►
From YouTube: IETF105-DMM-20190724-1000
Description
DMM meeting session at IETF105
2019/07/24 1000
https://datatracker.ietf.org/meeting/105/proceedings/
A
A
A
So
when
starting
a
presentation,
please
declare
any
idea
that
you
have
and
for
section
504
of
c30
978
47
48,
and
if
you
are
asking
questions
and
if
you
are
aware
of
any
IPR
policy,
please
state
the
same
you're
bound
by
this
saw
by
these
requirements.
Now
second
thing
is
essentially
note.
Well,
please
be
aware
of
the
IEP.
Are
policies
with
respect
to
you
know
how
the
current
conduct
that's
one
thing
and
finally,
some
you
know
blue
sheets.
You
know
please,
everybody.
A
A
The
link
to
the
working
group
is,
you
know,
on
the
posted
here
and
the
next
is
the
agenda.
It's
on
the
site
as
well,
and
a
couple
of
things
I
think
a
quick
review
of
where
we
are.
We,
the
first
document
that
these
are
the
working
group
documents,
a
quick
status
update
on
where
we
are
first,
one
is
the
user
plane
protocol
and
architecture
analysis
on
3gpp
Phi
J
system.
Alright,
this
is
Phi
GU,
plane
analysis,
it's
a
0,
2
version,
I.
Think
I've
seen
some
discussions
in
the
Mahler
and
I
think
it's
progressing.
A
It's
still
in
the
early
stages
right,
but
overall,
authors
are
serious
and
there's
a
good
amount
of
you
know.
Work
is
happening
right,
I,
think
yeah
working
on
participation.
We
need
reviews,
but
I
think
it's
okay.
So
far,
the
reviews
are
okay.
On
this.
The
next
document
is
segment
routing
ipv6
for
sr
v6
for
mobile
user
plain.
This
is
the
SR
v6
mobile
use
a
plain,
0
5
version
of
the
document.
This
is
almost
you
know,
coming
to
closure
right,
I
think
you
know
we
have
for
both
these
documents.
A
You
know,
there's
a
scheduled
in
a
presentation
slot
you're
going
to
hear
more.
You
know
more
on
these
documents
now.
Moving
on
to
the
FPC
document
earlier
FPC
document
was
one,
but
last
time
there
was
a
view
that
the
FPC
document
should
be
split
into
that
the
yang
model
should
be
because
many
people
felt
this
is
a
you
know.
Document
in
the
size
of
two
hundred
pages
is
unreadable
right.
That
is
the
feedback.
Even
I
tried
to
reach
out
for
some.
A
C
Question
so
like
this
is
kind
of
like
a
I
kind
of
pushed
on
this
a
little
bit
because
I
think
the
potential
review
pool
for
these
two
are
different
and
I.
Think
somebody
can
look
at
the
document
like
with
like
a
move,
mobility
perspective
and
then
like
sign
off
on
it,
and
then
we
can
see
if
the
yang
model
matches
this
in
what
we
can
put
that
in
the
groove.
C
D
C
Read
this
thing
right
and
I:
don't
think
everybody
has
to
read
the
yang
module
right
like
that
slide
has
to
do
the
review
separately,
see
if
the
yang
is
well
formed
in
one
day
and
well-designed,
and
all
we
have
to
do
is
check
if
what
we
want
to
do
is
good
and
if
the
yang
model
matches.
So
that's
like
three
different
things.
We
can
do
independently
that's
right
and
of
why
I
was
trying
to
push
for
this.
Okay.
C
A
Absolutely
I
think
based
on
our
discussions
last
time
it
already
it
that
process
that
step
already
took
place,
yeah
so
yeah,
so
essentially
that
the
FPC
document
now
got
split
into
two
documents.
One
is
the
the
base
specification
and
the
young
yang
pod
is
separate,
which
is
the
yang
for
protocol
for
FPC.
Now
that
the
document
size
are
somewhat
manageable
and
you
know
each
with
a
different
focus
or
it's
somebody
with
some.
You
know
understanding
familiar
with
the
yang
structure,
semantics
they'll
be
able
to
review
the
yang
thing
with
more.
A
You
know
from
the
point
of
view
of
the
semantics,
and
all
of
that,
so
is
it.
This
is
a
good
move
in
our
opinion.
So
now,
hopefully
we
should
be
to
move
them
forward
right,
the
FPC.
Unfortunately,
the
FPC
document
in
the
process
you
know
has
to
be
republish
because
it
also
has
the
yang
stuff
that
needs
to
be.
A
You
know
I
think
now
we
are
waiting
for
Charlie
to
North
us
to
no
posting
new
version,
so
so
this
is
the
these
are
the
documents
now,
the
other
three
other
documents
which
are
completed
working
a
blast
call
on
is
the
the
distributed
mobility
anchoring
and
the
proximal
abusive
extensions
for
DMM.
Both
are
you
know
we
have
completed
the
we
have
sent
it
to
Suri
if
they
or
ad
it's
in
a
TQ,
yeah
Suresh
yeah.
C
So
I
was
hoping
to
get
this
done
before
the
idea,
but
we
just
came
in
it
just
a
bit
too
late
for
me
to
review
it.
So
I
hope
to
get
this
both
of
them
out
of
my
queue,
probably
in
within
the
next
two
weeks,
at
least
ad
evil
out,
okay
and
hopefully,
there's
not
much
to
fix,
but
once
it's
done,
I'll
set
up
E.
So
I'm
not
gonna.
Wait
for
an
interview
on
this
because
I
think
like
I,
can
take
care
of
this
lecture
of
myself.
C
Mobility,
it's
just
by
waiting
some
discuss
resolution,
so
I
managed
to
have
some
discussions
on
that.
There's
probably
something
to
do,
but
it's
not
clear
what
it
is
so
I
do
like
sit
down
with
like
Danny,
hopefully
probably
like
discuss
like
you
know
some
ways
to
address
things,
but
it's
I
probably
won't
be
able
to
do
it
till
Friday,
okay,.
E
C
A
C
C
C
A
B
B
Here's
summary
of
the
update
from
fourth
version
to
fifth
version,
the
naming
the
change
the
the
commented
on
the
mailing
list.
I
took
the
naming
for
the
the
transformation
function
from
ipv4
gtp
to
SR
basics
from
TM
team
up
to
TM,
gtp
40.
It
is
not
clear
what
the
function
looks
like
and
next
one
is
to
reflect.
The
comment
on
the
Marianist
which
the
ue
session
is
unclear.
B
So
I
we
also
to
take
the
the
terminology
of
PDU
system
instead,
so
name
is
change
as
done
so
next
seeds,
status,
mapping
rule
have
been
changed
to
reflect
the
new
format
to
adopt
a
new
argument
to
put
the
mobile
session
mobile
mobility
session
information
in
the
sit
so
I'm
in
detail
the
fit
format
of
the
ad
GT
and
MGT
p6,
and
for
e
in
the
direction
of
the
SL
basic
to
GT.
Gt
you
to
align
the
format
of
the
odd
mob
session.
That's
one!
B
B
B
So
during
the
hackathon
we
figure
out
that
we
need
one
more
function
to
support
drop-in
mode,
so
the
drop-in
mole-
it's
not
described
this
document,
but
let
me
explain
a
little
bit
idea
is
the
cut
pocket
comes
into
the
GDP
node
gtp,
node
and
capital
pocket
by
a
GTP
hitter.
Then
a
service
node
transferred
gtp
packet,
tracer
basics.
This
is
the
normal
case,
but
in
drop-in
mode
that's
the
it's
a
basic
packet
transits
again
to
the
gtp.
You
that's
introduced
a
service
user
friend
with
no
impact
to
the
Assad
gtp,
the
past
end-to-end.
B
B
B
One
of
the
use
case
is
to
put
some
function:
it's
not
the
the
CVP
Network
function
it
to
be
kind
of
the
arbiter
of
function,
Riker
so
called
like
the
function
deploying
so
far.
It's
in
six
interface
beyond
that,
but
once
the
packet
translates
to
SRB
six,
all
of
the
function
deploy
so
far
in
the
m6
is
located
on
the
in
the
middle
of
the
path
of
the
mobility
path
that
we
1yd
another
is,
of
course,
the
traffic
engine.
It's
easy
to
see
and
make
sure
the.
C
Like
I
understand
why
it's
getting
converted,
but
I'm
just
worried
about,
why
are
it
because
the
insects
you
can
just
pump
out
the
IP
packet
right
right
right
next
I'm,
just
wondering
like
you
know
what
is
a
user,
so
the
two
things
I
thought
of
was
like
either
you're
having
a
hierarchical,
UPF
right
because,
like
the
other
one
might
be
true,
GT
will
UPF
right
and
the
second
almost
roaming.
Those
are
the
two
things
I
was
thinking
of
is.
Is
that
like
similar
to
what
it?
B
Idea:
okay,
okay,
so
on
please
so
I
hope
to
clarify
the
major
and
minor,
including
minor
Chris
comment
from
the
mailing
list.
So
to
make
sure
the
all
of
the
the
comment
would
be
appropriately
reflected
to
the
draft.
So
in
the
in
the
end
of
the
this
work,
I
would
like
to
ask
the
the
working
group
last
call.
B
We
can
have
discussed
shown
like
the
dropping
mode
almost
covered
by
the
existing
function,
except
the
one
Gigi
tp6
thought
did
the
eye,
because
the
GT
p
6d
doesn't
put
the
endpoint
of
the
gtp
you,
but
back
the
gtp
header
from
the
deepest
abyss
to
gtp.
So
in
for
the
GT
p6
case,
we
need
the
original
and
the
point
others
of
gtp
use
so
that
others
should
be
preserved
at
the
ends
and
the
last
sheet
in
the
SR
header.
So
that's
why
we
need
to
define
the
new
function
to
cover
this
drop
in
mode.
B
Another
point:
it's
whether
we
preserve
the
UDP
source
port,
that's
original,
to
make
sure
the
transparency
of
the
original
gtp
parkette
I,
don't
think
we
need
to
preserve
original
UDP
source
port
because
he
gtp
U
is
one
of
the
unidirectional
atonium
protocol.
It's
not
needed
to
to
define
the
tunneling
session
by
using
source
port,
but
it
makes
sure
that
some
stray
bit
of
the
road
balancing
by
using
source
port
entropy,
but
it
it
not
be
requirement
to
keep
the
original
source
port
transparently.
We
can
calculate
the
hash
to
put
this
or
UDP
sources.
B
Certain
number
of
people
think
dropping
mode.
Is
it's
useful
visible?
I
I
was
happy
to
work
on
that
and
also,
if
someone
doesn't
understand,
doesn't
find
the
any
goodness
to
cover
the
drop-in
mode.
I
think
we
can
keep
the
raft
as
it
is.
So
please
comment
on
that
now
or
after
the
presentation:
okay,
I.
B
B
What
is
B
so
discussion
here
is
unchanged
in
LB
scenario
should
be
a
case
at
least
the
augmentation
to
the
use,
because
the
entry
into
NSR
basics
user
plan
case
are
not
necessary
to
use
the
augment
system
if
the
seat
count
indicates
a
quest
for
an
session,
but
in
case
that
once
it
agree
a
once
it
aggregates
the
multiple
PDS
system.
Augment
system
also
help
to
avoid
barrage
of
acid
former
that's
discussion.
So
here
we
propose
the
resolution,
its
add
text
to
clarify
the
argument.
B
Augmentation
is
required
in
case
that
one
said:
I
agree
it
much
PPD
session.
That's
why
prop
order
please
next
so
Nick
so
next
slide
also
the
augmentation.
So
the
comment
from
menus-
it's
please
elaborate,
dissipate
more,
since
this
function
is
likely
not
to
be
instantiated
party.
This
session
argument
session
helps
the
UPF
to
perform
the
function
which
require
public
QP
q,
fi
and
o
video
session
granularity.
So
this
discussion,
it's
worth
it
I,
would
get
multiple
video
session
down
to
the
same
policy
argument
session,
help
to
indicate
which
PD
session
packet
will
apply.
B
That
course
such
a
usage,
they
bought
a
us
at
rest,
etc.
So
proposal
suggests
illusion.
Here
is
the
add
some
text
like
in
case
that,
once
it
aggregate
multiple
video
system
to
the
above
sentence
above
sentence
means
since
the
is
a
dysfunction
they're,
actually
not
to
be
instantiate
blah
blah
blah.
So
this
is
sad
point.
So
next,
please,
let's
repeat
the
next
point:
is
it's
our
policy
in
the
section
5
dos
3.1
interworking,
with
ESL
basics?
Gtp?
The
comment
is:
how
does
the
assault
gateway
run
citrus
to
DA
a
distinction?
Others?
This
must
be
positioned?
B
Doesn't
it
so
discussion
here
is
in
this
case
the
DA
receiving
GTV
pocket
is
a
seed
of
the
end
omgt
p6
d
associated
to
a
citrus
described
in
this
section,
6.3
the
deceit
Calgary
at
much
perfidious
system
so
that
it
shouldn't
be
position
basis.
The
proofs
all
proposed
resolution
aids
the
modify
the
text
on
the
sole
bread
to
align
with
section
6.3.
That's
our
proposal.
So
next
one.
B
Please
so
next
one
is
the
terminology:
the
anchor
the
comment
is
the
draft
seemed
to
use
the
term
anchor
or
anchoring
quite
a
lot,
but
what
exactly
is
means
it's
much
unclear.
It
looks
to
me
that
the
meaning
change
during
the
progression
of
techies
and
depends
on
the
context
as
it
appeared
to
be
such
key
term
of
the
door
I,
would
recommend
that
you
define
it
somewhere
in
this
document,
a
bit
of
fear
and
discussion.
B
A
Cushion
under
directory,
so
if
you
look
at
the
definition
of
mobility
anchor
as
we
and
based
on
our
prior
work,
our
mobile
IP
v4
v6
proximal
ipv6
anchor
is
a
topological
from
routing
point
of
it's.
An
anchor
of
the
IP
address
right
so
ease
the
service
expect.
Is
it
changing
that
those
the
the
fundamental
aspect,
or
is
it
defining
a
new
behavior
or
is
it
still
an
IP?
On
the
ID
point
of
view,
the
U
is
such
an
IPS
anchor.
That's
a
topological
entry
and
exit
point.
B
D
B
Teammate
that
would
be
applicable
for
the
the
the
other
all
of
the
the
mobility
networking.
How
does
that
I
think
we
need
to
cover
not
only
for
I
Bianca,
but
also
the
access
network
anchoring,
like
s
gateway,
mag,
something
like
that.
So
in
current
terminology,
in
IETF
it
doesn't
cover
the
such
kind
of
the
access
area
on
college.
So
I
could
cover
that
things.
B
A
I
think
the
way
I
look
at
it
is
anchor
also
anchored
doesn't
mean
it
ties
to
a
specific
protocol
or
people.
You
know
my
you
know
all
right,
I
think.
The
way
we
said
was
in
the
past
anchor
means
the
the
u.s.
IP
address
is
topologically
configured
on
that
node
from
the
IP
point
of
view,
that
is,
for
the
rest
of
the
world.
That
IP
address
belongs
to
that
node,
or
at
least
that's
the
entry
point
that
for
reaching
that
IP
address
now
LM
a
is
a
form
of
an
element
when
we
defined
a.
A
What
we
said
was
LMA
is
also
an
IP
anchor
or
home
agent
is
also
an
IP
anchor.
So
that's
how
we
put
it
in,
but
now
I
think
your
definition
of
IP
anchor
or
whatever
in
the
context
of
this
document.
We
need
to
maybe
discuss
this
and
see
how
it
fits
the
existing
definition
or
existing
understanding
and
how
what
we
are
modifying
here.
The.
D
B
G
D
This
draft
is
not
going
to
modify
the
functions
of
those
UVs,
the
signal
weighting,
the
data
cavity.
Yes,
the
actual
function
of
those
UPS
defined
by
the
3gpp
architecture
and
you're,
not
causing
any,
of
course,
regardless
of
earlier
very
going
with
the
IP
and
arcane
cars
of
mobility
and
cars
of
service
and
cars
doesn't
matter,
they
all
knew
yes,
Louis
and
we
are
all
versus
reading
each
other.
Yes,.
H
G
A
But
one
issue
with
that
approach:
reduction
with
UPF
is
a
3gpp
Tom
all
right
in
general.
You
know
idea.
We
don't
have
that
definition
right.
So
maybe
we
can
say
anything,
but
as
long
as
we
provide
a
proper
definition
of
or
functionally
what
it
is,
if
somebody
looks
at
it
from
3
or
IP
point
of
view,
if
you
can
say
UPF
is
so-and-so,
then
we
are
good.
D
Monitor
but
again
like
one
one
way
we
can
actually
generalize.
This
is
basically
where
the
anchor
point
to
me
at
least,
is
a
function
that
we
are
providing,
whether
it
is
a
policy
enforcement
function
or
something
like
that
right,
that's
a
rule
of
anchor
point,
so
you
want
to
actually
go
through
every
single
type
of
banker
that
you
can
find
and
basically
wanted.
You
skates
in
this
draft
fair.
C
D
D
A
I
I
think
anchor
is
a
Isabel,
well-defined,
terminus
meaning,
and
it
goes
beyond
that
traffic
traverses
that
particular
point
to
reach
UE.
So
it
has
more
policies
for
charger
with
and
monitoring
whatever.
So
since
we
talked
about
traffic
offload
low
traffic
of
functions,
so
that
term
doesn't
hold
it
exactly
anymore,
because
not
all
traffic
in
packets
need
to
traverse
that
kind
of
view.
I
So,
in
order
terms
of
the
men,
we
also
try
to
look
for
alternative
technologies
and
came
up
with
the
DP
and
EPA,
which
was
a
bit
of
an
anchor
I,
did
open
note,
so
that
could
be
a
node
or
then
Gateway.
The
terminus,
GDP
and
second
routing
kind
of
extension,
Harris.
So
I
think
we
can
go
around
these
three
terms,
all
right,
but
we
need
a
clear
definition.
I
A
J
So
we
writing
our
document.
One
of
the
3gpp
see
people
tell
you
get
corrected
me
on
this
part.
What
I
would
request
is
a
minor
comment.
Don't
change
the
definition
of
what
3gpp
does
in
this
fairness?
We
don't
have
to
redefine
that
stuff.
For
example,
the
PD
assertion
anchor
generally
the
original
P
gateway,
NTA
deployments
and
the
UPF
slot
of
Oh
said
one
is
the
IP
IP
address
assignment
to
the
UE
and
also
bitrate
enforcement,
lawful,
intercept
all
this
stuff
happens.
There
are
new
functionality
in
the
Phi
G,
which
is
introduced.
J
A
B
Next
one
is
the
TTP
message
support,
so
the
comment
AIDS,
we
assume
message
type
is
TTP
message:
TPD
you
and
mrs.
type
number
two
five,
oh
five
and
cannot
transfer
gtp
PIN,
which
used
gtp
message.
Echo
message,
type
one
and
deeply
message
echoed
primates
type:
two.
Should
we
add
capability
to
translate?
Gtp
echo
I
am
pissed
ICMP,
six
or
add
stored.
It
be
message
type
in
some
place,
so
discussion
middle.
B
We
will
cover
gtp
message
in
this
world
or
not
where
we
agree
to
cover
in
the
marker
is
a
random
type
of
the
gtp
message
in
another
documents.
So
proposal
resolution
here
is
cupboard,
III
message,
including
and
marker
in
another
document.
Yes,
I
really
appreciate
the
someone
frontier
to
write
down
that
document.
I.
A
B
So
next
exercise
please,
so
we
done
the
the
major
point
and
we
categorize
the
list
of
the
commented
minor,
but
someone's
easy,
a
measure
or
not.
But
let's
qualify
the
minor
comment
in
section
five,
one
one
uplink
packet
flow
into
Russian
mode.
The
comment
is:
what
is
a
spec
table
where
the
lookup
it
to
be
done?
Does
this
mean
that
to
support
more
is
a
brain?
There
need
to
be
an
additional
two
cup
table.
A
discussion
here
is
this
mentioned
typical
behavior,
which
took
up
Phebe
to
find
next
up
between
reading
usual
so
proposed.
K
B
One
please
in
Section
five
one,
two
downlink
packet
flow
interaction
mode.
In
this
case,
you
probably
need
a
additional
lookup
table
to
map
the
destination
address
of
the
ue,
with
others
seed
with
the
gene
lb.
The
discussion
is:
there's
no
need
to
look
up
the
you
ers
at
all,
so
that
she'd
encode
a
surf.
The
society
so
proposed
resolutions
keep
the
current
a
gift.
B
That's
really
straightforward!
Next
one,
please
section
five,
two
five,
two
one
uplink
packet
flow
in
hassle
mode.
The
comment
is,
what
is
others
V
shouldn't
it
be
others
D
what
it
t
ID
used
in
this
case,
so
discussion
here
is
address.
D
a
means!
Actually,
you
you
goddess,
eat
to
mapping
with
a
be
as
the
eye
for
web.
B
A
B
Next
slide,
please
section
5,
2,
enhancer
mode,
the
comment
aids,
how
much
overhead
does
the
use
of
the
multiple
seed
introduced
and
what
is
the
impact
to
hit
the
compression?
The
discussion
here
is
which
each
seed
had
16
bite,
but
we
can't
state
exact
side
above
ahead.
Do
it
depends
on
the
amount
of
the
seed
used
for
traffic
engine
also
support
programming
when
it
comes
to
header
compression
sr
b6
is
a
brain?
Doesn't
impact
lock
on
radio
interface,
so
proposal
resolution
is
keep
the
current
a
gist.
B
Maybe
in
the
spring
working
group
in
this
time,
I
discussed
some
sort
of
idea.
That's
it
compression
solution.
He
came
into
that.
So
next
one
is
six.
Twelve
five,
three
one
one
uplink
packet
flow
in,
inter
working
with
ipv6
DTP.
The
comment
rate
dated
one
instance
of
the
end
and
gtp
60-seat
papi.
Do
you
type?
How
is
the
PDO
type
in
run
by
use
of
tid?
The
discussion
here
is
dt
p
6d
seat
would
be
instantiate
puppeteer
session
type
basis.
The
proposed
resolution
is
keeps
us
current.
The
kissed,
the
next
slide.
B
D
B
B
So,
on
we
cross
to
end
five
three
one:
three
scarab
is
a
winter
walking
we
are
gtp,
6t
Eid
is
covered
by
the
Geno's
be,
and
you
tf2
same
tid
may
appear
for
different
genome
B's.
How
the
gtp
echo
works
for
these
cases
discussion
ways
as
the
last
sheet
indicate
generally
address.
It
makes
clear
that
Geno
don't
be
scoped
in
which
the
tid
is
kept
unique,
so
proposal
resolution
is
keep
the
current
exist.
B
Okay,
okay,
so
end
them
up
in
Section.
Six
to
is
a
mapping
tab
user
to
step.
One
Mobile
is
a
friend
specific,
or
is
it
general
for
any
SR
basics?
If
only
for
more
is
a
friend?
How
is
the
populated
discussion
waits?
The
segment
can
be
used,
dinner,
general
curry
for
NES
a
basic
functionality.
However,
in
the
context
of
a
mobility,
a
unique
and
mapped
segment
will
be
instantiated.
The
proposed
solution
is
how
to
populate
the
mapping
table
is
out
of
the
scope,
so
on
next
one
I
think
would
be
respond.
B
It's
also
the
end
map
in
Section.
Six,
two
to
the
comment.
Wait.
What
is
the
segment
that
is
in
line
three?
Is
it
segment
list
of
the
original
packet
or
is
it
new
segment
list
of
the
end
map
sorry
function?
Discussion
eight
signal
is
here
in
pseudocode
should
be
citrus.
The
new
mapped
seat
is
bound
to
the
citrus.
The
proposed
resolution
is
add
some
text
to
clarify
the
Bob
okay.
B
So
that's
all
we
qualified
we'd
like
to
clarify
the
comment
and
the
resolution,
so
our
next
shot,
a
a
next
step,
is
working
with
rascal.
After
when
we
reflect
see
a
substantial
resolution
to
the
draft
and
also
we
need
implementation
otherwise
mentioned
at
first,
we
already
have
a
PPP
code
and
also
the
p4
code
as
well,
by
kentaro
I,
but
we
need
more
implementation
and
I
also
be
continued.
The
hakusan
project
on
the
the
next
ITF
meeting.
So
if
you
are
interesting
to
Hakkasan
to
be
joined,
a
please
calm.
Thank
you.
That's
all
thank.
A
J
Hello,
everybody,
my
name
is
Uma
I'm,
going
to
talk
about
transport
network
of
air
mobility
for
Phi
G.
We
going
to
talk
about
the
new
addition
we
made
to
the
original
proposal,
so
first
I
will
just
lay
out
the
what
what
problem
that
is
being
solved
and
what
the
new
addition
that
we
made
is
will
be
talk.
My
colleague
John.
J
If
you
are,
if
we
are
sending
a
PD
you
session
for
the
slice,
you
need
to
make
sure
that
SLA
guarantees
for
the
end-to-end
has
to
be
observed.
Whether
the
transport
network
will
give
you
that
are
not
that
need
to
be
factored
and
during
mobility
from
G
note,
B
to
G
note:
B
are
from
UPF
to
UPF.
You
need
to
make
sure
that
transport
network
also
offers
the
same.
Qos
characteristics
for
the
end-to-end
budget
need
to
be
factored
so.
This
is
the
missing
part.
So
how
to
address
this
is
being
described
here.
J
There
are
two
approaches
described.
The
second
approach
is
added
recently
by
John
and
the
second
aspect.
What
we're
discussing
is.
There
is
no
clear
standard
mapping
function
to
map
this
ran
and
coordinate
work
to
the
transport.
So
there
are
various
ways
for
slicing
early
slicing
solutions
in
2016
for
the
LTE
networks.
There
are
ways
people
are
used
to
use
to
do
this.
There
are
only
two
batters
there
how
they
map
to
the
PD
sessions
to
the
underlying
transport.
J
There
are
various
ways
people
pay
people
are
doing
in
national
deployments,
so
we
were
thinking
to
do
a
standardized
mapping
which
is
applicable
to
Phi
G,
also
because
a
lot
of
parameters
are
changed
and
the
cue
of
Phi
is
inside
the
gtp.
U
extension
header
so
creating,
as
I
said,
the
creating
a
reference
architecture
to
integrate
the
transport
backhaul
network
in
Phi
G
service
based
architecture.
J
So
this
integration
will
help
you
get
the
transport
network
functions
exposed
to
the
mobility
functions.
If
you,
if
you
want
to
mobility,
functions
to
other
network
functions
in
the
3G
control
plane,
if
you
want
the
transport
functionality,
whether
this
particular
transport
slice
can
give
this
much
QoS
at
this
much
latency.
That
can
be
based
on
that
the
traditions
can
be
made
so
under
and,
as
we
said,
the
providing
clear
mapping
function
to
integrate
video
sessions
to
the
underlying
transport.
Also,
this
describes
a
new
path,
routing
mechanism
that
is
called
preferred
path
routing.
J
J
Service
mapping
also
like
suit,
are
explained
where
it
can
be
replacing
gtp,
but
traditionally,
transport
or
traffic
engineering
do
only
the
underlay
transport
traffic
engineering
that
is
like
RSVP
TAS
are
MPLS,
so
PPR
is
a
new
transport
mechanism
where
it
can
do
a
path
steering
without
any
overhead,
and
it
also
additional
benefits
PPR
brings
in
and
how
this
fits
into.
The
Phi
G
mobility
has
been
explained
in
this
document,
that
is,
to
validate
the
reference
architecture.
J
We
put
forth
how
this
mechanism
can
be
used
with
PPR,
but,
however,
the
reference
our
HR
put
forth
can
be
like
to
any
te
technology
existing
in
the
idea
from
rsvp-te
to
SR
MPLS
to
SR
v6
as
a
pure
transport
underlay,
not
the
service
programming
and
the
pure
transport
under
layer
service
six
can
be
used.
So
that
is
explained
here.
So.
A
J
F
J
L
L
J
J
That's
it
the
changes,
the
major
change.
What
we
introduced
was
we
come
up
with
one
approach.
First,
in
the
first
three
sections
of
us
three
versions,
then
later
on,
John
frum
took
it
from
the
different
point
of
view,
and
he
said
that
we
have
to
minimize
the
changes
in
the
architecture.
So
we
talking
about
this,
but
these
are
unresolved
issues.
That's
why
we
want
expose
this,
and
we
won't
eventually
image
these
two
approaches
into
one
approach.
Eventually,
that's
so
that's
where
we
asked
for
working
group
adoption.
We
are
not
ready
yet.
E
J
So
it
doesn't
touch
MPLS
layer
if
one
of
the
de
technologies
we
try
to
do
whatever
the
IETF
has
developed.
So
many
te
technologies
right
from
rsvp-te
to
s
our
MPLS
tests,
our
v6
and
the
PPR
we
are
working
on.
We
don't
want
to
touch
anything
there.
We
were
just
seeing
in
the
PPR
how
we
can.
The
reference
architecture
can
be
applied
to
the
PPR,
the
new
path,
routing
mechanism.
We
introduced,
that's
what
we
are
trying
to
do.
We
are
not
changing
anything
in
the
MPLS.
M
J
That's
good,
so
qci
is
an
LTE
LTE
deployments.
That's
that's
it's
that
that's
where
exactly
in
some
deployments,
based
on
the
QCA
and
qsr
functionality,
mapping
is
happening.
In
some
cases.
People
are
reading
judy
GTP
you
once
the
GTP
encapsulation
happens,
UDP,
so
space
mapping
are
being
used.
Udp
source
based
mapping
to
gia
underlying
rsvp-te
tunnels
are
sr.
Mpls
tunnels
are
being
used.
There
are
so
many
ways
people
are
deploying.
We
just
trying
to
see
whether
we
can
standardize
this
for
Phi
G.
There
is
a
q5
field
in
the
GTP
header.
J
M
M
C
J
C
The
user
of
that
would
be
like
somebody
working
on
the
Phi
J
architecture
right
so
I'm,
just
thinking
like
how
do
we
communicate
this
exists?
And
how
do
we
come
in?
How
do
you
find
out
if
this
meets
their
needs
right
and
a
couple
of
things
right
like
it's,
like
a
nose
ring
before
right
like
so
what
what
we
can
do
is,
like
you
know,
send
earliest,
instead
in
saying,
there's
existing
right
and
but
but
again
like
you
know,
I
I
keep
repeating
this
for
a
lot
of
things.
C
Somebody
on
the
3gpp
side
needs
to
do
it.
You
said
you're
talking
with
somebody
at
cd4,
right,
I
think
this
also
needs
to
be
worked
on
in
3gpp
to
include
this
and
the
second,
so
one
of
them
is
earliest
and
is
something
we
can
do,
and
the
second
thing
is
like
what
you've
done
in
the
past,
for,
like
previous
generations,
like
Ettie,
like
what
we've
done
is
like,
we
had
a
dependency
less.
So
this
is
a
differential
s.
C
There's
like
a
it
used
to
be
a
spreadsheet,
but
now
it's
like
more
formal
way
of
like
tracking
this
stuff.
We're,
like
you
know
if
3gpp
is
interested
in
this,
like
it
goes
into
something
we
track.
So
we
had
a
coordination
meeting.
I
was
there
in
the
meetings
on
on
Monday
gi-give
board
meeting.
So
the
idea
is
like
this
goes
on.
The
dependency
list
so
like
like
3gpp,
knows
the
status
of
this,
and
we
also
know
like
if
3gpp
is
interested
I
think
that's
really
the
interesting
thing
that
could
be.
C
J
A
So,
just
to
add
to
what
you're
saying
in
3gpp.
So
if
there
are
multiple
groups,
we
need
to
consider
there's
sa5
from
a
management
context
that
is
pretty
much
in
line
with
some
of
the
interfaces
here,
the
SA
to
work
on
the
architecture.
Those
folks
don't
think
that
anything
is
needed
at
this
point.
That's
not
like
so
there's
there's.
C
D
C
C
Person
to
talk
to
in
in
the
SI
groups,
but
I
I
can
ship
it
off
to
like
a
wide
distribution
as
far
as
potentially
interested
ship.
But
after
them
they
say
nothing
lot
of
times.
They
say
nothing,
that's
okay,
but
the
idea
is
like
to
be
cautious
and
say
like
send
it
to
the
largest
distribution
overall.
J
C
You
know,
I
cannot
tell
you
right
or
what
I'm
saying
is
like
it's
better
to
like
stop
early
because,
like
the
3gpp
process
can
happen,
and,
finally
because
it's
probably
like
a
three-month
delay
from
which,
when
we
start
something
it
has
to
3gpp.
So
we
had
to
go
through
a
meeting
cycle.
I
had
to
work
on
the
liaison
like
ship
it
off
and
just
like
a
3gpp
meeting
cycle
before
somebody
looks
at
it
right.
So,
whenever
you're
ready,
you
talk
to
thank.
J
C
What
I'm
saying
it's
like
when
you're
ready
to
like
take
this
one
step
further,
you
go
talk
to
receptor,
Sun
and
defying
right
and
they
pull
the
most
moved
consensus
and
they
ship
it
off
to
me
and
and
I
do
the
stuff,
because
there's
like
a
process
for
this,
and
it
all
takes
time
right.
So
thank
you.
Thanks,
okay,.
A
A
Think
so
so
then
I
guess
the
main
things
that
are
changed
are
in
Chapter
2.2
and
let
me
identify
what
exactly
that
is.
You
know,
so
it's
it's
a
mapping
between
the
3gpp
5
G's
slices
to
the
transport
slices
elaborating
on
that
a
little
bit
so
the
5
g
q,
s
and
slice
information.
That's
carried
in
NS
s,
AI
what
is
offered
or
is
offered
in
the
PDU
context,
to
a
ue
that
does
not
have
a
direct
relation
to
what
a
transport
network
provides
a
transport
networks.
Clients
are
the
3gpp
operators.
F
A
They're,
offering
slice
or
slice
services
to
a
3gpp
operator,
whereas
a
3gpp
PDU
session
are
slice
offers
to
a
Yui,
and
these
these
have
a
you
know,
end
to
em
kind
of
relationship.
It's
not
a
one-to-one
of
any
sort.
So
they're.
Definitely
so
we
are
mapping
between
these
two
just
to
take
it
a
little
bit
further.
If
you
take
those
segments,
for
example,
if
you
have
an
N
3
segments
that
may
be
provided
by
one
transport
provider
and
another
segment
may
be
provided
by
another
one
in
not
necessarily,
but
it
could
be
worse
versus.
A
A
The
point
in
B
is
related
to
I
mean
how
it
was
solved,
partly
in
4G
and
I.
Think
I
might
just
on.
This.
Of
this
discussion
came
up
with
the
MPLS
question.
To
part
of
it
is
that
what
we
are
trying
to
map
through
and
identify
here
was
probably
solved
within
a
node
which
terminated
the
3gpp,
as
well
as
a
transport
function
like
an
MPLS,
so
the
translation
was
done
within
a
node
and
it
was
okay,
but
as
was
identified,
insurance
Acamas
and
and
the
the
draft
in
on
user
plane
analysis.
A
G
A
So
you
know
we
need
a
reliable
means
to
convey
this
information,
so
any
examples
of
the
l2
issues,
just
yeah
I
guess
the
l2
issue
would
be,
for
example,
if
we
put
let's
say
a
VLAN
or
something
like
that
on
a
in
a
data
center.
What
would
it
mean
in
a
backhaul
network?
We
need
to
have
explicit
translations
for
all
l2
to
all
other
relatives
or
at
least
a
popular
actors.
A
So
that's
much
cleaner
is
to
have
a
means
by
which
you
can
convey
this
information
across
that
segment
and
that's
what
I
think
we'll
try
to
cover
in
this
point
here:
the
we're
introducing
a
mapping
identifier
or
a
context.
It's
a
mobile
transport
network
context
I
think
some
of
it
was
described
in
the
previous
slide.
So
it's
it
Maps
a
class
of
service.
A
So
this
is
generated
by
I
mean
this
mapping
can
be
done
ahead
of
time,
not
necessarily
when
a
session
arrives
and
it
is
generated
by
the
TNF.
Now
this
function
is
probably
in
in
a
part
of
the
3gpp
management
domain,
the
NSS
or
some
such
things,
which
we
don't
want
to
get
into,
but
I
think
the
essay
5
which
works
on
om
and
management
are
probably
looking
to
adopt
something
like
a
CTN
kind
of
work.
So
you
know
they
have
very
similar
specifications.
A
So
there's
that's
a
point
of
coordination
really
need
in
the
future,
so
how
it
generates
or
figures
out
which
slides
need
to
be
mapped
is
something
that
3gpp
should
solve.
We
only
say
here
that
there
is
a
mapping
that
needs
to
be
installed
in
the
transport
domain
and
it
serves
of
the
contract
between
the
two.
As
mentioned
earlier.
It's
not
a
one-to-one
Association,
because
that
just
would
not
even
scale,
and
we
don't
want
to
add
additional
delay
and
processing
when
we
want
to
establish
a
session.
A
There
must
be
an
estimation
of
what
is
needed
ahead
of
each
individual
session
establishment
and,
of
course,
we
want
an
identifier
that
scales.
So
the
way
this
one
scales
is
Oh
n
square
with
respect
to
the
number
of
sites,
but
not
I
mean
it,
and
that
n
square
is
much
lower
than
the
number
of
flows.
So
as
an
example,
if
we
have
three
traffic
classes
and
insights
and
we
end
25
sites,
then
we
need
only
about
900,
that's
just
an
example.
N
A
So
this
figure
is
trying
to
show
both
the
3gpp
functions
and
transport
functions.
Everything
in
that
blue,
not
a
box
but
house
rule
like
box,
is
part
of
3gpp
functions
and
I've
identified
three
steps.
So
I'll
try
to
go
into
sequence,
to
say
what
happens
it's
roughly
in
time
sequence
also,
but
not
exactly,
because
there
are
a
couple
of
control
loops,
so
the
TNF,
as
we
described
earlier,
may
be
part
of
some
3gpp
function.
We
only
care
about
the
functionality
that
it
interfaces
with
an
SDN
controller.
A
What
we're
not
showing
in
the
figure
is
how
an
SMA
for
a
UPF
figures
out
that
it
is
this
empty
and
C
that's
needed.
That's
their
would
be
part
of
3gpp
networks
and
hence
not
described
in
detail
here,
but
there's
some
way
that
an
SMF
figures
out
that
this
empty
and
C
can
be
associated
to
this
PDS
session
and
that's
what
is
then
m2
or
the
n4
interfaces
when
there's
a
PD
you
session
establish
not
so
so.
This
is
all
happening
before
a
traffic
flow
starts.
A
So
at
the
point
that
a
traffic
flow
starts,
the
you
P,
you
use
a
plane
nodes
and
just
for
reference
that
you
supply
nodes.
Are
the
G
note
B
and
a
UPF,
so
it
could
be
n,
3
or
n
9.
In
this
case
they
already
have
the
empty
NCI
teen
identifiers,
and
at
that
point
the
UPF
is
going
to
map
the
empty
NZ
identifiers
and
send
it
out,
along
with
the
PD
accession
packet.
A
The
PE
router
is
going
to
inspect
that
MTN
C
identifier
and
deliver
the
resources
that
were
configured
earlier
and
all
of
that
is
provisioned
as
and
one
of
the
techniques
we
may
use.
As
you
know,
the
PPR
that
Omar
described
earlier
in
terms
of
how
that
transport
network
is
going
to
deliver
those
resources
or
you
know
how
it
is
being
routed
and
so
on.
A
So
one
of
the
problems
is:
if
the-
u
PMF
adds
this
identifier,
there
must
be
some
way
to
carry
it
in
an
IP
packet
in
each
IP
packet
and
as
mentioned
earlier
layer,
two
ones
are
not
suitable
because
you
may
cross
multiple
networks.
Gtb
extension
is
really
quite
bad
actually
in
this
case,
because
it
would
take
a
long
time
to
inspect
where
that
MTN
CID
is,
if
it
is
chained
in
a
gtp
extension.
So
actually,
that's
not
a
good
idea.
A
Other
exchange,
like
dscp
or
ipv6
flow
labels
are
not
immutable
and
hence
not
suitable.
Also,
we
are
considering
a
few
here.
It
could
be
gue
or
SR
v6
and
one.
The
flow
here
shows
the
gue
extension
to
add
the
MTN
CID.
So
essentially,
if
you
have
a
packet,
the
PE
router
sees
with
this
empty
and
CID.
It
adds
the
appropriate
forwarding
information.
That's
needed
I
presented
this
yesterday
and
T's,
and
one
of
the
comments
we
got
was
to
also
consider
the
option
of
adding
a
fast
ticket,
and
that
is
something
we
will
look
at.
A
A
Will
it
impact
when
you
guys
and
do
this
mapping
here
right,
I,
think
this
is
an
area
where
the
ITF
has
to
do
things.
I,
you
know,
will
have
to
specify
things
here.
The
other
things
I
suppose
are
more
information,
I
mean
or
some
other
things
are
more
informational,
but
here
they
mean
3gpp
or
anybody
else
would
have
to
rely
on
at
least
partly
IETF
specifications.
So.
J
One
of
the
problem
is
approach.
It
is
specified
in
the
version
3.
We
were
not
trying
to
change
anything
in
the
gtp
packet,
so
we're
trying
to
use
such
maybe
here
if
I
filled
in
the
chiquita
extension
header.
But
the
problem
is
in
this
scenario:
speed
router
has
to
delete
with
deep
into
the
gtp
extension
and
aftermath
of
the
transport
tunnel,
which
is
lot
of
routers.
Can't
do
that.
So
that's
why
this
identified
was
brought
in
so
which
will
easy
of
the
routers
to
laughs
quickly
to
the
transporters
right.
A
Yeah
so
yeah
I
mean
that's
a
point
that
I
missed.
Adding
you
know
about
the
qfi
I
mean
I.
Think
I
must
clarify
that.
Just
say
that
you
know
it
it.
It
can
provide
a
mapping
between
the
bearer
on
the
downstream,
which
may
be
of
some
use,
but
on
the
upstream
you
don't
have.
So
that's
that's
the
rationale
for
all
of
this.
A
A
So
if
somebody
says
John,
you
can
just
index
into
the
gtp,
you
is,
if
the
decoding
part
or
what
is
it
so
I
can
I
can
say
that
the
challenge
is
simply
that
transport
devices
don't
look
so
deep
in
to
figure
out
on
every
packet
or
even
if
it
had
to
it,
would
be
rather
expensive
to
parse
some
location.
That's
you
know
not
a
very
clear
point:
I
think
the
parses
are
going
to
be
looking
at
the
extension
header,
and
you
know
the
transport
would
be
able
to
look
at
autism.
B
Certain
mishima
so
on,
okay,
so
on
the
the
the
one
of
the
idea,
the
bits
are
bite
in
the
header
to
put
the
empty
CID.
The
I
think
you
already
mentioned
that
yes,
L
v6
at
one
of
the
transport
with
the
use
of
play.
But
what,
if
we
put
the
MTS
CID
in
this
a
PC
user
brain
part,
I
mean
the
seat
can
be
indicate
to
the
NTSC
kind
of
things.
I
think.
B
A
A
Think
that's
more
or
less
what
we
have
so
at
this
point,
we're
looking
for
comments
and
suggestions
and
we'll
have
to
work
as
OMA
mentioned,
to
unify
these
different
approaches.
In
particular,
I
think
we
got
some
feedback
on
the
slide
7
from
T's,
but
you
know
would
be
happy
to
get
more
input.
I
think
Eric
gave
us
some
input
about
the
12
bytes.
J
K
A
G
A
K
K
K
J
G
K
B
Any
comment
on
the
that's
the
discussion
on
when
we
could
see
some
do
information
on
the
ipv6
to
indicate
the
yes
wise.
We
I
think
the
SMF
need
to
control
the
end
point
of
the
year,
the
digital
session,
so
that
may
touch
the
change
of
the
HTTP
architecture.
I,
don't
think
it'd
be
fit
correct,
I'd
say
with
the
architecture
if
you
cut
the
top
semantics
in
ipv6.
That's
my
comment.
Okay.
B
So
my
my
another
comment
on
this
idea,
so
other
Amida,
you
mention
that
it's
not
touched
to
the
agrees
to
touching
using
architecture
to
be
changed.
It
must
be
a
much
more
generic
solution
to
see
the
pocket
to
map
the
to
the
specifics,
rising
think
it'd
be
not
to
mobile-specific
it's
much
more
general.
B
So
if
you
want
to
proceed
your
idea
to
see
the
yeah,
that's
mapping
solution,
so
I
think
it'll
be
e
to
touch
to
the
some
sort
of
the
VPN
solution
working
group
to
propose
your
idea,
because
it
kind
with
the
appear
Rooter
behavior
to
to
look
up
the
ID
to
find
the
appropriate
it's
rising
in
the
course
I
know,
I
think
that's
much
more
smooth
way
to
proceed.
Your
idea.
A
B
A
There
and
thanks
thank
you.
Thank
you,
John
any
final,
you
know,
that's
that's,
okay,
yeah,
so
I
think
feedback
again
I
think
are
it's
interesting
topic.
We
need
more
discussions
and
I
think
you
know
the
number
of
touch
points
here.
I
think
is
do
feedback,
and
all
of
that
is
Suresh
pointed
out
right,
but
I
think
it's
truly.
The
ball
is
in
the
authors
coast.
Really,
you
know,
do
more
work
all
right.
D
A
A
P
Q
H
H
We
added
a
thrice
service
type
in
the
section
4.1,
and
this
SSD
is
defined
by
sweetie
pp
and
it
represents
type
of
the
network's,
for
instance,
and
so,
if
we
define
the
for
type
four
types
of
the
standard:
SSD
EMA,
pu,
r,
LC
m,
io,
t
and
b
TX
and
also
a
it-
is
a
PCP
stage.
Information
and
you
are
areas
related
requirement
and
a
variation
aspect,
and
I
will
show
you
the
details
of
these
things
at
the
followings
right,
and
this
is
the
fuchsia
p
state
information.
H
A
H
F
H
Go
back
to
the
URL,
see
so
country
2005,
0
1.
You
find
the
three
types
of
redundant,
a
transfer
and
first
one
is
this
one
redundant
use
a
frame
pass
using
the
two,
our
connectivity
and,
as
you
can
see
from
this
diagram,
so
you
establish
a
two
redundant
video
session
and
to
the
data
network,
peers,
master,
master,
hang
and
the
PFR
and
of
be
at
the
stamp
around
and
ups
to
and
as
for
the
up
ringside
communication,
you
II
just
replicate
the
pocket
and
the
send.
H
This
kind
is
duplicated
pocket
to
the
Becerra
and
stand
wider
and
has
were
wrong
and
UPF
long
and
UPF
just
transfers
the
incoming
packet
and
say
a
same
in
same
way
as
a
noble,
impede,
PD
and
sorry,
no
normal
use
a
prank
or
great
body
and
in
this
case,
is
so
by
using
such
kind
of
redundant.
It's
also
if
there
is
issue
or
the
pocket
rolls
at
user
brain
path,
break
between
the
master
rank
and
data
network.
So
this
data
network
can
receive
a
URL
AC
services
pocket.
H
The
brocaded
package
choose
any
suit
only
one
and
su-22
and
after
the
UPF
received
such
this
pocket
and
UPF
will
eliminate
the
tunnel
pocket
and
sender
pocket
choose
data
network
via
NECN
six
interfaces,
so
in
this
case
is
if
there
is
pocket
doors
and
transport
network
pass.
For
example,
in
321
we
can
continue
the
you
are
irresistible
ease
by
using
the
answer
eternal
to,
but
of
course,
we
need
to
provide
disjoint
transport
network
paths
for
each
answer
eternal
and
respond.
H
H
H
So
in
order
to
in
order
to
distinguish
this,
if
we
get
pocket
at
the
receiver
side,
so
in
this,
and
as
was
the
gtp,
you
country
is
a
3gpp
considered
to
use
gtp
sequence
number
field,
which
is
will
be
encoded
in
basic
CPU
header
and
in
this
example
case
you
PF
without
a
duplicated
video
duplicate
the
income
pocket
from
Datanet
work
and
attach
the
same
sequence
now.
But
this
case
in
case
number,
one
sequence
number,
one
and
saint-jean
of
the
B
and
also
I-
think
this
is
a
complete,
complicated
scenario.
H
But
in
case
of
the
sad
sad
one
we
have
another
consideration.
So
in
this
case
is
that
the
synthesize
there's
no
difference
from
previous
one,
but
there
is
another
consideration
of
the
intermediate
UPF.
So
so
in
case
you
thought
video
session
on
Caillou,
PS
and
same
duplicated,
pocket
with
the
same
sequence.
Number,
and
also
so
yes
must
transparent,
refers
IP
packet
with
same
G.
If
you
never
choose
that
gob,
this
is
the
Cochino.
The
me
must
distinguish
the
repeatedly
placated
IP
flow
for
30
minutes
process.
H
H
A
B
N
Q
A
I
I
Right,
so,
very
briefly
about
the
background,
you
to
not
steal
and
repeat
what
we
see
you
time
and
repeat
what
we
discuss
already.
So
if
we
look
at
the
end-to-end
communication
between
you
ease,
which
can
go
beyond
mobile
phones
and
services
running
in
a
cloud,
there
is
a
3gpp
architecture
which
has
the
control,
plane
and
user
plane
functions
which
can
work
as
such
an
anchor.
We
discussed
at
the
beginning
of
the
session
and
other
UPS,
which
can
use
be
used
for
uplink
classification
and
traffic
offload.
I
So
on
this
particular
interface,
there
is
a
lot
of
work
being
done
in
this
caste.
That's
related
to
the
study
in
3gpp,
and
so
Teru
did
a
lot
of
work
here
to
discuss
alternatives
to
GDP.
What
we
are
focusing
on
here
is
an
interface
which
is
not
really
considered
in
in
3gpp,
because
it
just
assumes
plane
PDI
roarings.
So
this
is
the
n6
interface
between
the
previous
session
anchor
and
the
actual
service.
I
So
to
focus
on
this
use
case
for
some
because-
or
you
is,
let's
call
them
you
ease,
we
may
have
to
move
services
closer
to
the
vehicle
itself,
so
this
is
mobile,
edge,
computing
kind
of
directions
and
since
the
Mobile's
are
attached
to
a
cellular
architecture,
also,
these
vehicles
need
to
have
a
UPF
serving
as
previous
session
anchor
or
local
breakout
point
configured.
So
in
this
example,
we
have
a
UPF
which
relays
traffic
to
the
vehicle
from
a
central
cloud,
but
also
from
an
edge
cloud
here.
I
So
in
some
cases
in
particular,
if
we
want
to
maintain
low
latency
communication,
we
have
mobility,
probably
high
speed
mobility.
If
we
talk
about
highway
roads
where
cars
are
moving,
so
we
may
need
to
relocate
also
the
edge
cloud.
Second,
we
also
need
to
relocate
the
Associated
UPF
for
efficiency
reasons.
If
we
do
that,
we
deviate
from
the
plain
routing
rules
if
we
want
to
maintain
the
u.s.
IP
address.
I
So
the
scope
of
this
draft
first
of
all
describes
the
use
case
in
the
operational
aspects.
Second,
we
propose
how
such
data
plane
overlay
on
the
n6
interface
can
be
integrated
and
aligned
with
policies
in
the
5g
control
plane,
because
the
UPF
is
under
control
of
the
cellular
control
plane
and
then
six.
We
want
to
have
additional
controllers
to
enforce
the
the
routing
policies
here.
I
So
in
this
draft
we
want
to
look
mainly
at
some
architectural
and
operational
aspects
and
then
define
the
data
models
that
are
required
on
this
interface
between
the
controller
and
the
data
plane.
Note
second,
between
the
controller
and
the
5g
control
plane.
As
you
can
see,
you
will
leverage
a
standard,
3gpp
function,
which
is
the
application
function
as
a
binding
element
between
the
5g
control
plane
and
the
n6
state
opening.
I
So
one
of
the
key
concepts
is
that
we
consider
two
types
of
deployments
here
so
I
said
we
need
policy
enforcement
points
for
traffic
steering
on
the
two
ends
of
the
n6
interface,
so
one
is
always
aligned
with
the
data
network,
so
we
have
here
which
can
receive
and
enforce
policies
for
traffic
treatment.
On
the
other
sent
side,
we
have
two
options,
so
the
UPF
may
receive
the
upstream
traffic
treatment
policies
directly
from
the
mobile
core.
I
We
are
the
n4
interface,
that's
what
we
call
tight
coupling
currently
we'll
look
more
at
the
loose
coupling
options
for
M
EC,
which
you
can
see
later.
Why
we
do
that
and
here
basically,
we
keep
the
n6
traffic
treatment
policies
and
the
associated
control
independent
of
the
mobile
core,
which
has
some
advantages.
I
So
the
latest
update
was
unfortunately
only
before
I
kiev
104.
So
we
described
operational
aspects.
We
support
the
two
different
and
describe
the
two
different
options
loose
coupling
company
and
we
need
to
put
a
bit
more
energy
into
elaboration
of
the
mature
information
model.
As
we
said
during
the
last
ITF
meeting,
we
wanted
to
have
an
update
of
the
draft
with
a
little
bit
of
a
brush
content,
because
we
have
a
lot
of
description
in
so
we
want
to
clean
up
that
and
drive
the
modeling
and
operational
aspect
description,
but
ahead.
I
Unfortunately,
we
could
not
do
that,
so
we
don't
have
an
updated
draft,
but
our
current
focus
is
more
looking
into
the
various
aspects
of
the
food
system,
so
that
includes
Etsy
MEC
architecture
related
stuff,
as
well
as
3gpp
related
interactions.
So,
secondly,
we
focus
more
on
experimental
prototyping
to
learn
where
our
additional
gaps
and
where
we
probably
have
more
things
to
specify
and
also
have
options
for
standard
contributions.
I
So
one
of
the
options
or
one
of
the
tracks
here
is
that
we
contributed
some
of
these
aspects
to
Etsy
Mac
and
in
particular
the
loose
coupling
approach
has
been
discussed
and
and
value
their
us,
but
not
least
for
the
prototyping
and
and
trialing
so
POC
related
activities.
We
currently
do
in
one
of
three
European
Commission
granted
automotive
corridor
projects,
which
is
called
5g
Carmen.
If
you
want
to
have
a
look
to
that
one
or
talk
to
me
later
about
the
project
pieces.
So
very
briefly,
so,
as
you
can
see,.
I
The
red
part
or
the
orange
part,
is
mobile,
pink
all
related
functions,
so
we
have
the
mobile
Co
control,
plane,
the
user
play
function
and
the
application
function,
which,
in
that
case,
we
deploy
on
a
Mac
platform
manager,
which
is
part
of
the
green
part.
So
we
integrate
basically
the
Mac
platform
manager
mainly
to
access
the
traffic
treatment
policies
on
the
Mac
platform,
where
the
applications
are
running
on
top
of
it.
So
this
is,
as
you
can,
identify
the
n6
interface
between
UPF
and
the
actual
service,
and
this
is
what
we
want
to
have
control.
I
So
in
that
case,
it's
a
local
scenario
where,
for
example,
the
relocation
of
an
application
in
the
Mac
platform
is
required
for
load,
balancing
or
failure
reasons.
So
that's
a
problem
statement
that
came
up
in
its
emag,
and
this
is
what
we
can
use
as
a
solution
here
to
really
relocate
the
epic
into
a
different
platform
and
steer
the
traffic
using
the
to
policy
enforcement
points
accordingly,
more
interestingly,
for
the
week
the
scenario
is:
if
we
have
a
moving
car,
you
have
a
question
on.
A
I
Is
done
by
the
policy
controllers
according
to
the
scenarios
that
we
described,
so
this
is
basically
the
case
which
we
show
in
three
slides
that
the
car
moves
from
the
left
and
right
and
at
some
point
in
time
you
may
take
a
decision
to
relocate,
also
the
edge
serving
the
the
vehicle.
So,
as
you
see
here,
that's
a
starting
point:
the
vehicle
has
a
UPF
assigned
from
the
cellular
network,
and
traffic
is
routed
on
the
n6
interface
to
the
v2x
application
running
on
the
platform
here.
I
This
is
basically
the
end
scenario.
We
are
on
the
right
side
and
in
the
end,
we
want
to
continue
serving
the
vehicle
from
a
new
max
site.
The
we
could
got
a
new
UPF
here
and
the
tracks
application
also
has
been
relocated
in
between.
We
try
to
leverage
that
DP
and
overlay
on
+
6
for
this
inter
site
scenario,
to
basically
keep
that
kind
of
procedure
as
much
as
possible.
Independent
of
the
5g
control
plane.
There
are
scenarios
discussed
in
3gpp,
4
I,
think
it's
part
of
ultra
low
latency
kind
of
study
and
description.
I
So
here
they
play
a
lot
with
UPF
serving
as
anchor
nodes
as
uplink
classifiers
and
in
between
they
configure
a
lot.
These
UPF
stood
erect
traffic
to
the
new
site,
so
here
3
TPP,
so
it's
pretty
much
relaxed
by
just
adding
in
another
UPF
which
can
hopefully
serve
the
same.
You
IP
address
and
all
the
relaying
we
do
on
the
VPN
overlay
here.
I
So
that's
something
what
we
investigate,
how
far
this
is
possible
for
intradomain,
but
also
for
inter
domain,
because
the
mentioned
project
like
50
Carmen,
they
take
a
lot
of
interest
in
cross-border
scenarios
which
implies
into
operator
and
over.
So
there
are
a
lot
of
technology
issues
and
challenges
underneath
for
inter
pyramid
over
which
is
not
optimally
deployed
today.
I
One
last
picture
here
so
for
that
kind
of
deployment
in
particular,
if
we
talk
about
cross
domain
handover
and
edge
relocation.
So
in
the
solution
we
are
investigating,
we
involve
an
orchestration
layer
because
here
its
advantages,
if
you
can
anticipate
that
such
a
movement
is
ahead,
so
you
can,
for
example,
derive
this
from
the
movement
pattern
and
the
directions
of
the
vehicle
and
then
anticipate
which
network
is
the
best
to
serve
the
vehicle
in
the
next
one
two
minutes
and
then
prepare
resources
in
advance
through
to
optimize
that
kind
of
process.
I
So
here
we
move
the
orchestration
layer
and
that's
also
taking
relevant
standardization
effort
in
Etsy
and
if
we
into
account
because
they
look
at
the
inter
domain
interfaces
between
distributed
orchestrators,
so
that's
the
full
system.
We
currently
investigate
the
ITF
part
is
more
on,
let's
say
the
control
plane
to
the
data
plane.
I
That
kind
of
interface
where
we
today
discuss
and
refer
to
the
FPC
document
or
PF
CP
could
also
be
used
here
and,
on
the
other
hand,
between
the
AF,
the
application
function
and
the
mobile
core,
which
is
semantic
which
may
be
contributed
into
3gpp
later.
So
that's
something
we
would
like
to
describe
in
the
draft,
and
if
these
scenarios
and
deployments
are
some
of
interest
to
you,
we
would
appreciate
having
more
people
to
discuss
with
jon-jon.
A
Question
for
clarification
you
mentioned
about
optimizing
and
predicting
normals
so
that
the
policy
can
be
noted.
Is
that
I
mean
as
considering
the
day
latency
issue,
because
in
some
sense
the
policy
is
being
handled
by
the
control,
plane
or
management
plane
and
the
traffic
is
a
thousand
times
faster?
Yes,.
F
I
I
So
if
you
can
select
a
target
before
and
at
least
some
of
the
resources,
you
can
some
states,
you
cannot
set
up
in
advance,
for
example
as
soon
well,
if
you
don't
know
which
access
point-
or
you
know
be
the
the
vehicle
will
attach
to
so
that's
one
of
the
last
steps
in
our
scenario
that
we
have
to
consider,
but
we
can
prepare
the
resources
on
my
platform.
We
can
prepare
a
selection
of
UPF
and
DPN
in
advance
and
that's
something
we
currently
look
at
so.
I
For
that
prediction
scheme,
various
things
are
possible
and
that's
to
be
investigated.
Okay,
thanks
all
right
next,
similar
to
to
last
time,
we
still
need
to
revise
the
document
to
focus
a
little
bit
more
on
the
actual
scope
and
converge
and
suitable
notation
for
the
first
operational
aspects,
and
also
for
the
semantics
and
the
information
models
we
want
to
describe
yeah
still,
we
target
a
draft
update
and
if
you
have
comments
to
improve
that
draft,
we
appreciate
and
I
said
we
already
have
associated
contributions
in
ANSI
Mac
here
and
interested
in
adopting
this
work.
I
I
A
Right
yeah,
I
I
tend
to
think
this
is
a
very,
very
good
work.
I
hope
you
know,
let's
see
how
it
goes:
yeah,
Thank,
You
Marco
any
any
final
questions.
Okay,
we
have
two
more
presentations
that
were
both
for
Charlie
Perkins,
but
unfortunately,
last
minute
II
here
to
do
is
cancellation,
so
I
think
we
already
discussed
about
the
yang
okay,
all
right
on
the
FPC,
also
right.
So
there
are
no.