►
From YouTube: Improving our Scaling and Availability page
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
A
C
A
A
C
Topic
of
restructuring
this
my
idea
was
it
trying,
maybe
split
the
two
concerns
scaling
and
availability,
because
you
can
scale
without
necessarily
investing
in
availability,
and
so
we
could
have
to
sort
of
sections
and
then
have
a
good
table
for
each
I
can
take
that
up
in
that
one
epoch
in
length
here,
but
you
can
have
a
simple
table
of
what
we
recommend
for
the
different
user
counts
and
then
the
costs
associated,
for
example
like
estimated
off
the
AWS
for
GCP
and
then
similarly
for
availability.
We
can
go
through
and
talk
about.
C
Ok
great,
so
you
want
to
achieve
certain
availability
for
certain
kinds
of
events
and
we
can
go
through
and
talk
about
what
the
recovery
point
objective
is
with
recovery.
Time
objective
is
and
then
also
the
Associated
cost
as
well,
because,
again
you
know
running
geo
isn't
free,
and
so
it's,
if
you
think
you
need
that
for
your
business
continuity
then
great.
We
have
that
available
for
you,
but
you
know
I,
don't
maybe
they
don't
start
there
and
be
be
cognizant
to
the
fact
that
it
will
increase
your
cost
as
well.
Yeah.
B
A
B
A
Because,
as
sometimes
the
hae
causes
more
problems
than
in
solving-
and
that
should
be
super
super
clear-
then
there's
this
so
let's
split
the
page
up
and
answer
get
rid
of
the
maturity
levels
just
have
scaling
and
availability
in
the
availability
part
should
stress
kind
of
the
trade-offs
that
you
make
in
cost
and
complexity
and
maintenance
and
everything
else
and
in
availability.
Potentially,
if
you,
if
you
make
a
mistake,
then
there's
good
lab
components
and
configuration
instructions
like
what's
that
doing
here
we
have
a
components
page.
C
D
We
used
to
have
step-by-step
instructions
on
the
order
in
which
you
should
set
up
so
start
with
this
component.
Then
do
this
next
component
that
got
removed
recently,
but
there's
an
MI
to
put
that
back
up
so
I
think
I,
don't
know
if
it
belongs
on
this
page,
but
I
think
we
do
need
instructions
for
the
order
in
which
you
go
about
setting
up
AJ
yeah.
A
So
considering
the
domain
name
and
the
URL
for
this
page,
let's
make
this
page
about
high
availability
and
how
to
set
it
up
so
leave
this.
If
this
is
a
step
by
this
is
going
to
be
a
step
by
step.
That
makes
sense,
let's
create
a
scaling
page
and
let's
create
an
availability
page.
So
the
scaling
page
is
like
different
sizes
and
then
the
by
the
way,
the
scaling
page
would
that
have
these
ones?
So
should
we
call
that
the
reference
architecture,
pages
yeah.
C
I
think
we're
I
call.
It
I
think
we
can
call
it
the
scaling
page
and
have
they're
up
in
the
skilling
pages.
We'll
use
the
reference
architectures
as
well.
References
for
for
various
sizes
and
architectures
you
can
use
I
do
think
we
need
smaller
reference.
Architectures
like
I,
was
we're
missing
that,
like
two
node
configuration
or
you
just
got
two
of
them
to
handle
increase
low
without
going
to
fully
split
out.
A
C
C
C
A
C
C
C
A
A
C
The
only
other
detail
which
I've
learned
trying
to
derive
direct
upload
for
object
storage,
as
the
only
option
is
that
we
only
support
s3,
compatible
and
GCP
object.
Storage.
What's
rekt
upload
anything
if
you
want
to
use
something
like
OpenStack
or
other
digitalocean
providers,
you
have
to
have
NFS
because
it
requires
the
background
uploader.
So
just
in
context
for
un
so
does.
C
Double
check,
but
there's
there's
a
very
few
number
of
providers
that
we
have
direct
upload.
Support
for
ok
has
to
do
with
like
multi
product
loads.
I
believe
that.
A
A
Make
sure
we
have
a
scaling
page
with
the
reference
architects
make
sure
we
have
ability
page
that
includes
the
trade
offs
and
then
we'll
add
TCP
equivalent
and
that
NFS
make
it
super
clear
that
it's
not
required.
If
you,
if
you
can
do
that,
the
multi-part
upload
so
now
I'll
link
to
like
it's
the
thing
you
just
said:
hey,
multi-part,
upload
or
uploads
are
only
supported
for
a
limited
number
of
platforms,
and
things
like
that.
C
Thanks
yeah
I
think
I'm
the
solution
page
we've
done
a
lot
to
sort
of
move
things
over
here,
which
is
great,
then
maybe
keep
that
more
marketing
focused
you'll
just
be
like
exception
handling.
This
will
be
like.
Yes,
we
check
those
boxes
for
both
available.
You
know
scaling,
we
can
meet
your
needs
go
here
for
me,
more
information.
A
B
And
illuminating
in
upend
us
all
together,
it
was
the
thing
that
I
had
originally
written
to
show
here's
how
you
get
rid
of
it
and
then
it
got
moved
to
this
other
page.
So
that's
why
that
we
need
to
clean
up
that
link,
cool
yeah,
I,
think
I,
think
I
think
we're
slowly
Jeanne,
something
that's
more
useful.
A
A
If
we
have
all
these
small
installations
with
H
a
it's
probably
gonna
call
us
more
downtown
than
it
solves
and
that
and
it
creates
an
enormous
support
burden
on
us
and
we
get
a
ton
of
questions
and
a
lot
of
essay
and
time
time
trying
to
fix
these
things.
So
that's
that's.
Why
we're
improving
this
page.
C
Thanks
ed,
the
one
other
thing
just
a
note
here
is
it
released.
High
availability
is
that
we
previously
stated
we
could
do
zero.
Downtime
upgrades
on
single
nodes
wasn't
always
the
case,
but
the
way
unicorn
handled
updates
it
was
purely
graceful,
and
so,
as
long
as
you
weren't
upgrading
some
dependencies
you'd
be
fine
with
Puma.
A
That
makes
sense,
I
do
think,
there's
an
aspect
to
like
how
much
downtime
is
zero.
Downtime
is
really
clear,
zero
seconds,
but
anything
to
two
minutes
are
not
too
bad
compared
to
some
alternative
products.
You
get
lab
that
I've
hours
of
downtime,
but
thanks
for
that
contact
and
since
the
goal
is
to
prevent
load
on
our
support
teams,
it's
okay
to
be
a
bit
prescriptive.
We
can
like
people
know
they
have
options
and
they
can
add
AJ.
We
should
put
the
advice
out
here.
A
That
is,
if
you
do
this-
you,
because
otherwise
it's
gonna
lead
to
a
load
on
our
support
and
we've
we've
experienced
that
people
will
do
whatever
they
want,
which
is
fine.
That's
why
it's
a
open
source
and
we
shouldn't
encourage
that.
But
it's
okay
to
be
a
bit
more
prescriptive
and
just
have
our
best
guess
on
this
page.
Tell
people
what
to
do.
We
have
more
contacts
than
anyone
and
we
should
tell
them
what
to
do
because
does
it
make
sense
to
say?
Well,
you
can
do
this
or
this
well,
we
get
like.