►
From YouTube: This Week In Cassandra 2/5/2016
Description
Link to blog post discussed: bit.ly/1mhskwB
A
Are
live
all
right,
hello!
Everyone
welcome
to
our
first
this
week
in
Cassandra,
we're
going
to
be
talking
about
just
basically
stuff,
that's
happening
in
the
Cassandra
ecosystem.
Over
the
last
week,
we're
going
to
be
taking
a
look
at
blog
posts,
new
projects
in
the
in
the
blog
post
itself
that
accompanies
this
video.
You
can
see
any
new
job
postings
and
we're
going
to
take
a
look
at
some
of
the
JIRA's
and
merges
that
have
happened
in
the
open
source
code
base
of
Cassandra
itself,
so
this
is
in
particular,
really
exciting
for
us.
This
is.
A
B
A
A
bit
yep
and
we've
you:
can
you
can
there's
a
Q&A
tab
if
you're
looking
at
the
Hangout
you're
watching
it
live,
you
can
ask
questions
just
be
aware
that
there's
a
chance
that
we
might
not
get
to
it
in
time
or
might
show
up
a
little
late.
The
Hangout
thing
to
be
a
little
bit
weird.
So
if
we
don't
answer
your
questions
in
the
Hangout,
we
can
answer
them
on
Twitter
I'm,
and
that
will
you
know.
Maybe
other
people
have.
A
A
B
B
Absolutely
so,
if
you
look
at
what
he's
talking
about
it's
really,
there
are
some
really
fundamental
points
in
there,
as
if
you
do
data
modeling,
you
know,
of
course
you
need
a
partition
key,
but
then
this
is
a
more
of
a
Y
and
then
the
downside
of
what
you
wouldn't
and
actually
I
was
I
saw
some
really
cool
stuff
in
there
like
how
he
talks
about
secondary
indexes,
which
I,
don't
think
a
lot
of
people
consider
when
you're
using
a
secondary
index
and
a
partition
key.
You
do
get
some
benefit
from
that
yeah.
C
C
This,
the
data
model
I
ended
up
with
because
I
needed
those
guarantees
ended
up,
trying
to
use
single
partition
operations
and
I
even
did
a
meet-up
talk
on
it,
because
it
was
kind
of
an
interesting
thing
that
people
sometimes
miss.
You
know,
especially
if
you're
a
newcomer,
so
very
good,
very
good
information
for
sure
yep.
C
B
A
B
I
have
time,
yeah
knows
an
old
friend,
but
it's
an
old
hero
of
the
public
right
but
I,
see
annex
was
around
I
think
that
was
one
of
the
first
really
big
contributions
from
netflix
and
the
Cassandra
open
source,
because
it
was
a
really.
It
was
a
bit
of
a
wild
west
with
drivers,
and
this
is
for
years
ago
five
years
ago,
and
it
was
their
internal
driver
which
validated
a
lot
it,
but
he
did.
It
was
a
thrift
driver
and
of
course,
when
that
was
the
gist
of
the
article.
B
Is
it
now
that
thrift
is
well?
I
guess
is
deprecated
now
is
then
it
will
give
them
a
doubt
frozen,
there's,
no
reason
for
them
to
continue
spending
time
working
on
it
when
they're
moving
forward
the
different
different
plan
yep,
give.
B
Was
interesting
because
Ilan,
who
was
one
really
I,
think
one
of
the
main
drivers
for
the
asti
annex
he
and
I
were
talking
once
and
he
was
the
first
one
to
turn
me
on
to
this
idea
of
reusing
a
token
aware
driver
and
when
he
told
it
to
me,
is
one
of
those
things
or
I,
that's
kind
of
obvious.
I
guess
now
that
I
think
about
it.
Yeah
just
kind
of
like
shocked
me
for
a
minute.
I
had
to
stop
eating
my
pizza.
A
Other
driver
and
yeah
eliminating
the
eliminating
extra
network
hops
is,
you
know
whenever
you
can
take
advantage
of
that,
it's
absolutely
yeah
yeah.
So
what
else
we
got
here,
multi-threaded
program
to
count
rows
of
your
Cassandra
table:
hey
I,
like
a
program,
I
love
a
good
project.
That
does
one
thing
exactly.
A
B
You
know
that
that
is
an
interesting
problem
to
me.
Right
is,
and
so
Brian
and
I
were
actually
talking
about
this
before
he
did
this,
he
was
getting
ready.
The
funny
background
on
this
is
getting
ready
to
take
a
cross-country
flight
and
he
needed
a
project.
Yeah
like
you
know
about
that
jug
had
he
was
he's
like
what
is
what
does
something
stupid?
That
is
not
easy
to
do
and
counting
how
many
rows
yep
yep,
that's
their.
A
Justification
opportunity-
that's
perfectly
so
I
get
to
do
a
shameless
plug
of
my
own
blog
post,
working
working
with
Python
and
Cassandra,
and
the
object
mapper
right
now.
There
is
no
way
synchronous,
there's
no
asynchronous
programming
behind
it.
You
can't,
let's
say,
create
multiple
rows.
You
know
asynchronously,
it
has
to
be
synchronous
process.
I
kind
of
just
show
how
you
can
do
all
this
stuff
with
G
event.
The
underlying
driver
supports.
A
B
B
C
Big
pain
in
the
king
of
the
button
till
they
I
mean
it.
You
know
its
continuation
style
programming.
You
know
like,
like
callback
held
from
you,
know
from
JavaScript
some
other
languages
like
that
until
they
till
they
added
a
sink
and
await.
So
it's
funny
they're,
actually
thinking
about
having
those
keywords
to
to
JavaScript.
Now
too,
but
I
didn't
yeah,
I,
guess
not
being
a.
C
A
A
That's
it's
not
very
fun
or,
like
you,
write
all
these
like
little
functions
up
top
and
it
just
it
gets
really
hairy
this.
This
is
nice
because,
with
with
G
event,
it
just
patches
the
standard
library
in
Python
and
yields
on
any
I
oak
all
right,
so
you
can
have
these
micro
threads
and
basically
be
like
constantly
doing
stuff
actually
using
your
CPU
instead
of
just
waiting
around
yeah.
It's
awesome,
I.
C
C
B
A
Well,
this
stuff
I'm,
pretty
I,
think
I
think
this
is
understandable.
I
think
gmn
is
a
really
good
rapper,
that
abstracts
away
the
interview
ya
go
so
there's
another
there's
a
blog
post
about
one
of
the
features
in
Cassandra
30,
the
new
token
allocation
algorithm,
which
I
think
is
pretty
cool,
that's
on
the
datastax
blog.
Would
you
guys
think
of
you
guys
have
looked
at
this
this
feature
already?
What
do
you
what's
your
take
on
this.
C
So
I
mean
when
I
read
it,
I
will
admit
to
like
not
understanding
all
of
the
explanation
of
the
algorithm
involved,
so
but
I
guess
what
I
took
away
from
it
was.
It
seemed
like.
Basically,
the
idea
was
to
for
people
using
be
nodes,
which
was
everybody
like
on
on
new,
newer
versions
of
seen
her,
since
what
when
was
that
too
old,
where
that
was
going
to
be
the
default?
Who
went.
B
C
So
I
guess
what
I
took
away
is
that
it's
going
to
be
trying
to
be
smarter
about
how
it
how
it
distributes
V
nodes
around
there
around
the
cluster
and
where
you
know
what
machines
they're
distributed
to,
and
you
even
get
some
parameters
to
kind
of
play
with
to
give
it
hints.
You
know
to
kind
of
be
smarter,
especially
if
you've
got
an
unbalanced
cluster
like
where
you
got
more
powerful
machines.
Maybe
you're
phasing
in
some
new,
more
powerful
machines
into
an
existing
cluster
kind
of
thing.
I'm.
A
B
Kereta
does
is
say,
you
know
the
when
this
has
been
liked.
It
I
think
day,
zero
problem
with
v
nodes
that
I'm
just
glad
has
finally
solved,
because
that
code
was
pretty.
The
original
code
was
just
a
randomizer.
I
took
up
a
token
a
random
token
out
of
a
murmur
of
three
and
said
here
just
use
this,
and
that's
that
was
the
reason
we
had
256
tokens
in.
B
A
B
All
right,
so,
interestingly
enough,
we
didn't
this
isn't
in
here.
This
is
a
previous
blog
post,
but
we
just
had
that
blog
post
about
vinodh
and
jay
baud
alignment,
and
I
think
that's
great,
because
now
we
could
start
getting
into
more
intelligent
operations,
type
decision
or
we
say
I'm
going
to
create
so
many
different
j
bought
dis
and
I'm
gonna
match
it
up
at
the
number
of
v
nodes
and
so
you're
making
those
kind
of
11
decisions.
Instead
of
just
throwing
out
a
number
which
you
know,
those
are
fun.
A
Dude,
don't
make
any
sense
yeah
and
this
this
jury
you're
talking
about
I,
think
this
is
where
it
allocates
v
nodes
two
discs
and
it
prevents
a
whole
bunch
of
failure.
Scenarios
like
if
you
lose
one
disc,
you
don't
necessarily
affect
every
single
range
of
tokens
that
that
a
note
is
responsible,
for
it
only
affects
some
of
them,
so
you
can
just
replace
one
disc
and
it's
a
less
chaotic
operation.
Hopefully
it
also
rained
for
your
entire
data
center
yeah
and
hopefully.
B
A
B
I
think
it
would
get
you
away
from.
We
are
tanjung,
but
it's
a
good
point.
It
gets
people
away
from
having
to
do
craziness
like
raid
one
raid
0,
which
one
am
I
going
to
do,
because
I
gotta
do
something
with
my
disk
and
I'm,
not
really
sure
what
the
allocations
can
be
like
and
I
use.
J-Bot
isn't
going
to
be
right,
I
mean
there's
a
lot
of
questions
and
come
up
in
operations.
A
Definitely
the
last
thing
that
we've
got
in
here
is
logging.
The
generated
cql
from
the
spark
connector
I'm
I'm
actually
really
happy
to
see
this
because
I
had
I
was
just
talking
about
this
with
somebody,
so
basically
the
the
big
problem.
If
you're
not
familiar
with
the
spark
a
connector,
you
can
be
executing
these
big
batch
jobs
against
Cassandra
and
you
may
have
no
idea
what
cql
is
being
thrown
against
Cassandra
from
spark,
because,
if
you're
using
especially
if
using
data
frames,
it's
doing
a
lot
of
the
work
for
you.
A
So
this
this
is
opposed
from
ryans
fila
talking
about
the
different
ways
of
seeing
the
generated,
cql
and
I
I
think
this
is
really
helpful,
like
if
you're
using
spark
of
the
printer
protector
like
to
me.
This
is
like
you
should
absolutely
read
this
post
smite.
That's
just
my
point
of
view.
Read
it
right,
read
it
right
now,
a.
C
B
I'm
right
now,
I
mean
I,
get
it
but
yeah,
but
it's
it
I
guess
that
goes
to
the
same
problem.
It
is
when
you
have
algorithms
developing
some
query:
yeah.
It's
like
trusting
a
pit,
bull
good
dog,
good
dog
bit
my
arm
up,
so
it
could
happen
right
I
mean
you
could
get
a
really
wack
query
going
and
it's
because
it's
non
deterministic
you
get
a
good
chance,
but
as
a
practitioner
you
need
to
have
visibility.
Okay,.
A
C
A
So
we've
got,
we've
got
an
event
coming
up
I
think
we
should
take
a
sec
to
talk
about
Cassandra
de
la
on
februari
17th
East
during
the
LA
area.
You
definitely
you're
going
to
want
to
check
this
out,
especially
if
you
don't
have
a
lot
of
experience
with
Cassandra.
This
is
a
lot
of
it's
my
birthday.
B
A
B
B
A
B
A
Let's,
let's
talk
a
little
bit
about
some
jira
updates
right.
This
is
this
is
where
I
get
to
play
around
with
code.
That's
not
even
released
yet
and
there's
little
changing.
I
love
this
stuff,
the
first
one
I
want
to
mention
this
is
this
is
an
improvement.
So
it's
a.
It
sounds
like
a
small
improvement,
but
it
can
be
a
really
big
deal.
Just
fixing
improving
this
spark
Cassandra
protector
to
support
reading
for
materialized
views
that
just
didn't
work
apparently
with
30.
A
A
B
C
A
You
don't
yeah,
you
don't
need
to
be
memorizing
every
single
JIRA
I.
Let
you
think
it's
fun
this.
This
is
huge,
so
we
can
we'll
probably
spend
some
time
blogging
about
the
the
actual
implementation
details.
But
the
gist
of
this
is
that
you
can
do
prefix
like
wild
card
searching
Suffolk
searching.
There's
some
awesome
stuff,
that's
coming
into
this,
and
it's
fast.
This
is
fast.
C
This
you're,
just
like
cited
bit
yeah
I,
mean
I've
just
second,
what
everybody's
saying
I
mean
it's
super
exciting
like
being
able
to
do
a
like.
You
know
it's
like
bringing
us
closer
and
closer
to
that
to
that
SQL
kind
of
parody,
not
that
that
should
be
our
goal,
because
we're
not
you
know
we're
not
a
relational
database,
but
pretty
cool
too
to
be
able
to
do.
You
know
those
those
kinds
of
queries
that
you're
used
to
doing
near.
B
B
Just
it,
I
also,
I
think
this
has
always
been
an
issue
with
data
model
and
cassandra
is
people
tend
to
use
secondary
indexes
a
little
more
liberally
than
they
should
thinking
of
well.
This
is
the
right
choice.
If
I'm
going
to
try
to
bend
my
my
big
data
model
to
my
will,
but
if
you
use
them
in
an
incorrect
way
than
they
can
turn
into
a
real
downside
and
it's
it's
almost
an
anti-pattern
I
know:
I've
had
some
people
say.
B
A
But
here
there's
a
lot
of
really
good
performance
stuff
and
it
too
I
mean
I,
don't
want
to
look
too
deep
into
the
internals
like
I
said
because
it's
it's
pretty
hairy,
but
there's
some
really
really
clever
things
being
done
in
here
that
enable
really
really
good
performance.
You
still
have
the
problem
with
trying
to
do
secondary
indexes
across
you
know,
distributed
costs
like
a
distributed
database.
So
you
know
you
could
hit
every
machine,
but.
B
A
It's
it's.
The
impact
of
doing
them
on
each
machine
is
going
to
be
so
much
less
that
it's
absurd
and
the
implementation
you
know
I
keep
reading
about
it.
I
keep
looking
through
the
code
and
honestly,
it
just
gets
more
and
more
impressive.
Every
like
everything
I
learned
about
it,
it
looks
better
better.
Everyone.
B
A
I'm
really
I'm
really
pumped
to
try
this
stuff
out
with
spark.
So
next
week,
I'm
going
to
be
playing
with
spark,
and
this
and
and
seeing
how
how
much
of
a
performance
impact
we
can,
we
can
have
from
from
having
secondary
indexes
that
are
there
you
know
trivial
like
they
add
it
doesn't
take
any
longer
to
do
a
right
to
so
you
can
you
get
the
same
right,
performance
and
you're
going
to
get
just
better,
read
performance
and
flexible
querying
that
you
just
could
not
do
before
so
so.
A
Grouping
as
a
separate
one,
but
that
is
being
worked
on,
but
there's
but
arbitrary,
like
arbitrary
numbers
of
predicates
that
get
built
into
like
an
actual
like
tree
of
an
indoors
with
you
know
like
a
real
query:
planner
that
can
combine
operations
to
make
for
more
efficient
lookups.
That
stuff
is
all
like
slated
to
come
out.
I
think
it's
just
being
done
in
stages
right
now,
but-
and
I
actually
prefer
that
I'd
rather
see
like
be
done
at
a
time
as
opposed
to
like
these
huge
patches,
which
landed
everybody's
like
rebasing
for
weeks.
B
B
The
dream
of
tic
toc
right,
though,
how
we
do
tick-tock
releases
as
the
even
ones
our
feature
releases,
but
we
do
them
enough
that
now
we
can,
we
could
have
this
apostles
been
working
on
enough
to
get
into
three
dot
for
where's
functional
what
people
can
start
using
it.
I
sorry
hi
talking
about
the
there
is
a
company
out
there,
a
user
ready
to
put
into
production.
God
bless
him
for
that.
But
you
know:
don't
everyone
do
that
great
idea.
I
will
in
person
yeah
I
know
you
know,
but
you
know
that's
that's
great!
B
A
Yeah
so
I'm
I'm,
loving
the
stuff.
That's
coming
out
in
tick-tock
I
wrote
a
blog
post
about
what
happened
in
3,
2
I'll,
be
doing
the
same
thing
for
every
single
release,
just
covering
what's
happening
and
the
sazzy
stuff
that's
coming
out
in
34
and-
and
I
suspect,
every
feature
release
is
going
to
be
just
amazing.
So
this
is
there.
B
A
So
I
think
that
covers
all
that
we
had
in
here
40
is
that
we
want
to
talk
about
projects
and
things
like
that.
Do
we
have
any
Q&A?
A
B
That's
a
physics
question
because
jana
not
hace
be
the.
A
A
B
B
B
Sad
when
they're
frozen,
so
what
we
want
every
week
John
is.
We
would
like
to
we're
going
to
do
this
blog
post
every
week.
This
we
can
go
sooner.
Just
to
give
you
get
you
caught
up,
so
you
don't
have
to
try
to
track
it
all.
The
time
like
we
do
we're
kind
of
maniacs
for
this.
But
what
we
want
to
see
is
some
projects
blogs
things.
You
know
things
that
are
interesting.
They're
happening
community.
Do
this
weekly.
B
If
you
have
a
job
posting,
you
really
want,
and
we've
already
got
a
couple
for
next
week-
ready
go
ahead
and
send
those
our
way
just
at
Planet
Cassandra
on
Twitter
and
we've,
and
if
it's
more
than
what
you
could
do
in
a
tweet,
then
add
us.
There
hit
us
up
on
Twitter.
We
can
get
into
a
email
exchange,
but
we
want
to
know
about
stuff,
that's
happening
so
that
we
can
convey.
It
is
back
out
to
the
Cassandra
community.
B
A
A
So
all
right
anything
else,
I
can't
think
of
anything.
I'm
done
no
I'm.