►
From YouTube: CNCF SIG Storage Meeting 2019-10-23
Description
Join us for Kubernetes Forums Seoul, Sydney, Bengaluru and Delhi - learn more at kubecon.io
Don't miss KubeCon + CloudNativeCon 2020 events in Amsterdam March 30 - April 2, Shanghai July 28-30 and Boston November 17-20! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects
CNCF SIG Storage Meeting 2019-10-23
A
Good
morning,
we're
just
speaking
for
a
few
more
people
check.
A
A
C
A
C
C
A
A
Indeed,
so
anyhow,
if,
if
any
of
you
have
any
any
strong
opinions
on
this,
could
you
please
vote
yea
or
nay
for
any
of
the
particular
items?
My
vote
is
for
the
top
right
hand
corner
because
it's
a
clam
and
it's
with
the
CNCs
sort
of
logo
kind
of
thing
built
into
it,
which
makes
it
clear
soft
like
a
criminally
sick.
A
A
C
C
A
A
D
D
C
Believe,
I
Fiat
being
able
to
put
out
a
vote
for
like
what
the
top
three
of
these
would
be.
A
A
B
Documents
we
are
going
to
discuss.
I
asked
two
people
to
join
in
the
called
one
is
Deepti
she's,
a
maintainer
of
with
tests,
and
she
can
contribute
to
the
to
the
use
cases
and
the
other
one
is
safe.
Alpha
t
he's
been
doing
all
our
benchmarks,
so
he'll
be
able
to
contribute
to
the
benchmark
document.
E
F
E
E
The
right
dock
is
August,
Toni
self,
that's
correct,
I'm,
sharing
the
screen
so
you're
correct.
So
what
we
have,
let
me
start
start
with.
Those
are
at
noon
that
we
had
discussed
a
model
of
something
to
do.
We
want
to
give
some
documentation
for
users
after
they
read
the
storage
101
document
which
describes
what
is
block.
What
is
fire,
what
is
object,
what
is
the
database
and
so
on?
So
once
they
read
that,
and
they
have
a
preliminary
concept
of
what
storages
then
they
probably
want
to
ask.
Well
how
do
I
use
application
X?
E
What
storage
system
or
persistent
storage
should
I
use
with
application
X,
and
this
is
what
the
goal
of
this
task
is
and
what
we
wanted
to
do.
Is
we
have
a
lot
of
discussions,
I
think
just
before
August
27th
on
what
it
the
output
of
this
part
of
this
task
would
be.
We
have
first
that
it
would
be
a
document
that
has
many
applications
in
it.
We
also
had
discussed
on
the
output
being
a
github
markdown
pages,
where
the
community
themselves
would
you
know
we
were
started,
but
they
would
be
like
you
know.
E
The
community
would
continue
it
where
they
would
bring
an
app
document
and
they
would
describe
how
it
interacts
with
those
persistent
storage
services
that
are
described
in
the
storage,
101
document,
and
so
with
that
Simon
and
I
from
Simon
for
storage
and
I
got
together
and
kind
of
started
thinking.
You
know
just
brainstorming
on
what
would
it
take
to
actually
accomplish
this?
One
of
the
things
that
we
started
looking
at
is
that
we
may
need
more
information
in
this
is
what
we
have
this
section
down
here.
E
Almost
like
a
brief
for
each
application,
where
an
application
you
could
have,
we
could
have
a
matrix
of
an
application
and
and
how
it
deals
with
you
know
with
we
got
storage
and
then
we
got
applications
and
then
the
application
can
have
its
own
page
on
and
making
except.
You
know
not
making
twenty
pages
anything
like
that,
just
a
couple
of
pages
on
how
it
deals
and
probably
more
information
how
to
use
with
that
persistent
storage.
B
E
My
sequel
Cassandra
any
application
that
Kafka
any
application
that
we
think
of
today
and
our
implication
that
we
think
of
tomorrow,
so
we're
trying
to
to
start
with
something
simple.
You
know
sorry,
it's
gonna
be
iterative.
We
are
going
to
see
we're
gonna,
try
it
we're
gonna,
see
what
works
and
then
we're
gonna
cook
keep
adjusting,
but
it
yes
exactly
right.
If
I'm
a
customer
I,
don't
think
of
storage.
First
right,
I'm
customer,
if
I'm
a
customer,
then
I
think
of
my
application.
E
E
F
F
If
we
built
where
people
could
add
in
their
particular
storage
and
I
mean
we,
the
tool
can
identify
different
criteria
like
we
put
in
the
white
paper
about
storage
and
how
it
runs,
but
then
it
could
be
more
of
a
living.
The
problem
I
have
with
a
lot
of
these
docks
is
their
point
in
time.
Like
you
know,
we've
talked
about.
Oh,
we
shouldn't
put
that
in
or
we
should
use
this
example
of
this
particular
vendor.
F
F
F
They
could
go
in
list
the
criteria
of
what
that
storage
capability
is,
and
then
that
way
someone
could
use
this
tooling
to
say:
oh
I,
either
using
this
or
I'm
thinking
of
using
a
product
that
needs
high
performance,
they
could
search
kind
of
this
database
and
then
that
way
people
could,
you
know,
submit
a
github
to
update
their
new
project
as
it
comes
along
and
the
landscapes
more
of
a
living
document.
Then
these
point
in
time
document.
Sorry,
would
anyone
be
interested
in
working
like
that.
F
I
see
like
I'm,
hoping
that
we
continue
to
evolve.
The
sig
to
you
know
improve
the
landscape
for
storage.
We've
done
CN
CF
because
it's
you
know
compared
to
other
projects,
probably
pretty
slim,
and
then
you
know
you
would
be
able
to
take
these
sandbox
or
incubation
or
whatever
and
and
see.
What's
you
know
involved
what
criteria
has
I?
Don't
know,
don't
be
limited,
yeah
I,
don't
know
I'm
just
definitely.
B
B
E
A
I
think
we're
getting
where
were
conflicting,
two
different
things
yeah,
so
so,
first
so
first
of
I
think
we
had
agreed
that's
rather
than
calling
this
application,
we
were
going
to
call
it
a
use
case
so
that
we
can
use
this
for
sort
of
end-user
applications,
but
also
you
know
projects
and
things
like
that
and
and
specifically
I.
Don't
think
we
were.
What
we
wanted
to
do
here
was
to
get
to
a
stage
where
we
have
a
template.
A
Where
we
can
say
for
this,
you
use
games,
and
you
say
you
know,
the
use
case
could
be
Cassandra,
it
could
be,
it
could
be
Jenkins,
it
could
be
Kafka,
it
could
be
at
CD
or
Prometheus
or
whatever.
Then
we
would
for
the
first
articular
use
phase.
These
are
the
things
that
you
need
to
that
you
need
to
consider.
A
So,
for
example,
we
we
would
create
the
template
that
could
list
a
number
of
different
options
and
the
idea
it
would
be
that
people
pretend
the
community
grew
generators,
PR
is
to
say
okay,
so
for
Africa,
for
example,
if
you're
doing
Africa
weights
and
well
EBS
volumes,
for
example.
This
is
a
way
you
can
do
it
and,
if
you're
doing
at
CV
on
Google
client,
this
is
the
way
you
can
do
it
and
it
would.
But
it
would
include
the
same.
A
You
know
five
or
six
parameters
of
what
I've
defined
for
five
or
six
sections
from
the
template.
That
would
be
kind
of
mandatory
for
it
for
every
option.
For
that
use
case,
and
then
the
community
can
kind
of
build
the
sides,
but
I
think
the
first
thing
is:
we
need
to
lock
in
that's
templates
and
at
least
one
example,
and
if
we
do
that,
as
we
can
then
launch
it
and
we
can
get
the
community
to
to
contribute
right
because
we
we
could.
We
could
then
ask
the
different
projects.
E
E
A
A
And
then
you
could
have
another
one
with
three
droid,
many
or
whatever
else,
and
maybe
another
one
with
an
object,
store
or
whatever
else,
but
the
yeah
I
kind
of
agree
with
that.
But
we
do
need
to
also
keep
in
in
mind
to
make
some
provisions,
for
you
know
things
like
data
locality
or
how
do
you?
How
is
the
best
way
to
back
this
up,
for
example,
and
things
like
that,
so
it's
a
remainings
to
include
things
or
technologies
which
might
be
you
know
different
for
different
providers,
but
that's
probably
okay
to
and.
E
A
And
we
do
only
do
talk
about
point
in
time,
backups
and
snapshots,
and
things
like
that
so
but
of
course
I
think
those
things
are
covered,
the
the
thing,
if
you
have
there
about
sort
of
storage
appliances-
and
you
know
some
advanced
topics
like
nvme
over
fabrics
and
things
like
that-
I
think
they're
kind
of
interesting
but
I-
think
they're.
The
perhaps
maybe
for
a
version
to
more
advanced
kind
of
I
mean
I.
Don't
think
that
is
relevant
to
a
large
part
of
the
audience.
But
I'm
welcome
to
comment
on
okay.
B
E
A
Well,
so
the
reason
reason
why
the
reason
why
I
think
it
shouldn't
be
too
abstracted
is
because
for
this
to
be
useful,
it
kind
of
needs
to
be
it
kind
of
needs
to
be
relevant
to
a
use
days.
So,
for
example,
you
know,
if
we're
talking
databases,
is
it
a
good
idea
to
have
multiple
volumes
for,
say,
transaction
logs
to
split
up
the
transaction
logs
or
the
data
or
whatever
else
or
with
Kafka.
E
Maybe
I'm
just
thinking
out
loud.
The
goal
of
this
is
so
that
when
users
read
the
one-on-one
document
and
they
have
an
idea
of
what
the
applications
want
to
use,
they
can
go
to
the
storage,
vendors
and
ask
there
were
correct
questions
right
and
they
can
go
to
the
website
and
say.
Oh,
this
use
case
requires
the
following
things
or
suggest
the
following
things,
and
then
they
can
go
to
the
storage
vendor
and
said:
what
do
you
think
of
this?
E
A
A
We
could
put
like
a
simple
expiry
date
or
something
like
that
on
the
one
on
each
of
these
use
cases,
but
they
need
to
get
reviewed
once
a
year
or
something
or
once
every
two
years
or
something
like
that,
and
if,
if
it
past
the
expiry,
then
we
could
trace
the
pr2
to
archive
it.
Unless
anybody
wants
to
update
it
with
my
certain
time.
So
that's
we
can
kind
of
keep
them
current
I
think
it's
a
cool
idea.
I.
E
A
I
mean
what
would
be
absolutely
awesome
would
be
to
get
to
at
least
get
the
template
done,
with
maybe
some
reason,
for
example,
in
time
for
cute
card,
so
that
that's
so
that
we
can
we
can.
We
can
launch
it
in
the
sink
meeting,
as
we
have
a
few
can
hopefully,
that'll
encourage
some
of
the
community
to
contribute.
Okay,.
E
H
I
wanted
to
just
kind
of
echo
what
Lewis
was
saying.
We
just
really
have
to
be
careful,
because
this
is
territory
that
could
land
us
in
controversy,
so
as
much
as
possible.
If
we
could
avoid
kind
of
saying
hey,
this
is
a
specific
product
or
recommending
products
that
would
be
best
for
us,
as
generic
as
we
can
remain,
while
still
being
helpful,
should
kind
of
be
our
guiding
guiding
light.
E
A
A
A
So
so
the
idea
behind
this
performance
and
benchmarking
talk
is
to
provide
users
with
a
better
understanding
of
the
performance
of
storage
systems
and
to
give
them
some
information
on
the
tools
that
we
can
use
and
also
document.
You
know
some
of
the
some
of
the
very
common
pitfalls,
specifically
we're
saying
we're
saying
it.
You
know
we're
not
looking
to
publish
any
benchmark
numbers
here,
we're
not
looking
to
provide
any
comparisons
and
I'm
proposing
that
we
exclude
tools
for
distributed
benchmark
because
I
think
that's
probably
too
complicated
for
a
first
draft
of
this
document.
A
Making
things
I
have
sense
that
you'll
be
looking
at
when
doing
benchmarks.
So
things
like
you
know
through
high
up,
so
you
can
see,
etc,
but
also
things
like
the
variance
between
different
benchmark
runs
resources
and
those
benchmarks.
They're
going
to
use
the
scalability.
That
kind
of
thing
have
a
general
talk
about
the
architecture
and
the
topology
so
reminding
how
the
different
layers
in
a
storage
system
all
contributes
to
the
performance.
A
A
So
I'm
not
going
to
sort
of
read
through
all
of
these.
But
you
know
happy
to
have
lots
of
comments
here
and
we
you
know.
The
idea
is
that
for
each
of
these,
what
will
happen?
Maybe
a
couple
of
sentences
or
paragraphs
that
can
describe
the
issue
and
and
wide
by
it
impacts
the
performance
of
the
system.
There
are
a
few
areas
that
we
kind
of
discussed,
covering
some
advanced
topics.
A
Things
like
you
know
a
new
mind,
scalability
testing
and
things
like
that
which,
which
I've
kind
of
put
into
this
advanced
listing
here,
because
I'm
not
sure
whether
we
will
actually
go
through
it
or
not.
You
might
think
you
know
just
a
reference
to
it
as
something
we'll
do
in
the
future
and
then.
Finally,
there
is
a
section
on
on
the
benchmarking
tools,
so
first
off
things
that
we
need
to
sort
of
level
sets,
then
what
are
quality
level
setting
the
environment.
A
So
you
know
understanding
the
performance
of
the
cluster
in
if
you're
running
on
cities,
like
you
know
the
CPU
performance,
onions
and
network
performance
of
the
notes,
you're
running
on
and
some
tools
you
can
use
to
to
to
benchmark
close
and
then
volume
benchmarks,
where
I'm
proposing
we'd
focus
on
FIO
and
the
database
benchmarks.
Where
ivan,
who
are
proposing
that
we
focus
on
on
sis
bench,
so
I'd
love
to
have
sort
of
comments
or
feedback
on
this,
and
then
we
can.
A
We
can
break
up
the
different
sections
and
DIF
different
people
contributes
I'm
going
to
be
sort
of
driving
as
much
of
this
as
I
can
I'm
very
passionate
about
and
again
I'm
kind
of,
I'm
kind
of
aiming
to
get
to
a
point
where
we,
where
we
have
a
trough
that
potentially
we
can
publish
in
some
provides
for
for
a
few
concerts.
We
can
get
more
comments
from
the
land
from
the
community
and
more
feedback
from
the
community
at
that
stage.
So
there's.
B
A
B
So
there's
one
thing:
I
almost
forgot
to
mention:
I
was
I,
don't
know
if
you've
heard
of
a
company
called
percona,
they
actually
are
consulting
company.
That
offers
support
for
open-source
databases
they
yeah.
So
they
are
they've
done
extensive
work
on
benchmarking
and
they
said
that
they
would
be
twisted
in
contributing
so.
A
B
B
A
B
D
A
Absolutely
so
so
the
idea
would
be,
for
example,
with
go.
There
would
be
that
for
unit
oat,
so
I
think
we
we
we
described,
you
know
say
some
of
the
different,
the
different
workloads
that
we
would
be
talking
about,
so
we
would
give.
We
would
give
examples
for
about
defining
benchmarks
and
think
it's
benchmarks
for
those
specific
workers,
and
so
there
would
be
examples
of
running
it
as
the
results.
Sometimes
we
to
reproduce
after
the
results.
A
A
But
as
with
all
these
things
right,
we
we
we
have
to
you
know
it's
it's
good.
It's
gonna
have
to
be
we're.
Gonna
have
to
limit
the
scope
of
everything
that
we
can
do
because
some
of
these
products-
you
know
it
like
fire,
for
example-
hands
probably
about
five
hundred
introductions.
So
we'll
look
at
some
some
of
the
more
common
things,
yeah.
D
A
So
I
mean
long
story,
short
I'd,
love
to
I'd
love
to
get
comments
or
feedback
to
these
things.
You
think
we
should
add
things
that
maybe
you
disagree
with
and
if
anybody
is
interested
in
can
help
you
with
particular
sections
I'm
on
super
you
and
your
team
work
we're
interested
in
obviously
doing
some
anything
today,
stuff,
let's,
let's
mark
those
ayats,
and
maybe
we
can
start
putting
some
content
together.
A
B
So
safe,
safe,
we'll
share
the
Lua
scripts
turn
the
suspense
configurations
that
he
used
and
maybe
guru
level
guidance
on
that
so
that
he
can
take
care
of
I
I.
Actually
I.
Remember
you
sent
me
your
email,
but
I,
don't
know
where
you
sent
it
to
me
chill
cop
dot-com,
Alex,
homework,
slow,
ok,
I
will
make
sure
that
eyes.
I
said
you
I
connect
you
to
both
safe
and
the
percona
person.
I
will
find
out
their
email
and
then
you
can
take
it
from
there.
G
A
Yes,
so
so
super
hunter
mentioned
this,
and
that
would
be.
That
would
be
certainly
something
of
interest
we
needs
to.
A
A
About
sort
of
preparing
the
environments
or
for
benchmarking,
you
know
I'm,
obviously
aware
at
DBC
has
some
particularly
stringent
requirements
in
terms
of
you
know
the
numbers
of
clients
and
network
topologies
essentially,
and
things
like
that.
So
if,
if
we
can
make
it
in
a
simple
way
that
somebody
can
actually
do,
then
I
guess
it's
something
we
should
do
because.
A
That's
right,
that's
right
and-
and
you
know
the
reason
why
we're
covering
things
like
the
pitfalls
as
well
and
I'm
setting
up
the
environment
is
so
that
you
know
we
can
avoid
people
getting
sort
of
focused
results
right
because
of
loss
of
lost
track
of
the
number
of
times.
For
example,
I
know
somebody
has
done
a
database
benchmark
runner
on
a
one
gig
database,
which
runs
entirely
in
cash
ins.
You
know
get
some
ridiculous
results,
for
example,.
A
A
Unless
there
is
anything
else
to
to
cover
it,
those
are
the
main
things
we
had
on
the
agenda.
One
thing
I
was
going
to
ask
was
just
a
reminder:
if
anybody
had
feedback
for
parents
documents
now
is
the
time
to
to
provide
backs.
This
is
the
documents
that
we
discussed
last
meeting
for
the
project
review
process.
I,
don't
know
Aaron
if
you've
had
a
chance
to
my.
F
Know
that
the
TOC
was
actually
reviewing
it
and
adding
stuff
in
in
their
meeting
when
they
invited
me
over
so
I
haven't
looked
at
it
for
probably
a
week
or
so,
but
yeah
I
mean
I.
Think
since
we're
one
of
the
first
SIG's,
we
should
help
kind
of
set
some
I
don't
want
to
say
rules,
but
some
process
forward
about
how
projects
get
reviewed
in
criteria
and
then
what
happens
when
we
turn
them
over
to
the
TOC,
because
right
now,
I'm
feedback
I've
gotten
from
various,
produces.
F
They
feel
like
it's
kind
of
chaos,
and
it
still
seems
like
there's
a
you
know,
a
duplication
of
work
between
what
we're
doing
in
the
sig
and
then
projects
suddenly
presenting
as
well
to
the
TOC.
You
know,
so
we
need
to
figure
out
that
balance
so
yeah
any
feedback
or
suggestions.
The
doc
is
meant
to
just
kind
of
set
up
a
straw
man
and
get
people's
feedback
I'm
not
trying
to
dictate
how
I
think
we
should
do
things
so
I
people
have
the
time.
I'd
appreciate
them.
Looking
at
that,
please.
A
A
F
I
think
we
should
put
that
in
there
I
think
I
love
the
idea
of
it
and
I
think
all
SIG's
should
develop
kind
of
a
similar
template
based
on
criteria
for
their
specific
subject
matter
area.
So
thank
you
for
reminding
me
of
that.
I
will
put
that
into
the
document,
because
I
think
that
helps
projects
figure
out.
If
you
know
that's
even
the
level
this
should
be
going
in
if
you're
answering
those
initial
questions
etc.
So
yeah,
let
me
add
that
in
Aleks.
A
Okay,
silence,
I!
Guess,
that's
good!
All
right,
so
we'll
be
meeting
in
a
couple
of
weeks
area
will
put
the
agenda
forward
for
that.
For
that
call,
but
I
think
that's
going
to
be
the
last
meeting
before
before
Kubik
on
so
it
would
be,
it
would
be
if
we
wanted
to
against
the
use
case
templates
and
the
performance
drafts
out
for
Q
Khan.
That
would
be
a
really
good
time
to
check
once
those
documents.