►
From YouTube: Cloud Foundry Community Advisory Board [May 2017]
Description
Video from the monthly Cloud Foundry Community Advisory Board (CAB) meeting. Learn more at https://www.cloudfoundry.org/.
Agenda:
Calendar
Now every 3rd Wed of the month (see above)
Tooling for the call
See Zoom info above in and the #CAB Slack channel
CFF update
CF Summit North America in Santa Clara, CA - June 2017
Certification beta is now open to take exam (lower cost before summit)
You can also take the certification at Summit
Hackathon
CF Summit Europe in Basel, Switzerland
Projects updates
Runtime PMC
Porting Garden runC onto Windows
OCI buildpack proposal - initial stories started
BOSH
DNS v2 work is continuing
Support for BBR
Extensions
Managing new proposal input: BBR and Kubo
Community projects
Project Unik by Idit Levine (formerly of EMC)
BOSH Backup Restore (BBR) by Therese Stowell (Pivotal)
Kubo (Kubernetes + BOSH) by Eric Johnson (Google)
A
C
B
C
B
B
B
Now
it's
great
tippy-toe
is
fantastic,
as
you
know,
we
have
hours
to
be
here.
So
that's
cool
okay,
so
we
have
a
packed
agenda.
It's
really
fantastic
to
have
we
three
great
talks,
I
think
by
people
that
are
trying
to
improve
the
community,
so
we're
gonna
try
to
get
to
them
as
fast
as
possible,
but
as
we
usually
do,
we
get
a
quick
update
from
the
foundation
and
then
maybe
some
highlights
from
projects
and
then
we'll
just
go
straight
to
the
presentations,
a
timing
that
is
part
of
the
wash
to
me
so
I
do
everything.
B
C
D
Clara's
coming
folks,
June
13th,
if
you
haven't
registered,
you
should
contributors
the
the
codes
but
on
the
mailing
list.
Anybody,
that's
that's
sponsoring.
You
know
talk
to
your
your
events,
team
to
get
those
codes
and
otherwise
just
get
yourself
registered.
We've
we've
got
a
little
bit
of
news
that
we
we
had
opened
up.
The
claw
foundry
certified
developer
program
for
a
pool
of
beta
candidates
that
that's
all
done
we've.
We
know
what
the
passing
score
is
so
for
those
that
participated
in
the
beta.
D
Thank
you
we'll
be
coming
back
to
you
with
with
either
a
congratulations
you're
now
in
the
select
few
cloud,
foundry
certified
developers
or
here's
another
code,
give
it
another
shot
at
so
we're
very
thankful
for
their
help
through
that
process
and
we've.
Actually,
we
we
had
received
about
three
hundred
applicants
for
the
beta,
so
we've
anybody
that
had
applied
for
the
beta
that
didn't
didn't
make
it
into
that
group.
You
all
have
a
code,
that's
that
should
get
you
access
to
the
exam
at.
D
Actually,
you
know
below
our
cost,
so
this
will
be
really
limited
and
you
have
to
get
registered
before
summit.
Don't
forget
we're
also
offering
cert
at
summit,
there's
really
great
training
stuff
going
on
the
first
day,
so
if
you're
new
to
community
or
maybe
something
like
concourse
parks,
your
interests
take
a
look
at
that
and
yeah.
Then
there's
a
hackathon
which
I
think
it's
going
to
be
amazing.
We
already
have
a
good
number
of
people
registered
for
the
hackathons,
so
maybe
dr.
max.
You
want
to
just
do
quick,
quick
view.
B
On
that
yeah
take
you
a
ticket
chip,
so
one
of
the
things
so
Tim
iris
is
helping
run,
that
out
kind
of
emceeing
part
of
it
and
one
of
the
things
we're
trying
to
do
for
the
hackathon
is
to
allow
everybody
to
be
part
of
it.
So
that
means
well,
you
can't
actually.
So
this
is
the
cool
part.
The
first
day,
we'll
have
like
a
couple
hours
to
introduce
and
kind
of
give
people
ideas
of
what
we
are
looking
for
enjoy
and
then
the
hack
starts.
B
So
that
means,
if
you
go
to
presentation
in
your
board,
you
have
your
laptop,
keep
hacking.
You
know
what
I
mean.
So
basically
the
hackathon
is
during
the
two
days
and
then
we'll
be
selecting
so
there'll
be
a
date
where
you
come,
and
you
say:
okay,
this
is
my
hat
and
then
so
you
can
keep
going
to
the
conference.
B
Are
you
just
submitted
and
then
once
we
select,
you
know
I
think
I'm
I,
don't
know
exactly
I,
don't
want
to
say
exactly
the
is
it
top
five
or
top
three,
but
anyways
we'll
select
a
few
that
will
get
a
chance
to
present
during
Abi's.
You
know
keynote,
so
you
get
essentially
visible
time
to
and
then
out
of
those
presentation
will
select.
The
winners
and
there'll
be
significant,
prompts
so
how
significant
robots
go
ahead.
You
know
better
than
me:
robots,
robots,
okay,
cool.
C
A
C
B
So
basically
that
means
anybody
can
participate,
we're
not
trying
to
tell
you
go
to
a
presentation
and
not
watch,
but
if
you
go
there
and
you
board
and
knock
right,
so
that's
that's
the
opportunity
right.
So
so
that's
that's
the
hackathon
one
thing
I
want
to
remind
everybody
is
that
the
call
is
being
recorded
so
watch
your
language
now
that
I'm
saying
this
YouTube,
but
just.
B
B
C
B
Out
of
it
will
be
interesting
for
the
family
for
the
community,
so
that's
it
any
questions
about
cff
or
comments,
no
okay,
so
they
one
more
thing
about
the
summit
is
that
there
is
a
summit
for
you,
one,
that's
coming
and
it's
gonna
be
in
Basel.
Switzerland,
really,
nice
city
have
been
there
actually
this
year
and
I
should
be
exciting.
Our
Swiss
column
is
essentially
helping
running
with
the
foundation,
so
if
you've
never
been
to
Switzerland,
this
is
your
chance
to
think
about
that
cool.
Alright.
B
So,
and
so,
since
there
are
no
question,
I
want
to
make
sure
we
give
enough
time
to
do
presenters,
they're
all
here,
so
we'll
go
with
each
product
in
terms
of
like
the
core
PMC's.
If
you
have
something
to
highlight,
let
me
know
I,
don't
know
if
Zach
or
anybody
from
the
runtime
PNC
wants
to
highlight
something
for
you.
You
wanna
highlight
anything.
A
Well,
so
there's
a
couple
of
kind
of
cool
things
going
in
garden
and
I'll
highlight
each
other.
The
first
thing
is:
we
just
started
work
to
port
garden
onto
windows,
so
this
is
sort
of
the
next
generation
of
garden
windows,
work
and
before
there
is
sort
of
his
custom
implementation
with
the
garden
API
on
Windows.
A
We're
certainly
not
definitely
going
on
the
whole
path
back,
there's
a
whole
lot
of
ocean
to
boil
in
that
proposal
yeah,
but
the
garden
team
has
started
doing
the
first
couple
of
stories
along
that
path.
So
we've
started
to
exploit
filesystem
layers
to
improve
performance
of
bill
packs
so,
instead
of
them
just
being
using
the
docker
flow,
we're
now
trying
to
see
where
we
can
use
some
of
that
stuff.
So,
instead
of
untiring
droplets
I'm
into
the
running
container,
we're
going
to
pre
great
file
system
layers
and
then
create
container
on
top
of
them.
A
B
B
B
Oh
okay,
so
maybe
let
me
see
if
de
metrio
here
you
must
be
unmuted
few
here,
I'll
give
the
highlight
for
him.
If
he's
not
so
I
actually
joined
the
Bosch
team
for
the
past
two
weeks
and
last
time
he's
here
too
I'll
mention
two
things
that
were
he's
been
focusing
on.
Dmitri
has
been
getting
the
team
to
get
the
Bosch
DNS
version.
So
this
is,
if
you
don't
know
what
that
is,
you
may
want
to
look
into.
B
You
know
I
think
Dimitri
as
a
Bosch
notes
on
this,
but
essentially
it's
gonna
allow
us
to
remove
power
DNS,
and
it
also
allows
us
to
remove
console
as
a
consequence
in
Diego.
So
this
is
a
broad
set
of
work
that
should
make
it
a
little
bit
easier.
So
this
system
has
been
working
hard
on
this
and
then
also
just
mention
here
as
well
mentioned,
the
VBR,
so
Bosch
is
affording
the
DVR.
So
that's
another
piece
of
work,
that's
going
on,
so
that's
that
and
then
for
extensions.
B
I'll,
just
say
that
you
know
we
have
three
presentations
that
essentially
sort
of
helps
show
that
some
of
the
things
I
were
in
a
pipeline.
So
if
you
have
a
question
about
that,
then
you
know
we're
trying
to
put
some
structure
around
extensions
but
also
make
sure
that
we
invite
people.
So
that's
a
consequence
of
the
this
meeting
Burnsville
giving
people
a
chance
to
so
let
me
see
if
there's
any
questions.
If
there
is
none
you
you
remember
your
new
automatically,
so
you
might
need
to
unmute.
Okay,
I,
don't
see
anybody
chiming
so.
F
F
So
so,
when
you're
running
today,
Cloud
Foundry,
the
Blessed
flavor,
usually
you're
running
it
on
home
winter
and
we're
doing
a
lot
a
lot
of
effort
to
make
the
container
verse
very
am,
and
so,
where
is
a
secure
but
actually
you're
still
having
the
problem,
which
is
the
operating
system,
is
not
really
secure.
So,
if
you're
running
going
we're
going
to
you
basically
running
that
on
that
kind
of
stuff
right,
every
time
that
you're
running
it.
F
So
if
we
look
in
a
day,
they'll
burn
to
destroy
itself
or
any
little
kind
of.
For
that
sake,
you
will
see
that
it's
very,
very
complex
and
because
it's
very
very
complex.
It's
because
it's
trying
to
do
quite
a
lot
of
stuff
and
it's
trying
to
manage
the
drivers,
any
the
memory
management
and
they
protect
your
processes
and
so
on,
and
that
making
a
lot
a
lot
of
sense
when
they
actually
started
the
UNIX
operating
system
and
that's
the
problem
that
they're
trying
to
solve
is
time
sharing,
because
back
on
the
day
they
build.
F
F
There
is
a
lot
of
unnecessary
component
on
the
pointer,
for
instance,
this
show
today
they're
still
the
floppy
driver
running
there.
If
you're
running
on
the
cloud
you
still,
the
USB
driver,
which
mean
a
lot
of
company
that
really
is
not
needed,
are
still
included
in
your
operating
system
when
you're
running
again
you're
just
trying
to
run
your
application
in
Cloud
Foundry,
but
you're,
actually
getting
all
these
things
behind
the
scene
that
you
are
not
aware,
or
maybe
you
are
in
terms
of
memory
management
again,
the
purpose
of
the
operating
system
is
to
protect.
F
F
So
that's
the
main
purpose,
but
again
when
we're
running
today,
we're
basically
running
one
process
usually,
and
therefore
there
is
no
really
need
to
protect
between
the
processes
and
between
the
user
and
all
this
over
at
creating
us
a
lot
of
complexity
and
a
lot
of
vulnerability.
A
it
context
we
each
and
and
and
you
know,
multi
address
space
and
so
on.
There
is
so
much
unnecessary,
they
aim
management
layer
there
because
we're
only
running
one
process
right
now,
I
mean,
if
I
just
described
that
they're
stuck
it's
very,
very,
very,
very
big.
F
That
means
that
also
there
is
a
lot
of
surface
of
attacked
is
huge
and
if
the
surface,
if
attack
to
view
it,
you
can
actually
penetrate
from
a
lot
of
things.
The
last
big
attack
that
was
in
the
US
so
I
think
it's
actually
was
all
over
the
world
was
that
an
open
source
project
called
Mir
I.
Actually,
it's
an
open
source.
So
the
architects
like
created
a
community
to
take
the
internet
down
and
what
happened
there
is
that
they
basically
went
to
all
them
them.
F
The
Internet
of
Things
devices
and
those
Internet
of
Things
device
is
actually
running.
They
operate
the
system
and
therefore
there
was
telling
it
there.
So
no
one,
no
one
care
about
the
talent,
because
that's
not
what
basically,
you
know
they
secured
application
itself,
but
never
dealt
no
operating
system
and
therefore
what
happened
they
they
kind
of
like
took
over
and
BOTS
basically
created
an
army
of
Spartan
to
paint
on
it
down.
F
So
that's
something
that
could
happen
also
in
if
you're
running,
Cloud,
Foundry
and
someone
is
penetrated,
they
were
going
to
do
boon
to
itself
right.
It
doesn't
matter
out,
secures
the
container
itself.
Oh
awesome,
Cloud
Foundry
is
what
is
important
is
that
people
can
penetrate
okay
Aaron.
Basically
again,
it's
a
one-hour
talk,
but
basically
I'm
explaining
about
complicated
the
Linux
operating
system
is
how
many
line
degrees
and
again
I,
run
with
it,
because
I
don't
think
we
have
the
time.
So
sorry
I
described
a
lot
of
inefficiency,
very
quick,
but
I
did
so.
F
The
question
is
how
we
get
there
and
it's
quite
simple
right.
You.
Basically
it's
evolution.
We
started
all
the
way
for
the
mainframe
until
today.
What
we
basically
running
everywhere
is
the
Linux
kernel
again
now
leaders
can
run
everywhere
and
it
can
run
anything
and
because
it's
a
multi
process,
it
would
be
multi-purpose
operating
system.
There
was
a
lot
of
unnecessary
data
there,
so,
for
instance,
I
can
actually
take
the
last
destroyed.
Do
want
to
run
it
on
pentium
one.
It's
still
going
to
work
right,
I'd
only
to
okay.
F
F
So
if
you
want
the
chorus
getting
it
built,
but
you
know
skip
is
is
basically
a
way
to
build
something
like
that
to
build
an
operating
system
that
will
be
purpose
for
what
you
need.
So
by
the
way,
a
lot
of
the
design.
Actually,
that
is,
is
taking
from
unique
and
basically
what
you
do
and
you
you
get
in
a
container
and
creating
the
operating
system.
Is
there
any
question.
F
Okay,
so
that's
the
New,
York,
State
and
eunuch
colonel
has
a
very
similar
project,
it's
kind
of
like
the
team
in
Cambridge
in
charge
of
the
eunuch
and
Allah
the
one
who
actually
did
an
ox
kid,
and
the
purpose
is
the
same
thing.
It's
basically
the
same
today,
every
time
you're
running
it
operate
the
system,
application
you're,
always
running
the
kernel
and
on
top
of
the
kernel
you're
putting
on
the
iOS
libraries
always
right.
F
This
is
even
and
on
top
of
it,
you're
putting
in
a
application,
no
matter
how
many
application
you're
still
running,
all
the
stack
on
the
operating
system.
So
what
what?
What
unicorn
is
doing?
They
said
what,
if
we
will
take
only
the
application,
and
we
basically
will
take
only
the
parts
from
the
OS
libraries
and
the
kernel
that
needed
to
support
this
application.
So,
for
instance,
if
your
application
is
doesn't
need
network
you're
going
not
going
to
have
the
driver
of
the
network
or
any
libraries
that
will
light
it.
It
looks
something
like
that
right.
F
Only
the
black
part
is
going
to
be
kind
of
like
packaged
together
with
your
application.
It's
time
for
wrapping
your
application
and
and
that's
it-
that's
basically
a
bootable
image
and
now
in
order
to
make
it
even
simpler.
We
didn't
we
make
these
design
choice,
the
community,
making
eMac
design
choice
and
the
design
choices
that
we
are
only
going
to
support
one
process
every
one
user,
because
if
you
doing
that
not
a
lot
of
the
complexity
of
context,
switching
of
memory
management
of
protection
ring
are
not
relevant,
so
we
can
remove
it.
F
There
is
no
need
to
do
all
the
time
permission
check,
because
there
is
no
Monda
and
so
on.
So
now
we
can
actually
take
something
like
it
went
to
destroy
that
it's
a
minimum
one
gig
side
and
reduce
it
dramatically
to
something
like
men's.
Oh,
you
know
Oh
even
case
so
so
that's
the
ID
how
to
make
it
I'm
not
going
to,
and
we
don't
have
time.
Basically,
there
is
some
packaging
tool
that
using
the
linkage
and
know
how
to
take
what
it's
need.
F
Basically,
the
compiler
and
basically
wrap
it
there
as
a
images
and
in
what
we
did
reduce
all
of
this
right.
So
basically
there.
It
is
right
now
that
all
this
middle
layer,
we
should
operate
the
system
that
you
donated
and
so
on,
it's
kind
of
like
move
aside,
and
that
means
that
you
have
less
code
fewer
bags
and
easy
to
reason
about.
So
it
looks
something
like
that
right.
Basically,
this
is
a
stack
code
we
had
before
and
right
now,
basically
you're
running
a
nitro
visor.
You
can
run
also
in
bare
metal.
F
It
just
does
not
really
make
sense,
because
it's
very
small,
so
you
don't
want
to
actually
take
all
the
Machine
and
put
this
application
on
it.
But
are
you
using
an
eye
professor,
usually
and
and
people
will
come
and
ask
so
what?
If
we
will
take
the
container
and
put
it
on
bare
metal,
and-
and
this
is
a
good
question,
but
the
problem
with
that
approach
is
that
the
operating
system
did.
The
Linux
was
never
intend
to
serve
as
multi-application
as
servers
and
therefore
they
they're
not
going
to
leverage
something
like
they.
F
C
A
F
True,
so
so
it's
all
trade-off
right.
The
question
is
what
is
easier
and
what
it
may
give
you
more
benefit,
and
it's
a
good
question
to
I
said:
I,
don't
know
this
answer
I'm
telling
to
think
that
unique
and
would
be.
You
know
a
bit
better,
but
you
know
downloader
right
in
terms
of
advantage,
so
no
permission
check,
so
you
utilize
one
percent
of
the
of
the
hardware
for
a
for
the
actually
application.
Again
choices.
We
don't
walk,
we
click
on
their
and
son
isolation.
F
Basically,
micro-services
is
sharing
a
lot
of
stuff
assuring
the
memory
the
filesystem.
You
know
there
is
a
lot
of
advantage
with
that,
but
also
you
know
disadvantage,
which
is
it's
lesson
you
know,
you're
still
sharing
a
lot,
so
it's
a
secure
in
terms
of
site.
So
we
talking
about
the
difference
between
one
gig.
You
know
minimum
of
one
Giga
decides
when
you
look
in
the
regular
operating
system
to
case
right.
Basically,
the
size
of
the
application
is
the
size
of
the
unica,
and
the
short
time
is
the
boot.
F
Time
is
really
quick
and
that's
will
be
very
interesting
to
something
like
several
s,
because,
if
you're
looking
today,
people
reusing
container
for
for
basically
creating
a
hot
container
for
service
and
the
reason
they
do,
that
is
because
it's
taking
time
to
boot
with
unique
I
know
the
boot
time
is
much
quicker.
So
we
can
leverage
that
and,
and
and
and
the
surface
of
attack
is
tiny,
but
it's
also
custom-
and
this
is
this-
is
very
important.
F
If
you
manage
to
penetrate
to
one
of
them,
you
will
not
be
able
to
manage
to
penetrate
to
the
others
oo.
It
is
not
easily,
and-
and
the
last
thing
is
that
I
really
really
put
unicameral
is
the
right
approach
and
the
reason
I
hope
also
the
right
approach,
it's
a
Endon
or
or
because
I'm
sorry,
it's
alright,
because
I'm.
F
What
we're
trying
to
build
right
now
is
an
immutable
infrastructure
right.
This
is
what
we
pitching
all
the
time
we
wanna
build
an
immutable
infrastructure
and
that's
a
really
mutable
infrastructure.
You
cannot
ssh
to
the
machine.
You
cannot
that
put
some
patch,
it's
actually
a
really
beautiful
infrastructure
and
that's
what
I
feel
that
this
is
a
feat
for
microservices
architecture
and
I
mean
promote
performance,
but
again
I'm
not
going
to
go
into
details.
You
can
see
this
online.
The
performance
are
much
better.
F
A
A
On
a
VM,
well
it,
oh,
so
you've
got
to
book
your
house
right
your
bare
metal
house
right
and
now
you
can
do
two
things.
You
can
either
run
containers
there's
another
virtualization
right.
Well,
we
can
win
únicos
in
some
sort
of
hypervisor
and
my
question
is:
what's
the
I
know
that
unicorns
are
faster
than
using
a
hold
vm,
but
how
do
they
compare
to
a
container?
Which
is
really
that
the
question
here,
because
we
don't
run
the
ends
right.
F
So
let's
go
flex,
but
I
will
show
you
in
a
second
so
anyway,
the
only
thing
what
we
did
with
unique
is
basically
created
doctor
kind
of
like
a
doctor
command
line
for
a
unique
kernel.
It's
very,
very
simple:
to
use
and
I
don't
have
time
so
I
will
not
show
you
demos,
but
it's
basically
you're
building
it.
It
is
a
VM
and
therefore
you'd
need
to
tell
me
where
you
want
to
run,
because
we
need
to
compile
the
drivers
and
then
basically
we're
running
it
for
you.
F
So
when
you're
doing
run,
you
actually
running
it
for
you.
Okay,
we're
supporting
all
of
this
unique
kernel,
all
of
this
and
cloud
provider,
and
but
but
but
but
you
kind
of
like
pink
on
a
point
because
basically
I'm
kind
of
like
transferring
them
both
the
unicorn,
the
unicorn
for
you,
blue
transfer,
the
container
and
the
the
server
itself,
the
VM
I,
don't
have
time
to
do
it
there
more
I
will
talk.
One
second
will
quick
about
the
integration
that
we
did.
So
we
didn't
do
the
right
integration
mainly
because
of
time.
F
The
right
integration
should
be
on
the
garden
right.
Basically,
all
the
purpose
of
garden
is
the
fact
that
it
can
support
different
container
I
think
that
it
can
also
support
different
única,
like
basically
VMs
per
se.
Now
one
thing
that
is
very
important.
You
actually
know
that
so
today,
when
you're
running
your
your
infrastructure
first,
you
need
to
somehow
manage
the
VM
right
and
that's
done
in
the
Bosch
layer
right,
so
budgets
basically
manage
managing
your
VM
and.
A
F
Fair
enough,
ok,
but
but
you
need
to
run
something
to
put
the
container
on
top
of
it.
Then
you
need
to
manage
that
and
I.
Don't
think
that
you
need
to
do.
You
need
to
basically
have
something
like
a
lot
found
area,
kubernetes
or
whatever
to
actually
manage
their
the
container
itself.
That
means
that
you
have
two
different
to
two
minutes.
The
same
the
same.
You
know
the
infrastructure,
which
is
not
the
easiest
way
to
do
with
your
internal.
F
The
idea
is
that
the
application
is
the
VM
and
therefore,
basically,
you
spinning
it
holding
on
demand.
So
basically,
in
that
point,
when
someone
actually
said
I
want
this
application
be
spinning
a
VM
which
means
that
you're
not
managing
them
separately.
You
basically
manage
them
as
an
one
entity
which
one
basically
Cloud
Foundry
will
be.
The
one
is
managing
that.
Does
that
make
any
sense,
yeah.
B
F
I
mean
I
can
show
you
a
demo
what
I
wanted
to
show.
You
is
also
these
things,
because
I
feel
this
is.
This
is
kind
of
like
started
from
unicameral,
but
right
now
it
also
makes
sense
to
run
it,
as
in
you
know,
with
the
cloth.
Honoree
generally
is
basically
a
debugger.
So
in
the
beginning
it
was
a
debugger
for
uni
kernel,
but
now
is
a
debugger
for
microservices
period
and
what
we
created
is
a
project
called
squash
and
the
IDB
squash
is
that
you?
F
Basically
you
know
today
when
you
debug
your
micro-service
application,
usually
looking
at
log
you,
you,
printing
and
look
at
the
log,
and
the
idea
with
this
project
is
that
we
don't
think
that
you
should
do
that.
But
we
do
think
is
that
you
need
to
attach
the
value
and
kind
of
like
we
created
a
distributed
debugger.
F
So
we're
using
right
now
gdb
and
dlv,
but
you
can
have
your
favorite
and
because
we're
using
the
debuggers
you
can
actually
attach
it
to
getting
for
free
supports
for
every
idea
that
you
want
and
what
you're
basically
doing
is
looking
at
the
code
putting
a
breakpoint
deploying
hit
of
Cloud
Foundry,
and
then
you
actually
can
debug
it.
You
know
you,
the
breakpoint
is
kind
of
forgetting
you
see
all
the
stock.
You
see
all
the
stock
of
all
the
application.
Basically,
so
it's
a
microservices
debugger,
which
I
feel
will
be
very,
very
useful.
Okay,.
B
F
E
F
So
we
show
you
so
so
so
squash,
one
sec,
so
unique
itself,
a
very,
very
popular
project
that
it's
already
exist
on
a
the
Gita
and
does
you
see
people
are
storing
it?
People
are
watching
it.
The
graph
is
going
like
if
you're
looking
at
the
traffic
people
are
cloning.
It
people
are
walking
it,
it's
actually
very,
very
popular
project,
so
I
would
suggest
you
to
look
at
it
and
squash
itself.
F
Once
again,
it's
not
an
open-source
project
until
the
opening
to
pick
me,
but
I
will
open
it
probably
beginning
of
next
week,
but
it
will
be
open,
so
you
will
be
able
to
go.
There
go
to
the
pack
head,
go
to
the
platform,
for
instance,
look
at
Cloud
Foundry
and
see
the
integration
there
and
so
on
coupang,
that
is,
for
instance,
we
use
the
kree.
We
also
doing
some
magic
there
with
the
fact
that
we
wanted
to
be
able
to
debug
your
micro
service
from
the
beginning,
so
we're
basically
using
exactly
basically
deploying
the
file.
F
B
Thank
you.
Thank
you.
So
maybe
here
as
if
you
want
to
start
setting
up,
the
one
thing
I
will
say
is
I
mean
first
thing
is.
Thank
you.
A
great
presentation,
I
I,
really
like
I,
mean
one
of
my
goal
and
you
know
if
you
have
feedback
on
this.
That
you
know
is
to
question
assumptions.
You
know
when
you
look
at
cloud
foundry,
it's
great,
but
where
are
we
making
assumptions
that
we
can
question
and
I?
B
Think
idiot
is
essentially
doing
that
and
usually
huge
advancement
for
science
in
general
has
been
that
I
mean
if
you
look
at
any
questions,
sunshine
about
Newton
and
he
made
you
jealous
I'm,
not
comparing
it
to
this,
but
come
you
know,
questioning
assumption
is
extremely
important,
so
thinking
you
did
for
doing
this.
Okay.
So
if
there
is,
you
can
take
it
now.
You
have
15
minutes.
You
have
to
share
your
screen.
Oh.
B
B
E
Yeah
hi
thanks
everybody
I'm
a
product
manager
at
pivotal,
I'd
like
to
talk
to
you
about
Bosch,
backup
and
restore
so
backup,
addressed
its
backup
and
restore
for
Bosch
deployments
and
Forbush
directors,
and
the
way
it
works
is
that
Bosch,
the
bbr
America
VR
defines
a
contract
with
release
authors,
the
binary,
bbr
orchestrates,
the
backup
and
the
release.
Authors
implement
backup
and
restore.
E
So
it's
a
new
model
for
us
of
how
about
how
backup
and
restore
works
in
that
responsibility
for
writing
and
maintaining
the
scripts.
It's
with
the
release
author.
So
that
means
that
the
scripts
are.
The
backup
and
restore
functionality
is
tightly
bound
with
the
release,
so
it
doesn't
get
out
of
sync
and
then,
if
need
be,
the
release
author
can
do
things
to
make
their
release
more
backup
able.
E
So
here
is
a
sample
deployment
with
a
sample
flow,
so
EPR
sits
on
a
jump
box.
The
operator
types
B
be
our
backup
deployment,
B,
be
our
will
SSH
into
the
deployment,
identify
what
scripts
are
implemented
on
what
job
and
then
it
will
call
all
of
the
scripts
of
a
particular
type
across
all
of
it
all
of
the
jobs.
E
So
it
will
call
pre
backup
block
across
all
of
the
jobs
before
it
calls
back
up,
and
then
it
will
call
backup
across
all
of
the
jobs
and
then
it
will
call
post
backup,
unlock
and
then,
after
that,
the
scripts
are
called.
It
will
then
transfer
artifacts
from
the
deployment
back
to
the
jump
box,
and
then
it's
the
operators
responsibility
to
manage
those
artifacts
and
also
to
do
things
like
scheduling
and
in
encrypting
artifacts,
and
things
like
that.
E
E
So
here
is
an
example
of
how
it
would
look
like
to
back
up
Cloud
Foundry,
so
you
call
backup
on
the
director
and
Cloud
Foundry
and
then
it
does.
This
is
just
a
more
specific
version
of
that
other
diagram.
So
you
can
see
that
what
we're
what
we're
do?
What
we'll
do
first,
is
with
it
we'll
take
down
the
Cloud
Foundry
API
within
the
pre
backup,
lock
script,
so
that
we
don't
mutate
state
during
the
backup.
E
Then
we
will
call
a
backup
of
the
Boche
of
Bosh,
which
will
backup
Postgres
and
the
blobstore
ua
a
cred
hub,
my
sequel
and
the
blobstore
in
cloud
foundry,
and
then
we
will
call
unlock
and
then,
after
that,
all
the
artifacts
will
get
transferred
from
the
deployments
to
the
jump
box.
So
it's
a
way
to
take
consistent,
backups
of
a
distributed
system.
E
We're
looking
at
optimizations
like
calling,
although
the
backup
scripts
in
parallel
other
other
improved
communication,
orchestration
looking
at
providing
bbr
as
a
library
or
an
API,
and
then
looking
at
things
like
pluggable
artifact
destinations
like
s3,
for
example,
so
I
should,
which
shall
I,
give
you
a
demo
now.
Yes,.
E
E
B
C
B
Backup
restore
your
desktop,
so
what
question
for
you,
the
res,
is
as
part
of
this
flow
I.
Imagine
that
there
are
do
you.
Imagine
if
I
have
five
different
is
including
the
director
of
course,
and
so
on
that
they'll
be
those
backups
so
happening
in
parallel,
and
you
will
orchestrate
you
know
essentially
getting
a
snapshot
for
the
entire
platform
or
you
see
it
as
your
backup.
We
store
each
one
independently
and
they
they
they
are
sort
of
separated.
B
E
It
you
can
do
you
can
backup
all
the
things
at
once,
so
you
can,
for
example,
backup
cloud
foundry
and
bashed
at
the
same
time,
by
going
be
be
our
backup,
CAF
comment
director,
it
depends
on
how
sensitive
you
are
to
downtime,
because
when
you
lock,
when
you
call
lock
for
a
cloud
foundry,
it
takes
down
the
Cloud
Foundry
API.
So
if
you
have
a
long
period
where
you
need,
if
you're
very
sensitive
to
engineers
not
being
able
to
see
f.push,
then
you
might
not
want
to
do
that,
but
it's
completely
possible
and
we
will.
E
A
E
A
E
C
My
impression
is
that
what
PBR
does
is,
gives
the
structure
and
then,
as
30
said,
'let's,
then
up
to
the
release
author
to
decide
which
portions
of
that
structure
to
use.
So
it
seems
quite
plausible
that
the
teams
working
on
the
elements
of
cloud
foundries
that
would
be
relevant
might
look
at
how
they
iterate
what
they
store,
such
that
they
use
just
the
pieces
of
the
vbr
that
they
need.
That's.
B
G
Constantine
is
also
on
the
call
he's
been
one
of
the
core
engineers
he's
the
anchor
on
the
project
right
now
so
kind
of
diving
right
in
so
consumers
right.
They
have
a
lot
of
choices
out
there
right
now
for
things
like
safe
functions
as
this
functions,
absent
Rick
system
is
like
with
Cloud
Foundry
container
orchestrations
like
with
kubernetes
and
then
of
course,
there's
data
services
and
a
lot
of
this
stuff
sits
on
top
of
infrastructure
and
I'm.
G
G
G
So
while
you
may
have
an
application
that
can
handle
some
fault
tolerance
because
of
kubernetes,
if
you
were
to
lose,
say
a
virtual
machine
running
on
your
cloud
provider,
kubernetes
does
not
take
care
of
that
for
you
out
of
the
box.
So
this
is
where
we
look
at
Bosch
to
kind
of
come
in
and
solve
some
of
these
problems
for
us
with
for
kubernetes.
G
So
one
of
the
things
I
think
that
we're
looking
at
doing
as
far
as
where
we
want
to
take
the
project
is
provide
a
little
bit
more
high
availability
of
the
kubernetes
constructs
the
masters
as
well
as
the
worker
nodes,
and
then
the
sed
nodes
that
sit
alongside
that
scaling
is
another
where
area
where
we
can
help
kubernetes.
So
a
lot
of
times
today,
I
think.
If
you're
running
kubernetes
depending
on
the
infrastructure
provider,
you
provision
a
cluster,
you
define
a
certain
set
of
nodes.
G
If
you
want
to
scale
beyond
those
nodes,
you're
kind
of
stuck,
so
we
can
leverage
the
power
of
Bosch
there
as
well
to
provide
scaling
of
those.
For
those
of
you
are
obviously
very
familiar
with
Bosch.
It
already
takes
care
of
things
like
health
checks
with
a
resurrector,
so
it
can
bring
back
nodes
that
fail
and
then
obviously
upgrades
as
well.
G
So
Bosch
is
a
great
great
tool
for
being
able
to
do
rolling
updates
across
your
infrastructure,
and
so
that's
another
area
where
we
can,
where
kubernetes
will
be
able
to
benefit,
is
through
upgrades,
so
I've,
just
sort
of
evolution.
I
think
everybody
is
very
familiar
with
Google's
board,
at
least
conceptually
how
we
run
containers
internally.
That
was
some
of
the
concepts
there
were
led
into
Bosch
and
then
what
we're
doing
now,
obviously
with
Bosch
and
kubernetes,
leading
to
the
project
takubo.
G
So,
like
I
was
just
saying
day,
one
activities
right,
so
how
do
you
install
Bosch
or
sorry?
How
do
you
install
kubernetes,
there's,
obviously
many
ways
out
there
that
you
can
do
that
today,
so
Bosch
will
provide
yet
another
way
of
installing
kubernetes,
but
I
think
that
the
most
important
part
of
this
is
obviously
the
day
two
activities.
So
you've
got
your
cluster
up
and
running.
How
do
you
manage
the
BMS,
how
you
scale,
how
you
do
upgrades
of
kubernetes
and
then
obviously
adding
capabilities
for
high
availability,
high
availability.
C
G
Multi
AZ's,
when
we
originally
started
with
the
project
with
pivotal.
A
lot
of
this
was
to
make
kubernetes
available
to
clog
battery
applications,
so
the
evolution
of
the
project
I
think
when
we
started.
This
was
to
obviously
have
a
kubernetes
cluster
that
was
managed
by
Bosch,
but
then
sit
kind
of
alongside
of
cloud
foundry,
that's
also
managed
by
Bosch,
and
so
as
applications
would
be
set
up
in
kubernetes.
We
would
propagate
that
information
over
to
the
Cloud
Foundry
system.
G
What
we're
looking
at
doing
as
far
as
the
proposal
to
get
into
Cloud
Foundry
foundation
is
to
kind
of
backpedal
a
little
bit
away
from
that
coupling
with
Cloud
Foundry
and
look
at
creating
just
the
standalone
Bosch
release
for
kubernetes,
and
then
pivotal
will
be
continuing
to
make
that
available
as
a
product
through
an
on-demand
service
broker.
That
will
still
work
well
with
Cloud
Foundry,
but
we'll
be
able
to
kind
of
have
the
best
of
both
worlds
there
so
roadmap.
G
G
So
we
talked
a
little
bit
about
that
here
in
the
slide.
Networking
persistence,
adding
high
availability
support,
so
multi
master
configurations
for
kubernetes
sitting
behind
infrastructure,
load,
balancing
or
separate
load
balancers.
This
is
not
something
that
kubernetes
has
out
of
the
box,
but
relies
basically
on
the
infrastructure
configuration
for
kubernetes
in
terms
of
how
to
provide
that
load,
balance
capability
for
multi
masters,
and
we
can
do
a
lot
of
that
with
Bosch
sitting
with
the
infrastructure
we're
also
looking
at
where
kubernetes
is
headed
in
general
upstream.
G
One
of
the
places
that
they're
investigating
is
moving
towards
more
of
a
self-hosted
model,
so
instead
of
having
a
separate
@cd
cluster
on
its
own
set
of
the
ends,
they're
actually
looking
at,
how
do
you
bootstrap
the
sed
capabilities
inside
of
kubernetes,
and
so
kubernetes
itself
becomes
the
management
platform
for
the
SED
cluster,
and
so
what
we
can
do.
Moving
forward
with
the
Bosch
release
is
to
start
aligning
in
that
direction
as
well
and.
G
G
Architecture,
diagram
for
the
standalone,
open
source
version.
This
is
where
we're
headed
as
far
as
the
cloud
foundry
foundation,
CF
experiment.
What
we
want
to
be
able
to
do
is
obviously
just
have
Bosch
manage
the
kubernetes
cluster,
and
so
the
diagram
shows
you
know
the
sed
nodes,
multiple,
multiple,
multiple,
multiple
masters,
as
well
as
a
set
of
workers.
G
The
Kubo
architecture
with
Cloud
Foundry-
this
is
sort
of
where
the
project
sits
today,
so
as
services
are
brought
up
in
kubernetes,
the
team
that
worked
on
propagating
that
information
over
into
cloud
foundry
as
well
so
they're
using
route
sync
and
an
API
route
register
that
feeds
that
information
over
into
Cloud
Foundry
and
then
through.
Networking
Cloud
Foundry
at
or
kubernetes
apps
are
surfaced
through
Cloud
Foundry,
so
you're
actually
going
through
the
go
router
to
be
able
to
talk
over
to
the
kubernetes
cluster.
G
So
question:
what
does
this
mean
to
club
hungry?
Obviously
so
Cloud
Foundry
is
a
great
place
for
running
a
lot
of
apps.
There
are
probably
certain
cases
where
you
may
have
requirements
for
your
applications,
where
Cloud,
Foundry
isn't
necessarily
a
great
fit.
Kubernetes
could
be
a
solution
for
that.
G
So
I
think
what
this
does
is
that
this
allows
users
to
be
able
to
look
at
kubernetes
as
a
way
of
deploying
some
of
their
application
workloads,
as
well
as
on
top
of
Cloud
Foundry
and
then
through
the
work
that
pivotal
is
going
to
continue
doing
be
able
to
make
those
apps
available
to
each
other
through,
probably
even
the
open
service
broker
API.
Those
are
future
forward-looking
kind
of
places
that
we
want
to
see
the
project
go.
G
You
ventually,
and
so
the
pivotal
of
Google
kind
of
partnership
on
this
Google's
been
working
closely
with
Cloud
Foundry
ecosystem
over
the
last
year.
I
want
to
say
to
build
out
capabilities
for
running
Cloud
Foundry
on
top
of
GCP
and
have
developed
a
good
relationship
with
pivotal
along
the
way,
and
this
is
just
sort
of
a
natural
extension
of
where
we
were
going
with
the
work
and
I
think
it's
great
for
users
to
be
able
to
access
both
kubernetes
and
Cloud
Foundry
through
the
infrastructure
of
choice
which
would
be
Bosch.
G
C
E
G
Yeah,
so
we're
actually
looking
at
using
the
new
bosch,
DNS
capabilities
that
was
talked
about
actually
earlier
in
this,
so
we're
not
using
a
vanilla,
Bosch
director.
Currently
we
are
looking
at
sort
of
somewhere.
The
direction
of
wash
director
is
going
with
things
like
red
hub
and
pulling
in
the
DIA.
That's
capabilities
into
that,
so
those
would
probably
be
things
that
we'll
continue
to
leverage
and
not
rely
so
much
just
on
power.
Dns.