Re: [E-impact] Carbon-Intelligent Global Routing in Path-Aware Networks

[Brian E Carpenter <[e-mail]>]

On 11-May-23 06:56, Michael Welzl wrote:
> Hi,
> 
> On “induced traffic”:
> 
> 
>> On May 10, 2023, at 2:50 PM, Michael Richardson <[e-mail] <mailto:[e-mail]>> wrote:
>>
>>
>> Brian E Carpenter <[e-mail] <mailto:[e-mail]>> wrote:
>>> 2. How much reduction in service will society tolerate? (Example:
>>> yesterday while waiting for a bus during a heavy rainstorm, I noticed
>>> that the electronic sign at the bus stop had no data and displayed "Low
>>> battery", because it was solar powered. Will society accept such
>>> trade-offs, especially less trivial ones?)
>>
>> I think that this is really a very important point.
>>
>> If the effect of the next-stop information being absent from the bus shelter,
>> because of a rainstorm, is that Brian takes his car, then it's an overall loss!

Just to illustrate the complexity of such arguments, what I actually did was
take out my cell phone to check the next bus arrival. So add that into the
energy balance.

>>
>> In another part of the thread, an argument was made that the Internet
>> infrastructure is on anyway, whether it's transmitting cat videos or not.
>> That's an argument called: fallacy of composition.
>>
>> In transportation economics, it's called "induced traffic"
>>
>> If marginal costs of new uses are low (or zero, or even negative) then new
>> uses come up, and then the infrastructure becomes congested, and more
>> infrastructure is added.  When that happens to highways, it's a problem for
>> everyone other than the road construction companies. (There is always a lobby)
>> It's not clear to me that the same is true for Internet infrastructure.
> 
> No, it isn’t. The Internet is not generally congested:
> https://ieeexplore.ieee.org/document/9817041 <https://ieeexplore.ieee.org/document/9817041>
> Preprint: https://arxiv.org/abs/2206.06642 <https://arxiv.org/abs/2206.06642>

We've also tried our own version of multiple traffic lanes, i.e. differentiated
services (RFC8100 etc.), even including mopping up capacity in under-used lanes
(RFC8622). Deconstructing that to determine its impact on energy use would be hard.

>> But... there is also always a lobby that benefits.
> 
> That may be true, but let’s not use the “induced traffic” argument because that just doesn’t seem to map to the Internet’s reality.

What happens when significant numbers of users are inconvenienced by lack of
peak-hour capacity (e.g. for streaming)? ISPs add permanent capacity. So it
isn't so much induced traffic as induced capacity, and then I think the
interesting question is how much energy is wasted by under-used capacity.

    Brian