►
From YouTube: 2019-01-29 :: Crimson SeaStor/OSD Meeting
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
A
To
review
the
the
error,
Akash
I
have
been
been
reading
a
little
bit,
but
didn't
make
much
progress
if
it's
possible
Jimbo.
Probably
we
can
discuss
a
little
bit
of
line
that
here
Congress
at
your
convenience
and
I.
Just
let
me
that
lead
from
Ricardo
that
they
have
pretty
much
visualized
completed
the
missing
a
big
to
implementation.
Probably
you
can
catch
up
with
done.
I.
B
I
all
pass
and
remove
try
to
remove
the
cues
I
found
a
little
complex
because
in
the
existing
step
OSD,
when
they
need
to
wait
for
something
forum,
for
example,
waited
the
OSD
map
the
PT
created
he.
It
can
just
include
the
operation
and
the
to
our
other
other
jobs.
But
I
thought
us.
If
we
removed
the
queue
we
should
return
future
and
then
that
job
will
be
will
be
stopped.
So
we
couldn't
start
a
task
along
this
for
this
person,
because
we
needed
the
job
of
the
earth
waiting
for
the
future
to
be
available.
A
Partially,
actually,
I've
been
being
trained
where,
when,
when
Riddick
after
Riddick
present
asta
with
presented
us
with
with
the
fast
already
passed,
his
public
typing
of
with
existing
no
team
I
sketched
up
sketched
our
military
implementation
of
a
red
path.
Getting
pseudocode
on
in
a
paired
I
post,
the
link
in
his
his
PR
and
I
I
was
thinking
about.
If
we
we
can,
we
click
if
we
could
to
get
written
out
continue
cute
future.
A
But
it's
nothing
that
nothing
is.
That
idea
is
not
not
necessary
correct
because
because,
as
you
mean,
we
have
like
three
three
preconditions:
we
need
to
fulfill
before
the
tutor
okie,
for
example,
the
period
of
PG,
the
update
to
the
Opie
or
the
or
the
object
it
is
in
this
context,
in
this
context,
is
the
is
not
missing
or
is
not
being
held
by
a
bath
scrub,
big
log,
being
locked
by
a
scrub.
A
We,
we
cannot
assume
that
the
the
three
apprec--
the
three
three
precondition
is
still
still
can
be
still
few
field.
If
the
pending
one
can
be
triggered,
so
we
need
to
control
them
again
once
the
last
one
is
fulfilled,
is
am
I
clear?
Yes?
Yes,
yes,
so
we
we
definitely
need
need
need
need
to
do
the
same
task.
Well,
when,
when
the
last
compute
condition
is
the
pending,
complete
condition
is
no
fuels,
so
we
need
to
have
some
sort
of
a
cure
or
sort
of
things
yeah.
We
need
to
to
to
put
another
way.
A
B
B
If
wait
for
some
condition
is
not
available,
we
return
a
future.
All
the
job
is
scheduled
by
the
by
the
sea
star
reactor
and
we
have
to
multiple
entrance
to
go
to
the
I/o
past
for
the
existing
code.
They
may
be
the
entry
pass.
They
just
include
the
a
pressure
under
to
another
task
for
the
only
entry
they
can
do
multiple
tasks,
but
for
us
we
need
a
multiple
interest
to
trigger
the
tasks.
B
B
A
A
C
B
A
B
D
D
D
Think
it
depends
maybe
because
a
message
if
we
include
a
message
we
can
let
the
right
event
to
reap
all
the
pending
messages
together
to
form
a
very
large
buffer
and
then
send
out
together,
and
is
that
I
think
that's
why
it
is
faster
and
there's?
No,
there
will
be
much
less
continuations
in
on
fly.
So
that's
why
it
is
faster.
There
is
only
one
dispatch
event
that
will
wrong
again
again,
but
they
don't
chain
together.
B
D
A
D
A
D
A
I
single
regarding
to
the
turmeric,
adding
to
the
to
the
o'the
GOP
candidate,
end
to
use
the
the
Q
we
can
always
to
to
acute
the
pending
operation
in
the
in
the
event,
in
the
same
thread
where
we
we
fulfill
the
precondition,
but
anyway
we
can
discuss
this
later
on.
After
the
after
the
of
the
update,
okay.
A
C
Last
week,
I
did
a
some
tests,
I
limited
at
the
client
a
yet
and
the
using
prophet
or
ricotta
Cisco
test
Cisco
include
sis.
Read
this
a
message
and
the
cccc
poet
or
Peter,
and
from
the
and
from
the
returned
I
found
I
found
yah
cinemas
in
a
single
message,
she's
called
read
and
see
and
cysts
in
a
message
a
lot
more
than
see
system
message.
C
B
B
A
C
D
A
B
C
D
E
E
C
E
For
the
first
one
we
can
see
that
the
T
star
messenger
was
utila,
was
utilized
on
the
hull
of
tip.
You
would
expect
I
won't
say
the
the
Crimson
Messenger
has
has
taken
the
worker
they'll
expect
and
sister.
We
know
that,
even
if
there
is
no
IO
worker
will
be
consumed,
but
the
proportion
are
much.
Much
are
much
more
interesting
for
the
100
cake
test
case,
because
I
guess
we
have
Apple
to
upper
comprising
both
rough
messengers
have
saturated
CPU.
D
B
A
D
So
the
first
thing
I
need
to
say
is
that
I
implemented
another
path:
tour,
because
current
FIO
path
engine
can
only
work
with
a
synchronous,
messenger
client.
So
I'm
writing
this,
because
I
want
to
do
pure
crimson,
taupe,
crimson
performance
check
to
performance
test.
So
this
the
first
thing
I
did
and
then
this
is
the
performance.
So
that's
why
we
can
compare
the
performance
of
crimson
to
chromosome
and
are
seeing
two
extinct
and
compare
different
scenarios.
B
D
D
D
A
D
No,
not
exactly
because
we
have
many
messages
to
send
and
maybe
actual
right
socket
is
not
quick
enough,
though.
So,
if
we
just
change
feature
together,
there
will
be
a
lot
of
continuations
pending
and
pending
there
that
that
will
be
a
huge
overhead
there
to
the
reactor,
though.
Instead
we
just
include
it
and
then
finished.
D
A
D
D
C
C
D
D
D
C
D
I
think
it
means
more
about
latency
analysis.
So
if
we
limit
to
the
queue
depth
here
to
limit
only
a
small
number,
though
the
the
message
will
wait
for
each
other
too
to
complete.
So
so
that's
why
I
think
it
is
more.
This
number
is
more
related
to
latency,
so
I
think
in
the
remote
case,
the
chromosome
latency
is
a
little
slower
than
the
synchronous,
because
there
there
is
I
think
there
may
be
some
merging.
A
B
D
B
B
A
B
D
A
A
A
D
D
A
D
It
change
okay,
so
let
me
put
here
is
this:
is
proof
here:
I
have
approve
here,
so
I
haven't
strayed
implementations
in
the
messenger,
so
in
the
foreign
this
passed
in
the
foreign
dispatched
case,
I
send
the
message
always
from
a
different
course,
so
I
have
to
submit
to
the
connection
core
for
each
ping
message
and
the
result
is
only
68k
ping
pong
per
second
for
the
unit
test,
and
the
not
optimized
case
here
is
that
we
queue
the
continuations
for
the
rights
and
we
can
see.
142
is
better
than
than
this
one,
but
slower.
B
B
B
B
D
A
B
A
E
Here
is
at
the
moment
I'm,
mostly
on
on
making
optimizations
there
and
also
on
to
security
I.
Did
it
teams
that
I've
miss
me
I
am
misusing
the
DCM
mouth
need
to
make
some
changes
from
the
things
somehow
related
to
to
Crimson
I
stand
the
first,
the
amended
version
of
of
the
PR
killing
up
in
the
buffer
buffer
list.
A
E
A
You
know
that
Jason
is
trying
to
try
to
come
up
with
the
liberators
to
comparable
results,
go
compatible
liberators
and
I'm,
not
sure
how
how
how
it
will
work
and
how
will
the
impact,
how
your
chain
will?
You
will
be
involved
with
the
histor
his
efforts,
but
one
thing
I
can
I
I
can
assure
that
we
need
to
get
it
get
as
much
other,
but
more
as
many
as
impossible
of
your
ABI,
Inc
or
API
incompatible
change
before
we
we
release
a
little
reason
to
or
reliever
a
little
free,
but
I
see.
E
E
A
C
D
A
D
D
C
D
A
Can
if
you
can
always
grab
because
I
in
the
last
ten
course
the
end
up
I
were
raising
my
concern
regarding
to
how
far
we
are
behind
from
from
ready
to
implementation.
The
visual
inflammation
is
pretty
much
in
shape
and
will
be
in
will
be
almost
done
before
we
roll
out
knotless
and
Ricardo
was
very,
very
kind
to
to
offer
his
help
to
two.
So
if
we
can,
we
need
some
help.
We
can
again
he
can
help
us
regarding
to
the
promised
messenger.
A
C
A
C
D
C
D
D
Previously,
I
don't
have
so
much
performance
data
to
present,
so
I,
don't
think
them
there's
no
there's
so
much
things
to
talk
about,
but
those
that's
why
this
week,
I
was
mostly
working
on
two
together
performance.
Get
some
sense
about
the
performance,
though
oh
yeah
I,
think
I
will
submit
a
proposal.
Maybe
today.
A
D
E
E
C
E
E
E
A
Riddle
and
I'm
still
thinking
about
the
idea
of
hinding
the
difference
between
one-to-one
mapping
and
m2m
emiting.
My
concern
isn't
more
regarding
to
how,
if
we
it
will
make
the
difference
more,
it
helps
the
hide,
a
difference,
but
my
concern
that
it
might
encourage
us
to
think
another
difference
between
course,
communication
and
an
on-court
communication.
A
E
A
A
A
A
Keep
an
eye
out
to
try
to
get
get
in
touch
with
you.
Regarding
to
the
your
changing
of
a
share.
Are
you
I
ask
you
got
some
I
still
have
my
head?
I
run
all
right,
I've
sent
a
mail
to
maybe
we
can
have
a
discussion
of
line?
Okay
that
sound
good
to
you.
Okay,
thank
you
and
to
me
probably
I
we
can
we
can.
We
can
try
out
the
Q
approach
and
to
untended
how
we
can
we
how
we
can
continue
resume.
A
The
pending
penny,
Opie
I
was
thinking
about
analyst
dropper
mail
to
a
civet
ever
and
the
copy
you.
So
we
can
continue
on
the
discussion
to
see.
If
we
can,
we
can
come
up
a
bad
idea
and
have
a
more
spectacle
design
to
address
the
problem
we
need
for
sure
and
on
anything
I.
Think
then.
The
next
thing
is
to
pour
for
sure
is
to
to
since
sender
proposal
to
this
difficult
before
the
deadline,
yeah
and
I.
A
A
B
A
B
A
A
Yeah
we
can
we
we
need,
we
need
it,
but
what
I
mean
is
that
the
survey
say
is
already
been
implemented,
but
it's
not
exposed
as
the
sort
of
like
a
global
variable
or
something
when
you
always
need
to
pass
the
OT,
the
OT
pointer
down
to
where
it
is
it's
used.
For
example,
we
have
a
PG
like
or
PG
back-end
or
or
the
ultimate
at
where,
where
the
OT
is
because
it
is
consumed,
but
what
this
consumer
needs
is
is
better
a
service
which
can
can
offers
the
OC
map
right.