►
Description
Building a cloud-agnostic platform used to be a challenging task as one had to deal with a large number of different cloud APIs and service offerings. Today, there are many managed Kubernetes solution (e.g., Openshift, GKE, AKS, or EKS), it seems like developers could simply build a platform based on Kubernetes and be cloud-agnostic. While this assumption is mostly correct, there are still a number of differences and pitfalls when deploying across those managed Kubernetes solutions.
This talk discusses the experiences made while building the ArangoDB Managed Service offering across different Kubernetes distributions.
A
All
right
we're
we're
rolling.
Ladies
and
gentlemen,
thank
you
for
joining.
This
is
another
episode
of
the
openshift
commons
briefings
operator
hours,
and
today
we
are
fortunate
enough
to
have
with
us
aywa
kronshma,
who
is
the
team,
lead
and
main
architect
for
the
oasis,
managed
services
offerings
at
orangodb?
Hey?
How
are
you
today.
A
Really
good
really
good!
I
got
I
got
some.
I
got
some
some
final
ski
runs
in
this
past
weekend
up
here
in
in
the
new
england,
the
snow
pack
is
pretty
much
gone,
everything's
starting
to
shift
over
to
spring.
How
is
it
how's
it
over?
In
the
netherlands.
B
Yeah
we're
actually
well.
We
don't
have
that
much
snow
in
the
netherlands
to
be
honest,
but-
and
definitely
we
had
some
earlier
this
year,
but
spring
is
definitely
coming-
had
a
very
nice
weather
outside
right
now
and
we're
trying
to
get
more
out
outdoors
and
making
a
lot
of
fun
with
it.
B
Yeah,
what
we're
going
to
talk
about
today
is
the
challenges
that
we
have,
while
building
our
managed
cloud
offering.
We
are
running
around
with
the
oasis,
which
is
our
managed
service,
so
we're
running
around
with
b
databases
for
our
customers,
we're
doing
that
in
a
multi-cloud
fashion.
So
customers
can
choose
different
cloud
providers
and
there
are
lots
of
technical
challenge
behind
the
scenes,
because
not
every
cloud
is
the
same
and
we're
going
to
talk
about
what
actually
makes
the
difference
between
cloud,
the
different
cloud
vendors
and
how
we
overcome
that.
A
Sure
sure
you
want
to
pop
up
that
first
slide,
that
we
were
talking
about
earlier
yeah,
absolutely
no
problem,
and
thank
then
thanks
again
for
for
being
part
of
our
show.
You
know
we
we
host
all
of
our
software
partners,
who
we
work
with,
who
build
a
operator
for
openshift
and
orangodb
is
is,
is,
is
another
database
vendor
that
we're
proud
to
have
been
able
to
see
supporting
our
platform
for
their
database?
A
I
got
to
ask
you
a
question.
I,
your
your
logo
seems
to
be
an
avocado
for
the
company,
and
I
know
when
we
were
talking
previously.
That
you
know
oasis
is,
is
a
is
a
managed
offering,
as
opposed
to
just
the
database,
and
it
has
its
own
mark.
B
Yeah
so
there
there
is
a
story
behind
here,
so
you're
right,
the
orangutably
logo
itself
is
in
avocado
and
for
oasis.
We
wanted
to
merge
that
with
the
cloud
and
obviously
when
you
call
your
product
on
oasis,
we
also
want
to
have
customers
have
the
oasis
feeling
they.
We
want
them
to
feel
that
we
got
their
back
and
they
can
just
relax
in
a
very
nice
place
and
actually
work
with
the
database
and
let
the
worrying
do
to
us.
B
So
what
we
did
is
we
merged
the
typical
cloud
image
with
the
pit
of
the
avocado,
so
the
brown
dot
in
the
oasis
logo
is
actually
the
bit
of
the
avocado
surrounded
by
the
palm
trees
of
the
oasis.
A
That's
pretty
cool,
so
you're
you're
you're
about
seven
years
old.
I
think
you've
been
with
the
company
for
about
five
of
those
years.
What
kind
of
changes
have
you
seen
in
the
company
over
the
five
years
that
you've
been
there
is
it
is
it
growing
is?
Is
it
you
know
how?
How
is
the
how's,
the
state
of
the
union
over
at
orango
db,.
B
Yeah,
we
are
growing
very
much
at
the
moment
and
we're
growing
in
all
lots
of
different
areas.
So
over
time
we
have
grown
from
a
database
that
people
really
get
to
love
and
and
use
for
all
kinds
of
different
purposes.
Primarily,
documents
graphs
nowadays
also
graph
analysis,
and
we
see
that
growth
in
our
customers.
We
are
getting
many
many
many
more
customers
nowadays.
When
I
started
it
was
still
a
few,
but
we
now
have
very
large
customers,
and
we
can
also
see
that
the
actual
use
cases
of
our
customers
are
changing.
A
B
No,
the
company
actually
started
in
germany.
It
started
out
in
cologne
which
is
in
the
west
part
of
germany,
actually
not
that
far
from
from
where
I
live,
but
nowadays
we
are
really
distributed
company,
so
we
range
all
the
way
from
asia
to
russia.
Many
countries
in
the
eu
the
us.
B
So
I
don't
think
that
we
are
having
people
in
australia
at
the
moment,
but
we
may
get
there
as
well.
Yeah.
B
A
A
A
You
know
at
least
partially
the
old,
the
old
ways
of
at
least
going
out
and
going
to
kubecon
right
I
mean
we
were
we
were.
We
were
scheduled
to
go
to
kubecon
when
where
was
it
going
to
be
in
in
amsterdam,
right
yeah,
and
that
was
that
was
really
kind
of
a
bummer
that
that
right
back
then
there
last
march
everything
got
cancelled,
but
oh
well,.
B
Yeah,
I
think
that
we're
all
trying
to
to
cope
with
this,
but
actually
for
the
database
market.
The
the
last
year
has
actually
also
seen
a
lot
of
optics
because,
with
all
the
more
digitalization
of
products,
people
are
needing
a
way
to
store
all
their
business
and
they're
doing
that
in
databases.
So
for
us,
it's
actually
also
a
positive
side
to
the
whole
story.
A
B
Yeah,
well,
that's
always
a
very
good
question.
I
think
that
there
are
so
many
because
the
use
cases
vary
and
the
technology
varies
over
time.
B
So
we
have
seen
in
the
database
world
a
natural
progression
from
storing
fairly
fact
fixed
rows
of
data
and
slowly
migrated
to
more
nosql,
because
people
found
that
having
a
very
fixed
structure
is
not
very
pleasant
and
more
recently,
there
has
been
a
large
shift
towards
graph
databases
and
analysis
of
graphs
data,
the
data
being
stored
in
graphs
and
what's
what
are
I'm
going
to
be
specializes
in
the
special
place
that
we
have
in
there?
B
Is
that
we're
not
only
doing
documents
and
graphs,
but
we're
also
doing
that
at
a
very,
very
large
scale,
so
you're
not
limited
to
single
machines.
If
you
have
a
data
set
that
needs
many
machines,
you
can
still
do
it
and
you
can
still
have
very
performant
graph.
A
So
there's
there's
there's
your
graph
database.
There's
there's
sql
databases,
nosql
databases,
new
sql
databases.
What's
the
difference
between
graph
data
are
how
does
a
graph
database
compare
with
sql,
nosql
and
so
forth
and
so
forth?.
B
B
That
can
be
as
simple
as
the
traditional
social
network
example
where
you
have
people
and
you
people
have
connections
between
them
and
you
post
blogs,
and
there
are
comments
on
that,
and
all
of
that
can
be
thought
of
as
a
huge
graph.
B
While
you
can
do
that
for
a
social
network,
it's
pretty
obvious
but
think
in
the
world
of
iot,
for
example,
there.
You
also
have
lots
of
different
interconnections,
so
you
are
talking
about
objects
like
houses
and
sensors
and
actuators
and
all
kinds
of
devices
in
there,
and
all
of
that
can
be
thought
of
as
a
huge
graph,
because
there
are
lots
of
connections
between
them
where
the
graph
database
really
shines
is
in
asking
questions
about
such
a
graph.
B
B
Then,
if
you
want
to
think
of
how
you
could
write
that
if
you
have
to
do
all
these
lookups
with
a
traditional
sql
like
joining
approach,
you
would
first
of
all
have
to
think
of
all
of
that
up
front
and
it
needs
a
lot
of
interaction
with
your
database,
so
also
on
the
technical
level.
That
means
a
lot
of
programming,
but
it
also
means
a
lot
in
performance
with
the
graph
database
and
the
graph
query
language
behind
it.
B
You
can
ask
that
entire
question
in
one
go
to
the
database
and
let
the
database
solve
the
problem
for
you
and
it
will
just
go
through
all
the
data
that
it
has
and
give
you
back.
The
answer
and
where
around
to
be,
is
really
special,
that
it
can
give
you
that
that
answer
very
quickly
also,
if
your
data
becomes
very
large.
A
People
watching
this
right
now
will
be
able
to
put
questions
into
their
into
their
chats
and
they're
going
to
pop
up
over
here
and
our
our
producer.
Chris
is
making
sure
that
that's
happening.
Are
you
saying
that
that
a
comment
that
someone
made
on
youtube?
If
and
I
don't
know
what
their
back
end
system
is,
but
you
can
then
actually
use
their
back-end
data
store
to
run.
You
know,
reports
and
and
graph
that
up.
B
A
B
Yeah,
that's
that
that's
the
beauty
of
this.
It's
not
limited
to
a
single
use
case
and
a
lazy
one
answer
would
be
all
of
the
above.
But
you
to
give
you
a
couple
of
examples.
We
spent
all
the
way
from
tracking
aircraft
parts
to
iot
applications
to
utilities,
to
medicine.
B
Just
imagine
what's
the
kovit
crisis
and
all
the
research
around
that
is
doing,
and
you
could
even
think
of
all
the
resource
papers
being
written
around
govit
and
spin
that
up
in
a
huge
graph
in
that
way
advanced
square
the
science
behind
it
and
come
up
quicker
with
answers
to
this
whole
pandemic.
A
Oh,
I
was,
I
was
just
making
a
note
here.
Actually
I
still
am.
Maybe
I
should
stop
making
note
and
look
at
you.
My
mom
always
said
to
make
sure
you
look
at
people
when
you're
talking
to
them.
So
so
you
know
these
challenging
times
that
the
people
refer
to
it.
As
what
kind
of
impact
has
that
had
on
your
business.
B
We
were
already
a
distributed
company
with
people
in
many
countries
as
we
discussed,
but
I
think,
like
any
other
business,
we
have
learned
that.
Yes,
you
cannot
just
take
the
plane
and
go
to
your
customer.
You
have
to
talk
to
them
I'll
resume
or
whatever
medium
you
use
and
also
in
the
way
that
we
work
internally
with
our
engineering
with
our
sales.
How
do
you
coordinate
that?
B
B
It's
going
like
crazy
and
they
are
storing
data
and
what
we
see
more
and
more
and
that's
definitely,
a
change
of
the
last
year
is,
since
the
customer
also
doesn't
have
easy
access
to
their
offices.
They're
also
store
more
and
more
moving
their
data
into
the
cloud,
and
that
is
also
where
around
the
oasis
then
pops
up,
because
we
see
more
and
more
customers
choosing
our
managed
service
because
they
already
reason.
I
cannot
go
to
my
servers
in
my
office.
B
A
A
Yeah
interesting,
okay!
Well,
good!
So
why
don't?
I
turn
it
over
to
you
here
and
and
let's
let's
learn
about
your
your
title
slide
here
and
and
see
what
you
got
for
us
very.
B
Good
okay,
so,
let's
dive
into
this
multi-cloud
provider
platform
that
oasis
is
and
we're
here
also
on
the
invitation
of
the
guys
behind
openshift.
So
obviously
we're
talking
about
kubernetes
and
everything
around
there.
B
And
let
me
start
by
saying
that
not
every
kubernetes
is
actually
the
same,
and
although
people
may
think
that
way-
and
I
would
love
it
to
be
true-
we
found
in
our
endeavor
to
build
this
oasis
platform
that
it
is
not
really
true
and
the
devil
is
in
the
detail
and
what
I
propose
or
will
discuss
today
is
some
of
these
challenges
that
we
have,
and
I
will
give
you
some
examples
of
where
things
can
actually
make
a
difference
between
different
kubernetes
offerings.
B
So
the
thing
that
we're
going
to
talk
about
is
what
kind
of
challenges
you
have
when
you're
doing,
multiple
provider
support
and
we're
also
going
to
talk
about
kubernetes
as
an
abstraction
later,
but
not
every
abstraction
layer
is
perfect
and
we're
going
to
dive
into
the
differences,
and
those
differences
are
in
lots
of
areas,
the
typical
traditional
ones
like
security,
but
also
networking
and
obviously
for
a
database
storage.
B
B
So
a
little
bit
of
introduction
about
myself,
we
already
discussed
I'm
team
lead
of
the
aramidv
oasis
product
and
I
always
like
to
work
on
distributed
systems
and
actually
make
things
work
for
customers.
B
B
A
B
Which
you
can
do
if
it's
flat
right,
yeah
yeah,
okay?
Obviously
there
is
also
a
much
bigger
team,
so
we
are
just
some
names
and
and
faces
of
my
my
teammates
a
little
bit
of
an
intro
of
a
runway
b
itself.
So
we
already
discussed
it's
a
graph
database,
but
you
can
go
much
further
than
graphs.
You
can
also
store
documents
even
as
raw
as
key
values.
B
You
can
have
very
large
graphs
and
we
give
you
all
kinds
of
tools
to
make
it
easy
to
not
only
scale
that
data,
but
to
make
it
still
performance
and
that's
very
important
and
all
of
that
data,
whether
it
is
your
graphs
or
your
documents.
All
of
that
can
be
queried
with
a
single
query
link
that's
pretty
important
for
our
customers,
because
imagine
that
you
would
have
to
learn
three
four
or
five
different
query
languages
for
each
of
them,
but
it
would
be
really
annoying
this
way.
B
So,
let's
move
on
to
our
wrongdoing,
oasis
our
managed
platform.
So
what
do
we
do?
We
have
in
in
the
history
of
orangutb?
We
have
spoken
with
lots
of
customers
and
we
get
a
lot
of
comments
saying
us.
I
really
like
your
database.
It's
easy
to
use
it's
great,
but
please
run
it
for
me.
I
don't
like
running
this
database
and
that
actually
makes
a
lot
of
sense
because
running
a
database
or
any
stateful
load.
But
definitely
a
database
is
not
an
easy
thing
to
do.
B
You
have
to
look
at
all
the
details
and
you
have
to
get
all
of
the
details
right.
Otherwise,
you
have
problems
and
that's
exactly
what
we
help
our
customers
with.
So
we
run
databases
for
them.
B
So
we
run
a
customer
comes
to
us
and
say
I
want
to
run
an
orangutb
database
in,
for
example,
google
in
in
london,
or
I
want
to
run
it
in
aws
in
ohio
and
we
make
sure
that
that
database
is
started
there,
but
also
we
make
sure
that
it
keeps
running
there
and
we
monitor
it
and
we
make
sure
all
the
backups
are
in
place
and
all
that
stuff.
A
You
know
there's
this
concept
of
you
know
if
you're
using
you
know
some
kubernetes
platform
like
openshift
or
others
that
you
can
build
it
once
and
deploy
anywhere.
Yes,
that's
that's
not
the
case
with
gke
aws,
because
they're
they're
all
different
enough
that
it's
almost
like
having
to
support
multiple
different
platforms
for
the
same
application.
B
If
you
do
care
about
all
of
these
things-
and
I
think
most
people
actually
do,
then
there
are
all
kinds
of
small
variations
and
let
me
let
me
dive
into
those,
because
we
see
kubernetes
as
an
abstraction
layer,
and
I
often
make
the
comparison
of
the
kubernetes
and
the
promise
of
the
java
virtual
machine
as
we
had
it
in
the
90s.
That
promise
was
also
one
yeah.
B
You
can
write
once
and
then
run
it
everywhere,
but
the
reality
and
I've
written
my
fair
share
of
java
in
the
past
as
well
was
that
you
were
also
tweaking
like
okay.
What
platform
am
I
running
on
and
I
need
to
adjust
for
that
platform?
B
B
Obviously
we're
also
available
on-prem.
If
you're
running
openshift,
you
can
use
also
the
a
rangodb
operator
to
run
an
rdb
database
by
yourself,
but
that's,
not
oasis,
for
oasis
we're
running
on
these
managed
kubernetes
offerings,
one
for
amazon
eks.
B
B
B
So,
let's
talk,
amazon
amazon
is
actually
it's
a
very
stable
platform.
That's
awesome,
but
it
has
many
resources
that
you
need,
if
you're
creating
a
kubernetes
cluster,
it's
not
as
easy
as
to
say
yeah,
I'm
spinning
it
up
and
just
creating
a
cluster.
No,
you
have
to
bring
in
your
load
balancers
and
your
security
groups
and
lots
of
stuff
that
makes
it
challenging.
B
It
also
makes
it
easier
to
control
once
you
have
done
it,
but
it
takes
a
huge
step
to
get
there,
and
what
we
find
is
that
everything
that
you
do
on
amazon
is
working
really
well,
but
the
error
handling
is
a
bit
outdated.
Unfortunately,
it's
not
really
structured,
and
sometimes
you
have
to
resort
to
things
like
string,
parsing
and
so
on.
If
we
switch
to
gke,
the
apis
of
gke
are
awesome,
you
can
just
spin
up
and
the
kubernetes
cluster
like
that,
and
it's
extremely
easy
to
do.
B
But
the
big
problem
that
we
have
with
gte
is
they're
very
aggressive,
update
policy
they're,
just
forcing
you
to
go
to
new
kubernetes
versions
on
a
relentless
space
and
if
you're
on
a
managed
platform
you
have
to
follow.
There
is
no
choice.
Can
I
ask
your
question
something
stateful
so
like
we
do?
You
have
to
be
extra
careful
there
yeah
go
ahead.
A
No,
it's
just
you
know
having
been
here
at
red
hat
for
whatever
21
years
now
I
was
here
when
you
know
red
hat.
Linux
was
a
very
popular
distribution
version,
6
version,
7
version
8
and
our
our
update
cycle
was
really
fast
and
when
we
started
creating
our
enterprise
product
red
hat
enterprise
linux
well,
the
first
version
was
actually
called
advanced
server
2.1.
A
But
that's
people
probably
don't
remember
that
we
we
did
that,
because
our
customers
told
us
that
they
wanted
a
stable
platform
for
three
to
five
years,
so
they
didn't
have
to
keep
revving
their
apps.
All
the
time
is
that
kind
of
where
we
are
now
I
mean,
wouldn't,
is
wouldn't
that
be
a
major
problem
for
people
trying
to
build
apps
for
gke.
B
B
So
the
majority
of
the
updates
that
we
do
in
terms
of
kubernetes
versions
are
more
of
a
test
and
deploy
kind
of
approach,
but
there
are
definitely
also
issues
usually
in
every
minor
versions.
There
are
two
or
three
issues
that
really
need
changes
in
our
code
to
be
effective
on
the
new
version.
A
B
Yeah,
that's
I.
I
must
add
here
that
the
majority
of
things
that
require
changes
have
to
do
with
stateful
workloads
and
the
majority
of
applications
doesn't
have
that
much
state
in
it.
That's
what
you
use
the
database
for
so
the
majority
of
applications
are
not
going
to
have
that
much
problems
when
their
underlying
kubernetes
version
is
upgraded.
B
It
becomes
a
problem
if
you're
doing
more
stateful
things
or
are
you
more
heavily
integrated
in
the
network,
for
example
you're
on
the
edge
of
the
requirements
of
the
network
in
terms
of
your
security
or
your
firewalling?
It
so
things
like
that
yeah,
then
you
have
to
be
careful,
and
then
you
have
to
go
with
the
flow.
A
B
Okay,
let's
switch
over
to
to
microsoft.
Azure
we
haven't
touched
those.
We
can
see
that
it
is
clearly
the
least
mature
of
the
three,
but
they
are
very
responsive
in
in
their
support.
We
have
had
great
support
from
them.
The
biggest
problem
that
we
have
is
really
on
the
storage
side,
with
with
microsoft
azure.
B
It
has
to
do
with
the
attachment
of
persistent
volumes
and
resizing
of
persistent
volume
claims
and
so
on,
and
also
there
is
a
weird
behavior
of
the
cluster
router
scaler
on
microsoft,
azure
so
for
the
cluster
outer
scaler
make
sure
that
your
cluster
actually
grows
when
the
workloads
on
your
cluster
grow
and
for
us
that
is
a
very
useful
feature,
because,
as
soon
as
a
customer
says,
I
want
to
run
this
deployment
there.
B
We
make
sure
that
the
nodes
are
there
and
that
all
the
availa,
the
resources
are
there,
but
it's
not
something
that
we
do
ourselves.
We
only
say
we
want
to
have
this
capacity
there
and
the
cluster
outer
scaler
is
going
to
make
it
happen.
B
If
you
look
on
microsoft,
azure
the
cluster
outer
scaler
is
not
that
smart
was
dealing
with
zones
and
that's
for
us
a
problem.
We
want
to
go
across
availability
zones
because
of
the
high
availability
guarantees
that
we
give,
but
what
the
outer
scalar
for
azure
is
actually
doing.
Is
it's
just
saying
I'm
creating
your
next
node
in
the
next
region.
B
So
first
you
get
a
node
in
zone
a
and
then
in
zone
b
and
then
in
zone
c,
but
it
doesn't
take
into
account
if
you
have
a
certain
affinity
with
your
note
and
if
you
have
data
that
lives
in
zone
two
and
suddenly
the
node
is
a
scalar
is
going
to
make
something
in
zone
three,
it's
not
very
helpful,
so
there
are
lots
of
issues
there.
B
B
There
is
a
big
difference
in
how
performant
the
different
persistent
volume
offerings
are,
and
all
of
them
have
different
options
for
configuring,
and
you
can
choose,
of
course,
your
ssds
and
your
iops
settings,
and
so
on.
As
soon
as
you
go
in
that
direction,
it
becomes
provider
specific.
So
getting
back
to
your
previous
question
of
yeah,
where
is
actually
the
difference
and
can't
I
just
run
it
anywhere
yeah.
Technically
you
can,
but
if
you
want
to
have
a
volume
with
certain
characteristics,
you
have
to
specify
it
for
that
platform.
B
That's
not
something
that
kubernetes
is
going
to
give
you
and
also
the
characteristics
of
the
performance
over
time
are
changing.
The
majority
of
providers
have
an
interesting
feature
that
they
essentially
give
you
a
buffer
of
ilpf.
So
those
are
the
I
o
operations
that
you
can
do
per
time
per
unit
of
time,
and
if
you
exceed
that
buffer,
the
majority
of
providers
is
a
bit
lenient
and
say
yeah.
Okay,
I'm
going
to
it's
fine,
you
can
do
it,
but
not
for
a
long
time.
B
Let's
talk
about
volume,
resizing
volume
resizing
is
something
that
is
specified
in
kubernetes.
You
can
just
create
a
resizing
resource
of
your
persistent
volume
claim
and,
according
to
the
kubernetes
specification,
it's
going
to
resize,
but
it's
not
really
done
exactly
in
the
same
way,
and
that
makes
it
hard.
So
we
have
had
to
build
in
additional
codes
specifically
for
microsoft,
azure,
because
of
the
way
that
the
attachment
of
volumes
works.
The
rest
of
the
providers
can
change
the
volumes
on
the
fly.
Don't
have
a
problem
with
that,
and
here
it
doesn't.
B
The
big
question,
of
course,
we
are
right
now
already
well
into
2021,
and
do
we
see
an
improvement
there,
because
we
have
been
working
on
the
oasis
platform
for
two
and
a
half
years.
Two
and
a
half
years
is
pretty
much
a
lifetime
in
kubernetes
world.
So
will
it
get
any
better
and
I'm
afraid
not?
This
is
just
a
picture
of
container
runtimes
that
we
see.
B
Of
course
not.
All
of
them
are
available
on
the
managed
platforms,
but
they're
a
huge
option,
a
huge
amount
of
options
to
choose
from,
and
we
see
the
whole
kubernetes
space
is
evolving
in
a
very
rapid
pace.
There
are
lots
of
exciting
projects
popping
up,
but
are
also
deviations,
always
they're,
specific
good
and
bad.
A
B
It's
always
a
very
challenging
question
to
answer
my
personal
answer
to
that
would
be
no,
but,
let's
be
honest,
there
are
also
a
huge
amount
of
different
cars
in
the
world.
Do
we
need
so
many
different
cars?
People
also
have
a
preference-
and
I
think,
in
this
case
that's
also
the
case.
So
there
are
preferences
there
are
real
benefits
of
all
of
them,
so
there
is
always
a
use
case
where
one
fits
better
than
the
other.
A
Okay
was
it,
you
made
the
scott
mccarty
from
our
company
once
when
I
can
quote
him,
he
was
talking
about.
Our
platform
is
because
there
are
so
many
different
types
of
vehicles
and
each
one
is
is
multi.
You
know
is
purposed
for
a
different
thing,
but
what
did
he
say
he
said?
Openshift
is
a,
is
a
dump
truck
that
can
carry
28
yards
of
dirt
and
go
200
miles
an
hour
and
handle
really
well.
I
thought
that
was.
I
thought
that
was
kind
of
cool
anyways,
throw
that
out
for
scott
yeah.
B
Oh
obviously,
there
are
many
more
different
areas
that
we
can
discuss
and
but
there
are
lots
of
different
issues
that
we
are
dealing
with
like
logging
like
networking,
but
I
just
didn't
want
to
go
into
all
the
details.
Right
now
sure.
B
Of
course
I
have
to
I
invite
everyone
to
take
oasis
for
a
test
drive.
You
can
try
out
oasis
for
free
and
go
to
orangeydb.com.
There
is
a
big
button
there
where
you
can
try
it
out
and
you
can
just
have
fun
with
it.
B
A
B
Sure
go
ahead.
A
A
B
Yeah,
let's
start
with
the
way
that
they
deploy
that's
a
very,
very
prominent
difference
over
the
last
couple
of
years.
If
you
were
looking
at
it
four
years
ago,
everyone
was
running
their
database
by
installing
some
package,
some
linux
package
on
their
servers
running
everything
in
I
don't
know
something
like
systemd
container
or
scripts,
and
hooking
up
virtual
machines.
B
B
B
That
is
a
huge
shift,
and
it
also
it
changes
everything
from
the
way
that
we
distribute
our
product
to
the
way
that
how
we
can
support
our
customers,
how
you
get
logs
and
so
on.
That
has
been
a
tremendous
shift.
A
B
Well
depends
on
how
you
are
looking
at
that.
I
think
that
in
18
months
the
shift
towards
kubernetes-
I'm
not
saying
it
will
be
complete,
because
there
will
always
be
companies
that
don't
jump
on
that
bandwagon,
but
it
will
be
pretty
much
all
over
the
place.
I
think
what
is
really
changing
nowadays
is
two
aspects
of
the
story.
One
is
scale
and
the
other
is
what
they're
actually
doing
with
it.
B
Let
me
start
with
the
scale
aspect
where
even
one
and
a
half
years
ago,
you
could
see
graphs
that
were
spin
up
with
some
tens
of
thousands
of
nodes,
now
we're
seeing
graphs
with
millions
and
hundreds
of
millions
and
billions
of
nodes,
and
in
order
to
fit
all
of
that,
it
absolutely
no
longer
fits
on
a
single
machine.
You
need
many
machines
and
we're
easily
crossed
the
number
of
nodes.
What
we
even
didn't
think
of
a
couple
of
years
ago,
that
trend
is
going
to
continue
very
fast
in
in
the
future.
B
The
other
thing
is
what
they're
actually
doing
with
it,
because
you
don't
only
want
to
store
your
graph
data
and
query
for
it.
We're
seeing
it
trend
towards
more
analysis
of
the
data,
so
there's
a
lot
of
things
that
you
can
do
and
then
you're
touching
again
things
like
machine
learning
and
artificial
intelligence
and
doing
that
in
combination
with
graph
data
is
a
new
area
that
is
popping
up
like
crazy.
B
B
Yes,
that's
true
and
and
it'll
it's
not
only
how
you
store
it,
but
also
how
do
you
model
it,
for
example,
because
if
you
have,
if
you
think
of
your
traditional
machine
learning
like
just
simple
vectors
of
numbers
and
and
do
some
magic
with
that,
how
do
you
map
your
graph
data
set
into
something
that
can
be
understood
by
the
machine
learning
algorithm
that
we
already
know
that
challenge
alone
is
a
very
interesting
one.
B
Yeah
you
could,
but
why
would
you
it's
pretty
much
the
same
question
as
if
you're
standing
in
front
of
a
very
nice
bmw
can't
you
go
buy
a
horse
and
yeah
you
could,
but
there
are
so
many
more
options
that
you
have
nowadays
that
I
would
prefer
the
bmw
over
the
horse.
To
be
honest,.
A
Okay,
so
from
a
marketing
perspective
use
you
know,
I'm
not
sure
where
who's
who,
where
your
cmo
lives,
but
there's
probably
some
things
that
your
marketing
team
would
like
you
to
make
sure
that
you
put
out
here
so
that
it
negates
that
phone
call
five
minutes
after
we're
done
where
he
or
she
says
you
know
geez,
you
were
live
on
the
internet.
A
B
Yeah,
I
think,
the
the
the
part,
the
message
that
we're
trying
to
get
across
is
make
sure
that
you
can
actually
model
your
data.
You
can
store
it.
You
can
scale
it
and
get
your
query
stuff
done
with
it
and
get
the
actual
value
out
of
your
data.
That's
really
key,
and
what
is
the
best
way
to
get
that
done?
B
A
Okay,
and
so
what,
where
can
people
come,
find
you
folks,
like
you
know,
are
you
do
you
putting
on
you
know
a
user
conference
or
what?
What
what
what's
going
on.
B
Yeah,
the
best
way
to
get
in
touch
with
us
is
through
our
website.
Through
our
community
slack
channel.
We
are
very
active
there
and
we
have
a
large
community
of
active
rankid
b
users
there
as
well,
so
there
are
always
people
both
from
inside
but
also
from
outside
the
company
happy
to
help
you
out
there,
and
I
think
it's
really
key
to
experience
the
orangutable
product
for
yourself.
You
can
of
course
do
that
with
oasis.
B
You
can
do
that
as
simple
as
running
a
darker
container
and
playing
with
it,
but
make
sure
that
you
start
learning
right.
We
have
a
lot
of
great
examples
in
both
on
our
website
also
on
oasis,
for
example.
If
you
want
to
learn
about
how
you
could
do
fraud
detection
with
a
graph
database,
we
have
a
great
example
on
the
ways
of
how
you
could
do
that,
and
we
tell
you
all
the
interesting
queries
that
you
can
run
against
that
and
start
playing
that
with
that
data.
B
So
that's
that's
something
that
we
have
seen
in
the
past
that
if
you
want
to
try
out
the
database,
what
can
you
do
without
data?
Not
that
much
and
with
the
examples
that
we
can
now
give
you
you
can
just
get
started
right
away,
start
playing
with
it
on
the
more
graph
analytics
side.
B
So,
if
you're
more
on
the
ai
machine
learning
side
of
the
story
and
want
to
start
exploring
how
to
model
your
graph
data
for
machine
learning,
we
have
a
lot
of
notebooks.
You
can
interactively
run
them,
they
even
integrate
with
oasis.
If
you
want
and
just
play
around
with
the
story,
how
do
I
map
my
data
into
something
that
my
machine
learning
can
work
with
and
then,
of
course,
visualize?
The
output
again.
A
B
We're
definitely
if,
if
possible,
we're
definitely
going
to
send
people
over
there
and
right
now.
The
big
question,
of
course
is:
is
it
going
to
be
possible?
Nobody
knows
that
right
now,
but
we
are
trying
to
get
back
to,
let's
say
the
normal
routine
of
doing
conferences
and
appearances
all
over
the
place,
because,
let's
be
honest,
it's
also
just
a
lot
of
fun
to
interact
with
all
the
customers
and
all
the
prospects
out
there
right.
A
B
Yeah
thanks
for
having
me
and
good
luck
with
the
show,
all
right
appreciate.