?

Log in

No account? Create an account

Previous Entry | Next Entry

Could someone help me confirm or disconfirm an urban legend? I haven't found anything either way on Snopes.com or anything conclusive through Google.

The legend asserts that transferring 1 megabyte of data across the Internet requires the energy generated through combustion of x pounds of coal, on average. One of my graduate students, Raju Mantena, quoted a friend of his as saying that x was 1. I was very skeptical when he told me this in 2001. This article reports x as 0.5 (without any reference).

I'm tempted to run a "back of the envelope" calculation a la Bentley's Programming Pearls, but it isn't immediately clear where to begin.

Do you realize, though, that if even the second legend (1Mb == 2 pounds of coal) is true, a mere half-season of Lost, Smallville, or Enterprise would cost nearly a metric ton of coal to download at HDTV resolution using BitTorrent? The 13-episode run of Firefly (350Mb per 1-hour episode * 13 episodes == 4550Mb) would cost well in excess of a metric ton. That's a FastTrack (/me 50p to the pun fund) to the Special Hell!

The "hundreds if not thousands of dollars" warning that USENET newsreaders give would seem to hit home, especially for binaries newsgroups, eh, crypthanatopsis and darana?




Pre-nano news: I levied my standing army of KDD this afternoon and headed over to Talat Rahman's lab to talk about the newly-funded project we have together. Joining us were my new Ph.D. students Yonghua and Waleed, and the undergrads: Andrew, zengeneral, and istari_ala. More on the project in a moment.

On the way in, I popped in and said hi to Chii-Dong Lin. During the meeting, Brett DePaola came up and said hello, and on the way out, I ran into Bob Burckel. Small world! I guess Cardwell is just cooler.




Post-nano news: remember last year's Ig Nobel awards with the chanting of "nano, nano"? (If you look on my main LJ page, the "Nano" and "Giga" tronkie friends groups are an homage to this.) Well, Jon the Forwarder, Defensor Anguillae of teunc, posted this link to the 2004 awards a couple of weeks ago - I just thought I'd pass it along. Hilarious stuff! When they have finished a small technical procedure1, I will post a link.

:nirg: I think the Atkins Diet opera has got to give Atom and Eve a run for its money. Though a love story about a Bose-Einstein Condensate is way funny, too.

1 Have I mentioned how nice it is to have "small technical procedure" not denote the imminent destruction of the universe?

--
Banazir

Comments

( 14 comments — Leave a comment )
twinbee
Oct. 16th, 2004 11:23 pm (UTC)
it should be noted that not transferring information across the internet takes a lot of energy as well, what with all those routers humming and keepalive packets being shot out over lasers and satellite radio frequencies.
banazir
Oct. 16th, 2004 11:51 pm (UTC)
A fair point...
... that's a good point, but I think that "background data transfer" can be subsumed into the 1Mb/1 lb.

--
Banazir
zengeneral
Oct. 17th, 2004 01:45 am (UTC)
...
According to my calculations, I found that at worst case, each MB takes about 0.02133333333 lbs of coal per MB-month

banazir
Oct. 17th, 2004 09:52 am (UTC)
Care to recapitulate that?
... or are you like Gauss, whom Abel said "is like the fox, who erases his tracks in the sand with his tail"?

--
Banazir
zengeneral
Oct. 17th, 2004 12:35 pm (UTC)
economic analysis
1 short ton = 2000 pounds,
$75 / 1 short ton coal (worst case)
Good bandwidth deal, 50 Gb ~ $40/month
50,000 MB = $40 / month
1 MB ~ $ 0.0008
0.0008$ / MB * short ton / 75$ = 1.0666666666666666666666666666667e-5 short tons per MB

it actually, is much less since there is a long list of hands drawing from the pot where bandwidth is actually sold. The background cost and active cost are fairly close, so the 1MB/1lb would only make sense if the entire network was needed for only one MB. However, the network reaches maximum efficiency when more bandwidth is being used. So, BitTorrent maximizes the value out of the coal needed to power the internet.
banazir
Oct. 17th, 2004 12:46 pm (UTC)
Re: economic analysis
Wlokay, then. :-P
Yeah, I didn't see that one coming. :-)

You might as well measure the cost in calories required to feed zengeneral on a treadmill dynamo! Or open a tap over Orodruin... or smoke a mage of Earthsea out of his or her tower.

--
Banazir
jellybeanzulu
Oct. 17th, 2004 07:59 am (UTC)
one ton? that seems low..
After all, 9,000 lbs is closer to FOUR metric tons. But of course, you don't take into consideration that since it's BitTorrent, you probably give about as much traffic as you consume, so you get to double THAT. (this is with the 2 pounds per Mb number)

Now, buring a pound of coal gives off somewhere between 8500 and 14000 BTU per pound. [1] A BTU is about 1055 Joules. A Watt is a Joule per Second.

Alas, coal power plants aren't very efficient, and only get somewhere in the 30% efficiency range.[2] From there.. well, I'd have to look up what the voltages of 100-T or gigabit ethernet are, to do a calculation of how much energy it takes to change the signal at the transfer frequency enough to transmit the required number of bits.[3] (and add all the bits for TCP and IP headers...) Then, take that number, and multiply it by the number of routers in your transmission stream, and then take that number (joules of energy to physically signal), and divide by the above joules/lb of coal calculation.

Yeah, it's pretty simplistic to assume that all the routers between you and your traffic destination are using IP over ethernet, but it's already messy.. Who knows how ugly it'd be if you assume that several of the links are laser, ATM, or microwave...

Just powering the equipment is a much easier calculation.. Look at the wall-wart on your router to see how much power the transformer supplies, or check the power-supply of your local Cisco box to see its power rating.


[1] http://www.ucsusa.org/clean_energy/renewable_energy/page.cfm?pageID=60
[2] http://www.newton.dep.anl.gov/askasci/eng99/eng99187.htm
[3] umm.. This is where my 9 AM brain falls down.. I can't seem to find physical voltage signaling specs for ethernet.
penguinicity
Oct. 17th, 2004 10:06 am (UTC)
Re: one ton? that seems low..
Don't forget the air conditioning for all those electronics, too.
jellybeanzulu
Oct. 17th, 2004 11:08 am (UTC)
Re: one ton? that seems low..
no thanks, I have to draw the line somewhere. Otherwise, I'd need to add in calculations for the power to run the lights in the co-location facilities where the routers are, and the power for the maintenance guy to microwave his lunch. I'm not willing to do that.

If I went nuts with that kind of math, I can easily see adding the coal requirements for the steam-powered cars that the techs would be driving to work, in a steam-punk universe!
banazir
Oct. 17th, 2004 12:59 pm (UTC)
Re: one ton? that seems low..
no thanks, I have to draw the line somewhere. Otherwise, I'd need to add in calculations for the power to run the lights in the co-location facilities where the routers are, and the power for the maintenance guy to microwave his lunch. I'm not willing to do that.
Don't forget the caloric intake!

Seriously... the back of the envelope is only so big.
Bounding the computation at powering the systems and hubs seels reasonable. "Number of routers", though, isn't so easy to peg. You can figure on 1-30 hops, but what does the distribution of the number of hops look like? Does hop distance matter? I have no idea.

If I went nuts with that kind of math, I can easily see adding the coal requirements for the steam-powered cars that the techs would be driving to work, in a steam-punk universe!
Yay! Steam-punk universe!

(Don't forget the 800-year old Swedish curses that they read out of their e-grimoires. Hey, it's a legitimate use of the Internet!)

--
Banazir
jellybeanzulu
Oct. 17th, 2004 02:00 pm (UTC)
Re: one ton? that seems low..
Ok, I arbitrarily pick 17 hops.

As for steampunk that features Swedish grimoires, I think you're reading stuff that I'm not. (but wish I was)
banazir
Oct. 17th, 2004 05:14 pm (UTC)
YOU'RE the one who FOUND the grimoires!
Ok, I arbitrarily pick 17 hops.
So be it... hopper!

As for steampunk that features Swedish grimoires, I think you're reading stuff that I'm not. (but wish I was)
ROFL

--
Banazir
banazir
Oct. 17th, 2004 12:41 pm (UTC)
Air conditioning
It isn't all that expensive, not with the internal fans, which probably run the bulk of the power consumption.

I always think of water and air-cooled ones, but unless you are running a cryo-cooled Cray, it doesn't really come to that.

We had an NCR Teradata (WorldMark 5100M, exactly the same model as this one, one of six they donated to universities in 2000) sitting in our machine room for four years. It was quite the space heater, yet the annual air conditioning bill was still only a few hundred dollars over what it would have been.

--
Banazir
banazir
Oct. 17th, 2004 01:00 pm (UTC)
Thirty percent
Hey, Doug! Long time, no see.

I figured it was the inefficiency. BTW, one metric ton comes from:

4550 Mb * 0.5 lbs/Mb (2 megs per pound) == 2275 lbs

--
Banazir
( 14 comments — Leave a comment )

Latest Month

December 2008
S M T W T F S
 123456
78910111213
14151617181920
21222324252627
28293031   

KSU Genetic and Evolutionary Computation (GEC) Lab

Teunciness

Breakfast

Science, Technology, Engineering, Math (STEM) Communities

Fresh Pages

Tags

Powered by LiveJournal.com
Designed by Naoto Kishi