►
Description
https://go.dok.community/slack
https://dok.community/
With:
Wey Gu - Developer Advocate, Vesoft Inc.
Bart Farrell - Head of Community, Data on Kubernetes Community
ABSTRACT OF THE TALK
- Nebula Graph Demystified
- Graph on **K8s**
- **Know-How** of Graph Database
BIO
Open Source believer, builder, singer and Graph Magic advocate(to scale the magic to help wizards)
https://siwei.io/en/about/
KEY TAKE-AWAYS FROM THE TALK
- Why& How graph db would help real life problems
- Arch and Design the Nebula Graph Database and the Intro. of the Project
- Implementation of the Nebula K8s operator
- Graph Database Know How in actions on:
- Building an intelligent Voice Robot
- MetaData governance and Lineage
- Building a corporation shareholding relations system
A
A
As
usual,
we've
got
a
live
a
panel
next
week
now
databases
on
kubernetes
going
to
be
featuring
various
community
members
looking
at
different
databases
how
they
integrate
the
different
challenges
that
we
face
with
that,
and
also,
of
course,
keeping
in
mind
the
element
of
storage,
because
that's
super
important
too,
if
we're
talking
about
databases,
he's
got
the
link
to
that
right.
There.
Also
our
cfp
for
kubecon,
our
collocated
event
on
may
16th
is
open.
A
The
submissions
are
coming
in
we're
getting
some
really
really
good
stuff,
very
happy
to
see
increased
quality
and
the
people
are
listening
to
us
when
we
say
no
vendor
pitches,
no
vendor
pitches,
no
vendor
pitches.
Keep
it
focused
on
running
safely.
Workloads
on
kubernetes
you've
got
all
the
information
you
would
want
about
that
there.
Our
speaker
today
has
also
submitted
a
talk
for
kubecon,
so
hopefully
get
that
in.
A
Anyway,
you
know
how
we
do
it.
We
always
got
lots
of
things
going
on.
You've
got
all
the
info
there
and
links.
Today's
speaker
is
not
the
first
time
he's
been
with
us
and
the
last
time
that
we
had
him
with
us,
gave
a
really
really
good
talk
and
had
some
amazing,
tangible,
practical
examples
that
I
really
enjoyed
alright
being
able
to
relate
concepts
that
are
very
complex.
A
We're
talking
about
graph
databases
today,
we're
going
to
be
hearing
more
about
nebula
more
in
depth,
but
being
able
to
relate
these
things
to
everyday
things
that
we're
talking
about.
In
this
particular
case
it
was
about
basketball
and
different
stats
about
different
players,
hoping
that
comes
back
today.
Our
speaker
is
a
little
bit
modest
and
I
don't
want
to
be
too
modest.
A
You
know
I
have
the
the
joy
and
fortune
of
making
the
data
on
kubernetes
rap
videos
that
we
were
putting
out
every
week,
but
our
speaker
today,
I
think,
has
been
making
music
about
cloud
native
stuff
for
even
much
longer
than
I
have
so
without
further
ado
way.
Can
you
introduce
yourself
for
the
folks
who
don't
know
you
super
good
to
have
you
you're
joining
us
from
china?
What
time
is
it
there
right
now
how
you
doing
man.
B
A
B
B
Yeah
sex
part
about
opportunity,
and
also
regarding
the
music
part.
I
just
played
some.
You
know
weird
music.
When
I
was
middle
middle
school
I'm
I
feel
privileged
to
have
a
chance
to
communicate
with
you
about
the
music
and
I
am
a
developer
in
shanghai
and
I
love
open
source
and
before
I
jump
into
the
graph
world,
I
used
to
work
on
open,
infra,
openstack,
okay,.
B
B
B
A
A
A
A
B
B
I
used
to
play
with
the
open
opening
for
a
community,
and
this
is
my
github
twitter
my
website,
and
wanted
to
mention
that
today
my
slice
is
a
web
friendly
one
and
I
already
put
it
online.
So
you
can
remember
this
url
and
you
can
check
back
and
forth
during
my
talking.
So
hopefully
it's
easier
for
the
listeners
yeah
and
the
yeah.
I
have
this
in
the
second
slide
too.
B
You
can
take
a
picture
too,
so
the
topic
today
will
start
from
the
the
basic
thing
around
the
graph,
but
I
I
believe
you
already
have
a
lot
of
sense
on
them
already.
I
will
go
through
very
quick,
then
we'll
give
some
more
detailed
information
on
how
nabla
graph
was
designed
and
the
architect,
and
why
do
we
need
yet
another
graph
database?
There
are
a
bunch
of
them
there
out
there
already.
B
We
already
know,
and
then
we
will
to
help
us
to
make
more
sense
of
what
navel
graph
or
what
the
graph
database
really
are
and
why
they
they
they
are.
There
we'll
go
through
some
know-hows
and
I
will
show
you
some
toy
projects
with
some
demos.
Hopefully
everything
goes
well
and
in
the
end
we
are
data
on
kubernetes
naval
graph
is
naturally
was
born
in
distributed
way
and
is
cloud
native.
B
B
B
So
I
will
go
through
very
quick
that
graph
in
mathematics
refers
to
those
thoughts
and
those
connections.
So
basically
they
are
just
vertex
and
edges,
and
that's
why
the
team
behind
level
graph
is
called
ve.
Soft
v
refers
to
vertex
e
refers
to
edges,
so
this
is
definition
of
graph,
so
we
will
know
why
we
need
this.
This
kind
of
definition
later
I
believe.
So.
What
is
a
graph
database
so
graph
database
is
a
database.
B
That's
persists
in
the
data
in
the
form
of
graph,
which
is
the
vertex
and
edges,
or
we
also
call
it
nodes
for
the
vertex
and
in
most
of
the
graph
database
in
general,
we
have
a
new
concept
called
property
graph,
which
is
a
definition
of
the
graph
on
top
of
graph,
and
we
add
something
called
properties
and
you
can
understand
them
as
the
key
and
value,
so
you
can
add
key
and
value
in
either
the
vertex
or
the
edges.
So
that
is
basic.
What
a
graph
database
is
all
about,
and
why
do
we
need
that?
B
So
this
slice
is
reused
from
the
last
one.
I
will
go
with
real,
quick
and
the
the
first
perspective
is
regarding
of
the
schema.
So
it's
not
a
very
large,
but
I
will
explain
so
in
the
left
side.
It's
it
comes
from
a
picture
reference
representing
the
the
schema
or
the
definition
of
data
for
a
tabular
database,
a
typical
relational
database.
B
But
when
you
come
to
when
you
want
to
represent
the
relationship
between
the
two
tables,
like
you
want
to
have
the
role
relationship
between
a
person
and
the
cooperation,
you
have
to
create
another
table
that
maybe
name
the
role
relationships
and
to
stand
for
the
relationship
like
someone
is
playing
some
roles
in
another
corporation,
something
like
that.
So
in
theory,
you
can
do
that
for
all
the
time.
B
But
it's
not
that
flexible
when
it
comes
to
a
scenario
that
we
cares
a
lot
of
different
relationships
and
it
changed
day
by
day
so
like
when
you
are
introducing
yet
another
new
relationship.
There
comes
to
the
schema
change
and
some
data.
Some
existing
data
should
be
rewrite.
So
that's
quite
that's.
Quite
you
know
not
that
flexible.
B
The
graph
database
was
born
to
solve
this
kind
of
issues
and
we
defined
the
schema
or
we
defined
the
data
in
a
way
that's
or
relationship
oriented,
so
you
can
just
define
the
way
it.
The
the
human
brain
is
understanding
that
just
like
this.
So
another
issue
is
when
you
want
to
query.
Think
information
like
like
you:
you
have
a
table
named
the
player
basketball
player.
You
want
to
query
the
the
given
a
given
player.
You
want
to
query
all
the
other
player
that
serves
exactly
the
same
team.
B
So
this
query
in
graph
database
is
like
a
one
jump
query,
and
it
can
be
done
in
this
way,
but
in
rdbms
it's
one
jump
will
be
something
like
this.
You
have
some
a
couple
of
drawings,
so
it's
it's
kind
of
disaster,
but
when
it
comes
to
multiple
jumps,
it
will
be
even
worse,
obviously,
and
the
crucial
the
deal
maker
actually
is
the
performance.
So
graph
database
will
persistent
data
in
a
way.
B
That's
originated
the
connection,
oriented
that
that
means
when
you're
querying
multi-hoop
it's
quite
cheap
and
when
it
comes
to
high
throughput
high
concurrency
case,
it
will
be
worse.
The
situation
will
be
make
a
lot
more
difference
between
the
two
model
of
database
system,
so
this
is
just
a
a
reference
that
there
will
be
like
a
hundred
or
more
times
differences.
B
A
B
B
Why,
between
the
two
queries
from
this
awesome
guy's
face
and
we
love
the
the
mimi
of
him
right,
the
other
one
is
a
ritual
game,
so
I
am
old
enough
to
play
the
nes
when
I
was
a
kid
so
and
I
noticed
that
the
the
awesome
kelsey
hightower
so
when
he
initially
introduced
the
kubernetes
in
in
in
one
of
the
cube
cars,
if
I
record
correctly
to
show
us
how
kubernetes
working
and
how
kubernetes
can
help
so,
I'm
I
go
through
all
the
game.
Virtual
games,
I
can.
B
I
can
see
to
find
this.
Maybe
the
the
metaphor
here
is
is
impr
appropriate,
but
I
I
hope
it
somehow
will
make
some
sense
so
in
in
this
game.
This
is
game
called
named.
Snow
snowman
brothers,
so
they
are,
the
brothers
is,
is
helping
to
kill
the
the
monsters
to
to
save
the
princess
and
they
have
to
jump
from
one
layer
to
another
layer.
So
in
this
metaphor,
I
I
I
somehow
in
a
wilder
way
think
maybe
the
layers
can
can
be
mapped
to
the
the
data
tables.
B
So
in
a
tabular
database,
you
have
to
jump
from
one
to
another
when
you
want
to
query
the
connected.
Sorry
like
that,
you
have
to
jump
from
here
to
here
and
you
have
to
scan
in
some
distance
to
find
out
data,
and
that's
one
hub
so
in
some
real
and
you
can
optimize
them
like
you
have
you
have
the
very
fast
leg
and
that's
that
can
be
mapped
to
a
case
that
you
you
create
some
some
sort
of
indexes.
B
A
B
A
B
A
B
Thank
you
yep
yep,
so
we
know
that
in
certain
scenarios,
when
it
cares
about
multiple
relationships,
so
we
know
that
graph
database
should
exist
and
how
popular
they
are
already
now.
So
we
can
check
from
this
is
the
url,
but
you
can
check
from
yourself
db
engines
that
it's
showing
that
you
know
from
one
perspective.
It's
a
marketing.
I
think
graph
database
is
really
is
really
in
trending
and
even
in
one
perspective,
it's
the
most
popular
popular
kind
of
database.
B
So
I'm
not
going
to
show
this
now,
but
you
can
check
that.
Okay,
then
we
move
forward
so
nabla
graph.
So
why
do
we
need
yet
another
graph
database
and
you
can
check
more
from
this
url
right,
so
I
will
before
we
move
on.
I
will
give
a
brief
introduction
of
what
says
like
for
nabla
graph.
This
is
the
this
is
a
page
from
the
documentation
of
the
nabla
graph,
and
I
will
give
a
brief
introduction
of
its
architecture
and
design.
B
So
nablograph
was
designed
in
the
in
day
zero
to
be
distributed
so
and
the
way
it's
designed
is.
It
comes
with
three
kind
of
services,
which
are
the
storage
service,
the
meta
series
and
graph
service.
So
the
story
service
literally
persists
in
all
the
graph
data
and
it
used
the
raft
to
ensure
ensure
the
data
have
ability,
and
the
replica
and
and
graph
service
is
actually
stateless.
And
it's
it's
own.
B
It's
the
service
that
handling
your
queries
and
parse
your
queries
and
and
make
the
plan
and
the
call
call
the
storage
to
to
make
all
the
data
fetching
and
the
manipulation
and
meta
service
is
in
charge
of
all
the
method.
Information
like
the
credential
and
how
the
data
was
distributed
around
the
different
hosts
and
you
can
put
to
put
all
the
services
in
in
any
way.
B
So
this
is
the
ship
as
blocking
in
a
small
river
and
the
two
guys
trying
to
make
it
right
and
the
another
case
is
the
famous
one.
You
know
the.
I
don't
recall
the
the
name
of
this
ship
and
it
impact
the
world
in
last
year
for
a
couple
of
months,
so
it
looks
like
similar
in
some
way.
But
the
way
you
are
dealing
with
the
the
thing
will
be
significantly
different.
B
So
scale
makes
differences
so
and
nebula
is
good
at
scale
and
another
reason
that
naval
graph
exists
is
it's
open
source
and
with
the
help
of
the
community
naval
graph
in
the
very
initial
phases,
when
it's
not
that
mature,
the
community
users
help
us
to
prove
that
it's
working
and
it
can
be
working.
Fine
and
the
power
users
is
teaching
the
community
and
how
some
really
some
really
good
scenarios
that
you
can.
B
You
can
improve,
and
you
can
do
this
in
the
next
level
way
and
they
help
us
to
involve
really
good,
so
nabla
graph.
This
the
project
of
nabla
graph,
was
initiated
in
in
2018
and
in
in
2019
december.
We
already
have
our
very
first
production
cluster
deployed
and
actually,
when
it
comes
to
2020
june,
we
have
our
very
first
ga
release
and
we
have
our
2.0
ga
last
year
in
in
in
march,
and
we
will
have
3.0
in
coming
months
in
march
or
april.
B
If
I
recall
yes-
and
there
are
already
a
couple
of
a
bunch
of
giant
teams
in
different
scenario-
different
industries
is
using
adopted
nabla
graph
in
their
production,
including
a
lot
of
giants
team
that
dealing
with
a
huge
volume
of
graph
data
like
matron,
which
is
the
chinese
version
of
yelp
like
tencent
and
like
akulaku
of
which
had
and
huawei
and
a
lot
of
was
not
added.
Here
like
last
week.
B
We
we
came
to
know
that
tencent
music,
which
is
the
chinese
version
of
spotify,
is
using
nabla
graph
dealing
with
their
knowledge
graph
like
by
baidu
last
month.
They
shared
it's
their
experience
that
they're
using
they
to
manage
their
a
huge
amount
of
user
profiles
and
they
are
doing
the
id
mapping
thing
with
nebulograph
and
they're
doing
really
good
jobs,
and
you
can
check
the
the
ranking
here
in
db
engines.
So
most
of
the
nabla
graph
users
now
are
in
chinese
most
of
them.
B
So
when,
when
db
engines
are
making
this
ranking,
so
you
know
they
are
counting
on
twitter
linkedin,
which
most
of
the
chinese
developers
or
companies
are
not
using.
So
we
already
in
in
a
very
front
ranking,
so
we
are
working
on
the
to
one.
We
want
to
promote
or
develop,
advocate
our
user
to
the
to
all
over
the
world
and
from
this
year.
So
I
hope
you
can
check
out
the
naval
graph
and
check
out
the
community
yeah.
B
B
We
have
a
bunch
of
different
sdks
and
the
last
at
the
end
of
the
last
year
we
have
a
couple
of
community
users
donate
their
contribution
on
like
nebular,
php,
nebula,
node.js
and
nebular.net
not
listed
here.
So
the
the
ecosystem
is
quite
made
sure,
but
we
are
still
working
on
to
make
it
better
and
better
every
day.
So
yeah
welcome
to
check
out
naval
graph
community,
so
marketing
part
is
mostly
done.
I
will
show
you
more
graphings
from
now
on,
so
this
is.
B
B
So
very
first
one
is
it's
a
intelligent
agent
like
you
can
from
the
app
you
can
request
your
in
your
natural
language
like?
Is
there
any
mcdonald's
nearby
the
street
football
so
underlying
the
graph
database
is
helping
me
trying
to
make
this
thing
work
and
I
will
demo
even
a
similar
project
later
on
how
the
dots
were
connected.
B
Another
thing
is:
it's
more
interesting
to
me
is
like
when
you
are.
You
are
surfing
the
different
restaurants
in
yelp.
You
want
to
know
some
choose
some
nice
fancy,
a
restaurant.
You
want
to
have
this
for
this
dinner,
and
sometimes
the
system
will
gives
you
some
recommendations
based
on
very
complex
algorithm,
but
in
the
recommendation
systems
the
explanation
of
the
system
is
is
somehow
quite
tricky,
but
you
can
get
that
help
with
the
nebula
graph
or
graph
database.
B
So
this
is
a
a
sentence
in
the
in
the
searching
result,
namely
those
those
whom
from
beijing
enjoyed
korean
cuisine,
said
this
restaurant
is
great.
So
this
is
kind
of
explanation
of
why
this
restaurant
was
ranking
high
so
and
under
the
hood
that
made
twins
shared
they
they
make
all
the
the
people's.
B
You
know
the
profiles
of
people
in
different
perspective
like
where
they
come
from
and
they
in
his
in
history
they
enjoy,
which
type
of
restaurant
or
cuisines
they
make
everything
in
in
a
huge
knowledge
graph
and
when
they
have
the
recommended
results.
They
just
make
a
fine
path,
find
the
the
path
of
this
graph
like
this
and
when
they
have
a
path,
and
they
make
this
path
into
a
human
readable
sentences
as
the
explanation
of
this
recommendation.
So
I
think
this
is
a
very
cool
example
that
so
for
navigation
graph
database.
B
This
is
quite
simple
query,
but
in
the
recommendation
system
the
performance
is
quite
it's
quite
good
actually,
and
there
are
actually
the
similar
case
that
the
graph
database
can
really
help.
There
are
more
than
10
of
typical
use
cases
which
is
kind
of
very
mature,
but
like
the
social
network,
obviously,
and
the
risk
control,
you
can
detect
the
fraud
in
real
time
with
graph
database,
and
you
can
check
the
the
the
the
the
people.
That's
you
can
check
the
people
in
the
public
public
security
domain
and
the
the
knowledge
graph.
B
If
you
put
the
knowledge
graph
in
the
graph
database,
you
can
have
multiple
hoops
query,
really
quick
and
there
are
some
machine
learning,
plus
graph
text
fields
and
in
some
of
the
cases
you
can
leverage
graph
database
and
I'm
not
going
to
name
all
the
scenarios.
But
I
will
give
some
of
them
in
the
following
of
the
talk.
B
So
in
china
we
have
a
website
that
almost
everyone
knows
that
you
can
before
you
are
joining
a
company
you
can
search
for
on
the
company
on
on
who
is
actually
sharing
the
the
the
whole
share
of
the
company
who
is
behind
and
what
his
other
corporation
is
sharing
with
this
corporation.
So
this
kind
of
web
website
or
regulation
system
can
be
done
with
a
graph
database
quite
easily.
So
I
created
a
toy
project
based
on
a
fake
data
named
corp
rail
search.
So
you
can
check
out
the
code
here.
B
B
So
this
is
the
actually.
This
is
the
graph
module
I
showed
them
before
when
we
were
comparing
with
the
rdbms.
So
in
this
graph
we
have
two
kind
of
vertex
or
node,
which
is
the
person
and
corporation
so
person
in
person.
You
have
involved
name
and
age
and
in
cooperation
you
have
the
name
and
and
location,
and
there
are
also
I'm
sorry.
B
B
So
this
is
a
very
simplist
module
but
can
somehow
reflect
the
the
problem.
I
hope
so
and
the
with
this
data
was
defined
in
a
graph
database
with
nebulograph.
You
can
use
this
query.
You
can
go
from
one
to
three
steps
from
this
corporation
over
all
kind
of
edges
by
direction
with
this
query,
you
can
do
something
like
this,
so
you
can
from
this
this
cooperation.
You
can
find
all
the
hidden
relationships
between
other
persons
and
corporations
in
in
a
visualized
way.
B
So
this
is
the
interface
of
network
graph,
so
with
the
with
this
kind
of
broad
graphing
for
constructed.
So
what
we
can
do
more
to
make
this
more
like
a
system,
and
I
will
give
more
information
on
this
project
so
start
from
the
architecture
part.
So
it's
quite
straightforward.
Just
like
other
websites
system,
you,
you
have
a
front-end
backend
and
your
database
and
in
this
case
we
use
graph
database
as
the
database
so
front-end
in
in
this
project.
B
I
use
the
view
js
and
d3
to
render
you
know,
manipulate
and
render
the
information
of
the
relationships,
and
I
will
make
it
more
sensible
for
the
human
beings
and
the
backhand
is
handling
the
request
of
the
api
from
the
front
hand
and
make
the
queries
to
the
graph
database
and
make
graph
that
and
assemble
the
data.
That
friend,
can
can
understand
and
return
it
so
quite
straightforward.
B
So
the
code
is
quite
simple,
so
this
is
a
a
a
simplified
code
of
this
logic
in
in
backhand
part.
So
you
can
see
when
a
cred
api
call
is
coming
and
you
can
handle
this
from
this
function,
and
this
function
is
making
this
match
query,
which
is
a
graph
query
in
the
open
cipher,
and
it's
actually
help
you
making
the
relationship
that
user
cares
and
and
will
be
sent
to
the
graph
database.
B
So
then
we
will
move
on
to
oh.
This
is
related.
This
slide
is
related
to
the
fake
data
that
I
create.
It's
also
open
source
and
I
created
the
this
data
to
you
know.
You
cannot
use
the
real
data
to
do
this
kind
of
things,
so
I
shared
how
I
generated
this
data
with
the
very
random
rows
and
names
and
how
I
import
the
data
into
the
graph
database.
B
So
you
can
check
out
to
my
article,
my
blog
and
my
github,
on
on
the
corresponding
tasks,
and
this
is
showing
on
that
leverage
nebula,
which
is
also
a
utility
utility
eye
open
source.
You
can
spawn
a
whole
bunch
of
the
graph
database
plus
the
nebula
graph
studio,
which
is
the
gui
projects
together
in
this
one
line.
So
so,
with
this
setup,
we
we
can
start
up
the
you
know
we
we
can
start
the
the
real
the
real
code
part,
so
I
will
demo
it
and
the
code
is
here.
B
B
Okay
and
we'll
start
with
any
corporation
enter
boom.
So
this
is
the
toy
system
that
you
can
help
user
to
find
given
person
or
cooperation.
You
can
find
all
the
relationships
that
you
cares
regarding
the
the
role
working
in
the
corporation
and
the
corporations
different
relationships.
So
isn't
that
quite
cool-
and
this
is
the
data
that
returned
you
can
see.
This
is
the
data
written
from
the
graph
database
and
you
can
see
this
the
result
long
result
and
you
can
see
the
query.
B
Here
this
is
the
graph
query,
so
this
is
the
very
first
case.
I
I
I
walk
you
through
very,
very
quick,
but
you
can
check
out
the
code
all
the
detail,
codes
from
the
github
to
look
some
more
details.
If
you
are
interested
in
so-
and
I
also
put
a
demo
video
here
in
the
slice-
but
I
I
already
demo
to
you-
so
I
won't
play
it
here.
So
then
we
come
to
the
second
use
case,
which
is
the
I
call
it
siri.
B
So
it's
similar
you
know
to
the
the
other
agent.
So
it's
a
dialog
system
that
you
can
answer
your
questions
intelligently,
maybe
not
that
intelligent,
because
I
make
it
very
it's
a
very
toy
level
project.
So
the
code
is
here
so
we'll
go
through
it,
so
we
will
first
go
through
the
graph
modeling,
but
I
actually
I
will
skip
this
part
because
this
this
robot
this
theory
is
actually
only
can
answer
the
question
related
to
tim
duncan
as
bart
shared.
B
Yeah,
do
it
yeah,
yes,
and
you
know
the
data
behind
the
the
whole
system.
I
already
demoed
it
before
and
you
can
play
it
anytime
in
this
demo
page.
So
there
will
be
a
couple
of
playgrounds
set
up
there.
You
don't
need
any
credential
if
you
only
have
a
browser
and
you
can
play
with
nabla
graph
with
the
tim
duncan
that
that
has
said
so.
I
won't
show
you,
but
if
you
not
see
it,
you
will
have
more
idea
on
on
the
following
demos,
on
how
the
data
was
shown
there.
B
So
one
thing
is
that,
and
in
this
data
set,
you
have
a
couple
of
players
and
a
couple
of
basketball
teams
and
the
players
can
follow
each
other
and
and
the
players
can
also
serve
in
certain
teams.
So
there
will
be
a
simple
graph
and
this
system,
as
is
actually
a
voice
agent.
You
can
ask
something
like
how
does
tim
duncan
connect
it
with
the
xiaomi
or
who?
What's
the
relationship
between
xiaomi
and
the
audio
shark?
B
Something
like
that
you
can
handle
a
couple
of
different
type
of
questions
and
the
whole
system
will
do
the
job
for
you
and
I
will
dab
a
little
a
bit
dive
deep
with
that.
So
it's
similar
to
the
other
projects.
You
have
front-end
backhand,
so
I
will
give
more
information
on
the
backhand
side
that
in
a
typical
system
that
handling
the
dialogue,
you
have
something
called
intent
matching
and
the
semantic
processing.
B
B
You
can
do
that
with
the
nlp
technology
with
machine
learning,
but
as
I
know
that
when
I
was
checking
with
a
lot
of
users
in
certain
cases,
the
if
else
is
actually
quite
performant
and
efficient
already
like
in
in
the
hospital
you
you
want
to
question
which
which
room
I
should
go
to
check
with
doctors.
I
have
this
simple
symptoms.
So
if
else
is
enough
in
those
kind
of
situations
and
then
with
the
intent
and
the
entities
which
is
a
and
b,
you
will
send
them
into
the
actor
of
the
internet
actor
part.
B
B
I
already
created
a
a
cataconda
playground,
so
you
can
actually
check
out
this
and
by
clicking
your
your
mouse,
you
can
actually
play
with
the
siri
on
your
own
browser
so
and
and
going
during,
that
you
will
learn
more
of
what
the
knowledge
graph
based
the
qa
system
will
look
like,
and
here
I
just
play
a
demo
I
recorded
before
so
yeah
oops.
It's
not
that
qualified,
okay,
okay.
I
hope
this
looks
better.
B
How
does
team
duncan
and
lakers
connected?
How
does
synthetic
and
layers
connecting
okay
yeah
and
that
works
and
underlying
the
query
and
how
the
data
was
connected
was
already
demoed
in
last
talk
in
in
the
okay
community?
So
I
won't.
I
won't
show
that
for
you,
but
no.
This
is
how
it
looks
like,
and
I
hope
you
will
understand
more
on
graph
database
was
helping
on
this
kind
of
agent.
B
B
So
with
this
kind
of
system
you
can
manage
them
all,
put
them
together
and
you
will
get
all
the
insights
of
the
relationships
and
what
and
how
they
were
used.
You
can
have
the
statistics
of
all
the
different
table,
different
data
sources,
so
I
will
have
have
more
sense
in
the
later
demo.
Hopefully-
and
this
is
the
what's
amazing-
is
look
like-
I
will
have
a
live
demo
later
and
we
will
start
with
the
data
modeling
first,
but
you
can
check
from
the
amazon
documentations
as
well.
B
This
is
a
simple
live
version
like
in
this
module.
You
will
have
the
database
like
you.
Have
then
dynamo
the
hive,
the
crdb
in
in
your
data
lake
and
the
database
can
have
some
cluster
instances
and
you
have
a
production.
One
cluster,
your
staging
one
cluster
and
each
cluster
can
have
a
couple
of
different
schemas
and
each
table
belongs
to
one
schema.
So
is
that
making
more
sense
to
you
and
the
table
has
different
columns
and
table
can
be
used
by
users?
B
So
this
is
the
real
user
of
developer
data
scientist,
the
data
admin
administrator
and
you
can
have
a
lot
of
data
in
your
airflow
system
and
that
can
write
to
specific
tables
and
the
tables
can
be
owned
by
users
and
you
can
tag
your
tables
and
the
table
has
description,
descriptions
and
has
a
and
the
columns
has.
Statistics
like
your
specific
applications
will
consume
this
data
in
a
different
frequency
that
can
be
also
managed
by
this
kind
of
system.
B
You
will
pull
the
corresponding
data
from
different
sources
and
you
will
ingest
them
in
your
graph
database
and
also
you
will
build
the
index
in
elasticsearch
to
enable
your
weld
card
search
and
there
will
be
a
a
meta
service.
That's
just
like
the
backhand
service
in
in
previous
projects
has
shared.
It
will
take.
The
api
calls
to
query
the
metadata
from
database,
so
this
layer
is
making
it
decoupled
with
the
different
kind
of
graph
database
as
backhand.
B
So
amazing
is
open
source
and
it's
not
binding
to
any
databases,
and
actually
it
even
supports
rdbms,
but
the
performance
and
the
scenarios
with
very
limited
in
that
case.
So
in
front
end,
we
will
have
the
user.
The
fancy
user
interface-
and
these
are
the
corresponding
code
in-
is
using
a
mono
data.
So
you
will
have
all
the
code
in
this
single
repository
here
in
github
and
then
I
will
give
you
some
demos
of
this
system.
B
So
this
is
the
the
front
end,
so
you
can
actually
search
anything.
So
this
is
elastic-backed
and
we
will
have
all
the
components
of
this
metadata
and,
for
example,
it's
this
table
is
called
comes
from.
The
test
schema
and
the
table
name
is
test
table
one
and
there's
actually
a
hive
and
it's
tagged
in
gold,
and
there
are
a
couple
of
owners
here
now
you
can
see
it's
last
mode
updated.
The
schema
was
updated
in
this
time
and
you
can
check
the
frequent
users
and
you
can
see
it's
all
columns.
B
B
Just
like
the
the
basketball
player,
we
did
last
time
so
in
the
in
this
view,
you
can
see
that
this
table
have
a
lot
of
different
relationship
with
with
the
other
components,
and
you
can
actually
see
I'm
going
to
okay,
this
table
dynamo.
This
is
the
nemo
database.
Has
a
cluster
named
gold
and
gold
has
a
schema
named
task
gold
and
from
the
graph
database
you
can
double
click
to
check.
B
B
It's
consumed
by
another
afro
jobs,
so
this
is
the
the
ability
that
we
could
do
when
we
put
naval
graph
at
the
back
end
of
the
mansion.
So,
apart
from
the
ability
that
I
already
showed
you
on
a
query
ability
the
schema,
flexibility
and
the
performance
of
the
graph
database,
comparing
to
a
tabular
database
and
normally
most
of
the
graph
database,
come
out
of
box
with
a
fancy,
usable
visualized
tool,
and
that's
that
make
difference.
B
From
my
perspective,
if
you
are
only
making
a
mental
governance
system
and
your
ability
sometimes
is
limited
by
by
your
meta
service
or
your
backhand
service,
you
have
to
define
specific
api
to
enable
certain
time
of
a
certain
kind
of
metadata
relationship
check.
But
in
case
you
are
using
a
graph
database
or
even
a
graph
database
like
nebula,
you
can
have
naturally
the
graph
query
interface.
B
Is
your
dsl
or
your
yet
another
powerful
interface
that
can
be
used
by
a
non-uh
graph
database
expert,
but
they
can
still
have
the
ability
flexibility
to
create
different
kind
of
variety
of
insights
between
the
different
data
without
defining
introducing
the
loop
of
the
introducing
another
version
of
api
and
testing
the
api
using
new
device.
You
just
expose
the
needed,
the
read
permission
of
the
graph
database
or
even
the
graph
database
with
a
visualized
tool.
You
can
like
the
expert
the
data
expert
to
explore
the
insights
directly
in
this
way.
B
So
that's
another
angle:
that's
graph
database
helps
and
then
the
final
is
I
want
to
show
you
give
you
more
a
sense
of
the
query
underlying
of
the
system.
So
this
is
the
code
of
nebular
proxy.
So
it's
a
sub
module
of
the
meta
service
part,
so
nabla
proxy
is
is
using
to
decouple
you
know
to
making
all
the
works
related
to
nablograph
as
a
backhand.
So
this
is
a
query
of
table
information
and
in
a
real
world.
B
The
query
will
be
something
like
this
and
it's
something
like
you
want
to
query
from
this
table
with
this
id
and
you
want
to
in
one
query,
to
create
all
the
relationships
related
to
different
angle
of
relationships
in
multiple
group.
In
just
one
query-
and
I
will
demo
you
so
in
in
in
rdbms-
is
it's
not
even
possible
in
real
time
normally
so,
but
in
level
graph?
You
can
do
that
in
no
time
boom.
You
will
have
the
very
complex
result
in
this
in
no
time
in
real
time.
So
it's
just
a
demo.
B
So
the
code
is
here,
and
the
nebula
part
is
apart
from
from
reconciling
the
resources
inside
the
kubernetes
cluster,
and
sometimes
you
have
to
deal
with
the
nebular
class
itself.
You
have
to
call
the
nebula
cluster
in
the
ngql
query
language,
so
that
will
be
the
nebular
part
here.
It
will
call
something
regarding
the
data
balance,
for
example
here
so
for
more
information,
if
you're
a
kubernetes
expert
welcome
to
our
community.
B
B
If
we
are
interested
and
then
I
want
to
introduce
a
tour,
I
created
called
enable
operator
kind.
So
it's
a
small
tool,
help
you
to
set
up
kubernetes
in
docker,
which
is
named
kind.
We
all
know
it,
and
then
we
will
place
everything
up
in
this
kubernetes
inside
the
docker,
just
in
one
line.
So
after
you
query
this
line,
you
can
access
the
cluster
very
easy
and
you
can,
for
example,
you
can
you
can.
B
B
So
this
is
the
document
documentation
of
the
sample
data,
so
yeah.
I
think
we
finished
our
topics.
So
let's
brief
recap.
So
now
I
think
we
know
from
the
you
know
the
snowman
brothers.
We
know
why
we
need
a
graph
database
and
from
the
the
shape
we
know
that
nebula
is
excel
at
the
scales
and
in
the
large,
the
huge
volume
of
data
never
come
to.
This
scenario
is
for
you
and
how
to
we
walk
through
three
toy
three
projects.
As
a
example,
we
have
a
corp
rail
search.
We
help
the
siri.
B
You
can
ask
him
about
the
team
dunk
and
leveraging
the
the
great
awesome
linux
foundation
I
mentioned
you
can
manage
the
meta
data
in
a
quite
fancy
way
in
the
in
the
interface.
It
also
comes
with
the
api
and
they
support
defacto,
all
the
different
kind
of
metadata
sources,
already
out
of
books
and
there's
a
very
active
community
and
then
finally,
I
introduced
something
around
our
approach
on
the
kubernetes
nebula
operators
and
check
out
my
github,
my
twitter
and
my
blog,
and
this
is
the
url
of
this
slice.
B
I
want
to
share
one
announcement
that
today,
just
I
just
updated
it
before
we
start
naval
graph
announced
our
as
a
service
manager
service
cloud
on
microsoft
azure
today,
so
you
can
feel
free
to
check
it
out.
It's
in
open
beta
in
the
first
month
and
yeah
welcome
to
check
out
our
github
repo
here
and
that's
all
from
my
site.
A
Pretty
good
stuff
and
very,
very
exciting
news
to
see
that
the
open
beta
with
microsoft,
azure
wait.
I
don't
know
where
you
learned
how
to
do
presentations,
but
you
either
had
a
very
good
teacher
or
just
naturally,
or
extremely
gifted
so
good.
You
cover
so
many
different
topics
in
the
span
of
an
hour
and
we
get
the
full.
A
You
know
the
full
experience
and
also
really
seeing
what
makes
this
database
different
and
what
I
really
like
is
just
just
how
visual
it
is
and
and
brings
these
things
to
life
that
sometimes
can
be
a
little
bit
difficult
to
read
or
difficult
to
understand,
but
it
just
projects
the
information
in
such
an
easy
way.
It's
at
the
end
of
the
day.
It's
not
about
data,
it's
about
helping
people,
make
better
decisions,
and
so,
when
you
get
to
see
things
at
least
for
me,
perhaps
I'm
more
of
a
visual
person
than
others.
A
But
I
I
really
really
enjoyed
it
interesting
to
hear
as
well
too
about
the
work
that's
being
done
on
the
operator.
Folks,
if
you
want
to
get
involved,
you
couldn't
have
a
nicer
person
to
help
you
get
on
board.
So
definitely
reach
out
ways
very
accessible,
twitter,
linkedin,
slack
etc.
So
there's
no
excuse
to
be
shy,
just
jump
in
and
start
asking
questions.
That
being
said,
you
know
one
of
the
things
I
wanted
to
ask
or
that
we're
asking
all
of
our
guests
is
we
we
talk
about.
A
A
There
are
different
benefits
that
some
people
might
reference,
that
it's
better
to
have
all
the
technologies
in
one
stack
that
there
could
be
cost
saving
advantages
that
there
could
be
different
benefits
of
of
running
a
database
on
kubernetes.
In
your
particular
case,
in
your
opinion,
what
do
you
think?
What
do
you
think
is
the
best
the
best
reason
as
to
why
people
should
consider
doing
it.
B
I
I
really
love
kubernetes.
I
think,
if,
if
possible,
we
should,
and
if
you
are
running
infrastructure
by
on
your
own,
you
should
put
things
on
kubernetes.
I
think
we
we
all
know
we
all
hate
a
lot
of
people,
hate
kubernetes,
but
a
lot
more
people
hate
linux,
but
I
think
kubernetes
make
the
essential
abstractions.
B
I
think
the
the
most
valuable
thing
in
kubernetes
is
the
api
itself,
maybe
like
five
or
eight
years
later.
We
don't
need
a
so-called
kubernetes
of
today
now,
but
the
abstraction
and
the
the
idea
behind
the
abstraction
to
save
the
day,
and
you
can
have
a
better
sleep.
Everything
is
in
in
the
loop
in
you.
Take
the
people
outside
of
the
loop.
B
I
think
that's
the
most
valuable
thing
and
we
have
the
serverless
modernization
way,
maybe
in
the
future
most
of
people
even
don't
know,
details
of
kubernetes
not
like
today
they
care
about
more
on
serverless,
maybe
or
even
some
more
fan,
even
more
more
crazy
or
you
have
to
deal
with
the
no
code
case.
But
I
think
in
today
kubernetes
is
still
the
go-to
solution.
A
B
A
Was
just
checking
out
your
your
website
right
now
and
I'm
gonna
link
it
right
here
in
the
chat,
so
folks
can
check
it
out
directly
and
so
the
slides
for
today's
talk
will
be
uploaded
on
there.
But
wade's
got
lots
of
good
resources
on
here,
too.
So
definitely
check
that
out.
I,
like
I
said
I
just
dropped
a
link
in
the
chat,
so
you'll
see
that
in
a
minute.
But
but
that
being
said,
one
of
our
favorite
parts
of
the
livestream
is
where
we
get
to
share
the
artwork
all
right.
A
Was
talking
to
me,
while,
while
when
this
is
going
on,
he's
like
this
is
so
interesting
and
also
because
you
know
anka
is
an
artistic
guy
so
to
see
these
things
brought
together,
brought
to
life
both
with
your
slides,
but
also
with
the
the
graph
database
itself.
In
terms
of
how
we
get
to
see
things,
we
absolutely
we
love
the
video
game
reference.
You
can
never
have
enough.
You
know
original
nintendo
and
yes,
that
was
great.
I'm
gonna
check
out
that
japanese
soccer
game
by
the
way
tip
we
got.
A
We
got
you
and
yeah
aqua
decided
to
draw
you
in
a
spurs,
so
we
could
represent
tim
duncan
a
little
bit
too
anyway,
plenty
of
plenty
of
good
stuff
way.
It
was
such
a
pleasure
to
have
you.
I
hope,
we'll
be
able
to
have
you
again
in
kubecon.
Remember
folks,
you
can
submit
your
your
sessions
for
kubecon
in
the
cfp
that
we
linked
and,
if
not
jump,
into
slack
talk
to
people
like
way.
A
This
is
what
this
is
all
about:
celebrating
community
meeting
people
from
different
countries,
people
that
from
other
countries
that
make
sacrifices
and
stay
awake
until
two
o'clock
in
the
morning
to
do
these
talks,
but
it's
the
magic
of
doing
it
live
so
wait.
Thank
you
very
much
we'll
be
talking.
Thank
you
all
right.