►
From YouTube: SIG PM Intro - Ihor Dvoretskyi, Cloud Native Computing Foundation & Aparna Sinha, Google
Description
Want to view more sessions and keep the conversations going? Join us for KubeCon + CloudNativeCon North America in Seattle, December 11 - 13, 2018 (http://bit.ly/KCCNCNA18) or in Shanghai, November 14-15 (http://bit.ly/kccncchina18).
SIG PM Intro - Ihor Dvoretskyi, Cloud Native Computing Foundation & Aparna Sinha, Google (Any Skill Level)
About Aparan
Aparna Sinha leads the product team at Google for Kubernetes OSS and Kubernetes Engine. She started and co-leads the Product Management SIG to maintain an open backlog for the Kubernetes project on Github. Aparna is currently a secondary member of the CNCF Governing Board. She has worked in enterprise software for 15+ years, and was formerly Director of Product at NetApp. Aparna holds a PhD in Electrical Engineering from Stanford and has co-authored several technical papers during her research.
About Ihor Dvoretskyi
Ihor Dvoretskyi is a Developer Advocate at Cloud Native Computing Foundation, focused on Kubernetes-related efforts in the open source community. He is a Product Manager for Kubernetes, co-leading the Product Management Special Interest Group, focused on enhancing Kubernetes as an open source product. Besides that, he is participating in Kubernetes release process as a features lead for multiple Kubernetes releases. A Linux user since 2005 with a deep engineering background, Ihor has been responsible for projects tightly bound to the cloud computing space, containerized workloads and Linux systems. His recent speaking appearances include talks at Kubernetes Leadership and Contributors Summits, OpenStack Summits NA and APAC, OpenSource Summit EU, ContainerDays EU, Google DevFests and other industry-leading events.
Join us for KubeCon + CloudNativeCon in Barcelona May 20 - 23, Shanghai June 24 - 26, and San Diego November 18 - 21! Learn more at https://kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy and all of the other CNCF-hosted projects.
A
Right,
we're
ready
to
start
so
welcome.
Everyone
today
is
the
cabinet,
a
sick
p.m.
intermittent,
and
my
name
is
yours:
came
a
developer
advocate
at
CNC
FM,
one
of
the
colleagues
of
sick
p.m.
and
one
of
the
cofounders
of
sick
p.m.
and,
together
with
me,
is
Aparna
Sina
from
Google.
So
we'll
speak
a
bit
about.
First
of
all,
we'll
take
a
look
at
about
to
the
features
in
the
press
release
we'll
take
a
look
and
what
what?
A
A
B
How
many
of
you
don't
know
sig
p.m.
and
how
many?
How
many
of
you
don't
know
what
it
is,
is
anybody
who
doesn't
know
what
sig
p.m.
is?
Okay,
one
person,
two
people
all
right,
three
people
all
right
so
so
well
and
we've
actually
recently
changed
the
charter
of
sig
p.m.
so.
Let
us
just
give
an
introduction.
We
are
looking
for
folks
that
want
to
join
the
same
right.
Yeah.
A
So
we're
going
to
make
it
a
brief
introduction
of
history
about
6:00
p.m.
so
initially
it
was
the
working
group
that
has
been
established
around
two
years
ago
by
a
few
people
was
in
the
community
who
had
the
main
goal
of
codifying
and
making
making
communities
more
consistent
as
not
only
dopers
project,
but
also
ism
as
a
solid
product.
A
We
is
moving
to
with
marine
communities
as
a
solid
code
base
and
a
solid
product
for
the
end
users
and
for
for
other
communities
vendor
and
market
players,
and
we've
been
working
as
a
working
group
around
half
a
year
after
that
we
switch
to
the
sig
format
being
more
formalized,
being
more
codified
and
having
our
own
place
in
the
Kuban
ideas
ecosystem.
So,
on
this
slide,
you
can
see
how
their
communities
communities
organized.
A
So
here
we
see
a
few
vertical
groups
that
are
focused
on
some
components
like
technical
components
of
cuban
areas,
for
example,
secret
is
focused
on
application.
Development
on
Dryden
applications
of
top
of
kubernetes
is
called
see,
gaps.
Some
errors,
for
example,
networking
and
storage,
occurred
by
signal
work
in
1606,
but
we
are
the
project
wide
seek.
We,
together
with
the
sig
release
the
contributor
experience,
sick,
dog,
sick
testing,
sick
architecture.
B
We
should
probably
explain
what
the
PM
stands
for.
The
PM
is
kind
of
an
all-encompassing
term.
It
means
product
management,
it
also
means
project
management
and
program
management.
So
we
do
a
combination
of
activities
that
ranged
from
helping
to
define
the
roadmap
to
working
on
the
release,
to
helping
to
manage
the
program
and
that's
why
we're
a
horizontal
sig
that
works
across
all
of
the
technical
areas?
The
verticals
are
the
technical
areas
right.
A
So
initially
we
performed
as
the
product
management
working
group
and
product
management
seek,
but
now,
with
the
new
chart,
I
would
like
to
extend
the
era
where,
where
we're
working
on
and
in
which
errors
we
can,
we
can
offer
it
our
support
for
the
community
and
that's
exactly
it.
As
a
partner
mentioned,
it's
not
only
product
error,
but
also
program
management
and
project
manager.
B
A
B
So
we
actually
have
have
both
types
or
any
type
of
function
participate
in
sig
p.m.
so
we
have
had
developers
engineers
actually
some
of
the
lead,
maintainer
x'
of
the
project
that
participate
in
sig
p.m.
and
then
we've
also
had
program
managers
and
project
managers.
This
is
a
sig
which
is
inviting
and
has
work
interesting
contribution
for
anyone
from
any
background.
So,
if
you're
looking
to
make
a
contribution
to
the
kubernetes
project,
this
is
a
good
way
to
get
involved
right.
A
So,
with
the
redefine
portfolio,
we
we
have
actually
highlighted
what
what
we
are
going
to
work
on
and
the
new
chart
is
is
available
in
github.
It's
currently
in
the
stage
of
proposal
we've
been
discussing
it
during
the
last
few
months
within
the
community
and
you're
in
almost
every
6
p.m.
meeting,
and
now
it's
available
for
you
to
review.
You
can
simply
open
this
link
and
to
take
a
look.
What
what
we're
headed
to
these
days.
So
speaking
about
the
major
points
we
are
currently
formular
name
to
SiC
p.m.
A
B
So,
during
the
course
of
the
release,
you
know
we
keep
up
with
the
SIG's.
Sometimes
we
even
see
demos
to
see
if
that
future
is
is
actually
landing
or
not.
That's
one
piece.
The
second
piece
which
is
actually
more
marketing
related,
is
that
we
help
to
write
the
blog
and
all
of
the
press
updates
write
about
a
particular
release
and
and
that's
kind
of
a
complicated
process,
because
usually
there
are
50,
60
features,
features
in
the
release
and
some
of
them
by
the
way,
even
the
first
piece
of
creating
the
backlog.
B
It
is
such
a
broad
project
that
it's
like
herding,
cats,
to
get
everything
into
the
backlog.
A
lot
of
things
are
flying
under
the
radar
and
then
the
the
marketing
piece
of
writing-
the
blog
and
communicating
what's
in
the
release
to
the
press
is
actually
very
difficult
to
do,
because
you
have
to
take
all
of
this
technical
content
and
figure
out.
B
What
of
this
technical
content
is
actually
relevant
and
then
explain
it
in
a
way
that
makes
sense
from
a
use
case
perspective
for
a
user
and
also
is
meaningful
to
communicate
to
the
press
and
others
so
yeah
if
you're
technical
marketing
person
or
if
any
of
that
surface
area
excites
you.
This
might
be
something
that
you
would
enjoy
doing
right.
A
So
if
you
a
few
notes
about
6:00
p.m.
itself,
so
the
readme
is
available
on
github
here's
the
shortened
link
to
eat
our
human,
a
readme.
You
may
find
all
the
all
the
detailed
information
about
the
sake.
How
often
do
we
meet?
Are
we
actually
meeting
every
Tuesday
at
4:00
p.m.
4
p.m.
UTC
time
or
as
far
as
every.
A
Every
every
other
Tuesday-
and
we
have
we
have
a
futures,
this
myself,
Aparna
Kaleb
miles
from
Google,
and
we
have
J's
Dumars,
who
is
here
in
the
audience
who
is
driving
the
program
management
sub
project.
So
we'll
also
take
a
look
at
at
1
to
10
highlights
the
major
features
that
have
been
there.
So
Cuban
itis
1
to
10
was
the
first
release
in
the
current
year
and
we
had
around
25
new
features
compared
to
the
previous
releases
that
we
had
in
the
previous
year.
A
Add
25
new
features
Israel
to
a
small
number,
and
that's
that's
caused
mostly
because
in
one
to
ten
release,
the
main
focus
of
the
community
was
not
only
on
not
only
on
developing
some
extremely
new
functionality,
but
on
enhancing
the
existing
features.
So
many
features
that
been
delivered
in
kubernetes
one
to
ten.
Where
are
our
enhancements
of
the
previous
features
that
have
been
delivered
before
just.
B
One
thing
to
add,
with
regard
to
features,
managing
features,
making
sure
that
they're
in
the
backlog
and
instituting
sort
of
new
processes
around
say
a
stability,
stability
release
or
an
innovation
release.
These
are
things
that
are
very
project
management,
oriented
and
so
we're
always
looking
for
folks
that
are
good
at
process
and
can
improve
the
process.
A
Keep
in
mind
right,
right
and
again
in
communities,
one
to
ten.
We
had
a
few
major
errors
of
enhancements
and
a
few
major
errors
where
the
entire
communities
communities
moving
towards.
If
you,
if
you
saw
the
keynote
that
a
partner
made
today
about
communities
movement
as
a
project,
you
could
notice
a
few
major
items
there.
I
will
just
highlight
them
that
we
are
moving
towards
the
enhancing
communities,
storage
and
workloads
ability,
extensibility
of
communities
with
a
login
you
as
a
developer,
to
develop
some
new
functionality.
Atop
of
kubernetes,
for
example.
B
A
Coming
back
to
the
features
we
it's
not
highlighted
on
this
slide,
but
just
just
to
mention,
if
you,
if
you're,
not
aware
of
Cuba
neatest
features
process,
we
have
three
major
categories
features
there
are
alpha
features
beta
features
and
GA
or
stable
features.
So
alpha
features
are
some
branding
features
with
experimental
functionality
that
not
that
I
really
not
ready
for
the
real
production
usage
but
they're
showing
their
effort
within
the
community
where
we
are
moving
to
and.
B
A
A
Exactly
beta
features
are
more
mature
X
alpha
features
that
that
currently
have
a
biggest
ability
compared
to
the
alpha
stage,
but
sometimes
it's
it's
also
not
not
necessary
for
year
for
our,
if
you're
running
some
precise
or
real
production
workload
for
you
to
use
the
beta
features,
but
at
the
same
time,
usually
beta
features
I
enabled
by
default.
Yes,.
A
B
A
Can
I
can
make
not
a
real
deep
dive,
because
we
have
it
at
different
sessions
here
at
cube
con
that
I'm
making
a
real
deep
dive
into
this
errors,
but
I'll
highlight
the
magic
bullet
points
about
these
features.
So,
speaking
about
storage,
one
of
the
major
features
that
has
been
delivered
in
cabinet
is
1
to
10
is
CSI
enablement
as
a
beta
feature
CSI,
if
you're
not
aware
of
CSI
is
a
container
storage
interface
that
allows
you
to
want.
A
A
So
si
si
si
sighs
the
container
search
interface
that
allows
you
if
you
a
storage
vendor,
for
example,
if
you
are
developing
some
storage
functionality
for
communities,
it
will
be
much
easier
for
you
to
to
develop
the
storage
product.
Who
is
with
the
well-defined
interface
in
mind?
So
si
si
is
not
only
the
technical
implementation,
but
it's
also
the
set
of
specification
that
defines
for
you.
How
should
the
storage
implementation
top
of
cabinets
should
look
like?
A
B
The
container
storage
interface
which
which
Zig
storage
has
been
working
on
it's
possible
to
do
out
of
tree
plug-ins
for
kubernetes
and
the
interesting
thing
about
the
si.
Si
is
actually
it's
not
just
kubernetes,
it
actually
spans
other
container
orchestrators,
so
I
believe
also
may
sews
and
docker.
Okay.
A
A
A
So
previously,
if
you've
been
using
communities-
and
you
would
like
to
use
some
some
Williams
functionality,
it
meant
that
you
had
to
use
some
network
attached
storage
if
you'd
like
to
use
it
as
as
a
local
storage.
But
today
you
can
use
the
real
physically
attached
to
local
local
volumes
and
local
storage
to
your
cabinets
cluster.
It
it's
not
really
intended
to
to
store
some
precise
data
there.
But
if,
for
example,
you
have
the
use
case
of
SS
education
or
similar,
you
can
use
that
ya.
B
B
You
know
the
stat
that
I
like
to
throw
out
is
more
than
40%
of
clusters
are
running
stateful
applications,
the
most
popular
ones
being
read,
is
an
elastic
search
and
so
and
then
you
know,
there's
there's
more
coming
like
Cassandra
and
others,
and
so
a
lot
of
our
users
started
asking
for
local
storage
for
higher
performance
for
these
stateful
applications.
It's
that
has
been
a
huge
ask.
I
think
it's
now
in
beta
in
the
1.10
release.
A
A
So
another
area
where
we
are
moving
ways
with
the
major
features
in
this
releases
extensibility
and
Syrian
API
Gration-
are
one
of
them
most
most
important
features
here.
So
sergey's
is
an
ability
for
you
to
to
develop
some
search
party,
certified
applications
and
some
party
resources
using
their
existing
a
pair
of
cabinets
and
probably
have
heard
a
recent
announcement
from
from
one
of
the
biggest
contributors
to
cuban
areas
about
enabling
ment
of
the
operators
framework-
and
you
also
mentioned
give.
A
B
Operator
is
a
custom
controller,
so
you
write
a
custom
controller
for
your
particular
application.
We
intend
to
use
this,
of
course,
for
stateful
applications
to
make
it
easier
to
run
state
applications.
So
you
have
a
custom
controller,
and
then
you
have
the
custom
resource
definition,
which
is
a
definition
that
is
given
to
the
API
server
and
the
combination
of
those
two
allow
you
to
extend
the
kubernetes
api
in
whatever
way
you
want,
and
obviously
it
has
many
many
many
uses,
but
that
application
support
is
one
of
them
right.
Great.
A
A
A
It
was
still
available
during
the
few
releases,
but
now
with
kubernetes
113
its
following
finally
deprecated,
and
if
you
had
experience
of
runnin
some
sort,
but
your
sources,
you
had
enough
time
to
migrate
to
CRD.
That's
actually
one
of
the
greatest
symbols
of
how
do
we
follow
the
deprecation
policy
of
the
features
in
communities
and
the
final
era
is
about
the
Big,
Data
spark
and
machine
learning.
So
again,
Aparna
I.
A
A
B
Device
plug-in
work
that
also
was
part
of
the
1.10
release
just
having
given
that
view
of
the
futures
in
1.10
I
think
it's
worth
pointing
out
which
cigs
you
know
we're
responsible
for
a
lot
of
the
future
work
in
this
release.
So
obviously
sig
storage,
very
active
sig
and,
as
we
look
out
to
the
roadmap
for
the
rest
of
this
year,
continues
to
be
a
very
active
sig,
with
more
support
for
local
storage,
with
more
support
for
other
plugins
with
finalizing
and
bringing
to
GA
the
CSI
standard.
B
You
know
across
the
community,
so
sig
storage
is
one,
has
a
lot
of
activity
and
then
sig
big
data.
Actually
it's
a
little
bit
grayed
out,
but
sig
big
data
is
the
one
that's
been
working
on
spark
and
that
activity
is
is,
of
course
now
maturing
and
then
cig
apps
is
the
one
that's
working
on
all
of
the
operators,
so
you'll
expect
to
have
more
operators
for
things
like
Cassandra
and
Redis,
and
other
things
signal
is
always
active.
B
I
think
now
particularly
active
in
the
area
of
sandbox
containers
and
the
container
runtime
initiative,
or
this
sorry,
the
container
runtime
interface
sig.
Cluster
lifecycle
is
working
on
deploying
kubernetes
making
that
easier
and
is
working
on
a
downward
api
to
make
it
standard
to
to
basically
cluster
resources
underneath
kubernetes,
any
others
that
we
want
to
point
out.
There's
a
number
of
working
groups
that
have
come
up.
Those
are
the
ones
in
green.
Those
are
actually
most
of
them.
B
A
B
Well,
actually,
maybe
we
realize
some
of
the
feedback.
First,
we
we
spoke
to.
If
you
went
to
the
initially
the
opening
keynote
yesterday,
I
think
it
was
Dan
Cohen.
He
talked
about
the
end
user
community
of
of
the
CNC
F,
so
the
CNC
F
has
actually
got
thirty
five
member
companies
that
are
users
of
the
open-source
projects,
primarily
actually
kubernetes,
and
that's
a
tremendous
resource
that
we've
been
leveraging
to
get
feedback
about
the
roadmap
of
kubernetes
and
how
kubernetes
is
helping
and
where
it's
not
yet
helping.
So
we
spoke
to
them
recently,
even.
A
B
I
and
we
got,
we
gathered
some
feedback
about
what
are
the
features
or
what
are
the
capabilities,
the
value
proposition
of
kubernetes,
that
that
is
really
resonating
in
this
and
really
panning
out
in
your
in
your
company,
and
we
heard
three
things,
and
these
might
surprise
you.
Actually
we
heard
why
don't.
B
Let
me
ask
this:
how
many
people
think
that
it
was
deployment?
Velocity
raise
your
hand
if
you
think
deployment
velocity
was
one
of
them:
okay,
two
people?
Okay,
that
doesn't
seem
to
be
the
majority.
How
many
people
think
that
it
was
the
fact
that
kubernetes
makes
their
application
portable
that
it's
multi
cloud?
That's
the
big
benefit
that
they're
getting
out
of
it:
two
people:
okay,
how
many
people
think
that
it
was
predictability?
B
B
How
many
people
think
it
was
security?
It's
just
much
more
secure
I
got
isolation
from
running
containers,
okay
and
then
access
to
innovation,
okay,
interesting
cool!
So
what
we
heard
and
it's
it's
a
it's
a
select
sample
I
mean
it's
35
companies,
but
it's
you
know:
it's
big
companies,
its
companies
like
big
and
small,
actually
Goldman
Sachs
and
eBay,
and
New,
York,
Times
and
so
forth.
But
if.
B
B
Efficiency
I
was
a
little
bit
surprised,
but
but
yeah
one
of
the
end
users
said
that
they
went
from
VMs
to
using
kubernetes.
They
were
startup
and
they
went
from
four
to
five
percent
utilization
to
90
percent
utilized
I'm,
not
sure
if
I
really
believe
them,
but
ninety
percent
utilization
of
the
underlying
resources-
that's
huge,
but
everyone
seem
to
agree
that
by
going
to
kubernetes
I
guess
they
are
really
Struhl
II
bin
packing
and
densely
packing
their
containers
into
their
infrastructure.
B
No,
it's!
It
was
utilization
resource
utilization,
that's
what
they
pointed
out.
The
number
two.
The
number
two
benefits
is
actually
kind
of
not
on
this
list.
It
was
reliability.
Yeah
I
didn't
even
think
to
put
it
on
the
list.
It
was
reliability
that
the
underlying
infrastructure
can
die
and
or
parts
of
it
can
die
and
their
application
doesn't
suffer
and
that
that
had
that
had
a
BAE
had
noticed
a
meaningful.
B
You
know
lack
of
headache
from
from
moving
to
kubernetes
from
whatever
they
were
running
on
and
then
the
third
one
was
in
fact
developer
productivity.
So
I
think
everyone
just
sort
of
takes
that
for
granted
that
yes,
now
I
can
I
can
release
hundreds
of
times
a
day
if
I
want
to
I
can
just
do
it
very
quickly,
so
that
was
the
number
number
three
benefit
and
then
some
people,
some
people,
also
chimed
in
with
the
portability
multi
cloud,
but
a
little
bit
less
than
the
other
three.
So
it's.
B
B
B
B
B
See
see
that's
great
feedback
I
mean
particularly
about
it
being
more
complicated.
I
still,
I
definitely
hear
that,
particularly
from
folks
who
are
deploying
its
themselves
either
on
Prem
or
in
the
cloud,
and
a
lot
of
people
are
using
the
managed
offerings
like
gke,
and
then
it's
not
so
complicated,
but
I
think
we
still
have
a
lot
of
work
to
do
as
a
community
to
make
kubernetes
easier
to
consume
I
kind
of
touched
on
a
few
pieces
of
that
in
my
keynote
this
morning,
but
I
think
the
deployment
experience
is
not
something
I
heard.
A
B
B
B
B
Yeah,
you
know
actually
that
one.
It
surprises
me
less,
because
there's
a
lot
of
small
clusters
there's
a
lot
of
smaller
clusters.
It's
starting
though
there's
it's
I
mean
it.
You
know
kubernetes
is
in
is
in
it's
it's
it's
in
movement.
It's
it's
in
progress.
The
the
transformation
is
in
progress,
and
so
and
it's
very
fast
to
change
so
we're
starting
to
see
a
lot
more.
You
know
a
I.com
I
think
they
wrote
a
blog
about
their
5,000
node
cluster
and
there's
a
few
others
that
have
5,000
node
clusters.
B
I
haven't
really
seen
too
many
that
are
larger
than
that,
although
there
are
people
that
are
looking
to
have
larger
clusters
and
and
potentially
span,
certainly
zones,
but
potentially
also
regions
and
that's
kind
of
a
big
area
of
discussion
within
the
kubernetes
project
is
how
should
we?
How
should
we
provide
for
more
scale?
B
You
know,
should
we
basically
have
one
very,
very,
very
large
cluster,
or
should
we
have
many
clusters
and
so
there's
an
entire
set
called
cig
multi
cluster,
which
has
you
know,
been
developing
technology
for
managing
multiple
clusters
and
the
and
I
think
the
for
the
vast
majority
of
users
right
now.
The
paradigm
is
that
you
split
your
your
your
applications
across
multiple
clusters
and
there's
many
ways
that
that
they
split
that
they
split
them.
I
mean
users
split
it.
B
They
have
one
cluster
for
tests
and
then
another
one
for
dev
and
then
another
one,
for
you
know,
production
and
another
another
kind
of
split
is
they
have
one
cluster
per
region?
They
have
some
clusters
on
Prem
some
clusters
in
the
cloud
they
have
one
cluster
or
a
set
of
clusters
for
stateful
applications
that
maybe
are
on
a
different
version
of
kubernetes
and
then
another
set
of
clusters
for
stateless
applications
that
they
maybe
upgrade
more
frequently.
B
Those
are
some
of
the
divisions
that
I
have
seen
in
terms
of
multi
cluster,
but
then
there
are
some
companies
that
just
have
kind
of
like
one
app
or
one
big
app.
You
know
Niantic
Pokemon
goal
was
like
that,
and
so
they
don't
really
need
multiple
clusters.
They
just
need
like
a
massive
cluster,
so
so
you
get
both
of
those
and
then
for
the
multi
cluster.
One
sorry
for
the
one
massive
cluster
one,
the
the
the
new
trend
that
I'm
seeing
is
that
there
are
SAS
provider.
B
You
know
so
WordPress
engine,
for
example,
is
running
a
service
for
WordPress,
where
they
host
many
many
different
peoples
websites,
and
so
they
want
like
a
huge
cluster
with
multiple
tenants,
and
this
is
where
the
security
and
isolation
and
multi-tenancy
comes
in,
so
it
I
think
technically
a
very,
very
exciting
time.
So.
A
A
B
The
main
thing
that
actually
they
breeze
through
this
like
there
wasn't
a
lot
of
it's
like
yeah
customizability-
is
their
resource.
Efficiency
is
their
security,
a
lot
of
requests
for
greater
security
and
developer
friendly
experience
and
ease
of
management.
Those
were
the
three
that
came,
but
I
think
probably
resonates
with
all
of
you
as
well,
and.
A
Speaking
about
security,
we
have
to
mention
that
there
are
still
a
big
demand
from
the
customers
to
enhance
security
in
kubernetes.
At
the
same
time,
enhancing
security
in
communities
was
a
big
trend
for
entire
communities
came
in
during
the
last
year
or
two
years,
so
enterprise-grade
kubernetes
cannot
be
achieved
without
some
major
security
features
and,
for
example,
air
base
seed.
It
was
one
of
the
major
features
in
cabinet
is
one
that
vi
it
has
been
presented
around
a
year
ago
was.
It
was
the
beacon
major
of
feature
bullet-point
for
so
many
enterprise
customers.
So.
B
B
Scalability
is
a
really
great
question
like
what
level
should
we
scale
to
and
if
we
scale
the
10,000
nodes,
what
should
be
the
features
within
the
cluster
that
you
know,
customers
will
be
looking
for
and
what
we
want
to
do
as
a
cig
is
to
take
that
input
and
then
take
it
to
shape
the
roadmap
of
the
project
and
obviously
go
to
the
various
SIG's
that
are
working
on
these
topics
and
say
you
know:
here's
the
10
customers
that
are
looking
for.
You
know
this
particular
feature.
B
B
Yeah
yeah,
it
really
varies,
I
think
that
a
lot
of
it
is
also
cycle
based,
it's
cyclical.
You
know,
depending
on
where
we
are
in
the
release
cycle,
there's
a
lot
of
time
up
front
to
create
that
backlog,
usually
and
uhor
does
I.
Think
the
vast
majority
of
the
work
there.
You
know
just
making
sure
that
every
sig
puts
their
features
in
making
sure
that
the
futures
have
enough
detail
that
there's
design
Docs
associated
with
things
to
eat.
Why
don't
you
tell
us
what
you
do
so.
A
That
usually
happens,
so
again,
it
can
be
coming
back
to
the
topic
of
how
cabinet
is
released,
where
listen
personally,
we're
making
four
releases
per
year
and
recycle
is
approximately
three
months
and
the
first
month
is
the
period
that
is
called
pre
feature
free
stage.
So,
at
the
end
of
the
first
month
we
have
the
feature
freeze
and
until
that
period
of
time,
all
that
all
the
features
have
to
be
finalized
and
codify
it
in
the
kubernetes.
/
features
repo.
B
A
A
B
Would
say
it's
probably
about
four
or
five
hours
per
week.
You
know,
but
not
every
single
week,
but
you
know
you
know
kind
of
more
bunched
together
around
the
start
of
the
release
and
then
towards
the
end
of
the
release.
So
what
we
do
towards
the
end
of
the
room,
not
at
the
very
very
end,
but
just
before
the
end
is
you
know,
we
will
sometimes
host
some
demos
to
understand
the
features
we'll
start
working
on
the
blog
post,
we'll
make
sure
that
there's
documentation,
release.