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

Related to this, I have a query.

In network engineering, is there any developed language for talking about the differences between attributional accounting for the use of shared resource (i.e. how we distribute responsibility), and marginal accounting (how we talk about the consequences of a particular intervention)?

In sustainability industry with the GHG protocol, the language is consequential vs attributional 

https://ghginstitute.org/2021/04/21/the-most-important-ghg-accounting-concept-you-may-not-have-heard-of-the-attributional-consequential-distinction/

You can see this most recently in a bunfight with the creation of the software carbon intensity, between people arguing for one form of measurement vs another - see the link below in this github pull request as a proponent of marginal accounting (Henry Wattime) argues with a proponent of attributional accounting (Ciril Emaps)

https://github.com/Green-Software-Foundation/sci/pull/357

Using an attributional model can be problematic for the reasons Dan and Dom describe, but at the same time, that PR features a common criticism of using a marginal / consequential approach - that it relies on a counterfactual.

That is, for you to be able make a reliable statement about whether powering down equipment would have led to someone switching off a dirty form of electricity generation, you need to be able to demonstrate that it happened because of your intervention, which is really hard.

An equivalent for ICT?

I figure the closest I can think of is the world of CDNs - one of the benefits of content delivery networks is that they save capacity on busy links by serving the same file from a store, closer to the requester, using less congested routes.

Is there something like this consequential vs attributional  division when talking about network capacity, and capacity planning?

I feel like it might help when trying to make sense of the savings claims you see in papers in the original email message.

Chris Adams

> On 9. May 2023, at 15:02, Dom Robinson <dom@id3as.co.uk> wrote:
> 
> I have to say, Dan, I second you here.
> 
> As Dan alludes to, and as we are rapidly finding out evidentially through work at GreeningofStreaming, as far as we can tell, nothing in the networks at all, apart from some layer two transmission in Wireless networks (perhaps a couple of percent anecdotally) is “elastic” in terms of energy use relating to traffic transmission over networks. “Not using the Internet” does not save anyone, any energy anywhere in the network. It is on, provisioned for peak at all times. Routers, switches, amps are inelastic. 
> 
> Video payloads do create variable energy workloads in “layer 4” (compute) at the ends of the network, where they are being encoded and decoded, but the transmission of bits (all zeros - ‘no information’, or a combination of zeros and ones - ‘traffic’) doesn’t vary the network energy at all. 
> 
> It is increasingly evident to me that it is a misleading urban myth to correlate gigabytes of data transfer and carbon. I am looking for somebody who can actually prove anything to me to contrast this argument, and I have been looking hard for nearly 3 years now.. I acknowledge that it’s easier for accounting firms to correlate carbon to data transfer, because Network economics account for money in terms of data transfer. But there is no relation between carbon and GB.
> 
> In fact, counterintuitively there is a stronger argument that using “more data” commercially drives a swift migration to new technologies, which almost invariably, save energy once utilisation passes an equivalent point on the legacy network.  Newer IT is almost invariably more efficient.
> 
>  Obviously this does not consider embedded emissions in the rollout and physical deployment of such networks. However, in carbon intensity terms - and I am conjecturing here - if you took all the internet kit and burned it and turned it into energy, the energy release would probably (i would guess) run the Internet for a few weeks at best, so the operational energy is a much much more important element to consider in sustainability terms in the long run in my opinion.
> 
> #catamongpigeons :)
> 
> -- 
> Dom Robinson
> www.id3as.co.uk <http://www.id3as.co.uk/> 
> www.greeningofstreaming.org <http://www.greeningofstreaming.org/> 
> uk.linkedin.com/in/domrobinson <http://uk.linkedin.com/in/domrobinson>
> Meet >> https://calendly.com/id3as 
> 
>> On 9 May 2023, at 03:35, Daniel Schien <Daniel.Schien@bristol.ac.uk> wrote:
>> 
>> 
>> Sadly the paper draws the wrong conclusions - it makes the same mistake of Obringer et al. in "The overlooked environmental footprint of increasing Internet use. Resources, Conservation and Recycling, 2021" who first develop an attributional estimate of the monthly carbon footprint of a single user "their monthly carbon footprint would be 9.4 kg CO2e." and then claim that "Simply turning off the video, however, would reduce the monthly emissions to 377 g CO2e.” - which is not the case, because routers are notoriously inelastic.
>> 
>> Tabaeiaghdaei et al. combine a model of the dynamic carbon intensity (very small)
>> <image.png>
>> 
>> with an attributional model of the idle power consumption (dominating power consumption by far)
>> <image.png>
>> 
>> 
>> They then wrongly apply it to bitrates from a number of content providers to estimate savings. Because the idle energy consumption is fix - the below estimates of savings do not actually materialise.
>> 
>> <image.png>
>> 
>> 
>> These accounting mistakes just don't stop. Reviewers are not aware of it and let them pass. It also looks really impressive that Netflix network footprint could be reduced by 68%.
>> 
>> Daniel Schien
>> Senior Lecturer in Computer Science
>> Department of Computer Science | University of Bristol
>> Submit software engineering project ideas for 2022
>> 
>> bris.ac.uk/software-engineering <>
>> Watch: https://youtu.be/lU-ZsBDFWDI
>> 
>> Merchant Venturers Building , Woodland Rd Bristol, BS8 1UB
>> Book a meeting: https://outlook.office365.com/owa/calendar/OfficeHours@bristol.ac.uk/bookings/
>> 
>> From: E-impact <e-impact-bounces@ietf.org> on behalf of Dom Robinson <dom@id3as.co.uk>
>> Sent: 04 May 2023 11:34
>> To: e-impact@ietf.org <e-impact@ietf.org>
>> Subject: Re: [E-impact] Carbon-Intelligent Global Routing in Path-Aware Networks
>>  
>> Very interesting -thanks Hesham.
>> 
>> -- 
>> Dom Robinson
>> www.id3as.co.uk <http://www.id3as.co.uk/> 
>> www.greeningofstreaming.org <http://www.greeningofstreaming.org/> 
>> uk.linkedin.com/in/domrobinson <http://uk.linkedin.com/in/domrobinson>
>> Meet >> https://calendly.com/id3as 
>> 
>>> On 4 May 2023, at 11:14, Hesham ElBakoury <helbakoury@gmail.com> wrote:
>>> 
>>> 
>>> This is a new SCION paper which uses SCION path aware architecture to estimate and disseminate carbon intensity of inter-domain paths.  You can access the paper here: https://scholar.google.com/scholar_url?url=https://netsec.ethz.ch/publications/papers/green_routing2023.pdf&hl=en&sa=X&d=2221193519881579352&ei=L39TZNbNL4HBsQKHg4Mw&scisig=AGlGAw8bvu6JoiUYgq86kLYpZ0W1&oi=scholaralrt&hist=LiEm55wAAAAJ:6297799576989356312:AGlGAw-I9fbYdpGr-eDV-oh3lITV&html=&pos=2&folt=rel&fols=
>>> 
>>> SCION team designed CIRo, a system for forecasting the carbon intensity of inter-domain paths and disseminating them across the Internet. Using SCION code base the team implemented CIRo PoC to demonstrate the potential of CIRo for reducing the carbon footprint of endpoints and end domains through large-scale simulations. CIRo PoC  shows that it can "reduce the carbon intensity of communications by at least 47% for half of the domain pairs and the carbon footprint of Internet usage by at least 50% for
>>> 87% of end domain."
>>> 
>>> Hesham
>>> 
>>> -- 
>>> E-impact mailing list
>>> E-impact@ietf.org
>>> https://www.ietf.org/mailman/listinfo/e-impact
> -- 
> E-impact mailing list
> E-impact@ietf.org
> https://www.ietf.org/mailman/listinfo/e-impact