►
From YouTube: Cloud Native Austin April 2020 Meetup: Meshery and KUDO
Description
Event details - https://www.meetup.com/Cloud-Native-Austin/events/267784090/
Talks:
Meshery - The Multi-Mesh Manager
Lee Calcote, Founder, Layer5
https://twitter.com/lcalcote, https://twitter.com/layer5
KUDO - Kubernetes Operators the Easy Way
Gerred Dillon, D2iQ and Michael Tanenbaum, Solutions Engineer, D2iQ
https://twitter.com/tbaums
A
B
I
was
about
to
make
our
first
dad
joke
of
the
evening.
I
think
I,
genuinely
don't
pre
contemplate
these
things,
but
as
Jared
points
out,
we
are
we're
gonna.
Do
we're
gonna
attempt
some
live-streaming
here
at
the
same
time
that
we're
we're
gathering
around
the
gathering
around
the
fireplace
sharing
some
sharing
our
technology.
So
as
such
don't
say
anything
that
your
mother
wouldn't
approve
of,
because
we
are
not
only
are
we
recording
but
we're
live
so
or
we're
at
least
20
seconds
drag,
live
good,
okay,
all
right,
they're,
not
very
good.
B
So
what
about
we're
about
six
minutes
after
we've
got
a
small
collection
of
us
coming
together?
Hopefully
we'll
have
a
few
others,
but
let's
get
kicked
off
you
know
and
as
we
do
that
I
suggest
before
we
do.
Let
me
do
this
out
of
curiosity
I,
wonder
who
all
can
grab
the
live
stream
link
if
you
can't
grab
it,
I
went
ahead
and
shared
it
in
the
chat,
so
try
not
to
watch
yourself
talking
20
minutes
ago
or
20
seconds
ago
that
a
little
undo
you,
okay,
very
good,
very
good,
all
right!
B
B
We
tend
to
generally
hit-
maybe
hopefully
sort
of
the
introductory
folks
who
are
just
coming
in
to
cloud
native
things
and
folks
who
have
been
here
a
while
and
those
in
the
middle,
it's
kind
of
hard
to
hit
the
full
range,
but
between
a
couple
of
talks
we
might
have.
Hopefully
that
would
would
be
the
case
so
tonight
we're
going
to
do
a
little
bit
of
socializing
a
little
bit
of
get
know
each
other
hopefully,
and
some
of
us
are
very
familiar
faces.
B
We're
gonna
dive
into
Kudo
and
Michael
and
Jared
are
here
of
the
open
source
crude
oil
project.
To
tell
us
tell
us
about
how
to
do
kubernetes
operators,
the
easy
way
I
suspect,
not
the
which
I
guess
implicitly
means
there
are
some
hard
ways
to
do
that,
and
a
couple
of
other
projects
come
to
mind.
So
the
second
talk
is
about
measure
II.
It's
as
a
multi
mesh
manager,
we'll
talk
about
that.
B
Welcome
those
who
haven't
been
at
cloud
native
Austin
before
we've
been
well
gosh
cloud
native
Austin
wasn't
the
name
of
the
Meetup
when
we
started
it
was
micro
services
and
containers,
and
we
called
it
that
because
I
also
organize
docker
Austin
and
way
back
when
we
were
having
container
orchestration
Wars.
We
were
not
quite
so
comfortable
to
talk
about
talk
about
some
of
those
same
things
in
the
same
venue
anymore.
It's
one
big
large
cloud
native
ecosystem,
and
so
we
were
able
to
bring
these
two
meet
ups
together.
B
So
some
of
you
might
be
here
from
dr.
Austin.
Some
of
you
might
be
from
cloud
native
Austin,
your
respective
you're
here.
If
you've
got
something
to
share,
please
do
otherwise
I
think
I've
warned
in
the
past
that
you're
likely
to
get
a
continual
feed
of
having
to
listen
to
tule,
and
that
is
not
what
this
meetup
is
about.
It's
just
that
I'm,
the
only
one
raising
my
hand,
so
there's
a
forum
or
two
out
on
the
cloud
native
Austin,
meetup,
page
I,
believe
it's
there
for
dr.
Austin
meet
a
page
as
well.
B
Please
fill
it
in
with
with
any
any
ideas
or
things
you
might
want
to
present.
We
do
have
a
number
of
upcoming
events.
If
you
look
on
the
cloud
native
Austin,
meetup
page
there's
some
other
interesting
talks
coming
up.
They
got
warned
at
the
start
of
the
year
or
late
last
year,
when
we
got
rolling
again
that
we
were
gonna
have
a
bit
of
a
messy
agenda
sort
of
perpetually.
B
For
my
part,
that
tends
to
be
part
of
part
of
my
focus.
It
also
tends
to
be
kind
of
the
the
in
my
mind,
sort
of
a
next
step
for
a
lot
of
folks
kind
of
just
just
after
kubernetes,
and
so
mostly
because
it's
a
focus
of
mine,
we
you
were
forced
to
listen
to
it.
But
our
point
is
that
the
point
of
this
is
is
really
about
being
vendor-neutral.
It's
primarily
about
open
source
things
within
the
scene.
Cf,
all
of
the
I
don't
know
how
many
projects
are
we
closing
in
on
now.
B
Are
we
toured
with
40
we
toured
on
it
where
we're
getting
up
there,
that
there's
so
much
to
talk
about
so
so
there's
really
more.
We
only
meet
once
a
month,
but
there's
more
to
talk
about
than
we
could
possibly
get
to
you.
So
in
those
months
that
we've
been
meeting
last
month,
we
were
fortunate
that
I
gnw
had
sponsored
and
they
had
sponsored
some
sweet,
sweet,
local,
Austin
beer
and
d2
IQ
is
had
sponsored
this
month's
Meetup.
B
This
is
the
time
for
me
to
be
quiet
and
listen
to
everyone
else,
just
as
we
get
together
as
well,
no
not
just
a
local
community,
but
a
community
of
those
interested
in
some
of
the
same
technologies.
It's
a
good
time
to
raise
your
hand
and
and
stay.
You
know
answer
some
of
these
questions.
Just
kind
of
are
people
looking
for
jobs.
It's
a
great
opportunity
to
meet
others
here.
People
are
trying
to
convince
talent
to
come
join
them
now.
Please,
please,
speak
up.
Now
is
a
great
time
to
signal.
C
C
You,
you
might
be
familiar
with
with
the
recent
launch
of
VMware
tons
ooh,
and
so
you
know
we
have
a
lot
of
customers
that
are
exploring
the
service
mesh
base
and
worth
looking
for
people
that
can
help
them
another
journey.
So
I
can't
point
specifically
right
now
to
any
particular
opening
or
what
cities,
but
until
you
that
we're
we're
aggressively
pursuing
getting
people
the
right
positions
to
help
us
with
it.
Nice.
B
C
That's
a
great
question:
I
think
you
know
you
have
those
two
sort
of
opposing
what
would
seem
like
opposing
viewpoints,
but
when
you
think
of
tongs
ooh
it
brings
it
all
together.
You
know
tons
of
being
a
perspective
of
the
application
and
you
know
development
of
an
application
it
with
all
the
tools
like
application.
Catalog
was
just
formally
bitNami
or
even
you
know,
hefty
o
green.
C
What's
now
called
tons
of
kubernetes
grid,
but
then
you
also
have
the
perspective
of
the
infrastructure
and
the
layer
of
abstraction
from
from
the
complexity
there
tons
ooh
brings
all
of
that
together.
So
you
know
I
love
that
question.
How
are
they
blending
tons?
Ii
really
kind
of
brings
all
that
together.
Now
you
know
where
the
rubber
hits
the
road.
It's
still
like
early
days
in
terms
of
people
getting
to
know
each
other
and
when
working
together
and
planning
together.
But
it's
been,
it's
been
a
great
experience.
Nice.
B
Yeah
I
was
just
just
having
a
conversation
with
a
couple
of
other
vmware
folks,
I'm
just
morning
on
tong
zoo's
service
smash,
nice.
Okay,
oh
good,
go
back
here,
the
other
question
that
we
generally
like
to
ask
when
we're
talking
about
hiring
and-
and
is
it
more
or
less
say
what
else
is
going
on
anything
anything
interesting
that
has
happened
recently
that
somebody
would
like
to
share
who's
got
a
share.
D
B
I'll
share
it
earlier
today,
I
had
a
well
a
C&C
of
sandbox
proposal
presented
to
me
it
within
the
scenes.
Yes,
a
network,
it's
called
chaos
mesh
and
it's
a
project
as
from
ping
cap.
It's
the
same
company
that
brought
forth
PK
IV
and
so
they're
they're
accustomed
to
doing
open
source
projects.
Tk
IV
is
a
scene
CF
project
and
they're
looking
to
donate
chaos,
mash,
which
is
a
really
young
projects
about
four
months.
B
B
Good
yeah,
fantastic
yeah,
I,
very
happy
that
you
guys
are
here.
We
also
quick
item
last
time
we
met
we'd
said
that
open
source
101
I
was
coming
to
Austin
and
Todd
Lewis
came
by
the
Meetup,
the
organizer
of
the
event
and
gave
a
few
words
about
it.
That
has
gone
in
virtual
and
is
now
I
think
rescheduled
for
May
12th.
So
a
free
virtual
event,
but
if
you're
just
getting
into
opens
worse,
Todd
and
all
things
open
is
the
other
event
that
the
key
and
crew
run.
They
got
a
great
lineup
of
speakers.
B
B
A
Great
Thank
You
Lee.
Let
me
share
my
screen
here:
I'm,
not
gonna,
go
fullscreen,
so
y'all
get
to
see
my
deck
set,
app
Chrome,
but
that's
okay,
yeah!
So
we're
here
to
chat
about
the
Kudo,
the
coober,
an
universal
declarative
operator.
It's
it's
a
dynamic
operator
and
I
really
look
forward
to
continuing
to
spread
the
the
message.
A
So
you
know
who
I'm
a
principal
member
of
staff
at
a
d2
IQ
I,
initially
create
in
a
corporate
or
akuto
I've
done
Cooper
nation
for
a
long
time,
I'm
heavily
involved
in
not
only
the
Cooper's
level,
but
also
the
CN
CF
level
and
I'm
a
chair
on
the
operated
working
group
and
contributed
strategy.
In
a
few
other
places,
chopping
wood
carrying
water
everywhere,
I
can
and
trying
to
advance
the
mission
and
without
right
over
to
Michael
to
introduce
himself.
A
So
a
little
we've
a
bit
of
a
story
together
because
you
know
I,
think
I.
Think
when
you
hear
the
word
operator,
it's
a
very
loaded
term
and
by
the
way,
if
anyone
has
any
questions,
please
Bob
on
the
zoom
group
chat
or
in
the
YouTube
chat
and
Michael
and
I
will
answer.
That's
casual
feel
free
to
intervene
whenever
I,
don't
we've
a
bit
of
a
story,
because
we
hear
this
operator
term
all
over
the
place
and
we
only
we.
A
It
starts
from
a
very
deep
understand
about
what
kubernetes
is
right
and
and
if
you
think
about
what
cooper
neighs
actually
is
you
know
we
we
start
out
with
some
of
this
notion
of
declarative
data,
and
on
top
of
that,
we
layer
on
a
series
of
control
loops
or
a
really
level
based
triggering.
That
does
something
with
that
data
right.
A
We
compare
a
state
against
a
desired
state
and
we
advance
on
that
stateful
thing
with
all
those
controllers
in
kubernetes
right
like
and
the
core
of
that
is,
as
it
reads
out
like
if
I
have
a
state
of
something
I
care
about,
and
that
changes
and
I
have
an
inferred
state
and
a
specification
for
what
I
want.
I
always
want
to
be
advancing
the
state
towards
that
thing
right.
We
thought
that
controller
would
have
a
reconcile
reconciliation
loop
that
actually
performs
that
advancement.
A
So,
okay,
great,
we
have.
We
have
this
notion
of
controller.
We
have
this
think
on
operator
thrown
around
which
doesn't
sound
that
much
different
from
its
onset
and
even
in
the
operator
working
group,
where
we're
working
to
nail
down
a
definition
that,
since
it's
very
complicated,
if
you
look
in
the
TOC
mailing
list
and
the
app
delivery
mailing
list,
you
know
we
we've
had
a
lot
of
discussion
about
this
right.
You
know
us
and
a
lot
of
others
you
know
have
had
have
worked
through
this
problem
of
what
exactly
is
an
operator.
A
So
you
know
we
we
we
start
to
describe
this
as
okay.
We
have
something
that
is
data
with
something
that
advances
a
seat
on
it.
What
says?
Are
you
controller
right?
So
so,
what's
the
difference
really
between
a
controller
and
an
operator?
Well,
a
controller
reconcile
state.
It
doesn't
care
what
you're
doing
underneath
the
hood.
It's
thinking
about
its
own
domain
above
a
pod,
a
deployment,
how
do
I
advance
those
things
right
and
so,
in
a
general
sense,
a
controller
plus
a
declarative
resource,
meaning
an
operator
doesn't
make
sense
right.
A
We
we
already
know
that
we,
we
don't
necessarily
want
to
do
this
and
it
really
comes
down
to
application
awareness
right,
and
so
what
do
I
mean
by
that
like
it's,
it's
how
your
application
interacts
with
other
applications.
How
your
application
interacts
with
other
applications
depends
upon
if
I'm
FMO
Postgres
database.
Something
is
expecting
me
to
be
able
to
back
up
and
restore
myself
and
and
add
an
index
and
allow
connections
and
new
connection
point
all
that
you
know
how
I
scale
it's
a
it's,
not
just
a
matter
of
increasing
your
number
of
replicas.
A
When
you
talk
about
a
more
complicated
application,
if
I'm
gonna
go
scale
at
CD,
I
actually
have
to
call
an
API
endpoint,
the
membership
API
net
city
to
scale
up.
My
number
of
replicas
read
CD
and
the
same
for
scaling
down
right
and
then
also
like
what
your
application
needs
to.
No
one
needs
to
upgrade
and
recover
and
sorry
for
the
slides
getting
cut
off,
but
but
but
really
it's
it's.
It's
hailing
a
full
lifecycle
that
is
very
contextual
to
the
application
that
we're
running.
A
So
we
pause
on
the
could
have
team
and
we're
working
on
San
upstream,
as
well
as
part
of
the
operating
worker
group,
that
an
operator
is
a
controller,
there's
well
optimized
for
the
requirements
of
its
domain
and
there's
a
lot
more
nuances
there.
Then,
then,
we
have
time
for
this
talk,
but
if
you,
if
you
apply
domain
model
plus
a
controller
now
you
get
an
operator
right
with
life
cycle
as
a
focus
there.
A
So,
okay,
great
Jared,
you've
convinced
me.
Let's
look.
Let's
make
operators
out
of
everything
you
know
like
if
it
exists.
Why
don't
we
make
an
operator
out
of
it
right?
Postgres
sounds
great,
my
green
right
backups,
let's
add
a
Kafka
topic.
Why
don't
we
make
it
Kafka
topic?
A
CRT
sure
gives
us
like
routes
right
and
operated
for
all
these
things,
but
you
start
to
see
a
problem
when
you
approach
this
world
I
was
like
okay,
great
I'm
gonna
write
an
operating
for
my
database
and
okay.
A
My
database
is
a
hundred
thousand
a
million
lines
of
code
and
alright
I'm
gonna
write
a
hundred
thousand
lines
of
opera
or
10,000
lines
of
operating
code
to
actually
run
that
alright.
Well,
let's
try
to
drive
this
model
its
it's
at
right,
all
right,
good
cool!
My
model
is
okay,
ten
thousand,
let's
code,
two
thousand
lines
of
operator
curve,
well,
maybe
sure
an
operating
from
a
macro
services,
so
I
instantly.
A
So
this
this
this
relationship
starts
to
invert
upon
itself,
where
why
am
I
writing
tens
of
thousands
lines
of
code
to
operate,
something
that
is
the
smallest
possible
thing
right
and-
and
we
we
into
this
this
this
level,
where
suddenly
doesn't
make
sense
and
what
that
will
you?
You
know
what
the
heck
am
I
doing
if
you
look
at
a
very
concrete
example.
A
If
you
go
back
to
September
and
and
you
had
a
lot
of
home
trucks
and
a
lot
of
tooling
using
apps
V
1
beta
1
or
extensions
V
1
beta
1,
you
had
a
bad
day
when
they
dropped
ABS,
V
1,
beta
1
and
extensions
V
1
beta
1,
when
deployments
became
apps,
V,
1,
right
and
and
it
can
be
a
full-time
job,
just
keeping
up
with
kubernetes
releases.
If
you're
writing
software
on
top
of
kubernetes
and
your
domain
is
not
a
kubernetes
developer.
A
So
how
do
we
not
all
just
become
kubernetes
developers
and
and
maintain
our
domain
of
hey,
I'm
a
great
developer
working
with
the
sandra?
I
want
to
ship
cassandra
on
top
of
kubernetes
without
knowing
all
these
things
intricacies
about
the
api
right
and
and
really
what
we
posit
on
the
Karoo
teams.
We
have
an
accessibility
problem
with
orchestrating
applications
and
in
an
application
aware
way
native
to
that
tooling
right.
So
we
believe
that
operating
development
requires
too
much
Cabernets
domain
expertise.
We
think
that
operators
can
actually
go
really
far.
A
There's
there's
a
great
case
for
doing
a
lot
of
orchestration,
using
this
sort
of
application,
awareness
and
tooling,
and
really
want
to
enable
new
pathways
so
that
more
different
applications
can
be
built
in
the
kubernetes
native
way
without
having
to
have
15q
Bernays
experts
wrapped
around
your
organization.
So
that's
why
we
created
Kudo.
You
know
we
we
really
focus
on
like
like,
like
you,
should
be
experts
in
your
domain.
We
believe
that
you
already
are.
A
We
don't
think
you
should
have
to
be
actually
think
in
kubernetes
now
and
and
for
every
effort
in
the
future
we,
but
we
think
we
can
do
we.
We
can
do
that
with
just
using
kubernetes
in
and
of
itself,
and
so,
how
does
that
manifest?
Right?
Like
I
think
we
focus
on
the
orchestration
of
Carre's
primitives,
including
other
operators.
A
We
like,
like
we're
it's
a
tool
focused
on
shipping,
your
application,
alongside
its
runbook
right
I've,
deployed
my
application
now
I
need
you
backups
restores
upgrades
compactions
all
these
things
that
would
normally
be
part
of
a
run
book,
bury
it
into
your
into
your
app
and
use
your
native
tool.
Right,
like
like
you,
sequel,
might
might
like
the
my
sequel,
CLI
tool,
use
the
Cassandra
CLI
tool,
don't
rewrite
all
this
stuff
and
go,
and
we
really
are
promoting
application,
awareness
and
use
of
the
application
native
tooling.
A
Instead
of
writing
something
mice,
my
sequel,
dump
or
PG
dump
in
go
right
and
we
don't
think
we're
the
right
tool
for
every
app
to
operate
on
the
planet.
Just
a
lot
of
the
you
know
those
that
have
that.
We
also
believe
that
those
there's
room
for
and
we're
built
on,
QQ
builder,
as
well
for
exposing
other
primitives
that
something
like
Kudo
can
use
in
that
orchestration
environment.
A
And
you
know
we
have
community
cope
governance
in
a
roadmap
where
we're
heading
into
the
C&C
of
sandbox,
where
we're
really
in
the
administration
phase
of
that
and
Kudo
is
just
kubernetes.
So
quick
could
architecture.
This
is
actually
changing,
really
in
flux,
but
the
core
of
it.
You
have
some
Sierra
T's
that
define
an
operator
and
and
in
the
instantiation
of
those
operators.
You
have
a
controller
that
operates
on
those,
and
you
have
repository
model
that
we're
moving
over
to
OCI.
A
That
contains
the
operator
and
potentially
it's
imaging
in
the
few
images
in
the
future
for
air
gap
deployment.
So
the
the
core
piece
here
is:
we
have
an
operator,
we
have
versions
of
those
operators
and
then
you
instantiate
the
operator
at
that
version,
which
then
can
move
from
version
to
version
to
version.
But,
most
importantly,
we
have
a
single
polymorphic
controller
that
operates
on
all
of
this,
instead
of
instead
of
you
having
to
run
a
different
operator
for
every
single
instance.
A
So
as
of
today,
Kudo
is
a
stable,
we're.
Actually,
in
version
ahead,
I
forgot
to
update
this.
We,
you
know,
we
have
a
declarative
approach,
but
we're
still
manually
sequencing.
We
have
all
the
things
you'd
expect
to
be
able
to
actually
write
and
make
these
operators
and
and
our
base
functionality
is
very
stable
and
what
we're
experimenting
ways
to
where
to
where
we're
heading
towards
1.0,
in
conjunction
with
with
of
advisors
up
in
the
the
cnc
of
toc.
A
A
That's
a
massive
win,
you
know,
and
and
I
and
I
get
to
use
Valero
to
go
move
that
from
cluster
to
cluster
cluster,
so
that
that's
what
that
means
we're
actively
working
on
drift
detection
right
now
and
running
plans
were
supposed
to
events,
progressive
enhancement
of
helm,
charts,
that's
one
we're
working
very
heavily
on
you
know:
helm
is
a
fantastic
package
delivery
tool.
How
can
you
add,
backup
and
restore
to
the
to
the
bid?
Nami
Postgres
chart,
support
buttons
and
pre-flight
checks
is
almost
in
and
interoperable
something
koi
which
I'll
talk
about.
A
In
a
few,
we
have
a
great
community,
we
we
we
do
caps,
we
have
open
meetings
and
sync
to
open
governance
over
the
roadmaps.
You
know
that
we've
been
community
Orange,
it's
the
beginning.
We
continue
to
be
there,
there's
no
other
pathway
forward
for
us
and
as
part
of
entering
the
same
sandbox,
there's
there's
no,
no
change
in
that
we
were
the
QBs
podcast.
A
Somehow
we
became
the
first
recommended
tool,
we're
doing
workshops
we're
doing
talks.
We
we
won't.
We
won't
shut
up
about
this,
and
and
and
we
want
people
to
be
productive
building
operators.
So
so
we
evangelize
the
hell
out
of
it.
We
have
a
lot
of
operators
build
with
Kudo.
We
have
the
users
using
this
as
well
and
and
and
also
not
only
using,
but
also
contributing
to
various
parts
of
this.
You
know
we
have.
A
So
in
the
CN
CF,
not
only
we're
going
to
be
a
sandbox
project,
we're
driving
interfaces
so
that
you
know
we
want.
We
want
operators
to
all
work
together
and-
and
we
think
that
we
can
do
a
lot
better
working
with,
like
others
in
the
community,
so
that
operators
can
consume
other
operators
and
end-user
tooling
can
all
look
the
same
so
that
you
know
what
we
you
have
this
problem
right
now,
where,
if
I
install
five
different
operators,
I
have
to
learn.
A
Five
different
operators
and
I
have
to
learn
the
underlying
tech
I'm
using
so
creating
interfaces
to
help
smooth
that
over
so
that
tooling
can
be
generated
in
order
to
both
depend
on
and
use
operators.
No
matter
how
they're
built
and
and
we
work
with
partners
than
in
the
CN,
CF
and
and
in
computers,
communities
to
make
this
happen.
A
It'll
post
these
slides
as
well
I'm
not
going
to
get
every
bullet
point
because
I'd
love
for
love
to
get
to
the
demo
without
taking
too
much
time
out,
and
then
we
also
just
released
cuddle
or
provoke
we're
sending
this
out
to
sig
test.
It's
the
kubernetes
test
tool.
This
was
originally
kudo
test
and
it's
a
declared
a
framework
for
writing,
conformance
teton
test
for
QA
resources
and
does
assertions
based
on
partial
gamal.
A
So
if
I
say
hey,
you
know,
I
want
a
deployment
and
I
want
to
say
it
has
two
replicas,
that's
all
I
put
in
and
it's
at
CU,
TTL,
dev
and
I
run
it
and
it
works
right.
So
we've
partnered
we
released
this.
We
now
have
involvement
from
from
Red
Hat
and
a
few
others
to
start
trying
to
drive
operator
scoring
and
hopefully
drive
this
up
into
two
basic
tests
to
provide
a
really
nice
declarative
and
the
end
test
tool
for
kubernetes
and
can
raise
resources
if
you
use
comp
comp
test.
A
This
is
fairly
similar,
except
this
runs
on
running
resources,
not
on
your
your
static
templates
and
we're
using
it
to
drive
a
matron
matrix
of
kubernetes
distributions
versions
and
operators
to
drive
it
and
conform.
It's
a
release.
Model
of
you
know
if
I'm
an
end-user
and
ask,
does
this
operator
work
on
this
version
of
kubernetes
and
provide
these
features.
That
was
that
was
why
we
originally
started
this
cuto
test
and
it's
now
spawned
out
into
engine
cutoff.
So
it's
it's
there,
a
zero
two
zeros
out
today,
but
it's
much
more
mature
than
that.
A
A
It's
it's
thinking
about
in
the
census
of
gathering
data
and
presenting
it
to
you
and
maybe
passing
some
tests
right,
whereas
Connell
is
entirely
focused.
If
you
were
to
think
about
LIGO
test
or
Omega
or
r-spec,
or
something
like
that,
that's
what
cuddle
is
it's
it's.
It's
an
assertion,
tool
that
operates
at
the
end-to-end
level,
so
selenium
for
kubernetes
is
another
way
to
put
it
nice.
A
And
that's
that's
why
we
split
it
out
because
we
started
knows
other
people
really
wanted
to
use
this
as
well.
So
it's
really
doing
partial
assertions
on
yamo,
that's
running
in
your
cluster,
so
you
you
set
up.
You
know
you
set
up
test
steps
to
create
a
resource,
and
maybe
that
resource
is
an
operator-
and
you
say:
okay
well,
I
want
to
assert
that
this
deployment
got
created.
I
want
to
assert
that
this
service
got
created
and
it
has
these
properties
so
that.
B
Test
set
is
like
to
your
point.
Is
not
you
not
necessarily
ask
the
question
different.
The
assertions
are
made
on
the
current
running
config
of
those
kubernetes
constructs
those
objects,
those
those
controllers,
the
the
CRTs
or
and
the
assertions
that
are
made
and
they're,
not
necessarily
against
the
original
yamo.
That
was
that
was
handed
to
kubernetes
to
deploy
that
operator
or
those
CRTs.
That's
correct,
yeah.
A
Oh
actually,
oh,
let
me
pause.
I'll
share
my
screen,
our
our
Kafka
tests,
because
this
explains
it
super
well
so
otherwise
get
get
involved.
We
have.
We
have
all
this
stuff
problem.
Sign
is
up
we're
being
River
in
Kauai
CUDA
on
the
Cooper,
a
slack.
Let
me
let
me
jump
over
real,
quick
before
I
turn
over
to
Michael
just
to
to
reify
that
a
little
bit
so
here
I
have
zeros,
you
were
installed
and
it's
gonna
go
install
an
instance.
A
So
this
is
the
CRT
that
that
cuttle
wants
to
create,
and
then
this
is
the
assertion
that
I
make.
Is
that
not
only
does
this
instance
exist
with
this
status,
but
this
stateful
set
exists
with
these
parameters
in
this
status,
so
I
have
I.
Have
you
know,
I
have
a
test
step
where
I
do
this
and
then
I
assert
that
I'm
very.
B
I
think
so,
yeah
I
know
it
actually
does
some
yeah,
which
is
I,
have
phrased
the
question
poorly
saying
that
no,
the
assertion
is
not
about
your.
It
does
verify
that
your
original,
what
you
originally
declared
is
in
fact
represented
and
working,
but
it
doesn't
necessarily
take
that
original
definition.
B
Rather,
it's
using
an
assertion
that
you've
defined
to
say
hey
in
the
same
fashion
that
you
would
use
a
selenium
or
jmeter
or
some
other
sort
of
functional
tests
to
say
it
well,
this
is
if
the
app
or
the
CRD
is
in
fact
deployed
and
healthy
and
doing
well
as
I
would
expect
here's
the
response.
I
would
expect
back
or
like
here's
the
the
characteristics
you
know
the
of
200
design,
but
it
doesn't
necessarily
like
in
this
example.
A
No-
and
this
is
actually
something
we
worked
with-
the
Red
Hat
team
on
and
and
the
the
way
we
ended
up
settling
for
it
for
noun.
So
we
come
up
with
a
better
API
is
have
a
job
as
part
of
a
step
in
here,
and
that
that
does
that
and
then
assert
the
status
of
that
job
is
complete.
So
you
can
do
that
just
through
another
mechanism,
nice.
A
Things
move
around
on
me,
but
but
we
have
some,
you
know
we
set
the
manifest
and
you
can
have
some
some
precondition
commands
to
set
up
your
controller
and
it
shuts
everything
down.
You
can
even
set
up
kind.
You
can,
you
can
do
a
whole
bunch
of
stuff,
so
that's
a
cuddle
about
dev
where
we're
on
Monday
or
Tuesday,
it's
being
presented
at
sig
TAS,
the
next
big
test
meeting.
A
E
A
E
E
E
There
we
go.
The
only
thing
that
I
did
to
this
to
this.
Cluster
is
I,
just
ran
cube,
CTO
to
toe
in
it
and
that
installed
the
requisite
CR
G's,
and
that
Jared
was
discussing
before
just
to
get
the
thing
going,
and
then
you
can
install
the
Kudo
sub-command
with
with
crew
or
brew.
There's
a
there's,
a
huge
it's.
It's
really
not
very
complicated.
If
you
just
go
to
kudo
dev,
there's
a
bunch
of
avenues
to
actually
get
it.
E
The
fairly
small
cluster
in
Amazon,
with
just
a
single
master,
so
OOP
CTO,
Kudo,
install
zookeeper.
So
if
I
run
that
so
I
created
I
created
the
instance-
and
you
can
see
in
the
top
panel
that
zookeeper
so
Jared
was
discussing
earlier,
the
notion
of
bundling
the
logic,
the
the
application
specific
domain,
specific
logic
with
the
application
itself.
E
E
Install
Kafka,
which
has
zookeeper
as
a
dependency,
so
we've
got
our
zookeeper
cluster
app.
Now
we're
gonna
install
Kafka,
alright,
and
so
you
can
see
in
the
case
of
Kafka
the
order
in
which
the
pods
that
comprise
the
nodes
of
the
Kafka
cluster
that
that
matters,
so
the
first
one
has
to
come
up
it
has
to
register,
is
healthy.
Oh
no!
Something
failed,
I!
Think
because
I
left
a
PVC
open
my
bad,
but
this
is
a
good
example
of
how
keeping
things
kubernetes
native
makes
a
big
difference.
So,
let's
say
I
wanted
to
just
delete.
E
Kafka
we've
taught
by
virtue
the
CRVs
we've
taught
the
kubernetes
cluster.
The
notion
of
what
a
an
instance
is
so
here
I
can
just
do
not
cube.
Ctl
Kudo
delete
I
can
just
do
cube.
Ctl
delete,
but
anyways
I
mean
I'm,
pretty
sure
it's
because
I
had
some
PVCs
hanging
out
there,
but
anyways.
That's
you.
Could
you
know
you
could
apply
the
same
logic
to
something
by
Cassandra
and
it'll
go
ahead
and
deploy
Cassandra.
E
A
One
thing
to
know
here,
too,
you
know
Cassandra
is
spinning
up
serially
right,
so
if
you
were
to
approach
home
or
other
tooling,
these
would
all
be
deployed
and
you'd
be
seeing
for
actually
backups
at
office
until
everything
came
together
and
that
that
works
until
a
point.
But
if
you
need
to
start
doing
things
like
like
run,
preconditions
like
okay
I
need
to
create
a
CA.
You
can
self
sign
CA
before
this
this
launches
and
then
pipe
that
into
these
containers
and
and
really
start
to
do
orchestration.
A
We
have
to
have
to
be
fighting
through
a
massive
in
it
containers
its
I
need
to
call
an
API
once
this
has
launched
to
added
to
them
as
a
cluster
member,
before
it's
ready
and
and
so
we
were
really
optimized
for
things
like
Cassandra
Kafka
zookeeper.
All
of
these
complicated
stateful
services
that
would
be
otherwise
fairly
difficult
to
run.
A
E
D
B
A
nice,
alright,
ok,
I
boy
I'm,
just
you
know:
I
got
all
excited
about
cuddling,
I,
just
didn't
even
know
so
by
the
way
I'm
Jared.
Just
to
recap
the
I
know
it's
a
sidenote
sort
of
been
the
longer
conversation
about
Kudo
and
Michael
for
the
for
the
uninitiated,
and
you
know
I,
you
say
what
you
will,
but
for
those
looking
for
a
general
guidance,
other
operator
frameworks
or
and
or
I
think
I
just
said,
the
name
of
one
sort
of
as
I
am
asking
this
question.
I.
B
Think
I
highlighted
within
the
discussion
today
has
been
something
like
I
said:
some
interesting
design
objectives
that
were
significantly
differentiated
from
I
think
some
of
the
other
existing
approaches
to
creating
controllers,
to
create
operators,
and,
at
least
in
my
mind,
I
mean
and
I'd
love
to
hear
what
you
guys
sort
of
recap.
Those
highlights,
but
in
my
mind
that
was
one
of
those.
Was
this
the
notion
of
something
happen
in
our
operator
interface
or
a
uniform
way
to,
or
rather
I
think
it
was
a.
B
It
was
an
identification
and
understanding
that
I'm
hanging
Cooper
did
his
native
administration
is
really
kind
of
nice
when
you
using
sort
of
the
same
style
of
interfacing
with
the
system,
hence
in
part
part
of
the
popularity
of
custom
resources,
and
that
it
is
more
and
more
of
these
operators
and
these
custom
resources
come
about
I'm
having
some
uniformity,
by
which
you
know
the
user
experienced
by
which
you
an
operator
I.
He
a
person
interfaces
with
these
custom
resources,
something
like
that
was
one
of
those.
A
I,
wouldn't
necessarily
it's
one
of
those
things
is
a
a
couple
of
us
on
the
Kudo
team
are
also
chairs
on
operator
of
framework
and
SDK
and
and
and
so
one
of
the
goals
there
is.
Is
you
know
we,
you
know
we're
not
the
end-all,
be-all
tool,
we're
looking
to
build
kudos,
an
orchestration
to
it
right
like,
and
so,
if
you,
if
you,
if
you
were
to
look
at
kubernetes,
maybe
as
a
type
system,
you
know
we,
we
want
to
make
sure
that
gonna
Q
builder
or
operate
SDK.
A
Those
tools
like
creating
the
new
types
in
a
lot
of
cases,
and
then
you
know
kudos
building
the
modules
that
that
orchestrate
all
those
things
together
and
that
that
thing
is
very
much
in
its
like
infancy
and,
like
that's
a
very
like
nascent
concept.
It's
a
Koi,
the
kubarz
operator
interface
is
is
starting.
This
try
to
stab
at
how
do
we
make
all
this
tooling
work
together?
A
How
does
one
operator
consume
another
operator
so
that
you
can
start
to
build
out
this
ecosystem
and
on
the
flip
side,
what
are
the
behaviors
that
may
operate
or
expose
is
so
that
someone
can
write
tooling
to
consume
those
behaviors
and
that
that's
really
the
motivation?
And
it's
not
a
kudos
specific
thing.
We're
just
we're
just
really
trying
to
begin
because
they,
like
the
CUDA
projects,
is
all
about
that
sort
of
orchestration.
Koi
is
about
you
know:
building
the
next
steps
of
the
ecosystem
with
everybody
else.
A
D
E
Of
people,
you
know
it's
funny
a
lot
of
a
lot
of
the
projects
that
I
work
on
that
when
you
get
to
a
certain
size,
call
it
like
5,000,
10,000,
15,000
core
clusters.
It
means
you're,
running,
really
big
applications
and
then
everyone's.
Then
then
you
know
you
introduce
a
like
Kudo
and
they're
like
okay.
This
makes
a
lot
of
sense
of
like
wow.
E
This
is
such
a
good
idea,
and
it's
like
you
know
when,
when
when
Apple
gives
you
an
application,
they
don't
give
you
or
I,
don't
know
if
we
don't
pass
around
like
dot
dot,
py
files
to
each
other
right.
Software
is
packaged
right.
It
has.
It
has
certain
common
interfaces
that
we've
come
to
expect
as
as
computer
users,
so
like
in
the
world
of
kubernetes.
Naturally,
there
would
be
something
like
that
as
well
right.
B
B
A
And
and
we're
I
mean
even
if
it
nobody
has
questions
right
now,
please
feel
free
to
drop
by
Kudo
and
the
Korean
slack
drop
by
CNCs
a
gap
delivery,
slack
operator
working
group
like
we
are
heavily
active
there
and-
and
you
know,
where
we're
working
on
honing
all
these
thoughts
with
a
group
of
people
right
like
like
this,
isn't
just
you
know
me
or
a
group
of
us
being
crackpots
and
try
and
tell
everyone
what
to
do.
But
this
is
like
feedback
on
like
real
pain
points.
A
You
know
there
are
real
problems
as
the
sort
of
store
data
stores,
another
truly
moves
through
their
entire
application
lifecycle
and-
and
we
think
heavily
about
that-
and
we,
but
we
don't
know
everything
right
and
so
the
more
activity
we
have
in
that
the
better
we
can
make
decisions
that
enable
and
make
operate,
develop
more
accessible
for
everyone
else,
and
that
that's
really,
you
know
the
message
that
I
want
to
get
out
there
is
you
know
it's
fun,
you
know
we.
We
want
as
many
people
doing
this
as
possible
and
there's
our
great
question.
A
What's
significant
turn
declarative
if
you
think
what
declarative
versus
imperative
imperative
is,
is
list
taking
a
list
of
steps
of
how
I
want
something
to
happen?
How
I
want
to
achieve
a
goal?
Declarative
is
enemies,
because
this
is
a
very
good
point
about
Kudo
Kudo
takes
imperative
steps
and
and
enables
people
to
wrap
it
and
declarative
shell,
because
declarative
is
all
about
what
I
want
to
happen.
A
Tell
the
system
that
I
want
to
assert
this
state
system
figure
out
how
to
do
the
thing
for
me,
but
you
can't
do
that
without
an
imperative
thing,
and
so
so
it
could.
It
was
still
imperative,
rapid,
a
declarative
shell
to
enable
users
without
having
to
go
through
the
escape
hatch
to
to
have
an
opinionated
declarative
approach
to
operating
in
their
system
right.
Even
if
the
developer
has
to
write
some
imperative
things
to
get
there.
D
A
B
Fair
enough,
well
I
almost
hesitate
to
ask
this,
but
but
Wow
when,
but
but
in
person
a
lot
of
times,
hey
we'll
get
a
break,
we'll
go,
get
it
grab
a
beer.
We'll
go,
you
know,
go
hit.
The
restroom
I
figured
I'd,
ask
there's
anybody.
We
could
take
a
quick
vote,
anyone
looking
for
a
5-minute,
restroom
break
or
or
you
just
ready
to
move
on
I.
B
B
B
It
is,
as
a
side
note,
actually
there's
a
there's.
This
is
the
project
site
is
out
at
mesh.
Redial
there's
also
measure
dev
that
was
originally
registered,
and
but
it
didn't
just
it
was
a
it
was
like.
Kudo
is
the
only
other
project
out
there
that
seemed
to
be
using
the
dot
Desa
just
kind
of
had
the
interest
thing
know
that
that
was
like
part
of
the
thinking
about
what
went
where
to
get
a
project
page
and
how
to
get
it
up
and
going
reason.
B
I
wanted
to
bring
up
this
project
page
is
because
I
wanted
to
highlight
the
long-standing
effort
of
an
any
number
of
contributors
that
have
come
behind
the
project
as
a
matter
of
fact,
I'm
pleased
that
there
we've
got
a
couple
of
the
contributors
joining
us
this
evening
that
are
that
are
in
the
call.
Some
of
them
may
even
be
pictured
here
on
this
community
page,
so
just
a
very
much
soaked
community
first
project,
stewarded
by
layer,
five
and
stewarded
by
the
people
that
you
see
here.
B
B
B
If
you
will
with
one
on
service
mesh
patterns,
so
there's
so
we
will
see
how
that
goes,
we'll
see
if
we'll
see,
if
I'm
still
married
by
the
Internet
honestly
I
I
grabbed
a
hodgepodge
of
slides
that
I
wanted
to
explore
with
all
of
you
and
when
I
say
explore.
That
means
please
interrupt
me
with
thoughts.
This
is
a
not
a
webinar.
This
is
this.
Is
this
really
is
our
virtual
meetup
so
argue
and
ask
questions
and
and
learn
with
me
as
I?
B
B
B
Consider
that
the
value
proposition
of
faz
compute
of
serverless,
if
you
will
of
functions,
really
pretty
similar
to
service
mesh
the
service,
my
value
proposition,
but
that
one
is
for
long-lived
things
and
the
other
one
is
for
short-lived
event-driven
things
so
short-lived
event-driven
things
that
are
done
in
the
functions
but
sort
of
the
rest
of
the
world.
The
80%
of
the
world's
workloads
that
are
long-lived
process,
driven
sort
of
always
hot
I'm
service,
meshes
tend
to
address
those
services.
It's
just
that
there's
often
a
lot
more
pain
over
there
that
they
can
solve.
B
B
Maybe
this
is
I,
don't
know
if
this
is
the
last
concept
that
I
was
going
to
try
to
convey,
and
that
is
sort
of
the
importance
of
a
mesh
and
the
decoupling
that
happens
with
the
mesh
decoupling
between
developers
and
operators.
The
notion
that
service
meshes
are
empowering
of
both
of
those
personas
in
big
ways.
I
think
it's
pretty
obvious
to
those
that
take
a
look
at
a
service
mash.
B
If
you
want
to
know
what
I
would
have
mesh
dozen
specific
on
the
topic
of
service
patches,
there
are
quite
a
few
of
them
out
there.
It
is
messy
out
there.
It's
about.
20
of
them
are
more
that
we're
tracking
on
the
surface
mesh
landscape.
This
landscape
is
community
curated,
so
each
of
the
meshes
each
of
the
technologies
that
are
listed
or
have
been
almost
all
of
them,
have
been
added
by
the
stewards
of
either
that
open
source
project
or
members
from
the
company
that
has
that
product.
B
B
Certainly
some
surface
meshes
enjoy
much
more
popularity
than
others,
but
the
fact
of
the
matter
is
today
it's
a
multi
mesh
world
and,
to
the
extent
that
it
will
be
a
multi,
lash
world
or
it'll,
be
a
multi
mesh
world
for
a
while,
at
least,
if
not
sort
of
perpetually.
There
are
many
reasons
for
that
and
I.
Think
I
didn't
include
the
slide.
B
We
make
it
sort
of
embarrassing
for
people
to
raise
their
hand
and
say
well,
you
know
actually
I
run
nomad
and
actually
maybe
that
running
Nomad
very
well,
because
they
they
have
windows-based
infrastructure
and
they're
running
batch
jobs.
Nomad
does
great
for
that.
So
my
point
is:
there's
a
bunch
of
meshes
out
there,
they're
not
all
made
equal.
Some
of
them
are
a
lot
harder
to
use,
love
more
a
lot
easier
to
use,
much
more,
capable
and
powerful,
but
because
of
that
phenomenon
there
have
begun
to
become
abstractions.
B
Interface
is
abstracting
away
how
it
is
that
you
might
interact
and
integrate
a
service
mesh.
One
of
those
is
the
surface
mesh
interface.
The
project
that
was
I'm
speaking
about
tonight,
measure
E
and
the
community
layer
v
were
both
launch
partners
with
SMI
during
its
announcement
at
last
year's
cop-con
in
Barcelona,
SMI
I,
just
I
just
got
done,
approving
its
entrance
into
the
scene
cf
three
weeks
ago,
and
so,
if
you're
not
familiar
with
it,
it
is
you
know
too
quickly
to
tell
you
about
it.
B
It
is,
it
would
be
appropriate
to
think
of
SMI
as
being
analogous
to
c,
and
I
container
network
interface
or
two
CS.
I
container
storage,
interface
and
I'm,
not
sure
and
in
some
respects.
Part
of
that
means
that
there's
SMI
has
a
bit
of
a
lowest
common
denominator
specification
for
saying,
hey
the
B's.
B
You
know
here's
an
interface
that
be
hot,
you
can
use,
maybe
it's
to
get
metrics
out
of
a
service
match
that
you
can
use
SM
eyes,
interface,
to
get
those
metrics
and
behind
that
interface,
maybe
plugged
in
any
any
number
of
different
service
messages,
and
it
wouldn't
matter.
The
interface
that
you
would
integrate
with
would
be
the
same,
and
so
this
project
is
bound
to
the
constraints
of
the
cluster
of
the
kubernetes
cluster,
so
it
not
intending
at
least
right
now
to
address
things
outside
of
the
cluster.
B
Really
much
more
anyway,
it's
a
project
called
Hamlet,
it's
suited
by
VMware
currently
and
is
about
Federation
of
meshes.
It's
really
about
Service
Catalog
exchange,
the
exchanging
of
services
and
their
information
from
one
service
mesh
to
the
next,
irrespective
of
whether
or
not
that
service
mash
is
of
the
same
type,
irrespective
of
whether
or
not
it's
a
homogeneous
or
heterogeneous
deployment
of
service
wishes.
B
B
And
overhead
of
a
mesh
more
than
just
the
performance
and
overhead,
it's
a
specification
to
help.
People
understand
the
overhead
that
their
service
mesh,
that
running
services
on
that
service
mesh
incurs,
but
to
really
do
it
in
context
of
the
value
that
the
service
mesh
is
providing.
So
if
you
think
about
getting
additional
telemetry
getting
fine-grained
traffic
control
over
your
services,
getting
some
additional
identity
and
security
authorizations-
and
you
know
all
the
things
that
the
service
mesh
provides.
B
It
becomes
much
easier
to
measure
that
overhead
in
a
single
place,
and
so
this
specification
is
about
well
I,
guess
I
would
say
if
you're
familiar
with
app
decks
from
ages
ago,
if
you've
ever
been
into
application,
performance,
monitoring,
there's
an
index
and
it's
called
app
decks.
That
will
help
you
just
give
you
a
very
simple
scale
to
measure
how
well
your
application
is
performing,
and
that's
what
this
service
measure
performance
specification
intends
to
do
is
create
a
mesh
decks
so
that
you
can
very
easily
say
hey.
B
My
service
mesh
is
running
at
an
84
on
a
scale
of
1
to
100
and
that
there
would
be
a
formula
behind
that
and
as
being
able
to
speak
to
that
easily,
you
could
also
then
compare
over
time.
How
well
you're
doing
you
know
whether
or
not
you're
improving
in
terms
of
your
overhead
or
not,
and
also
compare
across
other
deployments
of
meshes
that
other
people
other
organizations
are
running.
B
This
is
just
a
quick
discreet,
so
this
is
a
quick
set
of
performance
tests
that
measure
II
has
run
and
kind
of
what
they
look
like
this.
These
slides
are
quite
this
data
is
quite
old.
Moreover,
not
only
is
the
data
old,
but
in
this
case
sto
that
looks
like
it's
much
more
of
a
resource
hog
than
the
others
is
in
fact
doing
so
much
more
than
the
other
ones
are
doing.
So
it's
not
an
apples
to
apples
comparison.
B
B
I'm
gonna
keep
moving
briskly
because
I
want
to
get
through
this
I
want
to
play
around
with
a
demo
and
see
if
we
can
do
some
new
things
that
I've
never
never
shown
it
about
the
project.
So
the
service
mission
is
in
their
architecture.
I've
said
many
a
time
that
if
the
service
mesh
project
doesn't
have
a
data
plane
and
and
a
control
plane,
it's
not
a
service
mesh
like
those
two
things
are:
what
comprise
a
service
mesh
in
this
architecture,
you're,
seeing
kind
of
some
optional
components,
I'm
an
ingress
gateway,
an
egress
gateway.
B
Some
meshes
have
them
some,
don't
there's.
If
you
are
born
of,
or
have
been
a
network
administrator
network
engineer,
you're
familiar
with
the
concept
of
a
management
plane,
which
is
essentially
what
mystery
is
and
for
some
that
it
can.
That
can
take
many
different
forms.
You
you
will
see.
There
are
some
other
management
planes
there's
some
management
planes
available
today
for
service
meshes,
and
there
are
different
ways
of
thinking
about
what
characteristics
and
management
plane
would
have.
B
B
Difference
between
something
like
a
management
plane
and
something
like
an
Orchestrator,
the
orchestrators
tend
to
be
fairly
talking
about
a
lot
of
the
same
mountains.
A
lot
of
the
same
types
of
infrastructure
objects,
maybe
just
orchestrating
them
higher
level,
whereas
management
planes
are
probably
a
bit
more
about
integrating
with
your
back-end
business
flows.
Your
business
logic
may
be
a
good
way
of
segregating
some
of
these
layers,
maybe
much
to
debate
within
there,
but
so
of
the
architectures
of
the
meshes
that
are
available
today.
B
B
The
way
in
which
sto
1.5
is
able
to
be
deployed
is
in
a
is
also
in
turn.
It
can
also
take
these
individual
control,
plane
components
and
can
bundle
them
into
a
single
binary.
I'm
gonna
have
most
of
them
in
the
same
binary
as
an
optional
way
of
deploying
a
couple
of
different
models
that
are
supported
in
part
or
not
in
part.
B
B
B
But
this
the
notion
that
there's
also
some
things
that
are
interesting,
that,
like
there's
different
deployment
models
for
service
missions,
one
way
to
deploy
sto
is
really
just
deploying
its
ingress
gateway
and
using
that
as
a
starting
point
to
moving
into
full-blown
F,
which
is
nice.
What
time
is
it
yeah?
And
then
you
have
question
of
the
slides
being
available
after
the
slides
be
available
after
the
Meetup
and
one
quick
point
on
that.
Just
since
it's
a
it
is
that
I
generally
put
those
slides
here.
I
was
sort
of
jet
to
general
locations.
B
B
Different
service
mesh
deployment
models
ingress
easy
way
to
start
just
you
know,
sort
of
steal
a
la
carte,
if
you
will
I,
could
borrow
from
Dan's
to
release
the
steal
product
manager.
Another
way
to
deploy
service
meshes
that
sort
of
fell
out
of
popularity
and
then
sort
of
came
back
in
popularity
and
so
I
guess
I
haven't
done
a
count
of
which
of
how
many
service
meshes
go.
The
route
of
using
of
side
carring
proxies
versus
how
many
service
meshes
go.
The
route
of
you
know
deploying
a
proxy
per
node,
but
this
was
linked.
B
Adiz,
linker,
DV
ones,
architecture,
linker,
DV,
formerly
called
conduit,
had
then
moved
over
to
the
side,
car
or
the
propeller
sidecar
proxy
model.
Some
other
meshes
have
come
around
and
used
the
proxy
per
node
model
pros
and
cons
to
both,
and
which
is
camera,
which
is
back
to
the
kind
of
my
prior
point
about
being
pleased
that
we
still
have
choice
when
it
comes
to
meshes
because
there's
a
bunch
of
different
use
cases
a
bunch
of
different
organizations.
B
This
book
was
one
that
I'd
written
some
time
ago
and
O'reilly
recently
asked
for
a
refresh
on
it,
and
so
so
I
think
I'm
gonna
get
pregnant
with
twins.
We're
gonna
refresh
that
book,
which
should
be
easier
to
do,
but
that
book
is
a
free
book.
You
can
go
download
and
but
talks
about
these
different
models
to
the
extent
that
people
are
interested
in
so
I
think
this
last
slide.
B
One
of
the
things
I
wanted
a
demo
mystery
to
folks
tonight,
and
one
of
the
things
I
wanted
to
do
was
show
a
bit
of
multi
cluster
capabilities
that
mesh
Ria's
manager
do
measure
e
is:
is
a
multi
mesh
manager
as
an
open
source
project.
It
achieves
the
core
infrastructure
initiative
for
open
source
best
practices,
which
is
great.
It
was
a
google
Summer
of
Code
project
last
year
and
is
again
this
coming
year.
2020
headed
likely
headed
into
sandbox
first
half
of
this
year
fairly
shortly.
Is
s
and
is
an
SMI
implementation?
B
Messy
itself
has
enjoyed
contributions
from
a
number
of
companies
that
want
their
services
were
represented
in
this
project,
and
so
these
it
right
now
mystery
supports
six
different
types
of
service
meshes
I'm,
the
ones
that
are
highlighted
in
yellow
are
adapters.
You
know
measuring
adapters
for
that
specific
service
mesh
that
were
written
by
those
companies
or
those
open
source
projects.
B
Those
teams
make
sure
that
they're
they've
got
performant
service
meshes
and
to
be
able
to
compare
where
possible
so
measure
ease
architecture
is
these
runs
in
the
presence
of
a
service
mesh.
It
does
like
lifecycle
management
of
a
service
mash,
but
it
comes
in
as
a
management,
plane
and
so
you'll
do
it
deploys
as
a
set
of
containers
that
you
can
deploy
outside
the
cluster
or
inside
the
cluster.
There's
reasons
to
do
either
it
will
spin
up
it.
Does
lifecycle
management
event
of
those
six
service
meshes?
B
It
does
some
performance
management
which
we
kind
of
talked
about,
and
then
it
also
does
some
configuration
management,
so
it'll
analyze
your
sto
config,
for
example,
and
tell
you
you're
doing
it
right,
you're,
doing
it
wrong
and
so
different
ways
to
deploy
massery
there
are
you
use
brew
or
bash
or
you?
It
runs
on
Windows
and
Linux
and
Mac
it.
B
The
community
just
got
done,
creating
a
scoop
package,
a
package
for
scoop
so
which
is
kind
of
nice.
I,
don't
run
Windows
myself
and
so
I
haven't
tried
it,
but,
but
we
are
what
we
are
going
to
try
tonight
is
hopefully
a
little
bit
of
multi
cluster
things
that
I
was.
There
was
a
user
in
the
community
from
Ericsson
that
was
asking
about
the
ability
from
essary
to
interface,
with
multiple
instances
to
do
some.
Multi
cluster
things
Multi
mesh
things.
B
Clearly
it
does
multi
mesh
things
that
rather
multi
clusters.
What
I
was
gonna
say
and
what
I
was
trying
to
to
to
tell
this
individual
was
battery.
Does
do
multi
cluster
things,
it
just
isn't
the
user
experience
that
the
project
would
want
to
have
right
now,
so
I'm
going
to
so
where
it
said
we're
at
7:25,
I
think
we'll
do
and
15
minutes
of
demo
and
say
and
see
if
I
can't
play
around
with
the
project
and
so
Jeff
a
fantastic
question.
B
F
Yeah
I'm
just
curious
stuff,
I've
kind
of
I've
been
following
you
guys
since
your
inception,
and
you
know
I've
learned
secretly
in
the
slack
and
watch
things
that
are
going
on,
but
you
know
just
trying
to
get
a
single
surface.
A
nice
not
boarded.
You
know
a
telco
which
were
unique,
beast
I
get
it
can
sometimes
be
a
challenge
and
so
managing
multiple
service
meshes.
You
know
sometimes
I
debate
like
if
the
complexity
is
worth
it.
So
you
know
how
does
that
complexity
get
mitigated
by
this
and
then
a
real
value
add?
F
Is
that
question
around
federated
clusters
right
if
individual
consumers
want
their
surface
mesh
with
in
there,
you
know
unicorn
cluster,
that
I
give
them,
but
I
need
to
potentially
expand.
Those
you
know
is
that
something
that
make
sure
he
is
looking
at
long
term.
You
know
that
gets
into
some
control
plane
stuff.
Admittedly,
when
you
start,
you
know
pointing
ingress,
is
at
each
other
dresses
to
he
dresses,
etc.
But
I'm
just
curious
is
that,
like
part,
the
long-term
vision
it.
B
Is
I
was
talking
with
well
I,
believe
we've
been
any
conversation
with
with
well
with
VMware,
specifically
and
with
Hoshi
Corp.
Secondarily,
on
this
particular
topic
and
there's
a
you
know,
a
lesser-known
service
mess
at
service
mesh
abstraction
called
Hamlet
that
was
sort
of
put
out
by
VMware
this
middle
of
this
last
summer.
I
think,
and
it
is
that
specification
is
about
facilitating
Service
Catalog
exchange
from
what
about
federating
meshes
in
the
way
in
which
is
about
interoperability
of
service
of
workloads
running
on
those
meshes
to
me,
that's
a
much
more
interesting.
B
You
just
needed
me
personally,
a
much
more
interesting
use
case
to
chase
after
and
help
enable
people
with
than
it
is
the
lowest
common
denominator
goal
of
an
SMI
and
that's
not
to
cast
any
shadows
on
s
mind
in
any
respect.
Mastery
is,
and
we
were
saying
earlier,
an
implementation
of
it,
a
partner
of
SMI,
but
so
in
a
long-winded
answer.
B
I,
there's
probably
multiple
things
about
what
you're
thinking
about
in
federating
service
meshes,
but
one
of
them
in
part
what
your
the
mechanics
of
doing
so
about,
like
gateway
to
gateway,
things
about
certificate
or
other
domain
of
trust
and
certificates
that
are
where
they
are
signed
and
and
where
they
are
trusted.
That
administrative,
like
outside
the
organizational
psychology
that
comes
around.
B
F
Yeah,
you
actually
hit
the
big
one
way
more
than
just
you
know.
Linking
gateways
is
specifically
multiple
clusters,
understanding
them
in
TLS
and
getting
at
zero
trust.
When
I'm
potentially
spamming,
you
know
diverse
workloads,
it's
like
you
said
when
you
know
a
moaner
a
has
their
cluster
and
then
you
interface
with
that
learner,
be
who
you
know
maintains
it
controls
their
cluster.
Setting
up.
You
know
smart
networking
and
some
type
instead
of
trust
model.
Is
that
be
very
compelling
use
case
you.
B
B
They
may
have
just
hooked
up
and
begun
using
that
service
and
I'm,
not
talking
about
in
context
of
you
know,
different
different
service
meshes,
but
rather
just
understanding
that
that's
a
phenomenon
and
needing
to
be
able
to
provide
people
control
around
that.
Then
it
gets
even
more
complex
in
context
of
multiple
meshes
and
so
yeah.
B
So
this
is
unrehearsed
only
as
of
30
minutes
ago.
So
will
this
will
be
quite
fresh
measure
e
itself,
I'm
on
OS
X.
If
she
can't
tell
by
various
things
on
my
desktop,
that's
reinstall
mystery
has
a
CTL
tool
like
any
other
projects
that
you
might
be
familiar
with.
When
you
go
to
start
measuring,
it
will
spin
up
the
measure
a
the
server.
That's
you
know
a
container
and
then
also
any
number
of
its
of
its
adapters,
and
in
doing
so,
when
those
come
up,
it
will
as
a
project.
B
This
demonstration
wasn't
going
to
go
so
far
as
to
get
into
because
these
are
roadmap
items
from
area
around
route,
around
certificate
authorities
and
the
different
deployment
models
that
there
are,
but
I'd
wanted
to
show
that
Mary
is
capable
and
sort
of
headed
that
way
so
currently
in
in
this
environment.
If
we
do,
we
take
a
look
at
the
kind
of
the
configuration
that
master
is
running
under
at
the
moment.
Is
the
notion
that
there's
a
handful
of
containers
a
couple
of
sto
adapters?
One
is
the
measure
used
to
adapter
another
one?
B
B
B
Turn
that
off
so
so
here's
my
Shuri
mesh,
we
will
well
unless
you
I'm,
connect
to
your
different
on
kubernetes
environments.
So
if
we,
if
we
go
here,
we
can
go
load
up
our
kubernetes
config
when
we
do
and
we'll
try
to
identify
you're
in
your
cube
config,
it
will
try
to
identify
your
currently
run
or
your
your
current
context,
which
in
this
case
is
docker
desktop
great.
B
With
that
connection,
you
can
see
some
details
about
how
it's
connecting
to
kubernetes.
You
can
then
do
a
quick
sort
of
a
ping
test
just
to
make
sure
mastery
is
in
connection
with
kubernetes
you're,
also
able
to
do
that
same
kind
of
a
ping
test
between
measure
E
and
its
containers.
So
this
isn't
measuring
pinging
an
instance
of
console,
but
just
naturally
communicating
with
its
own
console
adapter,
there's
only
adapter
for
console.
B
So
we
can
see
that
there
are
two
sto
adapters
in
this
environment:
one
running
on
port
10,000,
the
other
one
running
up,
running
up
port
ten
thousand
ten
and
and
as
such,
they're
kind
of
split
out
over
here
I
think.
We
just
said
that
maybe
we'll
use
this
adapter
for
docker
desktop
this
adapter
for
mini
cube,
so
I
think
we
just
said
and
docker
desktop.
There
was
he
steel
was
already
running
and
we
wanted
to
probably
get
rid
of
that.
B
Bear
with
me.
If
there's
any
amount
of
slowness,
because
I'm
running
multiple
kubernetes
clusters
with
mono
will
be
running,
multiple
ISTE
O's
with
maybe
multiple
sample
apps,
and
so
there
may
or
may
not
be
some
memory,
slowness.
What
we
will
see.
Okay,
so
I
think
it
looks
like
over
here.
We
also
have
book
info
as
a
sample
app
also
deployed,
so
we're
going
to
remove
that
application
off
this
cluster
as
well
and
kind
of
start
fresh
between
the
two.
B
So
we
got
sir
fresh
clusters
as
those
get
fully
deleted,
fully
removed,
I
believe
sto
is
fully
removed.
So
let's
go
ahead
and
install
the
latest
version
of
sto
with
MPLS,
and
we
see
that
coming
up
over
here
could
also
go
Mestre,
as
a
project
is
very
shortly.
It'll
have
much
better,
so
real-time
feedback
about.
B
It's
sort
of
that
the
community
is
building
out
kind
of
mesh
sync,
if
you
will
so
it'll,
be
a
bit
more
fluid
and
by
the
way
I
couldn't
par.
The
reason
I
started
out
the
the
introduction
of
nursery
tonight
and
with
the
highlighting
the
members
of
the
community
is
there's
one
because
I
just
I
will
say
thanks
again
to
those
folks
who
are
spending
their
time
here.
Some
of
them
are
on
the
call
I'm
here
tonight.
B
I
don't
want
to
embarrass
anyone,
but
so
ISTE
was
now
deployed
great,
let's
go
ahead
and
deploy
book
info
and
as
part
of
getting
that
sample
app
going.
Let's
also
apply
a
config
to
a
book
info
to
say
that
we'll
just
will
maintain
that
all
traffic
be
routed
to
the
v1,
the
v1
version,
one
of
the
reviews
service
inside
of
book
info.
B
B
B
And
so
here's
product
page
I'm,
sorry
this
application
book
info
and
what
we're
seeing
here
is
some
errors
on
the
page
and
that's
because
of
the
rule
that
we
just
applied.
That
said,
hey
you
can
only
go
to
v1
a
booking
phone.
So
that's
not
super
interesting,
but
what
I
want
to
do
is
go
over
and
have
necessary,
deploy
to
to
the
other
environment,
to
mini
queue
and
confirm
that
we
and
also
deployed
booking
phone
and
confirm
that
not
sure
he's
able
to
manage
the
two
of
config.
B
B
Where
Mestre
can
stand
needs
a
slight
bit
of
an
enhancement
to
make
part
of
this
a
little
easier
that
you
should
be
able
to
just
come
back
here
in
the
UI
and
switch
between
the
contexts
hasn't
the
easy
way
of
doing
that,
and
so,
if
there's
a
little
bit
of
enhancement,
that
needs
to
be
done,
but
as
we're
doing,
qcg
I'll
get
pods
now
we're
doing
it
from
the
kubernetes
environment.
I'm.
Sorry,
the
mini
cube
environment
yeah.
We
switch
its
context.
We're
over
here.
B
B
B
Roadmap,
there's
a
lot
more
dynamicism
to
be
involved
for
the
type
of
configuration
that
people
might
want
to
apply
from
this
UI.
It's
certainly
my
understanding
that
people
who
are
doing
infrastructure
as
code
are
going
to
want
to
interface
with
their
get
based
systems
when
they
apply
this
type
of
config
in
if
it
hasn't
been
sort
of
pre
reviewed
and
pre
approved
they're
going
to
want
to
do
that
in
their
their
get
flow.
B
So
my
sure
you
will
integrate
there
and
facilitate
both
of
facilitated,
hopefully
the
best
of
both
worlds
of
giving
people
a
visual
interface
in
which
to
understand
what's
going
on
with
their
mesh
and
then
let
them
apply
changes,
but
have
those
changes
reviewed
through
through
get
through
their
normal
computer
review
process?
Stl
is
deployed.
Haley
are.
F
B
I'm
taking
your
note,
but
that's
a
great
suggestion,
jacket
I
mean
is
part
of
that
then
yeah.
So
so,
if
you
had
those
terraform
providers
measure,
you
might
might
do
some
of
the
same
things
that
terraform
enterprise
might
do,
but
do
it
in
a
service
master
specific
way.
Is
that
kind
of
party
thinking
yeah.
F
I
mean
you
are
always
going
to
have
overlap
of
tools
right,
but
I
try
to
bend
the
tool
that
does
certain
things.
The
best
is
let
it
do
that,
but
I
do
a
lot
of
provisioning
on
the
include
side,
with
tariffs,
ones
personally
and
be
super
cool
if,
instead
of
provisioning
all
the
clusters
and
been
coming
in
and
applying
all
the
service
message
service
finishes,
you
know
I
write
the
entire
plan
and
check
it
in
to
get,
as
you
were
saying
so
now,
I
have
a
version,
control
infrastructure
moving
forward.
B
B
B
The
last
thing,
maybe
that
I
would
show
folks
and
and
as
we
near
kind
of
at
the
top
of
our
time,
is
a
quick
performance
test,
maybe
between
the
two-
maybe
not
that
they're
equal
right
now,
because
one
is
enforcing
a
routing
rule
and
the
other
one
isn't
so
right
now
we're
focused
on
mini
queue.
We
go
over
and
run
this
test
and
just
can't
use
the
default
parameters
for
this
test.
B
This
would
run
for
30
seconds
it'll.
Do
an
analysis
of
the
you
know
of
the
load?
That's
been
generated,
calculating
kin
latency
in
a
histogram
fashion,
with
throughput
as
well,
generating
those
the
P
P
P,
P,
99,
etc.
It
also,
then,
provides
the
ability
to
download
those
results
in
that
format
that
we
were
speaking
about
before
the
service
mash
performance
specification
format,
the
one
that
the
the
community
intends
to
create
mesh
decks
that
index
to
determine
your
to
assess
your
overhead
vs.
B
B
B
B
Run
the
test
run
a
performance
test
against
that.
Okay,
though
that
is
our
localhost,
so
I
am
hitting
it
over
local
host
here,
although
with
the
routing
of
the
given
that
measure,
he
is
deployed
on
a
bridge
network
in
my
dock
or
desktop
VM
and
external
to
the
kubernetes
cluster
I'm
gonna
use
my
my
max
wireless,
my
host
IP
address
instead
of
localized.
D
B
B
B
And
so
maybe
I
won't
be
able
to
show
you
guys
a
comparison
of
the
the
the
two
performance
tests,
but
so
so
I'm
gonna
I'm
gonna
play
with
this
for
a
minute
but
I'll,
but
I'll,
ask
as
I'm
typing
here
and
as
I'm
playing
around
other
thoughts
or
questions.
Are
there
other
things
that
people
are
seeing
or
have
questions
about
service
meshes
in
general
places
that
mesh?
We
might
go
things
that
it
might
may
or
may
not
do
Jeff.
You
have
some.
Those
are
some
great
suggestions,
just
very
nice
to
hear
so.
B
B
We'll
give
some
s
of
my
demos,
maybe
another
time
not
at
this
Meetup
and
speaking
of
this
meetup,
who
else
I
made
a
call
shout
out
earlier
for
requests
for
presenters
and
we're
generally
kind
of
always
looking
for
folks
to
raise
their
hand
as
we
as
I
say
that
I'll
acknowledge
that
we
do
have
I,
think
a
few
other
meetups
scheduled
and
kind
of
upcoming.
So
assuming
all
goes
well,
we
would
be
doing
a
linker
D
workshop
in
the
past.
We've
really
done
a
lot
of
workshops
around
his
deal.
B
B
Out
of
curiosity
in
your
environments,
who
doesn't
have
like,
like
in
your
environment,
who
doesn't
know
where
you
would
go
or
who
doesn't
have
a
collection
of
people,
that
you
would
go
to
to
ask
things
about
open
source,
whether
or
not
it's
okay
to
consume
it?
Whether
or
not
it's
okay
to
contribute
elsewhere,
like
what
what
the
rules
are
within
your
organization's.
B
D
Haley,
this
is
Chris
Callen
how's
it
going
I'm
at
IBM
and
I
actually
have
clearance
to
work
on
certain
open-source
things,
but
I
have
to
go
project
by
project.
You
know
and
get
each
one
blessed
whatever.
Let
them
know
them
you
know,
and
then
they
always
want
to
know.
If
there's
any
sort
of
conflicts
with
internal
stuff.
I.
Imagine.
B
B
F
B
I'd
reason
for
myself
personally
and
recently
changed
my
full-time
focus
to
be
essentially
full-time
focused
on
service
meshes
I'm,
trying
to
find
a
good
damn
slide,
because
it's
kind
of
helpful
in
answering
this,
and
that
is
the
part
of
that
consideration.
That
now
is
a
decent
time
is
yeah,
not
even
necessarily
answering
your
question
directly
or
speaking
to
characterizing
it
directly.
One
thing
to
just
look
at
is,
like
is
to
say,
like
if
you
believe
that
service
meshes
as
a
piece
of
to
structure
as
a
cloud
native
capability.
B
If
you
believe
that
they
would
be,
you
know,
generally
ubiquitously
adopted,
and
if
you
look
at
the
the
the
timeframe
by
which
so
the
underlying
cloud
native
infrastructure
layers
have
been
laid
down,
I
kind
of
went
when
those
projects
were
announced
when
they
sort
of
were
production
ready.
If
you,
if
you
were
to
use
their
one
dot,
is
that
that
signal
and
if
you
were
to
I,
think
characterize
this
in
context
of
like
when
these
things
achieved
mainstream
adoption?
B
It
will
be
that,
as
you
try
to
encourage
developers
to
focus
on
business
logic
that
just
means
similar
to
the
value
proposition
of
a
function
that
you
would
that
meshes
would
take
hold
and
it
would
be
generally
adopted
at
some
point
if
that's
true,
they
become
probably
commoditized
infrastructure
and
we
in
the
world
sort
of
climbs
up
to
higher
layers,
which
is
you
know
more
or
less
where
this
project
is
focused,
I,
think
Jeff,
part
of
your
question
was
it
was
hey
like
which
of
those
available
meshes
are
hearing
now
and
ready.
Yeah.
F
And
I
know
that
you
need
to
stay
neutral,
so
I'm
not
trying
to
put
you
on
the
spot.
It's
just
you
know
as
a
service
provider,
I'm
working
the
office
of
the
CTO,
so
I
look
at
lots
and
lots
of
feature
Meucci
technology,
but
when
I
make
recommendations,
I've
spent
a
lot
of
time
with
them.
Miss
do
consult
a
few
others,
but
you
know
uptime
for
service
is
everything
for
me
and
you
know
when
I
make
recommendations
when
I
start
trying
to
introduce
new
technology.
It's
always
on
you
know.
F
The
onus
is
on
me
to
prove
that
I'm
not
inducing
unnecessary
risk
and
that
the
value,
because
I
do
think
you're
right,
I,
think
service
messages
are
going
to
be
just
come
on
ties
you
already
started
to
see.
Like
you
know,
third-party
containers,
a
service
that
vendors
do.
You
know
kubernetes
management
stuff.
You
know
OP
to
click
button
here.
Take
it.
You
know
it's
not
as
broad
scales.
You
know
they
they
partner
with
like
one
and
they
try
to
like
refine
it.
But
I'm.
F
B
B
Swarm
itself
was
built
as
an
extra
capability
external
to
docker
engine,
and
it
was
about
a
year
and
a
half
or
more
in
somewhere
around
in
there
that
it
essentially
rebooted
again
with
a
new
code
base
and
was
brought
into
dr
engine
itself
and
made
an
even
better
user
experience
and
even
more
available
at
people's
fingertips
it
to
regress
in
functionality,
but
over
time
proved
itself
well
in
terms
of
it
doesn't
continue
to
get
the
same
level
of
investment.
I
think
Miranda
is
signing
up
to
do
some
of
that.
B
But
my
point
is
is
like
to
a
long-winded
answer
to
kind
of
characterize
the
notion
there
like
well:
hey,
that's
it
without
me,
saying
anything
explicitly
just
like!
Well,
that's
it
that's
interesting
like,
and
actually
you
see
that
with
a
lot
of
projects
you
see,
that
is,
do
as
another
project
just
didn't
significant
in
process
of
a
fairly
significant
refactoring
of
its
architecture
that
note
I'm
not
telling
you
Jeff
anything
that
you
don't
know,
but,
but
that
you
know
the
the
mixer
this.
B
Those
don't
enjoy
it
quite
as
much
adoption.
Some
of
the
organizations
behind
them
are
pretty
serious
about.
You
know
cause
an
example
of
like:
doesn't
it
console
itself
enjoys
a
ton
of
adoption
for
its
some
service
discovery
and
incrementally
some
adoption
for
its
incremental
into
service
discovery,
incrementally
some
adoption
for
it.
Some
service
mission
capabilities
for
its
connecting
capabilities
of
all
the
interactions
that
I've
had
with
hashey
core
culture.
B
Quite
a
few
you
know,
they're,
pretty
earnest
about,
are
pretty
serious
about
the
long-term
value
that
can
be
provided
through
that
aspect
of
console
to
the
service
mission
aspect,
so
I
hope
that
that
completing
between
the
lines
of
some
of
what
I'm
saying
like
that,
that
I
think
it's
people
like
as
you
go
to
success,
there's
a
number
of
signals
that
are
used.
Some
of
the
signals
are
kind
of
crap,
which
is
like
how
many
get
up
stars.
Does
this
thing
have?
B
Those
are
for
those
people
running
those
technologies.
It's
a
it's
an
evolution
because
they're
already
running
the
the
intelligent
proxies
lesser
risk,
both
in
terms
of
the
projects
that
are
there
and
the
features
that
they're
providing
and
then
people
who
are
adopting
him.
It's
not
a
so
yeah
I
guess
in
part
what
you're
asking
like
in
and
around
Sto
I
guess,
I
considered
as
an
example
like
I.
Consider
that
it
a
good
thing
that
is
Tod
has
that
all
that
effort
went
into
the
feedback
that
was
received
by
people
who
are
trying
to
deploy
it.
B
Engineers
are
attracted
to,
or
from
my
perspective,
attracted
to
complex
projects
wanting
to
engineer
the
nerd
knobs
just
for
doing
that,
but
it's
actually
as
unfair
to
say,
because
Jeff
I
actually
think
that
in
you
were
probably
one
to
say
that
I
would
have
put
some
words
in
your
mouth.
You
tell
me
if
this
is
accurate,
but
as
people
go
to
deploy
a
service
mesh
and
they
as
they
go
to
fit
that
mesh
and
the
way
into
the
way
that
their
networks
operate.
Currently
I.
B
F
I
mean
you've
got
a
unique
insight
via
the
principles
that
I'm
kind
of
my
thoughts
of
the
matter.
I,
don't
disagree
that
engineers
like
to
tackle
cool
technology-
that's
complex.
You
know
tickles
the
deep
breaths
recesses
of
their
brain.
The
struggle,
though-
and
you
know
honestly,
it's
my
number
one
driver
towards
cloud
native-
is
simplification
and
consistency,
and
you
know
a
like.
Let
you
know
you
turned
off
my
videos.
F
F
You
know
code
based
deployments
version
control
and
you
know
the
concepts
of
like
a
neat
ability
and
things
like
that,
and
that's
why
you
know.
Like
the
federated
cluster
thing,
you
have
a
higher
level
construct
like
my
straight
right
like
them
once
again,
and
that's
why
I
asked
about
a
provider
for
interfaces
with
my
serious.
You
know
just
putting
everything
in
a
template
that
you
know
my
operations.
Folks,
get
to
look
at,
you
know,
have
buy-in,
make
pull
requests
against
and
then
once
everybody's
agreement,
you,
let
it
machine,
deploy
it
and
it
looks
the
same.
F
F
Yeah
I
mean,
in
my
opinion,
a
lot
of
the
fun
like
just
in
the
kubernetes
space,
the
general
it
was
you
know
at
people,
you
know
software
dents
in
systems.
People
who,
like
you
know
thinking
I'm
working,
is
cancerous.
They
just
wanted
to
make
that
problem
go
away
and
you
know
there's
a
lot
of
agreement
there,
but
you
know
our
entire
and
he
is
like
it's
entire.
You
know
business
model
at
this
networking,
so
we
care
about
it.
We
need
to
understand
the
underpinnings.
We
need
to
figure
out.
F
How
do
I
take
that
vapor
in
the
egress
gateway
and
expose
them.
You
know
to
GSLV
or
something
so
that
you
know
everybody
within
the
global
network
can
come
in
access
the
services
they
need
any
know
so
I
mean
it's
it's
it's
a
tough
balance
between
me
wanting
to
be
a
purist
when
it
comes
to
cloud
native.
You
know
to
you,
you
know,
look
to
the
Kelsey.
Hightower's
of
the
world
would
be
like
yeah,
that's
really
really
cool,
but
in
reality,
I've
got
to
like
have
a
give
and
take
like
service
providers
are
notoriously.
F
Adamant
about
you
know
their
own
snowflake
way
of
doing
things.
We're
no
exception,
and
you
know
the
draw
here.
Is
you
have
something
like
my
Cherie?
It's
a
drop
down
operations,
folks
love
it
and
they
can
just
deploy
their
mesh.
It's
commoditized
I
can
do
the
work
on
the
backend
is
to
ensure
that
the
mesh
looks
you
know
feels
breeze.
The
way
we
need
it
to,
but
they
don't
have
to
deal
with
that
they
just
they
get
their
helm,
tries
to
get
their
operators
to
get
their
whatever
and
they
hit
go.
B
D
B
The
vision
there
for
just
how
easy
to
make
that
but
yeah
I
agree
with
you
and
actually
in
some
respects
when
people
ask
like
to
your
point
like
what
is
this
service
mesh
in
some
respects
is
sort
of
like
well
well.
Well,
networking
for
developers
or
networking
for
non
network
is
sort
of
I,
don't
know
much
ways
to
maybe
capture
it
in
next-gen
s,
the
end,
no
more
like
I
didn't.
F
B
B
These
are
great,
fair
enough
time.
We've
gone
totally
over
and
you
guys
got
me
talking
about
service
meshes
and
I
and
I
won't
shut
up,
I'm,
very
nice
that
Jeff
for
you
to
come
great
questions,
doing
good
insights,
and
actually
you
characterized
it
I
think
are
you
gave
the
caveat
upfront
about
being
within
a
within
service
provider
world?
And
that
being
you
know
some
of
the
challenges
you
guys
have
being
different
or
many
of
them
being
different
for
enterprises,
but
not
all
of
them.