Getting Ready for Energy-Efficient Networking (GREEN) BoF
IETF-120, Vancouver
Wednesday 24th July 2024
13:00-15:00
Regency C/D
Materials : https://datatracker.ietf.org/meeting/120/session/green
Minutes and notes : https://notes.ietf.org/notes-ietf-120-green
Chat room : https://zulip.ietf.org/#narrow/stream/green
Recording : http://www.meetecho.com/ietf120/recordings#green
BoF proposal :
https://datatracker.ietf.org/doc/bofreq-palmero-getting-ready-for-energy-efficient-networking-green/
Draft charter :
https://github.com/marisolpalmero/GREEN-bof/blob/main/GreenCharterProposal.md
== Chairs ==
Jari Arkko jari.arkko@gmail.com
Adrian Farrel adrian@olddog.co.uk
== AD ==
Mahesh Jethanandani
mjethanandani@gmail.com
== Agenda ==
== Minutes ==
Notes by Edwin Sutherland
Administrivia and chairs' introduction
Chairs (10 : 10/120)
[Adrian] Initially questions for clarification, then main
discussion later.
A word from the AD
AD (5 : 15/120) [Mahesh]
Jari: A couple of additional thoughts from the chairs (see slide)
An operator's desires and objectives
Luis Contreras and Emile Stephan (10 : 25/120)
There's clear concern from operators regarding sustainability. data
network energy use was 260-360 TWh in 2022[IEA]. There are also
social responsibility missions and regulations that fall on NOs to
be more sustainable.
Questions:
Jing Wang: Is renewable energy aspect taken into account in GREEN.
For example, monitoring of renewable energy.
Luis: Yes in CO2 footprint.
Suresh Krishnan: Have you thought about priorities?
Luis: Would need more discussion because priorities will be
different.
How does networking equipment use and report energy use?
Tony Li and Kabiraj Sethi (10 : 35/120)
Breakdown of power utilisation, main culprits are: NPU, Optics and
Fabric(backplane)
Power utilisation is a function of: Traffic volume, ambient temp
and altitude.
Reporting - data sheet, EMAN (not requested by customers)
Questions:
Jeff Tantsura: What optics where used for measurement?
Tony Li: We used a variety. Some line cards have power limits and
you have to work within those. If you do everything with ZR you
might have a problem. To make an apples-to-apples comparison, you
need to normalise the optics.
Kurtis Hemmel: Do these figures factor manufacturing costs [energy
costs of manufacturing]? Isn't 90% manufacturing and 10% run-time?
Tony Li: These slides are just operational cost. Not practical to
factor this in as folks are going to buy network devices regardless
and IETF doesn't have a lot of say over manufacturing.
???: What is the methodology for these figures?
Tony: I can only disclose what is on these slides.
Cisco Catalyst Switching/Routing Power and Energy Data Models by
Kabiraj Sethi and Andrew Lu
Power supply stats
These stats can be collected by API.
Cisco catalyst center is also looking at sustainability with
estimated carbon emissions. Where are they getting the carbon
intensity figures from for their estimates?
Jari: These were two interesting examples we picked. Mileage may
vary for other specific use cases and products (e.g., cloud or
mobile).
What did the EMAN working group do and not do?
Benoit Claise (10 : 45/120)
Energy Management (concluded WG), they done work on new energy
related MIBs.
Accounting for the total power received and provided by an entity,,
a network or a service was one of the requirements of EMAN.
Information model was built and explained in RFC 7326 appendix y.
Observations - Energy crises was not long lasting for EMAN to have
an strong adoption. Mainly Cisco Energywise was the primary
implementation.
EMAN did not cover YANG, Energy Efficient Management, Network-wide
energy efficiency metric, such as the power usage
effectiveness(PUE), carbon footprint, new usecase for eco design.
EMAN was a success in part but did not cover everything, thus there
is scope for more to be done.
What the proponents think the WG should work on
Qin Wu (10 : 55/120)
Data models for energy efficiency management
IESG proposed NETCONF/YANG
MIB is good for monitoring but not for contol functionality
YANG is more extensible
Do we need to define a new metadata DB related to Carbon Intensity?
Questions:
Hesham ElBakoury: Will the energy efficiency KPIs or metrics be
standardised by green BOF?
Jari: Yes, the intention is for Green WG to propose and standardise
some of these metrics.
Qin: Some other SDOs have existing methods we might re-use. Also
BMWG already work on benchmarking that may provide a foundation.
Jing Wang: (Fifth slide) Should the energy used by the equipment
measuring the network energy consumption also be included in the
overall figure for the network energy consumption?
Qin: Yes, we are looking at both device and network level metrics
Dean Bogdanovic: Industry is using a term called NegaWatts (Note: N
as in November) to measure the reduction in energy used by applying
new mechanisms. Energy utility, have multiple energy sources - makes
it hard for us to measure carbon efficiency. Should look at what
energy companies are already doing with respect to NegaWatts. We
should use metrics used by the industry and not invent our own.
Qin: Look for the balance between power consumption and
performance. Our framework works regardless of energy source. We are
interested in power efficiency. Our YANG models should allow to
examine the metrics so tht they can be used by any application.
An introduction to the draft charter
Marisol Palmero (10 : 65/120)
Purpose develop energy effiency metrics, methods for measuring
energy consumption in networking devices and optimising energy
efficiency across networks.
Short term goals:
Milestones
Open discussion (50 : 115/120)
Dean Bogdanovic: Customers are asking vendors for energy measurement.
There's an economic incentive to know this measurement and work on
reducing energy consumption. There is demand by the market for this
measurement. I fully support formation of a WG and would contribute.
Gen Chen: Impact of energy efficiency on operator service is out of
scope. Would we consider bringing this in scope in the future as this is
important for operators?
Alexander Clemm: Support formation of a working group. We tried in the
past to bring this kind of work into other groups, and that didn't work
out. Needs a dedicated group which can have critical mass.
Robert Wilton: Support the formation of a working group. Have seen
drafts struggling in different places without a proper home. Does this
stuff matter? Yes it does. There is demand for it. Is IETF a good place?
Yes we have the experts to build YANG models and control capabilities.
Even if we have short-term deliverables, having a focal point for energy
related discussions would be useful, is that this group or E-impact?
Need to be clear on the split of responsibilities.
Jan Lindblad: +1 to Rob's point regarding need for a forum for energy
topics in IETF. There's a lot of things present in the world showcasing
how they are reducing power consumption but no standardised way to
report this. We need a agreement from multiple vendors on having a
standard way of reporting energy metrics.
Chris Seal: In support. Not sure how to charter this. There are some
items that are out of scope that we need to rethink. For example routing
and algorithms, would this be easy to recharter if we want to work on
such work? I agree with the priority list, but if it is hard to
recharter, then we should include the low priority things now.
Xiao Min: I support forming a WG. Suggest taking security mechanism into
acount. Specifically, why apply energy controls to networks, we must
ensure customers network are not heavily affected such as during special
events such as (burst traffic, line card maintanance)
Stuart Card: In favour of goals here. Concerns of assumptions and
resources that may be required. If installing equipment that saves
energy results in higher costs. The competancies needed include
expertise in thermodynamics of economies (not typically found in the
IETF).
Suresh Krishnan: In favour of WG. We should spend effort on priority. I
want a tighter scope so we can get work done ASAP. We need to get
something out quickly.
Tim Chown: He sees two paths, one is measurements and another is
control. Tim believes they should be two different groups. This group
should focus on measuring. Supports the YANG work at once.
Ali Rezaki: I support. Need to quantify effort/resources needed to make
measurements. Basline and target setting is out of scope, but we need to
be mindful of demonstrating the usefulness of our work. Where in the
network should we focus on optimising? So it is critical to get data
from the network. There seem to be a gap between what operators are
asking and what the charter is offering, so perhaps this is covered by
prioritising?
Jing Wang: In support of a new WG and would participate.
Joe Clarke: Echo Rob. Support the formation of this group. Willing to
participate in a new WG. A tight scope is useful. Learning from EMAN,
the need for energy management is different regionally, so we need to be
very proscriptive of what we want vendors to implement. Let's measure
what out there first with basic controls for the operators to say what
is on or off. Then iterate on what comes next? +1 on Tim's comment on
some of the later stuff happening in other WGs. didn't see traction in
OPSAWG because it has larger focus - we have enough critical mass now to
pull this out to a new WG.
Benoît Claise: Believes it important work to be done. Assume if we
re-build EMAN with YANG would it be good enough for Operators? Look at
the requirements on 1-2 use cases and avoid mistakes of EMAN in trying
to cover too many things. Carbon footprint would be a bridge too far.
Dean Bogdanović: We should keep it simple for the energy consumption.
Heard several proposals that are boiling ocean. Energy consumption per
bit? Energy efficiency rating for network equipments. He will put out a
draft about the current units the industry is talking about.
Laurent Ciavaglia: +1 with Ali's points. Do we [this group] have an
obligation of reporting energy measurement or to optimise consumption?
Eliot Lear: I support forming a WG, and the charter is in pretty good
shape. +1 with Dean's point. We have been talking within the context of
network devices, but we should make sure the model we deliver can be
expressed in non-YANG devices.
Ali Rezaki: Question about deployment strategy? Can that be part of the
scope of the WG? Example, rather than implementing these metrics on all
network elements would measurement probes be more practical?
Georgios Karagiannis: Agree with forming of this WG. +1 agrees with Dean
on simplicity of metrics from other industry. We need to think about
life cycle assessemnt. We need to make it clear in which stage of the
life cycle assessment the output of our work maps to.
Gyan Mishra: In support of WG. Goals and ideas of charter are good.
Gathering accurate metrics is important to reducing carbon footprint. It
is also important for network operators to calculate their cost service
delivery.
Adrian:
A recharter is really cheap. Do not be too concerned about high layer
use of the data for now. Let's get the foundations and tools in place to
put us in a position to build on this.
Show of hands [see poll tool]: 53 people willing to actively
contribute
AD:
There is clear interest in having a forum where the discussion for
energy management can happen. This was apparent by the number of people
who came to the mike and ~50 who said they supported the formation of
the WG, with a handful adding caution to the formation, without saying
no its formation. In addition, several of those folks who came to the
mike indicated a desire to contribute to the WG, either in the form of
providing comments or writing drafts. The next step is to discuss the
charter of the WG and come up with a list of deliverables that can be
achieved in the next year or two. The initial charter should be viewed
as a stepping stone, and not the end game for the WG. The charter can be
tweaked to include items that were discussed today, but deferred for a
later time.