The WING Blog

The Web and Internetworking Group at BU/CS

Nov

12

Serious numbers

By Mark Crovella

When times are mysterious / Serious numbers / Will always be heard

- Paul Simon

I’ve been reading “Sustainable Energy – Without the Hot Air” It’s by David MacKay — yes, the same David MacKay that wrote the terrific book “Information Theory, Inference, and Learning Algorithms”.   I got started on the topic after reading and thinking about “Networking in the Long Emergency” (which I previously posted about).    Once you start poking around in the related literature, all signs point to MacKay’s book.

The first half of the book consists of a fascinating assessment of energy consumption and production, broken down into categories like cars, planes, heating, light, etc.   The power of the book is that it makes clear where the big problems are, and which issues are just distractions.

On the right is a graphic from the book that shows the breakdown of average energy consumption per person in Britain.ConsumptionStackMacKay

What jumps out from this chart is “Jet flights.”   Surely this “average person” is jet-setting around the world, no?   Actually, no.  This bar corresponds to exactly one intercontinental trip per year. That’s it.

Well, this must be because air travel is so inefficient, no?   Again, no.   The book does a wonderful job of explaining the physics of transport and where the energy actually goes.   To boil it down: the problem is not that it’s air travel.   The problem is simply how far you travel when you go to another continent.

Of course, this makes sense.    My commute is about 40 miles per working day.  At 250 working days per year, that’s 10,000 miles in a year.   For comparison, next month I’ll attend CoNEXT in Tokyo.   The round trip is 14.290 miles, for a five day trip.

So what does all this have to do with networking?    Nowadays, there is a growing research effort on “green networking” and “greening the Internet.” My take is that to really make a difference in favor of responsible energy use, we should be beefing up the Internet.    I take anywhere between 4 and 8 trips per year, often traveling overseas 3 or 4 times in a year.   That kind of travel absolutely dwarfs all my other energy consumption combined.    What if I could eliminate one or two of those trips through the use of telepresence?   That would actually make a significant impact on my carbon footprint.

What do we need to make the routine use of telepresence commonplace?  The social / human factors answer is that we need telepresence to provide an acceptable substitute for the personal experience of being in a room with a group of people.    And we need to carefully identify the meetings where being together with a group of people socially is an essential component, and distinguish those from meetings where the social component is an afterthought.    For example, I think that some conferences fall in the first category, and some meetings fall in the second.

The engineering answer is that we need a quantitative improvement in the Internet, and a qualitative improvement in HCI.    At present, the Internet is not quite up to multiway-telepresence as a reliable service on a global basis.   But the current rate of network buildout, if it continues, will take care of that.    The real issue is the HCI component: how we think about our displays.   For telepresence to be routine, we need to stop thinking about displays as tools that sit on our desk and allow us to manipulate computers.   We need to think about displays as part of building infrastructure — something that is designed into every office space and has standard properties that can be relied upon to make the telepresence experience consistent.    We know what to expect in terms of social and nonverbal cues when we are in a room with someone.   We need to figure out how to make that consistently available over the Internet.

So, the bottom line for me is that while “greening the Internet” is important, ”Internetting people” will make a bigger difference in the long run…

4 Responses so far

I am a convert when it comes to telepresence from the day I saw a demo at Cisco about 6 months ago. The interfaces were amazing (but not yet seamless).

Some random thoughts/questions:

Do you really mean HCI? It seems to me that what needs to develop better are Human-to-Human Interfaces (HHI) — it is funny to think about it this way…

For an *average* person, how many miles would be eliminated if we take out all conference and business travel that could be done through telepresence? Another way to ask that question is: of the 30 kWh/d in jet flights, how much is due to travel that could be replaced with telepresence? According to dated statistics from the .com era (https://ntl.custhelp.com/app/answers/detail/a_id/252/~/percentage-of-air-travel-for-business-vs-other-purposes), about 40% of US air travel is for business. If telepresence cuts that in half, then we are talking about a 20% reduction (about 6 kWh/d — more than the energy needed for light).

The impact of the Internet on greening is far more significant if one looks at other current/potential uses. Here I am thinking about the Internet facilitating global-scale commerce and optimizing logistics (think about what Amazon is doing to “shopping”, what VoD is doing to DVD rentals, …)

I like Internetting people, except that I am not a social scientist to pretend that I can make a difference there ;)

I think that transformative Human-to-Human interfaces have to be led and created by technologists; indeed, this is really what a large subset of HCI people actually work on. While social scientists can contribute by vocalizing and evaluating what might work and what might be needed, their role is secondary, and in some ways subservient to, technologists.

Taking about green networking, we just had one session in CCW’11 on the topic and you can find the talk slides at:

http://committees.comsoc.org/tccc/ccw/2011/program.html

According to Azer’s calculations, 3% is due to business travel, which can be reduced by telecommuting/telepresence. Studies show that ICT also contributes about 3%. So it looks like it’s worth “greening the network” especially that telepresence will impose more demand on the network as well.

Mark et al.

Glad to see your thoughts and interest on this extremely important topic. As you know, I was fortunate to be able to spend a good deal of time last year exchanging ideas on sustainability with our colleagues at U. Cambridge. Led by Jon Crowcroft, those folks have a broad program on ICT sustainability, which includes topics on energy efficient networks, data centers and most importantly (IMO) on green buildings (this effort is being led by Ian Leslie). The point is that as soon as you being to think about matters of sustainability, it immediately becomes clear that while keeping “our own house in order” is important (ie. making sure that ICT is as low energy as possible), that effort is likely better spent on using computational methods to improve sustainability in other areas.

One of the conclusions that Jon and I came to early on is that simply creating technology that can lead to lower energy use (ie. telepresence) is not enough. There are a huge number of social and cultural issues that will need to be addressed before the promise of technology can be realized. I have no answer to this other than to say, I look forward to being a participant the first virtual SIGCOMM.

Finally, let me leave you with a challenge that I think it very worthy of consideration. There have been many efforts by the NSF and other groups over the past years to develop “new Internet architectures”. With all due respect, I submit that these efforts have not resulted in anything particularly noteworthy or promising in term of solving a big existing problem. However, consider the design implications for a network that has an *energy budget*. When this simple constrain is put in place, I think it suggests a number of VERY interesting possibilities.

PB

Leave a comment