►
From YouTube: CNCF SIG Network 2021-03-04
Description
CNCF SIG Network 2021-03-04
A
A
Hello,
hello,
hey
hey
mr
rose
ken
to
the
many
unanswered
emails
that
I
have
back
to
you
or
or
to
your
latest.
One.
Yes,
is
the
the
answer
that
there
are
some
of
the
things
that
we're
going
to
talk
about
today.
B
B
He
wanted
to
update
it
and
submit
it,
and
he
asked
if
I
knew
anyone
else
that
had
interesting
things
going
on
in
the
space,
and
I
thought
of
you
so
so
you
definitely
don't
need
me
to
collaborate
on
it
and
you
can
go
in
with
you
know
you
and
you
know
whoever
you're
working
with
names
on
it
for
sure.
Okay,.
B
B
A
Very
good
we've
got
some
the
folks
that
are
on
the
phone
if,
if
you've
got
access
to
the
dock,
please
drop
your
name
in
record.
Your
attendance
we've
got
maybe
one
more
minute
before
we
get
going
in
in
this
minute.
It's
a
great
time
to
add
agenda
items
to
the
bottom
of
the
list
and
actually
seeing
daniel
here
is
great.
Daniel,
welcome
hello.
Speaking
of
of
unanswered
emails,
I'm
going
to
have
an
awkward
moment
here
in
about
10
minutes.
I
think.
A
Very
good,
okay
dan,
if
I,
if
I
may
daniel,
if
I'm
if
I
may
add
a
conversation
on
submariner
as
a
as
a
topic,
if
that's
how,
if
that's
helpful
or
wanted
today,
but
if,
if
not,
then
you
know
feel
free
to
take
it
off.
A
Oh,
you
know
I'm
gonna,
I'm
dropping
submariner
in
as
a
as
a
topic
assuming
that
that's
desired,
but
okay.
C
Yeah
we,
my
plan
for
today
was
just
to
say,
hello
and
then
try
to
figure
out
a
time
to
give
a
presentation.
I
wasn't
planning
on
doing
the
presentation
today.
If
that's
okay,.
A
Sounds
sounds
good
thanks,
nice
nice
to
have
you
daniel
yeah
daniel,
since
we're
shooting
the
breeze
for
just
the
the
briefest
of
moments.
Anybody
else
here
with
you.
C
Yeah,
I'm
trying
to
kind
of
help
with
our
cncf
effort
been
pushing
that
for
a
while
now
just
hurting
sheep
yeah.
There
we
go.
A
Very
good,
well
we're
folks
we're
about
six
after
let's,
let's
get
going,
if
we
could
there's
a
couple
of
you
whose
names
are
relative,
maybe
perhaps
fresh
to
me,
which
is
some
super
encouraged
to
see
anita
here
as
well
as
daniel,
and
then
we've
got
some
some
some
familiar
names
on
the
call
as
well.
A
This
is
what
I'm
about
to
do
is
slightly
atypical
for
most
of
the
cncf
calls
that
I
that
I
see
and
that
I
participate
in,
but
I
like
to
do
it
on
this
call
anyway
and
if
it
makes
anyone
feel
uncomfortable,
don't
don't
sweat
it,
but
I
wanted
to
so
daniel
and
I
just
had
a
brief
interaction
daniel
that
was
pleasant.
Hopefully
you
feel
encouraged
and
welcome.
I
thought
yes,
thank
you
good
good,
good
anita.
I
figured
it'd,
be
nice
to
say
hi
or
to
let
you
say
hi
to
everyone.
A
And
yeah.
A
D
So
hi,
my
name
is
anita,
I'm
actually
just
subbing
in
a
bit
for
one
of
my
colleagues
who
comes
here
more
regularly,
christopher
hansen
he's
sleeping
because
he's
teaching
in
india
time
class
now,
but
we're
both
from
arcsum
we're
a
cloud
native
training,
consulting
firm
and
we
just
like
to
stay
on
top
of
things
with
cncf
and
always
great,
to
observe
and
see.
What's
going
on
in
the
space,
so
hi,
everyone.
A
Hello,
hello,
nice,
very
good,
very
good
thanks
for
joining
yeah
and
folks-
I
don't
I
don't
say
this
enough.
Ken
lets
me
ramble
on
and
give
you
know
say
bad
jokes
and
that
kind
of
a
thing,
but
so
so
ken
ellens
and
my
and
myself
are
here
as
co-chairs
of
the
sig
network.
Just
to
familiarize
everybody
a
little
bit,
we
have
sig
network,
it
has
trying
to
think
of
not
very
strong
language,
but
it's
the
home
base.
A
A
It's
proposed
for
incubation.
It's
been
proposed
for
a
while
now
it's
out
for
public
review.
So
if
you
have
they,
I
know
that
they
appreciate
supportive
votes
and
feedback.
Otherwise,
so
everyone
here
is
welcome
to
go,
make
a
comment
or
go.
Go
show
support
another
thing
that
was
relevant
not
to
sig
network,
but
to
all
the
sigs.
Is
that
and
maybe
particularly
to
sig
network?
A
Daniel
we
actually
happen
upon
the
topic
of
submariner
fairly
fairly
quickly
daniel
for
those
that
might
not
be
familiar
with
with
the
project.
Did
you
mind
I'm
giving
to
just
letting
people
know
kind
of
high
level
what
the
project
is
about
and
we'll
put
a
link
to
it
so
that
people
can.
C
Yeah
sure
so,
just
in
a
few
seconds,
it's
a
multicultural
networking
project
we've
been
working
with
the
multicultural
networking,
sig
and
kubernetes
following
really
closely
their
cap
process,
implementing
multicultural
networking
and
service
discovery
right
behind
them.
Basically,
it's
just
a
healthy,
open
source
project
that
we're
hoping
to
get
into
the
cncf
as
well.
It's
kind
of
the
angle
that
I'm
here
for
yeah
happy
to
talk
about
other
details.
If
people
have
questions
but
we'll
give
a
longer
presentation
with
all
kinds
of
details
later.
A
Beautiful
and
daniel
the
canonical
link
to
the
project
or
the
the.
A
C
A
Gotcha,
oh
okay,
okay
boy.
That
seems
like
forever
ago
that
I
actually
saw
another
human.
A
If
you're,
you
know,
I
recommend
folks
to
peruse
the
slides
here
which
talk
about,
I
think
about
five
different
initiatives
within
the
working
group,
and
my
hope
is
that
we'll
talk
about
too
in
depth
today,
because
this
initiative,
I'm
sorry
because
the
working
group
does
have
a
number
of
sub-working
groups
and
much
to
chat
through
and
things
to
execute
on,
there's
a
there's.
The
cncf
sig
network
mailing
list,
which
I
believe
is
the
url
all
the
way
up
to
here.
A
Cncfc
network,
there's
a
a
separate
mailing
list,
a
sub
mailing
list,
if
you
will
for
the
service
mesh
working
group,
so
that
everyone
isn't
being
everyone,
who's
focused
on
cni,
for
example,
or
on
nats,
or
something
isn't
being
bothered
with
a
bunch
of
service
mesh,
specific
stuff
or
to
the
extent
that
they
want
to
be
that
they
would.
They
would
sign
up.
So
please.
You
know
if
you're
not
signed
up
on
that
list
and
you
find
what
we're
talking
about
today.
Interesting,
go
sign
up.
A
Tell
your
friends
two
of
the
initiatives
to
to
cover
today.
One
is
an
update
on
get
nighthawk.
I'm
gonna
introduce
people
to
nighthawk
who
aren't
as
familiar
of
which
there
may
be
a
couple.
Nighthawk
is
well,
it's
well
simply,
like
simply
stay
over
overtly,
simply
stated:
it's
a
load
generator
that
doesn't
do
the
project
justice.
A
It's
a
it's
a
performance,
it
does
performance,
characterization
of
layer,
seven
of
layer,
seven
traffic
for
grpc
for
http,
for
for
http.
Two,
for
you
know,
has
a
lot
of
features
and
there's
been
an
open
source
initiative
to
help
get
nighthawk
into
the
hands
of
many
and
into
one
of
the
other
projects
that
we
that
we
talk
about
on
the
meshery
project.
A
E
Yeah,
that's
right,
yeah
thanks
lee
for
calling
up,
so
I
just
want
to
give
a
quick
progress
update
on
what
we
had
where
we
had
left
last
meeting
so
I'll
just
share
my
screen
in
one
second.
E
So
it's
invisible,
it
is
all
right,
so
this
particular
documentation
is
being
used
to
sort
of
get
the
track
track,
the
design
spec
and
the
updates
and
the
ongoing
process
of
this
particular
project
and
get
night
off.
Last
time
when
we,
when
we
had
this
meeting,
we
discussed
about
the
the
area
of
concentration
on
what
that.
E
What
are
the
next
steps
that
we're
going
to
take
or
regarding
this
project,
so
we
discussed
a
couple
of
items
on
the
ci
process,
how
we're
gonna
publish,
build
the
artifacts
and
publish
night
off
releases
or
on
different
versions.
E
So
I
I
did
put
up
a
design
spec
about
it
over
somewhere
over
here
or
mentioning
some
of
the
key
features
like
the
artifact
conventions
and
oh
one,
sec
yeah,
basically,
the
whole
whole
design.
Spec
ideas
have
been
updated
in
this
portion,
so
for
someone
who's
not
checked
it
out,
yet
they
can
go
over
here
to
look
at
it.
So
right
now,
we've
we've
managed
to
take
a
couple
of
steps
ahead
for
on
the
ci
process.
E
Basically,
we've
got
a
repository
where
we
are
currently
building
a
custom
ci
for
for
publish
publishing,
artifacts
on
get
night
on
the
night
of
on
different
distributions
of
night
of
binaries,
so
so
yeah
like.
Basically
this
the
progress
is
on
that
we
created
a
custom
action
yeah.
E
This
basically
is
the
definition
of
the
action
which
takes
in
certain
parameters
like
the
repository
to
publish
to
a
version
of
the
night
up
to
be
released,
the
operating
system,
architecture
based
on
every
different
parameter
that
we
are
going
to
release
these
different
versions,
and
then
we've
got
also
the
ci
setup
ready
to
be
published
and
right
now,
we've
we've
added
support
for
only
two
of
the
operating
systems
which
are
ubuntu
and
a
mac
os.
E
So
these
are
the
two
workflows
that
have
been
established
now
but
ongoing.
Maybe
hopefully
the
next
meeting
will.
E
Be
so
yeah
like
that,
that's
basically
more
or
less
the
update
on
on
the
ongoing
task
of
getting
dogs.
A
E
Oh
okay,
I'm
sorry
about
that.
So
what
I
was
saying
is
that
next
time
we
meet,
we
might
cover
two
more
distros,
which
would
be
a
fedora
that
is
red,
hat
and
also
windows.
We
will
cover
for
windows,
distributions.
A
Any
if
there's
any
along
with
that
vinayak
you've
been
working
with
a
few
different
contributors
but
have
been
busy
behind.
Actually,
I
think
there's
a
video
of
you
being
busy
behind
the
scenes
building
out
the
the
project
website.
Do
you
I'm
going
to
put
you
on
the
spot?
I
did.
A
Do
you
have
an
update
there
that
you'd
like
to
share.
F
F
F
Is
my
screen
visible?
It
is
okay,
so
basically,
this
is
the
site
that
we
have
been
working
on
for
get
nighthawk
and
so
far
we
have
achieved
all
this
on
the
site
and
we
plan
on
to,
I
think,
implement
a
docs
collection
for
this
website
and
a
customer
review
section
in
the
future,
and
I
also
a
couple
of
more
things
are
on
the
list.
For
example,
a
video
section
to
demonstrate
some
of
the
things
and
resources
for
get
nighthawk,
and
example,
implementations
other
than
this.
F
We
are
currently
looking
currently
making
a
decision
on
what
the
logo
for
get
night
talk
is
supposed
to
be.
We
have
a
few
drafts
of
it,
ready
and
currently
doing
a
informal
vote
on.
F
A
Nice
there's
a
link
to
those
draft
logos
inside
of
the
meeting
minutes,
so
so
for
anyone
who
has
comments
or
who
has
a
design
that
they
think
should
be
be
the
design
that
represents
get
nighthawk.
Please
please.
A
A
Good,
I
was
hopeful
that
there
are
a
couple
of
other
project
maintainers.
That
said,
they
couldn't
make
it
today,
but
they
have
been
responding
asynchronously,
one
of
them
in
particular,
really
avoids
any
synchronous
communication
at
all
costs,
so
so,
but
anyway,
pleased
to
see
the
engagement
pleased
to
see
the
support
from
nighthawk
maintainers.
So.
A
Any
questions
or
comments
on
get
nighthawk
for
abhishek
for
vineeth
urbanaik.
A
As
we
go
to
transition
to
the
next
topic,
I'll
I'll
say
this:
if,
if
nighthawk
is
somewhat
unfamiliar
to
you,
there's
a
couple
of
really
neat
things
that
that
load
generator
is
capable
of
that
is
part
of
the
genesis
or
part
of
yeah
part
part
of
the
logic
and
the
genesis
behind
get
nighthawk
itself
the
world,
the
the
rest
of
the
world
needs
to
get
their
hands
on
it.
There's
some
interesting
ways
to
characterize
performance.
A
So
speaking
of
performance
characterization,
the
next
topic
is
service.
Mesh
performance
and
sunku
is
here
with
us
perfect,
simply
there's
a
couple
of
folks
that
that
I'm
gonna
put
you
on
the
spot
as
well.
This
is
entirely.
B
A
Know
I
don't
know
that
this
is
fair
or
not,
but
there's
some
of
the
folks
that
are
on
the
call
like
blake
who've,
been
around
the
project
for
a
long
time
and
others
just
coming
to
to
learn
what
it
is.
Do
you
do?
Do
you
mind
doing
a
practice
introduction
of
what
the
project
is.
C
G
That's
good,
so
we've
been
discussing
about
smp
last
a
couple
of
times.
I
guess
the
idea
where
we,
the
discussion,
came
into
the
point
that
there
are
quite
a
few
load.
Generators
like
like
they
mentioned,
and
and
also
what
I
found
running
some
benchmarks
on
service
mesh
type
applications
that
there's
some
gaps
with
respect
to
how
performance
measurement
could
be
done
between
layer,
3,
tuning
versus
layer,
7,
unique
and
with
respect
to
load
generators.
G
So
in
general
the
idea
is,
you
know,
so
we
need
a
place
where,
where
we
could
discuss
some
of
the
methodology,
standards
or
features
of
how
performance
benchmarking
should
be
done
for
service
mesh
and
discussing
with
lee
we
found
smp
is
a
good
place
where
we
could
collaborate
and
build
on.
B
G
And
to
the
point
that
this
could
be
a
sandbox
project
within
cncf,
so
from
that
perspective,
smb
provides
the
the
right
constructs
with
respect
to
where
it
is
going.
One
in
terms
of
automated
deployment,
for
example,
measuring
using
measuring,
but
the
main
thing
is,
it
does
provide
integration
of
different
load.
Generator
tools
versus
different
service
meshes
along
with
it.
So
from
that
sense
you
know,
smb
can
be
a
place
where
we
could
discuss
and
build
on
that's
about
what
I've
seen
so
far
you
can.
You
can
fill
in
the
rest
early.
A
Yeah
so
see
it's
all
down
like
that:
zero
prep,
zero
warning,
it's
all
downhill
from
there
like
you'll,
hear
that
was
fantastic
to
try
to
add
a
little
more
flavor
to
to
what
sunko
had
just
said
is
that
if,
if
you're
in
the
meeting
minutes-
and
you
take
a
quick
look
at
the
service
mesh
working
group
charter-
well
or
you
can
look
at
the
charter-
or
I
guess
the
slides-
maybe
have
it
a
little
more
crisply-
I
can't
you
know.
A
For
my
part,
I
can't
tell
you
how
many
times
either
I've
been
directly
asked
or
I've
heard
people
ask.
What's
the
overhead
of
a
mesh
like
you
know
what?
How
much
is
this
going
to
cost
me,
and
how
do
I
characterize
that?
How
do
as
I
go
to
you
know,
use
it
more
and
more?
How
do
I,
how
do
I
make
sure
that
you
know
I've
got
an
ongoing
analysis
or
ongoing
hand
on
what
that
is,
is
did
I
make
the
right
selection
here
in
terms
of
technologies?
A
How
do
these
things
compare,
there's
so
many
variables,
and
so
you
know
there
needed
to
be
too
sunku's
point
like
a
forum,
for
you
know,
vendor
neutral
form,
for
some
of
that
discussion
for
the
advancement
of
some
of
the
studies
that
different
groups,
different
individuals
have
been
doing,
and
hopefully
a
common
way
of
capturing
and
describing
performance
characteristics
of
well
cloud
native
well
in
some
respects
cloud
native
workloads.
A
They
could
be
running
on
a
service
mesh
and
they
may
not
be
you
know
much
of
the
goal
here
is
centric
to
what
happens
when
they
are,
but
it's
like.
That's
probably
the
first
question
that
people
you
know
after
after
they
say
hey.
What's
the
overhead,
they
say
maybe
in
their
minds
one
of
the
easiest
ways
to
gauge
that
overhead
is
to
to
measure.
You
know
here's
how
their
app
is
running
today
and
here's.
A
How
much
that
you
know
to
characterize
its
performance
and
its
overhead
off
the
mesh
and
what
it
looks
like
on
the
mesh
and
then
they
start
to
tune
and
tweak
any
number
of
knobs
and
and
configurations
and
different
service
meshes
and
different
versions
and
and
part
of
the
part
of
trying
to
make
this
body
of
work.
Useful
is
after
having
engaged
with
a
couple
of
universities,
questions
around
what
applications,
what
cloud-native
workloads
are
represented?
A
Are
you
know
good
ones
to
study
and
use
and
disseminate
knowledge
on
which
ones
are
representative
of
you
know
the
most
common
workloads
that
others
would
have
so
just
even
the
topic
of
what's
what
are
good
sample
apps,
which
ones
are
available?
Where
are
they
available
from
cornell,
had
been
there's
a
professor
there
that
had
worked
on
a
few
they've
published
a
couple.
I
think
you
know
much
like
it.
Okay,
it
was
one
one
topic,
another
kind
of
topic
to
add
additional
flavor
to
what
sunken
was
saying.
A
Is
that
well
there's
any
any
number
of
configurations
of
not
just
the
infrastructure,
the
mesh,
the
what
it's
doing
your
applications,
how
many
nodes
and
clusters,
and
all
that
all
that
stuff,
but
there's
also
a
myriad
of
test
configurations.
So
is
this
a
silk
test,
a
capacity
test,
a
stress
test:
does
this
go
on
for
10
days?
Does
it
go
on
for
10
minutes?
A
You
know.
Are
you
how?
How
is
the
how
you
know,
there's
all
kinds
of
knobs
to
tweak
on
the
load
generators?
How
are
they
behaving
different
low
generator,
different
load,
generators,
measure
differently,
there's
just
there's
just
so
much
to
all
of
this.
A
That
in
large
part,
like
the
genesis
of
why
this
stuff
was
created,
you'll
often
hear
about
smp
and
to
suku's
point
like
the
one
of
the
first
tools
to
implement
the
spec
measuring
you'll
hear
them
talked
about
often
together
one
because
because
these
are
both
new
and
kind
of
driven
towards
part
of
the
same
goal.
That
part
of
their
shared
same
goal
is:
is
that
of
making
it
easy
to
repeat
and
run
these
tests,
but
not
necessarily
against
sample
apps
against
users
apps?
A
You
know
against
whatever
those
are
in
whatever
environment
they
have,
because
because
it's
all
bespoke,
you
know
to
their
environment,
and
so
it's
about
empowering
others
to
go
off
and
answer
these
questions
because
part
of
the
goal
and
part
of
what
I'm
hopeful,
that
you
know
sunku
and
the
rest
of
you
that
are
here
will
commit
to
and
and
that
we
will
collectively
go
off
and
do.
A
Is
know
defining
working
with
these
service
mesh
groups
or
or
those
that
aren't
running
on
a
mesh
fine.
What
are
what
are
those
representative
workloads?
What
what
I
think,
while
I
personally
consider
almost
any
almost
any
benchmark
almost
any
load
test
that
you
want
to
define
and
run
is,
is
valid.
A
Some
of
them
are
not
very
useful
because
you
know
no
one's
got
that
type
of
a
configuration
or
or
that
load
never
hits
their
services
in
that
way,
that's
important
where
I'm
you
know
excited
that
suku
and
the
rest
of
you
are
helping
advance
this,
because
those
are
things
that
that
many
of
you
know
a
lot
better.
G
Yeah
to
add
on
to
that
what
we
see,
especially
in
terms
of
blogs,
being
published
a
lot
of
folks
have
done
different
type
of
analysis
and
some
performance
numbers,
whether
you
see
istio
or
onward,
all
right,
so
there's
some
suggestions
of
what
could
be
done,
but
then
they
need
a
place
and
a
little
bit
more
traction
in
terms
of
you
know:
establishing
standard
patterns
like
you
mentioned
and
also
best
practices
to
see
how
best
to
characterize
a
mesh.
G
You
know
what
are
the
aspects
to
look
for,
and
that
is
not
there
today
from
what
I
see
across,
and
so
how
can
we,
I
guess
the
question
is:
how
can
we,
you
know,
bring
together
that
type
of
discussion
to
here
and
ensure
that
you
know
some
of
these
work
is
represented
across
multiple
meshes
or
methodologies
that
being
discussed
online?
G
How
can
we
make
it
uniform,
so
that's
easier
to
understand,
consume
all
right
so
and
one
other
thing
lee
discussed
was
having
this
project
as
a
sandbox
project
in
their
cncf,
so
that
it's
has
but
much
better
visibility,
make
it
easier
to
for
folks
to
contribute
and
also
participate,
but
and
just
open
to
ideas
as
to
see
can
it
bring
in
opinions
as
to
what
else
we
could
do
here.
A
Yeah,
everyone
else
is
getting
put
on
the
spot
today.
This
is
there's
a
you
know.
Maybe
I'll
have
one
more
comment
and-
and
that
is
like
soon
to
your
point-
that's
one
of
those.
A
H
B
H
D
H
A
There
thank
you
josh
that
that's
not
yeah,
yes,
actually
from
from
my
part,
I
would
love
to
for
others
to
bear
weight
on
that
or
to
express
opinion
on
how
important
the
latency
is
and
the
measure
you
know
the
long
tail,
latency
and
cpu,
maybe
cpu,
doesn't
matter
at
all.
H
I
will
say
I
just
haven't
had
to
troubleshoot
some
apps
recently,
where
we
were
running
into
capacity
issues
and
finding
out
that
it
actually
wasn't
things.
We
were
setting
cluster
levels
actually
application
level,
but
still
still
played
into
it.
H
I
mean
having
that
visibility
across
not
just
things
that
we
see,
obviously
the
kubernetes
or
backplane,
which
is
where
they
thought
the
problem
was,
but
within
after
the
java
app
there's
a
memory
and
heat
limitation
that
they
were
hitting
up
against,
causing
the
app
to
keep
crashing
out,
and
then
they
thought
it
was
kubernetes
limit,
causing
it
to
then
regenerate
the
pods
but
being
able
to
see
it
from
the
app
level
and
then
get
through
the
through
what
what
our
settings
are,
and
actually,
I
guess,
if
no,
I
don't
know,
is
snp
pulling
in
values
for
limits
that
have
been
set
at
say
the
cluster
level.
H
H
Would
that
app's
hitting
you
know
80
of
that
limit,
not
just
saying
hey.
This
is
the
performance
you're
getting
out
of
it,
but
also
by
the
way
you're
you're
pushing
really
close
to
a
boundary.
You
may
want
to
be
aware
of
to
increase
that
limit.
I
know
that
kind
of
visibility
is
very
helpful
for
us
ops
folks,
because
we
don't
know
when
to
go
increase
things
until
something
obviously
keeps
crashing,
and
somebody
calls
us
and
we're
like.
Oh
yeah,
you
keep
hitting.
You
know
limit.
Let
us
stuff
that
for
you
or
whatever,.
A
Right
because
you
may
be
hitting
the
upper
yeah,
you
may
be
hitting
the
upper
bounds
of
something
that
was
either
intentionally
or
maybe
unintentionally
configured.
But
it's
a
soft
limit.
That's
that's,
specified
and
then
causing
you
causing
you
issues
so
like
if
you're
not
yeah.
If
you're
unaware,
if
you're
unaware
of
the
fact
that
there's
a
governor
or
a
limit
to
find
that
then
yeah
you
may
you
may
be
off
trouble
right.
H
Yeah,
we
can
obviously
run
all
sorts
of
load
generation
tests
and
things
and
spec
out
and
figure
out
where
we
want
to
set
things
at
then,
once
you
actually
get
into
production,
it's
usually
a
different
beast
and,
like
you
said,
those
limits
that
are
set
are
sometimes
blind
to
the
application
owner
the
developers
they
don't
know
what
we've
said
at
the
operational
level
and
so
they're
being
called
in
trying
to
figure
out
why
the
app
is
not
stable
or
whatever.
Really
it's
it's
something
else.
H
A
A
You
know,
through
a
project
like
this
to
to
to
let
others
know
that
they
might
be
cha
on
a
wild,
goose
chase
that
like
hey
it,
you
know
some
certain
percentage
of
the
time,
you're,
actually
you're,
chasing
down
a
performance
issue.
That's
actually,
you
actually
configured
your
infrastructure
to
do
that.
A
Your
your
configuration
is
working,
you're
being
limited,
and
it
you
know
it's
like
anyway,
that
type
of
a
disco
like
a
that's
a
I'm,
going
to
try
to
I'm
going
to
make
a
note
about
consideration
for
that
and
asking
folks
that
are
on
this
call
and
elsewhere
to
to
comment
on
whether
or
not
to
include
limits
in
the
spec
and
then
even
to
the
extent
that
they
don't
I'm
josh.
To
the
extent
that
you
keep
coming
here.
A
A
You
know,
tests
that
have
been
run
and
results
and
things
that
they
can
see,
hopefully
on
on
something
of
a
regular
cadence,
is
to
empower
the
service
mesh
projects
themselves
to
measure
themselves
in
a
uniform
way,
use
and
publish
those
results
in
a
way
in
which
they're
in
hopefully
a
more
fair
way
than
they're
being
measured
today,
because
they're
being
measured
today
and
and
doing
doing,
these
measurements
is
pretty
pretty
difficult
and
those
results
being
published
and
a
lot
of
times.
They're,
not
they're,
not
it's
not
quite
fair.
A
It's
either
something's
being
compared
and
they're,
not
apples
to
apples
or
or
it's
not
being
compared.
But
someone
forgot
to
turn
debug
off
or
something
like
some.
Someone
forgot
to
actually
tune
their
configuration.
So
some
of
these
projects
are
getting
pie
in
the
face
when
they
shouldn't-
or
you
know,
egg
in
the
eye
and
they
shouldn't.
H
So
when
these
tests
are
running
are,
are
they
like
you
kind
of
set
up
a
traffic
simulation
of
what
you
want
to
run
through
it
or
are
there
like?
Is
there
like
a
real
world
mix
right,
because
I
I
I
know
what
you're
saying
I'm
being
a
networking
guy
all
the
time.
We
always
see
these
specs
of
what
throughputs
you
can
achieve,
but
that's
with
like
perfect
sized
frames
going
across
the
wire
like
they.
H
They
really
they
put
the
perfect
workloads
across
them
to
show
how
fast
something
can
work,
but
in
a
real
world
mix
we
see
a
lot
more
fragmentation
or
a
lot
more
a
lot.
You
know
much
smaller
packets
and
many
many
more
packets
that
need
to
be
handled.
Things
like
that.
So
is
there.
I
guess:
how
is
that
mix
that
mix
being
generated
and.
H
A
And
use
of.
B
A
Mesh
and
telling
the
mesh
to
introduce
some
chaos
is
one
mech.
One
way
of
you
know
mixing
up
the
tests
and
another
one
is
taking
advantage
of
the
load
generator
itself
and
its
ability
to
to
do
just
that.
Its
ability
to
inject
faults
or-
and
that's
very
much
so
squarely
in
the
wheelhouse
of
what
the
spec
should
be
considering
for.
H
Playback
capability,
so
this
this
is
actually
a
use
case
that
we're
looking
for
today
with
a
certain
application.
That's
in
an
environment
where
it
doesn't
have
the
dev
environment
that
we're
practicing
in
doesn't
have
the
same
access
to
components
that
the
production
environment
does
so
we.
H
What
we
want
to
do
is
just
take
like
a
packet
capture
or
record
the
traffic,
and
I've
done
this
with
other
products
on
the
market
that
sell
very
big
iron
appliances
that
can
record
a
certain
traffic
simulation,
and
then
you
go
and
play
it
back
in
your
test
environment
to
get
get
that
there
is
that
or
am
I
going
down
like
a
whole
rabbit
trail?
That's
not
even
a
part
of
this
products
capability
or
or.
A
You're
doing
a
beautiful
thing,
which
is
almost
exactly
what
cinco
had
done
last
time,
which
is
we
were
talking
about
services,
performances
we
were
talking
about
nighthawk
and
we
were
talking
about
measuring,
and
you
know,
part
of
his
question
was
hey
where's,
the
where's,
the
line
and
you're.
Actually
broaching
that
same
thing.
It's
like
you're
right
that,
like
hey
the
the
services,
has
a
specification.
A
What
is
it
capable
of?
Well,
it's
capable
of
capturing
and
characterizing
in
just
in
written
form
like
in
yaml
form,
capturing
and
characterizing
performance
and
how
you're
generating
load
what
your
environment
is
like
what
your
configuration
is
like.
You
know
like
doing
that
in
a
uniform
way,
so
it
so
it's
a
specification.
That's
like
unto
its
own.
It's
kind
of
static
in
nature,
it's
sort
of
like
you
know,
here's
a
document.
This
document
is
written
in
in
accordance
with
this
specification.
A
A
Smp
itself
isn't
isn't
a
capability,
but
it's
a
specification
that
other
projects
can
implement
and
the
first
one
to
to
have
implemented
is:
is
this
one
measuring
the
use
cases
of
like
traffic,
recording
traffic
playback
or
like
or
to
generate
test
after
test
or
like
performance
test
after
test
is?
Why
is
why
we're
talking
about
mesherie
nighthawk
and
the
spec,
so
the
spec
being
a
document,
a
standard
way
of
capturing
and
characterizing
all
those
concerns
and
considerations?
A
How
many
of
those
same
load
generators
generators
were
playing
back
the
messages
and
then
to
use
the
load
generator
to
to
do
it?
The
one
that
we're
spending
a
lot
of
time
on
is
is
the
load
generator
nighthawk
and
then
to
to
be
able
to
schedule
it
to
capture
the
output
of
the
tests
to
be
able
to
compare
the
tests
to
share
those
tests.
Centralize
them
do
studies
over
them
to
make
it
easy
to
repeat.
A
You
know,
have
repeatable
tooling,
to
measure
these
things
is
why
we
talk
about
measuring
so
yeah,
so
like
what
a
beautiful
you
just
sort
of
ran
the
gamut
of
the
projects,
the
whole.
H
Stack:
okay,
yeah,
that's
kind
of
that's
where
I'll
display,
my
ignorance
in
this
case
between
these
products,
knowing
they're
they're
set.
So
thank
you
for
explaining
that
that
helps
a
lot
so
with
smp
we're
talking
the
spec
itself,
like
you,
said
definition
of
what's
going
on,
or
what
to
do,
and
then
the
other
products
coming
into
play.
A
A
A
The
next
topic
was
going
to
be
donating
smp
to
the
cncf,
the
subsequent
donation,
that
I
think
that
we,
I
know
that
has
been
planned
for
a
long
time-
is
the
donation
of
measuring
as
well
to
the
cncf.
G
G
A
I
got
I
ended
up
getting
it
if
that.
G
Okay,
that's
great
yeah.
I
had
to
send
it
via
the
yeah
the
website.
I
wasn't
able
to
send
it
by
my
email
client.
So
I'm
not
sure
what
what
the
issue
is.
But
anyway,
if
that's
working,
find
a
way.
A
G
A
I
think
I
think
it
was
on
my
side
as
well.
Okay,
well,
fair
enough!
Well,
there's
you
know
by
by
number,
I
think
the
majority
of
you
that
are
on
the
call
have
had
some
amount
of
helping
hand
in
smp
or
or
maybe
just
learning
about
it
and
have
a
potential
vested
interest.
A
I
think
ken
I
think
has
already
is
already
helping
us
get
get
there
and
get
this
thing
submitted.
It's
a
you
know:
here's
an
open
opportunity
for
people
to
make
their
mark
on
an
upcoming
project.
This
one
has
been
socialized
with
the
cncf
toc,
I
think
more
times
than
they
probably
care
to
have
heard
about
it
and
it.
You
know
it
certainly
strikes
a
chord
because
it's
in
it
you
know.
A
From
my
perspective,
the
cncf
is
a
great
home
for
an
initiative
like
this
because
it
needs
to
be
in
a
vendor-neutral
place.
You're
just
trying
trying
to
the
goal
here
is
standardization
and
information
just
to.
A
A
And
well
our
this,
this
particular
dock
is
it's
actually
literally
just
a
copy
of
the
google
form
that
is
now
used
to
submit
cncf
projects.
Okay
and-
and
so
I
just
copied
and
pasted
the
questions
off
the
form
into
a
google
doc
figuring
what
no
better
way
than
to
invite
others
to
participate.
Then.
G
Yeah
absolutely
yeah.
I
will
go
through
that
and
respond
to
it
by
tomorrow.
Now
the
deadline
I
mentioned
is
23rd.
I
guess
right
so
review
by
before
that.
A
And
then
cinco
one
of
the
things
you
just
said,
though,
about
part
of
the
charter
and
the
vision
for
snp,
while
we
sort
of
reviewed
some
some
of
that
right
now,
I
don't
know
that
it
is
extraordinarily
explicitly
stated
prominently
like
it
is.
It
is
explicitly
stated
in
a
few
different
places
in
some
google
docs
over
here
and
maybe
in
a
blog
post
or
something
over
here,
but
in
my
mind
that
could
be
made
a
little
more
prominent
or
like
hey.
Is
it
an
ongoing
goal?
A
I
would
expect
like
once
s
p
were
to
land
into
the
cncf.
Is
it
an
ongoing
goal
to
to
publish
some
benchmarks,
or
is
it
a
goal
to
devise
what
what
the
group
considers
to
be
worthy
benchmarks?
Things
worthy
of
looking
at
is:
are
there
a
couple
of
iot
centric
ones?
There
are
a
couple
of
5g
centric
ones,
or
there
are
a
couple
of
and
to
your
point
about
vision
and
scope.
I
think
agree.
There's
room
for
clarification.
G
Absolutely
I
think,
as
part
of
the
proposal
or
the
questionnaire,
we
can
mention
that
also
we
can
also
not
possibly
write
a
blog
alongside
lines.
I
know
smp.
We
were
talking
about
publishing
a
blog.
Whatever
we
are
listing
here,
we
could
possibly
summarize
it
in
a
blog
and
go
through
that
and
publish
it
in
the
smp
site.
G
Yeah
internally,
we
had
to
go
through
quite
a
few
legal
reviews
or
some
reviews
before
you
know
I
can
put
in
my
name
but
yeah,
happy
to
kind
of
collaborate
here
and
to
find
the
procedure
within
my
company.
How
how
I
can
participate
here,
but
I
think
it's
a
no-brainer
to
kind
of
put
it
out
there
in
smb
sites.
So
it's
very
clear
for
anyone
who's
looking
at
it.
A
G
A
Hopefully,
hopefully,
all
the
rest
of
you
feel
like
that's
an
invitation
for
you
as
well
plea
plea
if
you're
interested,
please
do
same
thing
goes
for
get
nighthawk
and
for
mesherie
there's,
there's
there's
actually
a
lot
to
nighthawk
and
there's
there's
even
much
much
more
to
measury
much
that
hasn't
been
espoused
and
written
down.
So,
oh
one,
one
last
item
whoops
it
can.
I
don't
know
if,
if
you
mind
me
raising
up
this
topic,
the
josh
was
giving
some
fantastic
perspective.
A
You
know
from
is
ken
still
on.
Oh,
I
think
he
had
okay.
I
think
he
had
a
drop
good
anyway.
One
of
the
things
that
ken
was
privately
suggesting
is
that
if
you,
whether
or
not
you
know
this
or
not,
there's
a
there's
another
service
mesh
working
group
in
the
cncf
and
it's
the
it's-
the
end
users
service
mesh
working
group,
soliciting
input.
A
For
my
part,
I
think
that
would
be
a
wonderful
thing
to
do
to
ask
them
if
there
are
specific
types
of
benchmarks
or
like
to
josh's
point,
do
they
see
a
particular
pattern
that
really
causes
their
applications?
Issues
do
the
is,
it
is,
is
much
of
their
concern,
relegated
to
the
mechanics
of
a
canary
rollout,
something
that's
a
bit
more.
G
So
good,
no
along
this,
this
topic,
I'm
surprised,
there's
another
working
group.
Actually
I
I
didn't
know
till
now
so
is:
is
there
enough
synergy
that
can
be
combined
into
two
or
is
it
totally
different
charters.
A
No
there's
enough
synergy
I
would
have
been
it
is
against
my
will
that
there
are
two
or,
for
my
part
and
the
reason,
but
actually
there's
a
lot
of
logic
as
to
why
there
are
two.
It's
I
don't.
I
don't
actively.
For
my
part,
I
don't
actively
complain
the
the
concept.
The
notion
is
so
that
end
users
can
talk
amongst
themselves
freely
without
harassment
or
suggestive
comments
from
vendors.
A
It's
kind
of
the
is
the
line
of
delineation.
G
A
Yeah
yeah,
that's
right,
there's
a
whole
you're
right,
there's
a
collection
of
different
yeah.
I
think
they
are
called.
They
are
called
working
groups,
but
there's
there's
an
end
user
area
and
within
there
there's
a
few
different
working
groups
on
different
topics
that
get
together
and
and
yeah.
You
cannot
work
for
or
that's
part
of
the
divide
is
like
consumers
versus
producers
or
okay.
A
But
that
said,
what
an
opportunity
lost
not
to
have
some
of
that
exchange
to
be
able
to
solicit
the
complaints
and
provide
those
to
those
that
are
producing
the
tech
or
vice
versa.
A
A
On
that
topic,
as
we
go
to
wrap
up
here
anita,
it
strikes
me
that
you
might
be
in
you
might
end
up
interfacing
with
clients
or
or
you,
consumers
users,
as
you
guys,
go
to
assist
them
in
their
use
of
some
of
this
tech
I'll
plant,
a
seed
in
your
mind,
think
about
it.
You
know
if
this
is
an
open
forum,
for
you
know
some
of
that
feedback
about
where
people
are
really.
D
Sorry
about
that
double
mute
it
I
was
going
to
say
yeah,
I'm
not
as
much
on
the
technical
side,
but
some
of
our
consultants.
Obviously
they
are
and
they
would
be
interested.
So
I'm
taking
notes,
that's
part
of
why
I'm
in
the
meeting
so
I'll
make
sure
to
put
a
line
in
to
talk
about
like
call
outs
for
if
there's
any
kind
of
specific
issues
or
queries
that
might
come
up
that
might
be
relevant
for
you
guys.
A
All
right,
thank
you
all
same
time.
In
a
couple
of
weeks,
daniel
daniel
you're
you're,
whether
you're
asking
for
it
or
not
we're
penciling
you
in.