►
From YouTube: DASH Behavioral Model WG 20220616 (June 16, 2022)
Description
June 16, 2022
Upcoming Keysight CI/CD pipeline automation tests
Keysight/NVidia updates next week
B
Okay,
I'd
like
to
answer
that
and
and
I'll
to
share
our
progress
which
I'm
on
the
verge,
probably
well.
The
purple
request
forthcoming
pretty
soon.
Actually,
it's
coming
in
two
phases:
the
the
first
one
is:
we've
already
done
some
CI
python
automation
to
build
all
the
artifacts
and
actually
run
traffic
tests,
a
basic
UDP.
You
know
loopback
on
the
port
test,
as
without
any
configuration
right
now.
B
B
There
so
I
hope
to
be
able
to
present
that
next
week,
if
all
goes
well,
and
that
has
a
traffic
generator
component
in
it
and
the
one
I've
been
talking
about
since
last
summer
is
using
not
PTF,
which
is
very
popular,
but
it's
actually
using
an
open
source
framework
called
Snappy
that
can
scale
from
software
to
line
rate
and,
in
this
case,
we're
using
software-based
traffic
generator,
which
is
available
from
Docker
hub.
B
B
You
know
I'm
very
familiar
with
PTF
as
well,
but
if
you
write
a
lot
of
tests
in
PTF,
there's
no
path
to
line
rate,
so
we
put
a
lot
of
effort
into
making
this
a
scalable
solution,
and
so
that's
project,
one
that
we're
working
on
Project.
Two
and
I'll
I'll
need
your
help,
probably
Marion,
and
maybe
help
from
others-
is
to
put
the
scytherip
server
and
integrate
it
into
your
lib
site
and
I've
made
some
progress.
A
B
What
you
need
to
do
is
pull
python
client
library
from
pip
or
Pi
Pi.
It's
available
and
pull
a
couple
of
Dockers.
One
is
the
traffic
engine
which
you
can
spin
up,
multiple
instances,
one
per
port,
and
you
need
to
pull
a
Docker
based
controller
for
that.
And
then
you
can
write
python
script
to
spin
this
up
and
we've
got
that
entire
workflow
working
both
from
the
command
line
or
in
the
cloud
under
CI
CD
control.
B
So
this
this
pull
request
will
be
a
demonstration
of
this
technique
and
and
once
you
look
at
it,
you
can
see
how
it
goes
and
you
know
decide
if
you
like
that
approach
or
not.
But
it's
the
one
we've
been
proposing
for
a
long
time,
because
you
can
actually
use
the
same
client
library
to
control
line
rate,
Traffic
generators
using
the
same
Scripts
and.
A
Yeah
and
how
it
fits
in
with
the
simulator.
So
basically
we
have
one
Docker
running
the
simulator.
We
have
another
Docker
running
the
traffic
generator
and
then
you,
the
point
of
connection,
would
be
the
interface
and
you
can
swap
Simulator
for
the
for
the
physical
Hardware.
And
yes,
that's
the
idea.
B
Yes,
yes,
okay,
yeah,
yeah,
I
wish
I
was
a
little
more
ready.
I
was
up
till
practically
midnight
trying
to
debug
a
couple
things,
but
it's
it's
coming
along
pretty
good
and
and
next
week
hopefully
I'll
be
able
to
do
a
presentation
on
it.
But
the
idea
is
once
we
get
this
working,
you
can
do
a
check-in
of
you
know
anything
into
the
P4
is
under
serious,
Pipeline
and
other
than
doc
files
and
things
any
changes
you
make.
It
will
kick
off
a
build
and
in
the
cloud
and
do
a
full.
B
Another
thing
I've
been
working
on
is
improving
the
turnaround
time
of
the
builds
and
I've
actually
created
a
Docker
and
put
it
on
my
personal
Docker
hub
for
now,
but
eventually
we'll
have
it.
Hopefully,
in
the
Azure
container
registry
of
this
of
this
Docker
tool,
image
that
you
created
Marianne.
B
D
A
A
B
And
once
once
we
present
our
you
know
prototype,
then
then
the
group
can
weigh
in
and
we
can
decide
where
we
want
to
take
things.
You
know
what
and
people
always
have
new
ideas
and
new
requirements
that
we've
been
we've
been
using.
This
HCC
traffic
generator
quite
a
while
and
it's
being
used
by
some
Cloud
scale,
people
at
volume.
So
it's
pre-proven
and,
like
I,
said
the
same.
Client
interface
can
run
on
line
rate
Hardware
as
well.
B
D
Because
we
like
to,
as
you
know,
discussions
today,
we
have
no
this
whole
audience
knows
that,
but
we
have
several
mbas
that
are
up
in
our
cloud
and
then
we
generate
traffic.
D
You
know
in
our
Canary
and
other
places
in
the
cloud
so
that
we
can.
We
can
enhance
the
testing,
because
today
we
do
testing,
but
we
don't
have
all
the
flexibility
to
change.
You
know
it's
the
best
for
fuzzing
and
all
kinds
of
other
stuff
that
we
want
to
do,
and
the
access
should
be
able
to
provide
us
some
more
flexibility
to
do.
D
B
All
the
work
that
they're
doing
on
that
yeah
I'll
just
show
it.
Thank
you.
I'll
just
show
a
quick
down
the
page,
I
wanted
and
I'll
just
share
this
pretty.
Please.
B
You
can
see
that
that's
the
configuration
of
this
particular
test,
so
these
are.
These
are
Dockers
and
here's
the
behavior
model,
and
this
is
how
it's
bound,
but
this
could
be
a
physical
device
and
these
were
real
ethernets.
B
This
could
be
software
traffic
generator
running
about
and
you
could
be
binding
it
to
physical,
mid
ports,
so
you
could
actually
have
a
software
based
traffic
generator
going
out,
Nick
ports,
so
that's
one
mode
and
then
this
could
actually
be
replaced
by
you
know
a
traffic
generator
Appliance
running
it.
Whatever
speed,
you
want,
take
a
speed,
100,
Gig,
200,
800
and
and
the
same
client
in
API
run.
So
what
we
wanted
to
do
is
avoid
rewriting
tests,
so
this
approach
allows
you
other
than
the
scaling
factor
to
use
the
same
basic
approach.
B
So
when
we,
when
we
do
this
pull
request,
it'll
be
a
nice
kind
of
first
first
example
for
some
people
to
to
learn
about
how
to
do
this.
D
The
other
cool
thing
that
we're
doing
to
that
we'll
be
able
to
use
this
with
this
is
we've
had
traffic
sources
and
syncs
for
a
long
time
like
net
perp
Etc,
but
they
are
limited
by
the
connections
per
second
of
the
system
that
we
have
in
place
to
the
cloud
today
and
now.
What
we're
doing
is
we're
we're
actually
able
to
generate
a
lot
more
data,
a
lot
more
traffic
because
we're
actually
using
sirius
on
on
not
just
the
mbas
but
actually
the
source
and
sync
VMS.
D
So
we
can
use
a
lot
less
VMS
to
do
traffic
generation
now
than
we
used
to
in
the
past.
We
used
to
have
to
use
around.
You
know,
spin
up
around
20
and
now
we'll
spin
up
to
you
know
and
we'll
be
able
to
get
pretty
good
rates
in
the
cloud
to
point
at
the
mbas.
B
D
B
I
do
want,
to
you,
know,
disclaimers
this
particular
engine
and
this
client
right
here
these
are.
This
is
primarily
at
this
point
for
Layer
Two,
Three
streams,
and
you
know
we're
working
on
adding
layer,
Port
functionality,
but
it's
not
available
in
this
particular
set
of
packages
at
this
time,
although
you
can,
you
know
you
could
always
create
scenarios
where
it
kind
of
looks
like
connection
oriented,
but
that's
that's
work
in
progress.
B
As
far
as
the
true
layer,
4
connection
oriented
testers,
the
VM
versions
are
available,
but
those
are
commercial
Solutions,
whereas
this
right
here
there's
a
free
version
and
a
commercial
version.
So
you
know
those
are
details
we
can
get
into
later
yep
anyway,
things
are
things
are
rolling
along.
B
F
B
C
Okay,
so
Chris
had
asked
me
and
I've
done
some
poking
around
inside
internally
at
Microsoft
and
at
Prince.
This
has
to
do
with
you
and
I
about
the
the
need
for
a
Docker
registry
for
for
Dash
and
a
resource
for
GitHub,
Runners
and
I
was
in
the
building
yesterday
talking
to
Gohan,
and
he
said
that
he
felt
we
would.
We
shouldn't
have
an
issue.
We
should
be
able
to
put
something
together
like
that:
I've
never
done
it
before
and
of
course,
Prince
appointed
me
to
you.
C
So
if
you
have
someone
I
can
work
with
in
the
future.
Please
let
me
know.
C
And
I
can
I
can
also
ping
you
if
you're
on
mute,
that's
okay
and
then
everybody.
There
was
a
note
in
the
the
Google
group
last
night,
an
introduction
from
his
name.
I
I,
don't
know
if
he
dropped
off
here
there
was
a
Oh
madhu
devawa
from
open
next
offered
to
help
us
last
night,
and
so
he's
actually
on
this
call
so
I
wanted
to
introduce
him
say
hello.
Marion
Madu
had
offered
to
help
us
and
he
joined
this
meeting
and
Chris.
You
had
responded
to
him.
G
On
this
open
networking
and
they
started
some
tools
and
Rising
the
request,
so
I
thought
of
helping
the
community
in
the
best
way
on
the
open
source
side,
so
yeah,
some
peers,
I'm,
visual,
doing
them,
updating
the
dogs
and
living
and
helping
these
other
questions.
G
B
Thank
you,
yeah
yeah,
it's
interesting
modu.
Is
it
correct
that
you
found
the
dash
project
through
that
other
exchange?
We
had
on
the
htsc
support,
yeah,
exactly
yeah,
so
just
a
little
background,
because
it's
funny,
it's
funny
coincidence
how
things
work,
but
Madu
somehow
found
my
P4
project.
I
did
a
year
ago
for
the
P4
Workshop,
where
I
demonstrated
Snappy
connected
to
the
behavioral
model
as
a
as
an
exercise.
I
did
a
year
ago
and
Madu
found
that
project
which
was
kind
of
unmaintained
it
was.
B
It
was
like
a
POC
I
did
a
year
ago.
He
found
some
problems
with
it
and
we
were
interacting
it's
an
open
source
project
and
I
said
by
the
way,
we're
going
to
be
doing
the
same
thing
in
dash,
so
he
hopped
onto
the
dash
project
and
immediately
started
playing
with
things.
B
So
you
know
it's
a
happy
coincidence
and
and
Mandu
I
think
where
you
could
probably
help
once
we
get
this
pull
request
and
the
framework
stabilized
sure,
since
you've
shown
a
proficiency
already
and
an
interest
in
playing
with
the
snappy
scripts
to
do
traffic
generators
with
pmb2
I
think
that'd
be
an
excellent
place
to
contribute,
because
we
just
don't
have
enough
hands
to
do
all
the
work.
So
you
know
I'd
be
happy
to
talk
to
you
about
that.
Once
things
get
a
little
firmer
that'd
be
an
excellent
place
to
help.
G
C
So
nice
to
see
you
does
anyone
else,
have
anything
for
this
meeting.
C
Okay,
then
I'll
go
ahead
and
stop
the
video
post
to
the
YouTube
channel,
the
the
normal
stuff
and
Marion
did.
We
have
something
we
would
see
next
week
for
the
for
the
behavioral
model,
outside
of
Chris's
potential
presentation
that
I
can
put
on
the
agenda.
A
Yeah
there
are
updates
that
Prince
had
and
his
design
that
we
need
to
bring
into
behavioral
model
first
being
admin
State
for
eni.
The
second
is
a
slightly
different
kind
of
unit
route,
so
probably
I
will
I
will
have
a
sink
with
friends
and
I
will
introduce
those
changes
into
behavioral
model.
We
can
review
it
next
week.
D
When
are
we
going
to
have
the
overlay
ready
in
the
behavioral
model,.
D
B
A
Oh
connection
tracking,
yes,
so
there
was
a
delay
we
wanted
to
have
our
proposal
to
DNA
done
in
order
to
support
cases
that
are
relevant
to
dash
it
is.
It
was
being
delayed.
So
if
we
won't
make
any
progress
in
PNA,
we'll
just
Fork
their
architecture
and
do
it
first
in
our
simulator
and
then
when
it
eventually
will
come
to
to
the
DNA,
we
can
get
back
in
line
with
them.
But
that's.
A
We
had
basically
a
delay
in
the
proposal
to
p
a
architecture,
but
it
seems
to
be
taking
too
much
time,
so
we
will
first
do
it
for
Dash
to
take
higher
priority
and
have
this
working
in
the
simulator.
So
it's
like.
E
Yeah,
that's
also
Community
Based
Marianne.
If
you
can,
you
know
if
you
want
to
present
in
the
p
a
group,
you
know
you
can
always
schedule
the
time
with
Andy
or
I
can
inform
and
if
you
like,.
A
I
Yes,
so
Marion
I
was
wondering
if
there
are
any
issues
related
to
the
simulator
code
that
are
in
the
pipeline
or
because
I
can
see.
All
the
issues
that
are
open
in
the
backlog
now
are
related
to
the
P4
Dash
pipeline.
So
I
was
wondering
there
if
there
is
anything
that
is
in
the
pipeline,
related
to
the
simulator
code
or
any
issue
that
is
upcoming.
A
Not
simulator
directly
because
it
was
already
fixed,
so
we
have
match
types
that
are
new
to
the
simulator,
which
is
list
match
types,
but
they
were
added
already
by
kasim.
If
I'm
not
mistaken,
however,.
E
A
Still
have
we
still
have
missing
is
in
the
compiler
itself
to
support
it,
so
not
really
in
the
simulator,
but
we
have
gaps
in
the
compiler.
J
So
Marion
this
piece
is
needed
for
us
to
have
the
complete
ACL
implementation:
correct.
Yes,
okay,
so.
J
Have
just
you
know,
I
guess
placeholder
implementation
of
B4
it
is.
It
is
dependent
upon
that
and
then
you
know
subsequently,
once
we
introduce
the
V6.
That
also
requires
this
piece.
A
Thanks,
we
think
this
is
don't
think
that
affects
anything.
C
I
thought
that
for
the
V6
Echo
we
were
waiting
for
the.
Let
me
look
here:
I
thought
we
were
waiting
for
the
connection
tracking.
C
Okay,
sorry,
by
the
way
honey
I
created
the
issue
for
you
separately
here
on
line
nine
I,
don't
know
if
you
can
see
it.
C
J
E
J
Know
let
me
take
a
look
at
it.
Whether
or
not
V4
first
needs
to
be
tackled
before
I
can
reach
to
the
V6.
C
Okay
thanks,
oh
couple
of
pieces
of
good
news
guys,
a
couple
of
things
are
in
the
Dunn
State.
Now
we've
moved
a
couple
to
done
so
here.
Marion
moved
these
two
here
to
the
Dunn
state,
so
we're
making
progress
moving
forward.
So
that's
a
good
thing
and
probably
at
some
point,
we're
gonna
need
to
get
these
backlog
items
and
owner.
So
maybe
we
can
work
on
that
next
time.
G
H
A
Have
we
actually
merged
this?
It
required
a
fork
for
behavioral
model.
G
Oh
yeah,
that
is
actually
maybe
I
I'm
I'm,
not
sure
on
that,
but
the
branches
that
you,
whatever
we're
talking
about
earlier
Branch
I've
based
on
this
issue
now
I,
am
not.
A
We
need
to
Fork
the
simulator
repository
in
order
to
take
the
changes
for
to
support
list
matches.
We
don't
have
a
public
Fork
for
dash
community
of
the
simulator,
so
that's
Christina.
What
I
asked
you
last
time
we
talked
yeah
on
Monday,
no.
C
I
think
that,
though
Chris
was
this,
what
we
were
showing
kasim
how
to
do.
I
Yes,
actually,
that
is
what
about
what
we
have
any
change
of
emails,
so
yeah
I
got
that
Chris
and
thank
you,
Christina
and
Chris
for
that.
So.
I
In
that
they
were
saying
that
we
don't
actually
need
to
create
a
fork
for
under
the
Azure
project,
so
we
can
actually
create
a
PR
directly
from
our
development
Fork
to
the
Upstream
yeah.
So
that's
why
I
think
we
don't
actually
need
the
four.
I
A
I
I,
don't
think
anyone
will
accept
these
changes
and
to
Upstream,
because
those
matched
kinds
are
not
part
of
the
architecture
of
the
target
architecture
that
is
used
by
the
compiler.
We
are
using
a
modified
Target
architecture.
I
Okay,
so
in
that
case,.
C
A
I
B
Yeah,
that's
one
thing:
I'd
mentioned
kasim,
but
it
might
be.
Marion
pointed
out
a
subtlety
and
not
a
subtlety
said
you
can
do
this
process,
but
you
still
have
to
then
you
know
deal
with
the
process
of
upstreaming
and
I.
Didn't
I,
don't
know
what
the
barriers
to
that
are,
but
I
guess
it's
a
fundamental
architecture
thing.
E
Great
I
just
spoke
with
Andy,
and
he
says
that
the
changes
PR
that
you
have
raised
Marianne
are
a
fairly
small
change
and
it
makes
it
easier
to
implement
in
Open
Source
before
c
compiler,
and
also
BMV
too
so
he's
looking
into
it
right
now.