►
From YouTube: 2020 02 24 Database Sharding Working Group
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
Okay,
so
start
with
follow
up
from
last
week,
so
PG
11
is
now
on
nightly,
builds
grant
got
PG
11
through
the
10k
reference
architecture,
said
everything
went
well
and
the
he
gave
us
a
very
detailed
breakdown
at
the
bottom
of
the
issue.
There
were
only
I
think.
Overall,
it
was
better
and
there
were
like
one
or
two
end
points
where
it
was
slower,
but
not
significantly
from
what
I
recall.
A
A
C
A
A
Created
a
board
we're
starting
to
get
some
more
issues
related
to
the
sharding
working
group,
so
I
just
created
a
board,
so
we
can
track
and
we
can
see
where
there
are
blockers
what's
being
worked
on
actively,
what's
planned
out
far
into
the
future.
This
is
not
comprehensive
by
any
means.
I
didn't
go
through
all
of
the
issues,
but
at
least
we
have
a
bore
if
we
can
start
throwing
things
out
there
and
it's
just
based
off
of
the
sharding
working
group
label
at
the
moment.
E
E
Yeah
I'm
actually
reading
it
to
Silas
TV
and
it's
very
similar
to
what
we've
been
been
discussing
recently
and
yeah.
It
resonates
with
that
quite
a
bit.
I
would
like
to
explore
it
more
I
think
there
is
more
work
that
we
have
to
do
in
any
case
than
just
exchanging
the
database
system.
So
it's
I,
don't
think
there
is
a
metric
solution
where
we
just
changed.
A
F
Think
so,
we've
had
an
issue
in
the
past
for
science.
We
actually
talked
to
the
sun-sized
founders
years
ago
before
they
were
acquired
by
Microsoft.
So
there's
other
past
history.
Already
that
you
know
the
situs
founders
were
giving
us
tips
about
what
we
should
be
doing.
So
we
should
actually
link
those
issues.
Okay,.
F
F
A
G
H
G
I
think
that's
a
great
idea:
do
we
have
all
the
monitoring
and
staging
everybody's
it's
like,
because
we
can
run
the
performance.
That's
on
staging.
You
can
run
the
regression
test,
but
I
think
with
with
the
DB
changes.
I
think
the
first
signal
is
like
century
or
any
other
signals
that
you
get
in
the
static.
G
H
G
G
D
I
You're
typing
sorry
I
was
just
finishing
up
my
typing
thoughts
into
the
doc.
With
this
PG
upgrade.
Is
there
any
specific
problem
or
plan
for
the
there's
some
kind
of
collation?
Other
issues
with
Postgres
upgrades
I
know:
there's
I
was
I,
remember
from
previous
conversations
about
Postgres
after
grades
that
it
wasn't
going
to
be
smooth
due
to
replication
or
the
Lib
C
changes.
I
think.
H
B
A
I
This
is
actually
more
of
a
question
for
Eric
Oh,
more
business
type
growth,
yeah
cuz.
Basically,
like
you
know
this,
we
have
the
current
like
capacity
projections
based
on
current
use,
and
but
you
know
what
is
like
this
is.
This
is
more
of
a
market
market
share
question
is
you
know?
I
Realistically,
I
was
thinking,
you
know
what
is
our,
what
could
be
our
possible
potential
growth
rate
and
what
level
of
sharding
and
what
kind
of
sharding
techniques
would
let
us
operate
for
how
long,
because
you
know
just
thinking
kind
of
long
term
how
much
how
much
input
effort
it
is
for
various
charting
techniques
and
plans
versus
how
much
how
much,
how
much
opportunity
we
get
out
of
them.
Like,
like
Eric,
said
you
know,
he
wanted
to
be
able
to
open
the
floodgates
to
sales,
but
what?
What
is
that?
What
would
that
actually
look
like.
B
C
J
C
Looking
for
less
on
the
what's,
the
courier
look
like
and
more
like:
what's
the
five-year
plan,
as
first
projections
go
anyway,
he's
specifically
looking
as
like,
you
know:
sass
cuz
Eric's,
there
original
statement
was,
is
you
know
we
could
unleash
sales
on
the
south
side
of
the
solution
and
because
of
that,
that
would
cause
us
to
has
he
rapid
growth
in
a
multi-tenant
solution
for
scalability
super
important?
So
it's
kind
of
the.
J
But
it's
a
good
question
sort
of
like
if
we
do
get
and
if
we
do
get
thick
get
out
of
comm
to
be
underpriced,
ready
and
all
the
things
that
entails
right.
It's
not
just
about
obviously
scaling,
it's
all
the
other
features
and
getting
managed
and
in
other
places
that
we
have
caps
right
now
like
what
will
growth
look
like
so
I?
Don't
have
an
answer
right
now,
but
I'll
take
them
actually
I'm,
try
and
see
what
we
can
prognosticate
I
think
our
our
confidence
will
be
low
as
far
as
how
it's
yeah.
I
I
And
recognized,
and-
and
that's
and
and
that's
kind
of
my
basic
question
is
is-
is
how
far
are
we
gonna
have
to
go
and
like
not
not
concretely,
but
just
like
kind
of
hand-wavy
like
what
is?
What
is
the
total
available
market
to
us
as
a
company
from
an
enterprise
perspective?
There's
got
to
be
a
numbers
in
there,
because
otherwise,
we
you
know
are
we're
gonna
we're
gonna
gain
all
the
users
that
we
could
possibly
gain
in
a
certain
amount
of
time.
So
you
know
where's
where's
it
going
to
end
that
this
was.
I
This
is
actually
one
of
the
the
biggest
failings
of
my
previous
job
was
the
the
founders.
The
founders
thought
they
could
have.
You
know
infinite
billions
of
users
when
their
their
actual
market
size
was
wasn't,
was
not
that
big
and
so
I
want
to
make
sure
that
we're
were
not
shooting
to
get
to
another
solar
system
without
without
cause
and
I.
I
F
Our
limiting
factor
when
their
CIO
builds,
for
example,
or
disk
those
scale,
not
by
necessarily
number
of
users,
but
just
of
activity
right
so
I
think
it
would
be
helpful
to
kind
of
have
a
projection
of
our
current
usage.
If
you
assume
different
rates
right,
if
we
assume
you
know
this
current
trend,
what
does
it
look
like
if
we
sects
assume
it's
20%,
30%
more
over
a
year
after
year,
that
might
inform
it
better
than
trying
to
ask
the
business
case
like
how
what's
gonna
happen?
J
I
think
I
would
just
say
high
level
two
comments,
one
all
we're
gonna
to
is
I
I,
think
there's
a
general
recognition
in
the
company
and
I
can
be
wrong.
But
my
sense
is
that
that
we're
not
really
that
interested
in
trying
to
take
the
open
source
community
away
from
github,
and
so
it's
gonna
more
around
enterprises,
rather
than
trying
to
pull
the
open
source
community
to
get
that
and
so
I
think
we'll
end
up.
K
On
that
yeah
and
when
you
working
it
out,
like
I,
think
Stan
alluded
to
it
as
well
like
try.
Don't
talk
about
total
number
of
users.
Talk
about
because,
like
Chris
said,
we've
got
Christopher
sit
and
got
five
million
users
registered.
Our
daily
active
is
about
two
hundred.
Ninety
thousand
I,
don't
know
what
our
monthly
activist,
but
it
may
use
a
much
better
sort
of
metric
to
be
measuring
rather
than
like
total
users
so
like
when
we're
doing
all
of
that.
Just
thinking
and
they
use
yes.
I
Yeah
I
prefer
to
think
about
daily
active
for
for
peak
weekday
activity
because
it
kind
of
gives
us
a
better
gauge
for
what
we
need
to
keep
hot
and
memory
for
activity
during
a
day.
It's
like
you
know
it
it.
It
has
more
daily
active,
has
more
impact
on
our
our
peaked
peak
time
daily
use,
and
that's
that's
where
the
performance
needs
to
be
is
on
a
daily
basis,
not
on
a
monthly
basis.
Oh.
C
J
C
D
A
C
So
what
like
the
Goodell
EHA
team
has
been
doing,
as
is
they've,
been
basically
running
through
demos
twice
a
week.
Doesn't
it
enjoys
a
week
just
trying
to
give
you
the
primer
so
how
they're
doing
it,
and
basically
they
have
somebody
step
through
the
process
of
doing
you
know,
setting
up
giddily,
AJ
or
some
aspect
of
fiddly
hae
experience
associated
so
stand.
C
D
C
C
A
C
E
What's
what's
already
clear
from
from
Silas
to
be
able
also
firm
positioning
is
that
there
are
parts
of
the
application
that
sort
of
are
a
good
match
for
this
and
other
parts
where
it's
maybe
more
difficult
and
I
think
we
have
to
also
do
work
on
the
application
side.
Whatever
solution
we
have
on
the
backend,
it's
not
that
we
find
a
database
solution
that
sort
of
enables
us
to
use
charting
from
the
application
that
we
have
today.
E
J
C
Issues
that
you
keep
in
mind,
yeah
mm.
The
reason
to
do
the
demos
is
because,
like
dentist,
a
self
hosted
solution,
kind
of
you
get
a
feel
for
that
and
yeah
I
agree
like
if
there's
stuff
that
we
can
do
independent
of
the
DB
decision.
We
should
definitely
like
push
on
that
aspect
as
well,
just
just
so
that
we
can
make
an
iterative
process
hitter
of
progress,
cool
thanks.
I
Do
we
have
any
cell
posted
customers
that
are
anywhere
near
needing
the
kind
of
sharding
that
we
would
need
that
with
that
they
would
actually
need
that
would
date
they
could
they
can
improve
with
charting
or
or
yeah?
Is
there
is
there
any
kind
of?
Is
there
any
indication
that
customers
are
reaching
this
risk
scale.
F
J
J
So
those
are
my
question
marks,
because
my
main
concern
is
with
any
of
these
things
is
that
you
know
it
could
refer.
Saying
customers
have
to
run
site
is
to
be,
then
we
need
to
look
at
how
many
have
ever
tried
this
before
or
whether
they
would
be
interested
in
running
it.
You
know
whether
they
can
use
something
like
RDS
or
Aurora,
which
I
don't
think
they
can
to
take
over
the
load
of
some
of
these
things.
A
E
For
as
far
as
saddles
to
be
and
postfix
goes,
my
understanding
is
that
Saito's
is
a
an
extension
to
post
Chris
and
it
supports
a
subset
of
the
puskás
features.
Only
so
in
that
sense,
if
we
as
an
application
support
side
has
to
be,
then
we
should
still
be
able
to
run
on
Postgres
I.
Think
that's
one
of
the
points
that
makes
it
kind
of
appealing
to
to
look
into.