[irtf-discuss] Re: [E-impact] Re: [Green] Network Equipment Energy Efficiency Metric

[Mark Butcher <[e-mail]>]

SCI is OK as a concept but is 100% reliant on the underlying data being fed into it.

Thanks


Mark Butcher
Director
Posetiv Cloud Ltd

e - [e-mail]
m - 44 (0)7746 189321

On 14 Nov 2024, at 12:14, Hesham ElBakoury <[e-mail]> wrote:


Hi Mark,
Did you look at SCI (Software Carbon Intensity) metric which is developed by Green Software Foundation and standardized by ISO [1]. It  can be defined for any unit including a user.

The equation to calculate an SCI score is elegantly simple. This simplicity means it can be applied in a number of different scenarios.

[alt_text]

SCI = ((E *I) + M) per R

E = Energy consumed by a software system I = Location-based marginal carbon emissionsM = Embodied emissions of a software system. R = Functional unit (e.g. carbon per additional user, API-call, ML job, etc)

I am wondering who is using it?

Hesham
[1] https://learn.greensoftware.foundation/measurement/#:~:text=The%20Software%20Carbon%20Intensity%20(SCI)%20specification%20is%20a%20methodology%20developed,cars%20they%20produce%20every%20year.

On Tue, Nov 12, 2024, 10:27 AM Mark Butcher <[e-mail]<mailto:[e-mail]>> wrote:
Hello!

I’m definitely never going to claim to be a grid expert, but… understanding the grid’s dynamics, especially active, reactive, and apparent power is definitely essential if the objective is to get a better handle on power usage across networks. It’s obvs. not just about energy efficiency; translating this into carbon emissions gets even more critical when you consider how different power sources impact the grid's overall carbon intensity. I’ve had some interesting chats with operators this year around how they are planning to move from supply side to demand side shaping.

I don’t envy the Grid operators facing the balancing act between cost, stability, emissions targets, and resilience (especially as the subject of renewables come into play). Getting even vaguely accurate carbon intensity metrics from them isn’t easy and there’s lot of sweeping assumptions. What’s your view on the regional info shared by operators like the National Grid? I don’t think it’s perfect but it’s at least more accurate than a national average as it reflects their understanding of the flows/mix.

Starting with simpler elements, like SFPs, and then scaling up to more complex topologies makes sense. Tracking energy consumption at this granular level, from individual components up to entire network sections, could provide the data points needed to make informed decisions on reducing emissions, rather than just measuring usage (if we assume we can use intensity metrics from the operators).

Do you think breaking it down into energy consumption at the element, group, and network levels would give insights we could actually use and more importantly to act on? Or is this overkill without an agreed structure/approach on what to do with the data and how to interpret it?

Have you got any good examples where you’ve seen this work? I’m doing a lot of work in data centres on this subject and the networks teams are often excluded because of the complexity of their operations (outside of core DC networks).

Looking forward to discussing this further! I personally believe there’s definitely a lot we can do. It would also be good to consider how the data that’s collated can be rolled into the reporting within IT and the wider business.

Thanks

Mark



From: Dean Bogdanovic <[e-mail]<mailto:[e-mail]>>
Date: Tuesday, 12 November 2024 at 17:20
To: Mark Butcher <[e-mail]<mailto:[e-mail]>>
Cc: Michael Richardson <[e-mail]<mailto:[e-mail]>>, [e-mail]<mailto:[e-mail]> <[e-mail]<mailto:[e-mail]>>, E-Impact IETF <[e-mail]<mailto:[e-mail]>>, [e-mail]<mailto:[e-mail]> <[e-mail]<mailto:[e-mail]>>
Subject: Re: [Green] [E-impact] Network Equipment Energy Efficiency Metric

I’m really curious how many people understand how electrical grid operates,

the difference between active, reactive and apparent power in electrical grid. And how the grip operators are balancing power plants to maintain the needed power levels, how the operators decide between economic dispatch, grid stability and frequency regulation, environmental regulations and emissions , operational flexibility at the power plants etc.
Until we do, I would not speak about CO2 emissions and other complex energy and traffic engineering relations.

Lets first figure out how to report energy consumption by the networks, from very basic power consumers (SFPs) to more complex, like topologies. And provide the ability to provide the energy consumption information about an element, group of elements, part of network or even whole network to people who will know what to do with the info.

Dean

On 12 Nov 2024, at 14:59, Mark Butcher wrote:
What’s interesting is seeing how kWh/GB is so variable – either based on network demand, the reality of the power curve and most importantly for me, when translated into CO2e how the variability massively increases (time and location)


From: Michael Richardson <[e-mail]<mailto:[e-mail]>>
Date: Tuesday, 12 November 2024 at 14:32
To: [e-mail]<mailto:[e-mail]> <[e-mail]<mailto:[e-mail]>>, E-Impact IETF <[e-mail]<mailto:[e-mail]>>, [e-mail]<mailto:[e-mail]> <[e-mail]<mailto:[e-mail]>>
Subject: [Green] Re: [E-impact] Network Equipment Energy Efficiency Metric

Is is useful for network planners to know kWh/peak-GB for an entire infrastructure?

This doesn't try to model reaction of network equipment to increased flows,
but rather to model how infrastructure grows to accomodate higher peak loads?

There is also an induced demand issue.  Networks grow when they fill,
assuming there is revenue to support it.  This is akin to transportation:
      https://www.vtpi.org/gentraf.pdf

--
Michael Richardson <[e-mail]<mailto:[e-mail]>>, Sandelman Software Works
 -= IPv6 IoT consulting =-                      *I*LIKE*TRAINS*


_______________________________________________
Green mailing list -- [e-mail]<mailto:[e-mail]>
To unsubscribe send an email to [e-mail]<mailto:[e-mail]>

--
E-impact mailing list -- [e-mail]<mailto:[e-mail]>
To unsubscribe send an email to [e-mail]<mailto:[e-mail]>