►
From YouTube: Webinar | Oracle to Cassandra Core Concepts Guide Pt. 3
Description
Tired of timeouts? Cursing your cursors? Join the distributed revolution and bring your dev team into application nirvana. You won’t believe how easy it is to be code complete on your next big project. We will show you how to lead your devs away from the clutches of the DBA and be in control of their own data destiny. DIscover the methodology that will make your Cassandra project epic.
A
A
C
A
C
C
A
A
C
A
B
C
C
A
Let's
start
this
by
setting
the
Wayback
Machine:
let's
go
back.
C
A
That
point
I
had
given
up
on
Kobol
so
now
I
wasn't
really
concerned
about
it
that
we
yet
so
this
is
I.
Think
I
was
doing
this
all
the
time
data
modeling
Oracle
web
web
applications.
That's
when
it
really
got
serious
for
me,
and
it
was
always
the
same
thing
right
and
that's
you
can
go
way
back.
So
this
is
a
quick
sketch
of
me
and
Rachel
of
one
of
our
early
whiteboard
sessions
in
the
cave,
and
so,
if
you'll
notice
I
mean
this
is
a
funny
picture.
But
what
do
you
see
there?
A
This
is
an
ER
D
right,
one
demanding
relationships,
and
so
data
modeling
was
a
exercise.
You
started
first
by
looking
at
your
data,
so
this
is
the
process
and
I.
Think
you'll
probably
remember
this
is
like
you
take
this
data
domain.
Hey.
We
have
a
problem
in
our
data.
Here's
the
scope
of
all
those
things
that
we
need
to
talk,
order
to
deal
with,
store,
query
on:
let's
go
through
a
couple
of
design
sessions
and
work
up.
Our
data
mom
sounds.
A
B
A
If
you
take
it
way
way
back,
you
get
Peabody
and
Sherman
to
set
the
Wayback
Machine
to
the
70s
record.
Format
was
what
was
before
relational
format,
and
it
was
just
everything
within
a
huge
record
on
a
mainframe
or
multiple
files,
but
that
was
because
you
had
a
limited
space
and
with
relational
he
was
about
creating
the
relationships
between
data
that
then
you
could
form
together
using
a
query
and
the
normal
forms
gave
you
the
most
flexibility
for
those
queries
was.
A
A
C
A
A
A
C
C
A
At
all,
so
this
is
this
is
a
is
an
entity
relationship
diagram
of
killer
video?
Now
you
can
see
the
relationships.
For
instance,
if
you
look
at
the
upper
right,
you
see,
users
have
many
video
events
or
have
many
comments,
and
you
can
see
the
relationships
are
marked
by
there's
constraints
and
you
can
map
out
everything
here.
A
This
is
a
pretty
good
diagram
and
from
this
as
an
application,
developer,
I
can
look
at
this
diagram
and
say:
hey
I
know
what
queries
I
need
to
do
like
if
I
need
to
say
well,
I
would
like
to
know
for
a
user
all
the
video
events
for
a
particular
video.
You
know
that's
a
query.
You
could
ask
of
this
right
or
right.
A
B
A
A
Standpoint
and
actually,
we
went
through
a
design
session
on
this
and
I'll
post
a
schema
for
the
SQL
for
this,
but
yeah
that
we
went
through
a
design
session
like
this
is
how
we
do,
but
this
is
before
doing
anything
that
we
don't
even
know
what
the
applications
really
going
to
do
other
than
we
know
it
all
that
they
made
in
the
data
so
we're
building
the
model.
First,
okay,.
B
A
A
A
A
If
I'm,
if
I'm
getting
potentially,
if
I'm,
going
to
query
on
the
users
table
and
I
want
to
look
up
by
email,
mm-hmm.
C
A
Table
scan
yeah
so
by
indexing.
What
I'm
really
saying
is
I'm
presupposing
that
I
will
use
email
in
a
query
and
an
equal
query
greater
than
well
present
integrator
there
may
be
a
like,
but
this
is
just
saying,
I'm
going
to
query
on
email
and
then
foreign
key
constraints,
which
gives
me
this
linkage
between
users
and
a
video
if
you
notice
that
there
is
a
user
ID
in
the
video
table,
so
that
users
have
many
videos
right.
C
A
A
A
That's
why
you're
here
you
want
to
know
the
difference
and
some
I
gave
you
like:
here's,
here's,
your
old,
familiar
pastures,
so
to
speak,
and
now,
let's,
let's
see
what
we
have
a
fork
in
the
road,
we're
no
longer
going
to
do
that,
we're
going
to
go
a
different
direction!
First
thing
you
have
to
know.
B
Ahead
to
have
your
application,
it's
all
about
the
everything
from
the
application
to
the
fuse
and
there's
actually
stems
from
some
of
the
practices
of
denormalization
in
general.
So
when
I
used
to
date,
a
model
back
in
the
days
of
data
warehousing
and
do
my
star
schemas
and
all
that
I
would
spend
a
lot
of
time
and
I
work
for.
B
A
A
B
A
A
C
A
B
B
B
C
B
A
B
A
Product
people
make
sure
that
the
product
domain
people
are
involved.
This
is
this.
Is
the
people
involved
here,
because
we
want
to
make
sure
that
we
got
all
the
parts
we're
vast?
Yes,
okay,
we
do
need
to
have
a
user
with
him
first
name
and
last
name
check.
So
this
is
really
not
we're
not
getting
into
the
physical
data
mark.
This
actually
could
live
outside
a
Cassandra
as
well,
which
is
interesting,
but
will
quickly
change
that.
A
B
B
C
B
C
A
C
A
A
B
A
C
C
A
Yeah
we
we
now
have
a
pretty
good
idea
of
what
we
need
from
our
database
based
on
this
I
think
so,
based
on
all
that,
look
at
what
we
got,
we
actually
have
some
interesting
questions.
Now
we
can
address
with
tables
and
now
we're
progressing
towards
the
physical
data
model,
which
is
or
actually
create,
table
statement
so
like
when
a
user
logs
and
find
users
by
email
address.
I
need
that
yep.
C
A
Let's
get
into
the
reasoning,
so
I'm
gonna
I'm
gonna,
leave
this
there
for
when
we
put
this
on
month
SlideShare,
so
you
can
have
this
whenever
you
read
it,
but
well.
You've
already
should
have
gone
through
this,
and
this
is,
you
know.
We
know
our
queries.
We
denormalize
all
of
these.
These
are
like
the
data
model.
How
you
do
things-
and
this
should
be
from
last
week
right
and
let's
look
at
how
we
do
this
with
like
users,
look
directly
here's
a
physical
data
model
so
and.
B
B
B
Right
where
we
made
email
the
primary
key,
which
in
this
case
means
that
that's
how
it
gets
distributed
around
the
ring.
So
when
we
say
we
want
email
address,
equals
Rachel
a
dataset
com,
it
will
know
exactly
which
nodes
to
pull
that
record
from.
So
that's
how
it
becomes
extremely
efficient
in
a
distributed
system.
Right.
C
A
Is
a
critical
concept
is,
and
you
I
don't
want
to
just
make
this
a
blank
of
statement,
but
most
times
when
you
do
create
index
inside
of
a
relational
database,
you
could
probably
create
a
table
to
do
the
same
action,
and
it
gives
you
that
full
distributed
like
what
we
learned
last
week
with
a
primary
key
being
distributed.
The
way
it
is
and
how
it
can
hatch
properly
and
get
spread
out
of
them
are
much
larger
ring.
You
take
advantage
of
that
and
you
get
great
uptime
I'm
amazing
speed.
A
B
You
to
create
a
like
the
table
that
we
had
on
the
previous
slide
and
put
a
index
on
email
address,
but
we
know
that
index
is
fairly
unique.
So
if
you
think
about
it,
each
know
is
going
to
index
its
own
data.
So
note
80
up
there
is
going
to
take
all
the
data
for
the
cases
user,
ID
that
you
know
where
is
in
the
range
of
80
and
it
will
index
it.
B
So
if
you're
looking
for
specific
email
address
like
Rachel
at
data
sex
calm,
that
coordinator
note
isn't
going
to
know
where
that
name
is
this.
So
it's
going
to
have
to
ask
that
node
and
that
node,
that
node
and
that
node
and
that
node
and
that
node
until
it
finds
Rachel
a
dataset
calm,
because
it's
only
going
to
exist
once
and
that
seems
like
a
whole
lot
of
asking
for
not
a
whole
lot
of
return
and.
B
I
mean
yeah,
but
here's
only
one,
eight
eight
node
right
yeah,
but
that's
yeah.
That's
that's
a
beginning.
Nexus!
Tarter,
no
starter
ring
for
most
people.
Now
it's
that
being
the
case.
It's
the
it's
the
thing
that
you're
indexing
or
you're
looking
for
gender,
for
example,
gender,
might
have
five
values.
You
know
five
unique
values
associated
with
it.
So
therefore
you
know
to
bring
back
20%
of
data
when
you
say:
hey
I
want
everybody's
female.
You
bringing
my
pointers
on
the
data.
Okay,
maybe
that's
a
little
less
inefficient
right
going.
C
B
A
B
A
B
And
so
not,
but
not
every
question
is
going
to
fall
into
the
category
of
being
able
to
do
a
you,
build
your
own
index
so
say
if
you
wanted
to
do
a
query
where
it
says
we're
email
address
like
at
yahoo.com,
so
you
want
to
pull
out
all
your
Yahoo
addresses.
You
even
model
that
if
it
was,
it
was
very
important,
but
something
that
affects
enterprise
gives
you
is
the
ability
to
solar
and
leucine
indexes
right.
C
A
A
A
Let's
look,
this
is
just
fair
warning,
but,
and
your
scanning
is
important,
I
think
you
now.
You
know
the
reason
why
it's
not
just
no
now,
you
know
why
and
no
and
features
are
coming
3.0
how
to
really
interesting
things
that
are
going
to
change
this
story
of
it.
Even
more
so
I
mean
when
this
database
is
moving
in
a
correct
direction.
B
A
Right
that
can
take
a
blob
space,
so
yeah
there's
no
free
lunch.
It's
just
understanding
the
trade-offs,
so
I
I
have
this.
You
know
this
concept.
I
need
to
look
up.
Videos
by
user
and
I
created
two
border
with
it.
Well
so
I
have
this
user
video
table
on
the
right
and
what
what
I
want
to
point
out
and
what
we
taught
talked
on
earlier
was
this
whole
idea
of
denormalizing
data
where
normalising
was
about
less
data?
Duplication
is
like
one
copy
of
your
data
for
for
all
to
use.
A
B
A
C
B
B
It
down
right,
so
if
your
app
isn't
responding
and
we've
all
done
this,
you
say
forget
this:
did
you
go
to
the
Internet
and.
B
C
A
A
There
are
libraries
and
drivers
that
I
love
it
that
help
you
do
that,
for
instance,
eql
Engine,
which
is
now
part
of
the
Python
driver
data
stacks,
have
some
nice
features
for
maintaining
two
tables
same
time
with
single
data,
but
this
is
on
the
developer.
This
is
where
the
developer
have
to
be
in
charge
of
this
particular
problem
and
again
it's
for
what
you're
trying
to
accomplish
with
your
data
in
your
application.
So.
A
A
A
Just
a
different
way
to
do
it.
Yeah
well,
I
also
want
to
talk
about
another
issue,
and
that
is
when
we
get
into
some.
Last
week
we
learned
about
how
the
primary
key
works,
that
it
hashes
the
key
and
it
goes
to
a
single
node,
and
so
let's
look
at
this
data
model
where
we
have
the
latest
videos
bucketed
by
day.
C
C
B
A
A
C
A
A
C
A
A
B
B
A
A
A
A
A
A
Plan,
it's
spreading
that
out
that
load,
which
is
kind
of
what
we
wanted
in
the
first
leg
partition
key.
If
you
use
a
different
partition,
key
we'll
do
that
and
if
we
odd
bucket
number
it
will
guarantee
you
have
a
different
partition.
Key
every
single
time
gives
you
that
good
spread
on
your
nose
locality.
You
don't
run
into
hotspots.
It's
just.
A
A
A
This
assumes
that
you
will
be
on
and
if
you've
ever
ever,
had
your
sequences
get
completely
logged
and
in
Oracle
world
feel
your
pain
I've
tried
to
these
really
a
mess
whenever
the
sequences
get
whacked
somehow
or
the
cash
gets
messed
up
somehow
and
it
happens,
I've
had
that
them
in
a
few
times.
You
have
to
deal
with
this
problem,
so
this
doesn't
work
in
distributed
systems
a
and
so
when
you
ask
for
a
sequence,
I
know
what
you're
thinking,
how
do
I
find
unique
number
right.
A
A
C
C
A
That
it's
a
consistent
hash,
so
I
can
identify,
for
instance,
cream
and
but
every
library
does
it.
So
it's
a
128-bit
number.
So
that
means
that
there's
a
lot
of
them
and
the
guarantee
of
uniqueness
is
pretty
solid,
I'm
sure
there's
a
collision.
One
point
in
the
universe
at
one
point
in
some
time,
but
who.
C
A
A
B
A
C
A
C
A
A
A
A
B
This
was
always
fun,
because
this
is
I.
Think
a
very
one
of
our
big
clients
gave
you
a
how
many
page
document
of
skier
on
this
20.
B
Consistency,
oh
yeah
cap
theorem
and
not
so
consistent,
and
what
do
we
do?
The
whole
world
gonna
fall
apart.
Yes,
the
thing
is:
Cassandra
is
commonly
known
as
eventually
consistent,
but
I
like
to
say
to
the
tune
of
Li
consistent,
because
you
know
you
can
be
as
consistent
as
you
want
it
to
be.
You
just
pay
for
it
as
far
as
amount
of
time
does.
So
why
are
we
to
bring
it
up
here?
B
It
is
because
it's
part
of
your
application
designed
determine
how
consistent
you
need
each
one
of
your
reads
in
each
one
of
your
rights.
You
can
set
a
default
on
the
client
to
always
do
a
certain
type
of
consistency,
but
keep
in
mind
every
query
right.
You
control
how
consistent
you
want
that
right.
So,
let's
go
through
a
little
bit
about
what
that
means
right.
A
B
B
B
This
is
talking:
let's
talk
a
little
bit
about
consistency,
so
we've
got
our
famous
client
in
our
nose
and
we've
got
replicas,
so
replication
cracker
free
these.
These
are
concepts
that
we're
all
discussed
in
the
first
part
of
this
series.
So
if
this
does
not
make
any
sense
to
you,
you
might
want
to
go
back
and
wait.
B
B
A
A
Think
of
all
the
random
data
would
be
like
I,
usually
like
some
common
kind
of
concept.
It's
like
log
data
like
I'm,
writing
log
data
I
just
want
to
make
sure
it's
in
there
I.
Don't
really
need
consistency
to
very
high,
so
I'm
just
throwing
it
it's
a
database
as
fast
as
possible,
and
consistency
in
this
case
is
not.
My
concern
is
just
that
it
gets
written
to
the
database.
You.
B
B
A
C
A
A
A
B
B
C
A
B
A
I,
like
51%
committed
here,
but
what
does
that
mean
the.
C
B
A
B
A
B
B
A
A
very
good
transition,
so
when
we
when
we
go
right
or
right
arrow
application,
the
drivers,
a
critical
part
of
this,
obviously,
is
that
we
are
API
how
we
get
a
hold
of
Cassandra
with
our
application.
So
what
you
have
to
choose
from,
and
just
quick
tour
here
of
drivers,
its
data
stacks
we've
devoted
ourselves
to
trying
to
create
a
very
comprehensive
and
consistent
package.
A
C
A
From
data
stacks
and
in
the
community
we
have
drivers
of
closure,
dough,
Erlang,
and
there
are
others
that
I
see
popping
up
from
time
to
time,
but
these
drivers,
what
what
are
they
going
to
do
for
you
and
then
why
would
I
say?
They're
consistent,
like
the
experience,
is
consistent
by
what
here's
an
example
of
a
this
is
a
slice
of
Java
code.
B
A
A
They
do,
and
you
should
not
have
to
rewrite
your
application
code,
A
or
B,
deploy
a
new
configuration
file
and
man
I.
That
was
one
of
the
hard.
You
know
hard
thing
to
manage
in
production
environments
is
configuration
file
maker,
they're
all
correct.
What
I
built
it
am
I
covered
here
is
now.
This
is
part
of
the
consistent
driver,
experiences
that
we
have.
We
have
a
way
to
connect
to
the
cluster
survey,
adding
some
seeds
or
by
DNS
name,
and
then
we
can
specify
retry
policies
like
what
happens
when
bad
things
happen.
A
How
do
we
manage
that
and
then
load
balancing?
How
do
we
make
sure
that
the
data
is
being
spread
out
among
all
the
replicas
very
nicely
like,
for
instance,
in
this
case,
I
create
this
token,
where
policy
so
I'm,
actually
looking
at
the
partition
key
and
not
not
being
not
going
to
a
coordinator
I'm
going
to
write
the
to
where
the
replicated
and
I'm
keeping
it
in
the
same
data
center.
It's
got
enough
crossing
data
center
wine
yeah.
B
A
A
I
mean
I,
my
the
two
languages,
I
use
the
most
or
Java
Python,
but
I
use
the
Cassandra
driver.
If
I
don't
have
to
rethink
about
how
I
connect
and
do
things
to
the
cluster.
It's
different
syntax,
but
I
still
get
these
three
things
and
that's
the
consistency,
and
you
should
understand
how
this
work.
That's
about
all
I'm,
going
to
get
into
that.
What
I
mean
we
have
plenty
of
talks
about
writing
application
code,
but
I.
A
Language-Specific
talks
tons
of
information
on
this,
but
I
just
want
to
keep
keep
this
on
point
of
when
you're
doing
it.
This
is
where,
again,
you
can
even
work
with
your
DBA
s,
to
make
sure
that
you
know
the
contact,
points
and
load
balancing
policies
are
correct
for
what
you're,
what
you're
deploying
so
finish
up
here,
yeah.
B
B
A
Not
at
all
and
I
think
anything
I'm
living
proof
that,
but
we
have
you
know,
do
we
have
to
hear
some
roles
that
I
could
easily
define.
Just
you
know
by
looking
at
the
the
separated
separated
roles
from
when
you're
at
Oracle
developer,
Oracle
DBA.
These
are
the
things
we're
separated,
but
what
I
want
to
show
you
now
is
the
things
are
changing
right
now,.
A
C
A
Drivers
setup
like
I
just
mentioned
you
like
picking
the
right
consistency,
levels
and
things
like
that
capacity,
planning
used
to
be
all
DBA
yeah
developers
can
have
a
hand
in
that,
because
they're
going
to
be
able
to
create
tables
to
help
capacity
or
her
it
alright.
So
they
have
to
be
a
part
of
that
capacity
plan.
So
she
looks
private.
B
Yeah
well,
unless
see
why,
when
we
were
Patrick
and
I
were
on
the
plane
yesterday,
and
we
were
using
in-flight
Wi-Fi
losing
at
30,000
feet
in
width
and
how
bright
is
the
future
like
what
means?
What
is
the
job?
What
is
it
like?
Who
feels
good
and
RAF?
What
does
the
job
situation
look
like
for
people
who
are
using
Sondra
looks.
A
Is
a
this:
is
a
percent
of
growth
in
jobs
on
indeed
comm
taken
as
Rachel
said
as
of
yesterday
at
30,000
feet,
and
why
are
we
showing
to
us?
Well,
if
you're
listening
to
this
webinar,
you
were
probably
a
professional
in
the
field
and
I've.
B
B
A
We'll
put
this
online
before
we
before
we
get
to
we're
going
to
make
you
smarter.
If
you
want
to
actually
do
this,
this
is
all
living
online,
go
to
killer
video
calm
and
it's
Luke
Tillman
guy,
who
wrote
the
c-sharp
version
of
this
and
it's
coming
with
different
versions.
Soon
this
is
on
github
and
just
go
nuts.
It
is
really
the
application
was
really
cool.
All
the
scheme
is
there,
you
get
a
top
down,
there's
even
some
some
great
demonstrations
of
how
it
works
if
you're
an
application
developer.
C
C
B
A
Is
legit
yeah
and,
if
think
about
this
is
your
next
big
career
move
as
is
really
going
to
enhance
your
career.
I
gave
recruiter
emails
all
the
time.
I
need
to
Xander
people
think
about
this
in
that
term,
because
it
will
make
a
difference
in
taking
this.
So
these
three
webinars
we
have
give
you
enough
information
to
learn
more.
They
should
get
you
somewhat
dangerous,
but
go
to
Dave
Sachs
Academy,
Academy,
Dave
Dexcom
learn
as
much
as
you
can
plenty
of
courses
plenty.
B
C
A
A
A
Nev's
are
called
hints
and
the
hints
are
stored
locally
on
those
nodes.
Until
that
node
comes
back
online
now,
there's
a
configuration
value
that
allows
you
to
choose
how
long
those
nodes
will
store
like
if
we'll
start
doing,
it
won't
do
like
a
FIFO
buffer,
but
it
just
won't
collect,
hints
anymore,
and
the
default
is
three
hours
and
and.