►
From YouTube: Building Apps with Cassandra Python Driver
Description
With the new Python driver for Cassandra it is easy to build integrations and apps that use Cassandra seamlessly as a back in. This session will explore what it takes to build the app and the features available with the new Python drivers.
A
Hello,
everyone
and
welcome
to
this
week,
Cassandra
community
webinar,
my
name
is
Christian
haske
with
datastax
and
I
am
delighted
to
have
with
me
this
morning.
Eddie
Santa,
Lee
Eddie,
has
been
a
longtime
supporter
of
Cassandra
and
the
community,
and
currently
is
the
CTO
of
big
data
and
analytics
at
CSC
make
sure
to
follow
Eddie
on
twitter
and
also
friend
them
on
LinkedIn
as
well.
A
If
you
would
like
to
ask
us
a
question,
please
use
the
Q&A
tab
inside
a
webex
type,
your
question
in
there
and
we
will
get
through
just
as
many
of
them
as
we
can,
depending
on
how
much
time
we
have
at
the
end.
So
once
again
use
your
Q&A
tab
inside
a
webex
and
ask
those
questions
there.
So
Eddie
we
have
possible
to
you
and
you
should
be
in
control.
Thank
you
very
much
for
joining
us
today.
Take
it
away
Thanks.
B
Ok,
so
yeah,
as
you
mentioned,
I,
have
been
working
with
Cassandra
for
a
long
time
since
thought
six
I
put
into
production,
with
my
old
team
I've
recently
joined
CSC,
to
render
big
gain
and
analytics
and
right
now,
a
lot
of
that's
very
focused
on
legacy
technology.
Integration,
dude
we're
doing
a
lot
of
fun
things
Sandra
now
too,
so
they'll
be
more
coming
out
of
that.
B
The
app
specifically
I'm
going
to
talk
about
today
is
first
time
to
talk
about
the
driver
and
how
it
works
and
look
there
at
the
app
itself.
But
I'm
going
to
talk
to
you
is
app
that
was
built
to
do
integrations
between
swan
and
Cassandra.
There's
an
old
version,
the
applets
filter
and
get
up
today
that
uses
the
old
approach
with
a
lot
more
painful
things
and
the
new
versions
been
built
with
the
container
Python
driver,
it's
being
tested
right
now
and
will
be
updated
in
the
getup
suit.
B
It's
also
the
slide
deck,
so
we
can
jump
right
in
so.
The
first
thing
is:
where
is
the
driver
with
all
the
data
sex
drivers
years
to
get
out
calm
side
with
datastax
python
bus
driver
Tyler,
who
wrote
this
at
least
most
of
it
is
on
here?
Basically,
if
you
walk
through
this,
you
have
all
the
dependencies.
It
has
a
normal
review
files,
everything
all
tied
to
it.
So
if
you
want
to
use
it
is
actually
pretty
easy.
There's
even
a
link.
A
B
So
obviously,
a
lot
of
people
are
going
over
the
control
interface
rather
than
using
the
older
approach
with
the
first
side.
Some
people
still
use
thrift.
So
my
happy
and
supports
the
first
side
as
well.
If
you
need
to
do
it
that
way,
but
since
a
lot
of
people
are
really
able
to
c
ql,
I
wanted
to
move
it
over.
B
So
one
of
the
key
features
is
out
here
from
the
configuration
perspective
is
setting
up
connection
pooling
during
automatic
no
discovery.
The
old
way
was
really
painful,
I
had
to
put
in
a
set
of
nodes
and
then
I
needed
place
on
to
basically
do
all
the
smart
thing
decide
what
note
to
connect
to
read,
use
a
pseudo-random
algorithm
which,
as
we
all
know,
doesn't
work
particularly
well
and
causes
a
few
little
issues
with
nodes
getting
hit.
Far
too
often,
if
it
netherwood
went
away,
python
just
ignored
it
going
forward,
so
we'd
ever
come
back.
B
So
a
lot
of
the
features
in
here
with
the
connection
pooling
you
can
just
put
in
a
couple
of
C
nodes
in
the
case
of
the
cluster
configuration
my
cluster,
which
mirrors
this,
has
five
nodes
in
it
running
a
u.s.
I've
got
got
one
through
five
I,
just
feed
it
without
one
or
two.
When
the
connection
happens,
it
actually
goes
on
auto,
discovers
additional
nodes
in
the
cluster
and
spreads
out
the
connections
across
those
that
is
super
super
valuable.
B
A
B
Readme
from
the
actual
driver
itself,
basically
just
set
up
the
call
sir
put
the
nodes
and
connect
to
it
issues.
Some
kind
of
query
do
something
with
it
in
and
out
dot
and
then
close
it
out.
So,
like
I
said
you
have
the
two
nodes
in
there
and
it
will
also
discover
the
rest.
So
three,
four
and
five
will
be
available.
They
will
be
queried
as
well.
Each
doesn't
have
to
put
them
in
the
configuration
setting,
no
reason,
I
put
one
and
two
and
there's
two
have
to
just
in
case.
B
One
is
offline,
so
some
of
the
key
features
that
I
use
and
I
think,
are
really
really
valuable.
It
didn't
exist
before
there
are
some
extended
cluster
attributes
in
this
particular
apple.
I'm
talking
about
I'm
not
using
the
key
space
one,
but
in
an
app
that
I
helped.
Someone
work
on
for
their
particular
use
case
is
doing
some
advanced
analytics
and
pulling
some
stuff
into
our
bizon
and
doing
some
work
there.
B
They
want
to
restrict
heavily
the
key
space
that
can
be
connected
to
from
the
gap,
so
they've
set
a
default
key
space
and
they
do
not
allow
people
to
get
to
where
the
code
is
executed
from
they
have
a
UI
wrapped
around
it.
So
this
one
key
space
is
the
only
thing
that
can
be
queried.
You
do
not
have
to
worry
about
people
bouncing
to
other
key
spaces
or
doing
anything
against
them,
since
they
can't
touch
the
actually
Python
code
itself.
They
don't
have
the
ability
to
do
anything
creative
about
it.
B
They
use
basically
a
gateway,
server,
the
roads,
web
services
and
built
a
little
you
Iraq
remit,
but
this
allows
their
analysts,
who
are
allowed
to
ca,
t
subset
of
data
that
lives
in
a
specific
Keith
base
to
only
be
able
to
call
that
from
within
the
code
itself.
So
you
basically
hard-coded
a
key
space
and
that's
the
only
key
space
will
be
workable.
B
It's
just
the
really
quick
version.
Just
in
case
you
don't
know
the
cassandra
nomenclature
t
spaces
at
the
top
level,
which
you
can
think
of
as
like
a
database
instance
from
a
relational
database
space
other
than
you
have
column
families
under
that
which
relate
to
the
relational
database
table
ives.
So
when
you
look
at
those
you're
restricting
it
to,
basically,
this
would
be
equivalent
in
the
sequel
environment
of
restricting
it
to
a
particular
database
instance,
so
that
you
can
control
we
got
to
what
from
the
application.
B
You
can
also
turn
on
a
lot
of
other
things
from
in
there,
there's
I
think
about
60
or
70,
and
you
could
look
up
with
different
configuration
options.
One
that
I
use
is
the
metrics
enable
the
controls,
metrics
collection,
whether
it's
on
or
off
the
SSL
off
hidden.
So
a
lot
of
the
environments
that
I
worked
in.
Why
it's
not
in
my
app
is
not
it
by
default,
turned
on
a
lot
of
environments.
B
Note
very,
very
important
in
certain
compliance
environments
and
certain
Network
boundaries
where
people
crossing,
especially
for
public
and
private,
makes
information
load
balancing.
This
is
by
far
and
away
my
favorite
feature.
That's
in
here
the
fall
for
that's
around
Robin,
but
it
can
be
configured
and
load
balancing
before
like
it
was
mentioning
earlier.
There
would
be
hot
spots
and
notes
for
queries,
so
a
query
would
be
issued.
B
It
would
hit
a
particular
set
of
nodes,
and
those
nodes
that
were
in
the
list
would
basically
get
hit
frequently
and
you'd
start
in
it
started
to
see
a
pattern
of
that
happening
and
getting
a
lot
of
load
on
a
particular
note
set.
So
with
this
without
having
to
write
a
bunch
of
interesting
things
and
bites
on
to
do
it
myself,
it's
already
pre
built
into
the
driver.
B
You
know
what
may
not
connect
ever
get
a
connection.
Failure
message
back
meet
with
the
old
way
of
doing
it.
This
way
you
can
set
the
timeout,
it
will
automatically
detect
the
failure
in
the
timeout
range
and
you'll
immediately
get
a
response
back
telling
you
that
it
timed
out
so
that
you
can
catch
that
error
and
let
the
client
know
there
was
something
wrong
rather
than
just
letting
it
sit
there
and
run
a
lot
of
times.
B
Make
a
mistake
running
a
query
as
I'm
sure
we
all
know
the
ROE
factory
piece
for
this
is
extremely
important
for,
like
the
splunk
specific
path
that
was
written,
but
it's
important
to
a
lot
of
the
people
that
I
talk
to
you
in
a
space
to
be
able
to
figure
out
how
the
data
is
going
to
be
returned.
What
the
format
is
going
to
be
from
the
result
rose
so
that
you
can
write
handlers
to
be
able
to
do
whatever
it
is.
B
This
will
let
you
know
exactly
what
how
the
rows
are
being
returned
and
in
what
the
format
is
for
those
rows,
so
that
you
can
handle
them
in
an
optimal
way
to
be
able
to
get
into
whatever
your
analysis
processes
as
quickly
as
possible,
so
obviously
the
options
returning
as
a
tuple
or
named
tuple
or
a
dict,
or
an
order
dick
the
old
way.
You
had
very
limited
choices
on
how
to
make
this
return,
and
you
had
to
handle
the
fact
that
sometimes
it
didn't
return
them.
B
The
lead,
like
this
I've
friend,
several
thousand
queries
through
this,
and
when
it
once
this
is
set
from
the
decoder
side.
Every
time
I
get
the
results
on
the
right
format
and
I
can
handle
them
and
I
never
ever
throw
exceptions
which
makes
me
very
happy
they.
One
of
the
customers
that
uses
this
verse
in
the
app
right
now
is
doing
a
lot
of
the.
B
Both
the
machine
data
stuff
is
coming
in
this
Blanc,
as
well
as
pulling
some
stuff
in
firmware
relational
databases
lookups.
So
if
you're
doing
that
that
many
separate
queries
with
basically
the
equivalent
of
a
join
to
get
all
the
data
set,
then
if
you
start
seeing
any
kind
of
performance
related
issues
or
issues
with
you
missing
records,
it
would
highly
impact
their
business
process.
The
way.
A
B
An
exception
or
that
it's
not
being
processed
at
all
and
worry
other
things.
That
is
great,
especially
for
my
particular
use
case.
But
again,
another
couple
of
customers,
I
know
they're
doing
stuff
with
a
driver
is
being
able
to
do
the
execute
async.
So
you
stole
the
execute
command
or
method
the
same
way.
You
used
to
there's
an
available
execute
async
command
here,
which
basically
will
let
me
don't
want
to
do
other
things
while
I'm
waiting.
A
B
That
you
can
actually
run
the
query
with
yay
Singh
command.
You
can
basically
create
a
response
object
and
then
you
go
through
and
walk
through
that
and
go
do
different
things
analytics
whatever
it
is
you're
doing
with
them,
but
you
can
process
them
as
they
come
through,
so
you
don't
have
to
wait
for
the
result
sets
and
a
blocking
up
to
do
it.
However,
if
you
want
a
force
blocking,
you
can
actually
call
the
object
and
force
blocking
on
it
too.
B
So
you
could
say
execute
this
is
a
sink
I
wanted
to
do
these
15
other
sayings
once
I
finish.
These
15
I
will
then
connect
to
the
object
and
force
blocking
to
go
ahead
and
wait
until
I
get
the
last
one
of
my
results,
so
you
may
want
to
build
somewhere
to
put
them.
You
may
look
the
metadata
that
you've
returned
from
command
and
you'll
build
something
else
to
move
the
data
to
you
may
do
you
know
a
number
of
different
things,
but
those
things
can
be
done
in
the
process.
B
Why
the
the
query
is
actually
being
executed,
whether
it's
a
read
or
write,
personal
matter,
but
why?
It's
being
executed,
you
can
go
to
other
things
in
your
code
and
then
come
back
so
also
super
important.
If
you're
doing
multiprocessing
approach
to
get
better
speeds
and
performance
times,
there's
a
bunch
of
perv
data,
that's
does
it
store.
Then
they
get
outside.
There's
a
link
to
it.
B
But,
more
importantly,
in
those
cases
is
because
it's
harder
to
find
java
developers.
These
days,
you
can
do
a
lot
of
the
work
at
the
Java
drivers,
so
the
pythons
I,
having
equivalent
functionality
or
even
better
functionality
in
some
cases
in
the
driver
and
being
able
to
you,
know,
use
a
skill
set
that
is
much
less
competitive
in
the
market
space
to
go.
Do
these
things
within
an
environment
really
important
so
again,
datastax
get
a
bio,
/,
Python
best
driver
index
2.html.
B
B
You
want
to
do
hi
through
ten
operations.
Multithreading
is
interesting.
Multiprocessing
is
far
better
but
make
sure,
if
case
you're,
not
a
Python
master
and
spent
lots
of
time.
Doing
this
made
sure
that
you
understand
the
implications
of
object
passing
with
multiprocessing
because
it
will
bite
you
there's
even
a
comment.
A
B
In
the
documentation
about
object,
passing
and
and
multiprocessing,
but
go
look
it
up,
go
google
it!
It
will
make
your
life
painful.
If
you
try
to
do
multiprocessing
and
try
to
pass
objects
between
gold,
it
will
not
go
well
for
you.
So
do
pay
attention
to
those
caveats
and
use
multi
process
so
how
I
use
it
my
github
site
down
here
this
one
of
the
standard
app.
Unfortunately,
it's
still
the
dot
8
version
which
relies
on
the
old
drivers,
which
will
show
you
exactly
what
you
don't
want
have
to
do
anymore.
B
The
dot
two
versions
will
be
its
imminent.
I
am
waiting
on
a
couple
of
test
results
from
customer
added.
Do
a
couple
updates
to
it
recently
and
they
they're
doing
some
final
testing
so
I'm
just
waiting
for
them
to
get
time
to
get
back
to
me
to
make
sure
that
everything
was
fixed
and
we've
got
all
the
bugs
out
because
of
the
fact
that
this
is
going
to
end
production.
Most
people,
environment
I
want
to
make
sure
that
everything
is
fully
tested
narrated
ago.
B
I'll
also
make
sure
that
I
get
it
into
the
swung
App
Store
as
well.
Once
the
diversion
sees
about,
but
this
new
approach,
the
code,
it's
got
rid
of
about
40
or
50
lines
of
code
from
the
scripts
heard
a
whole
bunch
of
the
exception
and
corner
case
stuff
cause.
We
could
do
the
row
factory
setup
so
with
all
those
things
it's
made
it
much
better.
It's
much
much
faster
I
did
not
implement
multiprocessing
in
there
on
purpose,
but
that's
mostly
because
I
am
a
little
afraid
of
how
someone
might
use
it.
A
A
B
B
To
do
this,
if
you
want
it,
you
can
play
around
with
it,
you
can
do
it
you're
a,
but
I
won't
be
checking
that
into
the
get
on
site.
Just
because
of
what
might
happen-
and
I
would
definitely
recommend
building
your
own
playground,
so
you'll
build
grab
virtual
box
and
put
a
three
node
cluster
even
on
your
own
macbook.
It
works
perfectly
fine.
I
do
it
all
the
time
I
have
an
external
USB
drives.
It's
got
all
my
VMS
on
it.
B
I
grew
them
are
virtual
box
and
I
can
do
a
free
to
cluster
I
could
do
purse
testing
and
I
can
go
test
everything
and
it
doesn't
impact.
You
know
the
normal
workload
I'm
doing
for
development
to
build
an
app
or
whatever
else,
building
integration
do
something
custom.
So
I
was
just
spin
up.
Your
own
thing
go
play
with
it.
Look
at
all
the
various
tools
are
available
for
Python.
B
B
Them
in
because
they
have
the
Python
skill,
they
could
do
a
lot
of
this
stuff
do
not
have
to
write
a
lot
of
Java
code
to
make
it
work.
I
have
a
couple
other
apps
I'm
working
on
they're,
not
related
to
swamp,
for
this
I'm
working
on
them
for
set
of
customers,
but
as
soon
as
they're
done
and
cleaned
up
and
take
all
the
stuff
outlets
can
relate
back
to
customer
object
is
checked
into.
B
A
Thank
You
Eddie
very
much
indeed
so
we
do
have
some
questions
coming
in,
but
just
a
reminder.
If
you
would
like
to
ask
Eddie
a
question
right
now:
please
use
the
Q&A
tab
inside
of
WebEx,
so
we
will
get
through
as
many
questions
as
we
can
Eddie.
First,
one
nitesh
asks
once
we
install
state
start
enterprise.
All
Cassandra
we
get
or
Python
requires
us
to
be
able
to
run
cql.
A
B
That's
the
plain
color
driver,
the
difference,
I
500
words
and
the
link
to
is
not
felt
by
default.
So,
yes,
you
will
have
to
install
the
driver
separately
and
there's
a
couple
dependencies
on
there,
but
they're
pretty
simple
things.
None
of
them
are
ones
that
you'll
have
to
worry
about
security
reviews
in
a
that's
before
so,
yes,
it
has
to
be
installed
separately.
It
is
a
totally
separate
component
and
does
not
come
with
either
the
commander
or
the
DfE
install.
A
B
B
Looks
very
different,
I
mean
a
strannix
is
built
for
a
lot
of
particular
functionality
to
Netflix
that
a
lot
of
people
have
adapted
you
different
things.
With
its
it's
more
of
a
heavyweight
app
that
lives
in
front.
The
Python
driver
is
strictly
a
driver
layer
with
an
API
so
that
you
can
call
directly
go
to
Dublin,
have
nearly
the
functionality.
B
It's
not
built
for
a
lot
of
the
advanced
stuff
that
aster
annex
was
built
to
handle,
but
it
is
just
basically
a
low-level
driver
to
let
you
use
the
Python
API
and
call
down
to
the
driver
and
let
the
driver
handle
all
of
the
complications
and
intricacies
of
connecting
to
a
cluster.
So
there's
very,
very
different
y
con
drivers
is
a
specific
driver
functionality
just
to
give
you
basic
control.
A
strannix
is
more
of
an
application
built
on
top
to
handle
a
lot
more
advanced
functions.
So
that's
my
take.
A
B
Yeah,
probably
not
the
best
named
function
to
be
honest,
yeah,
that's
that's
great
feedback
for
Tyler
to
change
it
to
something
else.
Really
all
clustered
got
shut
down
does
to
shut
down
all
open
connections
to
the
cluster,
so
specifically,
if
you're
doing
multiprocessing
or
multithreading,
depending
on
what
you're
more
comfortable
list.
If
you
want
to
close
down
all
the
sessions
to
that
cluster,
all
at
once,
no
matter
where
they
happen
to
be
calling
cluster
shutdown
shuts
down
all
clustered
instance
connection,
it
does
not
actually
shut
down
the
cluster,
all
those
again,
how
that's
scary,
cuz.
B
B
A
B
Well,
so
the
quick
answer
that
is
not
now
that's
one
of
the
most
requested
features
for
the
app
and
I.
Don't
know.
If
there's
I
can
mask
Tyler
you
we
can
think
Tyler
and
find
out.
It
is
one
of
the
most
requested
features,
though,
and
I
know
it's
on.
The
list
of
apps
or
extensions
that
are
going
to
happen
is
to
build
the
I
think
it's
been
whistling
front,
page
I'm,
looking
to
see
if
there's
anything
in
the
JIRA.
B
Yes,
it's
listed
as
becoming
next.
The
features
to
be
added,
see
extensions
for
encoding
decoding
messages,
so
it
is
coming.
It
is
highly
requested,
I,
don't
know
when
it's
coming.
I
don't
see
cheer
for
it,
so
I
couldn't
tell
you
when
it's
going
to
happen
other
than
to
paint
tyler
and
maybe
follow
up
after
this.
So
if
you
want
to
email
me,
I
can
follow
up
and
find
out
from
tyler
how
far
along
that
is,
but
I
know
it's
highly
requested
so
being
able
to
use
the
C
extensions
is
definitely
in
the
plan.
B
Yeah,
so
it
actually
adds
all
the
failover
stuff
built
into
the
driver,
so
it
handles
payloaders
for
you,
it
will
mark
those
dead
and
then
it
will
reconnect
two
nodes.
It
is
when
the
API
documentation
and
Tracy,
where
it's
at
Cassandra
policies.
So
if
you
look
at
me,
api
documentation
online
and
go
into
the
Cassandra
qualities,
load,
balancing
stuff
which
I
talk
about
I,
didn't
talk
about
the
rest
of
it,
but
there's
a
capable
of
your
marcos
up
and
down
and
then
reconnecting
to
that
house.
So.
B
Recovery
stuff
is
built
into
the
driver
itself.
That's
why,
when
it
reconnects
the
next
time
it
actually
goes
and
interrogate
the
cluster
again
to
find
out
what
nodes
are
up.
You
can
also
set
reach
our
policies
in
groups,
so
there's
a
whole
lot
of
advanced
things.
I
didn't
talk
about,
unfortunately,
bit
there's
a
lot
of
policies
that
you
can
set
and
a
lot
of
them
are
related
to
reconnecting
to
the
dead
host
about
downs
and
then
making
sure
load
balanced.
He
is
working
and
making
sure
that
you
autodiscover
the
notes.
B
A
B
Well,
it
can
guess
what
we
work
with
Django
I,
don't
know
the
answer
on
ORM
I
haven't
tried
it.
So
I
can't
tell
you
but
I
know
absolutely
there.
The
the
front
end
that
I
hope
the
customer
build
is
all
tango
base
because
they're,
the
big
Django
user.
So
my
guess
is
yes,
but
I
can't
say
for
sure
the
ORM
works
because
I've
never
actually
tried
it,
but
it
definitely
works
with
Django.
That
is
a
front-end
that's
been
built.
B
I
will
see
if
I
can
get
folks
reverted
to
open
the
Django
one
and
just
push
it
off
somewhere.
They
may
or
may
not
their
challenge.
Sometimes
a
you
will
like
to
go
out
if
it
adds
value
for
them
so
but
I'll
ask
if
they
can
all
you
know.
If
you
want
to
email
me
again
and
for
follow-up
or
you
know
you
direct
messaged
me
on
twitter,
happy.
A
A
Okay,
great
his
question
that
I
will
answer
all
so
we
get
this
question
every
time
and
I
should
have
addressed
it
up
front.
So
every
webinar
we
do
is
recorded
and
made
available
by
a
planet
Cassandra.
So
this
webinar
today
and
all
previous
webinars
are
available
there.
This
webinar,
along
with
any
slides,
will
be
available.
We
usually
try
and
turn
things
around
very
quickly.
So
within
about
24
hours,
we
will
post
them
to
planet.
A
B
So
it
will
discover
the
entire
cluster
so
that
universities
would
be
discovered.
As
part
of
the
note
now,
it
uses
the
distance
to
decide
which
ones
to
query
first
and
the
load
balancing
if
you
set
it
up
that
way,
there
are
policies,
don't
be
set
it
up
correctly
to
make
sure
that
you're
hitting
the
local
nodes
first,
but
it
will
indeed,
if
I
note
that's
local,
isn't
available,
it
will
take
the
further.
A
B
It
will
tell
you
how
to
set
then
all
under
Glee
that
one's
under
cluster,
rather
than
under
policy
that
that's
under
the
cluster
settings,
but
yeah
yeah
it's
under
cluster
settings
it,
but
it
does
that
automatically.
So
when
it
sets
up
the
connection
pool,
it
looks
at
the
distance
setting.
So
as
long
as
that
ball
configured
correctly
in
your
cluster,
it
will
go
to
the
local
first,
but
it
will
always
cover
the
entire
cluster.
So
worst
case
scenario,
every
local
mood
goes
down.
You'd
still
be
able.
A
B
A
Okay,
well,
I,
think
I
think
we
have
I
think
we've
gone
through
all
of
these
questions
now
so
Eddie.
Thank
you
very
very
much
for
joining
us.
Today.
We
have
two
upcoming
webinars
to
remind
everyone
about
Boris
wolf
from
comcast
will
be
talking
about
some
of
the
things
he's
been
working
on
at
the
comcast
Silicon
Valley,
Innovation,
Center,
building
a
message,
bus
on
top
of
Cassandra
and
then
Aaron,
Morton
and
Patrick.
Mcspadden
will
be
here
on
me
15,
and
they
will
be
talking
about
designing
a
Cassandra
solution
both
of
those
webinars
great.
A
If
you
would
like
to
submit
a
an
abstract
to
speak
at
the
summit,
we
have
exactly
one
month
left
with
the
call
for
papers
open.
You
can
find
that
link
on
the
the
datacomm
website
as
well
so
Eddie.
As
always,
it's
a
pleasure.
Thank
you
so
much
for
joining
us
and
thank
you
for
your
continued
support
of
Cassandra.