►
From YouTube: Meshery Community Meeting - July 26th 2019
Description
This week: A new @nservicemesh architecture proposed by @premsankar. A proposal for additional @smi_spec integration that @sheriffjackson will wrangle.
Join the community calls and learn more about Meshery at https://layer5.io/meshery
A
A
A
B
A
A
B
A
A
A
But
fill
my
role
as
a
as
a
as
a
mentor
to
the
extent
that
that's
even
possible
with
you
I
think
my
advice
is
to
stop
working
three
jobs
at
a
time:
okay,
I'm,
sorry,
I'm,
so
good,
we're
fine
after
we've
got
a
number
of
folks
that
are
joining
me.
This
is
fantastic,
I,
don't
know
if
I
called
it
out
last
week,
I
don't
think
I
did
while
that
was
happening,
but
we
had
record
attendance.
There
were
11
people
on
the
community
meeting
last
week.
So
that's
that's.
Encouraging
to
me.
A
A
You
can
you
if
you
want
to
run
that
product,
you
can
run
it
on
prem
or
optionally.
Alternatively,
you
can
octarine.
You
will
provide
a
a
managed
to
control
plan
for
you
and
in
this
case,
for
how
they've
written
the
adapter
for
mystery,
it
will
leverage
their
managed,
control,
plane
and
so
they've
set
aside
a
public
sandbox.
If
you
will
or
a
public
account
that
anyone
can
go
use
to
do
the
features
of
function.
I
do
the
performance
tests
do
the
provisioning
we
do.
A
The
other
functionality
is
a
mystery
in
that
sandbox
so
that
it
so
there
is
public
facing
credentials
that
the
world
can
use,
and
so
this
adapter
focuses
a
little
bit
differently
anyway.
I
think
people
were
interested
in
that,
and
so
that
was
last
week,
and
that
was
much
of
last
week.
That
was
a
great
discussion.
I
told
time
that
I
would
do
a
call
to
action
on
that
to
the
community.
Here,
I
was
spending
the
briefest
amount
of
time
a
couple
days
ago,
testing
the
octarine
adapter.
A
The
call
to
action
here
is
for
each
of
you
to
go.
Give
it
a
test
now
that
I
say
that
I
think
there's
one
piece
of
instructions
missing,
so
I
will
take
it
upon
myself
to
make
sure
that
all
of
the
instructions
are
there
for
that
adapter
and
just
a
message,
the
channels
so
that
anyone
who
wants
to
go
try
it
out
just
before
we
move
it
into
sort
of
officially
into
the
measuring
release.
Guys
do
everyone
can
test
it?
Okay,
hey
socko
reminder
we're
gonna.
A
Do
you
guys
in
trying
to
help
be
a
steward
of
the
circus
master
community,
we're
going
to
put
out
a
newsletter
that'll
talk
about
service,
mesh
interface,
SMI
it'll
talk
about
the
google
Summer
of
Code
program
that
we've
got
going
on
and
you
as
the
most
prominent
individual
within
that.
If
there's
a
quote
or
a
statement
that
that
you'd
like
to
include
in
that
newsletter,
don't
worry
it's
only
going
out
to
a
few
thousand
people
who
have
subscribed,
but.
A
Good,
the
other
thing
that
I
think
last
bit
of
general
news
that
I'm
excited
about
I'll
click,
the
link,
but
we've
got
sort
of
the
Jesus
I
guess
the
the
oligarch
Anand
who's
on
the
call
had
joined,
studying
at
uem
in
Jaipur
and
now
Subin
has
joined
with
us.
Also
from
the
same
school
and
motto
is
also
joined
from
the
same
school
in
motto.
You
just
landed
your
first
PR
about
half
an
hour
ago.
A
So
congratulations
on
that
as
part
of
that
I've
been
also
established
under
yeah
yeah,
looking
good
by
the
way
and
I
tossed
in
a
sneak
I.
Think
you
think
you
addressed
the
first
issue
and
then
I
tossed
in
a
I
was
sneaky
and
costed
another
one
so
but
I'm
just
bringing
this
up,
because
there's
a
flyer
that
there
is
an
outstanding
ask
for
a
community
member
who's,
front-end
centric.
A
All
of
you
know-
and
we
talked
about
this
I
think
a
week
ago
or
two
but
you're
all
familiar
with
this
site
layer.
Five
dial
slash
mystery
and
it's
you
know
it's
the
prettiest
form
of
a
page
that
represents
the
project,
there's
also
the
github
readme
x'.
But
we've
noted
before
that,
it's
time
for
us
to
bring
forth
a
dedicated
product
project,
open
source
project
site,
and
so
the
domain
mesh
read
dev
has
been
registered.
A
If
you
go
to
it
today,
it
just
that
whole
domain
just
redirects
to
this
page,
but
the
call
there
is
to
find
folks
who
are
who
might
be
willing
to
contribute
and
who
are
maybe
front-end
in
nature.
Actually
I'm
yogi
here
messaged
me
last
night
and
saying
that
he
thinks
you
might
have
found
someone
that
would
like
to
do
it
but
yeah
come
leave.
Your
mark
tell
your
friends
to
come.
Leave
their
mark,
so
I
was
working.
A
A
C
A
Well,
let
me
highlight
this
one
briefly,
because
it
will
be
quick
and
because
mr.
Jackson
is
on
the
call
with
us,
and
so
there's
a
there's,
a
link
here
to
a
very
use
matter
fact
you
might
already
be
looking
at
it
very
quickly
put
together
suggestion
about
further
integration
with
SMI
we
had
had
Tarun,
who
was
also
a
google
Summer
of
Code
intern
who's
working
most
closely
with
the
linker
D
project
and
he's
been
thinking
it
for
the
most
part,
the
guy
behind
SMI
metrics.
A
A
B
D
D
A
I
really
would
I
think
that
you
are
abreast
of
the
status
of
the
sto
to
SMI
integration.
We
ended
up
when
Tarun
was
on.
We
ended
up
doing
a
little
bit
of
brainstorming,
a
brow
about
his
thinking
in
terms
of
the
approach
to
collecting
those
metrics
and
and
I
want
to
say
we.
We
helped
him
along
a
little
bit,
but
but
yeah
I
got
to
imagine
he
hasn't
gotten
to
it.
Yet
so
yeah.
D
And
I'm
gonna
I'm
gonna
I
will
looking
at
the
those
regular
SMI
calls
look
before
pucon
on
it.
It's
gone
a
little
quiet,
but
I've
got
some
some
spec
additions
that
I'm
gonna
be
adding
and
I
just
need
to
know
some
for
next
week.
Along
about
that,
although
I'll
definitely
begin
to
to
that
so
I'll
pick
that
up
as
an
agenda
item
to
find
out
what
situation
is
and
I'll
report
back
next.
Next
next
time,
Thank
You
Nick.
A
It
me
yeah
you
making
this
a
happy
Friday.
This
is
great
yeah,
yes,
I
would
phrase
it
well.
It's
kind
of
an
interesting
thing.
I
would
say
that
there's
been
almost
complete
radio
silence
from
from
some
very
good,
some
folks
who
I
know
or
whom
I
like
very
much,
and
so
no
doubt
there
are
other
things
happening.
I'm
you
know
unaware
of,
but
it
but.
D
Think
I'm
guessing
I,
think
everybody's
just
being
super
busy
hope
sets.
Am
I
but
I
still
think
it's
a
really
good
snack
he's
done.
I
literally
put
aside
two
weeks
from
Monday
to
to
be
digging
into
this
and
to
be
just
pushing
along
both
the
console
integration,
as
well
as
the
generally
the
the
spec
home,
because
coupe
colonies
will
be
pasta
pro
Qing
and
it's
just
on
a
little
quiet.
So
I'm
gonna
hustle
a
bunch
of
folk
next
week,
he'll
not
very.
A
Good,
okay,
speaking
of
well,
this
is
well
I,
expect
in
some
future
form
related
I.
Think
I
had
mentioned
to
you
and
maybe
in
passing
on
this
call
that
I
drafted
a
new
sig
for
network
or
for
traffic.
It
was
really
just
reincarnate
that
okay,
so
I
didn't
mention
it
before
anyway.
This
just
reminds
me
Ken,
Owens
and
paying
me
yesterday
saying:
hey:
do
you
want
to
host
a
maintainer
z--
track
or
what
happened
so
yeah?
A
A
A
A
The
other
thing
that
was
and
had
been
in
the
works
for
quite
some
time,
I'm
pretty
pleased
about
it,
was
the
displacement
of
the
command-line
utility
measure
II
that
was
written
in
bash
Sacco
and
Anand
had
been
focused
here
as
well.
Went
through
him
and
wrote,
rewrote
that
in
to
go
into
measure
ECT
l.
It
is
pretty
great.
A
All
right,
let's
so-
and
this
might
be
you
know,
I
think
for
both
Bremen
and
Knicks
benefit
is
probably
worth
highlighting
this
again
that
so
we
did
get
some
time
on
the
CNC
F's
cluster
that's
been
committed
to
us.
We
have
not
performed
a
testing
yet
in
part
Prem,
because
we
wanted
to
wouldn't,
because
we've
got
a
little
more
test
planning
to
do
to.
A
We
wanted
to
try
to
lend
support
for
the
four
NSM
before
performing
that
test,
because
it
would
be
fantastic
to
have
NSM,
including
that
that
that
longer
and
and
at
scale
set
of
tests.
That
will
do
the
other
thing
that
we
did
kind
of
a
third
thing
that
we
were
hoping
to
change
up
a
little
bit
is
that
you
know
today
measuring
includes
a
few
of
the
sample
apps
that
the
different
service
measures
use,
and
there
you
know
to
varying
degrees,
representative
of
workloads
that
organizations
might
have
their
in
from
I.
Believe
it.
A
If
I
recollect
correctly,
it's
from
Cornell
they've
been
working
on
well
a
performance
benchmark
tool.
That
is,
you
know,
I,
don't
think
it's
necessarily
the
goal
of
that
project.
To
have
a
very
well
put
together
some
tool
in
this
regard,
performance
benchmark
tool,
but
rather
what
they
focus
a
lot
of
their
time
on
was
a
sample
app
or
a
set
of
sample
apps.
A
B
D
Like
okay,
something
else
to
just
to
think
about-
and
this
is
something
I've
been
thinking
on
the
other
day-
is
I'm
doing
some
experimentation
at
the
moment
on
a
very
simple
application
which
allows
you
to
change
together,
micros
like
fake
micro-services,
so
you
can
kind
of
chain
together,
multiple,
multiple
fake
micro-services,
but
each
of
the
fake
microservices.
You
can
configure
it
to
respond
in
a
particular
particular
way.
So
you
can
say
right.
D
This
service
is
going
to
take
a
median
time
of
one
millisecond
and
95,
percentile
of
or
or
or
whatever,
and
so
theoretically,
what
that
would
allow
you
to
do.
Is
it's
pretty
much
regardless
you'll
get
really
consistent,
behavior
out
of
the
the
testing
of
the
the
system,
and
it
will
be
a
RPC
as
well
as
HTTP,
so
that
had
like
it's
an
experiment,
I'm
just
kind
of
working
on
at
the
moment,
but
yeah.
My
intention
is
to
try
and
create
like.
D
A
Let
me
and
actually
I
was
kind
of
I,
don't
know
Prem
if
you're
biting
your
tongue,
when
you
heard
him
say
can't
I'm
sorry
chain,
as
the
key
word
I
figured
you'd
be
supposed
to
but
yeah,
but
but
you
know
how
we
are
all
we
characterized
it
nicked
it
like
hate
when
you're
doing
performance
testing
a
part
of
the
real
pain
in
the
butt
about
it.
It
is
all
the
variables
that
aren't
necessarily
within
your
control
and
in
part
way
what
it
sounds
like
you're
doing
is
for
the
applications
that
are
being
load
tested.
A
Let's
put
them
in
a
vacuum,
make
them
behave,
I
have
to
have
control
over
their
behavior
and
make
them
behave
consistently.
Such
that
you're,
able
to
you,
know
any
variances
that
you're
seeing
are
much
more
of
the
microscope
is
pointed
on
the
surface
mesh
itself
than
it
is,
then
it
so
happens
to
be
whatever
the
app
is
doing.
Whatever
the
application
code
is
doing,
yeah.
D
Yeah,
absolutely
absolutely
so
so
things
like
where
you
get
latency
through
through
routing
and
stuff.
Like
that,
my
my
intention
is
that
you
would
be
able
to
have
like
you
know
you
could
you
could
set
up
tests
which
would
have
like
round-robin
load,
balancers
or
maybe
you're
gonna
use
like
routing
and
things
like
that,
and
you
can
kind
of
test
those
components
on
the
service
match,
because
the
application
which
is
being
routed
to
is
is
well
given.
D
That
will
obviously
be
a
snapping
point,
but
but
it'll
it's
gonna
perform
pretty
well
because
it's
not
doing
any
work,
so
you
can
control
the
latency
and
yeah
hoping
hoping
that
will
be
an
interesting
way
of
kind
of.
You
know
testing
some
of
those
some
of
those
scenarios
without
without
worrying
that,
if
you're
running
a
real
application,
you
get
halfway
through
the
test
and
you've
added
10,000
database
records,
the
database
starts
slowing
down
and
then
the
performance
of
the
application
degrades,
as
opposed
to
be
the
network
GamePro.
But
it's
purely
an
idea.
A
That
last
part
resonates
with
me,
oh
and
make
sure
I
get
that
down.
They
did
like.
You
certainly
have
seen
people
perform
at
length
tests
of
service
measures
trying
to
highlight
or
identify
either
memory
leaks
or
or
things
that
aren't
being
garbage
collected
or,
and
people
pointing
the
finger
at
the
service
mesh
components,
but
in
fact,
depending
upon
what
application
you're
using
they
may
become
bloated
as
well.
So
to
speak,
like
the.
A
B
A
B
A
D
D
D
So
you
can
just
energy
Murli
the
delay
that
you've
got
the
ability
to
do
like
50
cent
if
I
1999,
if
you
want
to
create
a
light,
realistic
traffic
or
you
can
kind
of
you
know
if
you've
got
a
service
which
exists,
which
you
know
what
those
percentiles
are,
you
can
actually
configure
those
in
in
it.
It
will
behave
with
the
same
kind
of
curves
random
variants.
It's
a
pretty
simple,
pretty
simple
little
little
thing,
but
I'm
thinking
you
could
you
know
I
can
build
up
like
pretty
complex
topologies.
B
D
By
basically
just
doing
Devin
all
right
the
way
down
the
stack
and
I'm
I'm
gonna
have
G
RPC
in
there
as
well,
because
my
intention
is
that
you
should
be
able
to
go
HTTP
RPC
job.
You
seem
to
HTTP
GL,
you
see
the
jail
you
see,
you
know
then
love
a
block
because
that's
kind
of
like
you
know
typical
stacks
for
all
and
of
course,
GRP
see
potential
we'll
perform
different
different
service
meshes
and
with
different
configurations.
A
B
D
A
Yeah
I
mean
yeah,
and
then
anything
about
a
lot
of
this
is.
Is
that
what
you
end
up
finding
is
there's
a
school
of
thought
and
I
think
that
they're
both
correct
but
there's
a
school
of
thought
that
says
well
hey
if
you're
testing
everything
in
a
vacuum
like
well,
that's
not
real
you're,
not
actually
going
to
capture
the
the
actual
signature
of.
What's
going
like
the
actual
thing,
that's
gonna
crash
and
burn
and
then
happen
and
and
there's
another
disclose,
thought
that
says
well.
C
Actually,
learn
quite
testing
the
reason,
a
some
bunch
of
us.
We
are
also
doing
a
bit
of
research
for
the
inch
computing
eleven
to
leverage
the
network
service
nation,
and
we
have
been
doing
a
lot
of
fast
performance,
metrics
and
probably
was
liberated
one
of
the
what
we
are
doing
is.
We
are
also
looking
to
see
if
you
can
basically
bring
in
a
bit
of
hardware
acceleration
so
that
the
service
mesh
doesn't
suffered
because
you
have
the
hardware
acceleration
engaging
a
spectacle.
I'll
probably
share
some
flights
soon.
B
C
C
C
Okay,
the
reason
I'm
asking
yourself
I,
don't
want
to
copy
body
to
death
on
a
Friday
morning
with
things
that
you'd
be
knowing
already.
So
let
me
probably
quickly
start
with
a
quick
introduction,
so
network
service,
even
though
it
has
a
word.
So
it's
mission
it
it's
a
big,
tough
terminal
towards
this
mesh
does
today.
C
C
Whereas
if
you
look
at
the
network
applications,
they
typically
need
more
than
one
a
professed.
That
was
one
issue.
The
second
issue
is
most
of
the
publications
today
the
enterprise
would
they
would
go
beyond
or
beneath
the
adultery
and
which
is
essentially
l3
and
able.
So
these
are
the
two
challenges
which
in
a
way,
was
a
big
trend
to
migrate.
C
Some
of
the
network
applications
from
that
of
VMs
to
that
of
kubernetes
Bay
sports
I
used
to
be
working
at
Ericsson
before
so
we
try
to
have
soft
work
closely
with
the
kubernetes
community
to
get
this
enable,
but
I
mean,
since
the
concepts
are
quite
well
defined,
and
it
was
quite
well,
it
was
Apple
does
for
us
to
take
that
network
service.
Nice
was
born.
C
A
brought
Nicky
from
Cisco
started
it
along
with
much
of
teaching
from
that,
as
well
as
me,
so
much
difficulys
infrastructure
layer
from
n
0
to
L
3,
and
it
basically
follows
the
service
mesh
pattern
which
every
one
of
us
familiar
with
and
what
simple
things
one
of
function
and
what
you
want
to
do
is
you
want
to
just
have
have
it
exposed
to
extreme
road.
So
what
happens
is
when
the
network
function
comes
in?
It's
a
network
function.
Example.
This
one
should
be
a
firewall.
C
Did
you
commonly
use
call
it
can
be
a
scintilla
cess
wife,
a
toaster,
a
soffit
is
available
right,
don't
want.
It
concerns
the
infrastructure
that
is
part
of
the
network
service.
Mesh
exposes
it
to
the
external
world,
wherein
it
understands
view
from
the
definition
on
what
of
the
service
endpoint
citrix
posts.
C
It
goes
and
registers
of
the
kubernetes
registry,
as
we
speak
today
and
the
clients
that
wants
to
use
it
then
essentially
look
up
for
the
service
as
well
as
get
a
service
standpoint
and
then
starting
hoping
the
cost
that
exposed
by
that
network
service.
Now
the
network
service
can
be
simple
or
complex
depending
upon
what
they
envy
other
ones,
and
the
second
thing
is
secure
from
the
networking
background.
As
Lee
was
mentioning
earlier,
the
chains
of
the
service
function
chain
is
a
very
common
functionality.
C
What
it
does
is
it
basically
creates
a
connection
or
change
between
a
bunch
of
network
functions
to
do
a
particular
activity.
Just
to
give
a
simple
examples:
let's
assume
that
you
are
installing
a
or
you
are
setting
up
a
small
business
office
at
home
right
now.
You
definitely
expect
you
need
a
fire
wall.
You
need
probably
like
a
more
social
aspects
like
that,
and
you
want
to
create
a
network
component
Bickford
for
your
business,
because
you
will
have
multiple
users.
You
have
need
subnets
for
the
assets
right
right
so
now
as
well
conventional
world.
C
That
all
effect
is
now
brought
into
the
cloning
in
world.
This
particular
concept
axis
mesh
essentially
allows
you
to
package
all
of
it
together
and
then
offer
it
as
a
service.
What
it
means
is
I
can
say:
okay,
I
know,
I
need
a
community
option,
a
which
is
just
a
simple
like
a
simple
network
connection
option
D
would
essentially
be
safe
until
containable
option.
C
would
be
with
firewall
enable,
but
at
the
end
of
today,
all
of
it
is
packaged
and
then
published
into
as
a
service
and
then
little
the
end.
C
A
A
Derive
from
network
service
mesh-
and
it's
probably
just
my
ignorance-
my
ignorance
said
that
leads.
Leave
me
with
questions
on
other
public
clouds
that
maybe
don't
give
tenants-
quite
the
perhaps
not
quite
the
level
of
access
that
they
would
need
for
to
take
advantage
of
some
of
the
capabilities
of
network
service.
Mesh
I
think
that
you
spoke
to
that
previously
like.
If,
if
certain
hardware
capabilities
are
present
and
assemble
pick
those
up
and
leverage
DB,
TK
or
PPP
they're,
not
your
present,
then
maybe
it
wouldn't
do
this
thing.
Actually,
my
question
was
intended
to
be.
A
B
C
That's
just
push
that
a
great
list,
and
because
some
because
currently,
if
you
look
at
the
service
corridor,
be
temple,
place,
I
mean
Enterprise
in
particular,
they
don't
really
play
along
with
the
congregation's,
because
further
distances
core
infrastructure
is
essentially,
they
are
ready
to
go
with
any
company
that
offers
own
as
a
package
right.
So
that
is
where
they
always
go
for
honor.
They
bought
solutions
lead
from
any
enterprise
networking
company
like
juniper
or
Cisco
right,
and
they
just
use
it
answers
things,
there's
a
bit
of
disruption
that
is
happening.
C
I
want
to
call
mention
it
more
here
right
now
with
5g
coming
into
play
right,
even
though
5g
may
look
to
be
so
this
product
play.
It
is
also
becoming
more
of
an
enterprise
play
the
reason
being
there
are
technologies,
the
radio
level
that
enables
services
enterprises
to
build
their
own
infrastructure,
including
the
radio
component,
and
just
to
mention
to
you.
If
you
go
to
Amazon
Web
service
and
they
just
look
at
the
marketplace,
you
already
see
network
functions
from
Jeany
Perez
could
be
readily
available
as
a
software
component.
C
The
second
thing
is
also
continuous,
like
a
foam
networks
that
are
bringing
in
a
lot
of
the
what
we
call
as
the
mobile
backhaul,
because
it's
a
wall
packet
core
functions
they
have
brought
in
as
a
software
component
into
that
of
the
AWS
at
the.
To
sum
it
up.
What
it
means
is,
there's
a
lot
of
disruption
that
is
happening,
and
eventually
things
are
becoming
more
cloud
centric
and
just
to
add
to
it.
C
So
that
is
where
such
an
infrastructure
replace
networks
of
this
mesh
plays
a
very
important
role,
because
you
do
have
on
one
side.
The
network
functions
are
becoming
more
like
a
common
software.
On
the
other
side,
you
need
to
have
those
constructs
as
soon
as
the
infrastructure
that
will
enable
them
to
leverage
it.
To
allude
to
the
example,
what
I
was
mentioning,
particularly
the
infrastructure,
has
to
be
ready
in
order
to
host
those
network
functions
right,
so
that
is
where
the
it
becomes
much
much
more
important
to
focus
on
l
0
to
L
3.
C
Now
it's
up
to
the
enterprise
so
that
they
have
appetite
to
do
it
or
that
they
are
the
interest
to
do
it
or
it
can
be
a
startup
that
can
essentially
probably
become
more
of
a
study
to
provide
a
box
solution.
Picking
all
of
this
conference
and
probably
trying
to
say,
hey
enterprises.
Here
we
have
a
solution
for
you.
We
don't
need
to
worry.
We
can.
C
C
In
fact,
there
is
a
lot
of
development
that
is
happening
in
the
hardware
side,
particularly
DPD
Kas
already
movie.
These
are
some
of
the
technologies
that
them
to
get
not
really
to
the
speed
that
they
expect
but
near
speed,
and
you
don't
really
feel
the
different
side.
So
so
that's
that's!
That's
why
the
whole
industry
is
moving.
The
works.
A
A
B
C
C
Looked
at
okay,
one
of
the
things
I
said,
so
we
are
also
so
this
is
just
a
half
a
second
probably
explain.
So
what
happens
if
the
beeping?
It
is
just
a
sampling
function.
There
is
a
like
a
adapter
that
sees
on
each
of
the
compute
node.
It
understands
how
the
network
cells
mesh
works.
So
the
moment
a
network
function
comes
up,
it
basically
gets
it
goes
and
registers
it.
This
is
after
registration.
We're
in
a
client
wants
to
use
this.
It
has
brought
the
handle.
C
It
goes
to
the
particular
node
where
it
is
hosted
it.
This
is
basically
a
complete
handshake
to
get
the
being
established
and
then
basically
what
happens
is
there's
a
time.
There
is
an
interface
that
is
created
in
the
in
each
of
these
pots,
and
then
there
is
a
tunnel
which
is
created
between
the
Bourdon
and
into
the
user
forum.
Then
that
of
the
network
function
so
that
the
data
plane
crash.
We
can
start
going
and,
as
you
are
mentioning,
these
are
the
enhancements
that
are
done
on
the
EVPs
less
obvious.
A
C
So
they
started
with
the
sidecar,
but
now
it's
also
Union
container.
So
it's
quite
interesting
because
what
happens
is
the
choices
are
quite
varied
and
basically
the
it
is.
What
and
exactly
what's
happening
is
the
infrastructure
is
available,
but
now,
if
I'm
a
network
function
developer,
what
I
do
is
I
leverage
the
libraries
that
are
provided
by
network
service
mesh
and
then
I
start
developing
my
network
function,
and
that
is
how
it
is
so.
C
I
can
probably
leverage
the
sidecar
pattern
and
then
start
using
it
to
add
one
more
aspect
to
it
is,
for
example,
as
I
said,
it's
L
0
to
L
3,
but
then
what
we
are
looking
to
do
is
we
are
looking
how
everything
covered
in
the
sense
we
need
to
enable
all
the
service
mesh
tube
right.
So
that
is
what
currently
Fredrik
is
working
on
in
the
Envoy
integration.
So
the
moment
all
my
integration
was
done,
you
would
essentially
be
able
to
leverage
the
power
of
networks
of
this
mission.
B
C
And
now
I
saw
it
sort
of
British
thanks
again
to
help
me
understand
more
about
the
adapter
part.
So
here's
a
here's
where
fashion
is
working
on
I.
So
what
we
intend
to
do
is
we
intend
to
probably
have
a
sample,
be
enough,
there's
already
available
in
Denison
rupees
loaded
off
the
deck
of
the
machete
platform.
Now
these
sample,
we
enough,
can
be
thought
of
as
the
applications
in
the
general
context,
and
now
there
is
two
models.
C
One
is
the
whole
of
Amazon
uses,
G
RPC
ten
points,
which
means
we
can
essentially
develop
a
client
that
can
coexist
with
that
of
the
network
function
and
then
develop.
The
MSM
are
not
treated
essentially
go
and
start
manipulating
the
client
so
now,
depending
upon
how
we
want
to
test
it,
the
client
can
coexist
with
the
network
function
or
we
can
probably
get.
Those
API
is
external
to
that
of
MSM.
B
C
The
NSA
monitoring
away
becomes
the
client,
and
then
it
would
essentially
invoke
now.
One
important
thing
has
we
can
probably
use
measure
is
ability
to
do
testing
more
so
from
the
performance
for
aspect?
Wherein
only
thing
is
what
I
was
also
going
through.
Some
of
the
measuring
code
is
to
understand
what
hooks
can
be
used
to
test,
specifically
the
the
functionality
of
the
MSM
performance
part.
A
B
A
This
is
get
a
pic
pictures
really
help
well
I
mean
I
hadn't,
really
considered
for
the
derivatives
of
the
use
case,
where
we
were
focused
on
a
provisioning,
the
surface
mash
and
be
provisioning
a
workload
on
top
of
it.
You
know
a
sample
a
half
a
lot
of
times
and
whereas
in
network
service
meshes
case,
you
didn't
you
could
conceptually,
you
know,
displace
the
sample
app
with
a
sample.
Be
enough.
Yes,.
B
A
Right,
yeah
and
then
count
to
your
point:
I'll
try
to
regurgitate
it
to
make
sure
I
understand,
is
sort
of
potentially
two
approaches.
One
either
in
some
respects
have
where
we've
got
NSM
client
here
have
in
some
respects.
It
could
be
that
the
that
that
is
more,
that
that's
more
or
less
the
NSM
adapter
or
yes
performing
that
yeah.
C
Yes,
absolutely
it
can
exist
with
the
not
externally,
because
here's
the
two
models-
parent,
the
NSF-
is
being
useful
if
the
client
or
if
the
client
is
within
me
within
the
same
node,
what
happens?
Is
it
leverages
a
capability
cook?
What
we
call
us
the
memory
interface
right
instead
of
using
the
socket
a
typical
socket
to
get
out
of
the
network?
We
basically
use
the
memory
interface,
so
that
is
much
more
powerful.
C
C
C
So
so
what
happens
is
also
we
as
part
of
it?
We
need
to
see
how
we
can
load
either
VDP
data,
petrol,
VST
adapter,
and
this
whole
model
thing,
because
data
plane
has
the
intelligence
to
know
how
the
whole,
how
optimization
can
be
done.
So
it
comes
as
a
complete
package.
So
what
I'm
not
put
the
details
on
it?
Probably
I
will
probably
further
on
details
on
it.
How
the
real
department
will
look
like,
so
what
I
can
do
is
I
can
do
scenario,
one
both
clients
lecture
addresses
from
the
same
compute
node.
A
A
A
Hadn't
spent
enough
time
immunity,
irrespective
of
the
adapter
discussion,
just
I
hadn't,
spent
enough
time,
considering
all
the
unique
use
cases
that
in
a
sim
has
its
own
style
of
service
mesh
offers,
and
so
that's
that's
one
exciting
thing
and
then
to
yeah,
hey
the
the
having
a
mind
like
yours.
Consider
the
the
realm
of
what's
possible.
Yeah
well,
is
exciting
because,
because
you
know,
good
I
was
just
about
as
comfortable
to
say.
Yes,
I
am
familiar
with
network
service.
Smash
familiar
is
a
cushy
term
I,
whatever
I
would
describe
it.
A
A
C
To
see
whenever
they
given
talk
about
it,
I
differentiate
but
I
basically
talk
about
the
differences
between
both
of
it
so
that
people
can
tune
into
more
of
because
when
we
talk
about
service
message,
it's
a
quite
a
different
and
applications,
and
you
talk
about
sidecars.
We
talk
about
the
ingress
controller
aspects
like
that
right.
So
if
I
don't
differentiate
them,
I
see
a
lot
of
people
are
me
when
I
started.
Initially,
we
thought,
if
part
about
the
difference
we
see
stars.
Where
is
my
controller?
Where
is
my
whether
the
traffic
coming?
A
Maybe
you
know
maybe
Nick
can
identify
with
this,
that
yeah
hey.
In
fact,
when
you're
saying
service
mesh
you're,
not
you're,
not
actually
just
saying
sto,
which
people
make
the
mistake
of
doing
this
like
well,
yeah,
that's
you
know
one.
You
know
one
very
popular
super
top-of-mind
implementation,
but
actually.
C
I'm
doing
a
lot
of
thoughts
about
link
early
and
other
things,
mainly
because
I'm
now
putting
my
hat
on
how
some
of
it
can
be
leveraged
for
the
edge
computing
world,
so
I'm
learning
a
lot
from
how
different
each
one
of
it,
and
why
is
that?
There
is
a
bit
of
performance
difference
between
both
of
them,
because
it's
we
can
in
general,
if
we
say
that
hey
performance
differs,
but
then
the
implementation
it
has
its
own
sets
of
reasons
why
each
one
is
simply
marking
a
different
weight
range
and
so
on.
I.
C
A
I
can't
tell
you
how
many
times
I
reference
the
surface
mesh
landscape
on
layer,
5
in
the
east,
EO
up-and-running
book-
you
want
it
when
it
gets
published,
you
will
see
I
tried
not
to
hawk
it
too
much
yeah
I'm
stoked
about
one
seen
its
face
again
and
Nick.
Some
of
the
things
that
you
described
more
like.
A
Clearly,
we
don't
have
enough
bandwidth
to
actually
meaningfully
do
some
of
the
things
I've
been
talking
about,
but
to
hear
you
say
it
gives
me
hope
so
and
then
present
boom,
yeah
I,
actually
kind
of
think,
not
that
we
wouldn't
want
to
share
all
of
this
on
a
community
call.
We
absolutely
do,
but
but
some
of.