►
From YouTube: IETF112-NMRG-20211108-1200
Description
NMRG meeting session at IETF112
2021/11/08 1200
https://datatracker.ietf.org/meeting/112/proceedings/
A
Okay,
so
hello,
everyone
again
hello
for
those
could
join
the
unemoji
meeting
and
welcome
to
this
meeting
that
I
will
culture
with
lauren
today.
A
So
before
we
we
start
with
the
missing
just
some
notes
as
usual
or
if
you
can
move
to
the
next
slide.
Please.
A
Okay,
so
so
far
the
easy
note
where
are
again
audio
and
value
recording,
so
just
to
let
you
know
that
the
rtf14x
recording
of
online
and
in-person
meeting
working
with
your
video
and
photographs
and
please
just
recording
online
in
particular,
for
this
online
meeting
as
you
participate
online
and
that
you,
when
you
turn
your
camera
on
or
you
make
microphone
on,
then
your
concern
to
appear
in
such
recordings
next
slide.
Please
law.
A
Share
privacy
and
code
of
conduct,
or
as
a
participant
participant
in
or
attended
to
any
hierarchy
of
activity,
knowledge
to
that
right,
written
audio
and
video
and
photography
because
of
meetings
may
be
made
public
personal
information
that
you
priorities
will
be
handled
in
accordance
with
the
privacy
policy
and
the
link
provided
in
the
slide
as
a
participant
or
attendee
you're
going
to
work
respectfully
with
other
participants
and
please
contact
the
ombuds
team.
If
you
have
questions
or
concerns
about
this,
you
may
also
refer
to
the
mention
irfc's
here
for
my
information
next
slide.
A
So
the
goal
of
the
irtf
the
iota
conducts
research.
It
is
not
a
standard
development
organization,
the
irc
focuses
on
longer
term
research
issues
related
to
internet,
while
the
parallel
organizations
idea
focuses
on
short-term
issues
of
engineering
and
standards-making,
while
the
rtf
can
publish
informational
or
experimental
documents
in
the
arc
series,
its
primary
goal
is
to
provide
development
of
research,
collaboration
and
teamwork
in
exploring
research
issues
related
to
internet
protocols,
application
architecture
and
technology.
A
So
it's
already
said
so.
This
is
an
online
meeting,
so
the
session
is
being
recorded
when
you
are
not
presenting.
Please
mute
yourself
and
put
your
video
off
and
when
speaking,
please
start
by
stating
your
name
and
affiliation.
Here
you
have
some
links.
In
particular,
we
have
the
link
to
the
notes
that
you
can
also
participate
into
the
kodi
md
link
and
know
that
after
the
meetings,
the
video
will
be
put
online
in
the
youtube
channel
of
the
ietf
next
slide.
A
So
the
agenda
for
today,
so
there
will
be
the
short
introduction
from
pharmacy
lg
chairs
that
we
will.
We
are
we
already
started,
but
that
will
continue
and
after
that
we
will
have
our
first
set
of
presentations
regarding
the
status
of
draft
about
the
in
about
ibm,
one
on
network
management,
intent
and
the
other
one
with,
inter
connection
intents
and
as
a
bit
of
follow-up
of
the
previous
meeting.
A
Also,
we
have
a
second
set
of
presentations
related
to
digital
twins
with
a
first
presentation
that
is
not
directly
related
to
a
draft,
it's
implementation
of
a
project,
and
then
we
have
two
predictions
related
to
a
draft
and
at
the
end
we
have
another
presentation
about
an
ai
framework
for
network
management
that
will
be
presented.
A
So,
just
in
the
in
this
introduction
just
to
make
you
a
kind
of
status
regarding
our
research
group
documents.
As
you
know,
we
have
two
documents.
I
think
there
isn't
a
bit
of
type
in
the
second
document,
so
the
first
one
is
about
the
to
our
body,
ibm's
personalized,
with
concept
and
definition
in
this
segment.
A
That
was
the
classification
so
just
to
let
you
know
that,
as
you
may
have
seen,
this
document
will
be
sent
for
the
ihf
chair
review
and
we
receive
receive
the
review
for
for
both
of
the
draft
in
september
and
we've
as
kind
of
results.
First,
a
set
of
discussion
mailing.
B
A
Regarding
the
different
commands,
just
for
clarification
and
see,
oh,
we
will
proceed
with
different
commands,
so
actually
there
are
one
the
first
one
about
a
concept
and
definition
draft
is.
Our
revision
is
under
preparation
to
take
into
account
the
different
command
and
the
other
one
for
the
classification
graph.
There
are
also
been
an
updated
version
that
has
been
proposed.
A
We
still
some
minor
command
to
be
to
be
addressed,
but
mostly
for
this
draft.
The
the
next
step
will
be
to
to
see
to
it
will
be
under
through
the
mailing
list
and
with
the
new
revision
to
see
if
they
fulfill
the
commands
we
receive
from
ahf
chair
review
before
we
proceed
further
with
the
drafts.
D
Okay,
yes,
I'm
editing
live
the
update
to
change
the
mistake
in
the
slide,
so
this
will
be
corrected
soon
yeah.
So
the
two
documents
are
in
discussion
between
our
tf
chair
and
the
shepherd
and
the
authors
we
have
going
through
a
round
of
different
exchange.
The
updates
have
addressed
part
of
the
irtf
shared
comments,
but
we
still
have
to
go
for
with
it
for
a
round
of
alignment
for
that,
we
hope
this
will
be
progressed
soon.
D
C
D
Okay,
olga,
I
suggest
we
have
maybe
a
round
of
discussion
between
us
so
that
we
can
see
how
we
address
the
the
comments,
if
you,
if
you
like,
I
can
be
available
this
week
to
discuss
this
with
you.
D
Okay,
if
I
may
just
add
something
as
a
co-author
of
the
draft
on
concepts
and
definitions,
we
have
gathered
also
to
see
how
to
address
best
the
comments
from
the
rtf
chair.
So
there
is
a
a
new
update
of
the
draft
pending
and
we
we
have
a
few
proposals
to
address
the
positioning
with
respect
to
the
research
work
being
done
on
ibm
and
to
better
position
the
draft
with
respect
to
that.
So
we
hope
to
be
able
to
release
the
new
version
again
this
week.
Still.
D
I
hope
this
will
be
addressing
the
comments
from
from
collin.
E
I
attempted
to
say
a
long
time
listener
first
time
caller,
but
I
won't
so
yeah
if,
if
it's
useful
to
have
a
to
schedule,
a
call
to
talk
about
these
comments,
then
I'm
happy
to
try
and
fit
that
in
this
week
or
next
week.
D
Okay,
perfect
thanks
colin.
I
think
we
will
of
course
keep
you
in
the
loop.
I
think
we
have
something
to
do
now
on
more
on
the
offer
size
to
make
sure
the
the
updates
are
satisfactory
and
then,
of
course,
we
can
check
with
you.
A
Yeah,
so
just
for
future
imaging,
so
we'll
continue,
of
course,
with
virtual
meetings.
We
try
to
have
it
monthly
and
we
will
try
to
continue
it
monthly.
We
were
not
so
much
regular
in
the
last
month,
as
you
may
have
seen.
A
I
think
there
was
a
summer
break
and
then
we
have,
of
course
I
think
we
are
all
a
bit-
let's
say
environmental
and
by
various
remote
meetings,
that
we
have
each
on
your
home,
but
anyway
we
will
try
to
resume
a
bit
more
the
different
activities
we
have
in
the
group.
In
addition
to
this,
of
course,
iatf
meetings.
A
In
particular,
we
want
to
resume
the
activity
regarding,
let's
say,
the
actual
network
and
editing
associates
license.
There
are
some
propositions
that
we
have
a
workshop
around
these
topics,
so
I'm
working
on
virtual
meetings,
so
we
try
to
to
organize
it
and
also
regarding
the
a
research
document
after
summers
we
we
had
input
during
summers,
but
we
did
not
yet
consolidate,
and
so
we
have
to
also
to
organize
dedicated
meeting
to
make
it
more
I'd,
say
more
consolidating
to
have
something
also,
which
can
be
reviewed
and
more
integrated.
A
Let's
say
so
anyway:
yes,
there's
also
the
mailing.
B
A
Course,
different
tools
that
you
haven't
collaborated
with
so
do
not
hesitate
to
use
it
or
to
also
propose
if
you
have
any
ideas
or
topics
you
would
like
to
discuss
in
the
group.
It's
also
still
is,
of
course,
open
to
20
ids.
D
A
It's
it's
network
management.
B
D
F
Hello,
everyone
I'm
from
china,
mobile
and
I'm
a
colleague
of
danielle.
Can
you
hear
me?
Yes,
we
can
hear
you
well
so
so
chef
should.
I
stand
up.
F
Okay,
so
hello,
everyone.
So
this
is
a
third
time.
Third
update
of
the
original
draft
network
measurement,
intent
and
I'll.
First
briefly
talk
about
the
the
updates
of
the
draft,
and
then
I
want
to
call
for
some
suggestions
on
the
stage
of
the
draft.
F
So
next
page
please,
so
we
can
see
in
the
diagram
that
this
time
we
there
is
not
very
much
many
updates
compared
to
the
draft
presented
in
the
last
meeting
and
the
major
update
is
the
procedure
we
the
whole
procedure
in
the
in
the
diagram
at
the
right
side
of
the
slide.
We
can
see
that
in
the
original
diagram
in
which
the
model
ni
policy
is
a
an
independent
part
of
the
whole
procedure
and
only
communicate
with
the
translation
module,
which
we
think
is
not
very
reasonable
and
accurate.
F
So
the
the
left
side
leads
the
original
procedure,
and
so
we
we
think
we
move
the
nmi
policy
model
inside
into
the
whole
procedure,
where
it
connects
the
nmi
translation
model,
compliance
assessment,
module
and
or
frustration
model
together.
So
it
is
the
break
now
which
we
think
is
very
reasonable.
F
Now,
because
the
in
original
procedure,
the
output
of
translation
is
some
policies
and
the
opposite
of
complex
assessment
is
the
result
and
some
decision,
whether
data
analysis
or
result
is,
is
matched
with
the
network
measurement
intent
and
the
input
of
auto
whereas
and
the
input
of
orchestration
model
should
be
some
policies
or
actions
according
to
some
templates,
so
they
are
not
very
matched,
so
we
moved
the
nmi
policy
model
inside
the
procedure
and
and
now
the
outputs
of
translation
model
and
compliance
assessment
model
will
all
be
translated
into
some
policies.
F
F
So
this
time
we
so
this
is
the
major
update.
We
think
we
should
modify
the
whole
procedure
and
we
think
that
the
so
next
side,
please.
F
I've
already
stated
about
the
study
about
the
the
updates
in
this
time
and
since
the
the
the
draft
has
been
updated
for
three
times,
and
we
have
already
shown
the
necessity
in
defining
and
satisfying
the
network
measurement
intent.
F
We
could
know
how
network
behave
through
the
process,
and
so
we
think
that
the
draft
is
relatively
stable
right
now
and
we
are
asking
for
the
permission
of
chairs
and
other
participants
if
the
threat
could
be
adopted
or
or
it
if
it's
still
in
early
stages,
and
if
it's
not
so
mature,
we
wonder
if
we
could
start
writing
the
on
the
draft
that
the
in
intent
based
network
use
cases
because
we
think
network
management
intent
is
a
very
it's
a
very
important
use
case
of
ibn.
F
So
we
have
sent
emails
to
chess
for
several
times,
but
we
haven't
replied
get
any
reply,
so
we
wonder
if
how
to
promote
this
draft.
So
if
anybody
is
very
interested
in
the
topic
and
you
can
send
emails
to
us,
we
can
promote
together.
So
that's
it
yeah.
Thank
you.
D
Okay,
I
don't
see
anything
also
in
the
chat,
also
so
to
address
your
your
request
for
research
group
adoption
or
the
next
step
of
the
document.
So
I
mean
there
are
different
things
we
could
clarify.
First,
a
bit
I
mean
we
in
the
research
group.
D
D
Nevertheless,
in
the
energy,
in
the
last
couple
of
years,
we
have
defined
a
procedure
to
assess
the
maturity
of
documents
that
you
would
like
to
adopt
as
research
group
document.
So
we
can
point
you
to
these
procedures
and
if
there
are
requests
from
you
to
do
that,
we
can
start
the
research
group
call
for
adoption
the
things
we
are
trying
to
look
for
maturity
in
terms
of
research
group
documents
are
based
on
several
aspects.
D
So
essentially,
if
the
work
is
mature
enough,
if
it,
the
scientific
work
that
is
being
addressed
is
complete
and
what
types
of
research
output
are
produced.
We're
also
looking
I
mean
for
technical
soundness
of
the
overall
documents
and
also
if
the
document
is
in
good
shape,
and
the
final
thing
we
are
checking
is
if
there
are
enough
support,
either
from
the
authors
or
other
participants
in
the
research
group
to
actually
move
the
work
forward
up
to
a
kind
of
completion.
D
I
mean
what
will
be
the
if
we
can
complete
a
publication
of
this
document
in
what
will
be
the
form
reading
by
the
authors,
so
that
we
can
publish
the
outcome
of
this
research
as
part
of
an
outcome
of
the
energy.
D
Just
a
comment
from
my
side
on
this
intern
based
network
use
cases.
I
think
this
is
good
because
we
had,
in
our
ibm,
workbench
like
them
to
study
different
use
cases
for
intern-based
networks,
especially
after
we
had
the
initial
concept
and
classification
draft
being
laid
down
with
which,
where
we
stand
now,
so
it's
good
to
have
this
the
track
measurement
intent
as
a
proposed
use
case.
D
What
is
still
unclear-
and
we
have
to
work
this
out
as
a
group-
is
whether
we
want
to
have
a
collection
of
individual
use
case
drafts
on
different
topics
or
if
we
want
also
to
have
a
kind
of
a
single
document
that
takes
multiple
use
cases
and
tries
to
reflect
on
either
commonalities
differences
and
learnings
across
the
use
cases
we
initiated
that
sometimes
ago.
We,
it
was
not
clear
which
way
we
go
with
the
group.
D
I
think
now
we
have
two
to
three
different
use
case
draft
that
we
can
maybe
try
to
work
together
and
see
if
we
want
to
keep
them
independent
and
or
also
to
consider
taking
part
of
them
into
a
common
document.
D
So
I
think
it's
we
are
at
a
nice
step
in
terms
of
progressing
this
work,
but
it's
also
up
to
the
research
group,
in
fact,
to
decide
the
authors
and
the
research
group
to
decide
what
that,
what
they
don't.
They
want
to
do
next
with
this
document,
and
I
will
also
the
floor
to
jerome
to
know
if
he
has
anything
to
add
on
this.
F
Yeah,
thank
you
lauren.
I
really
agree
with
you
and
that
that
we
can
work
together
with
other
companies
in
promoting
the
internet
network
use
cases,
but
I
think
that,
right
now,
if
there
are,
as
you
said,
there
are
two
to
three
different
use
cases
listed
in
the
working
group,
but
I
think
we
we
are
studying
from
different
dimensions.
You
know
the
different.
F
D
Okay,
thank
you.
So
a
couple
of
things
we
have
a
comment
in
the
chat
from
charles
that
mentioned
that
there
is
an
existing
draft
in
mef
metro
intellect
forum
that
may
be
of
interest
to
the
offers.
So
there
is
a
link
provided
to
that.
I
will
put
that
into
the
the
notes
of
the
meeting.
So
if
authors
want
to
refer
to
that
later
on.
D
And
charles,
if
you
want
to
make
a
live
comment,
don't
hesitate.
We
still
have
a
couple
of
minutes
to
help
our
offers
here.
The
second
thing
I
would
like
to
mention
is
that
I
think
the
first
step
for
the
kind
of
ibn
use
case
draft,
like
network
measurement
intent
is,
would
be
nice
also,
maybe
to
connect
explicitly
to
the
classification
you
see,
because
we
have
seen
the
path
that
it's
useful
to
try
to
position.
D
I
mean
we
use
the
the
work
learning
classification
and
if
we
have
across
different
use
cases
common
references
to
the
classification
work.
This
could
also
be
very
useful.
I
think,
as
an
output
of
the
group.
G
G
The
sd-wan
as
the
use
case
and
then
from
the
point
of
view
of
applications
that
are
identified
by
the
networks
measuring
performance
of
them
in
the
context
of
sd-wan,
so
a
bit
more
specific,
perhaps
than
what
you
have
here
with
the
intent-based
networking,
but
I
think
a
lot
of
intent
based
networking
does
focus
on
the
applications
on
the
network,
and
so
that's
one
way,
sort
of
of
look
of
interpreting
intent
or
capturing
intent.
So
that's
why
I
mentioned
this
draft
may
be
of
interest.
G
I
think
it
it
could
be
worth
looking
at
the
era.
What's
the
link
to
the
draft?
That's,
it
is
a
draft.
It
hasn't
yet
been
published
as
a
mf
standard,
but
it
will
be
relatively
soon
so
anyways
now
would
still
be
a
good
time
to
have
a
look
at
that
and
provide
some
feedback
based
on
it.
D
Thank
you,
charles
useful,
always
to
have
a
different
perspective.
I
think
authors
will
consider
that
in
the
discussion
or
revisions,
thank
you
and
just
to
mention.
Also
robert
robert
wilson
also
mentioned
a
useful
reference
that
could
be
added
to
the
document
which
is
on
the
network
telemetry
framework
from
ops
area
working
group.
D
D
Okay,
so
just
because
we
still
get
some
inputs
on
the
chat,
so
I
invite
authors
to
look
at
the
chat,
but
we
also
mention
something
and
everything
will
be
posted
anyway
in
the
minutes,
and
now
we
switch
to
the
other
presentation.
Thank
you.
H
A
H
H
So,
regarding
the
draft
status,
there
was
a
first
version
presented
already
in
itf
108
in
july
the
last
year.
This
version
presents
an
update.
I
will
comment
on
on
some
of
the
points
later
on
and
also
just
to
highlight
the
fact
that
paolo
lucente
is
also
added.
As
author
of
the
of
the
draft
we
power
on
myself,
we
have
already
identified
new
content
that
is
expected
to
be
included
as
a
new
version
for
itf
113
and
today,
essentially,
we
will
comment
on
what
is
present
now
in
the
in
the
draft.
H
So
next,
please,
okay,
regarding
the
motivation
and
the
objectives
of
of
this
work.
So
essentially
what
we
are
perceiving
or
what
we
are
seeing
is
that
the
present
mode
of
operation
of
in
the
interconnection
world,
essentially
a
this
interconnection,
is
only
conceived
as
a
pure
interchange
of
ip
traffic.
H
So,
no
more
than
that,
these
environments
of
interconnection
and
for
peering
and
transit
essentially
require
long
interactions
between
parties
is
typically
static
in
the
fact
that
you
establish
a
session
and
essentially
maintain
that
session
along
the
lifetime
of
the
of
the
agreement
for
interchange,
ip
traffic
and
well
is
somehow
based
on
a
manual
operation
at
all.
H
So
the
the
situation
is
changing
in
in
the
fact
that
the
operators
are
now
starting
to
deploy
computing
capabilities,
and
this
is
coming
with
the
advent
of
a
cloud
and
edge
computing
and
also
with
the
event
of
a
virtualization.
So
a
existing
model
for
interconnection
is
limited
to
that
interchange
of
ip
traffic.
So
we
were
wondering
how
to
leverage
on
intents
for
for
evolving
such
a
way
of
interconnecting
and
taking
into
account
those
other
capabilities
that
allow
two
requests
for
more
advanced
services.
H
And
here
we
are
considering
services
like
the
ones
that
could
permit
different
stakeholders
to
deploy
a
virtualized
services,
for
instance,
in
different
domains,
or
to
leverage
in
compute
capabilities
of
any
other
domain.
In
order
to
instantiate
later
on
and
whatever
builder
network
function
or
container
or
so.
No
and
clearly
here
an
example
of
that
could
be
to
instantiate
a
service
function
chain
in
a
multi-domain
environment,
so
how
we
are
trying
to
address
and
to
explore
such
kind
of
scenarios
leveraging
on
these
novel
capabilities
that
exceed
just
the
pure
interchange
of
ip
traffic.
So
next,
please.
H
So,
regarding
the
the
summary
of
the
draft
contents,
so
again,
the
target
is
to
leverage
on
the
and
intent
mechanisms
in
order
to
enrich
the
interconnection
request,
so
going
further
of
that
pure
traffic
interchange
and
the
scenarios
of
applicability
we
mentioned
before
the
one
of
the
the
typical
one,
let's
say
between
operators,
but
also
is
clear,
probably
important
to
highlight
the
situation
or
that
this
is
also
being
fostered
by
the
advent
of
ie
networks
where
the
there
is
this
idea
of
interconnecting
private
networks
to
public
networks,
and
so
on
so
far,
so
it's
not
just
I
mean
moving
from
pure
appearing
and
transit
to
something
else,
involving
further
capabilities,
the
compute
edge
and
so,
but
also
the
different
actors
that
are
not,
let's
say
in
nate.
H
That
probably
today
are
not
so
much
present
in
in
the
in
the
real
deployments.
Clearly,
multi-domain
network
as
a
service
request
or
multi-domain
network
virtualization
would
be
also
target
of
of
this
kind
of
interconnection.
Intents.
Regarding
the
modes
of
usage
of
these
interconnection
intents
we
would
have
the
baseline
one.
That
would
be
the
ip
traffic
interconnection,
so
clean
transit,
but
also
we
we
air
force.
We
are
foreseen.
H
H
Other
usages,
like
a
virtual
network
function
as
a
service
so,
and
here
and
I
can
an
example-
could
be,
for
instance,
the
deployment
of
packet
wall
capabilities
for
virtual
network
operators,
and
maybe
here
we
are
just
suggesting
for
what
other
tools
in
in
idea
could
be
used.
H
Maybe
this
service
function,
our
topology
model,
could
be
used
for
that
also
to
a
use
these
intents
for
interconnecting
computing
capabilities,
for
instance,
hitting
functions
or
containers
on
top
of
them
and
maybe
leveraging
also
in
some
other
drafts
that
have
been
presented
in
in
other
working
groups
in
atf
or
any
combination
of
the
all
these
usages
I
have
mentioned
before
so
essentially
to
make
and
broader
scope
of
different
interconnections
that
are
now
or
are
expected
to
be
allowed
now,
with
these
new
capabilities
in
in
operator
site.
H
So
the
benefits
of
of
this
intent
would
be
to
establish
a
common
and
a
normalized
way
between
service
providers,
but
also
private
networks,
in
order
to
automate
the
interconnection
and
look
for
a
simple
way
of
expressing
and
reaching
their
connection
request
again
going
further
than
the
pure
iep
traffic
exchange.
So
next,
please.
H
So,
as
next
steps,
we
will
work
developing
in
this
intern,
based
capabilities
for
interconnection
aspects,
trying
to
explore
more
details
on
on
that
and
some
for
the
details
not
present
here
in
the
slides,
like
the
policies
or
maybe
also
extend
to
different
kind
of
technologies
optics
and
so
also
to
request
comments
and
inputs
for
from
the
audience
from
the
research
group
audience
in
order
to
enrich
the
the
document
for
sure
and
again
also,
we
would
like
to
position
this
draft
as
one
of
the
potential
intent
use
cases
that
are
being
developed
in
this
research
group
and
well
here
probably
the
same
comment
as
as
before
and
the
presentation
before.
D
Do
you
do
you
see
the
main
I
will
say
segments
like
being
into
inter
carrier
exchange
or
it's
more
like
between
between
different
operators
at
any
point
or
between
different
actors
like
vertical
type
of
factors
and
operators?
I
mean
for
this
interconnection.
What
kind
of
environment
do
you
see.
H
The
the
initial,
the
starting
idea,
let's
say,
come
from
the
interoperator
interconnection,
but
we
we
think
that
can
be
extended
to
any
other
scenarios
as
soon
as
we
could
abstract.
Let's
say
the
way
of
requesting
this
intent.
This
capability,
sorry
so
probably,
could
be
applicable
for
all
of
them,
although
at
least
this
is
my
my
deal,
but
the
starting
point
was
for
sure
interoperator
interconnection,
yes,.
D
It
was
a
recent
publication.
I
don't
remember
if
it
was
a
cnsm
or
another
conference
that
was
addressing
inter
carry
exchange,
automation
and
I
think
they
were
referring
to
some
kind
of
level
of
abstraction.
So
this
may
be
a
reference,
a
useful
reference
for
this.
D
Anything
else,
no
other
questions,
nothing
in
the
chat.
So
thank
you,
luis.
We
proceed
with
the
agenda
and
hopefully,
let's
see
if
we
can
also
come
up
with
something
in
common
for
these
different
use
case
tracks
around
the
ibm.
Thank
you.
A
I
Okay,
so
good
morning
to
everyone,
I'm
cara
graceli
a
phd
student
at
the
university
of
bologna.
During
this
session.
I'm
going
to
show
you
and
present
the
activity
me
and
my
colleagues
have
carried
out
on
digital
twins
for
industrial
iot.
Networking.
I
So
so
what
we
have
implemented
is
a
digital
twin
that
we
use
for
as
a
cyber
range
environment
for
testing.
For
example,
we
carried
out
some
simulation
of
attacks
and
so
on.
Next,
please.
I
In
this
work,
we
have
exploited
our
expertise
on
virtualization
technologies
and
paradigms
such
as
nfv
network
function,
virtualization
and
software-defined
networking,
and
in
particular,
we
followed
the
etsy
management
and
orchestration
functionalities
and
framework
the
etsy
framework
covers
the,
for
example,
allows
us
to
design
the
architecture
we
tag
level
description
of
the
architecture
and
then
to
describe
each
building
block
each
element.
We
want
in
our
digital,
twin
and
the
topology,
so
how
we
combine
all
the
building
blocks
together,
we
can
interconnect
them
with
templates
called
descriptors.
I
We
will,
I
will
show
you
later
in
the
next
slides
the
types
of
descriptors
and
finally,
the
etsy
management
and
orchestration
allows
also
to
have
an
automated
deployment.
So
with
these
we
could
have
an
automated
deployment
and
configuration
of
all
the
components
of
the
digital
twin.
Next,
please,
this
is
in
this
slide.
You
can
see
a
picture
of
the
reference
architecture.
This
is
a
general
figure.
I
We
consider
some
of
these
elements
for
the
actual
deployment
you
can
see
that
we
can
have
one
or
more
production
plants
with
iot
connectivity,
one
or
more
gateways.
It
depends
on
the
protocols
we
are
using
in
the
in
this
environment
and
then
we
have
other
elements
that
are
not
related
to
the
real
production
the
to
the
production
plans,
but
are
more
operative
and
and
for
the
enterprise
part
of
the
network.
I
So
we
have,
we
can
have
vpn
servers,
we
can
nodes
firewall,
an
intrusion,
detection
system
or
production
system,
big
data
analytics
and
so
on,
and
in
our
case,
all
of
these
components
are
to
be
replicated
into
a
virtualized
replicas
with
virtual
machines
or
containers.
Both
approaches
are
possible.
We
selected
the
day,
one
based
on
virtual
machine
and
we
exploited
the
network
function
virtualization
to
emulate
the
network,
architectures
all
the
functions
that
are
called
virtualized
network
functions
in
the
term
of
nfv
and
the
application
into
wheelchair
replicas
too.
I
So
since
we
use
the
nfb
management
and
orchestration
framework,
actually
the
digital
twin
life
cycle
is
quite
similar
to
the
one
of
and
after
slice
we
we
have
different
phases
of
the
life
cycle.
One
is
a
pre-phase
is
not
actually
a
part.
Real
part
of
the
life
cycle
is
the
preparation
phase,
during
which
we
can
design
the
architecture.
I
We
must
design
the
architecture
of
the
digital
twin
with
descriptors
once
the
descriptors
are
created,
we
onboard
them
on
the
orchestration
platform,
and
here
we
can
have
also
some
additional
network
environment
setup.
For
example,
we
needed
additional
networks
to
have
access
to
the
external
world,
for
example,
and
manage
for
management
functionalities.
I
I
I
Here
you
can
see
the
software
we
are
using
and
that
we
have
chosen
for
the
this
project
open
source
mono-
that
is,
the
management
and
orchestration
platform.
I
Here
you
can
see
the
picture
that
sum
up
so
some
of
the
concepts
I
I've
talked
about
in
the
in
the
previous
slides.
We
started
deployment
from
eye
level
descriptors,
so
you
can
see
two
types
of
descriptors
here.
I
The
vienna
descriptions
are
the
one
for
each
building
block
that
this
type
of
descriptor
tells
how
which
are
what
are
the
characteristics
of
each
bnf,
so
each
element
of
the
digital
twin,
and
then
we
have
an
upper
level
type
of
descriptor
that
are
the
network
service
descriptors
that
describe
the
topology,
how
we
want
to
combine
all
the
vnfs.
So
we
want
to
combine
all
the
elements
in
our
digital
twin
and
then
we
need
also
to
upload.
I
I
Sorry
we
have
the
pro
okay
just
be
enough.
We
can
have
a
big
packet
inspector
intrusion
prevention
system
or
detection
system
firewall
traffic
shaper
vpn
server.
Some
of
them
were
already
in
the
reference
architecture
and
as
virtual
application,
we
can
have
different
type
of
also
protocols,
application
with
using
different
protocols
of
the
traditional
industrial
protocols
like
canopen,
mqtt,
modbus
and
lora.
I
We
implemented
the
elements
that
you
see
outlining
in
dashed
lines
in
this
figure.
We
have,
as
I
told
you
before,
we
needed
some
two
networks
for
to
be
present
before
the
instantiation
of
the
digital
twin.
These
are
the
blue
one
for
the
external
connectivity
and
the
green
one
for
the
management
functionalities.
I
With
the
instance.
We
create
also
a
management
console
for
this
purpose,
and
then
we
have
a
firewall,
the
other
one
are
all
bnfs.
We
have
a
firewall,
an
intrusion
detection
system
based
on
suricata
and
an
internal
router
plus
for
the
more
related
industrial
part.
We
have
a
virtual
machine
with
that
modbus
traffic
generator
for
to
simulate
the
possible
possible
cyber
attacks,
and
so
on.
We
used
a
kali
linux
machine.
You
can
see
also
this
one
in
the
figure.
I
Here
we
have
just
only
a
slide
that
makes
a
reference
to
the
one
of
the
model
usually
used
to
layer,
the
infrastructure
in
the
industrial
and
manufacturing
environment.
It
is
only
to
say
that
we
we
have
a
partner
related
to
the
field
and
production
plans,
the
one
for
operation
and
the
other
one
for
enterprise
which
actually
are
in
our
deployment
are
overlapped.
I
I
I
I
We
need,
of
course,
an
account
and
first
of
all
we
need
to.
I
just
I
will
stop
for
a
moment
with
the
video.
We
need
to
define
the
logical
connectivity
between
the
western
platform
and
open
stack
platform,
so
this
can
be
done
with
the
beam
accounts.
The
so-called
beam
account.
So
basically,
we
only
specify
the
type
of
the
beam,
the
tenants,
for
example
the
password
and
the
user
we
want
to
use
and,
in
particular
how
to
reach
the
beam.
I
I
I
This
is
the
list,
and
here
you
have
an
example
of
one
of
them.
In
the
descriptor
we
define
the
id
that
is
used
to
reference.
The
descriptor,
then
we
have
the
most
important
part-
is
the
video
the
deployment
unit
which
specified
the,
for
example,
the
cloud
image
we
want
to
use
for
the
vm
additional
configuration
files.
I
I
In
the
network
service,
we
put
together
the
components
we
want
in
the
in
the
final
deployment,
so
we
have
the
list
of
debian
apps
and
the
we
define
the
virtual
networks
link
the
descriptor
with
the
link
descriptor
that
are
that
corresponds
to
the
network
that
interconnect
the
bnf.
So
this
is
an
example,
and
we
can
define,
of
course,
also,
for
example,
ip
addresses
if
needed
or
other
parameters,
once
the
descriptor
are
ready,
we
need
to
upload
them
on
the
orchestration
platform
in
the
catalog.
I
I
And
finally,
we
can
launch
the
instance
now
that
we
have
the
descriptors,
so
we
need
to
define
we
need
to
select
the
descriptor
we
want
to
create
to
deploy.
I
We
define
a
name,
a
description
and,
most
importantly,
we
need
to
select
a
beam
for
the
deployment,
and
so
here
you
can
see
what
is
going
on
stack
during
the
deployment.
The
blocks
are
the
vms.
I
I
Here
we
are
performing
a
massive
reading
of
the
modbus
registers,
for
example,
from
the
kali
machine
to
the
other
node,
and
this
is
our
an
example
how
we
can
configure
the
ids,
the
intrusion
detection
system
to
detect
this
type
of
behavior.
So
on
the
right
you
can
see
in
the
right
part.
You
can
see
the
terminal
with
the
output
from
the
intrusion
detection
system,
and
this
is
older.
A
Okay,
so
looks
looks
not
thank
you
again,
kiara
for
the
presentation
and
neither
is
more
comment
than
mailing.
This
referral
to
you.
D
Actually,
I
I
do
have
a
question,
because
if
you,
unless
correctly,
you
have
set
up
different
virtual
environments,
but
how
do
you
relate
that
to
the
aspect
of
a
digital
tweet
and
having
the
opportunity
to
see
link
between
what's
actually
happening
on
the,
for
instance,
the
the
digital
plant,
the
factory,
and
I
mean
what
you
monitor
from
this
environment
and
how
you
can
run,
for
instance,
simulation
in
the
digital
tween
and
retract
on
the
on
the
actual
environment?
D
I
Yeah,
so
the
question
is
about
how
we
can
actually
interface
to
the
real
network
right.
This
is
actually
the
focus
of
some
of
the
other
research
groups
and
the
other
partners
of
the
project.
I
Some
of
them
provided
some
software
to
interface,
for
example,
with
the
actual
production
plans
and
industrial
networks
with
can
open,
for
example,
android
one
technologies,
and
we
once
at
the
inner
previous
deployment,
not
the
one
in
the
video
we
actually
included.
This
part
are
also
in
our
digital
twin,
in
particular
the
one
related
to
can
open.
I
There
is
also
in
the
I
think
in
one
of
the
figure
I've
shown
you
in
the
figure
you
can
see
also
a
node
with
the
can
open
protocol,
and
then
actually,
we
don't
have
also
a
real-time
monitoring,
because,
due
to
the
constraints
we
didn't
want,
we
wanted
something
completely
separated
from
the
real
twin
I
yeah
here
you
can
see.
Also.
I
can
open.
D
Do
you
have
other
typical
scenarios
or
types
of
functionality
that
you
you
simulate
into
your
digital
twin,
or
is
it
working
progress
just
a
little
bit
the
breadth,
the
number
of
it's
a
functionality
you
are
able
to
simulate
or
what
kind
of
functionality
you
simulate
for
this
kind
of
environment.
I
We
actually
had
some
industrial
partners
in
the
project
and
we
are
getting
now
feedbacks
from
them,
so
we
are
waiting
for
them
in
order
to
maybe
try
different
types
of
attacks
and
scenarios,
it
depends
on
what
they
are
interesting
actually,
so
we
are
waiting
for
them
and
we
are
working
now
on
this
part,
because
this
is
the
last
part
of
the
project.
So
we
are
waiting
for
feedbacks
in
order
to
proceed
with
the
other
testing.
J
I
You
can
hear
me
right.
This
depends
actually
on
the
available
hardware
and
resources
on
the
cloud
platform.
I
would
say
yes.
I
D
K
Okay,
hello,
everyone-
this
is
from
china
mobile.
My
printation
is
a
and
efficient
data
collection,
medicine
for
digital
network.
A
related
draft
can
be
found
on
the
website.
K
K
However,
the
existing
method
collects
the
full
amount
of
data
from
the
physical
network
for
monthly,
and
it
does
consider
the
problems
such
as
timeline,
efficient
storage,
resource
log,
computational
efficiency
at
a
restaurant,
bad
waste
resource
caused
by
data
transmission.
Therefore,
it
is
important
to
propose
a
lightweight
at
the
efficient
data,
cluster,
aggregation
and
correlation
method.
Next,
please.
K
In
order
to
solve
those
problems,
this
draft
introduced
and
efficient
and
lightweight
data
collection,
aggregation
and
correlation
method.
Three
network
sense
inscri,
is
instruction
to
a
physical
network
to
collect
data
or
demand,
and
then
the
physical
network
complexity
structure
such
as
laundry
representation,
a
time
machine
streaming,
element
of
physical
network,
complex
data,
aggregation
and
correlation.
K
K
Next,
I
will
introduce
the
details
of
draft.
The
main
modules
involved
are
showing
figure.
The
digital
network
consists
of
physical
network
at
twin
network.
K
K
This
page
will
introduce
data
storage,
central
physical
network;
it
has
it,
it
has
two
functions.
Firstly,
it
can
store
dates
such
as
performance
indicators,
operational
status,
blocks,
traffickers,
scheduling
and
business
requirements,
and
so
on.
K
The
tree
network
includes
the
instruction
management
center
and
data
storage
set
instruction.
Management
has
two
functions,
one
the
other
head,
one
one
hand
it
may
use
to
match
the
restriction
or
telemetry
streaming
element
in
the
physical
network.
K
On
the
other
hand,
is
it
is
used
to
adaptively
configure
data
collection
instruction
according
to
the
clinch
requirements
of
the
data
storage
set
in
the
tool
network
at
the
search
for
iep
address
to
set
the
instruction
among
these
registration
information
include
various
key
information
such
as
the
ip
address
of
the
data
storage
data
in
the
physical
network,
data
type
and
various
index
names
of
the
data,
this
source
name
and
data
size
and
so
on.
K
And
the
function,
the
data
storage
center
into
network
is
mainly
used
to
store
the
effective
information
after
data
processing
and
the
knowledge
representation
returned
by
the
telemetry
streaming
element
in
the
physical
network.
K
K
The
registration
information
includes
the
ip
address
of
the
data
storage
center
and
the
data
type
and
data
source
of
the
data
sites
and
so
step.
Two,
the
data
storage
set
intel
network
sets
the
data
collection
request
to
the
instruction
management
setup
step
3.
According
to
the
date
class
request,
the
instruction
management
center
intelligently
creates
a
registration
in
information
for
addressing
and
configure
the
data
processing
structure.
K
Okay
step:
four:
the
instruction
management
center
in
the
train
network
sets
the
course
corresponding
structure.
According
to
the
query,
results
to
the
telemetry
streaming
element
in
the
physical
network,
step
5
out
after
restoring
the
instruction
the
telemetry
streaming
element
in
the
physical
network
will
pass
and
execute
them.
According
to
the
instruction
at
the
query,
the
local,
the
old
data
stored,
the
query
function
can
be
performed
by
the
complex
inventor
processing
injury,
which
resource
or
telemetry
adder
processes.
It
with
all
queries
provided
step.
K
K
Okay,
stable
seven,
a
data
storage
set
of
physical
network
performance,
knowledge,
representation
or
local
data,
for
example,
in
a
resource
description
framework
form
also
sets
rounded
to
a
telemetry
streaming
element
for
knowledge,
representation,
step,
8
data
storage,
set
of
physical
network
push
state
or
knowledge
to
telemetry
streaming
element
in
step,
9,
telemetry,
streaming
element
aggregates
at
the
correlates.
K
K
K
K
A
Thank
you
for
the
for
the
presentation.
Is
there
any
comments
or
questions.
D
I
have
a
question,
but
I
like
to
hear
from
the
floor
first,
I
have
also
one
okay,
so
thanks
for
for
this
work,
it's
interesting
their
direction.
In
fact,
my
question
is
is
also
a
comment.
I
think
it's
an
important
question
in
digital
twin
environment
to
consider
see
the
performance
trade-off,
to
get
accurate
data
about
the
real
environment
and
to
balance
that,
with
the
cost
of
collecting
this
data,
okay,
so
the
fidelity
from
the
need.
So
two
things.
D
D
We
really
need
to
be
very
close
to
the
real,
because
the
idea
is
to
be
able
to
replicate
and
simulate
so
that
we
really
see
how
the
real
one
will
behave
and
be
able
to
use
that
to
make
useful
simulation
and
scenarios
to
be
able
to
help
or
to
support
the
real
environment.
So
I
think
there
is
this
question
of
how
far
we
can
get
into
saving
on
the
data
collection
and
data
representation.
D
Without
compromising
this
twin
approach,
so
this
is
more
common.
The
second
thing-
and
it's
a
bit
my
question
related
to
the
research
that
we
can
do
in
this
in
this
work,
because
I
think
you
touch
a
lot
about
the
methods
to
collect
data,
to
aggregate
data
and
to
convey
data
between
the
the
real
environment
and
the
tree.
D
So
I
think
this
is
very
relevant
as
a
mechanism,
but
I
was
thinking
about
more
the
the
let's
say,
the
algorithm
or
the
methods,
in
fact
to
to
understand
what
kind
of
data
we
need,
because
I
don't
think
there
is
a
single
approach
that
could
fit
every
needs
in
a
digital
tweet.
Depending
on
the
simulation
you
want
to
run.
D
You
may
need
a
very
coarse
data
on
a
let's
say
intermediate
basis
or
you
may
need
to
have
very
high
fidelity
data
on
different
set
of
metrics,
and
even
sometimes
you
need
to
anticipate
on
the
collection
of
some
of
some
new
measurements
in
order
to
have
them
ready
when
you
want
to
run
the
simulation
in
between.
So
my
question
is:
do
you
have
a
do
you
play
or
do
you
have
knowledge
about
different
methods
or
different
research
being
done
into
processing,
how
we
can
in
fact
evaluate
a
bit
in
advance
or
very
dynamically?
D
What
is
the
need
expressed
on
the
tree
in
terms
of
data
collection
and
how
we
can
in
fact
have
methods
that
can
optimize,
so
I'm
looking,
especially
for
optimization
methods
about
what
to
collect
when
to
collect,
etc,
so
that
we
get
a
very
good
interplay
between
the
real
network
and
the
connection,
the
measurements
we
do
on
that
and
the
need
of
the
of
the
twin
in
terms
of
having
the
right
data
at
the
right
time
to
make
the
right
simulation.
K
D
Yeah,
okay,
I
will
we
can
discuss
offline
about
my
input
and
suggestions.
Okay,
thank
you.
We
have
also
shane
in
the
landline.
J
For
this
chapter-
and
I
I
I
I
really
value
your
comments
and
surely
we
data
collection
of
data,
aggregation
of
that
data,
correlation
or
data
or
all
data
related
domains
are
are
really
a
very
big
big
facts
for
building
digital
twin
network,
and
currently
the
scope
of
this
structure
is
only
for
some
case
of
the
data
collection
in
in
that,
in
those
cases,
there
is
chance
to
aggregate
all
correlation
or
compromise
some
data
raw
data
to
to
get
more
efficiency
and
also
can
achieve
a
high
fidelity
to
build
the
twin.
J
So
maybe
we
we
next
version
we
need
to
narrow
or
clarify
the
scope
of
our
of
this
discharge.
J
D
Good,
thank
you,
yeah.
We
definitely
need
to
work
together
to
progress
in
this
work.
We
can
discuss
offline.
Of
course.
Thank
you.
Thank
you.
Okay,
just
for
the
sake,
mohammed
also,
I
mentioned
a
comment
in
the
chat.
I
think
we
are
slightly
on
time
for
the
agenda.
Let
me
just
check
so
mohammed.
Put
on
my
first
comment.
D
I
don't
think
there
is
an
answer
for
the
first
question
raised
by
laurel,
as
it
depends
on
the
application
making
using
the
making
use
of
the
digital
twin
if
your
app
is
for
training
purposes
or
for
running
what-if
scenarios,
routine
changes,
for
example,
you
don't
need
a
full,
real-time,
synchronization
or
sync
between
them.
Okay,
so
these
were
comments
of
from
mohamed
in
the
chat,
and
we
will
reflect
that
in
a
minute.
J
J
Actually,
this
is
the
fifth
time
to
present
the
drops
in
this
research
group,
as
we
have
presented
since
ietf108
the
object
tables
of
this
chapter
encodes
to
define
digital
tree
in
network
networking
field,
to
sketch
a
reference
architecture
for
digital
network
and
to
define
to
identify
key
values
of
digital
team
networks,
as
well
as
research
challenges,
and
also
to
document
typical
application
scenarios.
J
Here
I
I
shall
see,
give
many
thanks
to
expert
running
pedro
and
or
others
as
main
changes
in
dash
five
distortion.
We
we
mainly
clarified
the
difference
between
digital
twin
network
platforms
and
the
legacy
network
management
systems,
refined
the
description
of
the
reference
architecture
and
added
references
to
digital
twin
network
research.
J
J
This
table
in
this
slide
and
next
slide
lists
the
major
comments
and
our
action
taken
here.
I
will
not
describe
them
one
by
one
in
details
next
slide.
Please.
J
J
J
The
main
difference
compared
to
other
network
management
system
is
the
use
of
interactive
virtual
rail
mapping
to
build
closed-loop
network
automation
through
the
real-time,
deep
interaction
between
the
physical
network
and
these
two
networks.
The
digital
twin
network
platform
might
help
the
network
designers
to
achieve
more
simplification
automatically
resilient
and
for
life
cycle
operations
and
maintenance.
J
Digital
twin
network
allows
for
integrating
ai
based
mechanics
mechanism,
emulating
the
behavior
of
network
sections,
not
under
consideration
but
relevant
for
the
considered
section,
also
for
considering
several
parallel
options
or
causes
of
action
and
for
incorporating
real-time
monitoring
data
from
physical
with
other
real
or
virtual
data
sources,
and
also
for
repeating
and
reproducing
specific
situations
under
controlled
variations.
J
This
slide
shows
the
collaboration
on
the
reference
architecture
in
the
three
layer
architecture
we
provided,
the
top
the
lowest
layer
is
physical,
physical
network
and
top
layer
is
a
network
application
and
the
intermediate
intermediate
layer
is
the
network
digital
twin.
This
layer
includes
three
key
subsystems:
data
repository
service
mapping
models
and
the
digital
entity
management.
J
J
That's
like
this,
so
about
two
slides
away.
Well,
I
just
described
the
major
con
updates
from
that
from
last
meeting.
So
for
next
step.
We
believe
that
current
staff
is
quite
steep
after
five
times
of
update,
so
we
are
requesting
document
document
adoption
by
mrg,
so
welcome
to
join
our
work,
and
any
comments
are
welcome
and
since
time
permits
I'd
like
to
share
two
more
sliding
appendix
so
next
slide.
Please
thank
you.
J
J
The
first
is
data
collection,
including
diverse,
existing
tools,
for,
for
example,
smp
netcom,
telemetry,
int,
etc,
and
in
innovative
new
tools,
for
example,
sketch
based
measurement,
and
also
that
the
tools
can
can
include
semantic
and
aggregation
mechanisms
for
data
integration
and
action
translation.
J
This
is
first.
The
second
technology
is
data
storage
and
the
services
unified.
The
data
repository
to
effectively
effectively
store
large-scale
and
heterogeneous
network
data
and
to
provide
these
services,
including
fast
search,
batch
data
handling,
conflict
avoidance,
data
access
interfaces,
etc,
etc,
and
the
third
is
date
modeling
for
small
scale
network
network
simulation
tool,
for
example,
iso
i3
gms3
can
be
an
option
for
large
scale
network.
J
J
The
fourth
technology
is
visualization,
including
displaying
the
network
topology
option
operational
status
in
multiple
dimensions
and
finding
grand
granularity
and
using
interactive
visualizing
to
help
users
better
understand
to
use
and
explore
the
network
interfaces
and
protocols.
J
J
J
D
D
D
Because
we
have
to
go
also
through
the
request
for
the
call
for
adoption
shang.
Are
you
back
online.
J
Oh
sorry,
sorry,
my
my
network
is
broken.
Sorry,
I
used
this
light.
We
just
want
to
shoot
more
samples
on
applying
digital
to
network
in
academia.
There
are
several
researchers
in
computing
biology,
networking
and
also
in
industrial
iot,
and
this
sharia
gave
us
another
example
and
the
industrial
industry
more
companies
are
investing
solutions
of
digital
networking,
including
aeronautics
network,
rva
and
spider.
D
Thank
you
shane.
We
have
kieran
online,
please
kieran.
D
L
Okay,
I
just
have
a
quick
clarification
question
that
when
we
define
digital
twin
network
is
it
are
there
multiple
instances
that
we
are
abstracting
from
a
physical
network
or
is
it
a
representation
of
the
entire
physical
network
in
one
shot.
J
Okay;
okay,
in
our
opinion,
that
when
we
set
the
digital
twin
network,
it
should
contain
both
twin
network
and
the
physical
network
is
our
opinion
and
also
the
the
point
in
this
job.
D
Okay,
thank
you.
So
you
you
express
request
for
a
call
for
adoption
as
a
research
group
document
and
as
said
also
for
the
previous
draft,
we
follow
a
let's
see
a
group
procedures
to
validate
some
criteria,
to
see
a
bit
the
feeling
of
the
research
group
on
adopting
this
as
a
research
group
document.
D
So
usually
we
take
it
offline
on
the
mailing
list
that
we
have
time
to
collect
expression
for
from
all
the
participants.
I
just
share
my
viewer
more
as
a
participant
right
now.
I
think
it's
interesting
to
see
an
architecture
document
for
the
trump
digital
twin.
D
As
you
say,
the
document
has
matured
for
for
some
time
now.
So
I
think
it's
a
good
time
to
have
this
call
for
for
research
group
adoption,
but
I
I
like
to
see
through
the
answer
of
the
different
criteria
where
we
want
to
go
with
this
architectural
document
and
if,
let's
say
it's,
comprehensive
and
mature
enough
in
in
the
coverage
for
as
an
architecture
for
natural
digital
twin
and
if
we
can
lay
down
based
on
this
architectural
document,
a
set
of,
let's
say,
research,
questions
or
challenges
that
could
feed.
D
In
fact,
research
into
the
research
group
on
these
questions,
because
we
have
seen
that
there
are
also
other
drafts,
other
projects
working
on
that.
So
I
think
it's
an
interesting
trend
and
I
think
we
need
also
maybe
to
articulate
a
bit
some
of
the
research
directions
that
we
could
explore
in
the
research
group,
and
I
think
this
architectural
document
could
be
a
foundation
to
identify
or
highlight
such
directions.
So
this
is
for
me
as
a
participant,
some
feedback.
I
would
like
to
provide
for
a
protocol
for
for
myself
adoption.
D
I
will
put
that
in
a
into
the
mailing
list,
but
I
will
leave
also
maybe
to
steer
this
discussion.
A
Thank
you
lauren.
Thank
you.
I
think
yeah.
This
draft
has
been
also
presented
multiple
times.
I
think
it
was
very
interesting.
The
the
last
live
presentation
with
the
different,
let's
say,
let's
say
enabler
that
you
need
for
digital
networks,
and
I
think
it
can
be
really
useful
for
actually
articulating
this
discussion
of
what
what
we
need
actually
for
having
this
having
these
digital
twins
so
yeah,
I
think
it.
A
It
could
be
a
very
good
good
point
also
to
have
this,
as
already
serving
as
a
purpose
for
the
discussions
and
maybe
more
formally
now
for
the
for
the
confirmation.
So
as
long
as
so
there
is
a
kind
of
let's
say
it's
not
formal,
but
it's
something
that
we
define
in
a
number
g
as
a
procedure
for
the
corporate
reduction,
so
as
as
we
as
we
already
did
in
the
for
previous
documents.
A
So
if
I
think
the
first
thing
as
usual
is
as
to
to
have
the
different
feedback
and
then
the
mailing
list
and
update
the
draft
and
then,
if,
after
after
this,
maybe
the
next
round
of
discussion
you
might
have
with
maybe
use
laura
or
his
other
participants
that
provide
you
some
some
some
discussion,
some
comments.
After
this
meeting
we
can.
A
We
can
then,
of
course,
send
the
call
for
adoption
of
the
mailing
list,
and
you
know
it
will
be
done
somehow
offline
on
the
mailing
list
to
have
to
ever
do
the
opinion
of
more
people,
because
most
of
the
times
have
more
people
that
express
our
opinion.
In
that
case,
regarding
the
quality
of
the
document,
the
scope,
all
these
cases
can
support
the
group
and
so
on
so
yeah.
So
as
we
say,
yeah
there
is
a
there
is
procedure.
A
Okay,
there
are
some,
I
think,
quite
a
lot
of
commenters
in
the
chat.
Maybe
we
can
copy
paste
and
in
the
minutes
for
further
discussion,
because
I
think
we
have
to
move
to
the
next
explanation.
J
B
B
B
Thank
you,
okay,
so,
in
general,
from
from
the
context
of
this
draft
and
this
framework,
so
we
we
know
that
automating,
the
network
management
with
ai
enables
computer
networks
to
any
processor
to
go,
and
we
are
currently
dealing
with
the
exponential
growth
thinking
for
supporting
iot
and
this
last
detangling
functions
that
we
are
seeing.
So
there
are
a
lot
of
hurdles
that
must
be
resolved
in
network
management
that
the
ai
can
can
do.
B
Okay,
so
similar
mechanisms
have
been
to
to
define
to
to
to
to
to
achieve
these
ai
functions,
but
they
must
interpret
in
some
way,
so
we
we
need
some
framework.
That's
the
the
key
of
this.
B
This
draft
is
to
define
this
framework
that
is
needed
to
to
ensure
the
seamless
connection
among
all
components
and
participants
in
this
network
management
like,
for
example,
we
must
define
one
ontology,
so
everybody
all
components
are
talking
about
the
same
concepts
and
things
like
that.
We
need
to
define
some
kind
of
interfaces
for
different
components,
so
the
operation
must
be
more
or
less
well
known
by
all
of
them.
And
finally,
we
need
some
method
to
assess
the
correct
handling
of
of
of
those
elements.
B
So,
for
example,
somebody
can
validate
if
one
element,
one
component
is,
is
actually
doing
what
it
must
is
supposed
to
do,
and
actually
also,
for
example,
for
some
developer
or
integrator
to
find
out
that
that
the
specification
of
components
is
actually
what
they
need.
Not
just
the
company
is
working
as
expected,
but
also
the
feed
the
company
fits
with
the
needs
of
some
software
next
slide,
please.
So
the
the
background
notice
is
quite
stretch
forward.
We
we
begin
with
the
definition
of
particular
computer
and
ecosystem.
B
Today
they
have
high
degree
of
flexibility
and
reliability
and,
as
you
all
know,
know,
sdn
and
fb.
Conceptually
are
centralized
or
only
conceptually,
but
they
can
be
distributed.
Implementation
and
they
reuse
functions
so
as
software,
so
we
we
are
now
talking
about
enormous
flexibility
and
almost
amount
of
functionality
that
we
can
manage
to
to
in
order
to
build
the
network
okay.
So
this
is
increasing
the
complexity
of
underlying
systems
and
and
they
must
perform
dynamic
adaptation
in
order
to
optimize,
be
optimized
for
for
different
workloads
and
different
scenarios.
B
There's
a
slight
piece
in
addition.
Additionally,
from
from
the
background,
we
also
have
that
I
want
to
remain.
This
is
that
we
are
all.
Everybody
is
very
familiar
with
the
concept
of
machine
learning
and
how
it's
been
it
has
been
applied
into
the
network.
But
again
I
I
I
want
to
insist
now
that
ai
is
not
machine
learning
and
there
are
a
lot
of
benefits
from
ai
beyond
machine
learning
like,
for
example,
perception,
reasoning
and
planning
that
must
be
exploited,
but
for
for
network
management,
as
you
know,
many
many
people
know
us
already.
B
I've
been
in
collaborating
with
this
group
in
in
this
sense
now,
in
order
to
to
provide
some
kind
of
of
challenges,
or
at
least
some
kind
of
topic
for
discuss
in
this
in
this
direction.
So
I
retake
this
and
again,
I
insist
that
intelligence,
for
example,
is,
is
different
from
intelligent
in
the
in
this
case,
we
are
dealing
with
the
data
gathering
and
management,
for
example,
for
building
one
digital,
twin
or
or
just
demonstrating
some
hypotheses
or
some
for
for
for
network
management.
B
We
need
a
lot
of
data
and
we
need
to
gather
that
data
and
we
need
to
do
that
with
some
systematic
methods,
so
they
are
also
needed
in
this
framework
that
I
am
discussing
today
and,
on
the
other
hand,
intelligent
emphasizes
on
how
that
data
is
actually
processed.
So
we
have
machine
learning.
We
have
other
kinds
of
of
of
interpreting
ways
of
interpreting
the
information,
so
we
do
that
integer
intelligently.
B
We
will
be
particularly
focused
on
entering
network
management
because
with
the
currently
there
are
many
focus
on
this
on
this
side,
and
many
interesting
workflows
are
being
dispatched
from
the
end-to-end
services
and
we
need
to
to
deal
with
them,
particularly
so
key
components,
and
on
this
framework
we
have,
of
course
the
data
framework
itself
is
part
of
the
is
how
to
to
to
define
the
data
that
is
being
acquired,
how
to
model
the
data
store
and
distribute
any
kind
of
data.
B
This
part,
for
example,
is
also
related
to
the
data
collection
that
has
been
previously
discussed
in,
in
terms
for
for
providing
providing
that
data
to
build
one
digital
twin.
The
data
framework
will
be
interconnected
to
this
kind
of
of
data
collection
also
functions
for
aggregating
data
in
the
sources,
and
things
like
that.
B
H
B
That
are
implementing
the
processing
of
the
data
individually
or
collectively.
We
have
many
schemes
or
many
organizations
of
this
kind
of
modules
that
are
sharing
information
and
that
can
do
individual
processing
or
collaborative
processing
in
this.
In
this
sense,
we
have
also
end-to-end
or
global
intelligent
tasks
that
must
be
addressed,
so
these
a
modules
will
collaborate
in
order
to
to
achieve
this,
this
tax,
okay
and
finally,
we
have
the
what
we
define
the
hub.
B
We
will
be
we'll
receive
all
the
data,
knowledge
and
and
all
decisions
etc,
and
will
be
shared
again
with
all
modules,
so
every
all
information
will
be
shared
in
this
hub,
so
all
modules
can
access
the
information
previous,
for
example,
previous
reports,
previous
notifications,
previous
data,
current
data,
etc,
and
also
after
one
module
has
finished.
Its
work
will
provide
this
information
to
this
hub.
This
has
been
this
not
just
the
hub.
This.
B
All
that
process
operations
and
processes
must
be,
let's
say,
organized
no,
by
by
this,
this
af
framework.
That's
a
slide,
please,
oh
as
an
overview
of
this
operation,
so
we
we
will
first
have
the
exposes
like
this
there
we
will
first
have
the
I
I
know
say
it.
The
data
collection
subprocess
is
again
just
summarize
it
here:
it
retrieves
the
performance
measurements
from
the
computer
or
network
system
that
is
being
managed.
B
Second,
we
have
a
resonance
process
that
receives
the
collected
data
and
and
receive,
for
example,
additional
context.
I
also
mentioned
this.
That
would
be
some
kind
of
information
that
is
provided
by
external
elements
that,
from
outside
the
network,
I
call
them
those
external
events
or
external
element.
This
is
the
sorry
standalones,
let's
say
well,
and
that
information
will
be
used
by
finding
to
find
out
the
rules
that
are
governing
the
current
situation
of
the
network.
B
Then
the
third
through
process
would
be
the
solving
subprocess.
That
is,
that
guess,
the
the
situation
and
rules
previously
obtained
to
build
some
dynamic
knowledge
graph.
That
means
to
gather
together
all
the
knowledge
that
represent
the
current
situation
of
the
network.
The
knowledge
graph,
of
course,
must
follow
some
particular
ontology,
as
I
defined
before,
and
it's
it
is
worked
in
this
solving
process.
B
The
knowledge
graph
the
base
is
worker
in
order
to
find
an
acceptable
situation,
so
the
transform
the
the
the
actions
that
must
be
applied
to
this
knowledge
graph
in
order
to
find
one
acceptable
situation
will
be
later
in
the
in
the
next
super
process
by
the
planning
show
process
will
be
used
to
correct
the
network
or,
to
let's
say,
adapt
the
network
to
some
current
situation
or
future
problem
to
prevent
future
problems.
Okay,
and
how
to
do
that
is
the
final
show
process.
B
The
planning
process
will
will
get
the
actions
that
might
be
executed
in
order
to
to
achieve
this
coverage
as
a
type
of
situation
and
plan
them
how
to
apply
them
into
the
networking
for
to
ensure
they
are
correctly
enforced
in
the
network.
B
In
addition
to
to
the
this
operation,
we
in
the
in
the
draft,
we
also
define
some
concepts
that
must
be,
let's
say,
researcher
and
also
organized
in
this
kind
of
framework
like,
for
example,
this
particular
support
for
the
closet,
loop.
Okay,
all
managers
or
management
solutions,
especially
the
automated
management
solutions,
have
are
aware
of
some
kind
of
closer
loop.
So,
but
we
need
to
define
particular,
particularly
how
these
closely
loop
can
be
implemented.
B
B
Like,
for
example,
incorporating
a
target
entry,
automated
intelligent
methods,
that
means
that
we
have
some
kind
of
target
definition
for
the
for
the
network
management
solution
and
the
network
introducing
processes
will
just
try
to
find
and
get
all
the
information
required
in
order
to
to
to
accomplish
that
target
is,
is
not
just
to
watch
the
network
to
see
what
is
happening.
No,
it's
the
opposite
in
this.
B
In
this
case,
we
first
we
first
establish
a
target,
and
then
we
work
in
order
to
find
out
evidence
or
find
out
any
any
data
that
is
required
to
demonstrate
or
to
achieve
the
targeted,
general.
Okay.
Next
concept
is
the
external
level
detectors
again.
I
insist
this
to
get
I
discussed
about
them
previously,
but
is
to
include
in
the
ai
framework.
B
We
need
to
model
how
information
from
outside
the
network
is,
will
cross
the
boundaries
and
and
be
introduced
inside
the
the
network
management
solutions
and
how
to
correlate
that
can
that
information
can
be
correlated
with
with
what
is
happening
inside
the
network.
Let's
say
performance
measurements,
so
all
of
these
will
be
also
the
or
must
be
also
some
some
kind
of
defined
in
this
framework.
Let's
slide
this.
B
And
two
other
concepts
are
the
anticipation
of
network
requirements
instead
of
just
exploiting
ai
for
predicting
what
we
are
currently
doing,
for
example,
for
machine
learning,
we
are
predicting
what
can
be
the
the
near
future,
but
instead
of
just
doing
that,
we
can
focus
on
what
I
I
was
defining
about
anticipation,
let's
say
to
find
the
situations
and
the
events
that
will
be
dispatched
are
will
be
surely
happening.
B
Where
we
are,
we
can
be
sure
they
are
happening,
so
we
must
anticipate
the
impact
of
the
in
the
network
of
some,
for
example,
let's
say
one
flash
cloud
event
that
is
happening,
so
we
can
know
exactly
what
what
that
will
trigger
in
the
network,
and
we
cannot
anticipate
to
that.
But
it's
not
just
prediction:
it's
not
something
that
we
can
fail.
No,
we
it's
something
that's
sure,
so
we
can
focus.
B
We
can
have
a
particular
focus
to
this,
because
we
will
have
much
more
time
to
to
analyze.
For
example,
in
the
case
of
digital
twins,
we
can
one
situation
that
we
anticipate.
We
can
input
that
to
the
digital
twin
and
see
what
hap.
What
can
happen?
The
easter
twin
can
say:
oh
well,
this
is
most
probably
to
happen.
It's
okay
or
not.
We
can
say
the
opposite,
but
in
these
cases
I
insist
is
anticipation.
B
Not
just
prediction
so
is:
is
it
can
be
a
bit
different
only,
but
at
the
end,
when
you
think
about
the
implications
of
in
the
in
the
definitions
and
is
very
different
and
finally
is
the
the
reasoning
I
support
for
intelligent
reasoning.
Is
there
are
many
methods
that,
for
example,
the
machine
learning
and
even
for
example,
particular
neural
networks
that
are
quite
good
on
obtaining
some
kind
of
of
classification
or
prediction
from
some
data?
Okay,
but
the
full
reasoning
process
must
be.
B
Must
be,
let's
say,
researcher
in
terms
of
to
to
to
be
modeled
and
to
be,
for
example,
complemented
with
metadata
and
things
like
that,
so
we
can,
for
example,
achieve
some
kind
of
explanation
or
find
out
some
explanation
to
to
the
actions
that
are
automatically
obtained
by
this
ai
framework,
insisting
in
the
wrestling
of
course
next
slide.
Please.
B
Finally,
it's
just
the
most,
let's
see
just
two
out
of
highlight
the
the
most
important
issues
that
I
come.
B
I
find
here
in
this
ai
framework
would
be
just
to
define
common
interfaces,
so
we
need
to
define
the
the
gaps
we,
we
don't
have
interfaces
for
allowing
several
ai
modules
to
talk
to
each
other,
and
that
is
important,
because
in
the
future
we
will
have
several
vendors
that
can
provide
specific
functions,
ai
functions
that
we
we
will
go
in
one
direction
or
another,
and
and
most
of
them
can
be
able
to
talk
to
each
other.
So
it's
important
to
this
define
these
interfaces,
including
the
ontology
I
previously
discussed.
B
D
Okay,
yeah:
you
have
a
lot
of
having
good
good
inputs
to
the
discussion,
but
we
have
to
to
be
straight
on
just
very
quickly
oligarchy
or
something.
D
C
Just
one
comment
like
I
think
it
would
be
beneficial
to
add:
maybe
the
intent
driven
approach
to
the
ai
framework.
You
know,
because
I
do
see
that
you
have
kind
of
mentioned
other
contexts
in
one
place,
but
all
and
common
interfaces,
but
I
do
believe
it
would
be
beneficial
to
say
how
it
relates
to
the
intent
driven
approach
and
how
they
can
how
they
can
coexist
together.
You
know,
because
I
do
believe
there
is
a
intent-driven
approach
that
would
be
beneficial
in
the
ai
framework
going
forward.
B
Okay,
thank
you.
I
will
I
take
note.
I
will
try
to
in
to
let's
say
interconnect
this
with
the
intent
based
approach
we
are.
We
are
also
already
you
know
in
this
in
this
way,
but
I
will
particularly
take
a
look
to
that.
Thank
you.
Thank
you.
D
Okay,
we
really
have
to
to
cut
because
others
may
have
sessions
to
join.
Thank
you,
everyone
for
the
participation
on
this
last
presentation.
From
petal
I
mean
there
is
a
draft
also
associated
to
that.
So,
if
you
have
any
comments,
you
can
bring
that
offline
to
the
mailing
list
or
reach
out
to
together
and
the
offers,
and
we
will
have
to
in
fact
consider
how
this
activity
falls
into
the
activity
of
the
research
group.
It
connects
maybe
with
other
activities.
D
A
Thanks
thanks
everyone
for
participating
and
yes
for
the
representation,
I
will
share
some
some
thoughts
in
the
mailing
list.
Another
hospital.
I
have
some
questions,
but
I
think
it's
too
late
now
to
understand.
So
thank
you
all
and
yes,
we'll
send
me
nudes
and
we'll
keep
in
touch
emailing
lists.