►
From YouTube: The Apache Cassandra® Corner podcast w/ Aaron Morton
Description
Multiple "Aaron" alert! Listen in as Aaron sits down with Aaron Morton to talk about his upcoming talk at the Cassandra Summit in December. They also discussed Aaron's work with the Mongoose project, and improving the Cassandra developer experience with a JSON API that includes Vector Search!
A
A
Hey
welcome
back
everyone
to
the
catchy
Cassandra
Corner
podcast.
Once
again,
I
am
joined
by
Aaron
Martin
Aaron.
Welcome
back
to
the
show.
How
are
you
doing.
B
Good
to
see
you
again
Aaron,
thank
you
for
having
me
it's
I
was
talking
with
my
wife,
the
other
day.
I
was
doing
this
and
then
I
kind
of
sorted
our
names
through
to
like
the
beginning
of
their
last
name.
I
was
like
I'm
still
in
front
right,
like
a
a
we're
on,
but
I
I've
got
M
at
the
end
right
right.
A
Oh
that's
funny
yeah,
so
so
for
the
the
folks
at
home,
who
haven't,
haven't
seen
the
other
episode
of
the
unit.
Why
don't
you
tell
us,
you
know
real
briefly,
just
a
little
bit
about
yourself.
B
Cool
so
I've
been
working
in
and
around
the
patch
against
Sandra,
since
2010.
I
started
playing
with
it
around
version,
0.3.4
in
2010
and
then
in
March
of
2011
sort
of
left.
My
job
here
in
New
Zealand
working
at
a
company
called
weather
digital,
which
is
the
the
big
VFX
house
right,
the
company
that
makes
Avatar
and
all
those
things
left
there
and
became
a
professional
Cassandra
person
in
March
for
2011,
and
there
wasn't
many
around
ran.
B
You
know
ended
up
with
a
company
with
a
stupid
name
called
the
last
pickle
did
that
for
nine
years
and
then
joined
data
Stacks
at
the
beginning
of
2020
so
clearly
knew
all
the
data
Stacks
peeps,
all
right,
yeah
yeah,
so
more
than
before
there
was
the
latest
accent
and
being
around
them.
So
yeah
yeah
be
the
commit
I've
been
on
the
PMC
and
and
nowadays
just
do
a
lot
of
work
on
trying
to
improve
developer
experience
and
apis
and
things.
A
Oh
excellent,
excellent,
yeah
yeah
we're
we're
definitely
going
to
get
into.
You
know
some
more
about
the
improving
the
developer,
experience
for
sure
and
and
yeah
it's
it's
interesting
that
you
you
talk
about
it
from
going
that
far
back,
it's
like
I
mean
you
know
better
than
anyone
about
how
the
what
the
develop.
What
the
developer
experience
looked
like
and
say,
like
Cassandra
1.0,
you
know
when
we're
working
with
Thrift
and
yeah
yeah.
B
So
yeah,
so
to
begin
with,
it
was
like,
if
you
remember,
like
configuration
and
schema,
was
in
a
file
right
like
bind
your
column
families,
if
you
couldn't
even
do
create
column
family,
you
would
just
put
it
in
a
file
and
it
would
load
that
and
you'd
go
so
I've
got
the
thing
and
then
you've
got
and
we
had
Thrift.
Of
course,
yeah
I
mean
Thrift
challenges
as
a
project
back
then
it
didn't.
B
It
wasn't
operating
well
as
a
project,
but
it
also
kind
of
boxes
into
a
corner
that
that
made
it
hard
to
even
just
talk
about
one
person
explained
to
someone
else.
What
they
were
doing
with
well
ever
did
was
you
could
say,
could
you
look
at
my
query?
Please
and
someone
else
would
be
able
to
understand
it,
and
then
you
could
have
a
conversation
about
what
it
did.
You
look
at
my
Ruby
code
and
then
someone
goes
no
I,
don't
know
how
to
do
Ruby,
so
you
can't
get
any
help
so
right,
right.
A
So
now,
with
the
work
that
you've
been
doing,
you
know
why,
don't
you
tell
us
a
little
bit
about
you
know
like
Json
and
Cassandra,
and
and
maybe
about
the
API
layer
that
you've
been
working
on.
B
Yeah
I
think
we
we
touched
on
this,
maybe
back
in
March
this
time
we
previewed
it
around.
The
was
called
like
Cassandra
fullwood
that.
B
Earlier
right,
yeah,
and
definitely
going
to
be
talking
about
this
at
the
Cassandra
conference
in
December
over
to
that
going
back
to
back
to
the
States.
For
that,
so
you
know
we
know
that
Jason's
a
really
important
part
of
the
developer
ecosystem,
but
it
it's
a
transport.
It's
the
things.
It
allows
you
to
do
the
flexibility
that
allows
you
to
do
that,
that's
great,
and
when
we
look
at
the
way
the
people
in
the
node.js
world
work,
a
lot
of
them
and
I
mean
a
lot
of
them,
use
an
object
document.
A
That's
crazy.
You
know,
I
had
I
had
Val
karpov
on
the
show
like
right
after
you
back
in
back
in
March
and
yeah
I
mean
we.
We
talked
about
similar
numbers,
so
you're
telling
me
we
had
like
another
million
downloads
of
that
thing,
since
I.
B
B
We
need
some
type
of
swim
Lane
to
understand
how
people
use
this,
how
people
want
to
score
it
query
it
modify
it
and
we
looked
at
Mongoose
and
said:
let's
make
the
best
back
end,
we
can
for
mongoose
Let's.
You
know
understood
it.
Has
a
fluent
query.
Builder
attached
to
it
called
M
query,
but
it
talks
about
schema
it
under
it.
You
know
constructs
queries
that
it
wants
to
send
to
a
back
end.
We
can
understand
the
schema.
We
can
understand
what's
important
to
node.js
developers.
What
features
are
less
important.
B
We
can
understand
the
types
of
queries
for
filtering,
the
projections
that
you
need
to
support
and
you
know,
and
the
types
of
update
operations
that
it
expects
sure
if
you're
you
know
a
node.js
Dev
doing
this.
Now
it's
going
to
work
for
node.js
if
it's
going
to
work
for
any
other
language
as
well.
What
we
did
was
that
if
we
make
a
really
great
back-end
for
mongoose,
this
will
be
really
useful
to
the
Mongoose
people
and
useful
to
other
people
as
well.
So
it
basically
gave
us
absolutely
yeah.
A
B
A
That
sounds.
That
sounds
super
exciting,
so
you
know
it
now
in
order
to
in
order
to
make
this
this
API,
you
know
a
reality.
You
know
you,
you
kind
of
needed
some
newer
things
inside
of
Cassandra
right,
some
of
some
of
the
newer
features
of
5.0
are
are
being
utilized.
You
know
with
this
with
this
API
layer,
you
know.
So
maybe
you
could
talk
a
little
bit
about.
You
know
like
how
you're,
using
like
storage
attach
indexes
or
like
kind
of
like
the
new
operators,
a
little
bit
yeah.
B
Yeah,
could
you
go
in
a
little
bit
of
that,
so
we're
kind
of
using
old
stuff
using
old
things
in
Cassandra
and
things
that
are
new
and
things
that
will
be
new
soon
go
in
the
Spectrum
we
created
this
idea
called
super
shredding
right
hand,
look
at
the
Json
document
that
you've
given
us,
and
we
understand
the
types
of
querying
filtering
that
you're
going
to
be
able
to
run
against
that
right.
B
You've
got
to
be
able
to
say
where
username
equals
error
and
you
might
want
to
say
contains,
but
you'll
find
all
the
documents
that
have
the
human
tag,
the
human
keyword
in
them,
but
don't
care
about
the
value
you
might
want
to
do
some
operations
understand
the
size
of
an
array
and
things
like
that
right
right.
So
the
inside
here
was
to
say
we
know
all
the
read
patterns
that
you're
going
to
want
to
run,
and
so
let's
shred
this
document
and
store
it
in
a
way
to
serve
those
read
pens
and.
B
So
what
we
did
was
use
the
collections
in
Cassandra,
which
I
know
is
a
bit
of
a
dirty
word.
You
know
they
don't
like
classic
classic
Cassandra.
Let's
just
get
go
on
and
off
disc
is
absolutely
fast
as
we
can.
You
know,
and
not
not
big
on
collections.
They
also
have
some
issues
with
ranged,
tombstones
and
things,
but
I'm.
B
Tried
them
all
the
time
and
and
some
things
and
we
think
the
functionality
it
gives
us
is
great
and
Sai
indexes
can
index
a
map
and
a
set
and
they
index
it
either
on
the
keys
or
the
values,
but
also
on
the
entries
in
the
map.
So
you
can
walk
up
to
a
map
like
you
could
have
a
row
in
Cassandra
that
has
a
map
of
just
string,
string,
yeah.
B
The
query
that
says
find
me:
you
know
all
the
rows
where
the
map
has
username
equal
to
Aaron,
oh
wow
right
so
I
could
do
that
for
doubles,
and
I
could
do
that
for
Day
times
and
I.
Can
my
hair
this
set
and
I
put
all
of
the
all
of
the
the
Json
parts
that
were
in
your
document
into
that
set.
I
could
walk
up
to
the
cql
and
say
find
me
all
the
rows
where
the
exists?
Where
is
human
exists
in
this
set
right
right?
B
The
Sai
which
we've
had
at
data
Stacks
will
be
pushing
to
get
into
open
source
for
a
while
now,
just
with
the
Slowdown
for
four
and
five
all
right
I,
you
know
which
to
us
is
a
little
bit
old,
but
he's
new
and
and
covered
in
cep.
Seven,
you
know
no
matter
what
it
is
going
in
there.
Some
enhancements
will
bring
into
Sai
there's
some
enhancements
that
you'll
also
see
coming
into
Cassandra.
B
There's
a
cep.
It's
called
cp29,
which
adds
kind
of
not
at
ore
and
logical
precedence
tree
right,
so
it
can
construct
more
complex,
wear
conditions.
B
B
B
For
sure,
but
that's
we
know
those
people,
we've
been
those
people
over
the
years
you
at
Target
and
all
this
stuff
for
it
yeah
yeah.
B
This
is
about
functionality,
so
those
elections
Sai
and
the
enhancements
that
are
coming
to
those
enhancements
in
The
Logical
logical
operations
you
can
put
in
the
where
Clause
that
bring
more
complexity.
Do
you
hear
that
and
then
a
chord
coming
to
give
us
improvements
in
lightweight
transactions?
So
you
look
at
the
tour
right.
B
The
the
cep15
you'll
see
these
really
fancy.
Looking
like
begin
trans
do
a
read:
have
a
condition.
You
know
abort
yeah,
one
remember
when
we
started
you
know
talking
to
people
about
how
Cassandra
works
and
just
explaining.
Well,
there's
no
abort
right!
If
you,
if
you
wonder
if
you
write
two
things
well,
yeah
one
of
them
Works
while
them
doesn't
you're,
just
gonna
have
to
live
with
it.
That's.
B
Up
yeah,
so
so
call
it
comes
in
with
that
and
that's
the
great
part,
but
it
also
can
be
the
the
implementation
for
the
compare
and
set
operations.
B
So
if
you're
doing
you
know
insert
if
not
and
and
update
if
those
operations
can
be
driven
by
a
chord,
they'll
be
faster
and
they're
important,
because
going
back
to
that
first
comment
about:
what's
the
experience
that
node.js
and
mongoose
builds?
Oh
sure,
something
called
like
find
one
and
update
there's
a
complex
updates
structure
right.
You
only
do
like
find
the
document
pull
it
back,
make
these
changes
and
write
it
lwts.
We
have
to
use
that
so
we're
going
to
pushing
Cassandra
in
ways
that
you
know
five
five
years
ago.
A
Yeah,
well
that
just
that
just
all
sounds
amazing
and
you're
right.
It
changes
Cassandra
and
you
know,
like
some
very,
very
profound
ways.
It
certainly
does
open
it
up
quite
a
bit
and
yeah
yeah
I
could
see
that
I
could
definitely
see
that
improving
again,
that
that
whole
developer
experience
aspect
where
you're
trying
to
build
something-
and
maybe
you
need
you
know
a
where
Clause
that's
a
little
more
complex
than
you
know
what
you,
what
you'd
normally
run
in
Cassandra.
Maybe
you
don't
need
you
know
five
millisecond
response
time.
You
know
that's
yeah,
that's
that's!
A
Okay,
yeah
yeah
I
was
going
to
say
too
that
that,
with
kind
of
all
the
hype
lately
that's
been
around
Vector
search.
Have
you
looked
at
that
at
all
for
the
for
the
API.
B
A
A
B
We're
exposing
the
power
of
Cassandra
and
the
capability
of
Cassandra,
but
through
this
different,
API
right
and
so
vectors
are
a
part
of
Cassandra,
yep,
yeah
I.
Don't
think,
there's
any
argument
around
that.
You
know
factors
are
a
part
of
Cassandra
who's,
going
it's
in
Cassandra
five
and
it
does
a
pretty
good
job.
I'll
put
it
too
yeah,
yeah
I.
Think
one
of
the
things
we
learned
over
the
years
is
that
it
takes
a
long
time
to
bed
down
a
database,
a
new
database
right,
yeah,
yeah,
18
years
to
get
a
bulletproof
database.
B
A
B
Yeah
so
yeah
you
can
you
know
in
your
documents
that
you
saw
and
if
we
got
time
I
think
we
might
pull
up
just
a
little
example
of
this.
You
can
just
store
a
vector
when
you,
when
you
define
your
collection,
you
say
I
got
a
vector.
It's
going
to
be
this
long.
You
can
store
your
vector
in
there
and
then
use
that
in
the
sort.
So
we've
been
working
with
Valeri
the
maintainer
of
oh.
B
We
work
with
him
to
see
how
we
can
make
this
be
just
feel
that
it's
at
home
in
the
Mongoose
experience
in
the
node.js
experience
and
yeah
the
tides
saying
things
you
can
do
with
Cassandra,
do
these
through
the
Json
API.
Do
them
using
Mongoose,
do
them
in
JavaScript
in
node.js,
and
it's
very
very
easy.
A
Oh
excellent
excellent
did.
Did
you
maybe
want
to
show
your
code
example
now
just
so,
we
could
maybe
walk
through
some
of
that
and.
B
Yeah
for
those
that
take
a
step
back
just
for
those
that
don't
don't
know,
Mongoose,
don't
know.
Node.Js
Mongoose
is
an
object
document
mapper,
you
define
your
schemas
runs
in
node.js
and
then
you
are
just
coding.
Gets
it
like
an
orm
yeah.
A
B
Normally
have
so,
let
me
share
this
screen,
yeah
I'm,
about
to
show
some
code
that
is
talked
about
in
a
blog
post,
that
I
think
we'll
put
out
with
the
socials
and
we'll
include
in
the
in
the
show
notes
or
something
like
that.
So
there's
a
blog
post.
That
explains
this
this
thing.
So
this
is
a
photo
sharing
site
that
one
of
our
new
starts
made
see
here:
they're
using
mongoose
and
they're,
defining
a
schema
for
a
photo.
You
just
load
uploaded
all
of
his
his
own
photos.
B
Here's
where
the
vector
is
defined.
That
and
he's
just
got
a
mongoose
validation
on
here,
because
he's
saying
that
this
is
a
vector
of
size
1536.
So
when
this
collection
is
made,
a
configuration
goes
across
and
then
he
also
has
one,
for
this
is
a
vector
based
on
the
description
and
we
have
one
over
here.
It's
an
embedding
of
the
actual
photo
itself
and
then,
when
you
want
to
go
and
store
an
image
we
go
down
here,
add
photo
so
we're
just
going.
B
Hey
I'd
like
to
get
a
new
photo
object
and
he
sets
up
the
properties
on
that.
We're
calculating
the
embedding
here
on
the
client
side,
storing
that
as
the
vector
field,
and
then
we
just
go
save
so
very
easy
way
to
get.
Data
in
this
is
the
same
here
for
searching
by
photo
similarity
and
then
on
this
site
and
again
we'll
Link
in
the
blog
post,
we'll
link
the
blog
post.
B
Here
we
can
do
a
search
here
by
the
content
and
here
we're
adding
in
a
sort
by
vector
and
so
we're
saying
photo.5
in
the
category
that's
chosen
and
this
Vector
description
as
to
what
what
sort
of
photo
you
want
to
get
so
in
this
examples,
I've
used
this
and
said
hey.
Can
you
find
me
photos
of
things
that
cows
would
eat?
It
will
come
back
with
photos
where
the
description
has
something
about
grass
or
things
like
that.
We've
also
use
this
for
the
search
by
photo
similarity.
B
So
this
is
basically
the
same
as
Google
search
by
image,
so
you
upload
a
photo.
We
calculate
a
vector
and
then
we
again
we
do
a
source
yeah
and
we
can
pull
back
images
that
look
similar
to
others
using
the
embedding.
So
the
great
thing
here
is
you've
got
all
the
power
of
Cassandra,
with
a
very
simple
way
to
interact
with
it
in
a
way
that,
for
node.js
developers
feels
very
much
just
the
way
they
like
to
do
things.
A
You
know
one
thing
I
noticed
too
is
that
you
know
what
my
node.js
is
a
little
rusty,
but
you
know
I
did
see
you
know
a
few
awaits
in
there.
So
is
this
working
in
like
a
like
a
reactive
sense
or
yeah?
Could.
B
You
go
look
at
that
so
again,
my
like
my
JavaScript
node.js
is
a
little
rusty
as
well.
So
yeah
this
example.
It's
a
it's!
A
next
JS
app
we're
looking
at
the
service,
we're
looking
at
the
server-side
controllers
that
that
the
next
JS
app
is
routing
things
to
and
then
making
calls
through
Mongoose,
which
has
got
all
the
await
syntax
on
it
that
doesn't
have
callbacks
totally
deprecated
out
of.
B
Then
that's
calling
us
over
on
on
our
side
over
the
the
the
HTTP
base
API
and
hitting
our
service,
making
the
call
so
yeah.
In
that
sense,
it
fully
integrates
into
how
node.js
developers
are
working
because
Earl
Harry,
wouldn't
let
us
do
anything
other
than
be
very
good
at
what
we're
doing.
A
Yeah
I'm
sure
I'm
sure
yeah
yeah.
So
now
you
have
a
talk
coming
up
at
the
Cassandra
summit.
Tell
us
a
little
bit
about
that
before
we
get
going
yeah.
B
So
that's
it
in
December
and
yeah,
going
I
think
the
title
is:
is
around
super
trading
and
storing
Jason,
so
it's
going
to
be
going
through
from
a
Cassandra
perspective,
how
we
were
able
to
do
this.
The
new
technologies
we're
relying
on
and
the
new
technologies
that
we're
going
to
bring
in
and
I
think
this
is
a
win-win
for
a
cql
as
well.
Yeah
we've
seen
the
C
E's
there
that
we
talked
about.
B
This
is
all
about
allowing
you
to
do
more
complex
things
and
seek
you
well
that
allow
you
that
kind
of
remove
a
little
bit
of
this
partition-centric
view
of
the
world.
Always
you
know,
that's!
That's
there
that,
as
we've
seen
it,
the
highest
performance
will
come
from
understanding
how
to
make
Cassandra
go
fast,
but
a
bunch
of
this
stuff
is
going.
There
are
times
when
we
need
more
functioning,
we're
leaving
a
little
bit
less
performance.
A
B
So
your
laptop
is
melting
down,
I'm
going
to
guess,
because
I
can't
you
much
of
what
you're
saying
at
all
so
I'm
gonna
say
you
said:
hey
thanks
great
good
to
talk
to
you
again
as
you've
got
an
awesome
name
and
looking
forward
to
the
Lord
to
the
Json
API,
which
will
be
in
September
in
the
public
beta
and
then
GA
later
in
the
year.