►
From YouTube: Node.js Benchmarking WG Meeting - Oct 16 2028
Description
Google Form for Questions: https://docs.google.com/forms/d/e/1FAIpQLSea-eYU11_GYrJJCAXPX7hUM-CWsDPgTdFt0Xse64OzU0OHRA/viewform
A
B
B
C
A
So
I
I
agree
with
this.
I
was
recently
looking
into
adding
my
own
thing
for
an
NPM,
install
job
and
boy
I.
Wonder
I'm
supposed
to
implement
this,
and
you
know
there
isn't
a
consistent
way
to
implement
a
benchmark,
and
so
it's
hard
to
put
a
new
one
in
and
it's
sort
of
a
discouragement
in
that
regard.
A
A
So
in
in
the
issue,
155
Odom
suggests
I,
guess
10
scripts,
but
you
know
some
know:
every
benchmark
will
have
a
script
to
invoke
that
says,
install
the
dependencies
and
every
range
with
scripting
folks
that
says
load.
You
know
you,
you
would
give
the
script
a
set
of
IP
addresses
on
which
it
can
deploy
any
dependencies
like
MongoDB
database,
or
you
know
all
the
places
you
can
put
a
micro
service
or
something
and
then
it
lets.
A
You
know
what
it's
ready
and
then
you
launch
the
client
and
you
use
some
config
file,
that's
in
a
universally
agreed-upon
location,
and
you
know
that
kind
of
thing,
and
you
know
each
of
because
the
different
benchmarks,
currently
all
do
that
kind
of
thing.
They'll
have
variations
on
the
same
theme
of
how
to
do
this,
but
it'd
be
it'd,
be
great
if
they
were
the
same
theme
instead
of
just
variations
on
it.
In
my
mind,
ya.
B
A
B
A
B
Okay,
I
can
less
there's
some
reason
why
it
would
be
better
to
do
it.
The
other
way.
That
seems
like
an
easy
way
to,
and
it's
kind
of
what
we
already
have
like
in
terms
of
like
the
the
chart
generation
there's
already
a
directory
for
each
of
them
and
I
think
we
for
I
think
we
already
have
a
directory,
possibly
miss
name,
because
they
started
under
experimental
and
are
still
under
experimental,
but
like
a
directory
where
they
they
go.
A
A
B
B
Benchmarks
is
the
directory
that
actually
has
you
know
the
explanation
of
the
benchmarking
infrastructure,
the
IDS
and
then
the
individual
sort
of
files
for
each
of
the
at
the
chart
generation
mm-hmm.
Now
why
the
chart
generation
separate
from
the
actual
thing
itself,
we
may
be
able
to
get
rid
of
that.
B
Tools,
tools
is
is
more
things
that
aren't
tied
to
any
one
benchmark,
so
those
are
actually
like,
except
results.
Charge
and
and
post
results
are
just
things
that
are
used
on
the
benchmarking
machine,
one
to
generate
the
charts
one
to
post
results
from
where
the
the
benchmarks
run
to
a
machine
that
has
in
a
sequel,
database
yeah.
A
B
A
A
B
A
B
B
B
A
Okay,
well
I'll.
Try
to
proposals
for
that
too.
Serve
sounds
good.
I'll!
Try
to
make
sense
of
what's
in
here
and
how
to
organize
it.
That'll,
hopefully
come
in
as
part
of
the
agreeing
on
scripts
and
stuff
sounds
good.
Is
there
some
reason
that
she'll
scripts
are
preferred
over
other
languages,
or
is
that
is
that
a
deliberate
choice,
or
just
it
happens
that
it's.
A
B
Currently
we
run
on
a
single
machine,
but
the
the
other
thing
we
never
got
to
was
actually
updating
our
ansible
template,
which
creates
that
machine
yeah
and
we
do
have
three
of
them
now
and
and
if
we,
for
example,
one
to
run
have
the
test
be
able
to
run
across
any
of
our
CI
machines.
You
know
add
new
dependencies,
just
sort
of
expand
the
set
of
what's
needed
there.
So,
okay.
A
B
B
D
B
C
A
B
B
D
B
D
B
I
still
think
there
is
an
issue
somewhere
in
here.
I
just
can't
find
it
right
now,
where
we've
said
like
here
are
a
couple
extra
things
that
we've
manually
installed
because
people
said
hey
I
want
this
installed
and
we've
never
been
able
to
get
the
ansible
part.
So
we
have
ended
up
just
going
ahead
and
manually
installing
them
yeah.
A
B
B
A
B
C
There's
155
standardization
once
we
so
machine
setup
is
one
thing
and
then
parity
thinking
what
it
means
to
be
creating
one
framework
around
all
this
workhorse.
So
there's
I
see
that
there
is
going
to
be
a
script
for
workload
which
actually
knows
exactly
what
to
run
not
only
what
to
run
but
what
to
test,
but
it'll
be
something
on
the
top
level
which
will
take
care
of
everything
else.
C
For
example,
right
now
we
have
F
P
dot
Sh
for
no
DC
and
I
believe
for
anything,
but
rows
also,
so
we
don't
need
multiple
copies
of
them,
but
I
think
making
this
whole
changes.
We
should
just
talk
what
exactly
how
exactly
we
want
to
have
that
structure
what
it
looks
like
end
of
the
day.
We
can
just
talk
on
that
issue
and
we
can
finalize
the
process.
I
guess,
yeah.
B
D
C
Get
difficult
if
he,
let's
say
want
to
add
some
other
workload
which
are
which
is
not
our
property
but
from
someone
else's
and
it
might
not
be
I,
don't
know
how
exactly
it
would
work
if
we
make
a
copy
and
save
it
here.
It
might
be
good
idea
to
just
do
a
git
commit
a
git
clone,
but
just
just
a
thought,
I
think.
B
B
A
B
B
C
A
A
B
B
A
D
B
A
B
B
A
B
B
A
B
C
C
Just
my
okay
I
use
the
same
one.
Last
time
there
was
another
iteration
we
had
Bataclan
here
and
then
I
just
kind
of
used
just
for
this
purpose.
So
I
was
I
narrowed
it
down
to
fifteen
days
sometime
in
February
yeah
like
it's
difficult
to
kind
of
go,
then
the
dates
are
like
not
very
clear.
The
chickens,
the
Commission's
are
not
really
clear,
which
exactly
that's
what
I
need
to
do.
Next,
maybe
we
got
went
to
do
a
little
manual
work
there.
C
Yeah
yeah
I
was
talking
to
Michael
other
day
that
the
one
of
the
think
that
one
of
the
reason
is
when
the
chicken
come
it
gets
merged,
take
much
with
different
date
when
you
I,
think
created
them
and
not
not
when
actually
get
landed,
there's
some
kind
of
it,
and
especially
when
the
commit
has
a
multiple
checking.
Multiple
commits
a
PR
as
money
yeah.
A
C
B
C
B
C
B
C
A
B
B
Okay,
so
I
guess
we'll
just
wait
to
hear
from
you
on
that
one
it'll
be
interesting
to
see
what
caused
that
you
know.
Big
difference.
C
B
C
C
So
it
has
multiple
component,
which
we
typically
is.
Potentially
big.
Application
need
like
it
has
a
database,
but
in
this
case
it
is
a
my
sequel
database,
not
MongoDB,
which
is
already
installed
on
our
system.
So
we
decided
we
will
run
my
sequel
as
part
of
the
talker.
So
when
the
word
code
starts,
there
should
be
a
free
port
load
which
we
started
it
away
somewhere.
Yeah.
C
C
One
question,
of
course,
before
yep:
one
problem
it
goes
to,
it
does
not
work
with
a
master,
a
node,
so
it
expected
some
code
in
old
who
goes
to
a
checks.
What
version
of
node
are
you
running?
If
it
is
not
empty,
is
question
then
it
does
not
work.
I,
don't
know.
How
would
we
test
the
canary
build
and
I'm
a
stupid?
It
works
with
six
equals
with
eight
it
will.
It
was
Swift
and
I
checked
all
of
those
yeah.
It
does
not
work
with
four
and
does
not
with
master.
C
Is
it
just
because
of
that
chat
yeah
they
used
to
be
a
flag.
The
check,
the
old
version
equal
to
false.
If
you
pass
that,
while
running
coast,
you
have
to
use,
ignore
the
version
check,
but
now
I
that
has
been
changed
and
I'm
not
able
to
figure
it
out
through
the
code
where
the
check
is
anymore,
but
it
just
does
not
run
with
with
a
master
branch.
C
C
One
thing,
and
also
it
runs
in
two
modes:
return
is
run
in
a
production
mode
where
you
need
to
enter
installed
production
and
then
it's
also
run
in
development
mode,
but
the
performance
difference
is
around
10
to
20
percent
between
those
two
modes.
So
they're
few,
like
few
tricky
things
about
this
workload,
but
I
think
we
should,
as
part
of
this
exercise,
we
can
figure
it
out
me
figure
it
out
what
exactly
needs.
So
what
we'll
do
we
clone
the
workload?
B
Right
I
would
suggest
actually
doing
it
as
patches,
so
basically
I
clone
the
repo
and
then
apply
one
or
more
paths.
Just
that
way,
it's
clear
what's
being
changed
and
we've
done
that
in
other
places.
It's
you
always
hope
that,
then
you
don't
get
broken,
but
it
has.
It's
worked
out.
Okay
everywhere
else.
So
yes.
A
C
B
C
Okay,
so
that's
a
good
idea:
I
can't
incredible
that
also
okay
I'll
go
up
on
my
own.
That
goes
yeah,
we'll
see
what
I
want
to
say.
Yeah,
you
know
all
this
somebody's
time
right,
yeah
I
was
look
at
the
issue.
Someone
has
to
respond
to
it
and
we
have
if
he
is
definitely
want
to
use
the
master
on
that
canary
then
react.
We
are
going
to
have
that
issue.
Yeah.
C
A
So
he
gave
us
a
here
in
his
post
a
reasonably
detailed
description
of
what
he's
doing.
I
don't
think
that
what
he's
doing
looks
like
what
most
people
do
with
node,
which
is
I,
mean
incorporating
it
will
be
useful
for
him.
There
are
a
couple
other
groups
that
I
think
are
doing
system
software,
II
stuff
that
are
in
the
node
foundations
optional
case,
like
Harvard
VB,
is
kind
of
the
database
and
they're
their
application.
That's
where
I
oh
intensive
and
uses
the
file
system
a
lot.
A
So
you
know
this
is
send
us
a
guidance
for
that
systems.
Lawful,
use
case
I,
don't
think
it's
a
huge
priority
for
this
group,
although
obviously
the
more
the
merrier
in
terms
of
the
the
use
cases
we
can
cover.
But
you
know
we
talked
at
the
session
at
the
benchmarking
discussion
about
how
that
particular
workload
is
probably
not
as
important
as
say
getting
an
NPM
install
job
in
and
getting
a
micro
services
based
version
of
existing
benchmarks
in
both
of
which
I'm
more
realistic.
For
you
know,
practical
call,
not
more
realistic
but
more
popular
I.
D
B
D
B
B
A
A
I've
seen
yeah
I
have
a
ton
but
they're
doing
not
super
complicated.
So
like
they've
added
a
thumb
test
case
and
the
test
case
is
like
a
client
hits
your
endpoint
and
you
give
back
the
same.
You
know
50
byte,
JSON
string
every
time,
it's
it's
better
than
a
micro
benchmark
because
they
actually
set
up
a
server,
but
it's
sort
of
a
micro
benchmark
out
of
a
server
right.
Okay,
well,
I'm,
not
sure
it's
as
interesting
as
it
would
be
to
get
here
and
no
DC
right,
yeah.