►
From YouTube: CNCF SIG Network 2020-04-16
Description
CNCF SIG Network 2020-04-16
A
B
B
B
C
C
B
C
B
B
B
D
B
I
think
it
depends
on
your
perspective,
because
the
those
are
vines
that
are
kind
of
growing
over
the
outside
of
our
I'm
in
a
pool
house
and
on
one
hand,
is
really
nice
pretty.
On
the
other
hand,
the
vines
make
it
touch
that
squirrels
can
quickly
climb
up
the
side
of
the
house
and
get
on
on
the
roof.
Yeah
cool.
B
C
C
C
F
F
F
F
F
F
C
C
C
B
B
A
Just
a
maid
come
on,
I
am
bouncing
between
both
this
meeting
and
the
cig
meeting.
But
if
you
need
me,
please
ping
me
I'm,
sorry,
I
be
a
contributor.
B
E
We
got
this
huge,
beautiful
building
in
like
way
out
west
of
st.
Louis,
and
it
has
MasterCard
big,
a
logo
and
name
on
the
on
the
fight
like
a
hundred
feet
huge
right
now
they
changed
name
by
a
year
and
a
half
ago,
buddy
they're
not
gonna,
take
down
that
little
going
and
a
small
like
this
see
a
small
see,
but
any
reason
get
them
not
sure
how
much
they
spent
on
that.
But
it
shows
a
lot
of
what
a
MasterCard
dollars.
B
B
B
B
With
that,
we've
got
two
agenda
items
if
we
will
get
to
the
second
one,
irrespective
of
whether
or
not
someone
from
the
contour
project
is
representing,
not
that
we
have
anyone
just
yet,
but
the
first
agenda
item
up
is
a
project
presentation
for
consideration
of
incorporation
into
sandbox.
This
is
the
chaos,
chaos
smashed
project
construed
by
the
good
folks
at
pink
app,
and
so
we
have
some
of
those
those
representatives
on
the
call.
B
Today
you
can
go
to
this
Google
Doc
for
okay,
I
guess
this
is
hadn't
seen
this
yet,
but
that's
part
of
the
presentation
actually.
So
with
that,
let
me
introduce
a
couple
of
folks
that
are
with
the
project,
so
we
have
Calvin
Wang
who's
here
on
the
call
Calvin
you
and
and
Ed
will
be
presenting
today.
Is
that
right,
yes,
yeah,
okay,
very
good
I!
Don't
need
to
further
introduce
some
you
guys
or
the
project
I
should.
B
G
Camera
seated,
yes,
we
can
okay,
yeah
hi,
hello,
everyone.
This
is
and
I'm
the
co-founder
and
the
CTO
of.
If
you
haven't
heard
about
us
before,
we
are
making
the
fourth
Torrance
musical
database
court,
ITB
and
I.
Think
two
years
ago
we
donated
the
underlying
humanity,
storage
and
DB,
which
is
I
kami
to
CN,
CF
and
currently
hi
kami
is
the
in
computing,
never
forget,
and
we
also
the
major
maintainer
of
tech
TV
and
today,
I'm
and
I
was
very
glad
to
propose
another
exciting
project
which
is
can
smash.
G
It
is
our
internal
health
testing
framework
and
my
talk
will
have
will
have
two
parts.
The
first
half
I
will
introduce
the
project
itself,
including
the
technical
stuff
and
some
information
of
this
project
and
the
second
heart
I
will
hand
over
to
Kelvin
his
our
community
manager.
So
he
will
talk
about
stem
community
status
and
the
open
source
governance.
Things
like
this.
G
G
Apology
seems
like
this,
so
that
means
we
build
the
state
of
the
disability
system
to
write
the
unicast
is
the
deterministic
test
is
in
Cosmo
I.
Think
it's
not
very
hard
to
get
it
and
I
see.
Another
big
trend
is
kubernetes
is,
is
winning
I
have
already
won
and
crannies
is
eating
the
world.
It
is
so
easy
to
Bill.
You
know
it's
pretty
system.
On
top
of
this,
somehow
you
just
you
know,
write
some
yama
file
and.
G
Deployment
fare
over
orchestration
things
like
this
for
you
and
we
really
lower
the
bar
for
developers
to
build
the
complicated
system,
but
I
think
it's
good
in
a
good
way.
Yeah
and
before
I
talk
about
the
cash
match.
I
want
to
share
a
little
bit
of
the
history
of
this.
This
project,
before
I,
even
created,
pink
and
I,
watch
a
medium
from
in
from
foundation.
He
be
I,
think
some
of
some
those
some
of
you
may
have
heard
this
project
before
they
are
talking
about
a
testing
car
simulation
testing.
G
At
that
time
we
don't
even
have
a
house
engineering
at
the
time
and
it
is
Tuesday
2014,
and
this
is
a
great
talk
and
very
insightful
I
suggest
everyone
loved
you
have.
If
you
interesting
engineering
or
simulating
cast
you
can
you
can
watch
that
that
video
well?
The
record
is
that
to
use
the
simulator
to
you,
know,
intentionally,
create
failure
and
see
if
your
system
can
now
today.
I
still
think
this
is
the
only
way
to
cast
the
lobotomies
of
the
Swedish
system.
I
think.
G
Maybe
some
of
you
watch
the
show
Silicon
Valley,
maybe
in
the
Cavs
engineering
they
have
a
checkered.
You
know
what
I'm
talking
about
under
under
this
idea.
They
are
actually
a
lot
of
open
source
of
the
layers,
but
I
think
all
these
tools
are
separated
from
each
other
and
their
goals
are
totally
different.
For
example,
the
Jefferson
is,
you
know
the
most
famous
one
and
but
Jefferson
is
more
care
more
focused
about
checking
the
consistency
or
correctness
of
a
transactional
system
like
a
database
seems
like
this.
G
By
the
way
we
have
official
Jefferson
testing
report
and
another
interesting
I
interesting
project,
the
name
of
these
projects
namazu,
but
no
module
is
more
focused
on
messing
up
the
runtime
layer,
like
the
file
system
like
system
Linux
kernel
like
JVM
since
like
this,
but
most
of
these
tools
are
not
very
easy
to
use
and
you
have
to
be
an
expert
is
everything
to
just
make
them
work?
It
is
really
painful.
G
Believe
me
I
think
there
is
no
such
a
framework
to
chain
up
all
these
separated
tools
and
expose
the
human
readable,
programmable,
user-friendly
interface
to
to
the
developer
or
the
the
QA
engineer,
to
manage
the
different
type
of
failure
and
apply
to
your
past
work
load
or
what
in
production,
water
and
another
big
challenges.
Kubernetes
and
the
container
itself,
for
example,
in
pink
at
our
internal
testing
environment
is
all
is
on
super
large
kubernetes
employment
and
we
have
different
type
of
tests
and
different
versions
of
our
database.
G
G
What's
even
worse,
some
tools
are
not
even
worked
well
inside
the
container
like
TC.
The
traffic
control
is
true
and
IP
table
for
use
pieces
like
this.
We
also
have
to
make
sure
the
the
pots
in
the
same
missing
the
same
physical
names.
We
are
not
affected
by
each
other.
When
we
apply
some
cows
tests.
G
So
here's
cache
mash
that
the
better
of
the
the
the
predecessor
of
cash
mash
was
our
internal
database
testing
framework.
G
G
Let's,
let's
see
how
how
easy
it
is,
we
will
use
Sto
as
an
example,
you
know,
sto
is
a
popular
service.
Mesh
framework-
everyone
knows
is,
and
let's
say
we
want
to
inject
some
cows
into
its
control
panel
or
a
control
plane
or
on
data
data
playing
to
observe
the
performance
or
state
ability
or
security
or
anything
you
want
to
confer.
Let's
see
how
we
do
this.
For
example,
you
have
a
sto
cluster
deployed
in
your
communities.
G
Just
two
steps
and
all
you
need
is
just
to
write
a
llamo
file
like
this
define
what
kind
of
action
you
want
to
inject
to
your
to
your
existing
deployment.
For
example,
I
used
a
pod
failure,
cows
and
that's
after
the
is
two
clusters,
and
the
first
step
is
use
cubic
inch.
Aero
apply.
This
is
the
AMA
file
and
you
can
see
the
this
is
the
dashboard
of
of
the
application.
G
You
can
see
the
metrics
provided
by
by
sto
the
metrics
calm
during
this
time,
and
if
you
want
to
stop
the
the
cows,
you
just
use
the
kuba
tillage
and
you
are
set-
and
this
is
alive
again.
So
it's
really
very
easy
to
use.
You
don't
have
to.
You
know,
write
the
coast,
you
don't
have
to
manually
doing.
G
And
we
can
even
control
the
direction
and
frequency
of
the
the
different
kind
of
calves
like
here
is
an
example
to
random
randomly
kill
the
paths
every
one
one
minutes.
So
this
is
example.
We
are
testing
pandb,
you
know
tell
you,
we
have
a
self-healing
feature.
If
you
randomly
kill
the
stars
now
you
will,
you
know
you
have
the
high
availability,
so
you
will
be
back
automatically
and
we
can.
We
can
just
randomly
kill.
You
use
cash
match
to
make
sure
our
outer
forever
works.
G
Okay,
I
will
talk
about
some
key
features
and
technical
design
off
of
Afghans
mesh.
The
first
thing
is
that
we
follow
the
clown
native
design
and
we
defined
our.
You
know
the
different
kind
of
failure,
different
kind
of
cows
wires
TRD,
and
so
you
can
the
developer
or
the
QA
engineer
can
just
use
Yama
file
to
to
defy
the
the
cows
experiment,
and
we
all
we
use
the
slide
card
pattern
to
inject
the
the
the
cast
pod
to
your
application
to
no
application
part.
G
So
it
is
totally
transparent
to
the
application
layer
and
another
we
also
besides
the
randomly
killed
parts,
we
provide
many
different
kind
of
cows
type
like
network.
We
are
edgy,
so
we
can
simulate
the
network
delay
and
lose
the
network
packet
dub
or
the
Nano
partition.
Since,
like
this,
because
you
know
we
are
building
the
dissipated
database,
so
we
want
to
make
sure
we
can
make
a
mess
up
the
network,
so
we
can
see
the
the
assistance
view
as
well
and
besides
network
beside
network.
We
we
can.
G
You
know
also
inject
some
IO
error,
wire
things
and
also
we
can
simulate
the
clock
screw.
It
is
very
useful
when
you
are
developing
a
database
or
developing
in
the
transaction
or
system,
because
we
use
timestamp
all
the
time
so
and
another
interesting
feature.
Another
interesting
cast
type
is
Colonel
thought
we
use
a
BPF
no
to
readily.
You
know
inject
some
failure
to
your
Cisco,
so
yeah.
It
is
very
interesting.
So
this
is
the
the
architecture,
the
architecture
of
the
cache.
A
G
Project,
basically,
it
is
just
like
I
said,
is
very,
very
clear
and
if
we
use
the
CRD
to
define
intent-
and
we
use
the
sidecar
to
inject
specific
constituted
paths
and
we
create
a
customized
controller
manager
to
manage
the
housing,
it's
it's
really
straightforward
and
currently
I
think
cows.
Mash
is
not
doing
enough
on
observability.
G
You
know
the
effects
of
cows
has,
you
know
currently
need
to
be
observed
through
the
metrics
of
the
application
being
test.
So
we
we
are
considering
giving
the
cows
mash
is
on
dashboard
to
improve
the
you
know,
the
officer
mobility.
In
addition
to
make
it
easier
to
use,
we
are,
we
are
considering
make
a
service
public
service
on
public
cloud
like
gke
or
or
or
UK
s,
so
so
that
the
user
don't
even
need
to
write
the
llamo
file.
G
If
you
have
already
deployed
your
application
or
micro
service
on
U,
KS
or
gke,
you
can
click
some
buttons
on
our
service
to
to
tray
some
cows
to
test
your
your,
your
your
application,
and
the
third
thing
is,
that
is
about
automatic
validation,
automatic
validation.
What
is
automatic
meditation?
You
know
we
inject
some
cows
to
application
and
okay
allocation.
May
you
know
the
the
and
you
have
to
check
the
the
output
of
your
application
to
check
if
your
system
going
well
when
you
apply
the
cows
and
we
have
already
implemented
automatic
validation
in
our
database.
G
D
D
We
will
talk
about
Padma
bada
it's
later
and
we
have
kept
a
steady
pace
of
commits
so
far
we
got
270
I,
think
it's
more
than
that
commits
in
internal
and
we
don't.
We
do
not
have
a
website
yet,
but
we
we
are
fully
aware
of
the
importance
of
documentation,
so
we
put
together
our
documentation
and
for
the
community
using
the
github
wiki,
and
we
have
a
monthly
meeting
monthly
committee
meeting
in
the
planning
phase
and
I
think
he
should
be
ready
soon.
D
D
They
are
really
flattered
and
hambo
humbled
about
the
Speck
nation
and
attention,
and
so
with
that
we
are
seeing
already
seen
increasing
adoptions
from
the
community,
as
ed
already
mentioned,
pink
hobbies,
the
first
and
the
perhaps
it's
also
the
heaviest
user
of
cosmas,
and
we
have
other
adopters
from
multiple
industries.
For
example,
we
have
dailymotion,
which
is
the
YouTube
in
France
we
have
sello
and
it's
a
digital
payment
solution
provider
and
inspires
a
very
big
cloud.
Computing
computing
service,
providing
China
and
the
Shogun
moto
is
a
intelligence
augmentation.
D
In
China
yeah,
we
are
pretty
exciting
about
that
and
the
on
there
the
contributors
are
from
these
adopters
yeah.
So
here
come
to
the
question:
why?
Why
do
we
want
to
donate
as
much
to
CSC,
yeah
I?
Think
it's
quite
obvious
to
us.
First
of
all,
it's
I
think
the
values
and
missions
of
cosmas
is
in
very,
very
good
alignment.
D
This,
since
f,
as
and
already
introduced,
cosmas,
has
adopted
a
very
climatic
design
so
that
it
could
easily
integrates
with
the
Cinzia
ecosystem
and
its
harbour
in
initial,
and
the
primary
intention
forecast
match
to
be
serving
as
a
universal
curse
testing
platform
for
the
district's
discrete
distributed
system
on
kubernetes,
so
that
a
cream
enable
resilience
is
observability.
I
think
this
also
these
two,
our
suited
to
the
most
important
qualities
of
the
community
projects,
and
also
our
ultimate
goal
is
that
is
that
we
would
like
can
smash
to
be
either
cosmic
chaos
engineering
standards
on
cloud.
D
As
a
newsroom
for
collaboration
for
a
very
young
project,
black
cast,
mash
I
think
it's
the
only
way
I
mean
to
to
collaborate
with
other
culinary
projects.
It's
the
only
way
for
us
to.
You
know
to
get
better
together
and
yeah.
Obviously
yeah.
We
really
like
to
help
more
and
more
developers
yeah
you
that
we're
coming
to
the
end
of
the
print
as
you
thank
you
all
and
if
you
like,.
B
D
B
That
great
guys
this
is
there
were
there
were
even
a
few
jokes
included,
which
was
set
up,
which
was
nice
kudos
on
a
quick
ramp
of
the
project.
By
the
way,
your
marketing
team
will
have
to
teach
me
some
tricks.
That's.
B
So
a
couple
of
it
all
start
with
a
couple
of
questions
now
invite
the
community
as
well
to
ask
if
they
have
any
so
well.
A
couple
of
things
I
might
have
gotten
so
I'll
say
this
and
we
have
the
majority
of
the
time
of
our
call
to
probably
talk
about
chaos
mess.
The
second
item
on
the
agenda
can
be
covered.
If
we
don't
get
to
it,
we
can
cover
it
offline.
So
I
just
wanted
to
do
some
housekeeping
that
minuto.
B
G
You're
doing
this
slice
I'm
talking
about
you
know,
the
problem
is
some
tours,
for
example
like
if
you
manually
change
the
IP
table
of
your
host
physical
knows,
it
may
affect
all
the
network
of
your.
You
know
the
all
the
passing
in
the
you
know.
If
you
can
our
so,
if
you
want
to
simulate
the
the
network
or.
G
B
Makes
sense
yes
or
the
level
of
granularity
of
chaos
in
the
past
I
think
yes,
monkey
that
had
been
a
bit
more
focused
in
ayahs
or
VM
land,
you
don't
really
one
of
those
life
is
hoping
or
I
was
well.
That
would
just
I
know.
This
is
a
difficult
one.
Ur
can
be,
but
it
would
be
discussing
other
projects
and
kind
of
contrasting
them
think
things
like:
there's
chaos
or
others.
G
B
Yes,
litmus
chaos
in
particular,
they
may
be
an
appetite
larger
than
the
actual
experiments
to
have
the
know
that
there's
a
chaos
hub
sori
indicates
or
would
hint
that
it
would
be
a
healthy
set
of
experiments.
I,
don't
know
that
that's
necessarily
the
case
today.
They
do
have
experiments,
for
you
know
in
a
was
speaking
with
my
signal
network
had
on
the
expanse
guys
use
that
as
a
support
for
today
are
in
large
part.
Why?
B
G
F
F
G
Yeah,
actually
I
think
I
think
really
posted
a
link
of
how
how
you
created
the
first
cows
experiments
your
your
for
your
exciting
application
and
if
you
want
to
create
a
new
kinds
of
new
kind
of
cows,
failure
or
like
you,
create
a
new
new
new
new
type
of
network
partition
or
something
things
like
this.
You
have
to
I
think
we
don't
have
a
API
layer
to
to
help
you
to
create
a
new
kind
of
cows
easily.
But
it's
pretty
I'm
saying
it
is
very
easy
to
apply
existing
cows
to
your
your
application.
G
D
And
Undertown
the
maintainer
yes,
now
we
have
document
about
how
to
add
a
new
type
of
chaos
into
the
cosmetic,
and
now
we
already
have
almost
tore
a
table
of
type
of
chaos.
I
introduced
it
I
canõt
buta
and
the
either
it's
a
little
harder
to
again
to
add
a
new
curse
curtain
type
of
chaos
that
you
just
met.
B
G
G
C
Hi
Lee:
this
is
Queenie
from
pick
up
I'm,
also
a
community
manager,
so
I
want
to
shy
me
regarding
how
we
differentiate
ourselves
from
other
chaos.
Engineering
projects-
we
didn't
include
this
in
this
slide
because
we
do
not
want
to
you
know,
compare
to
be
so
competitive,
and
but
we
do
have
some
information
in
place
and
I
would
like
to
introduce
them.
If
you,
if
you
think
that's
okay,.
B
C
Yes,
just
because
we
are
speaking
from
our
perspective
and
might
not
be
subjective,
so
yeah,
please
just
let
me
give
some
illustration
about
the
difference.
So,
basically,
Kia
smash
provides
rate
and
fine
grained
injection
capabilities
and
is
more
optimized
for
complex
applications
such
as
providing
network
partitioning
file
system,
IO,
interference,
time,
interference,
Curnow
injection
et
cetera.
These
are
very
important
of
for
testing,
complex
distributed
systems,
and
so
kyo
smash
injection
capability
is
completely
provided
by
resolve,
which
can
be
controlled
independently.
Littmus
itself
does
not
provide
many
Chios
capabilities
so
far.
C
Some
injection
capabilities
need
to
rely
on
other
kills
tools
such
as
network
delay
and
Puma
Puma
battery,
and
another
point
is
that
chaos
smash
is
more
cost
effective,
costly
face
and
easier
to
use.
Creating
a
chaos.
Experiment
in
chios
smash
only
needs
one
Yama
file,
and
there
are
only
a
few
fields
that
need
to
be
set
for
users.
It
also
provides
a
flexible
application.
Selector
for
lattimer's
ochio's
requires
two
configuration
files.
Where
is
our
BAC
de
llamo
and
experiment?
C
Experiment
llamo
many
of
the
configurations
and
are
complicated
and
part
of
the
reason
you
that
therapy
of
implementation
relies
on
creating
jobs
in
the
corresponding
namespace.
In
addition,
Tia
smash
will
also
provide
a
way
of
front
and
management's
of
the
experiment
in
the
future.
So
can
smash,
provides
chaos.
Dashboards
will
provide
the
dashboard,
a
combination
of
kills
in
event
and
application
monitoring,
which
will
make
it
easier
to
observe
the
effects
of
chaos.
Experiments
currently
only
supported
for
Taibbi,
but
a
universal
version
will
be
available
soon,
so
so.
C
The
last
point
is
that
Keo
smash
and
litmus
architectures
are
very
different.
Soquel
smash
deploys
a
demon
sets
on
each
node.
This
architecture
can
do
more
things
we
believe,
such
as
supporting
Nature
Network
injection
capabilities.
Chrono
injection
cover
blaze,
et
cetera,
I
hope
that
that's
again,
that's
from
our
perspective
and
just
ticket
with
them
with
some
sort
of
grain
yeah.
B
We
could
ask
for
I
think
in
terms
oh
I'm,
a
couple
other
quick
questions.
I'm
high
level
I
know
that
we,
since
we
had
the
time
we
got
a
chance
to
ask
a
bit
more
about
the
workings
of
the
project,
but
just
some
around
the
project
quickly,
so
kubernetes
native,
but
kubernetes
only
is
that
that
would
be
an
accurate
statement
for.
B
G
F
Have
a
quick
question:
/
suggestion,
so
I
thought
the
architecture
was
described
is
that
you
have
a
local
node
worker
node
Digimon.
Have
you
considered
being
able
to
test
the
infrastructure
itself,
so
I
know
that
this
primarily
is
targeting
the
application,
but,
for
example,
I
would
probably
like
to
test.
My
government
is
infrastructure
before
I
start
deploying
my
application
so
that
I
know
that
it
scales.
Well,
it's
react
properly
to
missing
note
or
to
latencies.
I
might
be
wanting
to
take
in
my
CNI.
You
know
performing
well.
G
B
Canon
noted
that
the
end
user
community
is
the
chaos,
engineering
and
testing
as
a
topic
and
as
a
set
of
tools,
probably
very
interesting,
the
end
user
community,
as
indicated
by
the
stars
through
projects-
probably
we
did,
let
me
I
mean
at
this
point-
maybe
harass
you
guys
about
this
same
question
again.
So
I
mentioned
litmus
chaos
as
like
top
of
mind.
For
me,
there
are
other.
There
are
other
projects
out
there.
G
B
B
B
B
B
Only
four
months
in
the
open
source
I
do
have
questions.
Are
the
diversity
of
my
contribute,
the
governance?
That's
there
today
that
sort
of
all
of
the
maintainer
czar
pink
folks,
and
that's
that's
a
very
organic
thing.
That's
how
projects
generally
get
started
that
makes
sense,
but
so
especially
kind
of
a
cig
review
which
part
of
that
I'll
use.
The
word
diligence
lightly.
Part
of
that
examination.
If
you
will
will
be
that
yeah
I.
Could
this
help
with
bit
just
this
whole
process?
B
B
I'm
also
a
big
chunk
right
here,
because
that
process
has
recently
evolved
and
I.
Don't
know
that
it's
entirely
solidified
I'm
on
paper.
The
process
is
essentially
exactly
what
you
guys
are
doing,
which
is
you
come
to
the
appropriate
City.
Give
a
presentation
that
say
grace.
Is
it
up
help
solicit
that
interest?
It's
certainly
I.
Don't
think
that
it
is
I,
think
it
now
that
you've
presented
here
I,
don't
think
that
it
is
inappropriately
generally.
What
would
happen
is
we
would
do
a
review.
B
I'd
say
generally,
this
is
a
relative,
a
new
process
that
review
and
that
sort
of
more
concise
presentation
and
perspective
coming
out
of
the
sig
is
a
good
artifact
for
other
TOC
members
to
quickly
look
at
and
gauge
their
interest
engage
their
willingness
to
support
so
that
that
artifact
I
think
will
be
helpful
to
you
guys
and
don't
think
that
if
people
read
those
other
conversations,
you
might
find
that
some
of
the
TOC-
oh
hey,
I,
haven't
heard
from
the
singing
yeah.
Let
me
let
me
hear
what
they
have
to
say
or
hey.
B
B
On
your
part,
no
other
than
to
be
so,
we
will
move
on
with
this
thing,
we'll
put
together
a
short
document
as
a
review
of
the
project
and
kind
of
some
recommendations
that
we
might
have
about
the
project
getting
there,
we
may
have
some
additional
questions
and
that
would
be
okay
to
produce.
Otherwise
there
no
got
it.