►
Description
Speaker: Chris "Mac" McEniry and Igor von Nyssen, Systems Architect at Sony Network Entertainment
Slides: http://www.slideshare.net/planetcassandra/c-summit-2013-stepping-through-the-lifecycle-of-a-service-offering-with-cassandra-by-igor
It's a fine line to walk for incorporating new technologies in an organization with 15+ years of legacy software. In this presentation, we'll look at the lifecycle and adoption of Cassandra from a skunkworks project to a full fledged service in a legacy organization.
A
All
right,
well
Mickey's
big
hand,
is
on
the
the
two
so
we're
going
to
go
and
get
started.
Hi
Chris,
McHenry
I
want
to
welcome
you
to
the
last
sessions
of
the
day.
Last
block
hope
it's
been
a
good
conference
for
everyone,
we're
here
to
talk
about
Cassandra
and
introducing
it
into
a
heavily
traditional
database,
oriented
organization
and
some
of
the
soft
stuff
that
we
had
to
work
through
to
get
it
going
because
I
figure
there's
plenty
of
other
sessions
about
the
hard
stuff
or
the
technical
stuff.
A
So
a
little
bit
back
background
on
us
I'm,
the
ops
guy
you
can
tell
by
the
black
he's
not
wearing
black
he's,
not
the
ops.
Guy,
though
that
being
said,
I
have
a
strong
CS
history.
So
I
do
like
to
dabble.
I
do
like
to
play
around
I've
been
working
with
Cassandra
since
06
and
I'm
going
to
let
my
Deb
guy
introduced
himself
hi.
B
My
name
is
Igor
von
Essen
I've
been
software
architect
for
most
of
my
life
professional
life
doneness,
mostly
in
the
e-commerce
space,
speak
a
little
Java
speak
a
little
Ruby
and
I've
done
cassandra
since
Wando,
mostly
introduced
by
Mackin
on
by
stumbling
upon
a
book
question
here
for
all
of
you
who
actually
has
Cassandra
in
production
right
about
half
the
room
who
thinks
cassandra
is
a
really
cool
technology
and
would
love
to
use
it
in
production.
I
come
on.
That
should
be
the
entire
room
right.
B
C
B
What
we're
talking
about
is
really
what
did
it
take
for
us
to
get
Cassandra
established
as
a
technology
alternative
in
an
organization
that
really
speaks
relational
databases
very
very
well,
and
that
also
speaks
one
particular
brand
of
relational
databases
very,
very
well,
and
so
max
go
talk
about
how
we
got
started.
Oh,
not
actually,
for.
D
B
Tilly
still
me
work
what's
our
organization,
what
do
we
look
like
right?
We
have
hundreds
of
millions
of
customers.
Sne
I
may
be
better
known
as
the
PlayStation
Network
right.
So
on
Christmas
morning
we
have
more
people
playing
on
the
network,
then
live
in
the
country
of
Norway,
and
they
all
want
to
do
this.
At
the
same
time,
right
we
have
billions
of
transactions
every
year,
shifting
pennies
from
leather
left
to
right.
It's
all
micro
transactions
and
we
have
these
Peaks.
B
What
does
it
feel
like,
though?
Sony
itself
is
a
very
large
organization,
huge
company,
and
within
that
we're
somewhat
of
a
start-up.
We
really
have
to
comply
with
everything
that
makes
the
company
great
any
a
three-letter
audit
known
to
man
and
and
at
the
same
time
we
also
have
tremendous
pressure
to
innovate
and
tremendous
pressure
to
make
money
because
you're
not
going
to
be
innovating
for
a
very
long
time.
If
you
don't
make
money
right,
but
now
it's
mike
stern
to
go
and
talk
about
how
we
get
started
on
Cassandra.
B
A
As
I
said,
I'm
the
ops
guy,
though
that
being
said,
I
do
like
to
dabble
I
like
to
explore
I
like
to
see,
what's
out
there,
I've
been
using
Cassandra
since,
before
the
developer
guy
had,
you
know,
had
been
using
Cassandra.
So
from
my
standpoint,
part
of
that's
me
part
of
that's
my
job.
To
go
out.
Try
to
see
what
new
technologies
there
are,
try
to
see
what
we
can
use
in
the
organization.
A
So
it's
all
about
you
know
exploring
and
bringing
that
in.
So
that's
what
motivates
me
as
I'm
looking
for
these
things,
I
got
to
try
to
decide
what
motivates
everyone
else
you
know
our
is
everyone
motivated
in
the
same
way?
Do
they
want?
You
know
that
personal
growth
to
them
with
the
curiosity,
or
are
we
really
just
trying
to
lessen
the
pain?
You
know
because
I
think
anyone
here
can
attest
how
painful
other
technologies
can
be
in
what
we're
trying
to
do
to
avoid
that.
A
So
you
know
what
is
that
pain
again?
I,
don't
want
to
be
woken
up
at
4am,
I,
don't
want
this
to
explode
and
as
the
ops
guy,
the
other
thing
that
always
seems
to
happen
is
I'm
left
cleaning
something
up,
I
try
to
avoid
that,
but
at
the
same
time,
if
that's
my
motivation,
then
we're
not
going
to
get
anywhere,
I
need
to
say:
how
do
we
find
something?
How
do
we
bring
it
in?
How
do
we
make
that
happen?
A
I
can't
keep
saying
no
just
because
there's
going
to
be
some
pain,
because
that
makes
me
the
bottleneck
and
just
like
the
bottleneck
in
any
of
our
technologies,
that's
not
acceptable.
We
have
to
optimize
and
work
around
it
if
we,
if
need
be
so
from
the
bottleneck,
standpoint
I,
think
about
what
some
people
have
been
talking
about
with
regards
to
developer
happiness,
what
platforms
can
I
build
that
makes
the
developers
be
able
to
build
what
they
need
on?
You
know,
on
top
of
that
and
extend
it
and
work
it
out.
A
So
again
you
know
from
that
standpoint.
What
does
a
developer
look
for?
Well,
you
know,
there's
always
that
new
shiny,
you
know,
seem
to
have
stuff
keys.
Dangling
in
front
of
faces
seems
to
happen
a
lot,
but
really,
you
know
there's
feature
comparisons.
There's
you
know.
Does
this
actually
work?
Is
it
mature
enough?
You
know,
and
does
it
actually
solve
problems
that
the
developers
are
looking
for?
Those
are
all
the
things
I
try
to
embody
when
I'm
evaluating
stuff,
but
again,
I
can't
do
this
by
myself.
A
So
as
I'm
just
building
a
platform,
it
really
depends
on
my
developers
to
help
me
out
and
figure
out
what
they
need.
Otherwise,
I
come
up
with
a
bunch
of
solutions
that
are
looking
for
problems
and,
while
that's
kind
of
cool
that
doesn't
necessarily
work
so
I
had
to
start
with
a
problem.
I
have
a
solution.
Now
we
need
to
go
where
the
problem
is
yeah.
B
So
that's
really,
if
you're
interesting,
if
you're
interested
in
introducing
Cassandra
into
your
organization,
the
one
thing
the
one
key
thing
is
finding
the
right
album.
You
really
are
looking
for
a
problem
that
solved
by
Cassandra
very
well
right,
don't
take
any
problem.
If
you
have
one
way,
you
could
simply
say
you
know
what
slapping
Akamai
cash
in
front
up
a
whole
bunch
of
static
data
go.
Do
that
right,
don't
waste
your
time,
building
more
complex
systems.
B
If
you
have
different
ways
of
addressing
them,
you
really
need
a
figure,
something
where
the
problem
that
you're
trying
to
solve
aligns
with
the
technology
and
is
not
solved
well
by
anything
else
that
you
could
put
up.
While
you
are
deploying
this
while
you're
developing
this
there,
everybody
has
their
own
favorite
technology,
their
own
favorite
way
of
solving
problems,
and
they
will
all
have
an
idea
about
how
to
solve
your
particular
problem
and
if
any
of
the
ways
that
they
can
come
up
with
in
solving
your
problem
is
better.
B
Then
you
ought
to
go
down
that
route,
so
picking
the
right
problem,
thinking
really
hard
about
it.
That's
the
biggest
way
in
which
you
can
advance
Cassandra
inside
your
organization,
so
our
problem
turn
out
to
be
a
speed
of
light
problem
and
we
couldn't
really
resist
the
pun
here
and
said:
oh
yeah
we're
going
to
throw
a
little
sea
star
at
the
C
square
it
and
then
our
problem
goes
away
right.
So
do
that
in
a
little
bit
less
cryptic
terms.
B
What
we're
trying
to
do
is
we're
trying
to
solve
a
problem
where
we're
providing
the
same
customer
experience
across
the
globe
to
everybody
on
the
planet,
but
we're
trying
to
provide
authentication
that
is
rapid,
I'm,
also
trying
to
provide
authentication
that
is
always
on.
If
people
can't
play
their
games
just
because
we're
not
able
to
check
their
username
on
that
password,
then
people
get
angry
rightfully.
So
we
learned
that
the
hard
way
to
see
you
guys
may
remember.
B
So
why
is
that
a
speed
of
light
problem
right,
speed
of
light
problem,
in
the
sense
that
if
you
look
at
the
world
from
one
particular
place-
and
let's
assume
we're
looking
at
the
world
from
San
Francisco
because
we're
here,
then
the
customer
experience
that
results
is
directly
correlated
to
the
distance
that
this
person
on
the
planet
is
from
San
Francisco
right,
your
latency,
your
access
time,
your
response
time
that
they
perceive
is
directly
correlated
to
the
speed
of
light
to
the
distance
that
results.
So
how
do
you
go
solve
this
problem?
B
Quite
simply,
you
put
a
lot
of
data
centers
in
a
lot
of
places
right
and
thats
good.
That
gives
everybody
the
same
distance.
That
gives
everybody
the
same
chance
of
having
a
very
short
connection
to
your
closest
point
of
presence.
Now
the
questions
why
this
Cassandra
fit
this
problem
very
well
and,
first
of
all,
it's
not
a
cache
problem.
It's
not
about
putting
things
that
can
be
stale
somewhere
username
passwords.
They
need
to
be
up
to
date,
they're
very
dynamic
data.
B
If
you
put
them
into
all
these
different
places,
then
what
you
really
want
to
do.
Is
you
want
to
make
sure
all
the
data
is
everywhere?
You
can't
really
just
chart
this
and
say:
oh
this
country
gets
their
data
center
in
this
country
gets
their
data
center.
What
do
you
really
want
to
do
is
make
sure
that
if
you
need
a
failover
from
with
their
one
data
center
to
another,
all
your
data
is
the
same.
B
At
the
same
time,
if
you're
playing
DNS
games
to
route
people-
that's
great,
but
it's
never
perfect,
so
really
looking
at
on
having
all
the
data
available
in
all
the
places,
Cassandra
synchronizes
that,
for
you
very
well
same
time,
if
you
have
all
these
points
of
presence,
what
you're
doing
is
you're
asking
for
lines
to
be
cut
right.
The
internet
is
not
stable
by
any
means.
So
anytime,
you
will
have
all
these
connections.
One
of
them
will
be
severed
practically
on
a
daily
basis.
There
cassandra
helps
you
again,
it
picks
up
after
separation.
B
It
survives
partitioning
really
well.
So
that's
a
problem.
That's
very
well
suited
to
be
solved
with
Cassandra.
That
was
the
ideal
situation
for
us.
So
how
do
we
move
from?
We
have
the
sexy
technology.
We
have
the
right
problem
now:
how
do
we
get?
How
do
we
give
birth
to
this
whole
idea?
How
do
we
get
it
in
the
production
you
gotta
go
and
find
allies
right
if
Mac
and
I
hadn't
found
each
other,
and
if
we
hadn't
worked
together
on
this
particular
problem,
it
would
have
never
ever
made
it
into
production.
So
what
we?
B
What
you're
looking
for,
is
you're
looking
for
somebody
who
sees
the
same
problem
for
somebody
who
recognized
that
this
problem
is
important
to
your
organization.
If
you're
trying
to
solve
some
peripheral
issue
that
you
get,
you
can't
get
any
dynamics
behind
it
not
going
to
happen
right
so
pick.
The
right
problem
pick
the
right
allies.
B
What
you
want
to
stay
away
from
now
is
people
who
are
excited
about
this
technology
who
see
the
world
the
same
way,
but
who
do
have
an
inherent
conflict
of
interest
and
that,
for
whatever
reason
they
may
be
the
greatest
people
to
work
with,
they
may
be
your
best
friends
if
their
day,
job
provides
them
or
put
them
in
a
position
where
they
end
up
being
miss
area,
Jekyll
and
Hyde.
That's
not
going
to
help,
but
it's
not
get
help
you're
going
to
help.
B
You
doesn't
mean
that
you
exclude
them,
but
it
means
you
will
not
relying
on
them
as
an
hour
s
as
an
ally
right
at
the
same
time
go
and
develop.
Your
mantra
develop
your
elevator
pitch
about
what
the
problem
is,
that
you
try
to
solve
again,
have
countless
conversations
with
people,
it's
a
Oh.
What
is
the
problem?
What
are
you
trying
to
solve?
Why
don't
you
really
have
just
a
solution?
I
do
you
really
just
try
to
pick
or
push
this
particular
technology.
B
You
need
to
be
prepared
for
those
conversations
and
you
need
to
be
prepared
to
give
an
answer
in
under
30
seconds
right.
It
really
helps
articulate
the
things
that
you're
trying
to
address.
It
really
helps
getting
people
behind
it.
Boil
it
down
to
something
very,
very
simple
hours
on
ended
up
being
latency,
latency
latency,
but
something
that
you
can
really
coined
the
term
that
you
can
rally
people
behind
right
and
right
now,
it's
time
to
get
a
little
uncomfortable,
I,
think
yep.
A
So
you
have
a
new
technology
you're
playing
around
it.
You
found
the
right
problem.
You've
gotten
some
allies.
You
start
rolling
it
out,
you're
going
to
be
outside
of
your
comfort
zone,
no
matter
what
you
want
to
do,
it's
it's
going
to
be
the
new
experience.
It's
going
to
be
knowing
what
the
new
operational
items
are.
It's
going
to
be
a
it's,
this
new
and
shiny.
What
does
that
actually
mean?
So
keep
in
mind?
That's
okay!
You
know
in
any
kind
of
situation,
it's
only
going
to
work.
A
If
you
take
some
experiments,
certain
experiments
are
inherently
going
to
make
you
uncomfortable.
So
the
other
thing
to
keep
in
mind
going
back
to
the
Allies
and
the
anti
allies.
You
know
the
proofs
on
you
when
it
comes
down
to
it,
you
need
to
be
the
one
who
they're
going
to
come
to
you
and
say
hey.
This
is
working
great
or
this
isn't
working
what's
going
on.
How
do
we
fix
it?
A
You're,
the
one
who's
going
to
have
to
answer
for
that
and
that
is
uncomfortable
as
well,
but
again,
you're
experimenting
that
should
be
okay,
the
key
through
all
that
is
to
stay
positive
and
one
of
the
ways
you
stay
positive
is
by
avoiding
the
naysayers.
There
we
go
so
there's
a
new
kind
of
racy
chart
out
there.
Well,
ok,
it's
a
cute
slot,
but
it
kind
of
comes
down
to
you
know.
I!
Can't
you
know
stress
enough
that
you
in
your
initial
implementations,
you
have
to
pick
the
right
people.
A
A
So
then,
the
quick
analogy
is,
if
you're,
trying
to
put
an
oak
tree
in
the
desert,
whether
that's
a
good
idea
or
not
you're,
not
you
know,
if
you
just
put
it
there,
it's
going
to
die,
but
if
you
start
to
build
a
greenhouse
and
Bill
derogation
around,
it
has
a
chance
to
live,
and
this
is
what
this
is
about,
making
sure
that
what
your
investment
is
you're
protecting
on.
So
if
you
live
in
a
desert,
this
is
what
you
need
to
remember.
A
Yep,
the
other
thing
at
this
time
in
the
life
cycle.
What
you
need
to
start
thinking
about
is
your
models,
and
you
know
you
can
go
through
and
be
like.
Yes,
cassandra
has
the
nodes,
it
has
its
data
centers,
it
has
its
topology.
That's
one
set
of
models.
That's
your
architectural
model,
the
other
things
which
are
the
ones
that
are
in
a
lot
of
ways
going
to
be.
The
proof
is:
what
is
your
deployment
model?
You
know?
A
Is
this
going
to
be
attached
to
multiple
applications
as
it
can
be
single
e
attached
to
a
single
application?
Does
that
bring
in
security
concerns?
Does
that
bring
in?
You
know
data
integrity.
You
know
sorry,
data
commingling
concerns
anything
like
that,
so
you
have
to
decide
how
your
interaction
with
your
applications
are
going
to
be.
You
need
to
also
decide,
especially
as
you're,
offering
this
out.
What's
your
support
model,
you
know
it's
one
thing
to
say:
hey
here's,
a
Cassandra
cluster
go
use,
it
and
people
are
like
well.
How
do
I
connect
to
it?
Okay?
A
Well,
you
do
this
or
go
talk
to
these
other
guys,
they're
already
connected
to
it.
You
know
both
these
formal
and
informal,
you
know,
community-based
or
on
your
own
support
models
are
valid.
You
just
have
to
decide
which
one
works
for
your
organization,
you
know,
and
so,
as
I
put
this
together
from
the
outside
from
a
platform
building
side,
I,
look
at
it
from
a
service
offering
model
which
really
is
the
the
audience
for
this
as
a
developer.
Who
needs
to
use
the
service
or
wants
to
use
the
service?
A
It's
going
to
give
them
a
quick
description.
What
it
is,
how
mature
we
feel
it
is
as
both
the
technology
itself
and
our
ability
to
execute
on
the
technology.
You
know
virgin
topologies.
You
know
sizing
how
to
connect
who
to
get
how
to
request
help.
You
know
all
these
things
are
the
stuff
you
start
to
fill
out,
and
these
are
just
kind
of
that.
The
high
level
items
of
what
you
need
to
do
doesn't
mean
you
have
to
resolve
all
of
these
at
this
time.
A
If
you're
not
doing
anything
important
with
the
data,
maybe
you
don't
need
a
dr
plan.
That's
acceptable,
but
you
at
least
have
to
go
in
there
with
a
conscious
effort
of
this
is
what
we
care
about,
and
this
is
what
we
don't
care
about,
and
so,
once
you
have
the
service,
offering,
you
can
say,
hey,
developer,
here's
a
service
offering
go,
build
something
on
it.
B
Right,
so
how
do
you
get
from
this
minimal
service
offering
to
something
that
you
have
deployable
in
production?
It's
important!
You
don't
start
with
the
crown
jewels
right,
so
you
picked
your
problem.
The
problem
sexy
great,
but
realistically,
as
in
developers,
a
development
organization,
you
really
have
a
very
small
understanding
of
what
Cassandra
is
what
it
could
be
right,
pick
something
that
is
easy
for
you
to
get
to
solve.
That's
small
enough,
so
that
you're
not
stumbling,
but
you
need
to
look
for
the
win.
B
You
need
to
look
for
something
that
will
work
so
you
take
a
minimal,
Viable,
Product
approach
to
this
whole
thing
and
so
at
you're.
Looking
to
start,
hopefully
with
transient
data,
we
picked
authentication
tokens.
They
live
for
about
20
minutes
great.
So
even
if
everything
goes
pear-shaped
and
we
lose
them,
yes,
we're
looking
at
something:
that's
not
nice,
but
it's
not
going
to
cost
us
the
company
all
right.
So
look
for
problems.
Look
for
applications
that
have
this
transient
nature,
while
you're
getting
eased
into.
How
do
I
store
data
in
Cassandra?
B
How
do
I
operate
the
cluster?
How
do
I
really
interact
with
all
of
this
right?
Second
go
pick
which
driver
you
want.
There
are
plenty
of
them
out
there
Hector,
as
Diaw,
knocks
on
the
new
datastax
driver,
look
at
which
one
figure
fits
your
development
methodology
best
look
at
which
data
models
you
want
to
drive
from
that
and
learn
about
how
to
model
data
within
Cassandra
write
the
same
thing.
You
want
to
look
at
how
data
gets
stored
in
Cassandra.
You
need
to
look
at
what
I
see
how
the
serialization
happen.
B
How
do
we
actually
use
comparators
this
whole
time
series
concept
and
then
look
at
all
the
terms
that
you
think
you
know
right.
So
I
looked
at
the
term
index
and
I
thought
it
I
nu
what
an
index
does
turns
out.
That's
only
half
true
when
you
look
from
a
relational
database
model
over
to
Cassandra
indexes,
especially
secondary
ones,
aren't
really
the
same
thing
that
you
used
to
free
search.
A
So
we
got
something
small:
it's
running
it's
going
in
production,
now's
the
time
to
start
getting
really
serious
about
this.
You
have
proof
you
have
something
that
shows
this
works.
You
know.
Take
the
training
wheels
off
go,
buy
a
bigger
bike,
you
know
get
out
the
daddy's
credit
card
and
you
know
pony
up
for
support
or
you
know
this
is
where
you
start
engaging
datastax
enterprise,
because
you
can
conceive
and
use
the
additional
use
cases
that
it
provides.
You
start
having.
A
You
know,
it's
no
longer
a
completely
theoretical
question:
it's
you
have
some
real
data
to
back
this.
Up
with
you
can
start
introducing
other
data
classes.
All
the
assumptions
you
made
in
the
first
place
you
can
be
like
hey,
you
know,
yeah.
We
had
transient
data,
but
now
I
have
a
backup
strategy,
because
I
know
how
to
do
snapshots
and
save
those
off
and,
more
importantly,
I
know
how
to
restore
them.
Keep
in
mind,
that's
that's
the
real
part,
the
real
reason
for
backups.
A
So
you
can
start
persisting
data
or
using
persistent
data.
You
start
introducing
your
data
life
cycle
management.
You
know
this
is
what
it's
going
to
look
like.
This
is
where
it's
going
to
tear
into
this
is
what
is
more
acceptable
data
start
introducing
all
of
the
you
know,
squirmy
data
PII
pci,
any
of
those
kinds
of
things.
This
is
your
chance
to
start
saying:
hey.
We
can
now
tackle
those
the
other
at
the
same
time,
in
doing
that,
you
have
to
acknowledge
your
gaps.
A
Okay,
if
you
don't
have
all
ten
occation
running,
you
know
the
authentication
agents
running
on
your
on
your
Cassandra
nodes,
you're,
probably
not
going
to
want
to
be
doing
pci
data
in
there
or
even
PII.
If
you
don't
have
at-risk
encryption
going
on,
maybe
you
avoid
the
pci
data.
You
know
look
at
those
gaps,
decide
which
ones
you
need
to
tackle,
and
you
know
introduce
your
security
team
to
what's
going
on
and
see
what
they
look
at
and
what
they're
concerned
about
yeah.
A
The
key
thing
is:
this
is
the
awkward
teenage
years
you
got
to
revisit
your
assumptions
and
revisiting
some
of
your
assumptions
you're
going
to
have
to
revisit
who
your
allies
are.
You
know
this
is
where
you
need
to
go
back.
Get
more
allies,
as
I
said,
we've
engaged
security,
we
started
talking
to
them.
We
just
engage
some
of
the
traditional
database.
People
start
talking
them
additional
application
teams.
You
know
we
originally
started
with
one
application
team.
Now
we
have
four
using
it
and
a
fifth
one
coming
online.
A
You
know
these
these
things
grow
when
they
grow
incrementally
until
they
hit
some
critical
mass
and
take
off.
You
know
the
downside
of
all
this
is,
as
you
engage,
more
people
keep
in
mind.
You
are
going
to
be
bringing
on
more
battles.
You
are
going
to
be
bringing
on
more
questions.
You
got
to
be
ready
for
them
and
again
you
got
to
acknowledge
the
gaps
that
are
there,
because
the
proof
again
is
going
to
come
back
to
you.
You
have
some
proof.
A
At
the
same
time,
you
want
to
solidify
you
know
all
the
other
items
on
the
service
offering
makes
you
again,
backups
have
some
plan
for
them.
You
know
authentication
deployment
models,
peacebuilding
that
out,
maybe
exploring
different
deployment
models.
You
know,
based
on
data
segmentation,
we
put
all
the
key
critical
data
over
here
we
put
the
non
very
cold
ones
and
other
clusters,
you
know,
figure
out
what
your
topologies
look
like.
From
that
standpoint,.
A
The
other
thing
you
know
you
think
it
works.
Whenever
you
have
issues
with
this,
you
can
always
go
back
and
again
expand
that
capital
that
you've
already
gained
up.
You
know
use
your
mantra.
You
know
again
with
us,
it
was
latency
latency
latency.
If
you
have
proof,
don't
forget
that
don't
rub
it
in
people's
faces,
but
don't
forget
that
and
then,
as
you
go
through
this
teenage
transition,
you're
obviously
into
the
adult
phase,
and
this
is
where
something
is
within
its
full
lifecycle
and
honestly
for
us
we're
a
large
company.
A
It
took
us
a
year
and
a
half
to
get
where
we
are
even
at
this
fit.
What
you
know
in
this
adult
phase,
I
can't
even
consider
we're
there
our
future.
So
the
other
thing
is,
you
got
to
be
patient
with
a
lot
of
this
and
don't
forget
that
you
know
in
large
companies.
You
know
transition
takes
time.
There's
more
people
there's
more
moving
parts.
You
can
go
back
to
the
keynote
and
listen
with
regards
to
you
know
the
endowment
effects
and
and
all
the
expended
you
know,
capital
sunk
costs.
It's
all
there.
A
You
got
to
keep
that
in
mind.
So
along
those
lines
you
know.
Where
are
we
now
I'm
going
to
steelers
equation
and
say
this
isn't
necessarily
a
latency
issue
anymore?
What
we've
seen
is
a
large
company.
Our
bigger
issue
is
our
mass
or
inertia
or
momentum.
Whatever
you
want
to
call
it
our
ability
to
change.
You
know
it
takes
a
lot
of
energy
for
us
to
get
a
deployment
out
the
out
the
door,
but
that's
because
we
have
this
mass.
A
B
A
C
A
Mean
the
initial
pass
honestly
is
all
self-taught.
I
mean
one
of
the
things
which
we
didn't
go
into
is
the
fact
that
Cassandra's
open
source-
you
know,
there's
no
budget
approval.
You
can
play
with
it
on
your
own.
You
can
run
a
single
instance.
It's
actually
fairly
easy.
To
start
an
instance.
You
know
maintaining
a
large
cluster,
that's
different,
but
starting
it
is
relatively
easy.
A
B
From
a
development
perspective,
we
didn't
start
on
what
Cassandra
and
ended
there
Cassandra
was
on
the
winner
of
a
large
evaluation.
We
looked
at
a
whole
bunch
of
other
technologies
who
looked
at
in-memory
data
grids.
We
looked
at
a
whole
bunch
of
other
no
sequel
databases
and
really
looked
at
the
use
case,
for
what
we're
trying
to
solve
and
said
which
of
these
technologies
solves
are
in
case
our
use
case.
The
best
and
Cassandra
was
the
winner
primarily
because
it
simply
scales
linearly
right.
B
D
A
So
I
know
the
the
other
customer.
You
know
the
other
application
team
customers
are
using
it.
One
is
the
use
case
is
largely
an
audit
log
usage
and
then
I
know
the
additional
off
teams.
There's
the
doing
the
authentication
tokens
on
the
on
the
server
side
and
then
there's
the
clients
of
those
authentication
tokens
which
end
up
being
other
applications
within
our
service
Phylly.
So
they
actually
have
their
tokens
inside
Cassandra
as
well,
and.
B
It's
really
simple:
to
find
the
adjacent
problem.
Once
you
have
the
first
once
off
successfully,
we
have
lots
of
people
that
are
looking
at
this
and
I
say
all
this
is
cool.
It
has
all
these
aspects
of
how
I
can
store
data,
how
I
can
look
up
data,
how
I
can
deal
with
distributing
data
and
all
of
a
sudden,
you
find
more
and
more
possible
applications
right
and
that's
when
ya,
when
you
start
to
build
a
momentum
inside
the
organization,
that's
where
it's
going
to
be
easy
to
find
the
next
problem
that
looks.
B
A
One
of
the
examples
of
that
is
right.
Now
we
have
one
team
using
you
know
going
back
to
the
the
new
team.
That's
using
it
is
doing
it
as
an
audit
log.
We
have
an
existing
events
flow
where
events
get
process
can
send
all
the
way
back
to
the
warehouse.
You
know.
One
possibility
is
those
event
logs,
look
exactly
like
the
audit
logs.
So
right
now
we're
already
drawing
comparisons.
So
it's
really
easy.