►
From YouTube: State of the Union by Matt Ahrens
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
Good
morning,
everyone
so
we're
gonna
get
started.
My
name
is
Paul
Daniel
Lee.
Welcome
to
the
open,
ZFS
developer
summit
2016.
We
have
a
lot
of
great
talks
lined
up
for
you
today,
so
I'm
not
gonna
waste.
A
lot
of
time
jabbering
without
any
further
ado
here,
is
Matt
Aaron's
to
give
the
state
of
ZFS
presentation.
B
Thanks
and
welcome
to
the
fourth
annual
open,
ZFS
developer
summit,
I'm,
Matt,
Aaron's
I
hope
to
create
this-
opens
the
efest
thing
as
well
as
ZFS
back
in
the
day
at
Sun.
So
first
I
want
to
thank
our
sponsors.
If
you
are
from
one
of
these
companies,
please
stand
up.
First
off
we're
gonna
go
platinum.
Sponsors.
These
guys
gave
us
enormous
amounts
of
platinum,
first
off
dad
Oh
anybody
here,
I
know,
there's
some
folks
here
from
dad.
Oh
thank
you
guys
and
from
deltax
yeah
thanks
thanks,
Adele
VIX
is
also
my
employer.
B
B
B
B
We
have
a
bunch
of
a
bunch
of
things
that
they're
financing.
Besides
the
stuff
that
you're
going
to
see
and
eat,
and
here
we
have
a
live
video
streaming
going
out
on
the
internet
right
now.
So
if
you
have
friends
and
colleagues
that
couldn't
attend
on
point
them
to
our
website
open
to
ZFS
org,
where
there's
a
live
stream
of
this
we're,
also
recording
it
it'll
be
on
YouTube,
and
this
year,
I
have
been
promised
that
we
will
have
audio
recordings
that
you
can
hear
out
of
both
of
your
ears.
B
B
Last
but
certainly
not
least,
I
wanted
to
thank
some
individuals
who
have
spent
a
lot
of
time
in
their
own
effort
on
this.
So
thank
first
off
thanks
to
Aaron
holding.
He
has
designed
the
open
ZFS
t-shirts
for
the
past
three
years,
so
you
have
him
to
thank
for
the
really
cool
scene
of
San
Francisco
there
on
it.
He
designed
all
of
our
branding
and
t-shirts,
and
unfortunately
he
couldn't
be
here,
but
thank
you
very
much
Aaron.
If
you,
if
you're
listening
on
the
livestream.
B
Next
I
wanted
to
thank
the
del
phix
p.m.
team,
so
these
are
the
folks
that
we're
doing
all
the
behind
the
scenes
work
to
get
the
food
here
on
time
get
the
work
with
the
venue
work
with
the
AV
folks
and
all
the
other
contractors.
So
thanks
so
much
come
out
here:
Catherine
Karen
in
and
open
a
station.
Okay
Catherine
did
the
vast
majority
of
work.
B
Thank
you
very
much,
they're
behind
the
booth
giving
out
the
t-shirts,
speaking
of
which,
if
you
did
not
get
your
badge
and
t-shirt
during
the
first
break,
go
out
there
they're
right
near
the
entrance
and
grab
grab
one
of
those
t-shirts
and
then.
Lastly,
thanks
to
our
volunteer
Julia
Cisneros
she's
out
there
Manning
the
table
right
now,
but
thank
you
very
much.
Julia.
B
Okay,
so
today
we've
mixed
up
the
agenda
a
little
bit
from
last
year's,
so
last
year
we
tried
to
do
18
talks
in
six
hours.
If
any
of
you
were
here
last
year,
you
might
be
aware
that
it
was
not
exactly
six
hours
so
this
year,
I
have
regained
a
little
bit
of
my
sanity
and
we
are
planning
to
do
a
little
bit
fewer
talks
in
a
little
bit
less
time,
leaving
more
time
for
breaks
for
folks
to
chat
with
the
speakers
and
with
each
other.
B
You
know
in
the
hallway,
between
the
talks
I
mean,
hopefully
we'll
be
able
to
stay
on
schedule
and
some
of
those
great
talks
that
we
didn't
have
time
for
today
we're
going
to
have
a
few
more
talks
tomorrow
at
the
beginning
of
the
hackathon
day,
so
those
are
going
to
be
updates
on
subjects
that
we've
covered
previously
at
the
conference
and
I've
asked
the
speakers
to
very.
Very
very
briefly
talk
about.
You
know
what
their,
what
their
project
was.
So
you'll
get
a
little
taste
of
things
that
we've
talked
at
previous
conferences.
B
B
However
small
and
kick
off
some
projects,
actually
a
bunch
of
the
stuff
we're
going
to
talk
about
later
today
are
things
that
were
first
conceived
at
previous
hackathons,
and
if
you
don't
know,
if
you
don't
have
any
idea
about
what
you
want
to
work
on,
don't
worry.
We
have
a
huge
list
of
ideas
there
up
on
the
website
right
now,
so
we'll
definitely
see
those
and
talk
about
how
you
can
go
about
doing
those,
even
if
you
have
very
little
experience
developing
ZFS
cool,
so
I
have
a
cool.
B
A
few
announcements
to
make
first
off
point
of
open
ZFS
has
always
been
to
make
it
a
multilateral
collaboration
among
a
bunch
of
different
companies.
You
in
organizations
working
together
so
now
we're,
but
we
haven't
had
any
real
formal
structure
around
that.
So
now
we're
formalizing
that
a
little
bit
by
I'm
having
the
open
ZFS
project
join
software
in
the
public
interest
as
an
Associated
project.
So
SPI
is
a
non-profit
foundation
and
they
hold
assets
like
leftover
money.
B
So
we'd
like
to
encourage
all
of
you
guys
here
to
join
SPI
as
contributing
members
as
well,
the
members
of
SPI
they
vote
in
the
board
elections
and
they
help
SPI
maintain
its
501c3
nonprofit
tax
status.
Okay.
So
hopefully
that
was
a
very
boring
announcement,
but
it
says
open,
ZFS
up
for
long-term
success
and
it's
actually
enabled
a
much
more
exciting
announcement
to
talk
about
that.
I
need
to
roll
the
clock
back
to
the
beginning
of
open
ZFS
about
three
years
ago,
like
any
good,
open
source
project.
B
The
very
first
thing
that
we
wanted
to
do
was
create
a
website,
and
we
got
this
great
domain
name
open,
ZFS
org.
It
was
obviously
our
very
first
choice
of
domain
names.
I,
don't
know!
Maybe
you
pronounce
it
like
open,
ZFS,
org
or
open
is
the
FS
network.
Whatever,
however,
you
call
it,
it
was
definitely
the
best
domain
name
we
could
possibly
have
hoped
for
so,
but
today
we're
now
seeing
the
new
home
of
open
ZFS,
which
is
open,
ZFS
org.
B
So
thanks
to
Bill
McGonagall
for
donating
these
domain
names
and
the
domain
names
will
be
held
by
s
by
SPI,
the
nonprofit
foundation,
which
is
one
of
the
reasons
we
were
able
to
secure
this
donation,
so
bill
wasn't
able
to
attend,
but
if
you're
watching
bill.
Thank
you
very
much
for
donating
these
domain
names.
B
B
Cool,
so
you
know
most
of
the
conference
we're
talking
we're
looking
forward
to
projects
that
are
in
progress
that
ideas
that
are
being
generated
of
what
we're
going
to
see
in
the
future.
I
wanted
to
take
a
moment
to
reflect
on
the
past,
so
I
see
that
the
strength
of
ZFS
is
really
in
its
versatility.
You
know
when
we
designs
the
FS
back
in
the
early
2000s.
We
couldn't
have
imagined
a
lot
of
the
use
cases
that
people
are
building
their
companies
on
to
use
the
FS.
B
Today
you
know
remember
back
in
2000,
2005
cloud
computing
hadn't
even
been
invented.
Yet
Netflix
was
a
DVD
mailing
company,
and
you
know
streaming.
Video
hadn't
also
was
completely
infeasible
out
of
the
time,
but
you
know
the
foundation
of
ZFS
as
a
very
flexible
storage
system
that
can
be
used
for
all
these
different
uses
has
enabled
enabled
those
kinds
of
things
to
use
the
ZFS
today
in
you
know,
each
of
each
of
these
products
that's
built
on
ZFS.
B
These
are
all
things
that
have
shift,
and
you
know
you
look
at
this
list.
A
lot
of
this
is
performance
work,
a
lot
of
stuff
that
we
do
is
DFS.
It
is
improving
performance
and
oftentimes
in
your
performance.
Optimization
involves
making
trade-offs.
You
know,
like
CPU
versus
memory
or
the
the
performance
of
reading
data,
the
performance
of
writing
out
data
versus
reading.
Reading
that
same
data
in
later
on
now
you
know
when
people
are
thinking
about
performance
most
the
time
they're
thinking
about
like
I'm,
trying
to
improve
performance
in
this
particular
use
case
right.
B
Usually
it's
the
one,
that's
paying
the
bills,
so
one
of
the
ways
that
we
maintain
this
versatility
of
ZFS
is
by
making
sure
that
any
performance
enhancements
don't
like
inadvertently,
degrade
performance
of
other
use.
Cases
of
ZFS,
specifically
like
we
don't
want
to
degrade
performance
of
any
other
important
workloads
by
too
much,
and
this
is
a
really
vague
and
I
wish.
You
watch
the
area.
There's
a
lot
of
judgment
involved
in
this.
So
I
wanted
to
give
you
a
couple
of
examples.
B
So
let's
take
a
look
at
the
three
on
compression
and
hits
that
we've
done
in
open
ZFS.
So,
quite
a
while
ago,
we
added
a
new
compression
algorithm
LZ
for
LZ.
4
is
basically
better
in
every
way
than
the
previous
default
compression
algorithm.
It's
compresses
faster
and
it
compresses
to
a
smaller
size
and
it
decompresses
much
much
faster.
So
we
made
it
the
new
default
compression
algorithm.
We
said:
there's
really
no
downside
to
switching
to
this
as
the
default,
but
you
know
even
then
we
only
use
LZ
for
when
you
ask
for
compression.
B
If
you,
if
you
just
use
the
efest,
you
don't
change
any
settings,
we
don't
enable
compression,
and
the
reason
is
that,
even
though
LZ
4
is
really
really
fast,
there's
still
some
CPU
cost
and
there's
definitely
some
important
workloads,
we're
doing
that
attempted
compression
and
getting
no
benefit
from
it
like,
for
example,
storing
already
compressed
video
files.
That's
a
performance
hit.
B
So
more
recently,
we
added
in
what
we
call
embedded
block
pointers.
So
the
idea
here
is
that
if
you
have
data,
that's
very,
very
compressible
that
can
compress
down
to
fit
into
like
a
hundred,
bytes
or
less,
then
we'll
embed
it
in
embed
the
compressed
data
in
the
block
pointer
itself,
and
this
eliminates
even
this,
the
need
for
the
minimum
size
I/o.
That
would
otherwise
be
needed
to
write
it.
B
So
we
did
this
with
a
really
specific,
pretty
narrow
use
case
in
mind
of
initializing
of
writing,
initialized,
but
otherwise
unused
blocks
of
databases,
and
it
worked
really
great
for
that,
especially
because
database
administrators
liked
benchmark
their
storage
systems
by
initializing
database
files.
So
this
works
really
well
for
it
for
that
use
case
as
well,
but
we
discovered
that
it's
actually
on
a
pretty
big
win
for
a
whole
bunch
of
different
kinds
of
metadata
inside
of
ZFS.
So,
for
example,
think
of
like
directories.
B
Maybe
your
home
directory
has
like
hundreds
or
you
know
thousands
of
files
in
it.
If
you're
like
me,
but
most
of
the
directories
in
your
file
system
probably
have
only
a
handful
of
files.
Think
of
your
like
SSH
or
like
dot
git
config
directories.
They
have
only
a
couple
files
in
them.
Most
directories
are
like
that
and
those
can
be
stored,
very
very
compactly,
which
has
a
huge
performance
benefit
for
things,
like
anything
that
needs
to
traverse
the
whole
directory
structure
so
like
find
or
doing
backups
things
like
that.
B
So
this
is
on
by
default.
It's
just
used
by
everything.
It's
a
straight-up
performance
win
in
an
unusually
large
or
unexpectedly
large
amount
of
use
cases,
but
looking
at
the
more
recent
path
past
with
compressed
arc,
which
just
integrated
very
recently
with
compressed
dark,
we're
now
storing
data
in
memory.
Caching,
it
how
if
it's
compressed
on
disk,
we're
storing
it
compressed
in
memory.
So
this
is
a
really
huge
win
being
able
to
store
like
two
to
three
times
as
much
data
in
memory
as
we
previously
could
in
getting
much
much
higher
cache
hit
rates.
B
Because
of
that.
But
this
involves
a
pretty
significant
trade-off
on
we're
pretty
directly
trading
off
between
the
as
we
use
the
compressed
arc
more.
We
as
we
use
the
compressed
dark
more.
We
need
to
do
less
I/o,
so
that's
the
savings,
but
we
need
to
spend
more
CPU
to
decompress
that
data
for
every
cache
hit
that
we
make
so
like
I
said
for
most
of
our
codes.
B
This
is
a
pretty
big
win,
because
you're
able
to
cache
so
much
more
date,
more
data
and
the
cost
of
doing
an
I/o
is
so
much
more
expensive
than
the
cost
of
decompressing
the
straight
from
memory,
but
there's
certainly
workloads
where
I'm
doing
that
decompression
is
a
significant
CPU
costs.
So
we
went
the
extra
mile
to
make
sure
that
when,
if
you,
if
you
know
that
for
your
workload,
you
really
need
to
use
all
the
CPU.
B
For
other
things,
rather
than
decompressing,
what's
in
memory,
you
can
turn
off
the
compressed
dark
and
you'll
get
the
same
performance
that
you
were
getting
previously.
So
this
is
a
case
where
there's
a
more
significant
performance
trade-off-
and
you
know
we
decided
to
the
default-
was
to
move
us
forward
and
use
a
more
CPU
and
save
more
I/o.
But
I
we
wanted
to
maintain
the
previous
behavior,
so
ZFS
could
still
be
used
in
use
cases
where
that
are
very
CPU
constrained
and
they
can't
afford
that
CPU.
B
B
So
I
took
a
look
at
I
when
I
took
a
look
at
all
of
the
code
in
ZFS,
so
if
you're
using
ZFS
today,
half
of
all
almost
half
of
all
of
the
code
that
you're
using
was
written
in
the
past
three
years
by
folks
in
the
open-source
community.
These
are
the
folks
that
are,
you
know
mostly
sitting
in
this
room.
B
Today,
some
of
them
are
sitting
in
this
room
today,
a
lot
of
them
are
behind
the
scenes,
fixing
bugs
improving
performance,
adding
small
new
features
that
enhance
usability,
so
I'd
really
like
to
get
in
the
talk
by
giving
a
round
of
applause
to
the
folks
who
have
put
in
this
hard
work
over
the
past
five
years.
To
make
this
happen.