►
From YouTube: IETF111-NMRG-20210727-2300
Description
NMRG meeting session at IETF111
2021/07/27 2300
https://datatracker.ietf.org/meeting/111/proceedings/
A
A
So
concerning
the
note:
well,
a
few
announcements,
so
audio
and
video
recordings.
If
you
participate
online
and
turn
on
your
camera
and
or
microphone,
then
you
can
sense
to
appear
in
such
recordings.
So
please
be
aware
that
this
meeting
is
being
recorded
as
we
started
and
that
you
agree
with
such
a
policy
right
now.
A
Privacy
and
code
of
conduct,
so
as
a
participant
in
or
attended
to
any
irtf
activity,
you
acknowledge
that
written
audio
and
video
and
photographic
records
of
meetings
may
be
made
public
personal
information
that
you
provide
to
iotf
will
be
handled
in
accordance
with
the
privacy
policy
that
is
available
at
the
link
that
you
can
see
in
the
slide
as
a
participant
or
attendee.
You
agree
to
work
respectfully
with
other
participants
and
please
be
contact
the
ombuds
team.
A
Now
a
few
words
about
goals
of
the
irtf,
so
the
irtf
internet
research
task
force
conducts
research.
It
is
not
a
standards,
development
organization
and
the
irtf
focuses
on
longer-term
research
issues
related
to
the
internet.
While
the
parallel
organization,
the
ietf,
focuses
on
short-term
issues
of
engineering
and
standards-making.
A
Online
meeting
etiquette,
so,
as
I
mentioned
already
previously,
the
session
is
being
recorded.
Please
keep
your
audio,
muted
and
video
off
when
you
are
not
presenting
or
speaking
and
when
speaking,
please
start
by
stating
your
name
and
affiliation.
This
is
very
useful
for
people
that
we
look
at
the
recordings
after
you
can
also
get
some
useful
links
in
the
agenda
that
is
available
on
the
data
tracker,
so
the
materials
for
decisions.
A
A
A
A
A
B
B
Actually,
this
is
the
first
time
to
present
and
drop
in
this
research
group
and
as
we
presented
previously
since
iot,
idf
or
108,
the
basic
idea
of
the
draft
is
to
define
digital
tree
network
in
network
and
communication
field.
Our
goal
of
the
draft
is
to
define
a
new
tool
for
efficient
and
intelligent
management
and
to
stimulate
innovation
with
optimized
cycles.
B
B
B
B
Piece,
this
slide
shows
the
clarification
on
mapping
definition
and
the
updates
of
continuous
verification
and
the
mapping
is
used
to
identify
identify
the
digital
twin
and
the
underlying
entities
and
establish
a
real-time,
interactive
relation
between
the
physical
network
and
the
tree
network,
or
between
two
net
three
networks.
The
mapping
can
be
one-to-one
vertical
mapping
or
one-to-many.
Horizontal
mapping
and
continuous
verification
is
an
extension
of
develops
practices
that
are
concerned
with
verifying
the
system
as
a
whole.
B
B
B
B
Please,
okay,
the
coming
two
slides
show
the
detail
reveal
comments
from
expert
ramen
cetera
and
our
initial
reply.
The
first
comment
is
about
reference
on
digital
twin
technology.
The
second
comment
suggested
us
to
cite
more
papers
on
definition
of
digital
twin
to
network,
and
the
third
comment
is
about
stressors
are
to
to
to
clarify
the
benefit
of
privacy
and
the
regulatory
compliance.
B
B
We
believe
our
work
is
an
expansion
of
network
stimulation
and
the
unique
difference
from
traditional
network
management
architecture
is
using
interactive
virtual
rail
mapping
to
build
close
loop
network
automation.
Our
concern
is
providing
survey
on
other
architectures.
I
might
lose
workers
of
this
dropped,
so
we
decided
to
just
add
a
text
in
abstract
and
intro
instruction
in
next
or
next
version
to
clarify
the
scope
and
the
relation
with
traditional
network
management.
Architectures.
B
B
The
session
built
a
platform
for
exchanging
ideas
on
research
and
application
of
digital
technology
in
network
and
communication
field.
Our
co-author
diogo
lopez
co-chaired
the
session
in
this
session.
Totally
seven
apprenticeships
presentations
are
were
provided,
for
them
are
about
basic
technologies
of
digital
network,
including
traffic,
modeling,
security
applications,
model
of
wireless
channels
etc,
and
one
is
about
architectural
concept
of
dtn,
which
is
very
close
to
our
jobs
and
providing
a
method
for
dtm
constructing
and
another
one
about
data
driven
routing
are
based
on
dtn.
B
B
Going
forward,
we
will
keep
investigating
more
use
cases
and
requirements
of
dtn
and
address
any
issues
within
the
next
meeting
and
after
that,
after
address
all
our
comments
and
issues,
we
will
request
the
research
group.
Adoption
call
for
this
document
welcome
to
join
our
work
and
any
comments
are
welcome.
Thank
you.
A
Okay,
I
I
have
a
quick
remark.
I
think
it
was
when
you
presented
the
layered
view,
I
think
in
in
in
the
previous
statement.
You
mentioned
that
there
could
be
different
way
of
having
a
relationship
between
the
the
real
network.
I
mean
the
real
infrastructure
and
the
digital
twin,
but
it
seems
like
in
this
layered
view
that
we
we
need
to
go
always
through
the
digital
twin
to
have
actions
or
I
mean
effects
on
the
physical
network.
B
Okay:
okay,
it's
a
it's
a
good
stretching.
Currently
we
it's
true
that
currently
we
we
only
folks
are
to
to
represent
the
virtual.
The
physical
network
and
the
services
running
on
the
facebook
market
in
in
natural
digital,
twin
layer
and
your
you
are
stretching
to
represent
the
applications
on
digital
network
is
not
well
considered,
but
we
basically,
I
think
it's
not
it's.
We
can.
A
C
D
C
Well,
basically,
before
I
started
just
a
a
comment
of
what
did
your
previous
question-
the
ones
in
some
somehow
involved
as
well
in
the
in
the
dta
effort,
precisely
whether,
where
they
believe
is
that
this
is
the
model
when
you
are
trying
to
control
the
network
through
the
digital
twin.
But
you
can
do
one
half
of
the
other
mechanisms
for
for
control
and
then
you
can
choose
it's
not
that
every.
C
F
C
So
can
you
move
on
next
to
the
next
slide?
Yes,
so
just
a
very,
very
quick
introduction
to
what
the
f5
growth
project
is.
It's
a
it's.
A
european
project
funded
by
the
european
commission
as
a
joint
research
project
among
several
european
companies.
Very
much
focus
precisely
on
demonstrating
how
5g
technologies
can
be
applied
to
improve
the
results
of
the
in
vertical
industries
in
real
production
environments
and
to
explore
the
combining
management
both
by
the
measure
providers
and
the
and
the
vertical
industry,
is
seeing
achieving
a
common
goal.
C
C
Can
you
move
on
please
so,
apart
from
these
well
precisely
to
enable
this
and
demonstrate
the
the
viability
of
these
of
these
pilots
and
the
and
the
use
of
the
the
application
of
different
technologies
in
network
management
and
interaction
between
the
application
needs
and
the
and
a
certain
degree
of
control
of
the
network?
So
that's
to
satisfy
the
application
kpis.
C
Can
you
move
on
please
so
and
for
this-
and
this
is
probably
the
most
important
thing
that
is
about
this,
this
environment
that
is
has
been
demonstrated
and
in
which
we
have
been
working
for
the
during
the
last
couple
of
years.
For
more
than
a
couple
of
years
is
that
the
idea
is
that
we
have
the
product
has
gone
in
experimenting
with
what
I
have
called
here.
C
Normally,
there
is
a
previous
loop
that
is
used
precisely
for
training
the
mechanism
using
machine
learning
techniques,
whatever
the
techniques
you
you
feel
like,
and
the
important
thing
is
that
there
is
a
third
aspect.
A
third
loop
here
that
is
is
about
precisely
controlling
the
flow
of
monitoring
data
that
is
entering
any
of
the
other
two
loops.
So
you
can
control
you
can
control
how
much,
how?
C
So
the
the
the
mechanisms
in
which,
with
with
which
I
use
it
used
by
the
con,
the
human
controllers
to
interact
with
the
with
the
twin,
has
a
particular
request
on
latency,
as
well
as
the
control
that
is,
that
is
enabled
by
the
digital
train
to
to
be
applied
to
the
to
rob.
Can
you
move
on,
please
the
the
demo
I
mean
or
the
demonstration
that
was
run.
C
Take
into
account
that
again
it's
a
this.
This
three-dimensional
loop
allows
us
to
train
different
modules
using
real
data.
That
is,
are
not
necessarily
the
same
model
that
is
being
executed
using
a
combination
of
data,
and
whenever
there
is
change
in
the
model,
the
collection
of
data
for
the
inference
loop
can
be,
can
be
changed
as
well.
Can
you
move
on
please?
C
So?
Basically,
this
is
just
to
illustrate
the
the
different
loops
and
the
and
the
different
flows.
I'm
not
going
into
detail
on
this.
But
if
you
look
at
it
well,
there
is
a
phase
in
which
always
a
separate
phase.
In
which
you
can
make
the
model
and
training
of
the
of
the
ai
model,
and
once
the
service
request
happens
and
is
the
service
request,
can
include
precisely
the
the
requirements
in
terms
of
the
ai
model
to
be
applied.
C
Can
you
move
on
and
well
basically,
the
the
third
slide
just
illustrates:
what
happens
when
when
a
particular
scaling
need
is
detected
and
well
in
this
part,
I
would
say
is
quite
typical
in
the
sense
that
this,
the
application
of
the
ai
to
make
a
life
cycle
decision
on
a
particular
services.
Again,
what
is
important
is
that
the
monitoring
jobs
can
be
once
the
decision
is
taken
and
the
metering
jobs
can
be
reconfigured
to
dynamically
adapt
the
the
data
flows.
C
Can
you
move
on
so
just
as
a
conclusion
and
is
basically
that
well,
the
idea,
let
me
say,
is
that
we
have
been
experimenting
with
this
three-dimensional
closed
loop.
We
have,
we
have
for
sure,
use
the
common
double
loop
for
landing
and
inference.
I
mean
the
for
training
the
model
and
for
applying
the
mobile,
but
the
there
is
important.
C
This
has
been
done
precisely
because
one
of
the
main
goals
of
the
project
from
the
beginning
was
precisely
to
provide
a
complete
data
monitoring
data
infrastructure.
So
that
implies
that
it's
not
only
that
we
have
the
infrastructure
that
precisely,
I
guess,
is
natural,
the
one
who
is
going
to
make
another
presentation
of
our
data
infrastructure.
C
It's
not
only
about
this,
but
it's
about
including
the
requirements
for
data
driven
operations
in
the
service.
Descriptors
themselves
is
supporting
monitoring
orchestration.
I
mean
deciding
where,
where
data
is
collected
and
which
data
is
collected
from
these
points
and
the
other,
the
idea
that
says
it
was
providing
models
about
the
data
flows
as
well
that
allow
us
to
abstract
the
models,
combine
them
and
make
a
a
much
more
powerful
feeds
of
data
to
the
to
the
ai
and
ml
modules
and
well.
C
What
is
in
practice
is
that
we
have
applied,
and
we
have
demonstrated
it
into
an
industrial
use
case
and
taking
into
account
that
we
are
talking
about
the
digital
training
and
industrial
digital
training.
Precisely.
We
believe
that
this
is
some
conclusions
that
could
be
interesting,
please
connection
to
what
to
what
chang
was
presenting
before
and
in
general,
to
the
application
of
of
a
set
of
network
services
able
to
support
in
the
future
the
the
use
of
digital
twins
in
whatever
the
industry.
C
Can
you
move
on
just
the
last
one
is
just
to
to
show
where
more
more
information
can
be
found.
There
is
a
paper
on
this
that,
by
the
way,
it
was
get
some
award
to
have
a
best
paperwork
just
just
to
make
a
little
bit
of
of
of
of
of
roasting,
and
this
there
is
a
at
the
demo.
The
detailed
demo
was
presented
in
the
infocomm,
and
you
have.
C
A
C
C
C
You
have
I'm
wondering
whether
you
have
any
goals
or
degrees,
reduce
the
election
or
yeah
yeah.
Sure
sure
sure
that's
that's
precisely
idea.
I
mean
I'm
trying
to
read
to
reply
to
john
gueng
hong,
if
I
pronounce
it
right
yeah.
The
idea
is
precisely
is
to
optimize
the
the
time
of
response
and
to
provide
a
much
more
agile
response
in
terms
of
how
you
apply
ai.
C
It's
precisely
this
diversity
in
how
you
monitor
and
feed
data
would
allow
you
to,
for
example,
to
experiment
with
alternate
modules
as
models
as
they
are
you're
applying
one
model
and
you
are
training
another
model
using
the
same
plan
data.
At
the
same
time
and
this
we
believe
this
is
quite
important
and
yes,
but
you
just
say
I
guess
the
slides
are
available
as
with.
C
Estimated
the
actual
amount
of
currency
resources
related
to
a
high
accident.
Well,
this
question
is
better,
I
would
say
it's
better
for
chiang
because
it
has
to
do
more
with
it
with
the
previous
one.
The
the
digital
chain
that
we
are
we
were
working
here
with
is
the
digital
training
of
a
robotic
arm
pedal,
not
not
a
general
of
the
of
the
digital
thing
of
a
network.
C
A
E
G
G
Piece
and
in
this
slide
I
will
present
the
background
and
the
induction
that
feature
of
this
total
network
we
refer
to
is,
from
the
first
junction
to
the
japanese,
digital
team
network
concept
and
the
ctn
is
defined
as
a
network
system
with
physical
network
and
network
which
can
imagine
so.
The
construction
of
events
requires
the
advantage
of
physical
network
to
update
the.
G
E
G
G
Next
slide,
please
so
the
main
pro
process
of
this
data
clock
collection
method
is
to
send
instructions
in
the
team
network
to
the
physical
network
to
collect
data
on
the
mind
and
then
the
physical
network
parses
and
excuse
instructions
such
as
data
cleaning
or
knowledge
representation,
and
then,
since
the
processed
or
represented
represented
data
to
the
dtn
next
slide.
Please.
G
G
G
So
this
slide
is
about
the
introduction
of
data
storage
center
in
physical
network.
It
has
two
functions:
firstly,
it
can
store
data
such
as
performance
indicators,
operational
status,
logs
traffic
scheduling,
business
requirements
and
so
on.
Secondly,
each
has
the
function
of
automatically
passing
the
instructions
sent
by
the
instruction
management
center
in
the
team
network.
So
next
slide.
G
G
G
And
after
that,
the
instruction
management
center
in
the
team
network
sends
the
corresponding
instruction.
According
to
the
address
to
the
data
storage
center
in
physical
network
after
receiving
the
instructions,
the
data
storage
center
in
the
physical
network
will
pass
and
excuse
them
according
to
the
instructions
such
as
filling
missing
data
data
association,
knowledge,
representation
and
so
on.
The
data
storage
center
of
the
physical
network
will
send
the
processed
and
represented
data
to
the
data
storage
center
in
the
tin
network
and
next
slide.
Please.
G
And
then
the
bandwidth
resource
consumed
by
data
transmission
is
greatly
reduced
and
next
slide
please.
So
in
the
next.
We
will
unleash
the
instruction
carrying
information,
more
detail
and
discover
more
concrete
examples
of
data
collection
and
concern
about
the
extraction
that
consider
about
the
extension
and.
D
Thank
you
for
the
presentation.
For
me
there
is,
there
is
a
something
that
remains
unclear,
because
one
of
the
main
motivation
is
to
reduce
storage
capacity
of
the
digital
twin.
But
so
it
gives
me
the
impression
that
the
that,
with
your
approach,
only
moves
the
problem
towards
the
physical
network
and
also
put
more,
let's
say,
pressure
the
physical
network,
where
you
have
to
do
a
lot
of
processing
that
can
also
have
some
disadvantage,
as
maybe
the
physical
networks
are
sensible
and
sensitive
and
should
not
be
maybe
overloaded
with
too
much
additional
processes.
D
G
A
H
H
H
Next
slide,
please.
So
the
the
system
standard
focuses
on
defining
mechanisms
to
to
exchange
context
information
in
an
interoperable
way.
To
this
end,
etsy
sim
standardizes,
the
the
so-called
ngsild
protocol
to
exchange
context
information,
and
this
this
protocol
is
composed
of
of
two
main
two
main
ideas.
First,
is
the
ngsild
information
models
which
are
based
on
property
graphs
to
to
represent
that
context?
H
These
are
the
the
components
that
are
in
the
identified
the
corridors,
the
the
context
broker,
which
stores
the
context
information.
Then
we
have
the
context
sources
which
produce
context,
information
such
as
junk
based
devices
and
then
context
consumers
which
query
and
subscribe
to
context
information
from
the
from
the
context
broker
next
slide.
Please
so,
let's
see
some
some
examples
of
modeling
data
sources,
here's
the
first
one,
which
is
a
promises-based
data
source.
H
Where,
in
this
diagram,
you
can
see
the
information
model
for
for
this
data
source,
where
we
have
an
entity
representing
the
the
promising
server
which
can
be
accessed
through
a
specific
endpoint.
With
some
credentials,
and
also
this
promising
service
has
some
metrics,
with
with
a
particular
name
and
a
set
of
labels
and
and
in
turn,
these
metrics
belong
to
to
a
certain
metric
family,
just
to
clarify
that
we
were
aligned
with
with
the
openmetrics
initiative
which
defines
this.
H
This
schema
for
metric
and
metric
family
entities
next
slide,
please,
and
as
a
second
example,
a
telemetry
based
data
source.
Here
in
this
diagram,
we've
defined
the
entity
device
which
represents
the
telemetry
service.
In
this
example,
we
chose
gnmi,
as
you
can
see,
it
captures
the
version
of
the
protocol,
the
supported
encodings,
also
it
has
an
end
point
associated
with
some
credentials
to
to
access
it
and,
on
the
other
hand,
this
this
telemetry
service
supports
some
some
young
modules
with
the
associated
metadata,
such
as
the
name,
the
revision
and
the
organization
next
slide.
H
H
You
can
see
apache
link
for
data
aggregation
and
all
this
is
possible
thanks
to
apache
kafka,
so
these
big
data
tools
can
publish
and
subscribe
to
to
the
monitoring
data
and
at
the
top
it's
what
we
can
call
the
the
context
plane
where
the
contact
information
is
stored
in
the
scorpio
building
block,
which
is
an
implementation
of
the
context
broker
the
on
the
left
hand,
side,
the
data
sources
will
produce
context.
Information
such
as
the
young
modules
supported
by
device
and,
on
the
right
hand,
side.
H
H
So
the
idea
is
that
the
data
consumers
would
discover
this
this
metric,
that
you
see
at
the
left
and
they
will
request
to
to
to
schedule
a
task
that
runs
on
template
to
collect
this
data
and
store
it
in
kafka.
Note
that
this
is
the
first
stage
of
the
data
pipeline.
This
the
let's
say
the
collection
there,
the
digestion
of
of
of
data-
and
this
will
continue
if
we
add
more
space
with
another
task
for
aggregation
and
other
tasks
for
the
delivery
of
data
next
slide.
Please.
H
H
H
H
So
the
idea
is
that
this
aggregated
data
is
combined
and
transformed
into
the
data
model
that
I
showed
previously,
and
this
is
delivered
through
a
rest.
Api
next
slide,
please
to
summarize
some
conclusions
next
slide.
Please
well.
Data-Driven
network
management,
as
as
we
see
as
we
saw,
requires
the
integration
of
a
wide
variety
of
data
sources
and
data
consumers.
H
H
And
potential
next
steps
that
we
identified
first,
the
combination
of
context,
information
from
different
mode
from
different
levels,
such
as
what
we
call
the
the
operations
level,
which
could
be
a
network
function
and
network
service.
Think
about
the
the
the
telemetry
based
data
source
that
I
represented
well,
that
device
entity
could
be
linked
to
a
network
function
entity
which
in
turn,
could
be
associated
to
a
network
service
entity
and
also
combination
from
from
the
business
level,
such
as
indoor
policies,
intense
slas
owners
of
of
domains
etc.
H
The
assurance
graph
could
be
could
could
could
be
captured
as
more
more
context,
information
and
also
the
the
semantic
data
aggregator
could
collect,
as
we
saw
could
collect
monitoring
data
from
different
data
sources
of
different
nature
also
also
discovery
of
context.
Information
such
as
when
a
new
network.
Sorry,
when
a
new
element
appears
in
the
network,
but
all
the
the
the
aggregator
has
to
to
be
synchronized
with
with
that
information.
H
A
A
H
A
A
D
D
D
So
you
may
have
seen
this
this
this
report
in
the
main
english,
but
just
to
to
smile
here
to
all
the
participants.
This
meeting,
it's
important
to
know
that
there
is
a
new,
let's
say,
version
3
of
the
document.
That
is
a
version
where
we
integrate
the
different
command
received
so
far
and
where
we
adjust
actually
the
list
of
challenges
so
do
not
edit,
but
also
do
not
read
the
version
2
anymore.
It's
not
today.
D
In
particular,
it's
important
to
the
same
organization,
because
maybe
at
some
point
we
decided
to
merge
this
channel
that
it
would
be
easier
if
we
clearly
see
what
has
a
different
part.
So
mostly,
there
are
three
parts
that
are
quite
common.
The
first
part
is
about
the
motivation
of
the
challenge
that
should
highlight
the
criticality
or
the
impact
of
the
problems
that
need
to
be
addressed,
of
course,
related
to
the
coupling
within
network
management.
D
We
should
have
also
section
that
summarizes
what
is
the
current
state
of
the
arts
related
to
the
challenge,
at
least
to
see
what
are
the
trends
and
basically
some
how
to
identify
the
gap
between
what
has
been
proposed
and
what
remains
to
do
to
to
fulfill
the
challenge
and
also
a
perception
that
will
be
more
descriptive.
Regarding
the
challenge
and
most
of
the
time,
we
think
that
the
challenge
you
have
a
main
challenge
that
you
want.
D
Of
course,
we
need
to
highlight
what
is
specific
to
having
our
domain
network
manager
that's
working
in
url
when
we
are
talking
about
the
challenge.
In
many
cases,
challenges
are
not
are
not
specific
to
the
top
manager
is
a
general
sense,
but
of
course
we
have
some
specific
part
in
our
domain.
That
makes
the
challenge
a
bit
different
or
to
be
addressed
differently
next
slide,
please.
D
After
this
meeting
and
then
realizes,
the
challenge
that
we
have
identified
is
that
somehow
we
have
worked
all
challenges
related
to
data
under
data
management
challenge,
and
I
was
also
very
very
happy
to
see
the
presentation
before
because
of
some
equal
to
what
you
have
in
this
challenge.
So
we
try
to
to
integrate
the
different
aspects,
at
least
that
have
been
already
providing
the
document.
D
D
Answer,
if
you,
if
you
remember
with
the
previous
specific
challenge
one,
whereas
you
know
for
how
to
plan
the
action
an
implant
is
provided.
So
basically,
we
generalize
that
for
the
use
of
ai
for
planning
action
not
only
for
extracting
knowledge
but
with
taking
action
into
a
network
and
another
was
more
about
the
use
of,
for
example,
nlp
techniques
to
interpret
intense
or
high
level
high
level
representation
of
goals.
And
so
we
some
more
integrate
these
and
extend
another
channel,
which
was
more
about
also
integrated
external
events.
D
D
D
D
D
Maybe
there's
also
the
question
automatically
for
the
television
that
could
be
probably
integrating
to
data
management.
Of
course,
a
ai
for
planning
affection,
and
I
said
before
that
we
refined
the
title
to
be
not
specific
to
a
pen,
but
the
text
still
need
to
be
refined.
Accordingly,
it's
not
done
yet,
and
it's
the
same.
Also
for
exploiting
is
not
technical
event
and
it's
a
kind
of
let's
say
a
mirrored
with
a
simple
copy
paste.
It
needs
to
be
more
activated.
D
D
Okay,
so
the
progress
maybe
more
focused
knowledge
in
the
last
month,
but
the
content
part
on
the
answer.
Let's
say
you
know
securing
the
challenge
so
for
the
management,
as
I
said,
so,
we
merge
multiple
challenges.
It's
a
large
coverage
for
this
challenge
with
multiplicity.
It's
a
specific
challenge
and
channel
try
to
follow.
D
Let's
say
the
representation
of
the
data
is
pipeline
and
see
where
there
is
some
challenge
related
data,
starting
with
the
definition
of
that
that
will
be
collected,
the
need
for
collection
mechanisms,
although
the
optimization,
as
far
as
the
collection
of
data
before
the
regarding
the
aggregation,
click
on
the
degree
item,
consideration
also
for
customer
customizable,
flexible
monitoring,
information,
the
need
for
level
data.
That
implies
some
or
maybe
the
problem
regarding
data
sharing
immunization,
but
also
security
and
privacy
aspect,
and
we
also
some
in
this
part.
We
have
also
some
premises
more
related
to
all.
D
D
D
So
for
the
next
step,
so
we
have
four
challenges
that
have
already
written
in
the
description.
As
I
said
before,
the
idea
is
that
we
focus
on
working
with
refining
these
challenges.
I
hope
that
you'll
end
up
focused.
I
know
this
is
some
of
that
that
will
try
to
progress
on
this
problem
if
you
can
to
have
a
little
session,
maybe
one
session
per
challenge
for
people
interesting
for
the
challenges,
so
it
should
be
more
small
group
and
also
as
there's
also
a
detailed
description.
D
And
proposition
also
is
to
use
github
to
work
on
this
document
for
using
versioning,
but
also
to
use
issue
tracker
and
github
to
follow
the
discussion,
because,
in
the
short
google
doc
is
not
so
easy
to
to
follow
the
different
comments.
The
different
let's
say,
discussions
same
comment
and
see
when
there
are
results,
so
the
proposition
will
be
to
move
this
to.
D
A
Thanks
jerome,
I
I
do
have
a
question.
I
don't
know
if
the
answer
is
known
yet,
but
so
right
I
mean
so
far.
We
have
been
using
google
doc
to
let's
say
draft
the
first
version
of
this
document.
There
seems
to
be
an
intent
to
go
towards
an
id
to
put
some
exposure
in
the
data
tracker,
but
have
you
already
thought
about
what
would
be
a
the
form
of
the
result?
Do
you
want
to
have
it
as
a
publication
in
a
magazine
or
a
journal?
Do
you
want
to
have
it
as
an
rfc?
D
So
yeah
we
discussed
this.
I
didn't
get
so
much
feedback
on
that
there's
a
few
years
that
people
are
all
open
to
the
different
firm
personally,
but
my
personal
opinion
is
to
go
more
for
magazine
formatting
in
a
journal
but
yeah.
This
is
only
based
on
on
a
few
feedback
that
I
received,
but
if
others
want
to
express
their
own
opinion
that
people
contribute
to
this.
D
E
D
D
A
J
Oh
okay,
hello,
everyone
online,
I'm
kyle
from
channel
mobile,
and
I
will
talk
about
some
updates
to
the
network
management
intent
draft
so
next
time
please
here
is
just
a
very
brave
recap
of
the
mri
concept
and
the
whole
diagram
of
mmr
framework.
So
ami
is
short
for
network
environment
intent
and
it
represents
users
or
network
operators
intent
of
the
of
the
network's
quality.
So
there
are
five
major
components
in
showing
the
graph.
J
So
next
point
please
this
time
we
there
are
two
major
updates
compared
to
the
version
zero
one.
So
the
first
first
one
is,
we
add
a
new
use
case
on
the
clustered
performance
measurement
intent,
and
this
part
is
added
by
colleague
on
joseph
fukula
from
huawei.
And
the
second
part
is,
we
add,
a
new
chapter
making
classification
of
mi.
J
J
And
this
time
we
add
a
new
part,
it's
about
spatial
accuracy
right.
So
the
use
case.
The
name
of
use
case
is
a
clustered
performance
management
intent
and
it's
referred
from
rfc
triple
a9,
and
it's
about
how
to
partition
the
network
in
proper
ways
to
achieve
accurate
measurement
result
without
affecting
services
and
at
the
same
time
and
at
the
same
time
reduce
to
reduce
the
resources
when
doing
network
measurement.
J
J
J
While
on
the
update,
we
divide
it
into
the
dynamic
mi,
intent
means
the
network
premises
remain
unchanged,
but
the
translated
policy
will
be
adjusted
according
to
different
environment
and
like
I
want
to
know
that
when
the
network
busy
so
and
when
you
do
some
network
measurement,
you
want
to
apply
these
intent
based
system
to
a
very
large
network,
a
very
large
scale
network.
The
result
will
be
changed
based
on
this
intent,
so
we
give
so
that's
why
we
divide
the
network
measurement
intent
into
two
categories,
so
net
next
slide.
Please.
J
A
Thanks
lehan,
let
me
check
the
chat.
If
I
see
questions
it
doesn't
seem
so
yeah,
I
think
it.
You
have
good
development
of
the
document,
that's
distributing,
so
I
encourage
you
to
continue
in
that
in
that
way,
and,
of
course,
if
you
want
to
consider
this,
as
I
mean
in
the
scope
of
the
ibn
use
case,
that
would
be
also,
I
think,
very
interesting
to
look
into
yeah
sure
thanks.
A
J
A
D
D
Almost
all
there
is
maybe
one
or
two
I
mean
one
that
I
sent
to
the
authors
and
also
sorry
before
that
we
we,
we
agreed
that
my
role
of
she
probably
should
should
evaluate
if
there
will
be
a
need
for
a
new
round
of
classical
rbz
of
or
if
the
the
way
they
could
they
address.
The
command
does
not
necessitate
unless
so,
in
my
opinion.
So
basically,
in
my
opinion,
now
there
are
still
some
minor
commands.
D
I
don't
think
there
is
a
really
need
for
for
getting
the
last
call
on
this
document,
because,
let's
say
before,
we
got
a
lot
of
positive
support
and
the
address
the
different
commands.
This
command
does
not
really
basically
modify
the
curves
of
the
document,
so
the
id
now
is
to
more
address
the
remaining
command
with
directly
between
the
authors
and
hi
and
then
go
for
the
next
step.
That
is
the
submission
to
a
higher
review
and
that's
the
status
of
this
document.
A
A
A
There
have
been
also
five,
maybe
a
bit
more
so
relatively
detailed
reviews
of
the
latest
version
of
the
document,
knowing
that
the
document
already
received
quite
extensive
comments
in
the
initial
version,
so
these
were
the
reviews
received
only
for
the
research
group
last
call
period,
a
version
3
of
the
research
group
document
was
issued
submitted
in
march
this
year,
which
contained
new
text
which
addressed
all
the
pending
comments.
There
was
quite
a
lot
of
ongoing
comments,
so
this
was
a
good
contribution
from
the
offers
when
I
was
checking
the
draft
recently,
this
last
version.
A
So
what
I
will
put
in
my
shippers
write-up
is
that,
of
course,
the
the
offers
made
a
very
good
work
to
progress
the
documents
according
to
to
the
comments
received.
They
have
addressed
all
pain
and
comments,
so
I
think
this
is
let's
say,
addressed
solved
from
this
point
of
view.
I
have
just
checked
that
there
are
some
minor
editorial
fixes
that
would
need
to
be
added
in
order
to
comply
with
the
irtf
rfc
series
stream
guidelines.
A
So
this
is
essentially
in
the
abstract
and
introduction
text
to
mention
that
it
is
a
product
of
the
research
group
that
it
represents,
or
not
the
consensus
and
the
amount
of
reviews
that
the
document
received.
But
I
think
this
is
not
has
nothing
to
do
with
the
technical
content
or
value
of
the
documents,
so
this
is
something
we
could
address
in
a
future
revision
without
major
issues.
K
A
A
Okay
sure
it's
just
some
details,
but
I
think
this
will
be
spotted
by
the
rog
review,
but
it's
my
other
thing.
We
need
to
address
thanks,
so
if
there
are
no
further
feedback
or
comment
on
this
point,
let's
continue
with
what
we
have
left
in
the
agenda
so
shall
maybe
you
want
to
continue
with
the
next
slide.
E
A
So
I
mean
we
already
took
some
time.
This
was
a
bit
on
on
the
home
and
my
my
site
to
to
to
come
to
this
conclusion.
So
I
think
now,
in
terms
of
the
research
group
status
I
mean
jerome
could
confirm-
or
not,
but
at
least
from
my
point
of
view,
for
these
two
documents,
the
research
group
has
gone
through
the
I
mean
all
the
full
process
in
terms
of
reviewing
them
improving,
delivering
the
the
latest
version
and
making
them
stable.
A
A
D
F
A
A
D
D
Don't
have
so
many
things
to
announce
at
this
time,
but
so
we
will
continue
as
usual
with
monthly
or
here
we
have
our
channel
meeting.
Of
course,
we
think
we'll
not
have
meeting
next
time
next
month,
except
if
there
is
really
new,
but
we
have
to
see
all
can
organize
in
that
case,
but
mostly
we
for
the
next
video
meeting.
We
we
already.
D
D
We
hope
to
have
the
next
one
on
the
next
meeting
face-to-face
but
we'll
see
november
in
battery
and
as
usual.
Yes,
if
you
have
a
new
topic
that
you
want
to
propose,
please
let
us
know
we
can
arrange
that
in
one
of
our
next
meetings,
not
more
from
my
side,
though
I
don't
know
if
you
want
to
add
something.
A
No,
I
don't
think
we
have
much
use
for
for
the
research
group
status
beyond
the
the
document
and
the
future
meetings.
Let's
take
the
time
maybe
to
to
give
the
floor
to
to
the
other
research
participants
that
are
proposing
work
on
network.
I
mean
related
to
network
embedding
so
just
to
situate
a
bit
before
this
meeting.
We
there
have
been
a
couple
of
new
drafts
submitted
in
the
context
of
the
nmrg
and
both
were
addressing
virtual
network
function,
embedding
problems
from
different
angles.
Let's
say
so.
A
We
discussed
with
jerome
about
the
agenda
of
this
meeting
for
this
session
and
we
want
didn't
want
to
put
too
many
topics.
That's
why
we
suggested
to
the
offers
of
this
draft
to
dedicate
an
interim
meeting
to
explore
these
proposals,
because
to
our
knowledge,
virtual
network,
embedding
was,
I
mean,
is
a
very
interesting
problem
and
has
been
quite
explored
in
the
last
10
years,
especially
there
is
a
lot
of
literature
and
existing
solutions
onto
that.
A
D
I
I
So,
as
you
can
see
here,
we
tried
not
to
move
the
nodes
that
have
been
already
well
placed
and
just
move
the
nodes
that
must
be
moved.
Okay,
the
current
state
of
the
art
is
mostly
based
on
on
on
linear
programming,
so
they
are
hardly
able
to
manage
the
size
of
new
winter
network
and
physical
network
that
we
are
talking
today.
So
we
need
new
approaches
to
to
resolve
them.
I
So
what
we,
what
precisely
I
am
proposing,
is
to
somehow
define
one
genetic
approach
to
resolve
that
and
one
let's
say,
abstract
genetic
approach,
as
you
can
see
here
on
the
right.
What
I
am
proposing
is
a
very,
very
simple
genetic
algorithm,
but
we
have
a
different
ways
of
creating
new
configurations
that
are
respecting
the
previous
configuration
if
needed.
This
algorithm
is
begins
with
an
empty
configuration,
but
it
can
begin
with
the
current
configuration
of
the
network,
so
we
are
concentrating
in
evolving
it
and
genera
genetic
algorithms
are
best
for
evolution.
So
what
that's?
That's?
I
So,
in
summary,
there
are
new
circumstances,
new,
a
scenarios
that
are
not
well
covered
by
current
or
previous
approaches
for
network
embedding,
and
also
there
are
new
approaches
like
genetic
algorithms
that
are
giving
better
results
that
previous
approaches,
so
when
we
put
both
together
that
justify
that
currently,
you
know
today
is
a
there-
are
some
needs
for
this
work
inside
the
itf,
especially
in
the
ir
nmrg.
I
mean
this
research
aspect
of
the
of
the
network.
We
must
research,
this
kind
of
algorithm.
I
We
must
research,
for
example,
as
I
can
reference
this
of
what
kind
of
mutation,
breathing
or
creation
of
configurations
we
can
do
in
order
to
improve
the
resulting
configuration
of
the
network.
That's
my
summary.
I
hope
this
gives
some
kind
of
illustrative
or
illustration
of
what
are
my
intentions.
I
will
just
finish
so
order
can
talk.
Thank
you.
A
A
A
E
A
A
A
Okay,
I
don't
see
if
something
coming
back
just
so
we
we
will
address
on
the
mailing
list
a
a
doodle
to
find
a
good
slot
for
this
interior
meeting
dedicated
to
vna,
and
hopefully
we
can
have
inputs
from
these
two
drafts.
Our
idea
was
also
to
prepare
a
bit
offline,
as,
as
we
said
about
trying
to
understand
the
new
context
for
breaking
vane
into
the
scope
of
energy,
to
understand
what
will
be
the
research
directions
supported
by
these
proposals.
D
A
Yeah,
okay,
thank
you,
everyone
for
attending.
Let
me
check
one
final
in
the
participant.
I
don't
see
your
son
coming
back,
so
we
will
reach
out
to
him
and
involve
him
in
the
preparation
of
the
upcoming
interim
meeting.
Thank
you.
Everyone,
thanks
to
the
presenters
thank
to
the
offers
for
this
document
have
a
good
rest
of
the
day
if
it's
still
day
for
you
or
a
good
night
for
the
ones
that
stay
up
late.