►
From YouTube: Raphael Carvalho - ZFS on OSv - OpenZFS Dev Summit 2014
Description
Platform Overview: ZFS on OSv (Raphael Carvalho from Cloudius)
A
A
A
B
A
So
we
can
cut
layers
of
protection
extraction
where
you
have
a
single
space
and
one
few
components
that
we
reused
with
zero
class,
so
cfs
is
used
as
our
default
display
the
file
system.
So
there
has
a
unique
file
system
that
we
use
on
osb
for
the
purpose
of
writing
history
that
is
communicated
with
this.
B
A
Will
be
our
part
on
my
currency
so
journey
speaking
osb
found
in
the
guide
and
that
allowed
us
as
a
right
rule
to
create
their
file
system,
where
you
would
where
the
elixir
has
relied
on
the
on
the
zfps
device
entry
to
communication,
where
the
cfs
facebook
is
a
device
that
communicates
with
the
zero
pass.
The
zero
support.
So
you'll
have
the
zero
test
device
complete
with
the
retail
class,
and
you
have.
A
Workloads,
so
so
so
I
in
this
slide,
I
would
like
to
explain
why
we
decided
to
use
the
affairs-
and
that
should
be
very
clear,
so
osp
was
designed
with
effortless
managing
capability
in
mind
where
you
will
not
take
where
you
wouldn't
take
a
lot
of
effort
to
manage
the
operating
system.
So
you
will
just
launch
your
instance,
and
you
will
not
take
lots
of
time
to
finish
and,
for
example,
tuning
your
operating
system
out
the
layers
for
your
specific
workload.
A
So
that's
the
milestone,
so,
as
you
can
see,
we
created
osb
in
the
in
the
beginning
of
the
past
year
and
we
started
at
that
time.
We
were
using
grandfather,
which
is
a
file
system
that
just
entered
right
related
to
to
test,
so
it's
an
irrigated
file
system
and
so
as
we
need
just
basically
the
file
system.
So
yes,
we
use.
We
use
the
dell
it's
compatible
with
the
bsd
license
in.
B
A
So,
as
you
can
see,
we,
we
got
the
fast
amount
in
in
the
middle
of
that
year
of
the
past
year
and
two
months
later,
we
got
the
read
and
write.
Support
from
cfs
were
improved
with
university
and
that's
great
because
cassandra,
for
example,
sometimes
sometimes
need
a
additional
file
system
to
take
a
maximum
advantage.
A
So
you
can
with
cfs
tools
it's
it's
much
better
in
that
manner.
Where,
with
a
single
command,
you
can
create
a
new
file
system,
and
so
it's
much
better
in
that
manner
or
a
lot
of
workloads
for
cloud
deployment
for
storage
deployment.
Zia
has
much
better.
So
that's,
basically
why
we
decided
to
use
zfs
and
that's
the
milestone
or
the
gfs
receiver.
A
A
A
So
from
memory
pressure,
the
operating
system
will
go
to
the
artery
plane
say:
oh,
I
need
x
amount
of
memory
to
be
given
back
to
to
the
application,
so
it
would
go
to
our
plane,
which
was
integrated
into
our
reclaimed
infrastructure,
where.
B
A
So,
as
I
said
also
depending
on
code
information
which
we
get
from
the
previous
pistol,
we
have
support
a
lot
of
that
whole
wishing
shirt
freebsd
had
had
to
part
a
lot
of
solaris,
so
the
next
step
so
and
one
and
then
the
final
one
to
get.
There
has
actually
worked
where
you
could
see
some
of
the
results.
B
A
So
here
is
the
visual
biosystem
layer
and
I
am
going
to
explain
it
a
little
bit,
although
I
have
explained
essentially
in
the
previous
slide,
so
visual
persistent
layer
is
the
layer
from
the
opening
system
that
allows
many
file
systems
to
coexist
and
communicate.
For
example,
in
the
same
system
you
will
have
x34
and
zero
class
and
other
files
as
well.
There
are
choice,
so
the
video
file
system
layer
is
where
the
file
system
consumers
go
to.
A
I
want
to
write
to
to
search
a
file
from
zlfs
for
you,
so
you
have
the
file
system
space.
So
you
have
root
a
glass
shield,
the
name
of
the
file,
the
directory
and
so
on.
So
visual
file
system
layer
is
basically
layered
that
allow
audio
file
system
to
put
this,
where
you
can
call
a
file
from
one
file
system
channel.
So
so
that
is
what
you
have
to
work
on
because,
as
I
said,
our
vfs
layer
changed
our
videos.
Layer
was
best
because
we
had
to
change
it
as
well.
A
So
we
had
to
go
to
zpl
and
modify
it
to
work
with
our
vfs,
but
it
didn't
take
a
lot
of
change.
We
wanted
to
to
keep
the
zpl
as
it
was
originally
and
we
are
trying
to
to
get
missing
with
the
frequencies
so
that
it
would
be
easier
for
us
to
pull
change
on
the
furious
new
chains
from
open
circus
as
well.
A
So
let
me
go
to
the
next
slide.
Are
everything
integration
I
explained
as
well
so
the
orange
so
the
orange
speaker?
You
can
see
that
there
is
the
general
reclaim
reclaimer
from
the
os,
which
is
the
infrastructure
itself,
and
you
can
see
in
the
right
side
that
if
they
are
reclaimer
and
right
below
the
other
things
that
I
also
said
so
that's
the
change
we
have
done
with
with
within
our
disease
version.
So
they
have
has
to
this
layer.
A
So
we
have
changed
some
of
the
our
reclaimed
qualities
as
well
to
to
inform
the
opening
system
how
much
memory
was
reclaimed,
so
the
korean
virtual
article
playing
function
doesn't
provide
how
much
so
that
was
actually
made
specifically
for
our
for
our
reclaimed
infrastructure,
our
reclaimed
cause
and
one
change
that
we
have
done.
That
is
very
interesting
and
we
would
like
to
share
it
with
the
other
platforms
as
well
is
to
is
something
called
art
builder
sharing.
A
That's
when
youtube,
you
remember,
your
goals
are,
and
you
don't
need
to
duplicate
that
welfare,
so
you
can
get
that
art
builder
and
map
interact
into
the
process
at
your
space.
So
if
I
don't
mistake
this
correctly
in
custom
games,
you
will
get
our
art
builders
and
lubricated
and
magnetic
directory
space.
Where
you
have
it,
because
it's
not
well
integrated
into
the
page
cache
itself,
so
as
we
have
control
of
total
control
of
osb,
so
we
could
change
our
page
cache
totally
to
be
well
integrated
with
art.
A
A
A
Respect
to
the
range
that
we're
trying
to
map-
and
you
will
get
those
art
builders
and
you
will
in
charleston
you
will
get
to
the
team
you
broker.
You
will
get
the
artwork
out
of
tmu
bulford
and
we
integrated.
We
exported
a
osb
function
to
map
those
the
arc
builder
directly
to
the
address
views,
but
I
should
really
say
that
it's
limited
to
really
for
by
the
time
you
so
above
all
right.
You
will
end
up
with
the
top
right
and
that
really
on
the
book
pages.
A
A
The
operating
system
would
get
into
a
facial
and
we
will
have
to
locate
both
buffers
so
at
that
time,
home
will
not
be
mapping
those
art
builders
anymore.
So
we
have
a
specialized
white
page
patch.
That's
how
we
call
it
that
whole
they
called
the
cop
art
builders.
So
in
that,
from
that
time
home
we
would
be
exactly
like
the
customers
that
that
has
art
brokers
located
into
the
admin
space.
A
So
we
are
actually
we
actually
have
a
claim
for
the
future
where,
where
you
could
write
directly
to
the
art
builders
that
would
map
into
that
in
space,
but
that
would
take
a
lot
more
effort,
so
by
the
time
you
meet,
we
just
have
that
feature
for
weeks.
So
all
right,
as
I
mentioned
already,
you
will
do
a
couple
rides.
A
So
right
here
is
where
I'm
going
to
say.
The
plan
for
the
future,
where
I
already
mentioned
earlier,
is
to
get
conceived
with
the
other
platforms,
especially
previously,
that's
where
we
get
zfs
from
and
I
hopefully
we
will
share
our
modification
with
the
other
platforms
as
well,
of
course,
that
those
that
would
be
valuable
to
them.
A
B
A
If
that's
another
platform,
but
we
lose
the
call,
we
can't
accommodate
better
because
we
can
batch
rights,
for
example.
So
so
the
blockchain
device
driver
was
specifically
designed
for
ec2
or
or
amp
or
ec2
from
amazon
for
those
or
even
google
compute
engine
as
well.
So
our
operating
system
was
totally
molded
around
the
cloud
environment.
B
A
Yes,
osb
was
based,
it's
basically
a
way
of
doing
docker,
but
not
already
so
we
are
trying
to
achieve
the
same,
but
going
through
other
methods.
So
usb
is
basically
to
try
to
provide
something.
That's
well
isolated.
A
So
when
you
have
a
single
process,
you
don't
need
to
care
about
isolation,
or
even
that
you
have
a
simple
process.
So
when
the
scheduler
wakes
up,
you
don't
need
to
do
all
that
contact
streets
anymore
between
address
space,
given
that
address
space
is
supposed
to
hold
hold
the
super
process.
So
if
you,
of
course,
you
you
have
been
threatened
with
the
usb
so
but
as
the
press
share
the
center
space
in
a
context
which
you
will
have
to
go
out
without
that
change
anymore,
so
everything
is
so
much
simpler
and.