►
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
Thanks
so
I
think
we
have
people
here
from
with
a
wide
variety
of
experiences
with
with
open
ZFS
in
my
talk
is
going
to
kind
of
reflect
that
in
that
I'll
be
covering
a
huge
bunch
of
different
things,
from
high-level
overviews
to
a
deep
dive
into
ZFS
send
and
receive
so.
Can
I
get
a
show
of
hands
how
many
people
here
have
used?
Zfs,
awesome,
okay,
preaching
to
the
choir
here,
and
how
many
people
here
have
looked
at
ZFS
source
code,
Oh,
quite
a
few
actually
cool.
So
probably
the
beginning.
A
A
You
never
need
to
run
fsck
and
we
also
have
Ende
Ende
data
integrity
with
checksums
that
cover
everything
from
memory
to
the
disk
and
back
so
you
always
know
if
you're
giving
the
application
the
same
data
that
it
originally
wrote
and
I
think
probably
one
of
the
biggest
features
for
any
since
Edmonds
in
the
audience.
Here
is
just
the
simple
administrative
model.
You
know
it's
very
easy
to
create
a
storage
pool,
create
file
systems
and
concisely,
express
your
intent
about
how
to
manage
those
so
like
this
data
needs
to
be
compressed.
A
This
data
needs
to
be
faster
or
slower
things
like
that,
so
this
diagram
kind
of
shows
how
ZFS
and
all
those
features
that
I
mentioned
fit
into.
The
storage
stack
of
all
the
other
software.
That's
running
on
the
system
on
the
left
hand,
side
here,
I,
have
kind
of
a
traditional
file
system
volume
manager
stack,
so
operations
come
in
from
you
know:
NFS
local
file
access
through
the
through
the
virtual
file
system
layer.
They
talk
to
the
file
system
and
in
the
file
system,
thinks
it's
just
talking
to.
A
You
know
one
disk,
but
actually
it
might
be
talking
to
a
volume
manager.
So
it's
saying
you
know:
I
have
this
disk
Kwok
what
disk
and
I'm
going
to
read
or
write
blocks
at
certain
offsets
within
the
desk.
Then
the
volume
manager
takes
those
reads
and
writes
to
the
disk
and
actually
splits
them
out
over
multiple
disks.
So
a
lot
of
information
is
actually
getting
lost
here
in
this
boundary
between
the
file
system
and
the
volume
manager,
because
it's
just
a
plain
disk
interface.
A
Similarly,
you
know
you
get
these
islands
of
storage,
where
you
have
one
file
system
sitting
on.
You
know
two
disks
and
other
file
system
studying
on
another
four
disks
and
the
the
space
usage
is
isolated
as
well
as
the
performance
is
isolated,
which
sometimes
could
be
a
good
thing,
but
oftentimes.
It
just
means
that
you're
leaving
performance
on
the
table
versus
with
ZFS
we've
kind
of
rethought.
The
interfaces
between
the
file
system,
ish
parts
of
the
software
stack
and
the
volume
manager
ish
parts
of
the
software
stack.
A
So
we
now
have
essentially
three
layers
within
ZFS.
The
top
layer
is
the
POSIX
layer
or
the
volume
manager
layer.
This
basically
deals
with
like
owners,
permissions
file,
lengths
things
like
that
and
then
on
the
volume
layer,
interfacing
with
a
scuzzy
target,
and
which
is,
you
know
much
simpler
because
you
don't
have
like
ownership
and
things
like
that,
and
then
the
data
management
you
it
deals
with
like
okay,
we're
writing
some
data.
So
we
need
to.
A
We
need
to
figure
out
like
how
this
tree
of
blocks
is
interrelated.
So
this
is
kind
of
the
core
file
system
stuff,
like
I'm,
writing,
I'm,
extending
a
file
and
so
I
need
to
add,
like
a
layer
of
indirection.
In
my
tree
of
blocks,
all
that
stuff
is
managed
here,
but
this
layer
doesn't
have
to
think
at
all
about
allocating
space,
which
is
normally
a
filesystem
kind
of
thing.
Instead,
we've
moved
that
down
to
the
storage
pool
allocator
layer,
so
this
layer
deals
with
both
block
allocation
and
freeing
as
well
as
block
integrity.
A
So
it
deals
with
like
mirroring
raid
Z
checksums
compression
everything.
That's
like
a
physical
aspect
of
the
data
versus
this
layer
doesn't
have
to
think
about
that
at
all.
It
just
thinks
like
I,
have
this
logical
piece
of
data
like
I'd
like
to
write
that
somewhere
to
the
disk
and
I
need
to
be
able
to
get
it
back
reliably
at
some
point
later
in
time,
so
it
sends
you
know,
maybe
I've
like
a
128k
chunk.
A
It
sends
that
128k
down
to
the
spa,
the
spa,
maybe
compresses
it
allocates
space
for
it
computes
the
parody
of
it,
writes
it
to
a
bunch
of
disks
in
a
raid
Z
group
and
gives
the
DM
you
just
the
block
pointer.
That
tells
it
how
to
retrieve
that
data
later
cool.
Any
questions
about
this
probably
overview
for
most
of
you
who
have
looked
at
the
source
code.
A
We
worked
on
the
project
for
about
four
or
five
years
before
releasing
the
source
code
in
2005,
so
we're
almost
to
the
10
year
anniversary
of
releasing
the
ZFS
source
code,
and
when
did
you
join,
you
joined
like
it
yeah
so
George
who'll
be
talking
later.
My
colleague,
he
started
in
2005,
also
working
on
ZFS
in
the
release
of
the
source
code.
You
know
enabled
the
fs
to
be
ported
to
a
bunch
of
different
platforms,
starting
with
FreeBSD
in
2008.
It
was
released
as
part
of
freebsd
7.
A
So
a
couple
years
later,
in
2010
we
had
this
kind
of
big
event
where
oracle
acquired
sun
microsystems
and
nobody
knew
like
what
was
going
to
happen
with
open,
solaris
or
ZFS
or
any
of
that
stuff.
So
Oracle
kind
of
compounded
those
fears
by
stopping
their
contributions
to
open
source
ZFS,
but
in
response
on
the
the
community
kind
of
came
together
and
formed
the
illumos
project,
which
is
basically
like
a
multilateral
collaboration
of
companies
that
are
all
using
what
once
was
open.
A
Solaris
I
mean
continuing
that
source
code
for
words,
making
improvements
to
that
operating
system
and
I
think
this
is
the
difference
between
Lumos
and
open
solaris.
To
me
is
that
open
solaris
was
a
sun
microsystems
project
right
and
basically
all
of
the
source
code,
even
though
it
was
not,
it
was
open
source,
but
all
the
changes
were
taking
place
that
son.
A
A
You
know
in
part
due
to
the
Oracle
acquisition
that
knowledge
has
become
more
widely
dispersed
in
the
industry,
and
you
know
so
there's
actually
several
several
companies
making
changes
in
Lumos,
obviously,
dell
fix
where
George
and
I
work
is
one
of
the
most
prominent
ones.
But
but
you
know,
there's
there's
quite
a
few
other
companies
that
are
also
working
on
in
the
most
and
now
that
we
have
all
scooped.
Now
that
we
have
created
opens
DFS,
you
know
open
ZFS,
the
idea.
A
Well,
that's
my
next
slide
I'll
get
to
the
idea
that
on
the
next
slide,
but
you
know
open
ZFS
has
really
broadened
that
umbrella
of
people
working
on
ZFS,
from
not
just
Lulu
most
and
FreeBSD
and
Linux
being
different
islands
of
people
working
on
it
to
open
ZFS
is
about
kind
of
forming
at
an
umbrella
organization.
Over
all
these
platforms,
making
sure
that
people
working
on
ZFS
on
any
platform
are
talking
to
one
another,
making
sure
that
CFS
is
a
great
storage
platform,
regardless
of
the
operating
system
that
you're
using
it
on.
A
So
this
kind
of
goes
over
what
I
just
said.
You
know
our
goals
are
also
so
that
I
talked
about
kind
of
open
communication
and
consistent
experience
using
ZFS.
Also
one
of
the
goals
is
to
raise
awareness.
So
that's
one
of
the
reasons
that
George
and
I
are
here
today,
and
you
know
why
you
we
have
a
great
data
thanks
to
ryan
and
the
other
organizers
for
creating
this
conference.
Is
you
know
we
want
to
raise
awareness?
A
We
want
to
make
sure
that
people
know
that
you
know
ZFS
isn't
just
some
proprietor
or
Oracle
technology
that
died
one
day
when
they
closed
sourced
it
we're
still
there's
lots
of
people
creating
products
based
on
open,
ZFS.
I
know
a
lot
of
you
in
the
room
are
doing
this
and
engaged
in
this,
but
we
want
to
get
the
word
out
to
everyone
that,
like
open,
ZFS
ZFS
is
great.
It's
open
source,
it's
available
on
whatever
platform
you're
already
using.
A
A
The
mailing
list
is
primarily
a
developer
discretion
mailing
list,
if
you
are
working
with
the
ZFS
source
code,
I
definitely
recommend
that
you
get
on
this
mailing
list,
in
addition
to
any
platform
of
specific
mailing
lists
that
you're
on,
because
this
will
get
you
informed
about
kind
of
the
latest
things
that
people
are
working
on,
maybe
on
a
different
platform
than
what
you're,
using,
as
well
as
announcing
events
like
this
conference,
okay,
and
we
also
have
other
periodic
events-
the
office
hours.
This
is
like
a
live.
A
Video
chat
with
the
open,
ZFS
developer,
roughly
once
a
quarter
or
not
for
a
year,
if,
depending
on
how
how
well
I
am
at
organizing
it
so
the
open
source
gfs
has
enabled
all
these
platforms
to
have
a
great
storage
subsystem
minutes
enabled
a
lot
of
companies
to
create
products
based
on
ZFS
that
wouldn't
have
been
possible.
You
know
in
the
kind
of
closed
ecosystem.
A
So
you
know,
that's
all
great
and
you
know
ZFS
obviously
came
from
a
great
foundation,
but
you
know
what
have
we
done?
What
have
you
done
for
me
lately
in
the
past
12
months,
we've
in
to
being
the
collective
we,
both
dell
flicks
and
other
contributors?
Could
you
have
done
a
huge
variety
of
projects?
I
will
say
them.
The
vast
majority
of
these
are
really
centered
around
performance.
I
would
I
would
categorize
it
that
way.
A
You
know,
including
even
things
that
kind
of
look
like
features
like
you
know,
larger
block
size
or
you
know,
new
new
settings
these.
These
are
pretty
much
in
service
of
performance,
so
I
think
one
of
the
big
themes
within
ZFS
is
like
the
feature
set
is
pretty
good
and
one
of
the
key
things
about
CFS
is
making
all
the
features
work
with
all
the
other
features
which
makes
it
very
flexible
and
able
to
you
know,
accomplish
a
lot
of
different
things
depending
on
how
you
want
to
use
it.
A
We
also
have
a
chromium
performance
work,
Georgia's
going
to
be
talking
about
compressed
arc
and
allocation
throttle,
and
all
these
things
are
actually
we
think
most
most
of
these
are
Dell
fix
projects
and
we've
actually
already
implemented
these,
but
they
are
not
yet
upstream
in
illumise.
So
that's
something
that
we're
working
on
now
right
now
and
then
there's
also
a
few
projects,
including
device
removal
that
are
works
in
progress
that
we
hope
to
complete
an
upstream
soon
very
soon,
as
well
as
other
cool
things
like
persistent
I'll
to
work.
A
/
shows
here
today,
if
you
have
questions
about
it
like
him
in
during
the
break
we've.
Also,
as
I
mentioned,
you
know,
evangelism
making
sure
to
get
the
word
out.
These
are
we've
gone
to
about
six
different
conferences
just
in
the
past
year
and
today
we're
announcing
the
upcoming
third
annual
10th
anniversary,
open,
ZFS,
developer
summit.
This
will
be
in
San,
Francisco,
October,
nineteenth
and
twentieth
it'll
be
similar
to
this
conference,
one
day
of
talks
and
then
one
day
hackathon.
A
A
The
new
edition
has
a
great
chapter
on
ZFS.
It's
a
really
good
overview
of
the
internals
of
ZFS.
A
little
bit
about
data
structure
is
a
little
about
a
little
bit
about.
You
know
why
we
design
things
a
certain
way
and
what
trade-offs
we
made.
There's
also
an
upcoming
book
I
mean.
Actually
I
will
show
you
if
I
can
find
it.
I
know,
there's
a
there's
a
tab
here
somewhere
with
it.
Oh
here
we
go
so
there's
an
upcoming
book
by
Michael
Lewis
and
Alan
Jude
called
freebsd
mastery.
A
This
is
primarily
about
using
ZFS
and
kind
of
what
the
different
features
are
and
how
how
to
use
them,
how
to
compose
them
together
to
solve
your
problems
more
so
than
a
technical
implementation,
but
I
think
this
is
also
going
to
be
a
really
good
book
that
I,
unfortunately,
have
not
yet
had
time
to
review.
So
I'm
I'm
very
hopeful,
but
I
hope.
A
C
A
So
in
terms
of
the
different
platforms,
I
would
say
the
difference
between
illumise
and
freebsd
is
very
minimal.
I'd,
say
they're,
actually
kind
of
the
closest
together
in
terms
of
the
source
space.
The
feature
set
performance,
you
know,
and
then
your
linux
is,
is
derived
from
those
there's
more
changes
than
Linux,
both
in
terms
of
their
a
little
bit
further
behind
in
terms
of
syncing
up
with
changes
that
are
coming
from
a
limos
and
they've
done
a
bit
more
work
to
customize
the
fs
for
their
platform.