►
From YouTube: DASH Workgroup Community Meeting 20220112
Description
January 12, 2021 Community Call
A
Beam
for
everything.
Next,
we
can
before
running
the
simulator.
We
need,
of
course,
to
compile
the
p4
code
and
the
target
is
like
this:
it's
a
series
pipeline.json.
A
B
A
This
will
start
this
simple
switch
simulator,
and
this
is
actually
something
that
I
was
asking
some
help
with,
because
it's
missing
a
few
things
like
those
special
matches
and
we
have
some
work
for
the
connection
tracking,
so
future
improvements
will
be
there
as
well,
but
for
now
it
it
is
running
a
reduced
set
of
our
dash
pipeline.
That
does
not
include
connection
tracking.
A
A
A
Where
we
can
run
the
simulators
you
like
so
the
command,
for
that
would
be
simple
switch
cli,
as
you
can
see
in
here,
and
now
it's
ready
to
receive
a
configuration,
and
this
is
actually
the
next
step
that
I'm
working
on
right
now
and
I
plan
on
showing
you
the
update
in
the
next
couple
of
weeks,
which,
instead
of
using
the
p4
runtime
native
tools
for
programming
the
switch,
we
will
actually
generate
a
sci
library
that
is
compatible
with
this
software
simulator
and
we
can
use
our
common
site
to
actually
program
the
software
switch
as
well.
A
But
for
this
demo
we
have
a
native
tool
being
used,
so
we
just
copy
this
configuration.
So
if
you
look
at
it,
it's
if
you
were
familiarized
with
the
before
implementation
of
dash.
This
is
just
adding
entries
to
a
variety
of
tables
like,
for
example,
the
red
direction:
lookup
eni
lookups
eni
to
vni
mappings,
ca
to
pa
mappings,
routing
some
prefix
in
here
and
all
of
the
resulting
mappings
like
what's
the
pa?
What's
the
overlay
dmac,
which
bni
should
we
use
and
so
on?
A
So
this
is
kind
of
less
intuitive
right
now,
but
when
we
will
have
a
side
to
side
to
p4
runtime
mapping,
it
will
be
much
clearer
because
it
will
be
exactly
what
sonic
is
going
to
use,
but
for
now
we
just
copy
that
configuration
you
just
pass
it
to
the
cli.
A
B
B
A
Right
yeah
so
we're
just
using
skappy
as
an
example.
So
here
we
have
we're
building
our
packet.
This
is
a
vxlan
packet
that
that
is
destined
to
our
appliance.
So
underlay
destination
is
the
appliance
ip
and
here
are
the
customer
addresses
one
on
one
one,
one
two
so,
as
I
said
no
connection
tracking,
for
now,
it's
just
one
packet
processing.
So
but
so
we
are
only
verifying
the
v-net
routing
and
a
mapping
parts,
not
the
policies.
A
But
we
can
send
this
packet,
so
we'll
run
tsp
dump.
Basically,
the
software
switch
right
now
has
two
interfaces.
It's
completely
flexible,
but
just
for
the
manual
use
I
created
by
default,
two
interfaces.
One
is
ingress.
Another
one
is
egress.
We
can
do
more
realistic
models
as
well
and
on
one
it's
coming
in
on
the
other.
A
A
And
it's
got
the
as
you
can
see:
source
ip
destination,
ip222122
and
so
on.
What
what
we
just
put
into
the
scapi
and
here
all
the
mappings
are
done.
We
have
our
vnet
vni
set.
We
have
our
overlay
dmacc
set.
We
have
our
c82pa
mappings
done
like
this
one,
one
one
two
address
maps
to
the
underlay
two
two
to
two
yeah.
So
this
is
one
example.
Probably
there
is
another
packet,
let's
see.
A
Yeah,
so
we're
just
changing
this
time,
ca
to
a
different
value
which
will
result
in
a
different
mapping.
A
So
this
time
we'll
get
a
so
before
it
was
a
destination
vni.
If
I'm
not
mistaking
this
time,
rules
are
saying:
please
use
the
source,
vni
and
yeah
so
that
that's
the
difference
basically,
so
this
is
how
to
operate
it
and
a
couple
of
words
about
the
next
steps
me
and
a
couple
of
microsoft.
A
Engineers
will
be
working
on
bringing
this
simulator
to
the
to
the
full-blown,
let's
say:
sonic
simulator,
where
we
can
spawn
in
the
instance
of
a
virtual
switch
with
all
of
the
sonic
interfaces
plus,
I
am
working
on
a
sci-2p4
runtime
translation,
so
we
will
have
psi
api
for
configuring,
this
switch
and
then
upon
improving
the
p4
code
itself.
We
will
get
more
and
more
functionality
from
this
simulator.
A
So
this
is
our
step
number
one,
and
in
a
couple
of
weeks
I
will
have
an
update
on
on
the
future
progress.
D
D
I
can't
actually
complete
the
process
and
I
think,
there's
another
dash
participant
who
had
the
same
problem.
I
did
so
if
you
can
go
back
and
look
at
those
comments.
If
anyone
else
wants
to
try
to
build
these
containers
too,
if
there's,
if
they
have
success
or
failure,
it
would
be
nice
to
get
some
user
feedback.
A
Yeah
so
there
might
be
some
difference
in
the
environment
and
it's
good
that
you
tried.
So
I
will
take
a
look,
let's
see,
what's
missing
sure.
C
Fantastic,
thank
you
very
much
and
it
looks
like
we're
running
at
about
time,
so
the
action
items
for
next
meeting
would
be
for
the
community
to
go
ahead
and
look
at
the
simulator
look
at
the
pull
requests,
michael
and
I
will
post
document
or
post
drawings
as
soon
as
as
soon
as
we
make
a
few
more
refinements,
we'll
put
in
the
request
to
do
that.
We
want
gerald
any
closing
thoughts
for
this
week.
I'm
sorry
christina.
F
I
don't
know-
maybe
I
I
missed
the
part,
but
I
I
had
understood
that
you
would
have
demonstrated
also
the
the
compiler
that
generates
the
the
psy
api.
Did
they
use
it
or
do
not
present?
Oh.
A
No,
I
I
did
not
include
that,
so
let
me
do
that
next
time,
along
with
the
generation
of
the
library
itself,
because
my
next
target
is
psi
to
before
anti-mapping.
So
we
can
use
site
to
program
that
so
I
will.
I
will
do
it
next
time.
F
And
another
question:
we
have
very
quickly
so
this
this
emulator
does
not
support
those
extensions
that
we
spoke
about.
So
how
do
you
deal
with
that?
You
simply
compile
without
that
piece
of
code
like
avoid
compiling
that
part
of
the
code
or
or
somehow
you
just
skip
it
or
the
the
compiler.
A
So
if
you,
if
you
notice
those
are
under
the
define,
so
all
the
connection
tracking
is
not
compiled.
F
B
Christina
I
have
a
non-technical
question
for
you.
I
couldn't
attend
the
last
two
meetings
and
I
was
trying
to
get.
The
recordings
looks
like
they're,
restricted.
C
I'll
take
care
of
it
right
gerald
leave
any
closing
thoughts.
G
And
I
just
want
to
say
that
I
really
appreciate
and
we're
looking
forward
to
really
hear
all
the
feedback
that
good
or
bad,
I
think
joining
us
on
the
chat.
I
think
there's
a
few
very
good
questions
and
yeah
I
I
we.
We
always
try
to
learn
from
the
community
and
see
your
feedback,
criticism,
support
or
like
and
dislike.
G
So
please,
let's
work
together
on
this
right.
C
Yeah,
oh,
and
I
think
someone
else
mentioned
something
too:
oh.
D
I
just
wanted
to
share
my
screen
briefly,
and
I
won't
spend
much
time
on
this,
but
there
is
there
are
a
couple
pull
requests
that
worked
on.
One
is
with
this
diagram
and
a
description
of
it.
This
is
the
p4
variations.
This
is
the
one
that
marion
is
working
on.
That
represents
it,
and
this
is
the
one
that
intel
is
working
on
and
I've
I've
talked
back
channel
to
to
vet
these
diagrams.
D
So
I
think
they're
fairly
representative
and
there's
a
description
of
of
all
this,
so
people
could
take
a
look
at
this
pull
request
at
some
point.
I'd
like
to
get
it
accepted
and
then
there's
another
request
that
talks
about
what
I
call
dash:
testing
maturity
stages
and
I'm
just
trying
to
lay
the
groundwork
for
how
the
data
plane
can
be
tested
bottom
up
in
various
stages
and
also
how
sonic
can
be
tested
top
down
and
I've
just
put
some
thoughts
together.
If
people
could
take
a
look
at
that,
so
lay
some
groundwork.
E
Okay
thanks
chris
anyone
else,
yeah
christina
one
last
question
for
leeward,
so
for
the
smart
switch,
all
the
use
k
and
the
packet
transformer
will
be
applicable
to
smart
switch
as
well.
G
Yeah,
that's
what
we
believe
right,
because
the
yeah
it's
just
all
package
transformations
right
and
smart
switch.
My
basically
hiring
a
a
hybrid
of
pipeline
right.
The
tool
pipeline
might
be
a
little
bit
different
in
between
the
switch
asic
and
the
and
the
dpu,
but
but
the
package
transformation
itself
by
definition,
is
independent
of
the
hardware.
Yes,.
E
Yeah
another
question
is
the
intelligent
of
where
you
place,
like
you
mentioned
earlier,
depends
on
the
resource
of
the
switch
versus
the
dpu.
You
may
place
your
connection
or
you
know
whatever
the
tunnel
programming
differently,
so
the
intelligent
will
be
an
order
agent
strictly,
and
nowhere
else
is
that
the
right
understanding.
G
C
Okay,
thanks
everyone.
I
will
release
you
to
your
10
o'clock
meeting.
I
appreciate
your
time
and
I
will
see
you
next
week,
please
reach
out
to
me
specifically
if
you
need
to
access
recordings
etc.
Just
give
me
an
email.
Thank
you.
Thank
you.
Okay.
I
appreciate
y'all,
okay,
bye-bye.