Internet vs. Travel

A few months back, I was asked the following question via email:

I am trying to find out how much power will be used, both in my home and on the systems computer hubs, if I were to use Skype for and hour and a half to people who live 20 miles away. Is it better than driving?

There are a few variables (whether the computer would be on or not if Skype wasn’t going to be in use, etc.). But one of the findings of a study I did a little over a year ago was that Skype or similar video chat software uses (in terms of its share of the Internet’s infrastructure energy consumption) about 65 MJ / hour, so here, for 1.5 hours, that’d be 97.5 MJ. You could also add in the power to run the two machines (probably laptops), which would be about 40W * 2 * 5400 seconds = 0.43 MJ, which is negligible by comparison. So let’s say 98 MJ for the Skype call in total. Gasoline has a primary energy of about 132 MJ / gallon, and the drive would probably consume about 1.5 gallons of gas, using a total of 198 MJ.

So it turns out that Skype would indeed save energy if we compare the electricity up front with the gasoline’s primary energy. There is a more complex question here about how to factor in the primary energy used to derive the electricity, and that varies widely by country and source. As a discussed a while back about the idea of transformity, we have to root our energy discussions in a common underlying unit. Odum’s solution for this was to use seJ, or solar-equivalent Joules, but it’s a bit cumbersome to work that out for everything. So we might scale up Skype’s contribution by a factor of 2x or 3x (roughly) to account for this. Even still, Skype would use roughly the same energy as driving, and if you were to consider going further than 20 miles then it’s definitely the case that video chat would use less energy.

So from a somewhat holistic energy perspective, it’s generally better to use electrically driven options rather than oil-based options because the electricity can be sourced in many ways that are more sustainable (and while they’re not by and large today, they eventually can be) and oil and fossil fuels generally are depleting rapidly and impacting the climate.

There is however the question of complexity limits. The Internet is a complex system, as I explored in a study of the Internet’s dependencies. And as Greer and others have observed, complexity limits can make a complex system unviable long before energy limits do. That said, it’s hard to say which of the two systems—the global petrochemical industry or the global Internet—is more complex, and so for the time being my inclination is to move bits rather than move atoms.

Leave a Reply


Responses to “Internet vs. Travel”

  1. …65 MJ / hour…

    18 kW? I’m a tad sceptical?

  2. It is a bit surprising, I agree, but it’s due to calculating the fractional share of the entire Internet’s electricity and embodied power. You can find the derivation of that value in the paper ‘The Energy and Emergy of the Internet’.

  3. Interesting paper, thanks.

    I particularly liked “For example, a gallon of oil that is used to make the components in a router cannot be substituted with an equivalent amount of energy supplied by a wind turbine, and this is a fundamental issue that we must face squarely given the energy challenges we face.” It always worth a reminder that oil is a useful chemical feedstock which is really too valuable to burn.

    The comparison in this blog post is a bit unfair, though, because it doesn’t take into account the embodied energy in the car whose life (and therefore “embodied power”) is much more dependent on the amount it used – how worn out a car is is usually evaluated more on the kilometres/miles it’s done than the years it’s been parked.

  4. Ed — that’s a good point. I should have considered the embodied energy of the transportation system (car, road, etc.). I’ll try and do that for future posts on the topic.

  5. I’m really skeptical of the 68 MJ/hr estimate. Given that the video encoding/decoding is done at each end and all the network does is shunt the data, my feeling is that this must be way off the mark, possibly by two orders of magnitude (I’d be guessing that 180 W would be a closer estimate than 18 kW)

    The paper you mentioned estimated that the internet uses up to 300 GW. That means that your skype call uses about 1/50000 (18 kW / 300 GW) of the internet’s _total_ power consumption. That would mean that 50000 concurrent skype calls would account for the internet’s TOTAL power use — that seems very wrong to me.

  6. Gus — Hmm…I think your calculation there is off by three orders of magnitude. Now the conclusion I came to in the paper regarding the number of Joules per byte was based upon accounting for all the energy and emergy used over the entire Internet, which may be an unfair thing to do. (The issue is, however, how else to account for that energy expenditure and not leave anything out.)