►
Description
Service Mesh Performance Community Meeting - July 22nd, 2021
Join the community at https://layer5.io/community
Find Layer5 on:
GitHub: https://github.com/layer5io
Twitter: https://twitter.com/layer5
LinkedIn: https://www.linkedin.com/company/layer5
Docker Hub: https://hub.docker.com/u/layer5/
A
We
are
fire
after
so,
let's
get
started
good
evening
good
morning
good
afternoon.
Everyone
welcome
to
the
asmp
community
meeting
today
is
22nd
july
and
we
are
just
about
to
get
started.
A
A
The
link
to
the
minutes
is
in
the
chat
and
yeah
also
jot
down
your
name
using
the
attendees
list
as
well.
Okay,
so
this
is
the
first
smp
community
meeting
after
s,
p
has
been
adopted
into
the
cncf,
so
the
onboarding
process
is
going
on
and
we
have
a
lot
to
lot
to
come
with
that.
Let's
move
on
to
the
first
topic
for
today,
lee
will
be
they
will
be
giving
a
project
charter
review
lee.
Would
you
like
to
get
started.
B
Yeah
sure
good,
it's
I
tell
you
what
it's
it's
really
nice
to
see,
all
of
you,
there's
11
or
so
of
us
on
the
call,
which
means
well
for
the
will
yeah.
I
guess
it
is
the
majority
of
us.
Are
you
know
new
to
this
project
and
you're
in
luck.
This
is
like
the
best
call
to
be
on
for
this
project.
B
B
I
think
that
in
the
from
my
perspective,
in
the
best
of
ways
that
is-
and
it's
not
an
undefined
question-
it's
defined,
but
it's
but
with
the
other
maintainers
willing
it's
something
of
an
an
open
question
or
an
intentionally
open
question
that
would
probably
evolve
over
time,
we'll
see,
but
for
the
majority
of
us
there's
many
of
you
who
are
coming
from
the
layer,
five
communities
from
the
projects
within
that
community
that
tickles,
my
that's
phenomenal.
B
Some
of
you
have
some
of
you
from
that
community
are
maintainers
on
smp.
Some
of
you
have
heard
of
s
p
and
haven't,
and
so
we'll
try
to
go
through.
So
we
should
just
really
quickly
talk
about
what
snmp
is.
I'm
sorry
smp.
Oh
my
gosh
we're
like
10
minutes
into
the
call,
and
I
said
snmp:
oh,
that's
horrible,
we're
doomed!
B
I
don't
know
if
it's
recoverable
at
this
point:
okay!
Well,
you
know
what
the
I
don't
know
if
it's
awkward
that
otto
and
I
are
both
laughing
or
that
just
shows
our
age
or
what
but
okay
good.
So
so
for
a
quick
brief
recap:
let's
share
some
slides
talk
about
what
smp
is
talk
about.
What
we
think
the
road
map
is.
B
B
That's
a
large,
that's
the
thick
of
what
we
wanted
to
cover
today.
There
are
a
number
of
administrative
items
or
a
number
of
items
around
the
project,
the
the
project
and
how
it
engages
in
the
cncf,
and
so
with
this
is
the
first
call
kind
of
post,
cncf
novendu
podocat
in
particular,
has
spent
time
trying
to
help
just
organize
around
the
project.
So
fresh
meeting
minutes,
hopefully
a
fresh
place
in
the
cncf's
bevy
instance
on
its
community
community.cncfd.
B
Hopefully
we're
establishing
reoccurring,
consistently
reoccurring.
Meetings
will
be.
These
videos
are
recorded
and
published
on
youtube.
Anybody
is
welcome
to
come
to
the
call.
Anyone
is
welcome
to
speak
up
and
say
things
and
suggest
things.
This
is
this
project
adheres
to
the
cncf
code
of
conduct,
and
so
it
for
the
vast
majority
of
you
on
the
call
it'll
run
really.
Similarly
to
the
rest
of
the
calls
that
we
have,
and
so.
B
I
will
you'll
probably
hear
me
say
this
like
more
or
less
a
half
joke,
probably
too
many
times,
and
that
is
that
the
project
smp
service
mesh
performance
is,
if
you
were
to
say
it
as
concisely
as
possible.
You'd
say
that
it's
a
specification
and
specifications
have
all
of
the
propensity
to
be
boring
like
especially
if
they're
really
meaningful
specifications-
oh,
they
just
go
on
and
on
and
people
bickering
over
whether
or
not
something
is
an
enum
or
a
string
or
whatever,
like
just
agnosium
for
better
or
worse.
B
B
So
interrupt
as
we
go,
this
isn't
about
me
speaking
to
everyone.
I'm
only
going
through
this
because
there's
a
large
portion
of
us
that
are
unfamiliar
with
the
project
and
so
implicitly
many
of
us.
Many
of
you
are
familiar
with
the
project
from
the
perspective
of
having
used
or
contributed
to
measuring
so
mesherie
implements
is
the
sort
of
the
canonical
implementation
of
this
specification.
B
We're
hopeful
that
other
projects,
other
products
will
implement.
The
spec
right
now
is
captured
in
the
form
of
a
set
of
protobufs
a
set
of
protos
and
it's
incomplete
it.
It
needs
a
bit
of
refinement
in
terms
of
that
making
sure
that
it's
abstract
and
not
specific
to
any
one
mesh.
That's
that's!
B
B
That's
probably
the
next
place
to
go
is
what's
lined
up
for
roadmap,
there's
about
four
areas
that
have
been
kind
of
outlined
for
roadmap
and
and
really
again
like
this
is
for
auto
and
utkarsh,
and
all
of
the
rest
of
you
to
weigh
in
on
what
that
roadmap
is
and
potentially
change
it.
I'm
gonna
share
for
the
moment,
not
because
navendi
was
sharing
the
wrong
thing.
That's
that's
very
much
the
right
thing!
B
B
B
B
You
know
this
isn't
dissimilar
from
what
you
find
in
other
cncf
projects,
if
you're
familiar
with
open
application,
model,
ohm
and
cube
bella,
om
spec
cube
is
the
implementation
spiffy,
inspire
notary
and
tough
or
tough
and
notary.
B
B
I'm
sorry
but
point
is
it's
not
uncommon
for
there
to
be
specs
and
implementations
sort
of
hand
in
hand
now
some
of
you
may
or
may
not
be
familiar
with
the
notion
that
an
article
was
published
on
the
project
yesterday,
and
so
we
should
include
a
link
and
congrats
you're,
already
internet
famous
before
and
that's
the
first
of
what
I
hope
is
really
an
ongoing
set
of
publications,
publications
written
by
others
sure
but
publications
written
by
each
of
you
or
all
or
all
of
us.
B
B
B
I'm
not
going
to
go
where
I
wouldn't
need
to
sit
here
and
sit
there
and
explain
it,
but
is
to
say
that
part
of
the
that's
a
good
question
to
reflect
on,
as
we
think
about
the
charter,
like
the
one
of
one
of
the
other
maintainers
on
two
of
the
s,
p
maintainers,
and
by
the
way,
just
there's
a
list
of
of
maintainers.
Now,
and
I
I
think
I
speak
for
the
rest
of
them
when
I
say
that
at
some
point
it
would
be
lovely
to
have
at
some
point
meaning
as
soon
as
possible.
B
It
would
be
lovely
to
have
other
names
down
as
maintainers
people
who've
been
around
and
are
contributing.
Some
of
you
have
already
contributed
to
s
p.
So
that's
and
for
my
part,
I'll
say
I'd
love
to
not
be
speaking
on.
This
call
I'd
love
for
otto
and
specific
to
be
doing
lots
of
interesting
things
here
and
and,
and
speaking
so
much
and
talking
about
all
that
he
has
to
be
asked
to
be
quiet
because,
because
we're
into
the
thick
of
it
one
of
the
the
by
the
way.
B
Just
since
we're
all
new
here
a
bit
and
we're
kind
of
getting
familiar
this
I
should
take
a
moment
to
say
I've
been
saying
auto
and
I've
been
saying
utkarsh
as
two
of
the
maintainers
otto.
Do
you
mind
saying
hi
and
letting
some
of
these
folks
get
to
know
you.
C
Sure,
hello,
everybody,
my
name
is
otto
van
der
schaff,
it's
quite
difficult
to
pronounce
so
well.
I've
been
working
a
long
time
for
a
long
time
on
nighthawk,
which
is
a
load
generator
and
well
load
characterization
tool
that
is
able
to
measure
latency
for
specific
workloads
and
yeah.
That's
something
that's
been
going
on
for
about
one
and
a
half
year
now
and
well.
C
What
I'd
love
for
this
this
project
to
be
is
to
be
able
to
if
we
can
get
this
spec
defined
in
a
way
that
well
nighthawk
could
be
one
of
the
low
generators
under
the
hood.
C
B
Yeah
yeah
very
good
and,
to
the
extent
that
it
matters
otto
is
well
recently
shifted
roles
into
a
role
in
with
with
red
hat
and
otto.
I'm
gonna
bastardize
this
but,
like
you
know,
more
or
less
doing
more
of
the
same,
maintaining
nighthawk
being
involved
around
envoy
being
involved
in
performance-centric
things
being
a
maintainer
of
smp
and
any
other
things.
C
That's
that's
right!
That's
right!
So
I
shifted
roles.
I've.
I
was
engaged.
I'm
now
engaged
at
red
hat
and
well
over
there,
I'm
working
as
a
principal
enfoy
engineer,
with
a
focus
on
performance.
So
I'm
you
know
doing
a
bit
of
both
and
well
yeah
before
that.
I
was
doing
it
from
my
own
company
and
that
was.
B
B
Well,
I
was
just
otto
was
just
trying
to
pull
up
your
company.
Okay,
good,
just
gonna,
take
quick
inventory
of
other
folks
on
the
call
so
good,
so
the
the
majority
of
the
everyone
else
has
had
a
chance
to
talk
to
everyone
else
on
this
call
shuvam.
I
know
you're
somewhat
fresh
on
some
of
the
projects,
but
but
fear
not
we'll
get
you
engaged
so
to
to
otto's
point.
Part
of
the
goal
of
today's
call
is
to
well
get
you
have
the
rubber
meet
the
road.
B
There's
there's
a
few
things
that
that
I
wanted
to
try
to
outline
some
objectives
so
that
we
can
stop
listening
to
lee
on
the
calls
and
start
listening
to
other
people
who
are
talking
about
the
work
that's
going
on.
B
I
think
that
the
work
so
by
the
way,
nick
jackson,
another
main
t,
another
s,
p
maintainer-
he's
at
hashicorp,
currently
he's
co-authoring
a
book
with
with
me
at
the
moment
on
service
mesh
patterns
which
we'll
talk
about
it
already
does
I
think
it's
an
early
release
now
already
talks
a
little
bit
about
a
little
bit
about
smp,
but
I'm
hopeful
that
we
will
have
published
a
few
things
before
the
book
is
done,
so
we
can
write
more
and
it
that
might
the
book
itself
might
be
a
little
bit
of
something
that
we've
discussed
one
just
for
everyone
to
take
pride
in
in
the
work
being
you
know
published
in
that
venue,
but
two
there
are
a
couple
of
you
on
the
call
that
I
wouldn't.
B
I
would
love
to
have
your
names
in
the
book,
because
some
of
the
because
you've
reviewed
or
written
portions
of
it
bits
and
pieces.
Some
of
you
are
the
best
representatives
of
what
s
p.
What
what
performance
in
this
regard
is
and
things
to
to
think
about
over
it,
so
okay,
where
to
start
on
the
topic
of
publication.
B
So
what
what
I
did
was
we
just
took
these
four
areas
and
split
them
into
four
individual
slides
so
that,
hopefully
we
could
spend
some
time
typing
some
of
what
we
think
the
objectives
of
these
areas
are
or
changing
what
this
says.
B
If
this,
if
we
don't
think
that
this
is
right,
so
so
so
guess
what
you
know
like
this
meeting
is
recorded
and
some
of
the
things
that
I'm
going
to
say
candidly
are
going
to
be
out
there
and
that's
fine
one
of
the
things
that
I
I
didn't
actually
get
a
moment
to
read
this
article.
But
one
thing
I
wanted
to
say
is:
I
think,
highlighted
in
he
here
the
first
paragraph
that
I
read
and
first
I'll
say
like
that
green
is
atrocious.
B
I
have
no
idea
where
he
got
this.
It's
just
the
green
in
the
background
is
horrific.
It's
a
violation
of
the
of
the
logo.
The
logos
by
the
way
are
now
property
well
or
are
now
in
the
cncf.
So
the
artwork
for
the
logos
and
and
we'll
have
more
rigmarole
around
what
we
can
do
in
the
cncf.
There
are
opportunities
open
to
each
of
you
about
giving
what
as
a
project
we
like
next
meeting
the
venue.
B
If
we
can,
it
would
be
wonderful
to
have
a
walk-through
of
what
resources
are
open
to
the
project.
Now
that
it's
part
of
the
cncf
and
and
what's
our
plan
to
empower
those
that
are
here
with
the
ability
to
go
use
those
resources
go
host,
the
webinar
go
talk,
it
kubecon,
smp
service.
Mesh
performance
is
now
listed
on
that
cfp
explicitly
like,
and
so
there's
there's
a
lot
that
goes
on
there.
The
only
reason
I'm
bringing
this
article
back
up
is
to
say
that
there
was
a
well
one,
hey
one
of
our
slides.
B
B
I
wanted
to
bring
this
up
because
there
was
a
statement
here
about
the
fact
that
bringing
potentially
those
two
projects,
or
these
two
or
three
projects
together
smp
smi
and
mescheri,
is
it
speaking
on
behalf
of
the
cncf.
It
is
an
alluring
thought.
B
B
B
This
is
twice
as
many
or
three
times
as
many
that
show
up
to
the
smi
call
like
smi
needs.
Some
help,
we're
trying
to
break
breathe
some
life
into
it.
It
might
get
some
life
breathed
into
it
through
smp.
B
In
terms
of
some
of
the
objectives
like
in
terms
of
like
publication,
there
is
the
ieee
has
a
magazine
called,
or
I
think
it's
a
magazine,
a
publication
called
the
bridge
and
they've
been
asking
for
an
article,
and
I
you
know,
I
think
that
so
I
think
that
it's
a
cross
between.
B
An
academic
publication
slash
just
a
long
article,
and
so
I
I'm
talking
about
this
right
now,
because
it
is
one
of
the
objectives.
I
never
wanted
to
write
it
myself.
When
we
signed
up.
I
was
always
thinking
of
autos
specifically
and
a
couple
of
people
from
and
a
couple
and
a
couple
of
others.
B
So
hopefully
we'll
end
up
auto.
If
we
do
this,
it
actually
needs
to
it's.
I
didn't
realize
the
deadline
for
the
draft
was
like
two
days
ago:
it's
not
a
deadline.
It's
like
they're,
anxious,
antsy
and
want
to
see
what
a
draft
looks
like
and
from
what
I've.
Seen
of
these
other
main
articles,
they
are
thorough
like
here
we
go.
Here's
an
example,
so
there's
an
abstract
in
the
introduction.
B
Yes,
it's
throw
yeah,
I
guess
it's
more
or
less.
An
academic
paper
is
what
this
boils
down
to
references
and
like
this
kind
of
a
thing
I
think
is,
is
something
that
we
would
ideally
be
doing
six
months
from
now
after
we've
done
some
further
study
and
and
have
more
research
under
s
p's
name,
but
we're
also
going
to
find
that
s
p,
I
think,
has
is
a
leaky,
a
leaky
project.
If
you
will
the
the
concept
of
it
is
leaky.
B
What
I
mean
to
say
is
that
we're
going
to
we
already
talked
about
nighthawk
for
just
a
teensy
bit.
I
think
that
we
want
to
talk
about
much
more
about
nighthawk,
specifically
and
and
some
of
its
advanced
capabilities,
some
of
its
tried
and
true
sort
of
oldest
capabilities.
If
you
will
just
in
terms
of
performance
characterization
but
but
there's
some
advanced
ones
as
well
as
well,
so
it's
elite
s,
p
is
leaky
in
that,
hopefully
we'll
we'll
end
up
bringing
together
a
few
projects
like
intentionally
intentionally.
B
The
cncf
is
wondering:
should
these
projects
come
together
because
they
are
interrelated
and
that's
very
much
so
the
point
so,
okay,
it's
like
I
don't
know
how
many
times
I'm
going
to
digress
in
this
meeting.
I
think
that's
about
the
ninth
but
auto.
This
is
something
that
to
follow
up
with
you
very
quickly
on
to
see
if
more
or
less
there's
just
existing,
more
or
less.
If
a
lot
of
the
things
from
your
last
service,
mesh
contact
or
other
materials
are.
C
Yeah,
so
yeah
nighthawk
is
able
to
select
support
some
research
with
this,
and
I
think
ritika
also
was
interested
in
collaborating
on
this
right
so
well.
I
try
to
pick
this
up
shortly
and
see
what
she's
up
she's
been
up
to
as
well,
because
well,
as
you
may
or
may
not
know,
but
I've
you
know
had
some
personal
issues
or
personal
family
matters
going
on
in
the
past
couple
of
weeks.
C
I've
dropped
the
ball
for
a
bit
every
on
this,
but
I'll
take
a
look,
asap,
okay,
yeah.
B
Great
there's
another
gentleman
to
introduce
and
I'd
we'll
see
if
we
can't
get
him
on
to
the
call
his
name's
pablo
who's.
A
former
employee
of
mine
he's
I
forget
his.
B
I
forget
his
phd-
is
in
either
physics,
mathematics
or
data
science
or
something
but
he's
focused
on
data
science,
and
so
he
and
I
had
had
him
working
on
this
very
thing,
except
it
was
described
a
little
bit
differently.
It
was:
are
there
algorithms
that
can
be
designed
that
will
optimize
your
network
topology
more,
like
you
have
any
number
of
nodes
in
a
network.
They
are
speaking
to
one
another
and
interacting.
B
You
know:
they're,
interacting
to
various
degrees,
with
higher
frequencies,
lower
frequencies,
higher
volumes,
lower
volumes
after
analysis,
what's
the
optimal
network
topology
for
that
environment
and
taking
into
account
that
some
of
what
was
being
analyzed
were
physical
network
layouts,
physical
switches
and
routers,
and
things
vms
and
containers
as
well.
I
think
it
well.
It's
it's
easier
to
take
action
on
the
output
of
those
algorithms.
B
When
your
workloads
are
ephemeral,
because
then
you
can
actually
manipulate
push
them
around
and
write
some
affinity
around
them
or
bring
them
closer
or
whatever
you
need
to
do
the
physicals,
but
anyway
the
algorithm
would
take
that
into
account
whether
or
not
something
was
movable
or
not.
Anyway,
point
is
we
need
to
get
pablo
on
this.
This
call
I'm
quite
a
nice
guy,
a
smart
guy.
He
wanted
to
help
in
publishing
this
as
well,
and
so
we
need
to
get
an
introduction
going
there.
B
Yeah
yeah,
you
know
yeah
it
yeah.
I
don't
know
what
some
of
the
latin
characters
or
the
greek
characters
that
were
used
and
some
of
his
algorithms
are
called
here's
interesting
work
anyway.
Well,
another
thing
we
that,
ideally
we
would
be
publishing,
and
this
actually
for
the
majority
of
you
that
are
on
the
call,
is
still
kind
of
bored
out
of
your
mind,
wondering
what's
the
project
about
this.
Actually
is
an
area
here.
So
one
of
the
resources
that's
open
to
service
mesh
to
this
project
is
the
cncf
has
labs.
B
B
Some
physical
hardware
from
packet
from
a
from
a
hardware
from
a
second-tier
sort
of
cloud
service
provider,
if
you
will
called
packet
and
as
such,
with
tooling,
like
nighthawk
and
like
mesherie,
this
project
can
first
def.
I
was
gonna,
show
the
info
about
the
labs.
But
I
guess
it
doesn't
matter.
This
project
can
first
define
what's
a
valid
benchmark
like
what's
a
valid.
B
What's
an
interesting
workload,
what's
a
valid
thing
to
test
and
study
and
examine,
what's
there
might
be
just
results
of
random
tests
to
go
talk
about
look
at
learn
from
publish
and
do
so
on
an
ongoing
basis,
and
that
means
that,
well,
how
are
these
tests
being
run?
Well,
what
the
two
of
you
actually
rude,
rocks
who's
on
the
call
just
got
done,
demonstrating
the
ability
to
pipeline
mesherie
in
a
github
action
and
have
it
perform
conformance
tests,
something
that
we
should
put.
B
I
believe
that
we
should
put
onto
the
roadmap
here
is
in
context
of
publishing
ongoing
benchmarks
or
establishing
benchmarks
and
publishing
ongoing
results
is
the
ability
to
automate
that
type
of
publishing
or,
to
the
extent
that
we
can
and
so
there's
an
open
issue.
I
believe
in
the
s
p,
repo
that
calls
for
a
github
action
that
would
invoke
mesh
meshery
and
nighthawk
their
perf
to
run
a
performance
analysis.
B
B
Can
you
do
me,
could
you
do
me
a
favor,
find
that
issue
and
make
note
of
it?
Well,
I
suppose
in
the
publications
section
here,
and
so
you
know.
B
Automation,
one
thing
and
so
otto
I
don't
know
if
this
falls
under
research
or
publication,
but
the
notion
that
we
would
define
interesting
benchmarks.
C
So
that's
but
yeah.
B
So
so
there's
something
about
smp
that
may
that,
even
though
it's
called
service
mesh
performance,
I
think
that
there's
a
lot
of
things
to
examine
outside
the
context
of
the
mesh
for
the
foreseeable
future.
Much
of
what
we're
focused
on
here
would
be
service,
specific
or
or
rather
in
in
context
of
a
mesh
part
of
that
is
actually
showing
what
the
difference
is
in
terms
of
these
performance
measurements
when
things
workloads
are
on
a
mesh
or
not
on
a
mesh.
That's
an
interesting
thing
to
characterize.
B
So
to
to
again
to
to
make
some
of
what
we're
doing
tangible.
The
thing
that
I
just
said
just
there
is
like
is
something
for
any
one
of
us
to
pick
up
grab
a
sample
workload
run
through
like
for
anyone
any
one
of
us
to
go,
define
like
what
is
an
interesting
test
and
if
it's
wrong
and
we
we
go
run
a
bunch
of
tests
and
and
even
publish
those
things
by
wrong.
B
Like
I
mean
if
it's
uninteresting,
it's
still
good
work
like,
I
think
it's
hard
to
it's
hard
to
do
an
uninteresting
test.
There
are
certain
types
of
tests
that
will
illuminate
and
provide
more
insight
than
others,
but
nearly
all
of
the
above
is
valid,
so
to
speak.
B
B
B
There
are,
I
think,
a
few
aspects
of
this
that
are
either
too
technology
specific
or
they
are
not
abstract
enough.
B
A
service
mesh
is
maybe
or
an
egress
gateway's
performance
that
those
are
optional
as
part
of
the
spec.
Not
every
service
mesh
will
have
an
egress
gateway,
some
will
some
won't.
Some
service
meshes
do
not
have
their
own
ingress,
and
so
like
here's,
a
good
question
is,
should
does
that
mean
that
an
ingress
gateway
is
optional
and.
B
And
there's
a
valid
s,
p
compatible
report
results
report.
If
there's
no
details
captured
here
like
there's
many
of
these
types
of
questions
to
go
through
like
hey,
should
we
as
we
look
at
this
spec
and
we
look
to
part
of
the
goals
is
to
get
all
the
service
meshes
involved?
Have
them
self
reporting
their
performance?
B
Have
that
going
into
a
central
dashboard,
ideally
again
I'll
use
meshri
as
an
example,
because
some
of
you
again
just
presented
this
yesterday,
meshri
runs
conformance
tests
and
each
of
the
service
meshes
are
being
encouraged
to
self-report,
their
conformance
against
smi
in
the
similar
fashion.
Each
service
mesh
can
be
encouraged
to
self-report
its
performance,
ongoing
and
sort
of
watch
the
performance
trend
at
release
after
release,
and
it's
the
thing
on
something
like
this
is
I
don't
know
that
a
table
a
similar
table
for
smp
like
it
would
make
sense
to
me
that
we
might
characterize.
B
B
B
I
think
that
listing
envoy,
like
explicitly
on
an
smp
chart,
can
make
a
ton
of
sense,
because
envoy
is
like
an
essential
and
core
component
to
five
or
six
or
seven.
How
depends
on
how
you
want
to
count
seven
or
eight
different
service
mesh
data
planes,
and
so
as
such,
since
it's
you
know
essentially
ubiquitous
highlighting
it
explicitly
might
make
sense
having
tests
that
look
at
just
data
plane
things
or
just
control,
plane
things
or
or
holistic
things
like.
I
think
these
are
all
valid
types
of
performance
tests.
B
I
said
that
smp
so
okay,
so
we
should
can't
navendu
or
utkash
or
anyone
can
you
all
help
write
some
of
that
down.
It's
embarrassing,
I'm
embarrassed
at
how
many
times
I've
said
that
very
same
thing,
and
it's
still
not
written
down
anywhere
so
about
that's,
probably
like
about
defining
different
benchmarks
and
what
would
be
interesting.
There's
a
dock
in
the
smp
community
drive
right
now.
That's
called
the
layer.
5
community
drive
because
you
know,
because
smp
is
young,
but
there's
a
folder
in
here
for
service
mesh
performance,
project,
navindu
or
someone.
B
We
would
do
well
to
capture
this
in
the
meeting
minutes
like
a
permanent
link
to
that's
that's
actually,
where
the
meeting
minutes
reside.
I
think
we've
had
prior.
This
is
really
pretty
funny.
Actually,
we've
had
prior
meetings
that
were
basically
this
meeting
smp
and
otto's
been
on
them
before,
as
you
can
see,
and
we
we
ended
up
talking
about
like
nighthawk,
was
kind
of
the
focus,
and
that's
actually
why
I
bring
this
up
is
to
say
that
I
was
just
talking
with
well.
B
I
think
otto
I
think,
he's
an
envoy
maintainer
d,
whose
last
name
I
don't
remember,
he's
out
of
the
philippines,
I
think,
or
out
of
singapore.
B
You
know
I'll
find
him,
but
point
is
there's
a
number
of
people
who
are
focused
on
c
plus
and
have
been
described
part
of
what
we've
talked
about
in
the
past,
and
I
guess
I
consider
that
part
of
that
adaptive.
I
mean
part
of
what
we've
talked
about
in
the
past,
about
adaptive,
load,
control
and
then
having
custom.
Adapter
load
controllers
is,
in
my
mind,
in
in
scope
of
smp
or
like
is
something
to
be
discussed
in
in
this
context.
Like
it's
that's
extraordinarily
interesting
to
the
project.
C
Yeah,
I
I
think
so
as
well.
I
think
that
through
that
technology,
you
can
answer
some
of
the
high
value
questions
you
know
so
earlier
you
mentioned
like
which
benchmarks
are
interesting,
and
I
do
think
that
this
one
might
answer
questions
that
are
directly
related
to
costs
of
the
end
user
and
that's
always
interesting.
B
I
I
just
said
something
yeah.
It
was
the
something
we
should
capture
about.
C
Research
yeah:
I
think
that
will
do
it's
good
to
know
that
this
component
is
like
independent
of
neither
slow
generator,
so
you
could,
with
a
little
bit
of
effort,
also
plug
in
other
load
generators
that
do
not
support
this,
so
that
might
be
useful
to
list
it
separately.
Oh.
B
Is
it
that
you're
saying
nighthawk
has
a
plug-in
model
for,
if
you've
written
a
custom,
adaptive
load
controller
you
can
plug
in
that
custom
controller?
Are
you
saying
that
whole
framework
at
nighthawk
is
actually
detachable
and.
C
The
the
adaptive
load
control
system
is
like
sort
of
standalone,
so
you
should
kind
of
reverse
this,
because
the
nighthawk,
while
they
live
in
the
same
project
nighthawk,
is
integrated
as
a
plug-in
like
a
reference.
One.
B
Okay,
look,
look,
I
I
do
this.
I
I
have
for
my
part,
I
think
I
have
such
an
appetite
to
go
and
do
a
lot
of
these
things
that
we
kind
of
talk
about
a
lot
of
them
and
and
that's
good
and
that
helps
us
form
a
charter
and
but
then
like
getting
down
to
brass
tax.
If
you
will
on
okay
well,
what
do
we
want
to
do
by
the
next
time
that
we
meet
like?
First
of
all,
when
is
the
next
time
we
meet?
C
B
Who
will
who,
who
has,
for
my
part
I'll,
say
like
this?
This
is
something
that
I
will
attempt
to
drive
focus
around
and
attempt
to
assist
in
getting
over
the
hump.
I've
got
because
it's
basically
I've
got
escalation,
emails
for
that
draft
and
so
other
than
that.
B
C
So
for
me
I'll,
try
to
I
said
this
last
meeting,
but
this
time
I'm
really
gonna
do
it.
I'm
gonna
try
and
propose
some
changes
to
this
produce
and
I
think
that
might
that
might
deal
some
things
to
discuss
next
time.
So.
B
B
Beautiful
rude
rocks
your
your
audio
didn't
come
through,
but
I
know
that
you
said
you
would
love
to
demo
a
github
action
for
smp
performance.
B
B
So
something
that
I'll
do
as
well
is
and
by
the
way
I
like.
I
hope
I
hope
that
everybody
gets
this
sense
and
that
that
this
is
a
young
project
and
you're
most
encouraged
to
engage
for
some
of
you
who
are
looking
to
add
significant
line
items
on
your
resume.
I
would
say
that
if
you
can
stomach
the
boredom
of
working
on
a
spec,
what
an
impressively
awesome
thing
that
is
to
have
on
your
that
ability
it
takes
an
architect
level
type
ability.
B
It
takes
a
the
title
that
otto
had
mentioned
before
the
principal
principal
software
engineer
principal
architect.
I
can't
remember
the,
but
it
takes
that
level
of
of
thinking
and
understanding
to
really
be
able
to
impact
this
type
of
a
spec
and
so
for
a
lot
of
you.
What
an
amazing
opportunity
that
is.
B
So,
for
my
part,
I
will
send
out
I'll
try
to
fill
in
and
please
if
anyone
else
has
ideas
I'll
try
to
fill
in
the
rest
of
what
I
consider
that
these
objectives
are
so
expanding
on
the
specification
good,
that's
like
so
part
of
this
is
hey.
We
need
to
have
an
earnest
discussion.
B
B
I
will
also
try
to
beautify
this,
because
I
think
that
it's
really,
I
think
it's
really
important
to
get
a
blog
post
out
about
this,
and
it
would
be
nice
if
you
know
there
were
you
know
there
were
other
names
on
that
blog
post,
sharing
that
byline
I'll
also
be
bringing
some
people
who
are
focused
on
c
plus,
and
I
think
that
we
would
do
well
to
have
as
an
item
on
our
next
meeting,
potentially
an
exploration
of
for
auto
to
re-explain
again.
B
You
know
some
of
the
things
that
otto
would
like
to
see
done
with
that
and
okay,
something
else
I'll
call
out,
and
this
is
and
by
the
way
this
doesn't
I
really
desire
for
this
not
to
be
me,
the
there's
an
en
there's
a
in
this
desire
for
this
not
to
be
me,
that's
presenting
much
of
the
work
that's
being
done
here,
so
I
think
that
there's
actually
a
service
mesh
con
cfp
going
on
I'm
about
to
end
soon
that
I
would
encourage
any
of
you
to
go.
B
Take
a
look
at
go
submit.
There
is
this
next
week,
there's
a
cncf
service
mesh,
end
user
group
that
meets
these
are
in.
These
are
engineers
from
cncf
member
companies
that
identify
themselves
as
technology
consumers
as
users,
not
as
technology
producers
like
these
aren't.
These
are.
These
are
not
maintainers
of
the
project.
These
are
consumer.
You
know
prominent
consumers,
they
meet
twice
every
week
and
I'm
sorry
not
twice.
They
mean
every
two
weeks
and
last
time
they
met.
They.
B
B
They
voted
in
mesri
yeah
as
the
and
I
was
on
the
call.
So
I
was
the
default
there,
there'll
be
a
presentation
of
meshary,
which
means
a
presentation
of
smp
on
this
call
this
this.
This
call
I'll
put
the
link
to
this
it's
on
the
second
tuesday
or
they
meet
monthly,
so
okay,
second
call
second
tuesday
of
each
month.
At
this
time.
I'll
put
this
into
the
notes:
we'd
love
to
have
any
of
you
there,
even
if
you
want
to,
if
you're
interested
in
helping
describe
smp
or
mesherie
on
that
call.
B
Okay,
as
a
matter
of
again
like
helping
us,
be
productive
and
get
some
things
done.
There
are
a
few
mailing
lists
that
the
project
has
now
that
I
that
I
think
we're
gonna
find
the
asynchronous
communication
through
mailing.
This
is
probably
more
advantageous
for
how
this
particular
smp
runs
because
time
zones
are
yeah,
because
that's
just
how
I
think
that
this
project
is
probably
going
to
be
driven
forth.
It's
a
little
easier
for
everyone
to
do
that
independently.
B
B
B
B
Navindy?
Do
you
want
to
do
you
want
to
wrap
it
up
and
take
the
this
last
topic
tell
us
about
some
of
the
like
the
the
mailing
lists,
what
those
are
who's
on
them,
some
of
the
things
that
we
can
do
within
the
cncf.
A
Yeah,
so
we
have
a
couple
of
mailing
lists
for
smp.
Let
me
try
to
show
the
show
displaying
this
so.
A
A
So
this
these
are
basically
the
these
three
main
lists
are
basically
the
smp
mailing
list,
so
we
have
a
community
mailing
list
and
a
maintenance
main
list
and
a
user
group
mailing
list.
So
we
we
haven't
used
these
mailing
lists
much
in
the
past,
but
that's
niche
that
needs
to
be
changed
and
oh
yeah.
So
I
think
the
melonies
are
self-explanatory
and
about
the
cncf,
like
adding
a
community
platform
in
a
community
group
in
the
cncf
community.
Is
that
so
we
have?
We
have
some?
A
We
are
going
to
submit
a
proposal.
To
add
us,
add
the
smp
project
in
to
create
a
community
group
for
the
smp
project.
So
one
of
the
so
there
is
an
open
position
for
a
co-organizer
in
this
community
group.
So
if
that
interests,
you
then
please
reach
out
and.
A
There
are
so
being
being
part
of
us
part
of
cncf
will
will
give
some
benefits
to
smp
as
a
cncf
sandbox
project.
So
I
think
I
I
will
on
that
on
the
next
meeting
like
to
let
everyone
know
what
what
we
can
do
and
empower
each
of
us
to
go
and
do
that
yeah.
Anything
else
to
add
lee.
B
Yep
there,
one
of
the
things
that
we
can
do
ongoing
as
well
is
begin
to
speak.
I
have
agenda
items
on
this
call
in
context
of
github
issues.
B
There
are,
I
think,
there's
12
open
issues
at
the
moment,
but
there
are
more
to
be
opened
up,
but
particularly
as
we
define
objectives
in
the
roadmap,
so
one
of
those
there's
there's
a
project
site
that
is
in
need
of
having
the
mailing
list
published
like
there
are
there's
the
there's,
the
ones
that
navindy
was
just
talking
about,
but
I
don't
know
that
those
are
published
and
that
people
know
that
they
can
either
sign
up
or
or
message
on
those
you
by
the
way
I
mean
just
like
that.
B
B
B
We
should
probably
adjourn
follow
up
with
an
email
that
recaps
what
we
discussed,
what
things
that
we're
going
to
look
to
do
next
time.
There
are
a
number
of
you
on
the
call
who
can
immediately
make
an
impact
on
the
site.
B
I
will
follow
up
with
the
other
two
maintainers
so
soon
and
I'll
also
probably
send
out
an
email
that
just
introduces
everybody
to
the
project
and
says
these
are
the
maintainers
here's
a
list
of
maintainers
here's
the
mailing
list:
here's
the
intro
to
the
project,
here's
a
couple
of
blog
posts
that
I
think
that
we
would
like
to
write
and
to
to
tell
people
about
hey.
We've
got
a
twitter
handle
like
to
help
introduce
people
to
the
the
rest
of
the
project.
B
B
Who
was
on
the
last
item,
intel
was
asking
on
about
rudraksha's
action
item
yeah
on
yesterday's
smi
call
intel-
or
I
should
I
shouldn't
say
intel,
but
I
mean
an
architect
from
intel
was
asking
about
so
marlo
weston
was
asking
about
the
ability
to
run
performance
tests
as
a
pipelined
thing,
and
we've
made
a
promise
to
her
and
and
she's
representing
a
service
centric
group
at
intel
of
about
20
people.
B
We've
made
a
promise
to
her
to
get
short
video
explainer
videos
over
and
to
then
ultimately
demo
this
as
well.
So
we
should
take
a
note.
I
don't
know
if
marlowe
has
subscribed
it
to
this
to
these
mailing
lists,
the
s
p
mailing
list.
So
that's
another
action
item,
but
anyway
I
figured
I'd
mention
it
because
people
are
asking
on
about
the
work
already
by
the
way
for
what
it's
worth
a
lot
of
the
time.
B
B
How
should
I
run
my
mesh
most
effectively
efficiently
on
the
first
answer,
the
first
question:
it's
like
well,
no
well,
we
sort
of
in
some
areas
we
kind
of
have
published
some
things,
but
broadly
we've
held
back
from
making
a
big
stink
about
which
one
is
faster
or
slower,
because
we
didn't
want
to
piss
people
off
in
part
in
part
also
because
it
was
hard.
Mashery
wasn't
ready
to
help
automate
that
testing.
All
of
those
things
are
no
longer
measury
automates
this
quite
well
or
and
needs
them
always
willing
improvement.
B
But
for
my
part,
I
don't
care
about
pissing
anyone
it's
time
to
go
piss
off
the
projects
and
get
them
interested
in
what's
going
on
and
expose
where
they're
fast
expose
where
they're
slow,
I
don't
mean
to
intentionally
piss
them
off.
I
just
mean
to
get
them
interested,
that's
what
I
mean,
but
the
fear
of
upsetting
them
is
gone.
Three
like
where
the
project
is
in
the
cncf
now
and
kind
of
bears
that
backing
and
has
the
open
resources
to
it.
B
So
so
we
should
define
some
tests
and
go
run
them
and
actually,
I
think,
to
help
answer
the
workload
definition
test
like
what
is
an
interesting
benchmark.
Well,
those
folks
I
mean
otto
probably
has
you
know-
can
talk
about
that
for
a
few
hours,
but
some
of
the
folks
at
intel
that
are
interested
as
well,
that
is,
their
field
of
focus,
is
just
performance
engineering
and
like
so
they've
got
lots
to
draw
on
for
what
is
interesting
to
them.
B
B
Okay,
yeah,
I
think
I
think
that's
it.
I
think
we're
yeah
great
first
call
nice
to
see
everybody
if
you're
interested
in
doing
something-
and
you
don't
know
what
that
is,
then
please
say
so
in
slack.
So
by
the
way,
there
are
two
slack
channels
for
this
project.
One
is
called
service
mesh
performance
in
the
cncf
slack,
so
novendou.
I
think
it
created
that
one
a
little
bit
ago
and
there's
also
the
performance
channel
in
the
layer.
Five
slack,
so
you
know
either
one
will
work.