►
From YouTube: DataStax: Reliable and Continuous Data Availability
Description
Where do big companies such as Adobe, eBay and Netflix go to to ensure their data is scalable, flexible, and secure? DataStax serves these companies and more by providing a platform for all their data to live on that is built on Apache Cassandra.
Billy Bosworth, CEO of DataStax talks about how and why top companies and many startups use DataStax for its reliability and continuous data availability.
http://www.datastax.com/
https://twitter.com/datastax
https://www.facebook.com/datastax
A
When
you
saw
me
interview
the
guys
who
make
this
light
app,
which
is
a
big
data
app,
they
built
it
on
top
of
Cassandra
I'm,
hearing
that
more
and
more
from
startups.
That's
part
of
one
of
five
trends,
I'm,
seeing
as
part
of
the
age
of
context
that
I
keep
talking
about.
You
know
more
and
more
sensors
were
more
wearable
competing
big
data
and
social
social
networks
that
keep
evolving
and
mapping
that
keeps
evolving
all
that
stuff.
B
Who
are
you
Billy
Bosworth,
the
CEO
of
company
called
datastax
I've,
been
around
databases
for
a
very
long
time.
First,
ten
years
of
my
career
was
a
developer
and
the
DBA
right
writing
a
lot
of
turnkey
systems
for
departmental
level
companies
and
met
these
guys
through
I
was
a
quest
software.
They
just
got
acquired
by
del
recently,
and
we
partnered
with
a
lot
of
companies
in
this
space
and
I
was
very
interested
in
what
this
change
was
going
to
look
like
for
the
database
world
in
particular.
B
A
Time
so
I'm
seeing
these
five
trends,
you
know
whether
it
be
wearable
computing
sensors
is
going
exponentially
more
and
more
sensors
every
year
or
maturing
social
network.
So
we
can
do
stuff
with
study
what
your
Facebook
is
or
your
Twitter
tweets
are
right.
I
mean
there's
companies
like
vin
tank
that
just
study
when
the
world
talks
about
why
and
that
happens
1.1
million
times
a
day
right
now
and
it's
going
up
right.
So
all
these
new
things
are
happening.
General
Electric
called
it.
A
A
You
when
you
start
your
career,
you
know
you
had
to
buy
a
mainframe
somewhere
right
and
put
some
software
on
there
and
pay
a
company
a
lot
of
money.
For
that
now
the
world
is
a
lot
bigger.
A
lot
faster
I
mean
1.1
million
Twitter
conversations
about
wine.
That
day
is
different
than
even
five
or
six
years
ago.
Right
I
would
probably.
B
Describe
it
like
this,
when,
if
you
came
to
me
five
years
ago
and
said,
hey
Billy,
let's
go
write
an
app
for
restaurant
reservations.
I
would
start
thinking
in
my
mind,
because
I
came
out
just
when
Oracle
was
coming
on.
The
rise
like
92
is
when
I
came
into
school
and
so
I've
been
conditioned
to
think
about
that
app
in
a
certain
way.
So
I
would
say:
okay,
robert
great
here's,
what
we're
going
to
do
in
my
mind.
B
I've
got
some
ideas
on
the
data
model
and
how
everything's
going
to
be
stored
and
I'd
say:
let's
find
the
openings.
Let's
find
people
who
come
in,
let's
find
the
openings
put
them
in
a
slot.
Great
we're
done
today's
world.
We
have
to
think
about
that
totally
different
you'd
have
to
come
in
and
we'd
have
to
say:
okay
Robert,
here's
what
we
need
to
understand
somebody
might
be
walking
by
our
restaurant
or
they
might
be
two
miles
of
our
restaurant.
They
might
be
having
transit
problems.
B
Maybe
there's
a
backup
in
traffic
and
they're
not
going
to
get
to
where
they
needed
to
be.
Is
it
raining?
Maybe
they'd
like
to
stop
in.
Do
they
have
kids?
We
have
a
great
opening
at
seven
o'clock.
You
know
for
a
special
on
kids
meals
and
we're
a
family
friendly
restaurant.
Maybe
we
should
let
that
person
know
that
as
they're
walking
by
our
restaurant-
and
maybe
we
should
give
a
movie
tickets
for
afterwards,
because
we
know
they
like
this
kind
of
moving
the
movie
theaters
two
blocks
away.
B
A
Apps,
even
toys
toy
talk
is
a
new
company.
They
just
got.
16
million
dollars
was
started
by
the
CTX
CTO
of
Pixar,
and
the
toy
itself
that
I've
had
is
going
to
know
what
the
weather
is
and
they're
going
to
make
a
DI
call
out
to
weather
com
or
something
and
know
that
it's
raining
outside
and
they're
going
to
change
the
behavior
that
toy
even
toys
for
changing,
because
absolutely
so
what.
B
B
B
Store
or
something
like
that
exactly
and
we
do
see
it
everywhere,
we
see
these
problems
and
these
problems
kind
of
manifest
themselves
around.
One
of
the
biggest
pain
points
where
we
play
in
with
our
technology
is
this
notion
of
continuous
availability.
We
have
become
a
very
spoiled
culture
and
we
have
become
a
very
demanding
culture.
Think
about
your
own
experiences.
B
If
you
go
to
get
on
a
site
or
a
nap,
then
you
have
to
wait
two
seconds
you
get
frustrated,
you
start
to
bail,
you
start
to
look
for
other
alternatives,
so
this
idea
of
an
app
being
continuously
available
is
a
paradigm
shift.
This
is
not
how
we
used
to
think
before.
We
might
have
one
backup,
server
or
very
small
cluster,
but
again
now
we're
at
country
scale
and
now
we're
a
global
scale.
How
do
I
ensure
that
that's
going
to
be
continuously
available
all
around
the
world?
B
Multiple
data
centers
is
now
the
norm
for
these
architectures,
so
businesses
have
to
make
sure
that
they're
going
to
be
there
and
available.
Be
it
better
be
fast.
It
has
to
scream.
So
when
that
application
is
talking
directly
to
the
device
or
to
the
customer,
it
has
to
be
able
to
operate
in
a
way
that
they're
not
going
to
lose
that
that
customer
loyalty
very
quickly
right
yeah.
B
So
all
those
problems
create
this
other
challenge
of
operational
simplicity
and
then
cost
you
can't
let
this
stuff
get
out
of
control
either
every
business
that
we
work
with
is
fighting
with
that
and
we
range
from
fortune
100
businesses.
We
range
from
household
names
like
Adobe
or
Netflix
down
to
startups,
who
are
just
trying
to
solve
a
brand
new
problem
in
a
new
way,
but
they
have
that
same
common
use
case
of
continuously
available
extremely
high
performance,
low
operational
costs
to
keep
my
overall
cost
down
so.
B
B
Availability-
and
this
is
something
you'll
hear-
Netflix
talked
about
a
lot,
even
though
Netflix
does
so
much
in
the
cloud
being
in
high
availability
zone
in
the
cloud
is
not
enough.
They
actually
spread
it
across
availability
regions
in
the
cloud
to
make
sure
if
they
take
a
hit
like
with
Sandy
or
a
power
storm,
or
an
outage
of
some
other
type
that
they
can
still
carry
on.
Even
if
you
lose
an
entire
data
center,
so
think
about
just
not
losing
a
couple
of
machines.
What
if
you
lose
a
data
center
yeah?
B
Right
and
I
highly
recommend
that
actually
we've
got
a
big
retailer
we're
working
with
at
the
moment.
That's
got
a
co-located
physical
data
center,
so
they've
got
the
same
database
now
spread
across
two
physical
data
centers,
but
then
it's
also
spread
across
across
to
cloud
environments
and
it's
the
same
database
in
all
of
those
worlds,
and
anybody
from
anywhere
can
hit
it
from
any
place.
So
if
you
lose
one
of
them,
it
doesn't
go
down.
So
that's
our
first
use
case
now
we
take
that
great
distributed
architecture.
B
Then
we
say
you're,
probably
also
going
to
want
to
do
some
searching
you're,
probably
also
going
to
want
to
do
some
analytics.
So
we
take
two
very
popular
open
source
technologies
called
Apache,
Solr
and
Apache
Hadoop,
and
we
bring
those
on
top
of
the
cassandra
architecture.
Truly
integrated,
not
just
loosely
connected,
truly
integrated,
so
that
line
of
business
when
they're
thinking
about
writing.
One
of
these
transformative
apps
has
everything
they
need
for
their
silo
to
build
that
app.
B
B
Internally
and
that's
key
to
what
we
do,
we
want
to
be
agnostic
to
the
underlying
destination
decision
and
that's
critical
actually
to
what
we
do,
because
we
don't
want
to
limit
architects
and
saying
this
is
a
cloud
only
solution
or
worse
only.
That
cloud,
which
I
think
is
the
worst
scenario
we
want
to
be
able
to
say
you
can
spread
this
across
your
physical
data
centers.
You
can
spread
it
across
your
cloud,
data,
centers
or
any
combination.
There.
B
A
B
Agencies
we
see
it
as
well
thing
yeah
same
thing
there,
but
you
hit
on
it.
It's
the
flexibility
of
being
able
to
take
those
baby
steps,
for
example,
let's
say
you're
a
traditional
business
and
you're
starting
off,
and
you
would
say,
I
want
to
control
it.
Maybe
just
philosophically
you
can't
get
over
it.
Well,
okay,
fine
start
in
your
data
center,
but
wouldn't
it
be
nice
to
maybe
if
you
took
a
couple
pieces
of
it
and
started
pushing
off
to
a
cloud
just
to
test,
you
can
do
that
with
our
technology.
B
A
Cassandra
was
developed
partly
at
Facebook,
and
solar
was
partly
developed
for
Yahoo
right,
though
they've
had
their
own
hardcore,
geeks
too
yeah
that,
but
we're
talking
a
lot
of
businesses
around
the
world
who
don't
have
a
hardcore
DBA.
How
I
key
is
this,
and
can
I
can
I?
Do
it
just
without
knowing
too
much
or
what
do
I
need
to
learn?
Who
do
I
need
to
bring
on
board
to
help
me
build
one
of
these
big
data
systems.
B
Yeah
great
question:
it
an
essence:
that's
why
we
exist.
We
want
to
make
this
technology
consumable
and
accessible
by
businesses
and
what
we
say
by
mere
mortals.
Sometimes,
when
you
talk
about
the
crowd
at
Google
and
Facebook
at
Yahoo,
it's
still
I
popping
what
they
do.
They
do
so
much
customization.
They
do
so
much
of
their
own
tweaking
their
brilliant.
The
average
company
is
not
in
that
business,
so
they
have
to
find
a
way
to
consume
this
technology
in
a
way
that
enables
their
teams
to
move
very
very
quickly.
B
They
don't
want
to
get
behind
in
the
administration
side
of
things.
So
what
we've
done
as
a
company
is
bring
all
this
stuff
together
a
package
in
a
way
where
you
can
be
up
and
running
in
minutes.
We
have
demos
on.
If
you
want
to
get
started
very
quickly,
we
can
get
a
cluster
up
on
amazon,
for
example,
in
I,
think
it
was
under
three
minutes
or
four
minutes
to
get
a
ten,
no
class
or
up
and
running.
If
you
want
to
do
it
on
premise,
you
can
do
it
on
a
couple
of
machines.
B
Rackspace
cloud
we
have
people
doing
it
there
as
well,
and
we
have
our
job
is
to
make
it
easy
to
get
up
and
running
and
get
started,
because
one
of
the
knocks
on
our
technology
in
the
early
days
was
it's.
It's
like
all
three
standard
deviation
guys.
These
are
guys
that
are
way
way
at
the
end
of
the
bell
curve
and
we're
bringing
that
back
into
the
masses,
because
we
do
solve
technical,
challenging
problems.
B
I
do
want
to
say
when
people
choose
Cassandra
one
of
our
customers
meta
broadcast,
we
just
did
an
interview
with
them
and
they
said
we
don't
choose
Cassandra
for
simple
problems.
These
are
complex
problems
where
we
need
amazing
performance,
so
we
have
to
balance
that
with
ease
of
use.
If
we
sat
down
and
I
said,
Robert
tell
me
about
your
app
you're
trying
to
build
okay
and-
and
we
start
talking
about
it
and
I,
said
okay,
what
happens
if
that
goes
down?
B
And
you
came
back
to
me
and
said
well
sometimes
I'm
up
in
five
minutes.
Nothing
really
not
a
bit,
not
a
big
thing,
probably
not
in
the
best
conversation.
But
if
you
tell
me
no,
no
that's
the
whole
point.
I
don't
want
this
thing
ever
to
go
down.
My
users
have
to
know
that
they're
going
to
be
able
to
get
it
and
get
it
fast
when
they
need
it
got
it
now,
we're
in
a
good
conversation
as
an
example.
Big
data
right,
but.
B
Right
yeah,
big
data
people
tend
to
think.
If
they're
not
immersed
in
it,
you
tend
to
think
volume,
Yeah
right,
thousands
of
or
millions
of
something-or-other
right
in
our
world.
Sometimes
we
have
customers
who
do
fairly
small
implementations,
but
they
gotta
spread
it.
They
got
to
have
it
across
datacenters.
They
got
to
have
some
of
it
in
the
cloud.
They've
got
to
have
some
of
it
in
multiple
clouds
that
gets
awfully
difficult,
if
not
impossible,
with
relational
technology.
B
It's
why
I
made
a
career
change,
because
I
saw
this
modern
transformative
app,
requiring
this
availability
in
new
architectural
ways,
the
relational
databases
just
weren't
designed
to
handle
they
just
worked.
They
still
have
their
use
cases
in
the
second
part
that
I
would
answer
to
that.
Is
it's
not
an
either/or?
We
have
to
get
over
that.
So
many
people
want
to
know
if
you're
using
this,
it
means
you're
not
using
that
right.
Modern
architectures
today
you'd
be
amazed
how
many
back-end
data
stores
are
in
place.
The
adobe
example
I
gave
you
with
their
audience
manager.
B
Product
they've
got
a
big
Hadoop
data
warehouse
that
they
use.
They
even
use
HBase
as
part
of
that
which
is
another
kind
of
the
real-time
accessible
technologies,
but
for
the
piece
that
touches
the
users
that
has
to
be
available.
All
the
time
has
to
be
lightning-fast
got
to
be
spread
across
gos,
that's
Cassandra
same
app
same
general
app,
and
you
see
that
on
the
backend
a
lot
of
times
and
it's
brokered
by
services.
B
So
the
really
smart
application,
architects
figure
out
which
api's
go
to
which
data
store,
but
it's
very
confusing,
unfortunately,
to
the
person
who's
trying
to
learn
about
this
because
they'll
say
wait
a
minute
you're
talking
about
you're
in
that
company
I
heard
such
and
such
was
in
that
company.
Yet
it
might
actually
be
the
same
app,
but
it's
done
on
the
back
end
in
different
ways.
So
we
have
to
get
out
of
this
mentality
of
either/or
and
understand
that
it's
a
multiplicity
of
back-end
data
stores
used
for
the
right
scenarios.
B
A
B
Have
one
thing
was
required
now
I
told
you
my
first
ten
years
were
developer
dva
and
after
that
I
went
to
work
for
tools.
Companies
and
tools
are
really
valuable
as
you're
trying
to
get
into
the
beat
of
the
bell
curve
and
what
we've
done
is
create
a
product
and
interface
called
opscenter
that
manages
not
only
the
cassandra
bit
but
also
the
integration
of
the
Hadoop
and
the
solar
so
that
you
get
a
very
simple
point-and-click
web
interface.
B
That
gives
you
nice
pretty
pictures
and
gives
you
your
alerts,
and
your
warnings
allows
you
to
do
provisioning
and
things
like
that.
So
that's
sort
of
the
window
to
the
product
is
our
product
called
op
center,
without
that
the
ease-of-use
gets
much
much
harder
and
that's
why
it's
important
for
us
to
continue
to
deliver
not
just
the
better
integration
of
the
technologies,
but
what's
the
face,
how
do
I
operate
with
it?
How
do
I
integrate
with
it?
That's
what
that's?
What
option
it.
A
B
Great
great
question:
this
is
where
cloud
providers
come
in
really
handy,
because
sometimes
it's
much
easier
to
spin
up
your
load
tests
in
the
cloud.
Then
it
is
too
you
on
your
own
data
center
so
step
one.
It
becomes
easy
to
do
your
own
testing,
step
2.
This
is
the
beauty
of
open
source.
There's
a
lot
of
published
results
out
there.
Some
of
it
are
from
customers,
like
you
saw
with
Netflix,
where
they
did
as
testing
needed
to
do
a
scalability
test
to
see
not
if
it
would
scale
but
how
it
would
scale.
B
What's
it
going
to
look
like
on
a
performance
curve,
and
that's
where
we
come
up
with
this
linear
scalability,
it
actually
is
a
straight
line,
which
is
a
good
thing.
We've
also
got
some
academic
research
papers
that
have
been
done
where
they
benchmarked
a
lot
of
these
things
at
scale
against
each
other,
to
help
people
understand
what's
the
right
use
case
and
then,
finally,
we're
actually
going
to
be
releasing
some
independent
benchmarks
that
we've
commissioned.
We
should
have
those
by
January
to
help
people
understand.
B
What's
this
going
to
look
like
its
scale,
because
you
hit
on
a
great
point,
it's
easy
to
get
started.
Sometimes
yeah
one
of
the
bigger
problems
is
people
will
get
started
with
a
couple
of
nodes
and
think
I
got
it
and
then
all
of
a
sudden
when
it
gets
to
50
no
200
notes,
200
nodes,
oops,
maybe
I,
didn't
choose
the
right.
Technology
goes
back
to
that
operational
simplicity
problem.
If
you
lose
that
at
scale,
you
lose
everything
you
lose
performance,
you
lose
your
cost
benefits.
You
don't
want
to
be
in
that
situation.
Yeah
can.
A
B
Swap
center,
so
op
center
gives
you
that
window
into
the
product,
where
you
can
start
to
watch
the
load,
you
can
start
to
watch
which
nodes
are
getting
hit
heavily
and
maybe
it's
time
to
expand
one
of
the
other
challenges
for
a
paradigm
shift
for
a
lot
of
the
relational
people.
We
were
taught
for
so
many
years
to
scale
up
and
as
the
the
Moore's
law
continues,
and
the
hardware
gets
more
powerful,
we're
tempted,
sometimes
to
think
throw
more
at
that
node
just
make
that
node
bigger
in
reality.
B
It's
all
about
distributing
that
load
and
that's
what
we
did
in
an
unnatural
way.
With
this
thing
called
sharting
in
the
MySQL
world,
which
was
a
very
unnatural
way
to
do.
We
forced
a
relational
database,
hache
dee,
instant
yeah.
That
was
a
more
elegant
solution
for
the
non
persisted
layer
for
the
in-memory
layer,
but
for
the
relational
layer
it
was
infil
when
we
tried
to
make
that
happen.
Just
wasn't
built
for
that.
A
B
Use
case
so
really
getting
bits
about
the
use
case.
I
would
say
that
it's
hard
to
imagine
a
company
today
who
doesn't
have
and
I'll
use
the
horrible
term
big
data,
but
it
is
hard
to
imagine
a
company,
that's
not
dealing
with
some
of
these
challenges
of
context
and
more
information
who
doesn't
want
more
data
like
what
are
you
comfortable
as
a
business
person
saying?
I
don't
need
to
know
that
about
my
customer
yeah.
Are
you
willing
to
let
any
of
that
go
I?
B
Don't
care
what
you
do,
I,
don't
care
how
traditional
your
business
is.
So
the
next
challenge
is
that
ok,
now,
let's
architect
the
apt
in
the
way,
we're
taking
the
right
tool
for
the
right
job
and
we're
putting
it
in
place
through
the
set
of
api's
to
make
it
screen.
That's
what
we're
seeing
today
very
cool.
Where
do
we
learn
more
about
you,
datastax
calm,
you
can
come
out.
We
got
a
lot
of
information,
a
lot
of
free
information,
white
papers,
videos
all
that
sort
of
thing
very.