Agenda for the IVY 117 WG session

Session:

Thursday, July 27, 2023

Session II (13:00 - 15:00 PDT)

WG Chairs:

Daniele Ceccarelli (daniele.ietf@gmail.com)

Qiufang Ma (maqiufang1@huawei.com)

Available During Session:

Meetecho: https://meetings.conf.meetecho.com/ietf117/?group=ivy

Onsite tool:
https://meetings.conf.meetecho.com/onsite117/?group=ivy

Audio Only: https://mp3.conf.meetecho.com/ietf117/31462.m3u

Chat Only: https://zulip.ietf.org/#narrow/stream/ivy

slides: https://datatracker.ietf.org/meeting/117/session/ivy

Notes: https://notes.ietf.org/notes-ietf-117-ivy

Avaliable After Session:

Notes: https://notes.ietf.org/notes-ietf-117-ivy

Recording: https://www.meetecho.com/ietf117/recordings#IVY

Slot Information

1. Introduction (20 mins)

   WG kickoff: introduction, scope, charter review

Rob Wilton: Well done, a formed WG. The charter now covers a core
network inventory YANG model which is quite short-term and focused, and
then after that we'll look at how to extend the WG and what it should
cover.

Alexander Clemm: Will the WG items also include virtualized components,
like a virtual router?

Daniele Ceccarelli: Definitely yes for a virtual router, no for a
virtual network function.

Alexander Clemm: Going to be a lot of gray areas.

Ahmed Elhassany: Exclusion of how network inventory could be used
worried me. We might end up with something that is theoretical without
implementation visibility. The usability of these models needs to be
included.

Daniele Ceccarelli: That's part of the WG scope (informative examples
describing how the inventory data could be used).

1. An Inventory Management Model for Enterprise Networks (20 mins)

   Presenter: Bo Wu

   Draft:
   https://datatracker.ietf.org/doc/html/draft-wzwb-opsawg-network-inventory-management-02

Diego Lopez: Some terms that have been propsed are somehow more
connected to topology than to an inventory. We need a revision about
this terminology.

Thomas Graf: Comment on the speed of the interface, we should find a
terminology more general and applicable to other components. The
operators would like to use two values of an interface: phyiscal
capacity and planned capacity.

Italo Busi: How you thought about putting the the termination point
attributes into part of component? We think different components may
have component specific attributes, I don't understand why we need to
navigate from the component to the TP to get attributes which are
component specific.

Bo Wu: It depends on the core network inventory will provide link
information, in that case, the termination point is a good candidate for
the interface attributes.

Med (from chat window): This mainly for the correlation purposes Italo.
But you have a valid point.

Jan Lindblad (from chat window): Often when there are many ways to do
something, and perhaps history too, it is customary to start with a
requirements discovery effort. Just a thought.

Olga Havel: Suggesting to follow layer-2 and layer-3 patterns in term of
how you agument the attributes. You say augmentation of the node but it
is obvious the physical attributes, same as the TP or the port.
You removed the link, how do you see the modeling of the link?

Bo Wu: the physical link has issues now, like 8345 is now only
supporting unidirection link, the other part, links are passive
inventory which controller cannot easily to get but needs to be manually
configured.

Olga Havel: I think you are right in terms of uninteractionals. So when
we add bidirectional, hopefully you could use them. We may need to have
some information about what the intent of the physical topologies vs.
just collection from the operational status.
The last comment: do you see it as a single network? Do you see any
needs to do any partition or domains of the network? if you are planing
to have multiple network instances, you cannot have a link between
multiple network instances.

Bo Wu: Not sure if network inventory should support mutiple network
instances.

Olga Havel: This is more of a question for you, is it a one network
instance that you are talking about.

Oscar Gonzalez de Dios: The first step of the WG is to define which is
the foundation of the network inventory, topology or a seperate
structure but link to the topology. We started an implementation with
ietf-network in Telefonica but it confused a lot of people. What items
from the ietf-network is useful for the inventory and can we take these
items out to a new structure? How much in network topology is not useful
but you need to carry?

Daniele Ceccarelli: Useful feedback.

Rob Wilton: As a participant, if we think vrtual devices are in scope,
these won't have physical interfaces, we need to choose terms generic
enough but still have to be understood.

Ahmed Elhassany: Inventory is defiend as a list of equipments connected
to a network in the draft. So Anything is not connected is not part of
the inventory, e,g, the backup device? But it should be included as
well. Inventory is just listing the things that we have, how they are
connected is part of the topology, not part of inventory.

Alexander Clemm: there used to be the entity MIB in RFC4133 which seems
to be related to this. Maybe look into what could be reused from the
entity MIB.

Bo Wu: Already reuse some entities in IETF hardware.

James Cumming: I saw license in this model which I know in the scope for
the WG, it seems sensible to pull that out into a software inventory
that could potential then be augmented as needed, rather than embedding
into the overall node where it may not apply.

Daniele Ceccarelli: It is useful if it's done quickly. The process is to
start something simple, basic and then build on top of that.
We are not going to adopt any of the draft being presented. The first
part of the session is information sharing, but we probally could use
some draft that fits most of the needs as starting point.

Rob Wilton: Maybe organize some interims or something to keep the energy
going rather than waiting 4 months.

Daniele Ceccarelli: Might be tricky to find a slot in the day that fits
everyone. We will find a way.

Marisol Palmero: There is a lot of similarities to the work we have on
the DMLMO. Going to present this later and see on the license part how
components are address here. We consider assets which will be good to
have further discussion.

1. A YANG Data Model for Network Hardware Inventory (20 mains)

   Presenter: Italo Busi

   Draft:
   https://datatracker.ietf.org/doc/html/draft-ietf-ccamp-network-inventory-yang-02

Daniele Ceccarelli: As a CCAMP chair, we knew it is not the right place
to move in this work in CCAMP, but there is a lot of energy so we prefer
to let this work done and then decide where this would fit.

Rob Wilton: You want to move fast with the existing draft to get use
cases but I think if you take the core of that out and bring it in
that's going to slow things down. How quickly do you need to current
module to be finished? Do you think it to be published as an RFC as an
agreement of base structure?
Another choice is to publish existing one maybe experimental to get that
done temporarily so you don't have the time pressure.

Italo Busi: Operators would like to have something more stable, rather
than experimental and temporary. We try to see if we can go through
hardware solution now and add the missing pieces in a BC way later on.

Rob Wilton: My concern is not having this as a pushing force to produce
the core model quickly and then come back and tweak that differently.
You can spend the next 4 months to decide.

Daniele Ceccarelli: The experimental sounds interesting, but agree with
Italo that it would be something temporary. The idea is to carve out
from the CCAMP WG and move it to IVY, but try to avoid having technology
specific parts preventing the core model to move faster.

Joe Clark: I like this draft better compared with the other ones. I
think it's a good first step and makes a good basis for what inventory
can be.

Thomas Graf: We should differentiate between an intended interface speed
vs. the physical capacity. The draft is also targeting capacity
planning, you have a more like an intent approach versus a more hardware
approach. It would be good to combine them to clearly see what's
physically available, how much do we want to provision and how that's
relates actually to the end to end VPN service.

Benoit Claise: I read this WG as being of inventory, like hw/sw. Maybe
we want to seperate inventory and how it is connected. Part of metrics
like how to get the termination point connected is a little bit on the
side.

Italo Busi: You need to vavigate between the topology view and the
inventory view. we can have use cases where people provide topology,
inventory and use case where people provide both. And when you have both
views, it's good to be able to associate them.

Benoit Claise: Right, I am wondering if this is part of this one. We
might need to upate the charter every month.

Daniele Ceccarelli: Tend to agree with Benoit. Focusing on something
narrow could help us being faster and effective.

Chaode Yu (from chat window): I think for the inventory data model, it
is better to have a separated root. In the system of operators, topology
and inventory are different concepts. Topology is used to provide some
functional information which could be abstracted. then for the hardware
information, it could have some correlation with the abstract topology
object. One hardware object can be abstracted into several topology
objects.

1. Asset Lifecycle Management and Operations, Problem Statement (10
   mins)

   Presenter: Diego Lopez

   Draft:
   https://datatracker.ietf.org/doc/html/draft-palmero-opsawg-ps-almo-00

Daniele Ceccarelli: You spent months to convince me that it didn't
belong here, now you changed your mind.

Diego Lopez: Now things aren't so clear to me.

Rob Wilton: Two questions: 1) how the asset and inventory relate; 2) The
document is quite long and complicated and not sure people really
understand it, suggest to split it into a common simple core and more
modularity in terms of these bits extending it.
Don't know which WG should go, IVY might be a good place to move
forward, OPSAWG is another choice but sometimes might not get the exact
insterested people to focus.
Surprised it wasn't adopted before.

Diego Lopez: Regarding the first question, asset is something
corresponds to act at something that is a unit of deployment or
acquisition. Assets are combination of hardware and software inventory.
This is something trick and probably we have to refine.
We are trying to seperate and make the whole thing not so complicated.

Daniele Ceccarelli: An VNF is an asset not inventory?

Diego Lopez: Yes, and probably the image of the VNF is part of the
inventory.

Rob Wilton: When operators come in and say we need to solve the problem,
and we should be listening. This work is insteresting.

Diego Lopez: Agree the OPSAWG could be too diverse.

Camilo Cardona: Highlight that the YANG module is defined in a way that
is flexiable. LMO should be able to refer to both hardware and software
and bring the concept of inventory in other documents and define their
entitlements, usage etc.

1. Thoughts on network inventory overview (20 mins)

   presenter: Qin Wu

Thomas Graf: On the right side of the slide #10, if these are not in the
current scope of IVY, where do you think this work is belonging to?

Rob Wilton: No answer now. At the very moment the scope for IVY is
really focused and narrow. And after that the WG charter is fine to
cover other stuff that's related to that. At the moment this might be
slightly early.

Qin Wu: When we define L3NM and L2NM model, we touched some logical
resources without a seperate logical network inventory model. The
logical resource coule be useful information and it's important to
establish the mapping with the logical network resources.

Camilo Cardona: A bit of confusing, because we are just discussing
whether the license is included or not but here the slide says it's in
scope. Is this an official slide?

Daniele Ceccarelli: The only offical thing is the charter.

Mohamed Boucadair (from chat window): @Cardano, formally the charter
says:
The IVY WG will initially focus on developing a core network inventory
model that can be used as a foundation by other models to establish
inventory models that are specific to different hardware technologies.
The following activities will be used to help achieve this goal. It
includes:
B. Hardware/Software components including licenses: Hardware and
Software component management to allow network operators to keep track
of which physical/virtual devices are deployed in the network, including
software and hardware versions as well as licenses/entitlement.
But what I'm hearing is that it seems that scope was too ambitious :-)

Rob Wilton: It is allowed to recharter, suggest not to try to recharter
now.
There is a lot of energy to fix this first problem and try to narrow
down on that problem.

Diego Lopez: Agree that there are many things currently out of the
scope, but might be an important matter that we have to deal with sooner
or later. It would help the normal operations very much.

Rob Wilton: When doing routing proptocols in RTGWG, there is a lot of
discussion how to get a fixed structure to fit everything in a
particular place. OpenConfig built a complicated structure while IETF
tried to keep the modules simple and use schema mount to build more
complicated hierarchy with more flexibility. Prefer not to group too
much things together.
Whether we should define a grouping of inventory to be reusable, might
be helpful to keep flexiblity and not have rigid structure.

Qin Wu: We try to figure out which is in and which is out of the
inventory core model. One choice is to align with Openconfig and define
all together, but agree we should allow the flexibility, schema mount is
one option.

Daniele Ceccarelli: It will be good to clarify as a section what's in it
and what's not in IVY scope in the first network inventory document.

Ahmed Elhassany: Not sure if agree everything here on what's in and out
scope. Probally we should propritize first in the physical things which
are the easiest ones to do. Another note is that it should be easy to
augment later.

Oscar Gonzalez de Dios: You put a lot of management in the scope of IVY
in slide #10. What is your expectation in terms of management for what
relates to inventory here?

Qin Wu: The use cases provided here are not just monitoring but also be
leveraged to manage or control. We did cover how to manage these
information to help understand the use cases of IVY. As a response to
Ahmed, system software is also in the scope, like image, openconfig does
model hardware and software component together to provide consistent
representation.

Dan Voyer: Will there be any mapping between what we are doing here and
the TMF?

Daniele Ceccarelli: From the terminology point of view, yes. More than
the terminology part there is not anything we can promise.

Adrian Farrel: Completely sympathize with the desire to have an
inventory of what's in the warehouse. We need to be clear on our scope
with what is deployed compared to what is on its way to being deployed.

Daniele Ceccarelli: We probably need a deployment status to distinguish
what's in the van, what's inside the site/room/rack, right?

Rob Wilton: Hopefully it may not match the inventory YANG model that
gets produced. The YANG model could just define the inventory and not
tied to a network.

Ignacio Martinez-Casanueva: Propose to start with glossary or taxonomy
to clarify all things and avoid the terminology clashes. Sometimes we
use different terms to refer to the same things.

Rob Wilton: Hopefully not to spend time to do that first and slow things
down, some begining drafts cover the terminology at the same time to
facilitate the understanding.

Alexander Clemm: Make it clear where we expect this model to implement
and live. The network devices is up to the level of the controller, but
anything at the level of OSS is getting into TFM land, we would rather
aviod that.

Nils Warnke: I quite have the oppsite opinion. We need a plan status
somewhere mentioned in the model, propose to keep this at least
optional. it's not necessarily geolocation but I can insert an element
into the network inventory even though it has the operational status of
planned and not active or whatever.

Italo Busi: On the planned vs. deployed issue, we welcome everyone to
join the discussion in CCAMP weekly meeting. We have to be careful about
the inventory data which is not exposed on the interface but is written
manually, which we think out of scope.

1. Open Discussion (30 mins)

Daniele Ceccarelli: How to proceed them? Probally start from a single
document which everyone can contribute is better, rather than everyone
submiting their own drafts and merging together. Good to have operators
drive the work, but doesn't preclude the vendors to contribute as
authors.
We will arrange a couple of interims to speed things up.

Qiufang Ma: Hopefully we can continue the discussion on the mailing list
and reach consensus over the mailing list.

Rob Wilton: I will leave that up to you to decide whether it's a couple
of interims or weekly/biweekly meeting. There is energy here, just don't
want to get slow.