►
From YouTube: DevOps, Idit Levine, solo.io, GeekWire Cloud Summit 2019
Description
Idit Levine, Co-Founder & CEO at solo.io, speaks on the DevOps track at the GeekWire Cloud Summit 2019.
A
Well,
that
will
be
an
interesting
talk
for
me
because
usually
I'm
very
doing
a
technical
demos
and
usually
I
have
a
computer
and
I'm
sharing
stuff
life,
and
that
would
be
different
so
fully
it's
the
right
track
in
you
will
fall
and
I
think
it
would
be
really
good.
So,
okay,
so
your
customer
and
every
customer
today
the
echo
system
is
trying
to
do
some
transformation
by
jewelry
and
most
of
them
the
one
I
talk
you
do
with
no
exception
is
starting
with
migration
right.
The
problem
of
we
all
have
a
monolithic
application.
A
This
is
kind
of
like
an
old-style
application.
We
really
want
to
move
to
micro
services
that
will
give
us
a
lot
of
benefit,
they're
all
trying
to
do
that
and
not
easily
right,
but
once
they
actually
going
to
do
that,
they
will
discover
that,
like
everything
in
your
ecosystem,
what
we're
doing
is
that
we're
solving
one
problem,
what
we
did
using
a
lot
of
new
problem,
because
the
way
we
solve
it?
A
So
that's
exactly
what
you
discover
when
you
move
to
micro
services
because
suddenly
simple
stuff,
like
communication
inside
your
infrastructure,
it's
going
to
be
a
little
bit
more
complicated,
because
now
you
need
to
explain
how
to
create
actually
to
micro
services
and
you
need
to
make
sure
that
it
will
be
able
to
communicate
to
each
other.
You
need
to
make
sure
that
they're
doing
it
securely,
of
course,
and
you're
also
trying
to
figure
out
what
the
heck
is
going
on
in
all
these
micro
services
environment.
A
So
for
that
you're
probably
going
to
use
something
like
service
mesh,
that's
exactly
the
problem
that
service
mesh
is
one
is
trying
to
adopt,
but
what
service
machine
is
doing
is
doing
something
more
interesting.
It's
actually
the
way
we
actually
implement
service
mash,
it's
actually
abstraction
on
your
network
and
if
we
actually
abstract
your
network
days
way
more
stuff
that
we
can
do
and
in
that
case,
I
think
that
a
lot
of
the
health
of
your
application
can
come
with
that.
A
So
that's
kind
of
like
Nellie,
deep
to
all
of
this
and
understand
what
you're
doing
now
at
my
company
solo
solo
I/o.
Are
we
actually
doing
a
lot
of
stuff
with
migration?
We
have
an
API
getaway
call
glue
that
focusing
on
that,
but
I'm
not
going.
This
is
not
a
food
vendor
talk,
so
I'm
not
going
to
talk
about
that
at
all.
A
If
you
want
to
go
check
it
out,
but
let's
understand
what
the
problem
that
we're
doing
so,
as
I
said
when
we
actually,
you
know,
including
micro
services,
there
is
a
huge
improvement
that
we
just
right,
I
mean
productivity.
People
are
going
so
much
faster.
I
can
push
so
much
but
in
the
same
day
an
immediate
needs
in
production.
The
designers
forced
me
to
be
right
right.
All
these
two
Pizza
team.
You
have
to
make
a
small
piece
of
code
that
relevant
to
a
specific
subject.
A
You
encapsulate
your
code
very
well,
you
make
it
very
flexible,
I,
don't
care
which
language
you
use,
which
technology
which
protocol
is
just
going
to
work,
which
is
pretty
awesome,
but
of
course
we
introduced
a
new
problem.
So
now
everything
is
in
the
network
and
network
is
a
complicated
stuff.
So
that
means
that
now
go
figure
out
the
dependency.
There
is
a
performance
always
latency
in
network
right.
How
do
you
figure
out?
Where
is
the
problem?
Is
inside
the
code?
It's
inside
the
pipe.
Where
is
the
problem?
Cost?
Of
course,
I'm
really
now
figure
out.
A
You
need
to
change
a
little
link.
Nothing
is
going
to
work
anymore.
Now
everything
is
distributed
all
over
the
place
and
I've.
Also
testing
go,
go
figure
it
out
to
test
something
like
that.
That
is
spread
it
around.
So
that's
basically
the
problem
of
video
anyway,
right
now
you
have
the
one
monolithic
application.
It's
pretty
simple:
let's
take
a
use
case
for
something
like
debugging
I'm,
just
attaching
a
debugger
and
see
what's
going
on
in
this
application,
but
now
I
have
more
application
and
the
code
is
spread
it
all
over
the
wrong.
A
So
all
the
stage
of
my
application
is
actually
spreaded
around
and
then
we
discover
that
there
is
way
more
than
six
micro
servers
that
are
shown
in
the
previous
slide.
This
is
what
something
some
companies
like
Netflix
and
Twitter
is
running
today
more
over
500
micro
services.
So
it's
way
way
more
complicated,
all
of
those
to
communicate.
A
That's
like
oh
wow,
and
then
you
see
something
like
that
in
Twitter,
right
people
are
sitting
and
how
they
all
do
we
debug
that
right
now,
how
do
you
actually
finding
out
what
to
do
with
this
and
and
now
you're
stuck,
is
different
right
before
you
add
monolithic
application,
you
have
you're
using
something
like
a
PM
for
the
matrix
for
logging,
maybe
Splunk.
It's
just
an
example.
Right
I,
don't
know.
Native
debugger
is
something
that
you
will
use
to
debug
today.
We're
not
debugging
anymore
with
troubleshooting
architecture
will
be
something
that
service-oriented
architecture.
A
Deployment
deployment
would
be
something
like
chef,
puppet
and
Siva.
Maybe
you
will
build
it
with
terraform
testing.
Kids
will
be
80%
unit
test
and
under
20%
integration,
but
now
it's
all
that
different
by
you
will
use
something
like
Prometheus,
because
now
you
need
load
and
scale,
and
you
didn't
manage
to
end
all
of
this
and
then
when
you're.
Looking
at
you
know
on
its
way,
it
is
something
like
clogging.
Now
your
logging
is
spreaded
all
over
run.
You
have
replicas
I,
don't
know
what
to
collect
them
and
to
make
sense.
A
So,
for
that
you
need
some
transactional.
Logging
I
need
to
put
somebody
in
this
and
then
make
sense
of
it.
That's
something
like
operate
elementary
or
open
tracing.
It
was
before
you're
naughty
bugging
anymore.
That's
not
what
we
do
in
your
micro
services
anymore,
one
of
the
things
that
we
did
in
my
company.
It's
created
a
orchestrated
for
debugger
in
order
to
make
that
actually
possible
micro
services,
your
infrastructure
and
is
your
protector,
and
you
will
use
something
like
elm,
for
instance,
or
any
other
development
specification
and.
A
All
about
the
integration,
because
that's
all
what
you're
running
right
now
so
again,
you
have
this
great
monolithic
application
now
micro
services.
Now
the
network
is
the
piece
that
is
very
problematic.
The
question
that
I
asked
before:
how
do
you
actually
make
sure
that
they
communicated?
How
do
you
make
sure
that
they're
doing
it
security
and
how
do
you
make
sure
that
you
can
actually
see
what's
going
on
inside?
The
cluster?
Is
the
big
three
problem
that
everybody
trying
to
solve
and
they
do
it
called
service
smash.
So
again,
what
is
service
smash?
A
Service
mash
is
basically
solving
that
exactly
tree
problem
in
term
of
security.
It's
giving
you
it
certificate
MPLS
between
the
micro
services
policy,
making
sure
that
everything
is
really
secure
in
terms
of
routing
again
it's
using
proxy
as
a
sidecar,
and
it's
basically
making
sure
that
you
will
end
all
all
the
routing
and
communication
between
them
and
in
terms
of
observability.
He
knows
because
of
his
special
architecture.
You
can
actually
spin
off
the
lock
to
a
location
and
they'll
end
the
matrix
to
make
sure
that
you
will
see
better,
ok,
so
again,
a
little
bit.
A
I
will
explain
how
it
works
too,
will
kind
of
like
level
out
everybody.
We
have
a
service
reporting,
but
caller
psycho
psycho
is
basically
a
proxy.
In
that
case,
it's
envoi
coming
from
the
left,
guy,
open
source
project
in
the
C
and
C
F,
and
what
it's
doing
it's
pretty
damn
right.
It
doesn't
know
what
to
do
much.
The
only
thing
that
we're
doing
is
we
tricking
the
application.
This
case
microservices
that
every
communication
to
the
application
and
from
the
application
we
playing
some
with
the
IP
table
is
going
to
go
to
this
proxy.
A
So
it's
basically
everything
is
going
to
him,
and
now
you
need
to
give
in
configuration
because,
as
I
said
it
pretty
dumb,
it
doesn't
in
order
to
do
the
stuff.
Someone
need
to
tell
him
what
to
do
and
therefore
you
need
some
control
print.
You
probably
will
need
one.
You
probably
will
need
one
in
the
API
gateway.
That's
something
that
coming
outside
the
cluster.
A
A
Ok,
so
that's
great!
So
right
now,
I!
Hopefully
everybody
understand
what
it
server
smash
again:
I'm
a
big
fan
of
this
technology
and
apparently
I'm,
not
the
only
one,
because
that's
everything
that
everybody's
talking
about
in
2019
right,
it's
like
service
man,
she's
going
to
be
the
future.
It's
going
to
renovate
everything
that
we
need.
It
solve
a
lot
of
problems
and
therefore
everybody
want
to
get
it
turned
to
play.
So
the
first
one
that
came
there
is
link,
add
a
link
here.
A
This
came
from
the
boy
and
guy
to
start
up
the
first
version
that
they
did
was
horrible.
The
sidecar
we
like
to
call
it
the
site
bus
because
it
was
so
big,
was
very
bad
Java
implementation,
but
then
sto
the
Google
guys
said.
Well,
that's
nice
important
to
fix,
but
to
be
fair,
it's
just
bad!
So
we're
going
to
create
our
own
version.
They
took
a
sari
and
void
as
the
sidecar
linker.
A
They
strike
to
fight
the
creator
on
an
option,
a
better
implementation
console
Connect
which,
from
the
Asha
guys
came,
they
really
want
to
focus
on
security
and
so
on
and
so
on
and,
of
course,
any
cloud
that
respect
himself
will
be
invaluable,
including
AWS.
So
now
mr.
customer
really
excited
about
the
opportunity
of
the
service
match,
which
one
am
I
going
to
pick.
What
is
the
difference
with
the
best?
A
That's
a
good
question,
so
what
people
are
doing
or
most
of
my
customer
is
doing,
they
are
exploring,
they
don't
know
so
they
starting
with
sto
they're,
looking
they're
playing
they're,
taking
a
team,
they're
learning
the
technology
three
months
later.
The
discover
it's
too
hard
because
it
is,
it's
really
really
complicated,
so
they're
going
to
move
to
Lincoln
D.
A
There
will
be
certain
it's
pretty
simple
and
working,
but
here's
the
thing:
it's
not
using
Android
as
a
sidecar
and
if
we
becoming
this
important
and
commodity
proxy
in
this
ecosystem,
there's
huge
community
around
it
will
continue
the
computer
as
well
to
avoid
and
therefore
we're
going
way
faster
for
them.
So
again
they
don't
want
to
bet
on
this,
so
they
will
go
with
console.
They
discovered
the
concept
they
didn't
even
do
layer
seven
yet,
and
everything
about
service
mesh
is
to
bring
those
network
to
the
layer
of
the
application.
A
There
were
seven,
ok,
I,
don't
know
if
it's
interesting
the
right,
so
they
will
move
to
a
patch
and
so
on
and
in
the
end
of
the
day
you
came
by
they're,
not
reading
your
service
smash
with,
which
is
really
a
shame.
So
we
thought
about
how
we
can
help
them
and
the
way
I
thought
is
what
if
they
were
only
going
to
need
to
learn
one
API.
What
if
the
technology
behind
it
will
be
just
technology?
A
We
don't
care
right,
it's
just
going
to
do
what
I'm
expecting
him
to
do,
and,
instead
of
being
so
complicated
like
sto,
because
I
still
coming
from
Google
engineers,
they
have
to
use.
You
know,
H
cases
that
happening
in
the
Google
infrastructure.
Most
of
the
people
doesn't
have
this
problem
in
their
infrastructure.
So
what
if
we
can
come
with
this
abstraction
layer
that
will
make
it
easy?
First
of
all
at
least
they
don't
need
to
learn
all
those
technologies
they
only.
B
A
To
now,
of
course,
by
the
way,
that's
the
big
cloud,
but
then
companies
like
Kong
and
other
that
an
API
get
where
they
have
to
play
there
right.
They
make
it
very
uncomfortable
for
them.
So
everybody
is
a
mess
right
now,
they're,
not
really
amazed
but
their
mess.
So
the
point
is
that,
in
the
end
of
the
day,
we
need
to
abstract
it.
A
A
So
that's
basically
what
we
thought
that
we
should
have
done,
and
we
did
it
and
lucky
ass,
Microsoft,
agree
and
Microsoft
is
a
great
company
and
it's
a
big
company,
not
like
a
small
one,
so
they
managed
actively
to
get
all
the
community
around
it
and
basically
launch
it
together.
We
call
it
service
mesh
interface
and
what
service
machine
difference
making
you
again
it's
an
open-source
project
is
the
community
project.
It's
basically
giving
you
the
ability
to
use
whatever
service
mesh.
You
want,
but
one
interface
for
that
and
this
interface.
It's
simple.
A
You
know
how
to
adapt
in
order
to
communicate,
so
it
doesn't
really
matter
which
one
you're
using
you
can
start
right
now.
It's
also
make
it
way
simpler,
so
the
beautiful
of
Microsoft
is
that
they
can
actually
make
something
like
that
happen
right.
We
can
write
a
code,
but
they
need
someone
to
actually
be
able
to
grow
influence,
Asha
cope
to
help
and
go
into
influence
Boyan
to
help
and
bring
all
of
us
together,
credit
and
all
the
guys
basically
make
sure
that
there's
a
community.
We
can
build
something
like
that.
A
Okay,
so
we
have
this
thing,
this
interface
and
we
have
all
those
services
managed
and
we
can
actually
use
it.
It's
pretty
awesome,
but
there
is
something
more
interesting
in
service
mesh
and
that's
the
way
it's
actually
working,
so
service
mesh
is
not
forming
only
to
be
a
problem.
That's
what
happening
today,
but
think
about.
What's
special
about
I
will
build
it
before
that.
A
Back
on
the
day
we
wrote
code
in
the
car
we
put
operation
code
example:
security,
routing
you
want
to
put
your
log,
you
need
to
put
some
library,
you
want
to
do
curse,
engineering,
you
put
library,
everything
is
important
library.
Now,
if
your
provider
is
upgrade
this
library
guess
what
you
need
to
have
a
new
version
for
your
micro
services,
even
though
you
didn't
change
the
code,
even
so,
the
question
is:
is
that
the
right
way
to
do?
A
And,
of
course,
not
and
like
everywhere,
the
best
way
to
do
it
is
actually
the
couple
and
that's
exactly
what
we
did
here
in
sir.
Basically,
what
we're
saying
these
cycles
responsibilities
to
be
in
charge
of
this
operation?
That's
their
job
and
the
business
logic
of
the
micro
service
need
to
be
the
business
logic,
and
that
could
be
an
application
team
that
writing
it.
They
don't
need
to
talk
to
the
operation.
It
should
be
differently
and
now
we're
bringing
back
the
power
to
the
operation,
which
I
think
should
be
there
right.
A
This
is
why
those
people
so
experienced
I
need
to
be
there.
So,
if
we're
doing
this,
there
was
way
more,
we
can
do
and
that's
what
I
think
is
special.
This
is
why
we
needed
this
interface,
because
we
want
to
put
stuff
on
top
of
it
right
and
because
this
is
like
service
mesh
will
become
commodity,
but
I
want
to
start
and
innovate
before
this
thing
is
settle
and
for
that
I
need
some.
You
know
place
that
I
can
actually
count
on
it
being
there,
and
this
is
this
interface.
So
what
can
we
put
there?
A
So,
of
course,
API
gateway
right
I
mean
in
the
end
of
the
day.
The
full
solution
like
I
show
should
be
the
service
mesh
that
will
take
care
of
the
communication
between
you
put
inside
the
cluster,
but
you
also
need
an
API
gateway
and
that's
very
important
and
they
need
seamlessly
to
work
together
and
they
need
to
understand
each
other
and
be
aware
of
each
other
and
that
one
thing
that
we
need
to
do.
The
second
thing
is
that
we
have
a
lot
of
ideas
on
what
we
can
do.
A
I
can
give
you
an
example
of
something
that
we
build,
which
is
basically
an
ability
we
did
it
together
with
lyft.
It's
basically
the
ability
to
record
every
request
that
coming
to
your
infrastructure
and
because
it's
going
through
those
sidecar,
we
can
actually
record
it,
but
we
don't
have
to
use
the
ones
that
are
good
now
think
about
it.
It's
mean
that
if
it's
going
to
the
database,
if
it's
going
to
the
s3
all
this
communication
in
and
out,
we
can
record
if
we
can
recall
in
the
end
of
the
transaction.
A
If
everything
good
we're
not
going
to
use
it,
we
just
want
to
toss
it
right.
We
don't
care
about
everything
that
is
good,
but
if
something
is
wrong,
we
can
actually
take
it
the
recorder
and
then
spinning
up
outside
the
production,
inject
everything
that
was
in
production
and
replay
any
bug
that
was
there.
This
is
an
example
of
work
that
we
did.
A
A
My
vision
for
it
is
like
iPhone
Android
right
in
the
end
of
the
day,
is
an
iPhone
right,
yeah,
it's
a
platform
and
in
the
platform
you're
getting
a
lot
of
stuff
you're,
getting
whatever
discoverability
and
a
lot
of
other
stuff,
and
this
is
what
you're
getting
but
you're
some
getting
some
application
by
basic
and
those
applications.
You
know
like,
for
instance,
your
calendar
or
your
email
you're,
just
getting
them.
I
think
that
we
should
get
those
two
building
block,
which
is
API
gateway
in
service
smash,
because
that's
the
building
block
to
abstract
your
network.
A
So
now,
like
an
iPhone,
if
you
have
this
ability
to
do
this,
I
think
what's
powerful
about
the
iPhone
is
the
fact
that
two
application
can
communicate
to
each
other
right,
like
I'm,
putting
a
calendar
invite-
and
it's
already
talking
to
my
Google
map.
That's
one
thing,
and
the
second
thing
is
that
it
can
be
extendable
by
the
app
store
and
that's
exactly
what
we
should
do
right.
We
can
just
put
something
like
Carolingian
ink
on
top
of
it
and
not
put
the
library
inside
your
code.
A
You
know
why,
because
you
can
just
inject
it
on
the
layer
on
the
proxy,
you
can
inject
latency,
you
can
inject,
you
know
you
can
inject
for
it's
whatever
you
want.
You
have
the
ability
to
do
the
right
now,
the
powers
in
your
hand
and
something
like
in
that
case
Canary
deployment.
You
want
to
make
sure
that
it's
you're,
the
new
version
is
working
well,
be
smart
about
it
put
a
bit.
You
know,
traffic,
shifting
and
so
on,
and
so
what
you
look
with
them:
I,
usually
I'm
doing
life,
but
now
it
will
be
this.
A
A
One
of
them
is
a
service
match
for
my
AWS
app
mesh
and
then
the
other
one
is
link
FD,
and
there
is
the
extension
that
we
could
actually,
hopefully,
that
we
work
yeah
and
then
I
can
install
another
service
match
something
simple
like
a
steel
and
only
thing
that
I
need
to
give
is
the
namespace
and
which
version
I
want.
That
is
just
going
to
have
to
work
very,
very
simple.
A
So
this
is
the
basic
one
that
I
need
in
my
infrastructure,
right,
I'm,
choosing
a
platform
and
click
on
this,
and
in
a
second
you
will
see
boom.
You
have
the
service
meshes
install.
So
now
you
have
sto
it's.
They
discover
everything
all
the
services
that
is
managing
all
the
configuration
is
small
smart
about
it
and
the
reason
it
should
be
smart
is
because
I
wanted
to
be
able
to
actually
go
now
and
install
an
extension
of
it
and
the
extension
can
be
something
like
Keala
from
reddit,
which
is
again.
A
Open-Source
is
giving
me
visibility
on
the
cluster.
It
could
be
something
like
an
API
gateway
like
ours,
glue
or
others.
It
could
be
something
like
Canary
deployment,
like
flagger
from
we
work
and
so
on.
So
the
idea
is
like
there
is
community
that
working
on
top
of
which
we
can
give
you
everything
you
want,
because
you're
going
to
use
the
server
smash,
so
I
mean
I
will
go
for
the
I,
don't
know
if
I
can
move
it
fast.
A
But
basically,
as
you
can
see,
your
key
Ally's
install
right
now
is
an
extension.
I
can
go
and
stall
others
and
you
can
put
your
own
extension
and
we
can
actually
as
a
collaborate
as
a
community.
The
second
it
will
finish,
I,
don't
know
that
I
can,
with
my
clicker,
to
move
it
so
I'm,
sorry
about
that
to
take
only
a
sec
and
that's
basically
the
idea:
okay,
okay.
So
what
next?
Okay?
So
it's
talking
about
this
future?
A
It
is
the
future
from
a
lot
of
people,
but
actually
there
is
a
lot
of
people
wanting
it
in
production.
This
is
the
future.
It's
something
that
it's
worth
exploring,
but
you
know
it's
like:
every
new
come
new
technology.
There
is
a
lot
of
gap.
We
solve
those
problem
years,
then
you
want
to
adopt.
So
here
isn't
some
example:
one
service
much
it's
really
really
hard
to
configure.
This
is
a
treat
for
some
of
the
guys
of
St.
Arnaud
basically
said
you
sort
kubernetes
it's
hard
to
do
not
try
to
do
with
last
year.
A
Good
luck
and
it's
right.
It's
hard
now
I
think
that
the
service
match
hub
will
help
you
with
a
lot
of
stuff.
The
SMI
will
help
you
with
a
lot
of
stuff
because
we
make
it
simpler
already
as
a
community,
but
we
can
do
more
when
you
actually
install
service
smash
to
the
a
you
know
when
you're
installing,
for
instance,
you
wanted
to
install
whatever
something
in
docker
you're,
not
starting
from
scratch.
A
When
you're
installing
service
smash,
the
configuration
is
starting
from
scratch,
so
we
can
take
advantage
for
the
some
experience
from
docker
and
basically
come
with
the
bundles
of
configuration
that
the
community
will
share
between
each
other
and
then
what
I
can
say.
I
want
this
service
measure,
one
of
those
attention
and
one
this
configuration
and
then
I
can't
week.
But
it's
already
build
layers
that
way
that
I
will
be
able
to
get
close
enough
to
where
I
need
it.
A
I
don't
need
to
start
from
the
beginning,
I
think
that
would
be
very
powerful
and
the
last
is
so
stay
tuned,
we're
working
on
it
and
it's
really
close
and
I
believe
that
it
will
be
eventually
a
plug
in
for
CI
CD
system.
The
second
thing
is
that
okay,
so
this
is
great
I.
Have
this
amazing
technology
in
my
infrastructure,
no
problem
right:
the
application
developer
is
actually
going
to
develop
these
things,
and
then
it
will
do
this.
A
This
is
an
example
for
a
configuration
llaman
configuration
that
you
need
in
a
very
simple
a
it's
called
the
the
bookstore.
It's
their
demo,
application
for
Sto,
it's
the
basic
one
right,
it's
really
basic!
So
what
you
can
see
here,
you
need
to
create
a
repository
for
my
steel
that
actually
has
configuration
for
all
of
this
and
then
I
just
click,
for
example,
on
the
networking
you
need
to
do
more
Yama
file,
so
all
of
this
is
yummy.
Fun
now
will
be
coming.
A
You
know
that's
what
we're
doing
we're
basically
dealing
with
tons
of
Yama
fire
and
trying
to
configure
that
do
we
really
expect
the
application
on
them
right
now
to
actually
understand
service
match
with
all
this
mesh?
No,
why
did
we
do
that?
We
did
it
in
order
not
to
take
this
power
from
them.
So
the
question
is
what
we
can
do
there
and
again
we're
working
on
it.
I
think
it
would
be
very
interesting,
so
I
mean
hopefully
it
was
interesting.
A
B
Thank
city
that
was
that
was
interesting
and
as
someone
who's
been
talking
about
proxies
and
the
power
of
a
proxy
and
all
the
cool
things
you
could
do
at
the
proxy
and
how
the
proxy
is
the
right
place
to
add
a
ton
of
value
that
you
don't
want
in
your
application.
In
your
business
logic,
it's
kind
of
slightly
warming
to
my
heart
to
see
this
sort
of
proxies
making
their
way
down
into
the
micro
service
space,
but
it
is
hard
you're,
right
and
I.
B
Think
as
I
was
listening
to
you
talking
listening
to
Tara
talk
and
listening
to
Nell
talk
that
there's
a
couple
of
things
that
have
to
happen.
I.
Think
to
take
all
this
amazing
potential
and
technology
to
be
applicable
to
everyone
to
be
applicable
to
kind
of
the
people
that
don't
have
a
thousand
engineers
on
tap
and
I.
Think
we
need
to
get
our
our
our
organizational
and
our
mindset
and
our
community
space
and
the
way
we
work
with
each
other
inside
organizations
together.
And
then
we
need
tools
like
this
to
make
take
away
the
complexity.
B
You
know
we
have
to
bear
to
operate
this
stuff
without
you
know,
you
know
a
thousand
PhDs
lying
around,
because
there's
tons
of
companies
out
there
that
just
can't
recruit
that
kind
of
talent.
You
know
we're
not
all
Google
in
when
not
all
these
people
so
really
really
interesting,
talk
to
hear
how
these
come
together
and
we
do
have
time
for
questions.
If
anyone
would
like
to
ask
one
if
not
I
hear
I
have
a
couple
more
questions.
A
A
B
Part
of
understanding,
quite
all
the
time,
so
one
thing
I
wanted
to
ask
is
where
you
know:
where
do
you
see
the
line
being
drawn
of
what
you
put
in
the
sidecar
proxy
layer
and
what
you
put
in
you
in
your
application?
What's
the
logical
kind
of
boundary
point,
what
should
I
never
put
in
my
app
and
what
should
I
always
have
in
my
in
my
I
mean.
A
Everything
we
can
move
to
that
sidecar
they
beautiful
about
the
sidecar
is
that
it's
configurable,
there's
no
code
involve
its
yeah.
You
know
it's
easier
in
a
point
and
therefore
everything
that
we
can.
We
should
put
there
besides
the
really
business
logic
of
the
application,
what
it's
doing
that
specifically
to
what
it's
trying
to
achieve
and
nothing
that
related
to
how
its
communicated
or
all
the
logs
is
going
to
work
or
to
troubleshoot
that's
all
operation,
and
that
should
be
in
the
sidecar.
A
A
Mean
I
mean
it's
a
good
question:
I
think
that,
as
still
as
Google
behind
them
and
they're
making
a
lot
of
marketing
noise
but
I
think
they
need
to
work
on
their
implementation,
because
it's
a
mess
right
now
they're
getting
to
the
right
place.
Hopefully
it
will
take
them
so
I
I
will
bet
on
them.
Even
though.