►
From YouTube: IETF 113 Host Speaker Series
Description
The IETF 113 Host Speaker Series talk was held on Tuesday, March 22, at 15:45 UTC in Park Suite 9.This talk discusses:
latest momentum for IPv6 deployment in carriers and enterprises,
challenges that network architects face in deploying and operating IPv6 networks, and
collaborations needed to move IPv6 deployment to a new level. The presenter will be XiPeng Xiao, Standard & Industry Development for Datacom, Huawei Europe.
A
A
Okay,
folks,
sorry
for
the
delay.
A
There
are
a
million
ipv6
talks
right,
so
I
first
want
to
answer
the
question
why
you
should
also
listen
to
this
one,
and
the
reason
is
that
that
I
think
that
most
ipv6
talks
are
given
by
ipv6
supporters
and
usually
the
ipv6
supporters.
Only
tell
you
the
good
thing
about
ipv6
and
I
think
that
what's
unique
in
today's
presentation
is
that
we
are
going
to
talk
about
a
lot
of
the
challenges
facing
ipv6
and
the
reason
why
I
feel
that
this
is
important
and
useful.
A
Is
that
just
like
you
know,
nobody
is
perfect.
No
technology
is
perfect.
You
know
every
technology
have
certain
problems
and
in
order
to
get
all
the
people
on
board
to
deploy
ipv6,
we
really
need
to
acknowledge
the
challenges.
Certain
challenges
facing
ipv6
document,
the
challenges
possibly
provide
a
guideline,
and
I
think
that
eric
linker
kind
of
like
set
a
good
example
for
us
in
rfc
1999,
because
I've
seen
1999,
basically
document
all
the
ipv6
security
excels
and
provide
guidelines.
A
A
So
with
that,
I
really
want
to
say
that
we
spent
a
lot
of
time
presenting
preparing
for
these
presentations
and
a
lot
of
people
providing
a
lot
of
you
know
either
statistics
or
provide
a
local
review,
and
the
list
is
long,
and
here
I
really
again
want
to
thank
eric
binker
because
he
provided
a
lot
of
you
know
very
detailed
feedbacks.
A
So
the
first
part
is
the
momentum.
While
this
is
both
for
the
ipv6
supporters
and
the
ipv6
deniers
and
by
ipv69
I
really
have
no
pixel
respect.
You
know
just
an
easier
name
to
to
refer
to
the
camp.
A
One
will
acknowledge
that
later
here
we
want
to
say
that
you
know
for
many
of
these.
Are
you
aware
of
the
ipv6
progress
in
recent
years,
because
10
years
ago,
if
you
are
consuming
ipv6,
you
are
more
or
less
justified,
because
at
that
time
there's
almost
zero
ipv6
users.
But
now,
according
to
these
google
stats,
we
already
have
36
percent
ipv6
user.
Some
people
argue
that
actually,
due
to
the
way
google
called
the
users,
many
ipv6
users
in
china
are
actually
mixed.
A
So
if
you
count
e
also
the
users
in
china,
it
could
be
as
high
as
45
percent
based
on
etsi
started,
but
while
ipv6
go
very
fast,
the
the
it's
very
uneven-
and
this
will
have
a
major
implication
later-
that
we
will
talk
about
so
what's
driving
these
ipv6
momentums
according
to
jeff
houston.
This
is
you
know
the
gdp
per
capita.
You
know
which
country
usually
are
more
enthusiastic
in
doing
ipv6.
The
user
grows
like
a
reliance
on
the
ipb
for
authorities
and
also
the
competition.
A
A
The
first
one
is
that
we
believe
that
the
loyaltineness
of
ipv6
value
chain
is
something
very
important
because
the
ue's,
the
network
and
the
application
trend
is
finally
ready
because
10
years
ago,
if
you
know
some
of
these,
the
content
are
not
already.
Then,
even
if
operators
deploy
ipv6,
you
are
not
going
to
get
much
benefit,
but
now
it's
very
different,
and
here
we
list
some
of
the
major
milestones,
but
the
slide
is
busy.
A
So
you
don't
need
to
read
that
all
I
want
you
to
know
is
that
it's
summarized
in
these
two
blue
bars
by
2011
ietf.
Lastly,
you
get
the
solution
ready
by
2017,
the
ues,
the
ues
and
the
big
application.
Like
you
know,
google
netflix
are
already
and
this
year
we
believe
that
there
is
another
important
milestone
that
many
people
may
not
realize.
Yet
is
the
public
cloud
support
of
ipv6
like
as
aws?
A
These
are
important.
We
will
talk
about
why
you
know
this
is
this
may
be
quite
significant
and
then
on
the
value
trend,
we
are
seeing
that
the
network
is
still
slightly
behind
the
ue,
the
content,
and
if
you
are
interested,
I
will
not
go
into
detail.
If
you
are
interested
even
later,
we
can
discuss
why
another
network
is
42
percent,
how
we're
computing.
A
That
is
that
another
point
is
that
I
don't
know
whether
the
ipv6,
the
denier
realized
that
the
price
of
ipv4
is
going
up
so
quickly
and
specifically
the
price
doubled
last
year,
and
if
you
call
it
long
2014
it
already
increased
seven
times.
A
So,
if
you
the
world
on
ipv4,
you
may
have
to
spend
a
lot
more
money.
If
you
have
to
buy
ip
default
address
or
for
some
operator
they
are
even
considering
you
know,
going
to
ipv6
quickly
and
while
ipv4
you
know,
price
is
very
high,
they
can
sell
their
ipv4
address.
This
is
important
one
and
then
the
public
ipv6
wait.
This
is
the
kind
of
the
other
momentum
that
we
saw.
A
One
is
this:
just
now
we
mentioned
that
aws
ages,
support
of
ipv6.
Why?
We
believe
that
this
is
maybe
very
important.
We
when
we
look
at
the
operators,
you
know
the
big
operating
six
terms,
the
small
and
medium
enterprise.
A
A
If
we
look
at
the
sme,
you
know
there's
a
huge
number
of
sme
they're
simply
in
ipv4,
but
we
believe
that
when
these
you
know
aws
as
when
they
support
ipv6
and
as
the
you
know,
sme
go
to
the
cloud
you
know,
aws
may
enable
them
to
move
to
ipv6
very
quickly
in
the
past.
They
may
have
to
do
a
lot
of
work
themselves,
but
now
they
simply
select
ipv4
or
ipv6
and
if
they
select
ipv6,
they
go
ipv6.
A
So
we
may
know
today
in
the
v6
working
group
that
discussion
will
you
know
how
to
move
the
sme,
and
we
believe
that
this
may
be
a
very
big
momentum
and
then
on
the
iot.
You
probably
see
the
number
already
there's
a
lot
of
iot
device,
then
5g
bring
new
beo
a
new
opportunity
and
then
obviously
there's
these
innovations
on
top
of
ip
v6,
srb6
and
other
things,
provide
better
traffic
engineering,
network
programming,
etc.
A
So
for
the
people
who
don't
believe
in
ipv6,
I
think
we
really
want
to.
These
are
some
of
the
recent
progress
not
to
mention
the,
for
example.
The
big
government
proves,
even
though
u.s
government
is
pushing
towards
ipv6,
chinese
government
is
fixing,
but
I
guess
maybe
ietf
people
don't
want
government
regulation,
so
I
avoid
them
here
and
the
second
part,
I
think
the
first
part
is
more
to
the
ipv69
and
the
second
part
is
more
to
the
ipv6
supporters.
A
And
here
what
we
want
to
say
is
that
for
the
ipv6
supporters
there
there
really
are
more
challenges
than
you
think,
because
you
know
we
as
the
the
supporters.
Sometimes
we
feel
that,
oh
you
know
our
work.
Ietf
work
is
done.
You
know
these
people
are
not
doing
ipv6
just
because
of
the
lack
of
the
motivation
or
because
they
are,
they
are
saying
all
kinds
of
skills.
You
know.
A
A
You
can
say
that
oh
2017,
maybe
too
old,
but
actually
you
know
based
on
what
I
hear
from
other
industry
forum.
If
I
compare
it
to
some
of
the
ipv6
challenges
in
in
december
last
year,
it
didn't
change
that
much
and
really
here.
I
think
it
may
be
difficult
for
you
to
to
see
other
things,
but
I
just
want
to
kind
of
like
quickly
classify
them
into
four
categories.
A
One
is
motivation.
I
think
that
for
ipv6,
this
motivation
is
always
one
of
the
biggest
xu
okay.
You
know
my
end
user,
don't
care
about
ipv6,
it's
a
lot
of
work,
so
you
know
why
supply
so
things
like
this
and
then
knowledge
is
also
a
big
part.
You
know,
okay,
you
know,
I
don't
understand,
ipv6,
there's
so
many
new
protocols,
it's
difficult.
So
I
think
the
big
car
ecosystem
is
the
value
train
that
we
talked
about.
A
So,
regarding
this
motivation,
knowledge,
I
think
that
you
know
these
slides
look
a
little
deeper.
I
think
that
maybe
we
focus
on
the
the
more
technical
part
is
the
technology.
Is
the
technology
so
that
the
legacy
device
is
the
ipv6
the
network
management
system?
I
think,
generally,
this
is
a
field
that
you
know
ipv6
network
management
compared
to
ipv4.
A
I
think
that
you
know
there's
still
some
work
to
do
and
the
oam
also
may
be
not
as
mature.
These
are
some
of
the
you
know,
common
complaints
from
the
field,
but
we
want
to
say
that,
yes,
these
are
issues,
but
you
know
the
industry
largely
you
know,
taking
action
to
to
deal
with
them.
So,
regarding
the
motivation
and
knowledge
you
know,
leading
operators
are
selling
that
experience
and
then
you
know
there's
many
ipv6
forums
councils.
A
A
They
really
did
a
lot
of
work
at
all.
They
really
did
a
lot
of
work
to
move
the
industry
towards
ipv6
and
for
the
last
part,
for
the
last
part,
we
are
going
to
talk
about
the
technical
challenge.
I
think
that
it's
the
focus
of
our
presentation
today
so
before
we
talk
about
the
bad
things.
First,
let's
let's
say
that
you
know
ietf
has
done
a
lot,
so
you
know
this
is
not
to
to
complain.
A
This
is
not
to
to
complain,
but
even
though
that
we
have
done
a
lot,
I
think
that
there
are
remaining
challenges
remaining
challenges,
so
the
first
one.
I
think
that
you
know
to
compare
the
ipv6
performance
versus
ipv4
may
be
very
relevant
and
more
telling,
because
between
ipv6
and
ipv4
there
are
so
many
things
that
you
can
compare.
There's
so
many
things
that
you
can
compare.
For
example,
neighbor
discovery
is
very
different
from
app,
but
we
believe
that
you
know
performance
is
really
the
very
important
or
maybe
the
most
important.
A
You
know
kpi.
So
the
first
one
is
regarding
the
raw
chip
time,
the
latency.
So,
as
you
can
see,
okay,
you
can
see
here
over
the
years
the
the
ipv6
road
trip
time
is,
you
know,
constantly
improving.
It's
constantly
improving,
but
today
it's
the
old.
You
know
2.5
millisecond
higher
than
ipv4
than
ipv4,
but
this
is
what
average
this
is
whatever
right.
You
know
later,
I
will
provide
you
know
kind
of
like
the
the
the
grain
of
salt
on
my
own.
A
You
know
this
data
and
again
this
is
based
on
jeff
houston's
status
in
ap
nics.
There
are
you
know
in
the
industry.
There
are
some
other
statistics
like
linker
that
all
have
some
statics
right,
ncc
have
statics,
etc,
but
we-
or
at
least
I
personally
believe
that
maybe
ap
nix
data
is
more
systematic.
So
we
are
so.
I
think
that
the
first
is
that
you
know
ipv6
and
ipv4
is
very,
very
close,
but
it's
slightly
higher
on
a
worldwide
average
scale.
A
So
I
think
a
very
good
question
to
us
here
is
that
what
are
the
top
three
factors
leading
to
this
improvement?
Why
you
know
over
the
years
why
this
is
improving?
And
if
we
understand
what
are
contributing
to
this
improvement,
then
I
think
that
you
know
we
can
provide
a
guideline
for
other
operators.
That,
oh
you
know
do
this
from
day
one
then
you
know
you
can
you
can?
You
know
avoid
certain
things.
So
this
is
the
logic
time.
A
The
second
one
is
the
packet
loss
rate
and
we
find
that
the
data
here
is
really
inconclusive.
We
provide
you,
know
two
data
points
here.
The
first
one
is
from
this
hindavi.
It
reported
ipv6
plr
actually
lower
than
ipv4,
and
this
is
they.
They
did
the
management
based
on
one
way
of
management
from
from
china,
and
it
says
that
you
know
for
them.
They
find
out
that
ipv6
packet
loss
rate
is
0.25
percent
for
ipv4
is
0.33,
so
ipv6
performs
slightly
better.
They
also
have
some.
You
know.
A
A
So
here
we
are
citing.
You
know,
rfc
7872
by
I
think,
by
fernando
by
zhang
lincoln
here
and
now
I
think
recently.
Eric
winker
also
have
the
so-called
jerk
james
talking
about
some
new
management
on
on
these,
like
extension
headers,
but
the
the
long
story
short.
The
result
is
that
you
know
in
2017
the
packing
loss
rate
for
for
the
plastic,
waste
fragmentation,
extremely
high
21
percent,
and
even
last
year
it's
still
eight
percent.
A
It
improves
a
lot,
but
it's
still
very
high
because
we
heard
from
many
operators
we
heard
from
many
operators
like
their
ipv
for
packet
loss
rate
or
their
packet
loss
rate
is
about
0.1
percent
and
we
take
it
as
ipv4,
because
mainly
most
traffic
is
still
ipv4,
but
you
know
there's
really.
I
think
that
we
find
in
the
industry,
but
there
is
no
data
on
ipv6
packet
glossary
or
at
least
ap
nic
didn't
provide.
A
A
We
we,
we
don't
know
how
much
of
this
overall
packet
loss
rate
is
coming
from
those
fragmentation,
extinction,
headers,
etc.
We
believe
that
this
would
be.
You
know
something
good
to
to
find
out.
We
we.
We
really
feel
that
you
know
if
we
can
dig
into
this
and
find
out
why
you
know
ipv6
is
losing
package
etc.
Again,
this
would
help
will
help
to
to
improve
the
performance
and
make
the
the
life
of
those
deployer
a
little
easier
and
the
third
one
is
the
tcp
failure
rig
again.
A
We
can
see
that
the
the
blue
one,
the
green
one,
is
ipv6
and
then
the
other
is
v4.
We
can
see
that
over
time
the
ipv6
tcp
failure
rate
is
like
again
constantly
improving,
but
I
think
that,
based
on
the
data
today,
ap
nick-
it's
still-
you
know
0.96
percent
and
ipv4.
You
know,
depending
on
how
you
look
like
it
may
be,
like
0.25
percent
or
some
people
jeff
houston
said:
oh,
it's
almost
like
zero
percent,
so
you
know
ipv4
is
almost
successful,
but
anyhow,
I
think
that
the
conclusion
is
the
same.
A
Tcp
failure
ipv6
is
higher,
so
the
question
is
that
is
it
because
of
the
higher
packet
loss
really
of
ipv6?
I
think
that
this
certainly
you
know
if
you
have
a
higher
ipv6
plr,
it's
quite
likely
that
the
tcp
failure
rate
will
also
be
higher
and
then
because
quick
is
using
more
and
more
in
addition
to
bill
to
measuring
tcp
failure,
we
also
be
measuring
quick
as
well.
So
I
think
that
these
are
some
of
the
good
questions.
A
Okay,
you
know
me
as
a
supporter,
ipv6
supporter,
talking
about
ipv6
challenges
here,
and
some
people
say
that,
oh
they
they
may
be
concerned
that
okay,
you
know
I
come
to
the
conclusion
that
we
shouldn't
do
ipv6,
and
I
want
to
tell
you
that
this
is
not
the
case
because,
again
that,
just
now
what
we're
presenting
is
the
worldwide
average.
A
Whatever
it
can.
You
know
the
average
can
hide.
A
lot
of
things
can
hide
a
lot
of
things
and
if
we
look
at
the
data
you
know
more
carefully,
we
will
see
that
you
know
the
world
average
can
be
jerked
out
by
a
region
badly
and
again.
This
is
the
same
steps
from
ap.
Nick
you'll
see
that
you
know
the
whole
world.
A
The
rtt
of
ipv6
is
slightly
higher,
but
it's
largely
you
know
because
of
you
know
swing
me
and
if
you
look
at
the
world
you
know
american
and
europe.
Actually,
you
know
ipv6
perform
better
than
ipv4,
so
I
I
really
feel
that
these
data
are
useful
for
for
communicating
with
the
ipv6
deniers.
You
know
when
they
say
that.
A
Oh
it's
bad,
you
can
you
you,
you
have
the
stats
to
to
tell
them
what,
if
you
look
deeper,
what's
the
case
and
also
there
are
reports
that
ipv6
can
perform
better
in
enterprises,
for
example,
we
provide
two
references
here.
You
know
apple
at
one
place
is
claiming
that
you
know
ipv6
can
be.
You
know,
40
percent,
better
in
the
in
the
data
center.
Again,
you
know
we
provided
the
reference
here.
You
can
check
out
and
facebook
is
also
very
enthusiastic
in
in
promoting
ipv6.
A
If,
if
you,
you
have
a
lot
of
talent,
if
you
can
do
it,
you
know
correctly,
but
it's
not
always
easy.
This
is.
This
is
a
complicated
machine
and
if
you
install
it,
you
know
correctly,
you
operate
it
correctly.
Then
you
perform
better
than
ipv4,
but
because
it's
due
to
its
complexity,
you
may
make
mistakes,
and
if
you
make
mistakes
then
it
don't
work
take
work.
So
I
think
that
this
is
the
concern
that
we
want
to
want
to
provide.
A
As
a
consequence,
I
think
that
this
is
also
another
thing
that
I
find
as
a
big
surprise
to
me
last
year,
and
it
may
actually
be
a
big
surprise
to
many
ipv6
bathrooms
is
that
many
operators
ipv6
deployment,
are
actually
in
overlay
and
in
the
underlay
in
the
router
et
cetera,
they're,
actually
tunneling
ipv6
in
nprs
etc.
A
So
you
know,
when
the
operator
tells
you
that
oh
I
do
deal
stack,
you
think
that
you
know
their
routers
are
transporting
ipv6
traffic
natively.
You
think
that
their
underlay
is
also
dual
stack,
but
that
may
not
be
the
case.
That
may
not
be
the
case.
This
is
a
big
surprise
for
me
last
year.
So,
first
of
all,
let's
let
me
clarify
what's
overlay
and
what
underlay
I
think
this
is.
A
This
is
a
notion,
it's
applicable
to
operators
and,
to
some
extent
also
applicable
to
data
center,
but
it's
not
applicable
to
for
enterprise
if
you
are
not
providing
a
cervix
to
other
people,
you
really
don't
have
this,
like
distinction
between
overlay
and
underlay,
but
any
household
operator
overlay
is
the
service
layer
and
it
involved
the
the
user
device
and
also
the
mobile
gateway
or
bng
or
etc.
This
is
your
your
kind
of
like
your
service
layer
and
then
underlays,
your
routers
etc.
A
Is
the
the
the
to
send
the
send
the
package
because
ipv6
is,
I
think,
the
the
number
one
reason
is
to
address
this
address
sorted.
You
know
to
provide
enough
address
and
if
to
do
that,
then
ipv6
the
overlay
is
sufficient,
it's
sufficient
and
why
operators
don't
want
to
you
know
many
operators
don't
do
native
ipv
b6
packaging
talk
is
because
if
your
router
is
still
stacked
really
deals
that,
then
you
need
a
a
control
plane
management
claim
data
playing
for
ipv4.
A
You
have
need
a
separate
play
for
ipv6.
However,
if
you
just
tunnel
ipv6
packet
inside
ipv4,
then
you
only
need
a
single.
You
know
data
and
controller
management
click,
so
many
operators
feel
that
okay,
you
know.
Actually
it
will
be
easier
for
me
to
maintain
just
one
per
plane,
then
two
separate
play
and
there
are
many
of
them-
are
using
6pe
set.
So
some
of
the
ipv6
veterans
thought
that
operators
ipv6
is
largely
down,
but
this
is
really
not
the
case.
A
First,
there
are
many
big
operators
still
doing
ipv4
and
even
for
those
already
deployed
ipv6
that
ipv6
may
still
just
in
the
overlay,
not
much
in
the
underlay.
So
this
is
again.
You
know
back
to
the
question
that
we
need
to
make
ipv6
perform
better.
We
need
to
identify
the
challenges
and
make
ipv6
perform
better
than
ipv4
otherwise.
Otherwise
you
know
there
are
still
a
big
part
of
operator
network
playing
remaining.
A
You
know
there
are.
There
are
many
consequences
so
here
how
to
make
ipv6
perform
better
than
ipv4.
There
are
these
answers.
The
first
two
came
from
jeff
houston,
the
first
two
point
that
came
from
jeff
houston.
He
actually
have
three
points,
but
I
had
to
delete
one
point,
because
you
know
that
could
receive
a
lot
of
protest.
A
The
deleted
point
is
that
he
said
that
you
should
avoid
stateful
ipv4
to
ipv6
transition
and
then
kind
of,
like
you
know,
is
clued.
For
example,
four,
four,
six
four
x,
lab
etcetera
I'll
receive
a
lot
of
you
know
protests
from
that
point,
so
you
know
in
the
end,
I
delete
it.
There
are
other
points.
The
the
next
five
points
you
know
below
the
below
the
the
white
space.
The
the
other
six
points
are.
The
five
points
are
you
know,
based
on
our
analysis.
A
First
is
move
content
closer
to
end
user.
This
will
improve
both
ipv4
performance
and
ipv6
performance,
but
there
are
data
showing
that
you
know
the
improvement
for
ipv6
is
actually
bigger
than
for
ipv4.
So
if
it
close
the
gap
or
even
you
know,
change
the
change,
the
wind,
there
are
some
explanations
and
but
we
don't
have
the
time
to
go
into
detail
why
this
is
the
case:
increase
link
mpu.
If
you
can
and
change
the
choose,
the
tcp
mss.
A
A
Like
announcing
ipv6
routes
and
a
lot
filtering
and
jeff
houston,
provided
some
example
that
you
know,
if
you
don't
do
it,
you
basically
provide,
for
example,
that
between
australia
and
singapore,
ipv4
is
australia
to
singapore
directly
back
to
to
australia,
but
for
ipv6,
pakistan,
australia,
to
singapore,
to
germany
to
us
and
then
back
to
australia,
then
obviously
you're
going
to
have
a
lot
higher
roadship
delay
and
possibly
a
packet
loss
and
then
there's
this
package
filtering.
A
A
So
this
is
the
second
part
and
the
third
part
is
looking
forward.
As
we
said,
if
you
look
at
the
world
average,
it
seems
like
ipv6
seems
to
be
a
little
worse.
But
if
you
look
at,
you
know
deeper.
Actually,
as
we
say,
if
you
do
ipv6
correctly,
ipv6
can
perform
better
than
ipv4
just
like
in
america
and
in
europe.
A
So
if
we
look
at
the
the
world
today,
I
know
that
this
is
a
busy
slide
and
you
will
not
be
able
to
tell
the
company
or
read
the
small
numbers.
I
think
that
there
are
only
two
impressions
that
I
want
to
get
want
you
to
get
from
this
slide.
First,
is
that
there
are
many
companies
with
very
high
ipv6
user
percentage
already
starting
from
the
you
know,
the
top
one
is
100
ipv6
and
even
the
bottom
one
is
still
like
57,
it's
still
57.
A
They
can
already
move
to
ipv6
only
they
can
already
to
move
to
ipv6.
Only
and
brian
carpenter
told
me
that
you
know
every
time
you
talk
about
ipv6
only.
You
should
immediately
say
that
ipv4
as
a
service
over
ipv6,
because
otherwise
people
will
say
that
oh,
this
guy
is
a
lyrical.
How
how?
How
can
you
be
you
know?
Ipv4
will
stay
for
many
many
years.
They
thought
that
ipv6
only
means
that
no
ipv4,
but
it's
not
the
case.
A
It
just
means
that
you
know
ipv4
will
be
long
as
a
cervix
over
ipv6,
but
it
will
disappear
from
the
infrastructure
layer.
So
it's
the
the
time
to
to
do
so,
and
here
you
know,
will
be
a
little
bit
more
specific.
So
at
least
from
operator
perspective
you
know
facebook
already
done
it.
I
think
that
the
hyper
scale,
the
hyper
skill,
have
a
lot
of
talent.
They
probably
have
you
know
the
best
talent
in
the
industry,
and
they
have
no
problem.
A
You
know
they
can
they
they
they
they
can
do
whatever,
but
for
some
operators
it
may
not
be
so
clear,
so
we're
kind
of,
like
you
know,
draw
some
picture.
A
A
Is
I
think,
ipv6
only
at
the
moment
in
the
industry,
for
some
people
it
doesn't
mean
the
ipv4
as
a
cervix.
Again,
you
know
it's
just
in
the
overlay,
and
here
we
want
to
make
sure
that
you
know
by
ipv6.
Only
we
not
only
mean
overlay,
we
really
mean
you
know
underlay
and
the
ipv4
the
packet
will
be
tunneling
inside
ipv6
will
be
finally
inside
ipv6.
We
believe
that
this
will
further
improve
the
performance
of
ipv6
and
make
it
better
than
ipv4.
A
A
A
Really,
we
need
the
help
of
ietf
and
the
leading
operators.
We
really
need.
First,
we
need
to
document
all
the
challenges
and
provide
a
guideline.
So
this
brings
me
to
the
summary
again:
we
want
to
emphasize
with
our
feelings:
it's
like
a
sophisticated
machine.
A
Only
if
you
can
install
and
operate
it
well,
it
will
be
better
than
ipv4.
Otherwise
it
can
be
worse.
Then
it
can
be
worse
than
ipv4,
and
in
this
aspect
we
really
need.
You
know
ietf
to
kind
of
like
provide
kind
of
like
better
user
guideline,
to
make
sure
that
you
know
people
will
have
a
higher
chance
of
you
know,
install
and
operate
it
correctly
these.
If
we
do
this,
then
we
can,
we
can,
you
know,
accelerate
move
to
ipv6
and
ipv6.
A
B
You
talked
about
there
having
been
some
work
done
to
measure
the
performance
improvements
and
that
performance
improvement
for
ipv6
being
larger
than
the
performance
improvement
for
ipv4.
When
you
talk
about
dealing
with
keep
edge
or
deep
in
the
network
deployed
content
key.
So
I
was
just
wondering
if
you
have
any
study
or
reference
material
that
I
can
look
at.
B
You
know,
because
that
is
an
area
of
great
interest
for
me
and
the
organization
that
I
work
for
and
we'd
love
to
we're
already
kind
of
semi-dual
stack,
but
we
deploy
in
a
lot
of
networks
which
are
till
to
this
day
large.
You
know
consumer
facing
isps
that
are
ipv4
only
so
we'd
love
to
be
able
to
show
some
statistics,
and
you
know,
try
to
help
move
the
needle
with
these.
You
know
medium
mid-sized
isps,
who
you
know.
I
have
just
not
got
onto
the
ipv6
bandwagon
yet.
A
Okay,
I
I
think
the
way
should
probably
you
know,
take
this
discussion
offline
because
it
will
it
one
is,
it
will
take
a
lot
of
time
and
the
second
you
know
I
cannot
remember
you
know
all
the
other.
You
know
the
the
the
reference,
so
I
can
send
you
the
reference
afterwards.
B
B
You
know
ammunition
so
that
when
we
go
to
these
providers,
who
are
reluctant
to
you
know
embrace
ipv6,
we
have
something
that
we
can
show
them
and
say
it's
in
your
interest
to
do
this,
even
though
you
know
you
don't
mean
you
may
not
see
a
pressing
commercial
need
to
do
it
right
now,
but
hey
here's
a
case
for
a
performance
improvement
for
your
customers
that
should
you
know,
help
you
make
the
decision
to
go
ahead
and
do
this.
C
Journaling
over
google,
as
an
enterprise
operator,
who
spent
some
time
operating
dual
stack
and
v6
only
network.
I
have
an
impression
that,
in
my
experience,
every
single
case,
when
you
see
v6
performance
worse
than
before,
is
because
of
happy
eyeballs,
because
v6
network
is
broken.
Nobody
monitoring
kid
because
of
happy.
Eyeballs
users
are
not
complaining
and
it
stays
like
that
until
you
turn
off
ipv4.
C
A
D
D
Next,
one
that
one
do
you
have
any
sense
of
why
oceania
is
so
bad
compared
with
the
other
regions.
A
D
E
Hi
mike
ackerman,
first
off,
I
want
to
say
that
you
covered
a
lot
of
information.
This
is
really
good
stuff.
I
hope
you'll
come
back
and
do
it
again,
it's
almost
inspirational.
Thank
you.
The
first
question
I
want
to
ask:
maybe
somebody
already
did,
can
we
get
copies
of
your
slides?
Yes,
fantastic
and
then
I
just
wanted
to
ask
one
specific
question.
I
got
your
email.
I
got
a
whole
bunch
more.
So
beware,
you
had
a
slide
that
said,
and
it's
right
up
there.
It's
that
ipv6
performs
better
in
enterprises.
A
We
see
it
in
this,
for
example.
Let
me
let
me
take
facebook
as
an
example.
You
know
facebook
we
provide
here.
You
know
this
link
to
a
youtube
video
in
this
youtube
video.
I
believe
that
the
person's
name
is
taught
well,
it's
a
facebook
person
talking
about
ipv6.
His
ipv6
performance
is
better
than
it
before
I
I
remember
using
a
nano
meeting
or
in
some
kind
of
meeting,
and
then
there
are
people
from
the
audience.
E
A
Want
to
know
that
those
people,
I
think
that
you
know
they
provide
some
kind
of
I.
I
will
not
say
that
it's
it's
a
very
direct
answer.
They
say
well,
for
example,
if
you
do
ipv6,
then
you
have
no
net,
but
really
nat
will
introduce
only
a
very
small
difference,
so
ipv6
is
better.
Is
it
really
because
of
net
or
because
of
some
other
reasons
we
didn't
get
a
a
very
clear
answer
from
that
youtube
video.
I
think
we
have
the
reference.
E
A
E
I
definitely
wrote
down
what
she
just
said
that
when
you
get
to
ipv6
only-
and
that
makes
a
lot
of
sense,
you
can
see
where
the
economies
of
scale
will
start
to
add
up
and
give
you
better
performance,
maybe
even
holistically.
So
I'm
just
looking
for
any
answers
anywhere.
I
can
get
it,
and
this
was
great.
Thank
you
again.
B
Yeah,
so
this
is
again
relating
to
the
difference,
so,
while
in
this
table,
you're
kind
of
showing
the
average
difference
in
the
rtts
and
and
yes
based
on
these
numbers-
looks
like
it's
a
massive
outlier
with
a
much
larger
average.
B
But
I
probably
am
not
as
surprised
by
that,
because
it's
quite
likely
that
the
overall
total
rtt
average
for
connections
in
oceania
are
probably
going
to
be
a
whole
lot
higher
because
of
the
fact
that
it
has
very
limited
capabilities
of
serving
traffic
locally
and
most
traffic
originating
from
there
is
reaching
out
to
some
other
region
of
the
world
in
order
to
get
served,
which
is
not
the
case
with
maybe
america's
or
europe
or
for
that
matter,
even
asia,
because
a
lot
of
that
traffic
will
be
within
the
region.
B
So
when
the
total
connection
rtts
are
in
the
range
of
40
or
50
milliseconds,
their
10
percent
difference
turns
out
to
be
4
milliseconds
perfect
with
oceania,
if
they're,
if
the
average
of
their
connections
are
mostly
going
to
be
in
the
300
and
400
millisecond
rtt
rate,
because
they're
all
going
outside
of
the
oceania
region
to
get
served,
a
10
differential
in
performance
would
still
look
like
a
very
big
difference
over
here.
So
I'm
my
question
is:
have
you
looked
at
that
as
a
data
point?
A
The
the
answer
is
no,
as
I
said
you
know
this
slide,
I
added
today,
you
know
upon
the
request
of
a
colleague,
but
I
I
can
offer
observation
is
that
you
know
if
you
know
jeff
houston,
how
he
how
he
collects.
His
statistics,
he's
basically
partnering
with
google,
and
you
know
putting
a
kind
of
like
an
ad
on
the
you
know,
chrome
browser,
and
then
this
chrome
browser
will
contact
his.
You
know
he's
he's
a
prober,
let's
say,
and
then
because
this
is
sinkholding,
this
is
stinkling.
A
So
I
do
the
the
observation
that
you
know
at
different
time.
These
can
you
know,
for
example,
tomorrow,
if
you
go
to
you,
know
jeff
houston's
side
and
then
take
another
snapshot.
The
number
can
be
quite
different,
so
I
think
that
this
is.
This
is
point
number
one
this
you
cannot
definitely
not.
You
know,
draw
a
conclusion
that
you
know
oceania
is
always
is
always
bad.
So
again
you
know
all
of
these.
You
know
every
statistics.
A
You
must
also
look
at.
You
know
kind
of
like
take
it
with
a
grain
of
salt
with
a
grain
of
salt.
You
know
really
tomorrow
we
when
we
look
at
the
the
one
and
just
now,
when
you
talk
about
the
the
maybe
you
know,
osuni
is
far
away
from
content,
but
again
we
are
here.
We
are
talking
about
the
difference
between
ipv4
and
ipv6,
and
the
long
distance
you
know
is
a
burden
for
both
ipv4
and
ipv6.
A
So,
arguably
you
know
it
shouldn't
create
a
lot
of
difference,
but
other
than
that.
Really
I'm
sorry,
you
know
I
I.
I
really
feel
that
that
for
ipv6
performance
there
are
a
lot
of
analysis
is
worth
doing,
because
we
really
you
know
know.
I
would
say
that
we
know
very
little.
If
we
we
just
support
ipv6,
and
we
don't
face
these
challenges.