►
From YouTube: CNCF SIG Storage Meeting 2020-08-12
Description
CNCF SIG Storage Meeting 2020-08-12
A
B
E
F
B
C
B
A
Thank
you.
I
appreciate
that
alex
yeah.
So
basically
it's
it's.
As
alex
mentioned,
I'm
I'm
running
meetups
that
are
and
forming
a
bit
of
a
community
around
data
on
kubernetes
and
trying
to
just
have
it
be
a
place
where
people
can
share
their
experiences
and
learn
from
each
other,
and
so
I
I'm
going
to
share
my
screen
real
quick.
I
mean
I
have
four
slides,
so
it's
not
going
to
be
death
by
powerpoint
by
any
means,
but
yeah.
I
guess
I
I
got
intro
from
alex
and
I'm
demetrius.
A
I've
been
primarily
playing
in
the
mlop
space,
but
kubernetes
was
something
that
was
very
very
on
the
top
of
everyone's
mind,
especially
when
you
talk
about
kubeflow
and-
and
it
was
like
hey,
there
needs
to
be
something
where
we
can.
We
can
all
come
together
and
talk
about
data
when
it
comes
to
kubernetes
and
potentially
evolve
mature
the
space
a
little
bit.
A
A
So
it's
a
light
lift
or
it
can
be
more
of
like
a
talk
where
you
share,
slides
and
all
that.
So
it's
it's
up
to
you
really
but
yeah.
Like
I
mentioned,
we
have
this
slack
workspace
and
I
can
share
these
links
in
the
chat
right
now
in
case
anyone
wants
to
join
and
then
give
their
their
two
cents
and
then
the
last
thing
is
like.
I
would
really
love
to
see
it
become
something
where
people
can
bring
their
own
initiatives
start
their
own.
A
Excuse
me
start
their
own
series
around
whatever
it
is
that
you
think
is
is
interesting.
Maybe
you
know
you
say
well
hey.
I
really
want
to
talk
about
security,
or
I
want
to
talk
about
something
that
is
is
top
of
mind
for
you
and
we
can
have
a
series
or
an
initiative
around
that.
So
that's
it
that's
my
my
pitch
as
we
could
say,
and
I
hope
to
see
see
all
you
there.
G
A
Oh,
that's
that's
great
to
know
for
sure
I
will
I
will
reach
out
to
them
also,
but
yeah.
It's
not
really
so.
The
interesting
thing
about
this
is
it's
not
really
working
towards
anything
specific
like
it's
not
like
we're
getting
together
and
we
are
creating
projects
around
working
like
working
groups,
we're
not
doing
any
open
source
or
anything
it's
more.
Just
like
you
were
mentioning
like
yeah
user
stories
or
best
practices,
things
that
you
like.
D
D
We
were
trying
to
create
a
catalog
essentially
of
of
common
use
cases
that
we
could
then
document
and
say
this
is
a
good
way
of
doing
this
thing
and
if
you're
doing
something
like
this,
like
here's,
a
kind
of
a
recipe
that
we
know
works,
and
I
don't
want
to
paraphras
or
put
words
in
your
mouth
louise.
Maybe
you
want
to
talk
about
that,
but
it
might
be
interesting
to
explore
the
overlap
between
these
between
your
meetups
and
this
initiative,
because
it
seems
like
there's
a
lot
of
common
kind
of
goals
there.
D
A
Yes,
yes,
that's
that's
exactly
right,
like
that
whole
idea
of
hey,
let's
get
together.
Let's
share
this
knowledge
because
it's
untapped
and
we
need
to
be
talking
about
it
or
writing
it.
I
mean
I
did
have
the
idea
of
after
the
meetups
taking
what
I'm
doing
is
I'm
taking
some
of
these
gems
that
people
talk
about
and
I'm
writing
them
down
and
putting
them
onto
medium
and
so
I'll
connect
with
you
offline,
at
least
to
see
about.
A
E
A
A
F
A
D
D
Historically,
this
group
is
focused
a
lot
on
the
actual
infrastructure
at
the
the
you
know,
layers
of
the
storage
stack
and
and
all
the
various
open
source
projects,
and
we
tried
to
get
this
sort
of
use,
cases,
analysis
and
publication
going,
and
I
think
it's
it's
sort
of
been
tricky
I'll
put
it
that
way,
but
I
I
would
love
to
see
an
ongoing
collaboration
between
the
meetups
and
and
the
sig.
You
know.
D
One
model
that
comes
to
mind
is
maybe
you
kind
of
give
come
and
give
us
a
10,
15
minutes
sort
of
summary
per
month
or
something
like
that.
It
says:
here's,
the
really
cool
stuff
we
did
this
month.
Here's
you
know
stuff.
You
should
probably
go
and
check
out
here's
a
podcast,
here's
this
that's
the
next
thing.
I
think
that
would
be
super
useful
to
this
group
and
then
I
don't
know
how
you
know.
Maybe
the
reverse
might
also
make
sense.
A
B
B
So
so
everyone-
I
I
just
wanted
to
to
do
a
a
a
quick
sync
up
on
where
we
are
on
the
the
performance
white
paper
that
that
we've
been
working
on.
So
I
wanted
to
I'll
just
share
my
screen
just
so
that
we
can
go
through
it
briefly.
Can
you
see
this?
Okay,
yes,
cool
all
right,
so
so
what
I've
done
is
I've.
I've
simplified
some
of
the
parts
because
we
we
have.
B
B
So
I've
started
off
with
a
simple
introduction
that
basically
says
always
test
your
own
application
in
your
own
environment.
Don't
rely
on
published
results
from
vendors,
and
also
I
put
in
a
link
to
the
white
paper
so
that
people
can
understand
you
know
the
the
the
general
attributes
and
terminologies
of
of
the
of
the
storage
environment
before
they
before
they
start
reading.
B
Things
like
the
effects
of
of
random
rights
and
right
amplification,
the
effects
of
encryption
on
benchmarking,
the
effects
of
the
topology
and
notes
to
talk
about
the
the
challenges
of
the
on
the
client
side
of
testing
the
performance.
Because,
because
you
know
very
often,
the
clients
can
be
can
be
the
the
bottleneck
in
in
these
environments.
B
Things
like
you
know,
the
the
performance
implications
of
you
know,
say
and
replication,
and
perhaps
data
protection,
and
also
you
know
some
of
the
things
to
keep
in
mind
in
terms
of
say
things
that
happen
in
the
background
in
in
in
storage
systems.
So
you
know
some
storage
systems
might
have
out
of
bands
compression
or
garbage
collection
or
or
something
like
that
that
can
affect
performance
and
then
a
little
bit
about
benchmarking
tools.
B
So
we
don't
have
we
don't
have
we
don't
have
a
good
information
on
how
to
run
the
benchmarks,
but
at
this
stage
I
am
kind
of
thinking
that,
rather
than
hold
the
document
for
any
longer,
there's
still
value
to
publish
the
document
with
with
this
information,
as
is
with
some
basic,
very
basic
pointers.
To
to
the
benchmarks
that
that
people
can
run,
and
then
we
can
iterate
and
maybe
provide
the
version
to
when
we're
ready
to
to.
B
C
B
C
It
will
mean
something,
so
basically
you
get
bare
metal
instances,
but
they
are
in
the
cloud.
So
we
what
we
are
doing
with
that
in
with
this
is
we
have
started
running
nightly
benchmarks,
because
community
has
been
asking
like
how
do
we
like?
How
do
you
make
sure
that
you
don't
introduce
performance,
regressions
and
stuff
like
that,
so
we
just
started
doing
that
just
like
a
few
days
ago,
we
just
got
the
environment
up
and
running,
so
we
run
a.
C
We
have
a
very
basic
system
stuff
that
we
that
runs
every
every
night
and
then
reports
on
a
slack
channel.
How
many?
What
was
the
qps?
What
was
the
latency
and
that
kind
of
stuff?
I
I
don't
know
if
it
is
completely
related,
but
this
is
just
like
a
very
basic
startup
kind
of
thing.
We
hope
to
build
more
features
around
it
and
stuff,
but
I
don't
know
if
that
can
be
used
as
an
example
for
how
to
set
up
a
benchmark
and
run
it.
But
this.
A
B
B
How
long
do
we
kind
of
delay
putting
the
document
out
to
wait
for
somebody
to
write
some
of
this
in
a
in
a
way
that
end
users
can
actually
consume,
because
I
I
I
I
I
think
we
might
sort
of
be
getting
trapped
in
the
in
the
sort
of
idea
generation
where,
where
you
know
there
are
lots
of
possible
options
for
for
putting
for
for
putting
content
into
the
document.
But
I
want
to
try
and
balance
that,
with
with
timeliness.
C
D
Right
like
I
would
suggest,
I
think
that
the
first
part
of
the
or
the
document
up
until
the
the
tools
section
right
at
the
end
is
looks
fantastic.
To
be
honest,
I
don't
think
you
need
to
you.
You
might
be
kind
of
focusing
on
what's
missing
and
what
took
so
long,
but
but
I
mean
the
document
that
you
went
through,
looks
fantastic
and
I
would
publish
it
as
is.
I
would
actually
make
the
following
suggestion
to
to
remove
the
last
few
sections
because
they're
incomplete
add
a
section
of
links
to
other
information.
D
A
D
One
sentence,
and
then
people
know
that
that
there's
a
volume
two
coming
and
then
hopefully
we
can
actually
kind
of
set
a
timeline
and
and
and
have
a
target
date
for
that
by
you
know
whatever
it
is.
It
doesn't
sound
like
a
huge
job,
but
you
know
having
said
that,
it's
it's
a
year
in
the
making-
and
I
guess
it's
it's
proven
tough.
So
so
maybe
it's
more
difficult
than
it
sounds,
but
but
yeah
it
would
be
nice
not
to
kind
of
wait
another
year
for
that
hopefully
and
yeah.
C
C
C
That
is
something
that
they've
asked
for
the
other
stuff
they've
asked
for
they
asked
for
as
a
as
a
problem
in
with
us,
but
I
think
it
applies
to
all
database
instances,
which
is
version
compatibility.
I
mean
these
go
into
functional
areas,
but
these
are
like
things
that
people
are
really
really
concerned
about
when,
when
adapt,
when
basically
adopting
a
software
project.
B
B
D
B
Yeah
in
in
the
performance
document,
we
we
specifically
sort
of
highlight
that
you
know
things
like
the
way
your
data
is
protected
and
your
availability
and
and
consistency,
for
example,
are
all
big
factors
in
in
how
your
performance
attributes
work.
So
so
we
kind
of
point
out
those
things,
and
I
think
that's
actually
that
that's
what
the
overlap
is
covered
quite
well
in
the
original
landscape
documents.
D
Point
I
think
that
the
original
document
did
deal
with
it
well
and-
and
maybe
we
just
dive
into
each
area
relatively
in
isolation
and
just
refer
back
to
the
document-
remind
people
that
these
things
all
influence
each
other.
But
this
particular
document
is
squarely
about
performance
or
durability
or
reliability
or
whatever
it
happens.
To
be
sounds
like
a
great
idea.
B
B
B
B
No
so
so
tikv
had
has
a
dependency
on
something
called
a
placement
driver,
the
repo
is
called
pd
and
that
repo
is
also
used
by
tidb,
which
is
obviously
you
know,
a
separate
project,
and
so
it
hadn't
been
kind
of
bundled
into
the
into
the
original
tikv
due
diligence.
So
the
the
the
project
team
have
sort
of
said,
look
in
order
to
kind
of
remove
any
of
these
concerns.
B
B
So
so,
hopefully
it's
not
a
big
deal.
The
the
the
other
item.
I
I
wanted
to
to
quickly
discuss,
and
you
know,
hopefully
this
doesn't
won't.
Take
too
much
time.
Derek
moore
has
who
had
previously
presented
the
proviga
project
is
looking
is
looking
to
to
move
forward
with
the
due
diligence
to
proceed
with
an
incubation
submission.
So
just
as
a
sort
of
mental
refresher
praviga
is
a
it's
a
project
which
is
currently
sponsored
by
by
dell.
They
have
it's
it's
a
it's
a
streaming
storage
product.
B
B
If
you
haven't
read
the
the
incubation
proposal,
you'll
be
it's
it's
it's
worth
reading
that,
because
there's
there's
actually
quite
a
lot
of
useful
detailing
there.
We
we
all
thought
it
was
a
great
project,
and
you
know
the
presentation
went
particularly
well
and
we've
we've
recommended
to
to
move
forward
to
the
to
the
dd
stage
with
the
toc.
B
I'm
waiting
for
one
of
the
tse
members
to
to
to
step
forward
to
do
the
to
work
with
us
on
doing
the
dvd,
but
but
I
guess
we
also
need
to
figure
out
who's
going
to
work
on
the
dd
from
from
our
end.
So
so
I'm
I'm
happy
to
to
to
help
out
myself,
but
I'm
I'm
quite
I'm
quite
time
limited
over
the
next
couple
of
weeks.
D
B
Yeah
exactly
you
know
that
said
like,
like,
I
said,
a
lot
of
the
a
lot
of
the
criteria
for
for
incubation
were
already
covered
in
a
fair
amount
of
detail
in
the
proposal.
So
so
I
don't
think
I
don't
think
there's
a
lot
of
work
to
to
sort
of
document
or
dig
into
into
the
detail
required.
But
you
know
we.
We
have
to
make
sure
that
we've
we've
covered
all
of
the
all
of
the
areas
based
on
the
guidelines.
Documents.
D
B
D
H
D
Okay,
yeah,
I
mean
we
can
probably
work
around.
You
know
within
reasonable
limits,
but
if
you
told
us
you
could
have
it
done,
you
know
at
the
end
of
october.
I
think
that
would
be
good
and
if
you
told
us
you
could
have
it
done
in
the
second
week
of
november.
I
think
that
would
be
better
than
not
having
it
done
at
all
for
sure.
B
I
B
I
B
Yes,
sorry,
that's
a
very
good
point.
I'm
I'm
sorry!
I
didn't
mean
to
mention
this
so
about
a
week
ago.
I
I
believe
open
ebs
project
that
is
currently
a
a
sandbox
project
has
put
in
a
a
proposal
to
to
move
forwards
to
incubation,
so
we're
we're
currently
looking
to
put
to
put
an
open
ebs
presentation
on
the
agenda
for
for
the
next
meeting
on
august
26th.
I
J
J
B
Data
lifecycle
project.
Yes,
so
if,
if
you're
making
a
a
sandbox
submission,
then
the
the
toc
have
a
new
process
where
they,
where
they
vote
on
sandbox
once
every
month.
I
believe
the
best
person
to
the
best
person
to
speak
to
is
is
amy
perrin
off
at
the
cncf,
but
you
should
also
be
able
to
look
at
the
the
current
status
in
the
guitar
repo
for
for
sandbox.
So
there
is
a
let
me
I
mean
in
fact.
Let
me
just
quickly
look
that
up
now.