►
From YouTube: Service Meshes, but at What Cost?
Description
Cloud Native Austin - CNCF Meetup - https://www.meetup.com/Cloud-Native-Austin/
A
Live
stream,
the
meeting
now
brace
yourself
one,
I
think,
you're
all
aware
of
the
fact
that
this
is
probably
going
to
be
recorded.
A
But
what
you
may
not
have
been
aware
of
was
that
now
that
we're
doing
a
youtube
live,
I
mean
the
my
massive
fan
base
is
about
to
pile
in
so
so
yeah,
so
so
so
be
prepared.
Good,
hey!
It's
the
september
edition
of
cloud
native
austin,
thanks
all
for
coming.
If
you've
been
to
one
of
our
meetups
before,
then
you
understand
that
it's
jam-packed
full
of
some
pretty
bad
jokes.
A
It's
also
a
place
where
we
do
things
fairly
informally,
as
well
as
try
to
socialize
a
bit
used
to
be
that
we
would
do
that
with
a
cold
one
or
two
in
our
hands
and
in
person,
and
since
that's
not
happening
now,
we
do
desire
for
this
not
to
be
a
webinar
style
thing.
Rather
this
is
a
meetup,
and
so
a
meetup
is
well
the
meetup's
informal
and
the
reason
that
I'm
here
is
most
one,
because
I
enjoy
sharing
things
that
I
learn,
but
it's
mostly
for
you.
A
It's
mostly
so
that,
through
your
learning
I
selfishly
get
to
learn,
and
so
it's
kind
of
kind
of
kind
of
the
the
theme
here
a
little
bit
about
on
cloud
native
austin.
It
was
well
not.
It
was
not
called
this
previously,
but
it
was,
I
think,
the
first
meetup
in
the
u.s
to
become
a
cncf
meetup
and
since
that
time,
five
years
later,
we've
got
at
least
one
other
in
austin.
A
That's
sponsored
by
the
cncf,
a
lot
of
things
and
technologies
to
talk
about
in
the
cncf,
since
we're
talking
about
the
cncf
actually
earlier
today,
just
four
hours
ago.
A
One
of
the
hats
that
I
wear
is
as
the
cncf
sig
network
chair
and
was
hosting
that
meeting
and
talking
a
lot
about
well
talking
a
lot
about
service
next
service
meshes.
Actually
that
special
interest
group
is
the
home
base
for
all
of
the
network
and
traffic
related
cncf
projects.
So
envoy,
linker
d,
core
dns,
grpc,
there's
a
there's,
a
host
of
them
and
growing
by
the
day.
A
As
a
matter
of
fact,
two
of
the
projects
that
we
will
speak
about
on
this
call
or
in
this
meetup
is,
are
likely
to
become
cncf
projects
within
about
seven
weeks
and
so
highly
appropriate
that
this
is
a
cncf
meetup.
A
A
couple
of
things
that
I
would
say
the
last
few
times
that
we've
met
there
have
been
some
others
speaking,
which
is
good.
There's
also
been
some
talks
from
me
and
that's
out
of
necessity.
Really
this.
This
meetup
is
as
much
yours
as
is
mine,
I'm
I'm
the
the
unfortunate
soul
on
the
other
end
of
like
organizing
and
and
trying
to
get
some
other
geeks
together
to
exchange.
A
But
that
doesn't
mean
it's
my
agenda,
so
I
I
guess,
I'm
trying
to
encourage
you
that,
like
if
you're
spending
time
here
like
this
is
this
is
yours
as
well.
Part
of
that
is
is
is,
is
just
sharing
and
asking
some
questions.
A
One
of
the
ways
that
you
can
share
is
by
is
by
giving
a
talk,
so
you
can
there's
a
simple
google
form
to
fill
in
if
you
think
you
want
to,
if
you've
got
something
to
something
to
say,
sometimes
people
have
want
to
sponsor
as
well
on
that
subject,
there's
other
ways
of
sharing
other
than
talking,
because
it's
kind
of
it
can
be
a
lot
of
prep.
A
The
other
way
of
sharing
is
just
things
like
this.
If
you're,
if
you're
looking
for
a
job
or
maybe
you're
you're
looking
for
some
talent,
this
is
a
good
place
to
exchange.
So
so,
like
so
like
right
now,
there's
a
question
for
you:
if
you're
out
is
anybody
out
there?
Looking
anybody
anybody
been
negatively
affected
by
kobit
19
outside
of
you
know
any
number
of
inconveniences?
A
A
Okay,
then,
the
next
question
here
is
well.
I
guess,
there's
not
a
next
question.
The
other
question
is
hey
who's,
who's,
who's,
seen
something
interesting.
Anybody
know
of
an
upcoming
conference
or
anyone
saw
something
in
the
news
they
thought
was
kind
of
intriguing.
B
Well,
snowflake
came
out
what,
yesterday
or
the
day
before
that
doubled
in
stock
price.
I
don't
know
what
snowflake
was,
but
some
type
of
technology
and
also
I've
noticed
an
influx
of
online
conferences.
This
fall,
I
mean
it
seems
like
end
of
september
october,
there's
just
more
coming
through
my
email
than
I've.
Seen
in
the
last
three
months,
yeah.
A
That
makes
a
lot
of
sense
to
me.
Yeah.
Have
you
curious
about
some
of
the
platforms
that
you've
seen
them
using?
Some
are
probably
fairly
atrocious
and
others
maybe
do
a
decent
job
of
making.
It
feel
like
an
event.
B
Oh
yeah
well
a
meet
up
the
other
night
with
ernest
and
them
use
some
tool.
I
don't
know
what
it
was
had.
B
You
had
some
sticky
pads
or
something
and
you'd
vote,
and
you
do
you
know
a
lean
coffee
and
you
put
stuff
on
sticky
pads.
Look
the
graphics
all
look
great,
but
I
had
no
idea
how.
A
B
Vote
on
anything
or
write
on
anything,
so
I
interacted,
but
I
couldn't
really
I
mean
he
had
like
30
people
on
the
call
and
it
seemed
like
they
only
had
about
eight
to
15
people
voting
or
being
able
to
vote
on.
What's
the
subject
or
you
want
to
continue
to
talk
or
not
in
a
lean
coffee.
One
that
I
really
really
really
really
liked
was
osf
did
three
one
day
conferences
this
summer.
B
I
think
about
two
three
weeks
apart
and
they
use
zoom
interface
and
but
they
also
had
a
some
kind
of.
I
don't
know
what
it
was
called
man
I
didn't
know
you
were
gonna,
ask
me
this,
but
it
was
really
great.
B
It
was
like
a
spreadsheet
that
it
was
no
big
deal
to
you,
click
one
thing
and
everyone
can
edit
the
spreadsheet
at
the
same
time,
and
when
you
click
it,
you
have
a
certain
color,
and
so,
if
you
have
something
you
wanted
to
talk
about
or
add
to,
they
had
a
list
of
what
they
were
going
to
cover
for
that
hour
and
you
could
interact
by
writing
as
the
person
who
is
moderating
the
talk
anyway.
That
was
really
cool,
really
cool.
It
was.
B
C
I
I
attended
the
what
was
it
cncf,
the
kubecon
eu?
I
just
can
barely
hear
you
all
right.
Let
me
see
if
I
can
turn
it
up.
I
guess
not
that's
about
it.
That's
all
you
get.
That's.
D
C
Okay,
yeah,
so
the
the
kubecon
eu,
and
then
I
also
I
was-
I
spoke
at
the
open
source
foundation
or
sorry
open
source
summit
from
the
linux
foundation
and
they
both
had
a
similar
platform.
But
I
thought
they
did
a
pretty
good
job.
They
make
it.
They
like
the
home
page
for
the
the
conference
was
as
if
it
was
a
conference
hall,
and
so
you
click
around
to
the
various
like
doorways.
D
C
Walk
around
it
was
pretty
nice.
I
I
do
enjoy
being
able
to
attend
those
conferences
where
otherwise
I
wouldn't
be
able
to
travel
and
take
time
off
work
and
all
that.
So
it's
pretty
nice,
it
did
get
a
little
overwhelming
because
there's
so
many
things
happening
all.
At
the
same
time,.
B
Yeah
yeah,
that's
exactly
what
I
was
talking
about.
It
was
just
like
that
yeah
with
a
bunch
of
sticky
pads,
but
I
missed
the
beginning
of
the
talk
like
how
do
you
I
mean
my
my
cursor
appeared,
but
how
do
you
write
on
them
and
how
do
you
vote?
Where
do
you
vote
on
on
the
the
tablet?
That's
the
only
problem
I
had
other
than
that.
I
was
able
to
follow
the
talk
and
the
sticky
notes
and
things
of
that
sort.
They
had
like
three
different
columns.
B
Yeah
there
was
there's
like
there
was
a
a
picture
associated
with
my
login,
and
I
clicked
on
that.
That
was
the
only
icon
I
had
on
the
left.
A
Nice
thanks
for
that,
I
wasn't
aware
of
mural
and
they
they
have
118
million
investment
to
to
make
a
real
time.
A
A
And
yeah,
so
many
apps
have
been
missing
the
sort
of
the
google
docs
like
experience
of
synchronous
multi-user,
you
know
collaboration
like
it.
Just
changes
the
game
that
yeah
that's
the
these.
Are
I
hadn't
considered
it
until
you
guys
brought
it
up
that
geez
there's,
probably
something
else
that
we
should
be
doing
here
other
than
slides.
A
You
know
other
than
like
static
media.
I
mean
we
we've
done
workshops,
and
so
you
know,
there's
there's
labs
that
we
go
through
and
people
and
you
know
anyone
attending-
is
doing
that
on
their
own.
A
Machine
have
to
get
this
some
good.
Something
good
to
think
about
is
yeah,
be
really
interesting,
actually,
just
even
specific
to
the
technology
at
hand.
Around
service
meshes
is
to
have
one
up
here,
sharing
that
we
can
all
look
at
and
and
actually
all
poke
at
and
sort
of
see
the
effects
of
one
person
making
a
configuration,
change.
B
A
Yeah,
as
a
matter
of
fact,
yeah
yeah,
it's
been
part
of
the
vision
for
one
of
the
websites
that
that
I
focus
on
and
that's
been
that
anyone
who
visits
the
site
would
understand
that
the
site
is
running
on
a
surface
mesh
and
that
they
would
be
able
to
kind
of
take
control
of
it
would
both
be
a
learning
thing
to
understand
how
a
service
mesh
works.
They
would
be
able
to
take
control
of
their
own.
A
It
would
be
a
choose
your
own
adventure
in
some
respects
as
to
how
they
navigate
the
site,
how
they
can
affect
changes
on
the
site
based
on
letting
them
affect
the
config
of
the
surface
mesh
it
it
needs
to.
You
know,
ideally
not
let
them
shoot
themselves
in
the
foot
or
if
they
do
give
them
an
escape
route,
but
I
think
that
that
would
be
pretty
intriguing
and
powerful
a
powerful
learning
tool,
while
people
are
visiting
your
site,
so
I
don't
mean
it.
I
mean
I
don't
mean
your
site.
In
general,
I
mean.
B
B
Yeah,
I
just
like
what,
whatever
what
was
that
one
we
used
to
play
three
mile
island
and
you
played
until
you
blew
up
the
world
or
there's
several
fighter
airplanes.
It's
like
hell,
yeah
here,
go
to
this
website
and
have
all
the
fun
you
want.
A
Yeah,
a
by
the
way,
so
bob
brought
up
actually
two
other
really
good
call
outs.
One
is
hashiconf.
That's
coming
up,
I'm
aware
of
a
couple
of
announcements
that
they'll
be
making
at
hashicoff
and
I
didn't
look
to
see
if
there
was
a
price
on
the
on
the
conference,
but
the
announcements.
I
think
it's
free,
because
it's
digital
nice,
nice
cool.
D
Mean
or
there's.
B
A
Yeah
no
working
on
a
project
with
with
one
of
the
engineers
in
their
cto
office
and
spent
a
lot
of
time
on
the
weekend
working
on
some
stuff
with
hashicorp,
so
yeah.
I
expect
it
will
be
a
good
conference.
This
is
a
great
call
out
as
well,
but
this
has
been
a
few
years
running
now,
right.
A
That's
something
they're.
I
they've
been
silly
enough
to
put
the
last
couple
of
talks
that
I've
had
there
into
the
the
the
volume
that
they'll
put
out,
which
I
think,
which
is
that's
it's
a
bright
idea.
It's
a
great
great
set
of
content.
They've
got
all
the
content
there.
They
can
help
elevate
it
and
yeah.
A
It's
a
good
one
as
well
like
that
that
one
I
gotta
imagine,
is
it's
either.
I
think
in
the
pa.
Has
it
always
been
free,
like
I
thought
it
was.
A
12,
okay,
yeah
nice,
all
right!
Well,
good!
I'm
glad
that
we
exchanged
in
part
because
now
there's
less
time
to
talk.
A
Yeah,
let's
do
as
a
matter
of
fact,
maybe
we
should
just
do.
Maybe
we
should
just
do
that
I'll,
stop
I'll
start
with
a
slider
too,
but.
A
I
don't
know
just
at
any
point
that
I
pause
too
long
or
gets
too
boring,
then
just
holler
out
and
we'll
just
switch
to
a
demo.
A
We
were
going
to
talk
about
this
today
that
surface
mesh
is
you
know
that
is
the
the
notion
that
service
meshes
bring
quite
a
bit
of
value
as
a
matter
of
fact
for
for
myself
just
kind
of
focused
on
them,
I
really
consider
that
they
bring
a
whole
heck
of
a
lot
more
value
than
the
vast
majority
of
users
are
currently
aware
of.
A
There's
some
uses
of
them
that
aren't
popularized
just
yet
those
uses
are
on
their
way
to
being
popularized
a
lot
of
the
value
that
and
I'll
try
I'll.
I
will
articulate
that
a
little
bit
later
in
this
talk,
I
think
it's
worthwhile
to
couch
what
I'm
alluding
to
in
context
of
the
notion
that
the
value
proposition
of
why
people
deploy
a
service
mesh
is
perhaps
not
all
that
distinct
from
the
value
proposition
of
functions
of
serverless
the
value
prop.
A
I
don't
know
how
to
I
don't
know
how
to
put
an
asterisk
or
a
caveat
next
to
that
statement,
but
but,
but
you
know
what
I
mean
and
that's
for
you
know,
works
well
for
event,
driven
things
which
you
know
is
some
certain
percentage
of
the
world's
workloads
and
that
the
much
much
larger
majority
of
the
world's
workloads
is
process
driven
things,
long-lived
processes
right
things
that
are
listening
all
the
time,
which
makes,
which
is
what
service
mesh
service
measures
address,
both
event,
driven
workloads
and
long-lived
processes,
but
probably
benefit
long-lived
processes,
a
bit
more
there's
more
more
ways
in
which
they
benefit
them
and
which,
just
so
happens
to
be
like
the
80
of
the
world's
workload
like
most
of
the
most
of
the
workloads
run
in
this
way,
and
so
I
bring
this
up
to
say
that
of
that
value
proposition,
I'm
going
to
articulate
it
here
in
a
moment
the
standard
value
prop
that
people
know
of
today.
A
It's
probably
there's,
probably
some
more
there,
so
there's
a
horrible
teaser
that
I
just
gave.
But
anyway
my
name
is
lee.
Most
of
us
probably
hopefully
had
a
chance
to
meet.
If
we
haven't
there's
some
of
the
some
of
the
talks
that
I
give
or
some
of
the
there's
there's
two
books
that
I'm
writing
right
now,
which
is
really
painful,
they're
they're
listed
here,
but
but
if
you
want
to
know
more,
there's
there's
links
or
if
we're
not
connected,
please
connect
up.
A
I
wanted
to
point
out
pratik
as
well
pratik's
a
phd
candidate
at
the
university
of
ut
austin.
Here
nice
guy
he's
currently
doing
an
internship
at
intel
on
more
toward
hardware
things
and
but
he's
been
engaged
with
layer
five
for
about
a
year
and
he's
been
helping.
Do
some
of
the
studies
that
we're
going
to
talk
about
today.
Some
of
the
studies
around
the
cost
of
a
mesh.
A
I
figured
I'd
toss
in
like
one
or
two
high-level
slides
to
say
mostly
that
service
meshes
are
a
thing
that
they
are
from
my
perspective
and
inevitability,
that,
for
many,
it's
probably
for
more
so
than
excuse
me
than
not
service
meshes
aren't
necessarily
a
thing
that
they've
got
their
hands
on
and
have
deployed
yet
maybe
but
yeah,
but
it's
but
they'll
they'll
you
they
us
will
get
there
it'll
be.
A
If
you
go
into
an
environment
today
and
and
people
aren't
using
containers,
yeah
there's
reasons
not
to
but
you'd
kinda
you'd
probably
be
find
that
strange
enough.
That
you'd
say
well.
Why
not,
if
they're
using
a
few
you're,
probably
using
an
orchestrator,
you
would
expect
that
there's
an
orchestrator
there.
B
B
A
A
Yeah,
a
hundred
percent
aren't
of
that
those
workloads
aren't
in
containers,
which
is
makes
a
lot
of
sense
it
it,
but
that
doesn't
mean
that,
like
all
of
them
aren't
using
containers
somewhere
or
or
anyway
be
careful
with
this
statistic,
yeah
that
you,
the
statistic
was
70
of
those
queried
were
not
using
containers
for
100
of
their
workloads.
A
Okay,
yeah
yeah.
No,
but
I
guess
in
part,
like
hey
your
point,
your
respective
of
the
the
numbers
like
you're.
I
totally
agree
with
the
point
which
is
like
hey,
there's
a
ton,
there's
any
number
of
valid
reasons
not
to
be
running
things.
B
Inside
yeah
and
then
I'm
I'm
I'd
still
get
approached
a
lot
about.
You
know:
hey
we're
going,
we've
never
been
in
containers
and
we're
looking
into
that
and
what's
the
kubernetes
and
how
do
you
where's
the
map,
the
other
night?
It
was
like,
how
do
you
figure
out
all
these
parts
and
everything
so
we
sent
them
to
what
cncf
their
little
page
of
all
the
software
he
he
said
he
had
joined
the
company
and
didn't
know
anything
about
it
and
the
company
wanted
to
do
that.
B
A
C
Yeah
that
one
yeah-
that's
that's
really
handy.
I
used
that
in
my
last
job
to
help
convince
the
cio
to
to
move
forward.
That
was.
A
C
And
service
mesh
is
on
there,
I
think,
towards
the
bottom.
A
Oh,
is
it
yeah?
It's
been
a
while,
since
I
it's
been
a
couple
of
years,
I'd
had
a
hand
in
making
that,
and
it's
been
a
few
years
since
I
looked
at
it
but
makes
a
lot.
It
was
certain
but
yeah.
A
It
makes
sense
that,
like
maybe
the
service
mesh
was
sort
of
down
down
the
road,
but
that
to
your
point,
like
yeah
containers,
being
there
sort
of
right
at
the
start
and
continuous
integration
kind
of
being
there
early
on
and
and
actually
I've
got
a
keen
sense
of
what
you
mean
by
the.
How
helpful
can
be
to
have
a
piece
of
an
asset
like
that
or
a
piece
of
literature
like
that
to
be
able
to
say
hey
using
using
crayons.
This
is
sort
of
what
that
looks
like.
C
A
Yeah
no
yep
really,
and
I
say
the
the
crayons
with
a
smile
on
my
face
because
I
need
I
need
crayon
drawings
quite
often
and
and
yeah
for
for
it,
for
you
not
for
that
direction
to
be
coming
from
a
source
of
authority.
If
you
will,
or
or
not
out
of
the
same
mouth
that
you're
using
to
tell
the
person.
C
Yeah,
it's
not
just
hey,
I
saw
you
know
jim
smith's
blog
post.
That
says
we
should
do
this.
A
That's
that's
really
thanks
for
saying
that,
because
the
sig
network
that
I
chair
it
it
like,
I
was
saying
before
it
is
it's
not
over
the
these
specific
projects
right
here,
but
it
has.
A
Governance
is
a
strong
word,
but
it
has
responsibility
for
all
the
rest
of
the
ones
that
you
see
in
this
section
and
yeah.
You
know
we
get
a
ton
of
feedback
about
the
landscape
and-
and
you
know
the
notion
that-
and
it's
fun
like
I
enjoy.
You
know
poking
at
things
as
much
as
the
rest,
it's
fun
to
say
like
it's
really
confusing,
and
it's
also
I
mean
some
of
this
is
about
I
mean
in
order
to
innovate.
You
usually
have
to
disaggregate.
A
You
usually
have
to
have
a
bunch
of
competition
and
figure
out
what's
going
on,
and
you
know
find
some
new
paths
forward
and
over
time,
a
lot
of
times.
The
cncf
isn't
there
to
pick
a
winner,
but
they
will
like
of
the
projects
that
are
further
along
and
are
doing
well.
They'll,
take
them
through
those
phases
of
sandbox,
incubation
and
graduated
and
as
they
go
through
those
phases,
they'll
just
visually
represent
them
a
bit
larger.
A
I'm
here
and
I
was
saying
before,
like
one
of
the
projects
we'll
talk
about,
it
is
measuring
and
today
and
it's
on
the
landscape.
Here
it's
about
to
head
into
the
the
cncf
and
so
yeah
I'll
go
I'll.
Go
look
at
the
cloudtrail
map
again
because
I
I
don't
see
a
link
to
it
immediately.
C
Yeah,
it
is
a
little
difficult
to
find,
but
also
I
mean
just
not
to
distract
you
from
your.
I
don't
talk,
doesn't
matter,
it's
the
that
recent
release
from
the
was
it
the
continuous
integration
team
yeah
with
their
radar.
C
Oh,
I
found
that
was
super
helpful
too,
because
you
know
that
that
chart
you
just
showed
the
actual
landscape
is
huge
and
having
that
radar
to
say
hey,
these
things
are
what
a
lot
of
people
are
using
and
here's
some
stuff
that
is
getting
more
popular
and
then
here's
some
stuff,
that's
pretty
new.
So.
C
A
Then
you're
right
that
it
probably
is
from
the
cncf
it
is.
I
gave
a
lot
of
feedback
on
that
at
the
last
technical
oversight
committee,
which
is
to
say-
and
I
don't
know,
I'm
gonna-
be
able
to
quickly
find
reference
to
the
radar.
A
But
it
was
to
say
that
it's
yeah,
just
like
the
landscape
is
the
radar
is
a
good
thing
and
a
bad
thing.
It's
frustrating
that
it's
not.
A
All
exhaustive
that
it's
not
an
all-inclusive
list
like
the
last
edition
of
the
radar
that
came
out,
was
on
observability
and
talking
about
tools
for
observability,
and
there
were
about
32-35
35,
end
user
organizations
that
had
participated
in
saying
here's
what
we
here's,
what
we
suggest
to
adopt.
Here's
what
we
suggest
to
wait
on
for
a
little
while-
and
I
just
I
wish
it
was
a
lot
higher
number
of
participants
and
a
more
exhaustive
list
of
like
options
to
choose
from,
but
that
yeah.
C
So
I
was,
I
was
also
a
little
disappointed
in
the
lack
of
it
wasn't
full
like
I
expected
it
to
be,
but
still
really
nice,
though
yeah.
A
No,
it's
still
quite
helpful.
I
mean
it's
your
point
earlier
about
the
trail
map
that
it's
another
one
of
those
things
that
you
that
can
really
facilitate
a
conversation.
You
can
really.
C
A
That
came
out
this
last
week,
I
think,
on
observing.
I
think
it
was
this
last
week,
but
it
was
on
observability,
and
so
the
format
is
a
little
bit
different
in
terms
of
how
they
wrote
it
down,
but
it's
still
the
same
three
rings
and
yeah
no.
D
Isn't
there
a
fourth
section
to
like
drop
stuff
or
you
know
like
it's
depreciated
or
whatever.
A
There
is
that
mo
on
this
radar.
I
don't
believe
that
they,
I
I
to
my
knowledge.
They
don't
account
for
that.
In
general,
though,
the
cncf
does
have
a
motion
for
exodus.
You
know
for
identifying
that,
like
hey,
that
that
project's
the
project
that
once
looked
promising
sort
of
its
time
has
come
and
gone
so
there
has
been
either
one
or
two
projects
that
have
more
or
less
sunsetted.
A
I
couldn't
recollect
what
it
is,
but
but
yeah
to
my
knowledge,
the
radar
doesn't
capture
those,
but
the
cncf
does
acknowledge
that,
like
yeah
thanks
for
this
yeah,
I
really
don't
much
care
about
the
talk
that
I'm
giving
today.
So
it's
just
like
this
interchange
is
much
nicer.
A
A
That
is
a
real
good
introduction
to
the
space
to
what
service
meshes
are
and
which
ones
are
out
there.
And
what
do
they
do
this
one
there's
a
second
edition
of
this
coming
out
in
like
three
weeks
so
so
my
recommendation
would
be
to
go
to
the
layer,
five
io
site
subscribe
to
the
site,
the
email,
the
mailing
list
and
then
get
notified
when
the
second
edition
comes
out.
It'll
be
free,
but
it'll
be
up
to
speed
fair
enough.
So
this
is
usually
when
people
articulate
the
value
of
a
mesh.
A
They
talk
about
kind
of
three
things:
the
fine-grained
traffic
control,
the
ability
to
intercept,
requests,
introspect
them
redirect
them
put
you
know,
kind
of
dismiss
them
do
do
all
kinds
of
things,
but
but
to
do
some
heavy
lifting
on
the
on
the
network
as
part
of
doing
that,
heavy
lifting
and
looking
introspecting
packets
and
requests
and
there's
a
lot
of
telemetric
signals
that
are
produced
logs
and
metrics,
and
traces
that
are
facilitated
and
people
are
really
drawn
to
that.
A
The
third
leg,
if
you
will,
is
around
security
which
has
to
do
with
identity
management
and
making
sure
that
every
service
has
its
own
identity
and
being
able
to
then
facilitate
mutual,
authentic,
mutual
tls
in
between
every
service.
That's
in
the
mesh.
A
I
I
tossed
the
fourth
one
on
here,
which
is
really
the
the
notion
that,
since
there
is
so
much
the
ability
to
observe
and
the
ability
to
control
a
lot
of
the
value
that
a
match
will
provide
is
happens
to
do
with
resiliency,
either
inducing
chaos
or
kind
of
controlling
chaos.
A
So
good,
I'm
a
little
bit
hesitant
about
sort
of
moving
through
this.
This
talk
because
I
want
for
any
of
you
to
interrupt
as
we
go,
and
I
also
I'm
not
sure
how
many
of
you
are
totally
fresh
on
the
scene
here
and
how
many
of
you
have
heard
this
before
the
short
of
it
is
even
if
you're
not
running
things,
even
if
you're
not
running
things
in
a
container,
you
can
use
it.
Those
services
can
be
on
a
service
mesh.
A
C
So
lee,
if,
since
you
yeah,
are
begging
for
interruption
yeah,
so
the
reason
I'm
here
is
because
I've
been
curious
of
service
mesh.
I've
played
with
it.
That's
it's
neat,
you
know,
but
at
work
we
are
actively
migrating
a
monolith
into
containerization
microservices,
splitting
things
off
and
moving
into
a
kubernetes
world
and
cloud
native
and
all
that
other
fun
stuff.
C
And
I
see
there
is
a
lot
of
benefit
for
service
mesh.
However,
you
know
we
also
have
a
lot
of
things
on
our
plate
and
how
I
guess
is:
is
there
a
point
where
it's
like?
Okay?
Well,
you
know
you
have
two
or
three
services
out
there
running
it's
time
to
start
breaking
the
ice
with
the
service
mesh
and
bring
it
in
earlier
or
wait
until
you're
stable
and
then
also
introduce
the
service
mesh
on
top
of
a
stable
migration,
or
you
know
where.
B
C
Good
but
yeah,
that's
exactly
it
too
right
when,
when
do
you,
when
do
you
trade,
the
balance
of
the
complexity
of
adding
a
service
mesh
versus
the
benefits
of
having
a
service
mesh,
and
when
does
that
fall
into
place?
And
I'm
not
sure
if
there's
a
specific
metric
like
brad
was
asking
for.
A
Other
totally
yeah
totally.
A
A
A
Don't
have
I'm
trying
to
be
a
little
more
reflective.
I've
had
four
different
service
mesh
maintainers
up
on
stage
answering
this
very
question.
Actually
a
few
different
times
like
this
question
gets
asked
essentially
every
time
that,
like
really
frequently
and
actually
the
bur
you,
you
probably
character,
probably
asked
it
a
little
bit
better
than
most
do
a
lot
of
times.
It's
a
fairly
pointed
question
of
like
so
how
many
services
do
I
need
to
be.
A
Do
I
need
to
be
running
in
order
for
me
to
know
what
where's
the
where's
the
point
of
inflection?
That's
like
okay,
I'm
running?
Is
it
five?
Is
it
ten?
Is
it
and
people
get
fairly
impatient
about
it?
Actually,
they
wanna
know
like
no,
I'm
just
you
know
like
ballpark
like
is
it
500?
Is
it
like
one?
Is
it
and
you
know
like
every
every
good
answer
to,
or
you
know,
to
any
complicated
thing
it's
well.
It
depends
now
that
it
depends.
A
Well,
this
is
a
perspective.
There
are
truths
in
here,
so
one
truth
is
even
if
you're
just
running
one
service,
you
will
derive
value
out
of
a
service
mesh.
A
That's
only
part
of
the
thing,
like
the
other
side
of
that
coin
is
okay.
Well,
I'll
get
value,
but
you
know
like
is,
but
is
the
return
on
investment,
but
there's
some
investment,
so
is
the
value
greater
than
the
investment
and
and
the
more
that
time
ticks
along
the
easier
it
is
to
say?
Yes,
it
is
the.
There
are
a
few
messages.
A
few
meshes
that
we
could
even
maybe
demo
today,
actually.
B
B
If
I
have
all
my
test
cases,
you
know
it's
going
to
be
better,
but
I
got
three
guys
and
we
got
60
days
to
get
a
prototype
out
and
then
we're
going
to
go
to
80
guys
in
six
months
and
then
we're
going
to
do
something
else,
and
then
we're
going
to
go
to
260
guys
and
we're
going
to
do
something
else,
and
there's
only
so
much
work
that
you
can.
You
can
slice
up
for
people.
So
if
there's
no
code,
then
who
cares
about
server,
smash
and
testing
and
security?
B
B
What
what
the
question
is
is
I
can't
do
all
the
right
things
at
the
beginning,
even
though
no
matter
who
you
are
because
one
person
can't
work
more
than
about
18
hours
a
day
it
you
couldn't
run
maybe
longer
for
a
couple
of
days,
but
other
than
that
you
know.
So
that's
the
the
direction
of
the
question
is
server
smash.
B
A
A
If
you
want
to
call
it
that,
I
got
to
imagine
that
if
you've
ever
been
there
and
you're
you're
busy
working
on
building
out
on
your
day,
job
and
building
out
business
logic,
applications
and
functionality
that
to
take
a
sprint,
a
couple
of
sprints
more
than
a
couple
of
sprints
to
like
to
set
up
and
invest
in
ci
and
get
the
system
going
and
the
scripts
and
doing
automation
and
like
all
this
and
then
it
breaks
and
then
actually
the
builds,
are
breaking
and
then
like
it,
didn't
look
pretty
on
this
side
of
the
wall.
A
If
you've
ever
been
on
the
other
side
of
the
wall
of
having
ci
deployed,
you
ain't
never
going
back,
you're
like.
Why
would
you
ever
wait?
Why,
yes,
it's
gonna
take
a
week
to
like
get
it
right,
get
all
the
various
levels
of
the
build
done
and
etc.
Why
would
you
ever
have
that
be
a
manual
process?
You
would
only
ever
like,
like
you
start,
a
new
project.
The
first
thing
you
do
is
set
up
ci,
so
you
don't
waste
that.
B
That
I
mean
I
a
lot
of
consulting
in
central
texas
and
made
a
whole
lot
of
money
going
in
and
solving
that
particular
business
problem
and
huge
corporations.
Apple
ibm,
am
amd
and
solve
those
business
problems,
but
even
they
were
hey.
I
got
only
so
many
head
count
to
to
get
something
to
go
and
that
that's
just
the
perspective.
I'm
asking
I
understand
what
you're
going
through,
but
you
don't
have
to
have
it.
Have
you
ever
been
to
a
site
that
doesn't
have
it?
A
Was
that
number,
oh
sure,
actually
allen?
Something
had
happened
to
your
voice?
It
was
pretty
low.
I
I
think
I
got
what
you
said,
which
was,
if
you've
ever
been
to
a
site
where
in
fact,
they're
not
running
service
mesh
yet
and
and
it
is
a
resource-constrained
environment
and
they
were
doing
just
fine
like
I
to
finish
the
thing
that
I
was
saying.
A
That
is
that
it's
the
same
thing
for
me:
hey
if
you're
running,
if
you're
not
running
a
mesh,
you're
running
it
one
service,
five
services:
do
you,
do
you
ever
need
an
api
gateway
in
front
of
it?
Do
you
ever
want
resiliency
to
have
a
virtual
ip
address,
or
maybe
to
have
an
automatic
retry
from
one
service
to
the
next?
Maybe
just
have
two
services
that
talk
to
each
other.
Do
you
want
load
balancing?
Do
you
want
some
rate
limiting
in
there?
A
Do
you
want
to
be
able
to
redirect
traffic
so
that
you
can
do
a
roll
out
of
one
version
of
the
software
to
the
next?
Like
there's
a
lot,
a
lot
of
use
cases
in
there
for
and
I
my
perspective
would
be
well.
A
You
know
most
of
the
time
you
probably
want
about
half
those
things
like
some
of
those
things
you
can
live
without
you
don't
necessarily,
but
the
other
half
it's
like
hey
you
great,
there's
solutions
out
there
that
are
point
solutions
for
those
things
that
I
just
rattled
off
service
mesh
puts
them
all
together
and
it's
it's
one
of
those
things.
This
is
why
I'm
saying
it's
kind
of
an
inevitability.
A
It's
like
hey,
just
like
everything
else
when
containers
came
around
takes
a
while
to
figure
out
what's
going
on
the
layers,
how
to
run
it.
The
fact
that
you
can
run
some
using
a
process
manager
or
run
some
without
some
using
a
daemonize
and
for
the
so
for
the
service
mesh.
It's
kind
of
the
same
thing
like
it's
there's,
a
reason
why
it's
it's
a
thing
and
has
an
industry
term
next
to
it,
because
it's
big
and
complicated
for
those
of
you
that
are
running
kubernetes.
A
Many
of
you
probably
feel
fairly
comfortable.
You
also
understand
that
there's
a
few
corners
of
kubernetes
that
you
probably
don't
feel
comfortable
on.
Maybe
you
haven't
gotten
to
yet,
and
hopefully
you're
comfortable
with
not
with
the
unknown.
Networking
is
difficult
and
service
meshes
help
empower
people
to
overcome
it.
These
are
some
generalities
to
try
to
answer
your
questions.
Better
is
mostly
what
I'm
saying
is
like
the
answer
to
bert's
original
question
is
I'd
start
now,
I'd
start
at
the
beginning,
so
that
you
can
consume
a
service.
Mesh's
functionality,
a
la
carte.
A
You
can
go,
deploy
a
service
mesh
and
not
have
it
do
anything.
A
Don't
don't
tell
it
to
do
anything,
it
can
still
be
there.
The
proxies
are
there.
The
traffic
is
running
through
the
proxies,
and
it's
not
doing
anything
you
can
tell
it
don't
generate
any
telemetry,
don't
mess
with
the
traffic?
Don't
redirect
it
don't
encrypt
the
traffic,
don't
create
chaos,
don't
it
can
just
be
transparent
and
sitting
there.
A
You
can
get
not
even
now,
I'm
not
I'm,
not
I'm
not.
What
I
am
suggesting
is
like
deploy
it
early.
A
What
I'm
also
suggesting
is
to
bite
off
those
pieces
of
functionality,
a
la
carte
to
be
comforted
so
that
you
can
get
comfort
with
it
as
you
go
so
that
you
can
see
also
the
return
investment
on
each
one
of
those
pieces
of
functions
that
you
enable
as
you
go
yeah.
A
I
think
that
that
reduces
your
the
mental
risk
or
you're
psychologically
the
risk
that's
involved
in
running
a
nice
set
of
services,
and
then
you
come
over
and
say:
well,
we
need
to
dramatically
augment
the
architecture
by
laying
down
this
new
layer
and
then
oh
and
then
it
is
actually
like.
A
You
know
any
number
of
meetings
and
people
bickering
and
showing
touting
one
architecture
over
the
next
and
allen.
For
your
case,
like
I,
I
think,
you're
a
stronger
individual
than
I,
because
those
environments
that
you
describe
I'd,
get
frustrated
and
bring
an
uzi
in
the
next
day.
Just
tell
everybody
you're
missing
the
point.
The
point
is,
you
are
resource
constrained.
B
B
I
could
get
basically
the
amd
yeah,
the
amd
project
30
years
ago
that
they
brought
me
in
to
solve
was
they
were
going
to
save
multiple
millions
and
millions
of
dollars
to
deliver
chips,
because
basically,
chips
take
three
four
years
to
design
and
debug
to
then
roll
out
and
go
through
fabrication
chemicals
anyway,
the
software
was
going
to
do
that,
but
they
only
were
given
six
people
to
do
the
job
and
one
phd
scientist
just
explained
to
them
kind
of
how
to
do
it
to
create
a
prototype.
B
To
then
go
and
say:
hey.
This
is
the
right
project,
so
those
are
the
kind
of
projects
I
was
brought
in
on
or
like
at
ibm.
There
was
political
issues
at
a
board
of
director
meeting
where
they
didn't
agree,
and
so
they
needed
a
moderator
to
solve
that
particular
problem,
and
it
was
all
about
everyone
was
doing
their
own
thing
and
trying
to
just
get
the
job
done
so
that
I
mean
I
run
across
that
a
lot,
and
so
you
you
only
have
so
many
people
to
do
so.
B
So
you
know
your
investment
is
one
week
to
do
the
match.
Well
great.
We
also
need
this
a
week
or
two
to
do
an
ide
to
train
all
the
people
and
when
we
need
processes
and
policies,
those
things
are
are
reality
to
large
direction
and
well
financed
in
in
the
finance
agrees
with
the
vision
of
the
the
the
deliverable
that
they're
going
to
go
generate
revenue,
I
mean
it
all
comes
down
to
dollars.
So
so
that's
what
I
was
asking
about
is,
in
you
know,
from
an
executive
level.
B
What's
my
return
on
the
investment
at
what
point
because
everyone
wants
to
be
first,
I
can't
hire
30
security
people,
which
is
the
right
thing
to
do.
I
can
get
one
you
know
so
anyway.
That's
kind
of
this
perspective
I
was
asking,
I
don't
know
enough
about
server
smash,
I'm
I'm
certain
with
the
cncf
basically
process
to
even
get
a
project
in
there.
It's
it's
got
to
be
well
validated
in
required
function,
but
there's
a
whole
lot
of
those
products.
A
B
Know
they
don't
give
a
about
a
server
smash,
they
want
to
see
what
their
flashy
crap.
So
that's
kind
of
what
I
was
yeah,
how
your
your
knowledge
in
mentoring
and
other
people,
someone
else
on
the
audit
can
remember
it's
like
yeah.
You
got
to
do
it
first.
Well,
a
whole
lot
of
things
have
to
be
done.
First,
so
think
about
it.
As
a
small
group
of
25
guys,
trying
to
like
hey
we're
going
to
put
this
together.
B
B
Too
much,
but
you
know
that's
what
most
people
I
mean
they're
going
to
have
to
go
to
the
investor
and
say
I
need
a
service
smash.
I
need
an
ide
and
they
got
to
be
able
to
explain
and
no
matter
what
you
are
in
huge
corporations.
They
brought
me
in
to
solve
these
problems.
They
weren't
funded
adequately
to
get
whatever
they
wanted
to
be.
A
B
And
they
knew
they
needed
to.
Another
example
is
how
many
people
write
the
code.
Actual
code
doesn't
have
anything
to
do
with
service
mesh
that
is
easily
translatable
to
do
an
international
delivery.
In
other
words,
are
you
writing
your
code
with
the
message
tables
and
all
your
errors
can
be
swapped
out
to
multiple
languages
around
the
world?
You
must.
I
don't
need
to
do
that.
Everyone
knows
english.
Well,
you
you
just
cut
out
certain
markets
like
in
europe
and
france,
it's
like
it.
B
You
can't
sell
to
the
military
or
the
government
there
because
they
only
accept
stuff
in
their
language.
So
you
know,
service
mesh
is
one
of
those.
Many
things
you
need
to
do
to
finish
the
product
to
you
know
be
the
best
tick
tock.
You
know
in
the
world
and
worth
a
whole
lot
of
money
and
you're
staying
making
that
damn
cent.
You
know,
because
you
invest
the
money
wisely
and
got
to
the
end
result.
So
that's
kind
of
what
I
was
thinking
about.
A
I'm
glad
that
we're
having
this
conversation
burton
you
might
have
a
couple
of
other
things
to
say
this.
Actually,
I
think
that
that
I
there's
a
new
talk
that
maybe
I
would
I'd
like
to
give
and
it,
and
it
might
be
just
be
like
hey:
what's
the
value
of
a
service
smash
like
what
would
it
when,
if
you
lay
one
down,
what
do
you
get
the
things
that
I
talk
to
in
the
slides,
and
maybe
it
should
just
be
done
through
pure
demo
of
like
here's,
a
mesh
as
a
matter
of
fact
like?
A
Let
me,
let
me
show
you
guys
this.
This
is
it's
really
nice
that
we're
having
this
this
conversation,
I
if
I
do
a
watch,
cube,
ctl,
get
pods
all
and
so
on.
My
in
my
environment,
I'm
running
docker
desktop
on
my
my
mac
here.
A
It's
running
kubernetes
turns
out
I'm
running
one
service
mesh
here.
Kuma
excuse
me,
which
is
relatively
small.
This
piece
of
software
is
a
management
plane.
Service
meshes
are
made
up
of
data
planes,
a
control
plane
or
a
data
plane,
a
control
plane
and
a
management
plane
and
mesherie
is
the
the
tool
that
we
use
as
a
management.
Plane.
A
Meshery
in
this
case
supports
eight
different
service
meshes
and
lets
you
quickly
deploy
them
so
for
open
service
mesh,
which
is
one
of
the
ones
that
well
not
one
of
the
ones,
but
the
one
that
microsoft
had
announced
a
few
weeks
ago.
A
Mestre
will
go
ahead
and
interface
with
your
kubernetes
environment
and
quickly
deploy
open
service,
mesh
or
osm
part
of
the
value
of
most
of
the
meshes
that
you'll
find
today
have
to
deal
with
the
fact
that
they'll
provide
some
telemetry,
so
they're
they're
deploying
a
small
instance
of
grafana
and
prometheus
for
this
particular
mesh.
What
we're
looking
at
in
this
mesh
is
just
is
their
control
plane,
so
they
have
three
different
deploy
pods,
as
I'm
assuming
three
different
deployments,
one
of
the
greater
pieces
of
value.
A
The
open
source
community
that's
building
measuring
is
building
is
building
a
visual
topology.
To
short,
to
show
you
real
time
traffic
to
let
you
manipulate
that
traffic
configure
it.
You
know
like
visually
design,
the
mesh
so
that
you
can
say
no.
I
want
to
redirect
any
traffic
that
has
a
jwt
where
alan
is
logged
in
we'll
redirect
him
to
the
the
gold.
You
know.
D
A
A
gold
subscriber
or
what
whatever
the
whatever
the
logic
is
there
but
yeah
it's
kind
of
going
back
for
a
second
to
like
hey
when
and
why
to
deploy
a
mash.
A
It's
an
interesting
thing,
a
lot
of
what
the
mesh
is
providing
that
people
are
aware
of
today
is
the
same
thing
that
you
can
get
other
places.
It
just
brings
them
all
under
one
central
point
of
control,
it's
kind
of
like
using
a
piece
of
monitoring
tool.
A
If
that
monitoring,
monitoring
tool
can
centralize
your
logs
and
your
metrics
and
do
snmp
and
do
http
and
do
like
the
more
of
it
can
be
in
one
a
single
pane
of
glass,
that's
kind
of
that's
kind
of
nice,
there's
no
context,
switching
between
the
logging
system
and
what
happened
at
that
time,
and
then
the
graphical
charts
and
grafana.
What
happened
that
happened
at
this
time
like?
A
Rather,
you
find
a
lot
of
value
when
you
use
a
data
dog
or
a
or
something
similar,
and
it's
showing
you
all
those
signals
kind
of
in
one
view,
so
that
you
can,
you
don't
have
to
context
switch
well,
you
consider
the
it's
kind
of
part
of
the
same
value.
Proposition
of
the
service
mesh
is
like
hey,
you
can
get
api
gateway
functionality
elsewhere,
you
can
get
load
balancing.
B
B
You
know
when,
when
you're
having
problem
when
you're
developing
the
code-
and
so
it
can
monitor
the
system,
the
processes
and
all
that,
but
if
you
buy
a
license
from
some
of
these
monitoring
products,
I've
seen
them
demo
where
okay
and
then
we
push
and
we
built
our
code
and
it
went
out
here
and
it
crapped
out
and
well
what
line?
Did
it
crap
out
on
the
monitoring
tool,
had
a
gui
or
an
add-on
or
a
plug-in
that
brought
it
up
to
the
actual
operating
system
and
hooked
into
their
ide?
B
It's
like
well,
here's
the
line
of
code,
and
here
oh,
I
forgot
to
initialize
this
or
whatever
they
did
yeah.
I
can
see
the
return
on
investment
from
I'm
trying
to
build
something
in
the
monitoring.
You
monitoring
your
resources,
your
utilization
of
your
product,
how
much
it's
going
through
and
how
much
network
traffic
is
going
through.
I
can
see
that
so
I
agree
a
hundred
percent.
C
Yeah
sorry
lee
I
was
just
gonna
say
I
appreciate
you
mentioning
that
just
go
and
deploy
it
and
then
break
off
little
pieces
as
we
go
because,
like
you,
you
you
reference
like
logging
and
stuff
and
ci
cd
and
those
other
things,
and
you
know
those
are
easy
and
and
well
known
to
be
small
pieces.
And
you
know
you
yeah,
it's
like
your
ci
cd
solution.
For
example,
you
get
that
right
away,
you
get
it
to
automatically
build
on
every
commit.
C
You
know,
that's
just
a
small
piece,
but
when
researching
service
mesh
and
stuff
a
lot
of
the
times,
it's
like
look.
You
can
do
this
and
this
this
and
this
this
it
just
goes
off
and
like,
oh,
my
god,
no
wait
hold
on,
let's
just
let's
just
get
everything
stable
and
and
everybody
comfortable
and
then
we'll
go
back
and
do
the
service
mesh.
It's
it's
not
well,
at
least
for
me
anyway.
C
It's
not
advertised
that
you
can
just
put
it
there
without
it
affecting
anything,
because
the
way
it
sounds
is
like
when
you
install
it
or
when
you
deploy
it,
you
have
to
configure
all
your
networking
and
all
of
this
other
stuff.
It's
like
man,
I
don't.
I
don't
have
to
deal
with
that
right
now.
Let
me
I'll
come
back
to
it,
but
the
way
you're
saying
it
just
just
put
it
out
there,
let
it
sit
idle
or
whatever,
and
then
you.
D
C
Maybe
later
we'll
we'll
start
adding
some
metrics
and
then
you
know,
then
we'll
do
some
circuit,
breaker
and
stuff.
That's
awesome.
A
Yeah
yeah,
I'm
glad
that
that
helped
yeah
it.
I
think
that
yeah,
you
can
imagine
why
all
the
literature
on
service,
mesh
or
piece
of
tech
will
say
will
a
spouse
or
it
will
promote
all
the
capabilities
of
the
thing
and
then
trying
to
convince
you
that
that
it's
quite
powerful
and
you
should
use
it
because
you
can
benefit
in
any
number
of
ways
and.
C
A
Yeah,
where
where's
step
one
and
so
yeah
that's
a
good
thing
to
note-
is
that
and
there's
there's
and
actually
this
is
funny
because,
like
a
couple
of
years
ago-
or
at
least
for
me,
like
a
couple
years
ago,
a
lot
of
times,
people
would
find
a
little
bit
of
discomfort
with
the
magic
sprinkles
of
a
service
mesh.
A
And
by
that
I
mean
the
notion
that
the
ability
to
sidecar
a
container
or
a
utility
container
next
to
your
application
container
was
a
bit
more
of
a
novel
concept
or
not
as
a
broadly
used
of
a
you
know,
of
an
approach
and
then
over
time
that's
gotten
people
kind
of
gotten
more
comfortable
with
hey.
Maybe
you
just
drop
in
a
utility
container
for
a
little
bit
just
to
maybe
debug
or
troubleshoot
or
some
and
now
the
approach
that
most
service
meshes.
A
Not
all
of
them
have
taken
is
to
leverage
the
notion
that
you
can
deploy
a
secondary
container
or
a
tertiary
or,
however
many
containers
you
have
in
your
your
pod
and
that
being
a
service
proxy.
This
that
you
would
sidecar
this
proxy
and
that's
part
of
what
that.
That's
really.
What
makes
up
the
data
plan
and
it's
what
makes
up
part
of
the
the
service
mesh
is
that
it's
transparent.
A
What
the
magic
sprinkles
were,
that
that
was
happening.
Automagically,
like
you,
took
your
regular
kubernetes,
manifest
your
application
manifest
and
did
cube,
ctl,
apply
and
and
all
of
a
sudden
there
was
a
second
container
next
to
your
other
one,
and
it's
not
like
that
was
a
bad
thing.
It
was
just
how
did
wait
a
second?
How
did
that
happen?
A
I
didn't
even
tell
you
know
I
didn't
instruct
kubernetes
to
necessarily
do
that
and
in
that
time
people
have
come
to
understand
the
mechanics
by
which
that
happens
and
become
more
comfortable
and
and
the
notion
that
these
service
proxies
are.
You
know,
they're
like
again
to
kind
of
draw
the
analogy
back
to
the
days
of
of
old,
or
maybe
the
current
days
for
for
some
of
us,
which
is
like
hey,
there's
a
layer
to
you
could
in
physical
networking,
you
might
think
of
it
as
like
a
anyway
there's
a
there's,
a
transparent
router.
A
That's
there
a
transparent
fire,
intelligent
firewall,
that's
there
that
can
do
a
lot
of
stuff
with
the
packets,
but
it
might
not
be
doing
anything
at
all
and,
as
a
matter
of
fact,
if
it
fails
someone
unplugs
the
power
cord
or
you
know,
if
the
service
proxy
fails
in
your
that,
I'm
sorry
like
the
control
plate,
there's
different
failure
modes
for
a
service
mesh.
So
maybe
I
but
but
what
I
was
trying
to
say
is
is
if,
like
let's
say,
the
control
plane
goes
down.
A
Those
proxies
are
sitting
there
running
I'm
doing
whatever
the
control
plane
had
told
them
to
do.
They'll
just
they'll
just
keep
doing
that
or
if
you
didn't
configure
them
to
do
anything
yeah.
I
guess
I'm
kind
of
repeating
what
I
said
before,
which
is
then
you
can
go
over
and
just
and
and
my
recommendation
would
be
to
just
seek
out-
and
maybe
alan
had
said
this
as
well
as
just
seek
out
first
a
bit
of
telemetry,
because
it's
it's
it's
read-only.
A
A
That
doesn't
mean
that
you're
telling
it
to
mess
with
the
traffic,
or
do
it's
just
doing
it's
just
an
accounting
machine,
hey
if
you've
ever
looked
at
apache
logs,
like
all
the
http
requests
that
come
through
an
apache
server
web
server
or
an
nginx
web
server,
a
lot
of
the
logs
you
you'll
get
out
of
a
service
mesh,
pretty
similar.
It's
just
like
it's
access
logs
and
like
who's
asking
for
it
and
when
and
so
those
that's
not
very
intrusive
or
that's
not
it's
you're
not
telling
it
to
mutate,
something.
C
Yeah,
that's
good
because
you
know
the
the
the
big
well,
I
don't
like
it
seems
like
the
big
selling
point
of
the
service
message.
We
can
control
all
of
your
network
traffic
like
well.
I
don't
want
that.
C
You
know
I
I'm
scared
of
putting
that
in,
especially
while
we're
in
the
process
of
migrating
we're
not
fully
comfortable
where
we're
at
and
you
know,
and
so
adding
the
service
mesh.
That
is
advertising
like
I
will
control
all
of
the
connections
and
routes
and
all
this
other
stuff.
Well,
don't
you
know,
but
just
just
putting
it
out
there
and
then
say.
Just
tell
me
what's
happening
is:
is
a
nice
sort
of
baby
step
into
into
the
world.
A
Yeah,
it's
illuminating,
can't
take
the
blinders
off
and
and
then
yeah
like.
Yes,
there
are
a
hundred
and
something
traffic
manipulation
things
to
do,
and
man
if
I'm
operating
it
and
I'm
operating
100
something
services
and
I'd
probably
poop.
My
pants,
like
you
know,
I
don't
want
to
be
the
one
and
I
don't
understand,
what's
going
on
with
this
yaml
and
what's
really
going
to
happen
when
the
mesh
takes
a
hold
of
that
yaml
and
does
what
I
think
I
told
it
to
do.
A
I
think
I
told
it
to
redirect
traffic
that
way,
but
it
wasn't
clear
to
me
that
there
was
a
depend
like
hey.
I
wanted
to
do
a
trial
run
first,
okay.
A
I
know
I
can
do
that
and
set
up
my
own
infrastructure
to
do
but
like
it
might
be
nice
to
have
a
management
plan
or
some
other
tool
that
can
show
me
what
a
dry
run
looks
like,
and
these
are
the
kinds
of
things
that
that
are
planned
for
for
measuring,
but
but
anyway,
when
you
do
get
to
the
traffic
routing
things
that
yeah
again.
A
Like
my
perspective
and
it's
funny
because
I
really
feel
like
this
is
kind
of
the
same
perspective
that
hey,
if
you're,
born
of
the
cloud
or
or
if
you're
mode,
two
or
if
you're,
devops
or
if
you're,
whatever
whatever
you
know,
all
the
all
the
things
that
we
call
it
that
you
it
takes
a
while
to
come
to
this,
that
small
frequent
incremental
change
is
for
some
extr.
When
you
say
change,
they
say
management,
that's
funny!
That's
the
first
time
I've
said
that
I
have
to
have
to
chuckle
at
myself.
A
You
know
like
and
so
to
them.
That
means
yeah
change.
The
last
friday
of
every
eight
months
will
do
a
change
like
and
actually
until
you've
lived
on
the
other
side
and
come
to
understand
change
every
eight
minutes,
because
it's
so
super
small
and
we're
changing
it
all
the
time
anyway.
That
means
we've
got
all
the
right
tools
to
handle
all
the
like.
Now
I
feel
so
much
more
comfortable
because
we're
just
you
want
to
change
something:
fine
yeah.
A
We
just
changed
something
six
minutes
ago,
so
you
know
go
ahead
and
change
it
again,
you're
like
I'm
that
maybe
extreme
the
minutes
that
I'm
using
but
but
for
some
it's
they're
not
that
far
from
that,
and
so
I
think
it
depends
upon
how
you
were
brought
up
or
maybe,
where
you've
spent.
Have
you
been
on
the
other
side
of
having
a
ci
system,
because
if
you
have
you're
and
anyway
meshes
meshes
even
they're
a
bit
they're
scary
or
they're,
I'm.
B
B
B
I
want
security
updates.
I
don't
want
them
to
change,
oh
well.
This
is
over
here
and
that's
over
here,
because
someone
decided
that
that's
the
better
thing
to
do.
I
want
to
choose
that
if
I
want
to
versus
I
mean
so,
I
want
security
updates
things
of
that
sort,
but
I
don't
like,
from
a
user's
point
of
view,
upgrades
to
my
phone
changing
all
the
little
tools.
B
You
know
just
a
little
bit
here,
a
little
bit
there
similar
to
remember
meetup
when
they
got
bought
out
and
everyone
loved
that
product
and
then
the
new
company
bought
it
and
it's
like.
Oh
we're.
No
longer
doing
this
and
people
complain.
Hey.
B
So
it's
just
a
different
method.
Now
you
know
it's
like
if,
if,
if
your
scotch
came
in
a
different
shaped
and
sized
bottle
every
time
in
a
different
level
label
on
it
every
time,
because
someone
decided
to
change
it,
it
would
be
kind
of
frustrating
when
you
walked
into
the
liquor
store
to
find
your
scotches.
I
wonder
what
the
hell
it
looks
like
now.
You
know
it's
still
scotch
on
the
inside,
so
the
product
still
in
general,
solve
problems
anyway.
A
A
Let's
start
there
and
and
then
as
I
I
agree
with
that
perspective,
the
I
will
espouse
the
other,
the
other
perspective
as
like,
well
an
acknowledgement
that,
with
that
everything
can't
be
static
there,
that's
it
does
have
to
change
and
given
that
okay,
then,
let's
lean
into
some.
B
B
They
have
they
created
standards
that
all
their
products.
So
if
you
go
from
a
a
word
processor
to
a
spreadsheet
to
something
else,
they
never
really
changed
very
much,
those
menus
that
we
were
all
used
to
pulling
down
and
they
added
functionalities
in
those
menus.
But
for
30
years
they
were
pretty
much
in
general.
You
know
you
could
on
the
left,
was
save
or
file
or
print
or
something,
and
so
they
evolved
their
product
and
added
functionality
but
kind
of
set
a
standard.
B
They
they've
evolved
over
a
time
and
had
different
menu
bars
up
there,
but
as
a
standard
they've
always
followed
that
and
that
got
a
lot
of
people
that
didn't
know
about
computers,
to
use
them
and
get
more
familiar
with
and
then
expand
to
their
other
product
lines
of
a
spreadsheet
or
a
powerpoint,
or
something
like
that.
So
so
there's
a,
I
think,
there's
a
compromise
on
you
constantly
change.
B
People
can
get
overwhelmed
by
that
because,
if
you
own
a
company,
the
marketing
guys
that
go
out
and
and
market
it
and
sell
it,
inevitably,
no
matter
how
big
a
company
will
sell,
something
you
don't
have
and
we'll
bring
it
back
and
say:
hey
we
got
this
million-dollar
deal
if
we
could
just
have
chickens,
come
out
of
the
screen
and
dance
around
the
desk
and
go
back
in
the
screen
and
just
something
like
that,
and
I
guarantee
them
that
we
would
have
the
chickens
run
in
six
months
and
just
like.
D
B
So
so
the
sales
team
always
sells
more
functionality,
but
that
kind
of
pissed
off
the
other
kind
of
businesses
that
didn't
want.
But
you
just
run
into
business
problems,
so
change
is
inevitable
and,
yes,
we
need
change.
But
overkill
of
change
to
the
meetup
example
created
a
lot
of
people
pissed
off
at
that
product
for
many
years,
and
finally,
people
ignored
and
got
over
with
it,
but
some
businesses
would
have
gone
out
of
business
because
they
did
so
many
changes
when
they
were
trying
to
build
their
market
share.
A
We're
about
at
the
we're
about
it!
This
is
thank
you
for
that
alan.
A
Thank
you
we're
about
at
the
time,
which
means
it's
about
bedtime
for
me,
because
I'm
about
about
out
of
energy,
the
there's
there's
a
ton
of
things
to
talk
about
in
this
with
this
tech,
and
I
really
enjoy
the
question
you
guys
are
asking,
and-
and
I
also
am
realizing-
that,
like
being
a
being
a
nerd
focused
on
focused
on
one
niche,
like
that
I've
gotten
the
luxury
of
kind
of
going
of
going
deep,
int
into
this
and
then
and
while
sticking
my
head
in
the
sand
and
focusing
on
it,
I'm
ignoring
in
large
respects
what
some
of
the
things
that
alan's,
highlighting
like
yeah
but
where's,
the
business
value
again
like.
A
How
are
you
articulating
this
in
business
terms
and
then
what
burton
is
saying
as
well
is
like
hey
we're
you
know
or
the
enviro
this
environment
is,
you
know,
generally
fall.
A
You
know
to
generally
sort
of
follow
suit,
with
the
notion
that
there's
a
series
of
tooling
that
that
you'll
you'll
that
most
generally
want
to
consider
and
use
and
service
mesh,
is
one
of
those
tools,
but
on
the
trail
or
where
does
it
land
on
the
trail
and
the
thing,
and
maybe
next
time
we
talk
next
time
we
have
a
meetup
that
does
talk
about
surface
meshes
is,
I
should
just
do
a
few?
A
Well,
you
know
simple
small
but
concrete
demonstrations
of
like
deploying
linker
d
or
deploying
or
deploying
a
couple
of
different
meshes
with
different
sample
apps
and
showing
the
time
that
it
takes
to
get
to
value
of
using
that.
B
That
would
be
great
side
by
side
yeah,
it's
like
a
service
mesh
and
here's
something
a
small
cluster
of
two
machines
or
three
machines
without
a
service
mesh
and
here's
linker
and
here's
a
couple
of
other
tools
on
top
side
by
side
would
be
great.
They
said
for
someone
to
go,
sell
this
to
their
business.
To
say
we
need
to
spend
more
money
on
this
now.
D
C
B
A
A
Thing
that
might
be,
I
don't
know
if
this
is,
but
here's
the
if
anybody's
got
so
so
there
there
are
sample
apps
that
I
readily
have
access
to
or
use
with
used
pretty
frequently
does
anybody
have
an
app
that's
meaningful
to
them
that
you
might
prefer
that
we
were
to
work
with.
You
know,
do
a
lab
on.
A
You
know
I
figured
I'd
ask
just
because
it
might
make
it
more
real
for
you.
You
know
in
that
sense.
A
If
something
comes
to
mind,
you
know,
let
me
know,
but
otherwise,
there's
like
four
or
five
sample
apps
that
that
we
that
we
do
whole
use
in
our
workshops
and
things
to
show
folks,
but
I
also
recognize,
like
sometimes
your
eyes
will
roll
when
you're
looking
at
the
sample
app
you're
like
yeah.
C
A
What
do
you
know
there
were
no
glitches,
or
what
do
you
know
like
that?
Was
yeah.
A
A
A
No
yeah,
no,
as
a
matter
of
fact
I
mean
just
send
over
your
will,
come
co-author
it
with
me.
Man
come
or
I
know
that
this
sounds
kind
of
funny
and
and
I'll
say
it
funny.
So
you
don't
feel
like
you
have
to
do
it,
but
but
come
by
line
it
with
me.
I'm
serious
I've
they've
been
asking
for
like
three
of
them
and.
C
I
just
that
that
would
be
great,
because
I've
been,
I
took
the
summer
off
basically
of.
C
Well,
not
necessarily
by
choice,
more
of
a
mental
health
break,
let's
say
sure,
but
yeah
so
it'd
be
great
to
to
work
with
you
lee
and
to
contribute
to
the
yeah.
A
A
I
hope
I
hope
it's
helpful.
I
know
some
of
the
and
I'm
frustrated
by
the
lack
of
even
the
the
lack
of
clarity
and
succinctness
around
the
response,
for
what
are
the?
What
are
the
signposts?
What
are
the
indicators
that
signal
to
me
it's
time
for
a
mesh
there's
a
lot
of
literature
out
there,
but
I
think
most
of
it
probably
does
what
you
guys
were
describing,
which
is
it.
It
tells
you
about
all
the
things
and
in
the
process
you
know
loses
all
of
us
and
so.
D
B
And
on
my
screen:
what's
the
you
know?
Basically
I
like
your
idea
of
being
able
to
do
maybe
side
by
record
one
and
time
it
with
the
timer
on
the
bottom.
Here's
the
service
mesh
and
we're
doing
I
mean,
but
I
wouldn't
I
mean
a
demo
that
really
ruins
enough
containers
to
look
like
a
business
value.
So
this
one's
running,
you
know
600
containers
or
whatever,
and
here's
with
the
service
mesh
and
then
here's
a
different
with
linker
on
top
or
whatever,
and
here's
one
without
and
the
business
value
is.
You
can
have
this
information.
A
Yeah
yeah
the
the
tool
that
we
were,
that
we'll
use
to
show
that
that
measury
tool
it
in
an
interesting
way
it
it
takes
a
sample
like
each
of
the
service,
meshes
generally
come
with
their
own
sample,
app
and,
and
it
supports
each
of
them,
and
it
actually
supports
like
taking
one
and
deploying
that
same
sample
app
across
different
meshes,
which
is
kind
of
nice.
Because
then
you
you're,
like
you,
gain
some
familiarity
with
the
sample
app
and
then,
as
you
deploy
it
across
the
other
ones.
A
You
can
kind
of
get
a
sense
of
you
know
the
difference,
because,
because
you've
got
a
consistent,
yeah,
okay,
a
lot
of
times
when
we
do
that,
we'll
talk
about
performance,
which
is,
is
a
point
of
interest,
but
but
actually
just
functionally.
A
What's
the
difference
as
well:
yeah,
okay,
fair
enough,
troy
thanks
for
pointing
this
out.
This
is
I
hadn't
seen
this
before.
A
They've
done
good.
I
oh
actually.
A
A
A
B
C
C
Believe
is
from
a
what
do
you
call
it
a
company
that
does
contracting
sort
of
work,
so
they
do
something
like
evaluation
and
stuff
on
each
of
those
tools
which
is
nice
to
they
write
about
it
and
say:
what's
what's
really
good
and
what's
not
so
good?
Oh.
B
A
No,
I
know,
as
a
matter
of
fact,
no
your
responses
actually
were
the
most
founded
in
reality
for
a
lot
of
organizations.
Reality
is,
I
could
give
you
know
two
craps
about
your
newfangled
whatchamacallit.
A
You
know
the
the
service
master,
the
what
you
know
that,
like
it
doesn't
even
yeah
anyway,
there's
a
the
the
fortunate
thing
for
us
nerds
is
that
digital
transformation
is
a
real
deal
like
in.
If
these
executives
that
are
uncomfortable
with
technology,
don't
don't
get
hip
with
it.
B
Not
not
mirantis
yeah
mirantis,
just
I
think
their
conference
was
this
week.
Maybe
maybe
it
was
this
yeah.
It
was
this
week
and
in
their
keynote
they
said,
yep
there's
a
lot
of
y'all
that
have
small
applications
and
don't
need
this
monster
orchestration.
We
are
not
going
to
drop
swarm,
we're
going
to
maintain
it
for
these
smaller
type
of
operations
that
don't
need
this
huge
orchestration,
and
that
was
pretty
interesting,
because
that
was
an
issue.
Last
year
people
were
investing
in
different.
You
know
that
technology
and.
A
I
do
yeah
yeah
see
sing
it
man
yeah.
I
totally
agree,
but
like
I
mean
I,
I
love
me
some
kubernetes
I
mean
I,
you
know,
I
think
they
all
have
a
place.
I
guess
is
what
I'm
saying.
B
A
B
A
A
B
A
We
didn't
hear
him
tonight,
but
yeah-
oh,
oh,
very
nice,
very,
very
good
to
see
all
of
you
can't
catch
you,
hopefully
in
about
a
month.
Okay,
that's
great
thank
see.
You.