►
From YouTube: GitLab 13.3 Kickoff - Enablement:Memory
Description
What we are working on for the Memory team in 13.3
A
If
you've
followed
us
in
the
past
releases,
one
area
that
we
even
working
on
is
to
help
improve
the
visibility
we
have
into
the
operational
characteristics
of
getting
applications
out
in
the
wider
field.
So
the
instances
that
our
customers
run
we'd
like
to
understand
basic
characteristics
like
how
much
memory
is
it
consuming?
How
many
CPUs
are
on
a
particular
box
and
how
many
requests
can
that
node
actually
handle?
This
can
give
us
a
better
idea
of
how
the
application
is
performing
and
where
we
should
target
our
investments
in
the
future.
A
Moving
on
forward
we're
looking
at
executing
on
performance
improvements
as
well
with
memory,
duction,
x'
and
one
area
that
we
found
is
that
we
found
a
particular
pain
point
here
with
the
cache
sequel
calls.
These
queries
can
help
to
reduce
the
database
load,
but
they
still
do
consume
a
fair
amount
of
memory
and
CPU
time
we're
essentially
going
to
recommend
treating
them
across
the
water
organization,
as
n
plus
1
like
queries,
so
we
can
find
them
find
the
worst
offenders
and
just
continue
to
optimize
them
and
remove
them.
A
We
actually
had
some
progress
here
on
13.2
with
a
particular
endpoint,
the
notes
endpoint
for
the
controller,
and
you
can
see
here
some
of
the
results
and
we
actually
had
pretty
impressive
performance
improvements
right
now.
We
had
around
a
23,000
accounts
of
the
calls
that
are
made.
This
resulted
in
about
45
Meg's
of
memory
being
allocated
and
I'm.
Sorry,
actually,
a
too
much
are
being
allocated
and
then
in
the
working
branch
that
we
have.
Hopefully
it
gets
emergent,
something
that
you,
if
not
it
will
be
entering
about
three
you
can
see.
A
We've
have
memory,
consumption
and
we've
also
smoothly
reduced
more
than
half
the
total
retained
memory,
and
you
can
see.
We
also
significantly
reduce
the
number
of
calls
as
well
for
the
database.
So,
all
in
all,
a
significant
improvement-
and
also
you
can
see
here
the
overall
time
it
took
to
complete-
is
also
steadily
reduced
as
well.
A
One
thing
we
have
found
is
that
one
of
the
big
drivers
of
the
slower
performance
on
the
user
side
is
actually
in
the
front
end
side,
not
necessarily
in
the
backend
side.
Although
you
can
see
we're
also,
as
I
mentioned
earlier,
focusing
on
improving
that
as
well.
But
one
of
the
areas
is
that
we
actually
send
quite
a
lot
of
data
down
to
our
users.
I
mean
in
the
form
of
images,
and
we
aren't
doing
a
very
good
job
of
shrinking.
A
A
We
currently
like
the
image
proxy
service
and
we're
go
ahead
and
work
on
a
proof
of
concept
to
help
us
understand
how
that
service
works,
its
operational
characteristics
and
get
a
better
idea
of
some
of
the
cost
involved
in
terms
of
CPU
of
resizing
in
terms
of
reductions
and
transfer
rate
and
other
information.
So
can
build
a
better
idea
of
what
we're
seeing
here
and
then
we'll
work
to
continue
to
improve
this
move.
A
The
next
few
releases,
with
things
like
improving,
adding
caching
support
along
the
way,
will
likely
rolla's
on
github.com,
initially
sort
of
gets
those
information
with
only
a
small
percentage
of
requests.
As
you
continue
to
improve
this,
but
overall
excite
about
the
opportunity
here
and
being
able
to
deliver
this,
then
only
by
com,
but
also
to
all
of
our
customers
as
well.
A
So
Kamino
is
working
on
this,
where
we
can
start
to
actually
have
a
ENL
file
definition
in
the
first
iteration
here,
so
we
can
track
what
it
is,
what
the
default
is,
who
owns
it
and
also
when
we
should
target
removal
the
feature
flag.
We
no
longer
need
it,
and
this
will
help
us
clean
up
these
flags.
We
will
ever
need
them,
so
don't
have
this
ever-expanding
matrix
of
configuration
options.
A
So
this
is
a
great
work
for
us
internally
and
once
we
have
this
for
my
little
bit
more
improved,
well
awesome
working
to
contribute
this
back.
The
larger
feature
fact
functionality
and
get
lab
so
that
we
can
have
all
of
our
customers
to
benefit
from,
and
with
that
that
is
the
major
items
working
on
delivering
here
in
the
33
release.
We
are
very
excited
about
them
again,
Ling
the
foundation
for
better
understanding
where
we
should
target
our
efforts
in
the
future.
A
Some
significant
performance
and
memory
reductions
on
the
back
end
side
with
the
cash
circle
queries
as
well
as
also
working
to
improve
the
front
end
performance
as
well
by
offering
resized
images,
so
we
can
reduce
the
work
that
browser
has
to
do
and
the
amount
that
has
to
be
transferred
over
the
wire.
Thank
you
so
much.
We
are
super
excited
and
can't
wait
for
this
release
and
stay
tuned
for
all
the
great
improvements
we're
working
on
next
release
as
well.
Thanks.