►
From YouTube: WebAssembly Filters Meeting (July 12th, 2021)
Description
WebAssembly Filters Meeting - July 12th, 2021
Join the community at https://layer5.io/community
Find Layer5 on:
GitHub: https://github.com/layer5io
Twitter: https://twitter.com/layer5
LinkedIn: https://www.linkedin.com/company/layer5
Docker Hub: https://hub.docker.com/u/layer5/
A
Okay,
hey
great,
so
hey
we're
it's
about
nine!
It's
about
four!
After
let's
get
rolling
it's
monday
july
12th,
which
reminds
me
solarwinds
is
going
to
announce
a
stock
split
today
I
should
probably
go
look
at
the
stock
market.
A
Random
side
note.
So
today
is
the
12th.
It's
the
web
assembly
filters
meeting
one
of
many
different
meetings
that
we
have
in
layer,
5
community
focused
on
service
meshes
and
related
tech.
It
won't
be
too
long
before
there
will
be
other
people
who
are
hosting
this
call
and
that
might
have
started
today.
A
Pranav
is
working
to
backfill,
a
missing
speaker,
a
missing
presenter,
and
so
today,
as
a
reminder,
since
last
time
we
met
is
that
last
time
we
met.
We
said
that
there
are
any
number
of
us
here
who
stand,
who
are
still
learning
in
this
area,
which
makes
a
lot
of
sense
to
me,
because
the
vast
majority
of
the
world
is
still
learning
these
newer
technologies
and
so
webassembly
as
part
of
that
core
focus
is
certainly
an
emerging
technology,
it
web
assembly
as
a
concept
and
as
as
a
wow
gosh.
A
What
would
you
as
a
as
a
write
once
run
anywhere
framework
technology
is
relatively
new
unto
itself,
but
even
newer
with
respect
to
service
meshes.
This
last
cube
con,
which
was,
I
don't
know
a
couple
of
months
ago,
or
so
had
its
first
web
assembly
day,
just
like
a
single
track
conference
that
was
associated
to
kubecon,
in
which
people
presented
that
same
thing
will
be
happening.
I'm
at
the
upcoming
kubecon
so
actually
opportunities
for
your
presentations
here
to
potentially
be
submitted
there
submitted
there.
A
Good.
Well
on
that,
so
we
had
a
number
of
topics
lined
up
with
respect
to
knowledge
sharing.
Mostly
people
were
going
to
go
off
and
try
to
digest
a
few
pieces
of
technology
and
share
what
they
could
in
over
the
last
week.
So
leading
us
off,
I
think,
is
samir.
A
Just
before
samir
jumps
in
real
quick
check,
there
are.
A
B
Yeah,
hello,
hello,
hello,
yeah,
my
name
is,
I
just
joined,
I
guess
yesterday
and
I'm
yeah,
so
I'm
pretty
new
here
and
I
just
right
away
wanted
to
get
started
with
learning.
So
I
was
suggested.
A
B
A
Yeah,
nice,
nice
good
to
have
you
anybody
else
that
hasn't
gotten
a
chance
to
say
hi
on
this
call.
That
would
lie
to
you.
C
A
A
I
might
have
pointed
out
some
faqs
about
like
about
how
someone
might
venture
into
that
type
of
an
internship,
and
I
wanted
to
quickly
point
out
that
there
were
two
other
individuals
that
were
on
that
same
community
call
talking
about
the
same
internship
and
guess
what
you're
the
last
one
standing
right
now
or
you're,
the
one
that's
showing
up
so
like
thanks,
yeah
yeah,
it's
good!
It's
good,
all
right,
fair
enough
without
further
ado.
A
A
A
F
E
D
F
Requests
so
please
do
yeah.
D
A
D
Yeah
so-
and
this
slide
is
already
in
our
community
drive,
so
anybody
can
follow
along
can
be
used.
D
Okay,
so
this
is
about
the
web
assembly
from
the
lens
of
service
message,
so
ideally
I'll
be
talking
through
how
exactly
webassembly
works
specifically
for
the
service
message,
and
this
is
a
very
fundamental
introduction
that
is
gonna
like
you
know,
help
people
to
understand
why
we
have
started
up
this
meeting
in
layer.
Five,
okay,
so
feel
free
to
like
your
edit
or
adding
more
info
as
you
as
you
go,
or
as
you
understand,
as
you
learn
more
so
yeah
before
this.
D
Let
me
tell
you,
like
you,
know
I'll
just
take
some
reference
which
lee
has
already
presented,
so
our
id
infrastructure
has
been
drastically
changed
by
last
few
years
so
wherein
we
have
moved
from
like
a
monolithic
to
micro
services
and
in
the
deployment
strategies
we
have
moved
from
vm
to
container
and
then
container
orchestration
with
kubernetes
and
then
finally,
service
mesh
is
the
one.
So
what
exactly
service
messes
are
doing
right
now?
It
is
enabling
the
service
fast
solutions,
which
is
only
for
the
micro
service
based
architecture
where
it
feeds
more
accurately.
D
But
yes,
so
you
can
use
service
mess
anywhere
wherever
you
want
to
connect
between
the
services
or
you
can
kind
of
decoupling
services
from
one
another,
and
still
you
want
to
have
reliable
connection
between
them
and
few
of
the
architecture.
Very
new
need,
like
you
know,
secure
connectivity
and
maybe
l3
l4
or
l7
l5,
like
you
know,
sort
of
connections
and
all,
and
that
is
where,
when
we
talk
about
service
message,
most
of
the
services
leverage,
the
functionality
of
one
or
the
other
proxy,
which
enables
them
to
effectively
work
in
a
microservice
environment.
D
So
when
you
talk
about,
like
you
know,
wave
assembly,
this
is
again
a
very
buzzword
right
now,
which
is
like
you
know,
very
portable
bytecode
format,
which
is
kind
of
a
like.
You
know
very
native
to
your
or,
I
would
say,
lower
level
language
or
your
machine
code,
and
it
is
very
much
effective
and
very
fast
because
it
is
it
is,
you
know
very
near
to
your
machine
or
machine
understanding.
D
So
that
is
where
you
can
say
it
is
in
your
native
speed,
so
why
we
are
using
this
wave
assembly
to
achieve
high
speed,
because
nowadays
everybody
wants
their
application
to
be
much
more
faster
and,
like
you
know,
it
is
as
a
language
which
adds
in
into
html
css
and
javascript,
but
it
is
not
kind
of
replacement
for
them.
All
those
things
go
hand
by
hand
so
which,
like
you
know
you
you
can
see.
D
Wave
assembly
provides
a
better
performance
for
all
the
scripts
which
has
been
like
mastering
in
the
we
have
delivery
networks,
or
you
can
say
in
the
client
side,
okay.
So
actually,
this
has
been
like.
You
know,
tested
in
many
more
browsers
which
have
come
recommendation
from
the
w3c,
which
is
ideally
the
original
web
assembly,
and
it
was
on
2019.
D
Okay,
as
I
told
this
was
first
innovated
to
deliver,
or
you
can
say
the
for
used
in
client
side.
But
again
now
it
is
everywhere,
so
you
can
use
it
in
the
server
side
as
well
as
in
the
client
side.
So
just
just
I
was
fascinated
how
it
was
got
birth.
So
that
is
why
you
can
see.
So
this
is
the
birth
of
web
assembly
and
it
was
announced
on
2015.
D
Has
been
ideally
like,
you
know,
tested
and
it
it
has
got
a
tremendous
kind
of
popularity
after
that
and
then
after
in
march
2017,
the
minimum
viable
product
was
declared
to
be
used
by
other
people
and
in
2017,
of
course,
it
got
safari
like
you
know,
release
wherein
it
supports
safari
and
2018
again.
The
group
probably,
is
the
public
like
working
traps
for
the
core
specification,
javascript
interfaces
and
web
api.
D
So
this
is
how
wave
assembly
got
introduced
and
got
word
and
now
why
you
are
using,
as
I
have
already
told
it
is
usable
it
require.
It
requires
a
high
level
code
and
very
simple.
It
runs
like
a
docker,
so
some
people
say
if
this
wave
assembly
has
been
if
it
could
have
been
originated
much
before
docker.
They
couldn't
have
gone
for
docker
to
run
their
application.
So
this
is
that
important
for
the
current
modern
applications
and
again
as
it
told
it,
is
language
agnostics.
D
That
means
you
can
write
in
any
language
and
you
can
just
make
your
own.
There
are
tools
wherein
you
can
convert
your
language
to
the
web
assembly
binaries.
So
it
doesn't
really
understand
any
language.
That
is
why
it
is
not
dependent
upon
any
language.
So,
as
far
as
the
statistics
goes,
it
supports
more
than
30
different
languages
and
again
it
is
a
very
low
level
or
low
language
web
assembly.
So
it
is
very
quick
to
execute
and
resembles
a
native
experience.
So
performance
is
very
good
in
fact
so
platform
agnostic
again.
D
So
this
web
assembly,
though
it
was
first
originated
or
was
originated,
to
use
in
a
client
site,
but
now
it
can
use
it
can
be
used
anywhere,
so
you
can
use
it
in
ios,
linux,
mac
and
windows
anywhere
and
in
your
server
too.
So
that
is
why
we
are
using
web
assembly
here.
So
this
is
a
generic
definition
or
you
can
say
generic
use
case
for
we
have
assembly,
but
let's
just
go
into
why
we
are
using
it
in
service
mesh.
D
So
as
of
now,
if
you
have
followed
this
slide,
so
you
you
might
have
understood
what
is
the
like,
you
know
extent
of
performance
or
you
can
see
like
you
know
how
easy
to
integrate
with
any
other
languages.
So
that
is
where
every
service
message.
I
told
uses
the
proxies,
and
these
proxies
are
mostly
written
in
either
c
or
rust.
Any
of
those,
and
it
is
very
easy
for
this
proxy
to
adapt
this
particular
ysm
library.
In
fact
it
is
for
every
language.
D
It
is
very
easy
to
add
up
this
kind
of
awesome
libraries
and
is
like
you
know.
Proxies
plays
a
very
important
role
in
the
service
message.
D
It
is
also
a
add-on
on
the
proxies
wherein
you
can
make
your
service
mess,
or
your
micro
services
more
effective
or,
like
you
know,
more
deliverable
to
any
of
the
workloads
okay.
So
what
exactly
this,
like
you
know,
envoy,
provides
for
a
service
mess.
It
provides
the
traffic
management
filtering
of
http,
request
policy
enforcement,
and
many
more
so.
These
are
the
like.
D
It
is
natively
provided
by
envoy
with
some
sdks
and
you
can
write
your
own
language
or
according
to
your
business,
need
so
for
an
example,
as
I
was
needing,
some
of
the
like,
you
know,
manipulation
in
the
jwt
token,
which
I
get
from
one
of
the
services
and
pass
it
over
to
another
one,
and
that
is
where
we
have
used
the
the
jwt
manipulator
awesome
binary
to
do
that,
so
this
service
must
provide
a
lot
of
thing.
So
do
your,
like
you
know,
do
the
work,
how
you
want.
D
Okay,
so
that
is
how
the
awesome
has
evolved.
So
this
is
now
in
evolving
state
in
service
message.
So
if
you
go
to
istio,
it
is
still
on
a
like.
You
know
alpha
release
that
so,
which
is
you
can
contribute
to.
You
can
learn
from
them
and
solo
dot.
Io
is
the
most
like.
You
know,
active
participator,
providing
this
particular
proxy
or
some
support
to
isd.
D
Okay.
So
how
exactly
awesome
provide
the
extensibility,
so
generic
application
binary
interface?
Is
there
so
we're
in?
Like
you
know,
it
is
very
easy
for
you
to
write
something
in
your
own
language,
convert
it
to
binaries,
which
can
be
kind
of
a
dot
awesome
file,
and
you
can
again
enforce
them
to
any
of
the
services
or
any
of
the
proxies.
D
So
when
I
talk
about
the
proxy,
it
is
mostly
the
envoy
proxy,
which
is
used
by
most
of
the
service
messages.
As
a
like,
you
know,
proxy
vm
or
you
can
say
the
proxy
side
card.
So,
whenever
a
request
comes
to
any
of
this
side,
car,
this
particular
filter
chain,
which
you
have
written
in
awesome
format
will
be
executed.
D
And
then,
as
you
have
written
all
of
your
logic,
you
will
get
a
fine
grained
request
so
which
is
more
secure
and
again,
as
I
told
you
can
write
it
in
over
30
languages
and
it
has
a
abstraction
to
allow
for
the
custom
run
time
to
be
used
and
it
takes
very
less
memory
so
which
is
like
you
know,
very
good
advantages,
because
it
doesn't
provide
any
overhead
to
your
micro
services
in
any
manner.
So
it
is
very
native
to
the
involved,
because
there
are
the
application.
Binary
interface
has
been
developed.
D
Okay,
so
where
exactly
the
proxy
awesome
fits
in
in
the
invoice
or
any
of
the
service
messages.
So
if
so,
so,
if
you
can
see
the
like,
you
know
proxy
was
from
here
so
which
is
like
you
know.
This
is
a
istio
service
mess
wherein
you
can
see
the
various
proxy
component,
which
is
being
there
as
a
side
car.
D
D
So,
according
to
this
particular
explanation,
which
I
gave
that
is
elaborated
here,
let's
say
this
is
a
listener
wherein
your
request
come
in
then
this
is
a
regular
envoy,
like
you
know,
command
or
you
can
say
the
regular
invoice
filter
which
is
marked
in
yellow
and
whenever
you
give
a
awesome
binary
to
it,
it
will
override
those
things.
So
for
for
this
particular
scenario,
we
have
the
http
connection
manager
wherein
we
are
pushing
the
web
assembly
okay.
So
when
you
put
the
web
assembly,
then
the
request
will
follow
from
here
then
go
here.
D
Then,
whenever
it
fetches
like
now
finds
a
web
assembly
module,
then
it
follow
the
route
along.
So
it
goes
here
and
here
and
here
and
finally
we
have
the
android
router
from
there.
It
will
go
back
to
the
end
point
of
this
manager
and
from
there
it
will
go
to
the
service.
So
this
is
what
happens
inside
and
why
proxy?
D
When
you,
when
a
request
comes
from
here,
so
starting
from
the
listener
from
to
the
click,
no
request
comes
out
of
the
envoy
and
enters
into
the
service,
and
there
are
few
fun
facts
like
it's:
it's
just
a
if
you
know
about
and
why
this
is
also
very
much
known.
So
you
can
do
only
the
patch
operation,
because
there
is
already
some
logic
written
now.
You
want
to
overwrite
that
which
is
a
patch
operation,
and
that
is
why,
so,
how
do
you
ideally
build
your
proxy
version
for
the
envoy
this?
D
There
is
a
like
no
cli
tool,
which
is
you
cannot
use
it
for
your
production
workload,
of
course,
but
you
can
play
around.
There
is
a
watch
me
which
is
from
the
solo,
and
you
can
just
go
ahead
and
download
the
cli,
and
there
is
a
very
much
yeah
so
sameer.
What
about?
What
about.
D
D
Okay,
so
this
is
ideally
in
our
layer,
5
dot
io.
Where
is
our
all
the
awesome
filter
has
been
like
you
know
listed,
so
what
you
can
ideally
do
is
you
can
go
here,
go
to
the
src
folder,
wherein
you
can
find
a
leave
dot,
rs,
okay!
So
in
this
leap.rs
whatever
you
see,
it
is
ideally
a
rust
code
and
it
get
when
it
is
get
built.
It
produces
a
library,
sorry,
the
binary,
which
is
a
dot
awesome,
and
this
awesome
you
can
in
like
you
know,
one.
D
D
Okay,
so
this
awesome
again
can
configure
with
your
service
mesh
so
how
to
do
it,
you
can
run
it
as
a
kind
of
one
second,
okay,
so
once
once
you
write
it,
then
you
compile
it.
Then
you
get
awesome
binary.
Then
you
have
to
inject
it
to
your
proxy.
So
how
to
do
it?
You
can
do
it
via
the
manifest
files,
which
is
again
part
of
this.
If
you
can
go
to
awesome
filter,
you
can
see
the.
D
F
I
guess
the
manifests
and
that
level
of
detail
they
can
slowly
just
absorb
the.
D
Get
started
with
rust,
then
it
is
here
which
is
a
like
sdk
from
proxy
awesome
and
you
can
write
in
and
there
are
list
of
invoice
filter
sto
so
where
exactly
you
can
do
the
manipulation.
So
these
are
the
filters
here.
So
you
can
see.
That
is
where
we
can
do
the
manipulations
yeah.
So
that's
all
okay.
So
thank
you.
G
Over
to
you,
can
I
ask
a
question
so
yes,
yes,
so
the
question
is
so
we
use
showed
about
a
compiler
right.
I
think
it's
a
compiler
from
solo.I
o.
So
similarly
there's
one
from
measurey
as
well.
F
F
For
example,
if
you
create
what
do
you
say,
a
wasm
filter
in
rust,
you
would
have
various
packages
like
wasm
pack,
wasm,
opt
and
so
on
which
actually
build
your
vosm
binary
from
the
rust
code,
just
like
compiling
a
c
c
plus
plus
project
similar
to
that
it
will
compile
your
rust
code
to
the
wasm
binary
now
wasn't
binary,
as
samir
mentioned,
is
quite
close
to
machine
machine
language,
but
not
quite
so.
There
is
still
a
bit
of
interpretation
to
be
done.
F
It
can
be
compared
to
java
byte
code,
and
that
was
me
cli
tool.
It
merely
calls
the
respective
what
is
a
tool
chain
to
compile
it,
and
on
top
of
that,
it
provides
you
interfaces
to.
F
I
guess
you
know
upload
it
to
your
own
registry
and
all
now
where
mesherie
comes
in,
is
that
regardless
of
your
sort
of
deployment,
and
so
it
once
okay?
So
what
we
are
doing
here
is
we're
creating
a
sort
of
interface
where
you
can
take
those
binaries
that
you
create,
so
this
can
be
either
manually
compiling
it
in
whatever
language
you
want
or
through
the
was
me
tool
or
whatever
poison
of
your
choice,
and
you
could
upload
it
and
you
could
then
apply
it
to
any
aspect
of
it.
F
So
I'll
be
demoing
image
hub
very
quickly,
so
you
guys
can
see,
live
how
the
filter
actually
works.
So
I
guess
I'll
explain
that
as
we
go
ahead.
So
if
anyone
has
any
other
questions,
no,
if
anyone
like,
if
there
are
no
other
questions,
I
should
continue,
but
please
do
go
ahead
in
case
you
do.
F
Okay,
no
one
going
once
twice
and
okay,
I'm
gonna
take
over
now,
so,
okay
hi
guys,
I'm
pranav
and
I'll
be
presenting
the
okay.
If
this
allows
me
to
share
my
screen-
okay,
yes,
it
does
so
I'll
be
presenting
the
imagehub
application,
which
has
been
there
even
before
I
joined
mystery.
So
this
is
a
dinosaur,
but
it's
a
very
good
dinosaur.
F
So
just
a
brief
recap
of
what
you
know
what's
going
on
behind
the
scenes,
essentially,
you
will
have
a
certain
service
mesh
over
here.
This
image
depicts
the
console
architecture
and
you
will
have
a
service
mesh
running
here.
This
is
the
control
plane.
I
guess,
and
some
of
these
service
meshes
use
something
called
envoy
proxies
now.
These
are
traditionally
used
because
without
these
proxy
site
cars,
normally
your
traffic
would
have
to
go
through
quote.
F
Unquote,
centralized
you
know
a
registry
in
order
to
talk
to
another
pod,
so
you
have
pod
a
or
foo
over
here,
and
you
have
part
bar
over
here.
Without
these
proxy
site
cars,
you
would
have
to
be
routed
through
a
central
thing
up
here,
somewhere,
it's
not
depicted
in
the
diagram,
but
with
the
proxy
side
cars
your
pods
can
directly
talk
to
each
other.
F
What
that
allows
you
to
do
is
it
allows
you
to
have
these
wasm
filters
which
I'll
be
getting
to
in
a
short
bit,
and
these
filters
will
basically
receive
whatever
traffic
in
or
out
according
to
what
you've
configured
them
and
that
code
that
you've
typed
up
each
packet
or
each
request
it
receives
your
code,
will
be
run
over
it
and
you
can
do
a
lot
of
crazy
stuff
with
that
then.
F
So
in
our
example
imagehub
so
one
misconception
is
imagehub
is
a
hosting
sort
of
project,
but
or
like
you
know,
it's
the
it's
where
all
these
awesome
images
are
hosted.
That
is
not
quite
true
image
hub,
as
you
can
see
in
the
description,
is
just
a
sample
application
to
demonstrate
the
capabilities
of
version
filters,
so
in
our
case
image
hub
it
samples
something
like
docker
hub,
so
over
here,
I've
logged
in
here
I'm
going
to
keep.
F
You
know,
grabbing
your
attention
and
getting
you
to
look
at
this
request
response
site
tab
over
here,
but
basically
through
the
measuring
management
ui.
Currently
I'm
running
these
processes
locally.
So
this
is
this
is
my
measuring
server.
This
is
my
measuring
adapter
and
there
are
a
few
custom
changes
there
are.
You
know
it's
still
a
work
in
progress
and
it
was
broken.
You
know
until
10
minutes
into
the
meeting
so
yay
for
it
working
currently.
F
If
anything
goes
wrong,
don't
blame
me:
okay,
where's,
chromia,
so
getting
back
to
chrome,
I
basically
used
mescheri
to
deploy
istio.
So
I
just
clicked
on
instio
and
you
have
this
managed
service
mesh
lifecycle,
so
I
deployed
steer
here
and
for
the
sample
application,
which
is
imagehub.
I
click
over
here
and
I
have
deployed
imagehub
now
traditionally
in
such
an
application,
you
would
have
a
sign
up
and
login.
So
let
me
just
sign
up
with
test.
Okay,
I've
already
created
an
account,
so
I'm
just
going
to
skip
it.
F
So
it's
going
to
be
test
and
test
and
yeah
successfully
logged
in
so
the
this
is
supposed
to
mimic.
You
know
how
you
have
various
images
on
docker
hub
and
you
can
hit
download
to
download
those
docker
images
similar
to
that
you
know
we
have
another
app
trying
to
mimic
that,
but
when
you
hit
download
you
only
get
this
pop-up
saying
image:
pull
request
in
progress.
Now
the
people
who
have
deployed
this
app,
they
don't
want
to
give
unlimited
downloads
for
each
user.
F
So,
right
now,
when
I
hit
sign
up,
I
can
select
the
plan.
Yes,
it's
flawed,
that
it
doesn't
ask
for
your
credit
card,
but
I
don't
think
that's
that
important.
So
the
enterprise
allows
thousand
image
pulls
quote
unquote:
image
pulse
tml
is
100
personal
allows
10,
and
when
you
go
ahead,
okay,
this
this
should
okay,
it
says
unlimited.
A
F
I've
limited
it
to
a
thousand,
so
this
gives
you
all
the
information
you
want
and
you
can
upgrade
the
plans,
but
this
is
not
something
we
have
to
concern
ourselves
with
going
to
the
hub.
Ideally,
you
should
be
on
able
to
click
this
download
button.
Since
I
have
signed
up
with
the
enterprise
account,
I
should
be
able
to
click
this
thousand.
If
I
had
done
personal,
I
should
be
able
to
do
it
10
times
and
then
a
warning
pop-up
would
come
in
now.
F
All
of
what
I've
told
you
traditionally,
if
you
know,
if
I
ask
you
to
create
this
application,
you
would
hard
code
the
logic
of
okay,
fine,
I'm
going
to
what
do
you
say?
Get
I'm
gonna.
Have
my
user
make
a
request?
He
is
going
to
send
me
some
authorization
token.
So
this
authorization
token
is
actually
a
jwt,
and
this
contains
your
username,
along
with
the
plan
that
you
have
so
now.
F
What
your
application
might
do
traditionally
is
it'll,
look
at
the
username
and
say:
okay,
fine
in
the
database,
okay,
yeah,
here's
the
username,
let's
say
pranav-
is
the
username.
So,
okay,
I
found
prana.
How
many
times
has
he
pulled?
Okay,
he's
pulled
nine
times
ten
times,
so
I'm
going
to
block
him
the
next
time.
F
So
that
kind
of
logic
you
would
be
coding
into
your
application
traditionally
now
the
magic
of
wasm
filters
is
what,
if
I
told
you,
you
do
not
need
to
include
that
code
in
your
application
at
all
and
you
could
let
a
wasm
filter
instead
dictate
that
logic.
So
you
have
your
rate
limit,
filter
and
you
have.
This
is
the
rust
code
for
it
and
over
here
after
running
the
compile,
you
know
compiled
sorry
the
wasm
tool
chain
build
tool
chain.
You
get
this
binary
over
here
and
this
binary.
F
You
know
using
certain
scripts.
You
then,
hopefully
we
do
like
soon
enough.
We
make
it
easier
for
you
to
apply
it,
but
currently
we
have
to
use
yaml
files
and
apply
them
manually.
So
once
you
do
apply
them,
I'm
going
to
show
how
measurey
can
automate
it
in
its
current
state.
So
I
have.
What
do
you
say
my
give
me
a
second
yeah.
I
have
my
filter
ready
to
go.
I
just
okay.
I
have
to
set
the
namespace.
F
I
apologize.
So
I
set
the
namespace
I
deploy
and
voila
we
got
our
you
know,
filter
up
and
running.
A
tiny
thing
that
I
need
to
do
right
now
is
just
give
me
a
second
see
if
you
can
get
one.
F
Yeah,
so
you
can
see
now
new
pods
have
spawned,
and
these
this
sorry,
this
pod
will
make
sure
that
whatever
traffic
comes
in
it
will,
you
know,
take
a
look
at
the
jot
filter.
I
at
the
jot
that
the
user
would
send
and
it
will
then
keep
help
you
keep
track
of
the
traffic
and
the
user
and
the
number
of
pulls
the
user
has
made.
So
why
don't
we
try
that
out?
So
if
you
can
see
currently
the
request
is
looks
pretty
normal.
F
The
response
also
looks
pretty
normal
and
you
can
see
it's
on
envoy
whatever,
and
now,
let's
hit
the
sign
up
again,
let
me
just
hit
reload
okay,
I'm
going
to
create
new
user
and
I'm
going
to
keep
personal
for
the
sake
of
the
demo
and
I'm
going
to
sign
up
and
of
course
it
fails.
Yeah.
So
remember
how
I
told
if
things
break,
don't
blame
me.
It's
broken.
E
E
F
F
F
F
Technical
difficulties,
basically,
it
was
working
right
before
I
switched
on
screen
share
and
what
should
happen
is
that
logic
should
be.
You
know
ideally
executed
on
the
wasm
filter
itself
and
I'll
just
show
the
okay,
no
showing
the
code.
It
might
take
a
long
time
so
I'll.
I
I
guess,
for
the
next
meeting,
I'll
save
this
and
I'll
get
it
working.
Okay.
A
All
right
I
mean
one
you
that
was
you
weren't
intending
to
give
the
demo
so
so
having
to
scramble
just
before
is
totally
totally
reasonable
and
then
two
actually,
the
the
main
point
of
I
mean
pranav
hit
on
all
the
main
points
of
what
of
the
purpose
of
the
demo,
which
is
I
mean
like.
Had
he
finished
it
off?
You
would
have
just
seen
either
something
being
rate
limited
or
not.
A
Rate
limited,
like
you'd,
have
just
seen
that
the
it
it
happening
in
action
which
would
have
just
been
the
ability
to
refresh
the
screen
or
not
like
click
the
button
or
not,
but
the
rest
of
what
he
had
said
was
really
the
important
bits
of
helping
people
grasp
the
concept
that
to
to
joshua's
recent
questions.
So
joshua
had
two.
A
A
Why
should
I
spend
time
coming
to
understand
how
this
type
of
thing
works
and,
what's
all
possible
and,
like
you
know,
what's
really
compelling
about
this,
and
some
of
it
goes
back
to
what
samir
was
saying
about
the
characteristics
of
webassembly
super
small,
quite
secure
portable,
except
like
a
lot
of
the
same
characteristics
that
you
would
like
refer
to
docker
as
helping
promote
and
and
part
of
the
reason
that
it
came
to
popularity
is
for
some
of
the
same
ways
that
you
would
characterize
web
assembly.
A
A
You
know
running
that
that
unity
game
in
your
browser,
like
that's
kind
of
ridiculous
and
so
okay,
fine
taking
that
compute
power
and
those
same
characteristics
over
to
build
the
in
network
intelligence
in
the
network
like
okay,
interesting,
so
so
to
your
question
like
okay,
what's
what
can
be
done
with
it
and
it's
like
well,
this
sample
application
that
pranav
is
showing
is.
Is
he
saying
well,
hey,
look
like
imagine
you're
the
product
owner
for
by
the
way
we
I
should
put
up
a.
A
A
What
the
hell
is
that,
well,
it's
actually
kind
of
what
pranav
I
mean
like
what
pranav
is
showing,
which
is,
or
so
it's
a
follow-on
to
that,
which
is
to
say
today
if
you're,
using
a
subs
as
an
example
of
power,
a
capability
you
might
get
out
of
a
web
assembly
is
is
basically
implementing
business
level
logic.
A
So
he
so
in
this
demo
app
it's
showing
that
anyone
who
can
click
a
mouse
like-
and
I
don't
mean
this
in
a
derogatory
way
in
any
sense
but
like
if
a
product
owner
product
manager
wanted
to
change
the
pricing
model
for
or
the
the
the
rate
the
like
switch
around
for
given
subscription
tiers,
how
much
activity
a
given
account
can
have
based
on
that
tier
they
can
go
over
and
mouse
click
their
way
to
changing
that
and
in
moments
have
that
whole
model
affected,
as
opposed
to
in
months
and
months
and
months
of
development.
A
You
know
development
team
sprints
to
go
change
that
logic
work
with
the
operations
team
to
go,
implement
that
maybe
that
logic
in
somewhere
else,
that's
quite
powerful
and
get
exposed
to
them
right
away.
It's
pretty
interesting
if
they
want
to
look
at
well.
Is
that
the
right
decision?
What
telemetry,
what
information
metrics
do?
They
have
to
understand
like
where
that
inflection
point
is
between
like,
if
someone's
really
gonna
buy
the
next
higher
level
of
their
subscription
of
their
pricing
model?
A
They'd
have
all
the
telemetry
right
there
to
say
well,
yeah
like
it
looks
like
it
would,
because
there's
any
any
number
of
these
people
and
we
can
tell
who
those
people
are
because
we
can
see
their
jot
or
we
can
see
their
session
and,
like
they're,
doing
these
behaviors,
and
so
okay,
like
this
becomes
a
product
in
this
sense
like
in
this
example.
A
This
becomes
well
a
development
teams,
tool,
a
service
delivery
teams,
tool
a
product,
manager's
tool
to
to
be
able
to
do
that
in
the
network,
with
the
existing
infrastructure
through
configuration
they're,
just
configuring,
the
way
that
their
product
is
being
sold,
whoa,
that's
pretty
cool
and
the
developers
go.
Thank
god!
Thank
you,
praise
god,
because
I
didn't
want
to
work
in
that
boring
crap
of
refactoring
the
subscription
model
based
on
a
whim
based
on
you
thinking.
It
was
going
to
work
better
and
you're
coming
back
two
months
later.
A
To
tell
me
you
want
to
change
the
model
again,
it's
like
you
know,
so
it's
it's
a
win
for
everyone.
I
think
to
round
off
the
answer
to
your
question,
which
was
a
little
more
generic
and
it
was
like
hey.
What's
what
else
is
the
value?
So
that's
interesting
great!
You
can
implement
some
business
logic,
okay.
Well,
what
else
is
there
valuable
about
web
assembly
and
the
way
that
this
works
inside
of
envoy
based
data
planes
at
the
moment?
Not
that
other
proxies
don't
have
plug-in
models
they
do.
A
If
you
take
a
look
at
there's
a
model
with
an
ambassador
for
plugins
that
are
written
in
golang,
there's
a
model
in
nginx
for
plugins
that
are
written
in
c
c
or
lisp,
or
something
else
there's
a
model
in
traffic
that
for
plugins
that
are
written
in
golang.
A
So
this
notion
that
you
can
extend
the
network
and
its
intelligence,
and
these
proxies
to
something
more
customized
is
pretty
interesting
and
then
there's
another
character
component
of
that
is
that
well,
okay,
yeah,
but
you
can
write
that
in
a
regular
in
a
native
filter
so
like
to
joshua's
question
like
what's
what's
better
about
webassembly
than
just
a
native
filter,
I
can
go
write
the
custom
native
filters.
Do
that
thing,
natively
to
onboard
it's
like
yeah,
and
that
can
be
the
right
answer.
That
could
be
the
answer.
A
The
reasons
that
someone
might
use
web
assembly
filter
instead
of
a
native
one
they
might
they
might
not,
is
actually
in
part
the
study
of
our
folk.
It's
like
it's
not
based
on
dynamicism,
it's
like
hey.
If
you
have
a
running
envoy
and
great,
it's
deployed
it's
running
and
you're
like
yeah.
We
need
to
add
this
filter.
We
need
to
have
a
filter
that
we've
written
to
do
this.
Okay,
we
can
build
it.
A
Natively
rebuild
the
envoy
image,
et
cetera,
roll
out
that
to
our
infrastructure
and
great,
like
we'll
bump,
you
know
we'll
rev
the
info
like
we'll
bump,
we'll
rod
a
new
version,
because
you
gotta
get
a
new
proxy
in
there
that
has
that
filter.
It's
like
well,
okay,
but
wouldn't
it
be
nice
if
we
could
just
dynamically
leave
the
same
proxy
in
place,
but
just
load
unload
this
sucker
and
then
like
also
dynamically
load
in
oh,
let's
load
up
three
and
maybe
based
on
something
that's
happening
real
time
like
oh,
this
thing's
happening
real
time.
A
Okay,
we
need
another
one
inserted
in
a
chain
behind
the
the
first
one
like:
oh
damn,
all
right
that
level
of
dynamicism
the
ability
to
unload
and
load
either
based
on
signals
or
or
some
human
like.
Oh
that's,
pretty!
Okay,
that's
intriguing!
Like
yeah,
there's
a
little
bit
of
a
virtualization
tax,
that's
going
to
go
on
between
something,
that's
native
versus
something!
That's
running
in
this
virtual
stack
machine
running
inside
of
a
web
assembly.
It's
like
well.
What
is
that
overhead?
A
Well,
I
don't
know,
go
watch
the
presentations
that
were
given
over
here
and
we
need
to
fix
this
domain
because
the
it's
currently
it
needs
internally
down,
but
there's
slides
that
talk
about
literally
exactly
comparing.
What's
the
difference
between
this
discreetly
studying
the
effects
of
individual
individual
traffic
control
functions,
so
what
is
the
overhead
of
native
versus
wason?
A
Well,
good?
God,
that's
a
complicated
question
answer.
Actually
that's
why
you
need
some
tooling
like
meshri
to
say
I
don't
know.
What's
your
custom
filter,
I
don't
know
you
know
you
test
it.
Okay,
here's
measuring
you
can
use
to
sit
there
and
test
and
run
those
tests
to
see
what
the
difference
is
great
and
then
what
if
I
wanna
like?
Okay,
great,
I
understand
the
difference
now
between
what
that
overhead
is,
but
is
that
normal?
Like
am
I
doing
it
really?
Well?
Am
I
doing
it
kind
of
piss
poorly?
A
Like
I
don't
know,
hey
samir
how's
yours
running
well
like,
wouldn't
it
be
nice
if
we
had
a
common
framework
to
exchange
and
talk
about
that.
So
that's
what
this
is
about
is
a
common
framework
for
characterizing
the
performance
and
understanding
that
and
exchanging
it,
which
is
part
of
the
the
need
in
this
ecosystem.
Another
need
is
in
part.
What
joshua's
asking
now,
which
is
you
know,
is
there
a
standard
across
various
proxies
for
writing?
These
extensions
and
filters,
in
other
words,
are
the
filters
portable
across
different
proxies
in
the
mesh.
Well,
that's
awesome
question.
A
A
That
doesn't
mean
that
they're,
like
this
community
in
specific
and
these
like
that's
something
very
much
desired
for
our
part,
because,
hopefully
just
like
web
assembly,
the
contributors
here
are
able
to
write
once
and
run
many
or
run
anywhere,
so
it
is
the
in.
So
there
isn't
a
standard.
A
F
To
interrupt,
but
I
got
it
working
now.
E
F
So
yeah,
okay
cool,
so
I
managed
to
create
that
personal
account
that
I
was
talking
about.
So
let's
go
ahead
and
log
in
and
yep.
So
if
I
click
this
now,
you'll
see
image
pull
request
in
progress
and
perfect,
so
you
can
see
over
here
the
rate
limit.
Actually
this
is
supposed
to
represent
the
number
of
times
you've
pulled
and
what's
happening.
Right
now
is
that
if
you
can
look
at
this
diagram,
okay,
I
apologize
if
my
laptop's
lagging
it's
running
everything
at
once,
so.
F
Okay,
so
over
here
this
basically
uses
a
wasm
vm,
which
you
can
think
of
like
a
jvm
and
it's
storing
the
variables
like
okay,
I
see
this
user
and
it's
keeping
track
of
all
of
that
in
memory,
and
so
when
it
sees
you
the
jot
coming
in
from
the
user
test,
and
it
will
just
increment
the
count
and
it
will
check
the
plan
type
as
well
and
because
it
knows
what
sort
of
plans
are
there,
you
know
you
pass
it
through
a
configuration.
F
It
also
knows
the
limit
for
that
particular
plan.
So
it
looks
at
the
plan
type
and
looks
at
the
number
of
requests
the
dude
has
made.
If
he's
made
too
many,
it
prevents
this
traffic
from
being
forwarded
and
it
just
blocks
then,
and
there
and
you
get
a
message
saying
hey:
this
is
you've
gone
over
your
limit
and
if
it
isn't
it
forwards,
it
and
the
container,
then
you
know
process
it
further
and
a
benefit
of
this
is
that
your
containers
don't
waste
time.
F
F
Oh
god,
it's
lagging
yeah.
So
you
can
see.
The
thing
has
gone
up.
I'm
just
gonna
click
this
a
couple
of
times.
F
Okay
early,
just
anything
you
would
like
to
say,
while
I
click
this
10
times,
no
okay,.
A
Cool
just
that
portability.
There
isn't
the
standard
with
respect
to
portability,
but
in
practice
the
image
hub
as
an
example
and
it's
filter.
You
can
see
them
deployed
across
two
different
meshes
and
so.
A
So
the
answer
I
was
giving
earlier
about
no
is
like
no
with
respect
to
a
standard
so
but
in
practice
like
given
the
commonality
of
an
envoy
based
data
plane
and
its
application,
binary
interface-
you
it
may,
where
we're
banking
on
part
of
that
being
true,
that
you'll
be
able
to
use
these
across
on
the
whitebase
database.
A
H
Know
yes,
lee?
Yes,
I'm
there.
A
A
So
so,
by
the
way
for
everyone
on
the
call,
just
like
it's
inevitable,
that
nearby
is
going
to
present
something
pretty
interesting
and
juicy
and
we're
going
to
ask
questions
in
about
20
minutes
is
going
to
fly.
You
know,
15
minutes
is
going
to
fly
by
or
something,
and
so
I
just
wanted
to
say
if
you've
got
another
call
and
you
need
to
drop
very
nice
to
see
everybody
today
hang
out.
A
If
you
can
I'm
a
nearby,
maybe
we
should
just
maybe
with
since
I
said
that
and
since
it's
like
one
minute
left
like
we
should
just
say
hey.
If
you
want
to
stay
around,
please
do
I'm
gonna.
I
know
I've
got
questions
and
things
to
learn
so,
but
if
you
need
to
go
see
you
guys
next,
we
see
you
all
next
week.
At
the
same
time,.
H
So,
okay,
guys
I'll
just
shorten
it
I'll,
not
take
much
of
your
time.
So
basically,
you
asked
me
to
prepare
a
demo
on
and
void
traffic
metering,
but
the
actual
demo
I'm
facing
some
issues
with
docker
and
all
those
stuff.
So
I
could
not
complete
it
so
probably
definitely
next
week
I'll
complete
it.
So
I'm
just
par
the
presentation
part
I'm
showing
today.
So
today,
I'm
talking
about
the
envoy
traffic
metering
and
first
of
all,
we
talk
about
what
is
ny.
H
So
ny
is
a
high
performance
proxy,
so
it
can
be
applied
at
different
layers
and
mechanisms
are
there
and
it
is
like
it
is
nothing
but
a
service.
Miss
implementation
and
istio
is
one
of
them,
and
so
definitely
we
use
this
to
very
much
for
the
subject:
mate,
implementation
and
basically,
what
we
do
is
the
traffic
handling
part.
We
actually
can
be
done
using
some
sort
of
filters
and
that
filters
or
a
chain
of
filters-
and
this
allows
a
couple
of
things.
H
One
is
access,
control
and
one
is
like
the
rate
limiter
and
also
it
can
help
you
to
have
some
billing
based
on
the
who
is
authorized
to
do
what
like,
based
on
what
subscription
he
has
taken.
So
all
all
those
things
can
be
taken
care,
and
then
I
talk
about
what
is
traffic
mirroring,
so
traffic
mirroring
or
shadowing
is
a
concept
by
which
we
mirror
the
copy
of
the
live
traffic
to
the
from
a
local
code
base
to
the
some
remote
code
base.
H
So
the
the
the
mirroring
the
basic
example
of
a
metering
service
is
mail,
archival
service.
So
basically
your
people,
like
art
providers,
like
barracuda
or
even
microsoft.
What
they
do
is
whenever
you
have
that
email
archival
service
there
is
that
option
to
mirror
your
archived
mail
for
an
user
to
remote
data
center,
so
that
remote
setter
data
center
is
generally
on
the
cloud
and
you
can
mirror
either
the
whole
of
the
mains
for
the
user
or
you
can
just
mirror
a
part
of
it.
H
So
that
is
what
we
do
as
part
of
the
enviro
traffic
mirroring
and
so,
as
I
said
that
it
can
happen
for
the
entire
traffic
or
the
selected
portion
of
the
traffic.
H
So
this
is
how
we
can
limit
the
rate
of
the
traffic
okay,
and
we
can
do
multiple
things
with
the
traffic
using
this
concept,
and
so
what
we
will
do
is
we
will
basically
apply
a
rule
to
push
to
a
portion
of
traffic
to
either
the
mirror
entire
thing
or
to
mirror
only
part
of
the
traffic
and
basically,
what
we
do
is
we
create
two
instances
of
the
http
pin
service
on
the
remote
server
and
that's
how
we
use
kubernetes
to
deploy,
have
the
service
running
parallely
and
then
we
whatever
wasn't
filter
we
create
so
we
use
istio
to
deploy
and
then
it
can
be
help.
H
H
So
basically,
what
it
does
is
we
always
need
to
keep
updating
us
and
the
drawback
as,
as
is
clear,
is
like
we
always
have
to
maintain
our
version
of
the
envoy
and
continuously
be
in
sync
with
official
distribution,
and
so
this
is
the
problem.
So
what
the
basic
best
solution
is
to
dynamically
load
new
filters
into
the
invoice
proxy
at
runtime,
so
this
will
be
adding
new
features,
new
capabilities,
and
we
can
write
our
user
defined
filters
to
achieve
this
kind
of
traffic.
H
I
can
say
traffic
monitoring
or
direct
relation.
A
lot
of
things
can
be
done,
okay,
and
so
anyway,
I
think
everyone
has
covered
this,
so
I
I
my
small
thing
would
be
basically
the
filters
whatever
we
are
creating
the
vessel
filter,
this
acts
as
a
proxy
and
whenever
normally
the
communication
happens
between
this
entity
and
this
entity
through
the
cube
kubernetes.
H
But
when
there
is
a
networking
issue,
then
our
system
can
go
for
a
toss,
so
that
is
where
the
proxy
helps
into
communica
helps
communication
between
this
entity
and
this
entity
directly
not
via
this
normal
flow,
what
happens
in
kubernetes
and
that's
that
that
is
the
place
where
we
are
going
to
work
for
the
vassal
filters,
and
I
think
people
have
taken
care
of
this.
So
I
just
put
this
so
I
don't
want
to
go
detail
into
this,
and
I
think
this
is
also
explained
so
the
end.
H
What
what
is
the
incoming
request
and
what
exactly
we
do
so
we
basically
once
we
apply
the
filters
we
route,
the
traffic
to
the
service.
So
basically
what
has
happened?
It's
a
filtered
traffic.
You
can
say
it
is
not
the
exact
traffic
it
wants.
It's
originally
coming
so
based
on
our
requirements.
We
can
apply
filters
and
then
route
it
so
that
that's
the
main
thing
and
okay.
So
I
think
this
is
also
covered,
so
I
think
and
okay
basic
strategy.
H
So
I
I
was
just
going
through
some
code
code
base,
so
I
have
not
been
able
to
do
much
on
the
code
right
now,
but
I
just
saw
some
code
base
where
in
they
have
written
a
filter,
vessel
filter
using
c
plus
plus,
and
they
were
talking
about
what
exactly
we
need
is
like
two
classes.
One
is
a
root
class
and
one
is
a
context
class.
So
the
root
context
is
like
like
based
on
the
like
the
root.
H
It
is
like
the
main
user
or
the
admin
user
and
the
lifetime
is
throughout
the
vm
and
the
filter
can
be
used
for
all
those
interactions
between
the
initial
setup
and
again,
the
interactions
between
the
proxy
and
the
different
entities
and
on
configure
can
be
invoked
by
the
root
context
to
like,
basically
based
on
the
configurations,
what
is
done
and
then
and
it
again.
H
If
the
plugin
contains
one
or
more
assembly,
filters
is
expecting
a
configuration
to
be
passed
in
invoice
proxy
again,
we
can
override
this
functionality
using
some
override
functions,
and-
and
so
I
have
not
done
much
so
I
was
just
going
through
the
code
and
probably
in
the
next
session.
Probably
I
will
just
give
a
brief
of
what
exactly
I'm
trying
to
do,
and
I
probably
will
give
more
context
on
that.
So
I
had
given.
I
have
put
some
code
base,
but
probably
I
am
not
fully
prepared.
H
I
just
prepared
a
part
of
it.
So
let
me
just
present
this
next
time.
Okay-
and
I
think
this-
these
all
things
are
already
covered,
so
I
have
already
taken
this
vasmi
and
I
have
installed
it
and
I
have
actually
created
a
small
filter
and
then
I
was
trying
to
push
it,
but
I
was
getting
some
issue
with
the
docker
and
that
is
where
I
am
struck.
So
probably
I
have
to
sync
up
with
pranav
and
see
how
to
deploy
that
stuff.
H
So
once
my
entire
setup
is
up,
I
think
I
would
be
more
comfortable
to
do
much
more
things.
So
that's
it
from
me
early.
I
think
I
have
not
taken
much
of
your
time.
Thank
you.
A
Nice,
that's
great,
that's,
that's.
Definitely,
that's
definitely
progress
there's
another
another
another
component
to
this
is
is
digesting
I'm
envoys
capabilities
itself,
like
its
existing
capabilities
to
perform
traffic
mirroring
and
then
yeah,
and
then
reflecting
on.
I
think
like
having
understood
that
one
of
the
questions
that
we
will
no
doubt
ask
ourselves,
is:
what's
it
missing
or
like?
A
Where
doesn't
this
work
or
do
people
need
more
controls
over
this,
like
part
of
what
you
were
saying
early
on
in
the
presentation
is,
I
envoy
can
mirror
the
entire
thing
or
a
subset
of
that
traffic.
It's
like
well,
how
granular
can
that
be?
Is
it
can
it
be
just
just
a
near
bonds
traffic,
because
that
that
could
be
a
very
common
use
case?
It's
like
hey!
I
want
to.
A
We
want
to
do
a
canary
as
an
example,
but
only
for
an
airborne's
traffic,
because
it's
okay
to
mess
his
up
because
he's
testing
and
developing
anyway,
and
so
or
you
know
like
so
yeah
so
digesting
what
envoy
is
capable
of
already
leads
us
to.
A
You
know
hey
what
what
you
know
should
if,
if
you
do
like
yeah
boy,
I'm
trying
to
be
concise
and
having
a
hard
time
I'll
say
that
one
of
our
youths
get
like
a
great
use
case
here,
and
this
has
to
deal
like
it's,
not
whether
it's
traffic
mirroring
or
just
traffic
capturing
kind
of
one,
in
the
same
in
this
case,
like
to
the
extent
that
we
can
capture
some
packets
and
whether
or
not
they're
actively
played
you
know,
replicated
and
sent
somewhere
else
actively
or
if
they're,
just
persisted
to
a
data
store
to
a
pcapp
file
or
whatever
I'll
make
a
quick
analogy
to
the
thing
I
think
would
be
really
helpful
to
a
lot
of
people
running
operating
services
on
a
mesh
it'll
be
one
of
those
compelling
things
where
people
say:
oh
yeah.
A
Well,
if
you
just
had
a
mesh
deployed
you
could
you
could
have
trouble
shot
your
issue
because
of
this
capability,
and
the
capability
is
something
like
this.
If
you
make
an
analogy
to
distributed
tracing
where
people
are
there's
a
lot
of
troubleshooting
and
a
lot
of
helpful
information
that
comes
from
traces,
you
look
at
each
of
those
spans
you're
like
and
you're.
Like,
oh
that's
a
massive
span
at
such
a
slow
service.
We
should
figure
out
that
sql
query.
A
That's
taking
19
years
to
come
back,
like
you
know
so
distributed
traces,
are
really
helpful
for
like
white
box
monitoring
or
like
deep,
deep
insights
and
okay,
but
there's
a
cost
to
to
like
to
tracing
those
like
hey
how
often
you're
going
to
sample
those
suckers
like
every
single
time,
maybe
if
you're
in
a
regulated
industry
or
something,
but
it's
like
yeah
go
ahead
and
like
go
purchase
may
as
well
go
work
on
a
purchase
order
now
for
those
hard
drives
you're
gonna
need
to
have
to
you
know
like
for
the
cpus
that
you
need
to
have
to
do
it's
like
okay.
A
Well,
we
actually
don't
you
know
for
a
lot
of
people,
it's
like
no!
No,
like
we
occasionally
want
to
take
a
look
at
those
distributed.
Traces
and
it's
like
really
important
for
the
lat,
but
like
the
trailing
last
three
days,
it's
pretty
important
for
us
to
see
those
but
like
beyond
that
we
could
have.
We
could
aggregate
our
samples
like
we
don't
need
the
high
fidelity
we
can
have.
A
You
know
lower
granularity
going
back
in
time
and
okay
great,
but
but
there
isn't,
there
are
some
vendors
in
the
distributed
tracing
space
that
handle
those
types
of
use
cases
and
make
that
easier
to
do.
They
also
have
worked
on
things
where,
oh,
no,
why
don't
you
just
sample,
keep
keep
those
samples
only
as
and
when
there's
an
error,
500
or
or
some
only
as
and
when
the
latency
gets
so
bad
that,
like
there's
something
going
on,
we
should
start
to
investigate
like
do
it
intelligently?
A
A
Can
you
play
back
that
same
thing
that
caused
the
trace
to
be
huge
or
that
that
is?
Can
you
play
back
this
same
like
as
you
go
to
troubleshoot
your
trace?
It's
like
great.
We
saw
that
issue
happening,
but
it's
not
happening
now.
It's
like
well,
oh,
I
can't
repro
the
issue.
It's
like,
oh
no.
Actually,
you
want
to
repro
just
click.
The
repro
button
click
the
replay
button,
like
so
great.
A
A
Great,
that's
like
really
close
to
the
answer
that
we
would
need
to
have
if
we're
going
to
capture
all
that
traffic
like
again
you're
talking
about
like
churning
some
cpu
churning
some
some
disc,
like
let's
do
that
intelligently,
and
so
that,
like
the
two
examples
are
kind
of
interrelated
and
and
to
your
point,
yeah
you
you
really
you
it
would
be
compelling
for
measuring
and
tooling
like
it
would
be
compelling
for
people
to
come
and
use
it.
If
it
has
that
kind
of
a
capability.
H
So
you
need
to
integrate
with
multiple
things,
so
I
think,
like
a
big
data,
kind
of
might
be
a
database
or
something.
A
Or
yeah,
if
the
system
gets
intelligent
enough,
then
actually
it
just
captures.
It
only
captures
traffic
on
an
anomaly
in
in
it
it
maybe
it's
like
perpetually
capturing,
but
most
of
it
is
going
to
the
bit
bucket
after
five
minutes
up
until
it
sees
at
500
and
then
it's
like.
Oh
wait.
Save
that
save
the
10
seconds.
A
Thanks
yeah,
I
know
thank
you
yeah.
This
is
great.
This
is
really
good.
Yeah,
I'm
pumped
to
see
everybody
learning
and
sharing.
So
anybody
have
anything
else
before
we
clock
out.
I
Okay,
not
much
exactly
just
that.
I
looked
into
the
point
that
is
in
the
meeting
notes
the
custom
styling
for
rjsf.
I
So
there
are
a
few
customizations
that
we
can
add
to
the
rdsf
components
like
custom
fields,
custom,
widgets
and
custom
templates.
So
there
is
a
array
field
that
we
can
use
to
change
the
way
the
form
looks,
and
there
is
an
object
field
that
we
can
change
to
decrease
the
nested
cards.
Basically,
so,
if
gaurav
is
on
the
call,
maybe
he
can
take
a
look
at
them
or
maybe
both
of
us
can
take
a
look
at
that
after
this.
A
Well,
hey
we'll
work
on
the
agenda
for
next
week.
In
slack,
some
of
some
of
us
already
have
homework.
So
we
are,
we
already
know.