►
From YouTube: Keynote: Adobe’s Internal Developer Platform Journey and Lessons - Srinivas Peri, Adobe
Description
Keynote: Adobe’s Internal Developer Platform Journey and Lessons - Srinivas Peri, Director of Ethos + Cloud Platform Engineering, Adobe
In this session, Srinivas Peri of the Adobe Cloud Foundation (Ethos) team will share the story of Adobe's journey to the cloud. Lessons learned, mistakes made, and the software built to run thousands of services from hundreds of teams on shared infrastructure in a secure and compliant fashion. Then we will look at how adoption of CNCF in general and Argo in particular is allowing the Ethos team to streamline and optimize offerings and operations while providing increased flexibility and control to our service teams, all without compromising the guardrails that make our platform successful.
A
All
right,
hi
everyone,
how
are
you
all?
Almost
there?
Almost
there
all
right,
I
am
very
excited
to
be
a
speaker
for
this
conference.
I
remember
last
year
the
first
one
virtual
ayah
from
my
team
did
the
presentation
and
it
was
one
of
the
best
sessions
rated
right
there
very
good
job,
and
then
we
wanted
to
better
than
that
this
time.
Right.
So
guess
what
I
think
there
were
a
lot
of.
I
learned
a
lot
just
in
the
last
two
days
talking
to
a
lot
of
you
guys,
really
really.
A
I
want
to
congratulate
each
and
every
one
of
you.
This
is
a
very,
very,
very
big
milestone.
Congratulations
guys!
All
right!
I
am
srinivas
peri,
I'm
a
director
of
ethos,
cloud
platform,
and
today
I'm
going
to
share
here
the
story
of
what
we
learned
all
right.
This
lights
are
moving,
I
guess
all
right
yeah.
So
before
we
go
deep
into
the
technology,
I
think
let
me
just
hide
the
high
level
we
I
have
thousands
of
services.
A
Adobe
has
lots
thousands
of
services
right
where
are
all
these
services,
so
there
are
three
clouds
document
cloud,
creative
cloud
and
the
experience
cloud
which
are
powered
by
lots
of
products
and
services
and
which,
in
turn
depends
on
lots
of
platforms,
machine
learning,
platform,
content
platform,
data
platform.
All
these
services
are
built
on
top
of
ethos,
which
you
see
at
the
bottom,
which
is
built
on
top
of
aws
azure,
and
we
also
have
a
data
center.
A
So,
let's
pause
here
for
a
second
right,
I
think
when
we
say
developer
experience,
what
exactly
do
we
mean
by
developer
experience
right?
Let's
take
some
personas,
so
you
are
a
developer
and
you
just
started
in
a
new
company
or
started
a
new
project
and
you
have
to
get
going
right,
so
you
probably
are
thinking
about
which
technical
language
you
should
use
and
how
do
I
get
started?
A
What
libraries
I
should
use?
What
are
the
secure
frameworks
and
everything
that
relates
to
about
the
local
development?
So
what
do?
What
are
we
doing
right
now
to
solve
the
pain
for
the
developer
right
at
adobe?
We
have
some
of
the
some
we.
We
have
a
tools
to
generate
the
initial
thing
and
we
also
have
developed
asr.
A
That
is
what
this
all
about,
and
that's
where
we
are
using
argo
a
lot
and
we
will
talk
about
it
and
last
one
operate
and
improve
again,
I
would
say
we
have
a
lot
of
tools
and
lots
happening,
but
still
again,
glass
is
half
full
and
half
empty
and
I'm
looking
forward
to
learning
a
lot.
What
intuit
has
built
around
mlaps.
A
All
right,
so,
let's
talk
about
little
bit
of
a
journey
that
we
have
gone
through
in
the
last
few
years.
There's
also
a
lot
of
journey
even
before
2015,
which
I'm
not
going
to
talk
about
because
there's
not
enough
space
there.
Okay,
so
2015.
If
you
think
about
what
happened
in
2015
right,
if
you
have
followed
adobe
is
one
of
the
successful
companies
who
have
already
implemented
the
subscription
and
wall
street
has
already
started
rewarding
us.
We
are
already
there.
We
have
a
cloud
company.
Everything
is
working
right.
A
So
what
is
there
like?
So
then?
It
was
all
done
by
500,
plus
aws
accounts,
10
000,
plus
jenkins
machines.
It's
all
working
right,
okay!
So
let's
do
something
about
it.
So
we
said:
okay
containers
are
also
coming
along.
Okay,
let's
start
a
platform
right.
We
started
our
platform
looked
around
and
saw
two
options:
right:
vcos
and
kubernetes
dcos
is
used
by
a
lot
of
big
companies,
apple
and
everybody
else
and
communities
nation
stages.
A
A
whoo
big
deal
right.
I
will
tell
you
a
little
bit
about
how
it
has
helped
us,
so
we
built
an
abstraction
for
developer
experience
and
we
got
all
the
dc
os
clusters
platform
is
all
building
right.
So
that's
great.
So
what
are
the
big
lessons
that
I
want
to
learn
from
here
is
built
in
abstractions
when
you
have
to
okay,
let's
move
on
building
a
platform
is
one
thing
and
getting
it
adopted
by
everybody
else.
Is
at
another
thing
right.
A
So
what
did
he
do?
We
have
done
something
what
we
called
as
a
platform
champion
right,
so
pretty
much
in
every
product
team
we
went
there,
we
sat.
We
went
to
that
location,
we
made
a
road
trip,
went
there
sat
down
for
a
week
and
then
we
told
them
that
we
are
coming
and
then
block
your
sprint
come
down,
come
here
to
your
laptop
and
they
leave
the
week
with
one
of
the
service
working
right.
A
So
that's
how
we
were
in
a
single
calendar
a
year
we
were
able
to
get
the
number
of
services
from
two
digit
to
four
digit
thousand
plus
services
in
a
single
calendar
year
in
a
dcos,
oh
cool,
awesome,
awesome,
awesome,
awesome,
great
story.
We
were
all
clapping
exactly
like
this,
then
what
happened
is
kubernetes
matured
guys
kubernetes
has
matured.
So
it's
time,
it's
time
to
oh
kubernetes.
What
about
this
kubernetes?
We
should
do.
Then
our
abstraction
came
in
very,
very
handy.
We
have
used.
A
We
have
built
a
lot
of
migration
tools
and
again
we
call
it
as
a
migration
as
a
product
and
we
were
able
to.
While
everybody
was
deploying
while
it
is
flying.
We
shifted
the
dcos
clusters
into
the
kubernetes
clusters,
thousands
of
services
right,
big,
big,
big
achievement
for
all
of
us,
okay,
which
is
great
okay,
that
clap
again
clap
again,
please
exactly
that's
that's
what
we
were
clapping.
Also,
then
what
happened
is
so
with
kubernetes.
We
were
also.
We
could
do
something
that
we
did
not
do
with
dc
voice
right
with
kubernetes.
A
We
gave
two
offerings
internally.
We
call
them
as
a
cast
and
pass
and
rohan
also
published
a
blog
around
this,
so
essentially
think
of
cash
as
a
paved
path
and
pass
go.
Wild
diy
give
your
name
space,
do
whatever
you
want
and
we
could
not
keep
up
with
the
pace
of
the
kubernetes
innovation
and
all
the
stuff
that
we
did
for
the
abstraction
behind.
It's
all
homegrown
tools
right,
so
we
had
two
offerings.
A
We
said:
go
use
the
diy
pass
path,
right,
guess
what,
while
this
all
is
happening,
the
number
of
services
on
the
paved
path
was
3000
plus
to
celebrate,
but
the
number
of
percentage
of
the
name
spaces
that
are
on
the
pass
are
like
80
of
them,
essentially
the
problem.
When
we
started
when
we
have
a
lots
of
jenkins,
everybody
is
doing
diy.
Guess
what
the
there's
a
spinnakers
out
there
there's
everything
out
there,
and
then
we
have
that.
A
Why
why,
and
we
looked
at
it,
gitops
is
coming
up
cool
awesome.
What
is
the
best
qtops
tool
out
there
right
and
then
definitely
we
want
to
this
time,
learn
from
what
we
have
learnt
in
the
past
and
then
kubernetes
and
then
kubernetes
native
and
all
that
stuff.
So
we
put
together
a
hackathon,
we
put
together
some
teams
and
then
we
were
able
to
build
a
solution.
I'm
happy
to
tell
you
all
that
we
are
going.
Ga
in
three
weeks
is
that's
where,
before
you
guys
clap
there
are
two
important
points.
A
I
want
to
kind
of
mention.
Okay,
whenever
you
do
a
project
like
this,
it's
not
a
a
fancy
technology
you're
going
in
a
room,
and
then
you
come
together,
hey
product,
you
want
to
take
it.
It
doesn't
work
that
way.
You
need
to
work
with
your
lighthouses.
I
want
to
call
out
if
the
lighthouse,
my
team,
lighthouse
teams,
are
here
my
two
lighthouse
teams,
which
are
sitting
right
there,
their
support
without
their
support.
It
does
not
happen
right
from
the
get
go.
They
are
telling
us.
This
is
what
we
need
this.
A
A
Second
thing
I
want
to
call
out
is
think
about
the
engineering
team
that
is
going
through
this
transition.
They
build
a
lot
of
software
and
after
some
time
they
are
moving
on
to
the
next
one
right.
It
takes
a
lot
of
guts
and
energies
to
be
able
to
don't
attach
it
to
the
stuff
that
you
are
doing
it
as
you
moving
along.
Do
it
enough
and
then
move
on
right,
I
I
want
all
the
my
adobe
argonauts
to
get
up.
Let's
give
them
a
big.
Please
stand
up.
A
A
All
right
is
it
feeling
good
so
far,
okay,
all
right!
Let's
keep
going
again.
I
think
there
are
a
few
more
things
that
I
want
to.
I
want
to
go
over
as
well.
I
hope
this
is
moving
on
yeah,
so
I
wrote
a
paper.
Please
go
and
take
a
look
at
it.
I
think
each
one
of
these
box
I
touched
on
some
of
them.
If
I
really
have
time,
oh
there's
still
two
more
minutes,
I
can
touch
on
one
of
them,
so
my
favorite
one
is
perfection.
A
Perfection
versus
progress
right.
I
struggle
a
lot
in
conveying
and
in
general
these
things
are
that
part
of
it.
I
think,
making
a
good
enough
progress.
I'm
not
saying
don't
write
unit
test,
don't
write
an
integration.
Please
do
all
of
that,
okay,
but
just
enough
so
that
you
are
able
to
kind
of
make
progress
on
it.
One
right.
I
think
that
is
some
of
the
important
things
that
we
have
to
learn,
because
some
of
them
may
not
stick
some
of
them.
We
have
to
go
away
right
and
for
the
developers
who
are
just
starting.
A
I
think
it
is
very,
very
important
that
the
way
technology
is
all
moving
along.
We
they
have
to.
We
have
to
invest
on
both
breadth
and
depth.
Take
one
thing:
scaling
deployment,
which
is
one
of
them,
go
deep,
deep,
deep
into
it
open
the
source
code.
Whatever
is
needed,
it
takes
10
000
hours
to
be
able
to
build
that
subject,
matter,
expertise
right,
you
need
to
build
one
of
them
and
you
need
to
also
build
a
depth
breadth
right
because
we
have
to
be
on
on
call.
A
I
mean
when
the
thousand
plus
developers
or
5000
plus
developers
are
asking
you
these
questions,
and
then
we
can't,
I
mean
we
have
to
rotate
things.
We
should
know
we
should
be
using
our
product,
so
I
think
I
think
we
are.
We
are
going
decently
well
there,
but
I
think
it's
an
ongoing
challenge
for
us.
A
There's
a
lot
there's
a
lot
we
have
to
kind
of
as
we
move
along,
I
think,
hopefully,
by
next
argo
con
when
I
come,
I
will
put
a
check
mark
on
the
second
one
and
probably
the
third
one
will
be
green
or
maybe
even
check
mark.
So
I
think
there
are.
There
are
quite
a
few
things.
There
are
at
least
four
things
that
we
don't
have
right
now
and
then
we
will
be
thinking
about
building
it
as
we
go
along
right.
A
We
still
are
the
infrastructure
provisioning
and
as
far
as
the
the
stuff
that
cross
plane
and
us
are
trying
to
address,
I
think
we
have
a
terraforms
all
over
and
then
that
is
something
that
we
need
to
do
about
it,
and
we
don't
have
a
unified
portal.
We
did
not
bring
all
of
them
together
and
backstage
is
still
not
cutting
for
us.
A
I
wish
there's
still
a
journey
there,
so
that
those
are
the
things
that
we
need
to
kind
of
do
and
what
ed
talked
about
and
what
you
guys
talked
about:
ai
and
observability
right
for
sure.
That
is
a
space
that
we
have
to
invest
on.
So
I'm
saying
please
help
us
help
each
other.
Let's
make
this
community
even
better,
even
bigger,
and
we
are
doing
for
those
of
you
who
registered
coming
to
adobe
office
today.
Tomorrow,
please
participate,
we
are
doing
some
brave
attempt.