►
From YouTube: Keynote by Dustin Kirkland
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Howdy,
my
name
is
Dustin
Kirkland
I
from
Austin
Texas
and
I
work
for
canonical
on
Ubuntu.
It's
in
fact
invited
here
by
Matt
and
a
team
to
talk
a
little
bit
about
ZFS
and
related
utilities,
as
you
might
find
them
in
in
a
bun
too.
So
first
off
I
want
to
pose
this
question
and
I'd
like
you
to
think
about
this
for
the
next
few
minutes.
A
While
I
talk
about
a
few
other
things,
but
I'd
like
you
to
think
about
what
are
the
three
things
for
the
Linux
users
in
the
room
understand
not
everyone
here
is
a
Linux
user,
but
for
the
Linux
users
in
this
room.
What
are
the
three
things
you
miss
most
from
your
Solaris
or
PSD
or
AIX
or
other
UNIX
days?
A
We
have
a
couple
of
offices
and
in
London
Boston
Austin
and
Taipei
Shanghai
and
Beijing,
but
for
the
most
part
we
were
this
sort
of
vagabond
traveled
the
world
hack
on
software
and
live
where
you
want
to
live.
We've
been
that
way
since
day,
one
it's
one
of
the
core
virtues
and
values
of
the
of
the
company.
A
So
let
me
ask
you
where,
where
might
you
find
a
bun
to
today
and
it's
it's
really
a
little
bit
of
everywhere,
especially
it's?
It's
funny.
Matt
mentioned
something
like
Netflix,
that
was
shipping
DVDs
and
what
you
know
a
few
years
ago,
certainly
not
more
than
a
decade
ago,
but
you
look
at
how
many
of
these
organizations
perhaps
you're
from
some
of
these
organizations
actually
so
many
of
them
have
basically
developed
within
the
last
five
years,
and
so
many
of
them
have
this.
A
This
resource
sharing
as
as
a
core
tenant
to
that
to
that
to
their
business
model,
and
you
know
they're
their
software
underneath
underneath
them
actually
aligns
quite
well
with
that.
So
you
know
you're
talking
about
ride-sharing
lyft
and
uber
you're
talking
about
house
sharing
Airbnb-
and
you
know
from
that
perspective-
we're
really
in
this
peer-to-peer
software
and
technology
age,
but
it's
actually
moved
into
the
real
world.
A
Now,
when
you
find
a
bun
to
is
actually
sort
of
the
the
secret
to
a
bun
to
success,
I
would
say
as
the
product
manager
responsible
for
it.
It's
the
fact
that
we
release
on
time
every
time
and
we've
been
doing
this
for
12
years,
we're
working
on
our
25th
release
of
Ubuntu
Ubuntu
1610
will
release
next
next
month.
October
of
2016
we've
done
that
we've
done
that
on
time
every
time
and
that
track
record
is
pretty
remarkable
for
releasing
an
entire
operating
system.
A
Every
fourth
release
of
Ubuntu
is
an
LTS
release,
a
long-term
support
release.
Those
are
the
enterprise
releases.
The
developer
releases
are
the
ones
we
release
every
six
months,
and
you
know
that's
what
maybe
many
of
the
people
in
this
room
might
gravitate
toward
for
a
desktop
or
a
developer
environment.
While
you
know
the
organization's
on
the
previous
slide
certainly
focus
on
the
LTSs
and
then
the
other
secret
to
I'd
say
a
bun
to
success
is
the
vast
variety
of
open
source
packages
and
software
that
get
released
every
six
months
on
time.
A
Every
time
that's
over,
55,000
binaries
are
built
from
over
26,000,
open
source
projects
and,
of
course,
we're
proud
to
have
CFS
as
one
of
those
and
integrated
into
the
Linux
kernel
as
of
1604,
and
we
can
do
that
on
quartz
containers
in
virtual
machines
and
across
a
wide
array
of
hardware
partners
and
that's
essentially
bun
2
everywhere
everywhere
on
time.
Every
time
we
put
those
into
public
clouds,
Amazon
Google
Azure,
of
course,
software
Rackspace,
Oracle
giant
cross
the
board.
We
ensure
that
abun
too,
is
in
those
clouds.
It's
working
well,
it's
its
q8.
A
It's
tested
well,
it's
up
to
date
and
and
secured,
but
also
in
the
scale-out
private
clouds
as
well.
That's
the
mais
au
s--,
the
kubernetes,
the
hadoop,
the
OpenStack
cloud,
foundry
a
bunch
of
running
under
those
clouds
and
also
in
those
clouds.
As
a
basis
we
run
on
some
of
the
biggest
supercomputers
in
the
world
from
the
IBM
Z.
A
This
is
the
chiana
to
the
supercomputer
in
China,
but
also-
and
this
is
coming
on
extremely
strong
right
now
on
internet,
a
device
type
Internet
connected
devices,
printers
drones,
robots,
that's
that's
a
boon
to
across
the
board,
and
it's
really
interesting
to
see
how
bringing
an
open-source
platform
to
those
devices
really
stimulates.
Innovation.
A
This
slide
is
only
in
here
because
I
was
told
the
slides,
they're
gonna
end
up
on
a
on
a
on
a
website
somewhere.
We
want
to
just
make
sure
everyone
knows
that
we
do
so
support
for
for
a
bun
all
right.
Let's
come
back
to
that
question.
I
posed
a
few
minutes
ago.
What
are
the
three
things
you
miss
most
from
Linux,
UNIX
or
sorry,
UNIX
BSD
Solaris
in
Linux.
A
A
Indeed,
indeed,
much
much
smaller
scope
in
many
cases,
any
others
all
right.
So
it's
a
bit
of
a
loaded
question,
but
that
that's
actually
you
know
that's
the
product
managers
dream
to
sit
in
front
of
a
bunch
of
smart
people
and
ask
them
you
know.
What
do
you
want
to
see
in
in
a
product?
And
here
the
product
is
the
OS,
the
Linux
operating
system.
So
that's
fantastic,
the
first
one
I
would
say
you
know
you
don't
miss
this
anymore,
because
it's
actually
well
integrated
into
into
Linux.
A
But
that's
the
concept
of
containers
right,
that's
jails,
that's
zones,
and
now
we
see
various
different
types
of
containers
that
are
part
of
our
everyday
experience.
Inside
of
inside
of
Linux,
of
course,
that's
Linux
containers,
Lexi
Lexi.
Everyone
is
familiar
with
docker
and
the
concept
of
application
containers.
You
know
we
finally
have
I
think
in
Linux
a
bit
of
taxonomy
some
nomenclature
that
helps
us
understand
that
not
all
containers
are
the
same
containers
right
that
that
docker
has
a
place
that
machining
their
system.
A
Containers
have
have
a
place
and
I'll
demo
that
here
in
a
second
the
second
one,
you
don't
miss
any
more
because
we
we
also
have-
and
that's
at
ZFS-
that's
really
really
advanced
file
systems-
a
file
system
inside
of
inside
of
Linux.
We
certainly
take
advantage
of
ZFS
I'll
demo
that
in
in
a
few
minutes
here
well,
we
take
advantage
of
ZFS
for
all
of
these
things,
snapshots
backups
instant,
start
time,
the
efficient
compression
and
automatic
repair
the
integrity
checking
it's
it's
brilliant.
It's
super
useful
and
we
make
use
of
that
every
single
day.
A
We're
good
okay.
So
this
is
an
instance
in
in
Amazon
in
NEC
to
where
I'm
you
know,
listing
a
zpool
that
was
automatically
created
for
Lex
deal.
Xt
is
a
container
hypervisor.
It
provide
system.
Containers
I've
also
got
docker
running
here.
Let's,
let's
actually
start
with
docker
so
I'll.
You
know,
sudo
docker
run
I'll
need
this.
A
Because
everyone
is
watching
right
now,
all
right,
hot
dog
I'm
in
docker
and
I'm
running
one
single
process.
That's
a
process
container!
We
put
docker
around
a
single
process
here,
I'm
running
bash
and
I
can
do
you
know
bash
e-type
things.
I
can
take
a
look
at
a
slightly
structured
proc
and
and
have
a
have
a
poke
around
if
I
exit
that
that
container
is
gone.
A
If
I
wanted
to
run
some
other
process,
if
I
run
top
I'm
running
one
process
inside
of
a
docker
container
and
that's
the
that's
the
best
practice,
that's
using
some
of
the
concepts
of
jails,
wrapping
a
process
and
keeping
that
process
separate
from
the
rest
of
the
system.
I
can
I
can
drop
back
into
another
container
and
then
you
know
apt,
update
an
apt,
install
and
and
throw
some
software
inside
of
that
container
right.
A
It's
gonna
create
dashing
bonefish
and
if
I
list
I've
now
got
a
dashing
bonefish
running
and
in
a
second
later
I've
drawn
an
IP
address
here
and
if
I,
if
I
exec
into
that
a
bash
shell
I'm
now
inside
of
that
container
and
if
I
run
PS
or
PS
tree
I
see
what
looks
like
a
machine,
certainly
more
than
than
a
docker
container.
It's
running
s
been
in
it.
It's
running
system
D.
At
the
top
of
that
process
tree,
we
forked
off
a
journal,
we're
running
cron
and
at
so
that
you
know
this
system.
A
If
it
stays
up,
it
can
do
cron
things
and
it
can
do
it
can
schedule
things
that
to
happen
later,
it's
drawn
an
IP
address
using
using
D,
H
client.
It's
actually
running
an
SSH
server,
so
you
know
I
could,
if
I
had
keys
exchange
that
could
SSH
into
localhost
or
come
into
it
from
outside
for
all
practical
purposes.
A
A
A
A
A
There
we
go
wrapped
a
little
bit
here,
yeah
exactly
so,
you
know
we
really.
We
really
love
seeing
ZFS
underneath
flex
D.
It
lets
us
do
some
pretty
amazing
things,
including
snapshots.
So
if
I
want
to
I'm
gonna,
do
an
Alexi
list
and
I'm
gonna
take
this
first
one
grizzly,
I'm,
gonna,
say
Alexi
snapshot.
A
Absolute
grizzly
and
let's
just
create
some
havoc
right
so
now,
I
have
no
been
directly
directory
if
I
exit
out
of
there
and
I
try
to
come
back
in
and
exec
bash.
Clearly
that
container
is
host
right.
Let's
imagine
something
has
gone
wrong
with
with
some
service,
but
if
I
want
to
restore
that
snapshot
back
to
a
point
before
I
RM
are
effed
all
of
user.
A
You
can
actually
snapshot
with
state
and
store
the
memory
as
as
well
and
that's
how
we
live
migrate,
containers
from
one
system
to
another
by
snapshotting
and
capturing
the
the
state
of
the
memory
and
the
CPU
registers,
and
we
can
seamlessly
migrate
a
container
from
one
one
to
another.
Zfs
is
fantastic
for
for
its
snapshot
capability
under
under
Lex
D.
Here
Matt
mentioned
Iost
at
we
do
enjoy
Iost
at
here.
A
Which
is
a
way
of
applying
some
stress
to
a
system?
Let's
say
it's
a
way
that
he
unfurls
and
attacked
some
really
interesting
bugs
in
and
around
linux
itself,
and
so
I've
got
a
little
cheat
sheet
here.
The
command
I
want
to
run
stress
in
G,
looks
like
a
couple
of
threads.
We're
gonna
just
start
really
killing
them.
A
A
A
So
that
includes
the
Berkeley
packet
filter
and
the
features
that
we
need
from
BPF
to
perform
really
introspective
tracing
into
the
kernel.
The
user
space
side
of
that
is
something
called
IO
visor
and
it's
part
of
the
BCC
tools
written
by
Brendan
Gregg
at
Netflix.
So
many
times
we
can
mention
Netflix
in
a
in
a
day
right,
so
Brendan
wrote
this
the
set
of
tools,
utilities
that
make
it
a
lot
easier
to
to
build
traces
to
insert
into
the
kernel
and
he's
shipped
a
number
of
interesting
tools,
a
few
of
which
I
can
demo
here.
A
What's
really
interesting
is
that
we're
bringing
those
tools
directly
to
abun
to
1604,
and
it
gives
in
many
ways
even
deeper
introspection
than
dtrace
has
ever
been
able
to
do
this
berkeley
packet
filter
started
as
a
network
raw
device
access
tracing
utility,
but
now
there's
access
into
all
sorts
of
interesting
places
in
memory
and
along
the
file
systems
and
various
drivers.
So
let's,
let's
poke
at
this
a
little
bit.
A
I'm,
just
gonna,
let
that
keep
going
so
down
here.
I
will
use
this
opportunity
to
also
introduce
a
third
application,
our
third
container
type,
which
your
application
containers.
So
we
have
a
new
packaging
system
in
Ubuntu
called
snaps.
A
snap
is
like
a
Deb,
but
it
actually
sort
of
wraps
a
container
around
a
given
application.
It's
different
than
docker
and
that
docker
has
its
own
file
system
and
every
docker
instance
has
its
own
copy
of
the
file
system.
A
Now,
BCC
and
IO
visor
and
tracing
in
general
requires
pretty
unlimited
access
to
the
to
the
kernel
right
so
that
one's
a
bit
of
a
different
case,
but
snaps
themselves
have
some
really
interesting
advantages
over
over
Deb's
in
the
way
we
package
software
into
a
gun.
The
first
advantage
is
that
we
can
very
easily
bring
those
older
packages
in
sorry
bring
those
newer
packages
into
older
abun
at
any
time
at
any
pace.
A
We've
always
had
this
difficulty
with
with
the
release
cycle
of
a
bun
that
we
were,
you
know
never
more
than
6
months
out,
but
in
6
months
some
really
neat
things
happen
and,
of
course,
over
2
years.
Some
really
interesting
things
happen.
So
snaps
give
us
an
interesting
way
of
canonical
bringing
new
software
into
a
bun,
but,
more
more,
more
interestingly,
any
developer
anywhere
making
that
software
available
and
available
at
their
pace
as
opposed
to
Ubuntu
space.
So
I
have
installed
here
a
couple
of
snaps.
A
Most
of
these
need
a
pseudo
in
front
of
it
and
most
of
them
I'm
very
new
to
and
so
we're
gonna
look
at
the
help
before
I
get
us
into
trouble
with
with
any
particular
1b.
I/O
latency
is
really
interesting.
It's
a
it's
a
block
device
as
a
as
a
histogram.
You
know
looking
at
the
latency
of
that
block
device.
A
So
if
I
run
sudo,
bio,
latency
and
I
need
an
interval
two
seconds
an
account,
we
should
see
some
block
device
latency
coming
up
and
and
where
that
latency
is
falling
from
an
from
a
histogram
perspective.
Of
course
we
can.
Let
that
keep
running
will
fire
up
another
one.
What
else
is
interesting?
Cache
stat,
so
sudo
BCC
cache
that.
A
A
Okay
and
interestingly,
we
can
see
our
our
stressor
running
as
well
and
what's
getting
what's
getting
hit,
what's
getting
touched
from
an
IO
top
perspective
or
a
file
top
perspective,
we
can
look
at
heart.
Iqs
mm
leak
I'd
actually
be
interested
to
see
if
anyone
wants
to
play
with
this
on
on
Anna
Bunty
system-
and
you
know,
give
us
some
feedback,
there's
a
couple
of
utilities
that
are
specific
to
ZFS-
that's
actually
built
into
BCC.
So
let's
have
a
look
at
those.
A
A
There
we
go
thank
you
for
making
this
a
very
busy
system
calling
so
yeah.
So
we're
excited
about
this.
This
is
this
is
a
really
interesting
pool
of
very
low
level
utilities,
we're
taking
advantage
of
some
interesting
new
features
in
the
Linux
kernel
and
we're
bringing
this
of
this
fantastic
user
space
from
Brendan
directly
into
the
Ubuntu
archives
through
the
the
new
snap
that
we've
created
for
that.
A
A
Can
certainly
go
and
get
clone
it
and
pull
it
up.
I
should
give
props
to
Brendan
Brendan
here.
Io
visor
is
the
name
of
the
project
BCC.
It's
interesting
I
had
a
meeting
at
Netflix
a
few
months
back
and
Brendan
was
asking
how
to
package
this
as
Deb's
he's
a
brilliant
kernel
developer
in
in
his
words
and
I
quote:
I
just
hate
packaging
I
want
someone
else
to
just
package
this
for
me
and
the
beauty
of
a
snap
is
that
a
snap
as
a
really
simple
package?
It's
a
tiny,
little
Yama
file.
A
That
explains
where
things
go
and
what
permissions
are
necessary.
And
then
you
run
a
utility
we've
created
called
snap
craft.
You
run
snap
craft
on
your
enamel
file
and
out
the
other
end
comes
a
dot
snap
file,
which
is
just
a
it's
just
a
an
archive.
It's
a
really
simple
little
little
tar
ball,
Alaric
I've
I
should
mention
that
the
back
end
snap
D
is
actually
available
on
on
basically
all
Linux
operating
systems.
It
makes
use
of
system
D
of
cgroups
comm
app
armor
to
sort
of
piece
things
together.
A
I
certainly
encourage
you
to
try
this
out
and
let
us
know
what
you
think
very
soon:
you'll
be
able
to
just
run
snap
install
BCC
on
any
into
1604
system
and
you'd.
Have
it
it's
a
you
know,
an
alternative
to
apt,
install
some
software
1604
ships
with
snap
D
installed
and
no
snaps
at
first.
You
know,
but
you
can
snap
find
and
install
software
that
way.
A
Right
so
that
question
is
directed
to
me:
can
Dustin
do
that
not
yet,
but
can
you
do
that
in
the
answers?
Yes,
absolutely.
This
is
back
to
the
to
the
github
page,
but
it
explains
that
a
sent
of
the
BCC
is
the
compiler.
It's
a
compiler
collection
he's
shipped
a
number
of
interesting
things,
but
you
can
you
can
drop
in
your
own,
your
own
tracing
and
it's
it's
super
dynamic
super
interesting,
and
you
know
this
goes
through
a
bit
of
how
it
works
and
where
it
works.
A
A
A
A
I've
got
five
of
them.
Lex
D
itself
is
a
REST
API
or
has
a
REST,
API
and
I
could
literally
type
lxc
move
dashing
bonefish
to
some
other
host
:.
If
I've
exchanged
the
SSL
Certificates
properly,
and
then
this
this
instance
gets
snapshot,
moved
restart
it
and
then
delete
it
at
the
source.
Okay,
the
slowest
part
of
that
operation
is
actually
in
moving
the
underlying
filesystem.
A
Okay,
now
we're
using
ZFS
by
default.
We
love
that
locally
for
really
fast
snapshots.
Okay,
as
it
turns
out,
all
four
of
these
machines
have
exactly
identical,
precisely
identical
starting
point:
the
abun
to
1604
starting
point,
it's
exactly
identical
and
in
fact,
it's
probably
already
over
there
on
that
other
system
and
whether
it's
a
couple
of
megabytes
or
a
couple
of
hundred
megabytes
I'd
like
to
not
have
to
our
sync.
A
B
A
Okay,
interesting
so
there's
a
second
part.
I'll
get
that
question
a
second
there's
a
second
part
that
to
that
and
there's
a
second
place
where
I
would
love
to
see
us
use
ZFS,
but
we
can't
today
and
that's
in
snaps
and
cents
and
snap
snap
Ubuntu.
So
we
have
this
other
version
of
Ubuntu
that
we're
working
on
right
now
called
a
bun
2
core.
It's
a
transactionally,
updated
read-only
immutable
operating
system.
So
your
root
filesystem
can't
change.
A
When
you
do
change
it,
you're
upgrading
it
and
then
we
just
splat
a
whole
nother
copy
of
slash
root.
On
top
of
that,
ok,
there
is
readwrite
space
and
that's
where
applications
live
and
each
application
has
their
own
bit
of
readwrite
space
where
they
can,
they
can
read
and
write
we
flip-flop
back
and
forth.
We
can
flip-flop
back
and
forth
between
any
version
of
the
OS
that
we
have
a
copy
of.
We
can
flip-flop
back
and
forth
between
any
version
of
the
applications
that
are
installed
as
well.
A
The
snapshot
aspects
and
the
copy-on-write
would
provide
some
really
nice
advantages
for
the
wholesale
upgrade
of
the
of
the
OS
for
the
rolling
back
and
forward
applications
as
well.
It's
a
little
bit
related
to
that
to
the
to
the
previous
one,
but
you
know
that's
we're
a
little
bit.
You
know
myopic
in
that.
The
operating
system
is
what
we
care
about.
There's
another
question.
A
A
A
A
A
Yes,
absolutely
so
I
I
should
show.
I
should
have
I
set
this
up
ahead
of
time,
so
I
I
didn't
the
question
was
the
question
was
the
the
setup
of
ZFS
for
Lex
T
and
if
this
is
available
in
1604,
so
a
couple
of
things.
First
of
all,
the
ZFS
kernel
object
is
part
of
the
abun
to
kernel.
Now
we
build
that
out.
A
tree
will
build
that
into
the
tree.
Colin
and
Aaron
here
are
largely
responsible
for
that.
A
The
utilities,
the
ZFS
Linux
utilities,
is
a
package
that
we've
also
brought
into
a
bun
too,
as
well
Aaron's
responsible
for
that
package
and
works
with
the
Debian
community
to
try
to
keep
those
two
in
sync,
the
integration
of
ZFS
and
lex
d
together
is
automated
by
a
script
that
we
wrote
called
ZFS
in
it.
I'm
sorry,
Lex,
T
and
knit
and
I
already
ran
that
on
this
machine.
So
I'll
need
to
find
another
place
to.
A
It
comes
along
with
Lex
D
itself,
so
apt,
install
XD,
apt,
install
XD,
actually
Lex
these
ships
as
part
of
part
of
a
bun
to
apt
installed
ZFS
will
bring
the
user
space
utilities
the
kernel
pieces
there,
the
user
space
utilities
is,
you
know
one
command
away,
we're
actually
working
on
bringing
that
that
into
the
default
into
the
default
image
right
now.
This
is
a
technical
detail,
but
right
now,
ZFS
Linux
utilities
depends
on
Python
to
seven.
A
We've
done
a
lot
of
work
to
take
Python
2
out
of
the
default
image,
so
we're
just
trying
to
make
sure
that
those
Python
utilities
still
work
with
Python
3
or
we
move
them
out
or
fix
them
and
upgrade
them
as
soon
as
we've
got.
That
and
ZFS
only
depends
on
Python
3.
We
can
bring
the
ZFS
package
into
the
default
instance
at
which
point
Lex,
D
and
ZFS
are
just
there
by
default.
Always
the
one
thing
that
you
have
to
do
to
make
Lex
D
and
ZFS
work.
A
Well
together
is
sudo,
Lex,
T
and
net
running
that
command
will
prompt
you
with
a
couple
of
interactive
questions.
Hitting
entered
will
do
the
right
thing,
but
it'll
ask
you
I'm
trying
to
think
how
quickly
I
could
get
this
set
up
somewhere,
I'll
demo
it
for
you
later,
but
basically
it'll
ask
you:
do
you
want
to
use
ZFS
or
our
directory
for
Lex
T
and
you
type
ZFS,
which
is
the
default,
then
it'll
say:
do
you
want
to
use
a
block
device,
or
do
you
want
to
use
a
sparse
file?
A
It's
just
a
backing
file
and
the
defaults
of
backing
file
because
that's
safe,
and
we
can
do
that
everywhere
for
the
demos
that
I'm
doing
here
and
almost
anywhere
I
set
this
up.
I
always
use
just
a
raw
SSD
as
the
backing
device
for
for
ZFS,
but
by
default
will
create
a
10
or
20
gig,
sparse
file
and
then
essentially
block
mount
loop
mount
that
file
and
then
create
a
zpool
on
top
of
that
that
block
file,
which
has
a
few
limitations
but
generally
works,
works
well
in
the
in
the
nominal,
the
nominal
case.
A
A
You
can
just
tell
it
to
use
that
existing
pool
I
doubt
I'll
be
able
to
start
an
instance
in
time,
but
the
SU
dual
XD
in
it
and
you'll
see
all
the
questions.
It's
a
nice
little
set
of
questions,
so
I'm
told
I'm
out
of
time.
This
was
fantastic.
Thank
you
all
for
ZFS,
and
you
know
thanks
for
listening,
you.