tool using traceroute. It lists the hops from
your machine to different parts of the globe. It also gives time (in
milliseconds) of transmitting the message duting the hop. At end of
each hop there is
 a computer which will store the message in memory, and resends it
later (once, or up to 5 times, if required). It can also block the
message, if it decides it is  a spam. It nice to send a 'message back
to the sender' informing him that
his snder adress is blocked. Often this is not done - since most spam
fakes (spoofs) the senders address - so 'nice notifications' become
part of the spam flow. (Al these problems are temporary - see the
'future of the email' at the end of this answer).
 
 http://www.tracert.com/cgi-bin/trace.pl
 
  PART 3: ANSWER TO THE WIDER (real life) QUESTIONS
 ==================================================
    Part 1 and 2 of the answer, combined,  will tell how much energy
is used to process a given volume of mail, sent to different parts of
the world.
   
   However, that minuscule number is  actually irrelevant  to the task you posed.
    The main environmental impact of telecommunication is not the
'waste heat' due to extra cycles needed for data processing, but the
cost of building the infrastructure  (that is, laying down cables or
optical fiber or building WI-FI towers, data centers ..).
   
   See e.g.
   http://www.wirelessinsightasia.com/article.asp?id=1528
   
   For this reason, the wholesale cost of the ISP service is a better
measure of impact of use of a website than energy consumption per
Megaflop or Megabyte.
   
  Let's look at those numbers: How much would a company pay, per
month, to operate a commercial website which would process, let's say,
a  million messages per day (via web mail)?
   
   Here is a sample of the OEM costs:
    http://www.tdl.com/~netex/isp/isp.html
    
    Of course, a lot depends how well the wholesale providers are
selected and a site and service designed.
    
    As an example, Google built its impressive worldwide network,
using Linux OS and redundant cheap hardware (REID disks):

"However, this does not seem to be hindering Google at all. In fact
the low cost hardware combined with the low cost OS is what allowed
them to grow so quick and remain a private company. They spent the
start up cash on brainpower and not hardware and software. ."
http://www.webmasterworld.com/forum3/26931.htm
http://www.manageability.org/blog/stuff/economics-google-hardware-infrastructure/view


  CONCLUSION RE EMAIL COSTS IT's IMPACTS - Today and Tomorrow
  =============================================================
 (We use Google email service as an example, other services are listed below).

 Considering that Google is offering free email service, called gmail,
using it's worldwide network (built initially to precess search
queries)
 https://www.google.com/accounts/ServiceLogin?service=mail
 - a service which is both web-based and alos has free forwarding (so
that it can be used like regular email) is telling us two things:
 
1) The cost of the telecommunication itself (energy use and
infrastructure) is so small that the company is compensated just by
placing a modest number of ads, next to the messages.
 
2) It will be impossible (for at least a few years) to compete with
the present offerings of the web-mail services (Google, AOL, Yahoo ..)
   SEARCH TERMS: (free)  webmail services
   http://www.pcmech.com/show/reviews/884/
   
   This means that our hypothetical company offering web mail service
commercially would have to operate on a value-added basis. It would 
use infrastructure which already exists, and to offer services which
are today missing.
   That would be, at this time, primarily customer education and spam
protection. It may be possible to build a niche business: let's say
mail services for farmers (in the US or Zambia ..)  or hardware
stores, or ..  which would use knowledge of and understanding special
needs of that branch of industry, combine spam protection, marketing,
computerized invoicing, tax returns  ...Company could provide software
to interface or replace current accounting software and databases ...
- and otherservices not offered by today's basic mail companies.
   
   So, the conclusion is:
==============================
 
   The actual environmental costs of email transmission itself are so
minuscule that they can be neglected when compared to other costs.
Therefore, the carbon dioxide savings would be simply the  carbon
saved by eliminating the energy used by old fashioned energy-guzzling
methods
  
   The task  of commercially viable service would be related to 
educating and converting companies which do not presently use email,
 to provide customer support (which todays in minimal or non-existent).
  
   
   We  need to keep in mind that scene is still changing very quickly and there
are generational differences and differences across the globe.

 Here, then is the last of the relevant
 
   SEARCH TERMS: future of email

    http://www.theregister.co.uk/2005/09/21/email_future/
    
    As appropriate for the 'information age', the decisive factors
turn out to be not in the consumption of energy and material, but in
something more elusive, something called information,
ease-of-use,custommer support, security and
dialectical paradox of privacy combined with access to information.

    
      http://reviews.cnet.com/4520-6028_7-6301361-1.html
      http://www.infoworld.com/infoworld/article/04/04/16/16FEfuturemail_1.html
      etc
      
      Once you've had a good look at these links
      please do rate the answer or, if needed, post a RFC (request for
clarification)
      
      Hedgie

---

Request for Answer Clarification by ashdezign-ga on 07 Aug 2006 04:22 PDT

All very interesting, I do have 1 question thouh. You give the figure
of 1 nano-watt per Megaflop with the subquestion of : how many
Megaflops does it take to process one "hop" of 1 Mb (Megabyte) of
email? And then go one to reference the traceroute network tool.

This looks good, very much like what I am looking for; but my question
about your answer is this:

How accurate is the "1 nano-watt per Megaflop" figure? In the
reference pdf you quoted it seemed to refer to a hypothetical system
as noted by the heading "What we will need" in that section. Is it
safe therefore to assume that conventional computers today use more
wattage per Megaflop?

---

Clarification of Answer by hedgie-ga on 08 Aug 2006 21:53 PDT

Power consumption of processors vary in a wide range, depending on
hardware and activity: a supercomputer, running a (non-stop, as they
do) number crunching will be more effficient (in terms of
energy/cycle) then a typical home computer (Intell Pentium or AMD
Athlon) used to retrieve and display web pages or mail.

Also, new computers have much improved power management (they go to
sleep, when not used) than previous generations.

 Here are some hard numbers for today's  PCs:

 http://www.techreport.com/onearticle.x/7417

Example and explanation
It shows 230W for pentium with clock rate 4.3 Ghz, which translates to

230/ 4.3 E9  W/cycle = 50 nanoWatts per cycle. 

SEARCH TERMS: Megaflop, flop

Unit 'Megaflop' was invented to measure number-cruncing performance 

Unit 'Mflop'
   1. (computing) a measure of the speed of a computer; one million
floating point operations per second
floating point operation (abbreviated as FLOP)

Unit 'flop' 
   1. (computing) any simple operation, such as addition,
multiplication or division, that a particular computer can perform
using a single operation

 
Question '  how many cycles per instructions ' can be made very complicated
as it depends on architecture of the hardware.:

http://guides.macrumors.com/index.php?title=Computing_Units&printable=yes
http://www.dgate.org/~brg/files/vsimII/website/softarch252.html
http://www.rwc.uc.edu/koehler/comath/43.html

For our purpuse we can aproximately take one flop as hundred cycles.

More data are available under 

SEARCH TERMS: CPU power consumption , power diddipation

 ('Search Terms' are words to type into a (Google) search engine. in different
  combinations,  to get additional article in the topic). E.g this:
  
Google query:
://www.google.com/search?hl=en&lr=&q=CPU+power+consumption&btnG=Search
will bring articles like:

http://www.goodwin.ee/sulo/Power2.htm
or this one:

"The power and heat issue is relevant to everyone else in a variety of ways:

    * Environmental: High power efficiency means lower energy
consumption becomes a significant environmental issue. Computers do
represent a significant percentage of electricity consumption not only
directly but indirectly with increased air-conditioning cost in
enterprise applications due to their added heat..."

"..As we noted in the Turion 64 article, hardware manufacturers have
recognized the importance of reducing power. High efficiency power
supplies are increasingly more common, and Intel and AMD both identify
performance-per-watt as a key benchmark..".
http://www.silentpcreview.com/article313-page1.html


Finally,this article
http://www.adobe.com/devnet/coldfusion/articles/performance_61.html

deals specificaly with 

E-Mail Handling Performance

"..Another important area of enhanced performance in ColdFusion MX 6.1
is in handling and delivering e-mail. Customers have long requested a
more robust mail handling subsystem, capable of sending very high
volumes of mail quickly, and ColdFusion MX 6.1 delivers. Macromedia
lab tests clocked ColdFusion MX 6.1 sending well over one million
e-mail messages per hour. This new found e-mail handling capability,
when coupled with the extensive e-mail handling enhancements, allows
for truly industrial strength e-mail applications. Whether for use
with high volume lists or just rock solid stability for any
application that needs to send e-mail reliably, ColdFusion MX 6.1
delivers..."
http://www.adobe.com/devnet/coldfusion/articles/performance_61.html

Please note that this  one is a 'merketing talk', not an objective coparison. 

That (mass mail) is also a  'science' in itself:
http://www.iventa.com/Email-Publishing-Tracking-System.aspx
http://images.apple.com/server/pdfs/Mail_Services_TB_v10.4.pdf.

Can one say 'junk science'? :-) 

 Analysis of those numbers and specialised mass-mail software, both
spam and anti-spam would require a separate question to be covered
well.


Hedgie

Just a free comment.
Your clarification was certainly a help.  Your question is very
seriously priced, so I hope that you get an answer, but I rather
wonder if one is possible.

It seems to me that the fractions of seconds involved in each stage of
the composition, transmittal and downloading of a single email would
be extremely difficult to segregate from the hardware's ongoing use of
electricity  - and would not be a significant amount in terms of
"carbon footprint".

But I may be wrong.  I hope you find out.

I would also think that an email would have a lot less of a carbon
footprint than millions of letters (or possibly phone calls) that it
has replaced. Also, if you're not going to write an email, you'll be
doing other things on the PC, so it's really not entirely possible to
know.

Thanks for the comments, You may be right I was thinking of simply
multiplying the servers carbon footprint by 3. In otherwords for a
hypothetical webmail service "X" User A composes an email and
transmits it to User B. Assuming User B is on the same webmail service
than a simplistic description of the route would be User A to server
for service X to User B on service X. If User B is on a different
webmail service than the multiplication would be by 4. User A on
service X composes an email and transmits to User B on service Y. The
route would then be User A to Server on Service X to server on Service
Y to User B.

This approach would be very easy for me to calculate, but how accurate
an assumption would it be? Would simply multiplying the the server by
3 or 4 be roughly equal to the exact answer or would it fall to far
short?

If we assume the purpose of the excercise is to account for the carbon
in order to offset it thus creating end to end carbon neutrality for
the webmail service how would one go about calculating that in an
honest fashion?

That sounds logical, but I have minimal knowledge of EDB/web workings.
Although in practice it obviously takes longer to send a longer email
or one with attachments, isn't this time watching the blips pass or a
percentage rod move greatly exagerated by the multi-tasking in the
network?
Someone who understands transmittal speeds could easily tell us the
actual time used to transmit a single page, say, 300 words.  I expect
that on a dedicated, private line, it would almost be instantaneous. 
To be generous, let it use twice the time on a busy multi-tasking line
for routing and handling the segmented message.  Seems like it would
still be almost instantaneous.