►
From YouTube: Cassandra Community Webinar | Keep the DB, Lose the A
Description
Christos Kalantzis, Engineering Manager of Cloud Persistence Engineering at Netflix, will talk about his transition from a database administrator to a database engineer working with cutting edge big data technology. He will provide expertise and insight into how he learned about new technologies and the use cases he now gets to work on.
A
Hello,
everyone
and
welcome
to
this
week's
edition
of
our
Cassandra
community
webinar
series.
I
am
really
excited
today.
To
have
with
me
Christo's
collapses
from
netflix
Christos
is
the
engineering
manager
of
the
cloud
persistence
group
button
at
Netflix
and
where
he's
going
to
talk
about
today
is
he
comes
from
a
heavily
relational
database
background
in
a
previous
life,
he
was
a
DBA,
and
now
he
has
become
a
database
engineer.
A
So
he's
going
to
talk
all
about
that
transition,
so
I'm
going
to
hand
over
to
Chris
thought
now,
but
before
I
do
I
just
wanted
to
remind
everyone.
If
you
would
like
to
ask
a
question
of
priestess,
please
use
the
Q&A
tab
in
side
of
WebEx
post
your
questions
there
and
at
the
end
of
today's
session,
I
will
go
through
those
questions
asked
resource
and
we
will
get
through
as
many
as
we
can
so
without
further
ado,
I
would
like
to
welcome
christos
calanctus.
B
B
So
we're
going
to
cover
this
topic
and
everything
it
entails
in
four
steps.
First,
we'll
start
with
a
brief
history
of
databases,
the
evolution
of
the
technology
and
the
challenges
we
faced
with
every
step.
You
know
a
wise
man
once
said.
You
need
to
know
where
you've
been
to
understand
where
you're
going
will
then
look
at
how
the
dissolution
of
databases
has
molded
current
DBAs
and
then
dilute
the
list
of
skills.
B
So
a
brief
history
of
databases
in
the
60s.
You
know,
while
a
lot
of
people
were
celebrating
peace
love,
you
know,
computerized
databases,
you
know
made
their
debut
in
the
sixties.
Since
computers
are
getting,
you
know
more
cost-effective.
A
lot
of
private
companies
were
using
them
more
and
more
and
relying
on
two
popular
database
models
existed
in
the
60s
one
was
the
network
model
called
Curtis
ill
and
there
was
also
the
hierarchical
model
called
IMS.
B
B
A
B
There
Edgar
cod
roe
taper
about
the
relational
database
model,
which
you
know
pretty
much
started
our
DBMS
revolution
in
the
70s.
Two
major
prototypes
were
created
based
on
this
model.
One
was
ingress
which
eventually
begat
ingress,
indeed
at
sybase,
which
in
turn
turned
into
CMS
sequel,
server,
there's
also
system.
Our
was
another
one
of
those
prototypes
and
the
modern,
db2
and
Oracle
were
really
modeled
after
that
prototype.
B
Also
in
the
70s,
the
pH,
and
he
came
up
with
the
entity.
Relational
model
I
mean
that
you
know
it's
basically
what
we
use
today
to
focus
on
designing
database
applications
instead
of
focusing
on
you
know,
the
logical
database
structure
now
you'll
notice
that
in
the
two
prototypes
I
listed
ingress
and
system,
are
they
both
used?
Different
query
language,
english
used
to
you?
Yell
system
are
used,
sc2
uel,
and
you
know
there
was
very
little
interaction.
B
If
you
want
it
was
very
hard
to
work
with
the
other.
Well,
that
was
all
six
mini
eighties,
so
you
know
a
lot.
We
catch
a
new
wave
pop
songs
and
drug
in
his
hair
and
clothing
styles.
There
was
also
SQL,
which
made
its
debut
the
structured
query.
Language
SQL
really
became
the
standard
way
to
query
all
these
database
system.
All
these
relational
database
systems,
which
we
you
know
still
use
today.
B
Another
thing
happened
in
the
80s
as
well:
I
mean
personal
computers
became
popular
became,
affordable
and
with
that
came
really
slow
database
products.
Some
of
them,
you
might
know
you
might
have
never
heard
of
which
quite
frankly,
makes
me
feel
kind
of
old.
There
was
paradox:
whatcom
SQL
and
my
personal
favorite
debase
yeah
I
miss
Davis.
That
would
be
a
tease
for
you.
B
Then
the
nineties,
camer
grunge,
replaced
top
and
also
at
this
point
database
is
really
you
know,
got
firmly
rooted
in
companies
all
over
the
world.
I
mean
I,
think
we
all
remember
the
3p
replication
model,
client
server
database
right
all
those
applications
mostly
served
internal
users
or
limited
customer
base.
You
know
it
only
had
to
support
hundreds
or
or
in
the
case
of
really
big
companies.
Maybe
thousands
of
users.
Well,
the
internet
came
along
and
I
really
changed
all
that
the
toward
the
end
of
the
90s,
the
internet,
exposed
applications
to
potentially.
B
Of
users
and
if
you're
lucky
you,
you
cashed
out
before
your
database
crashed
but
seriously
a
traditional
RDBMS
systems
were
being
pushed
to
their
limits.
I
remember
in
the
late
90s,
some
of
my
video
in
my
early
to
be
a
career.
You
know:
we've
been
wads
of
money
to
scale
up
those
servers.
We
hadn't
ran
faster
disks.
We
chose
servers
with
dual
power
supplies.
B
B
92,000,
cable,
you
know
in
in
the
web
two
point:
O
world.
We
became
the
product
companies
starting
to
realize
the
value
of
data,
both
the
Bulls
of
data
they
were
collecting
and
especially
in
the
value
of
the
data
they
were
not
collecting.
So
the
scramble
began
to
collect
big
and
I
mean
huge
amounts
of
data.
A
B
It
to
recommend-
and
you
know
better
movie
choices
too-
aren't
we
nice
some
new
techniques
to
store
this
massive
amounts
of
data
on
our
DB
masses,
which
were
still
in
the
early
two.
Thousands
was
the
prevailing
database
products
out
there.
You
know
new
techniques
were
created,
sharding
cable,
you
know
sharding,
which
was
good,
because
you
continue
to
use
your
existing
already
being
a
solution
to
store
these
massive
amounts
of
data
and
to
be
able
to
query
it
effectively,
but
I
mean
that
added
complexity
to
applications.
B
Now
you
know
the
location
of
visibility,
the
logic
to
store
the
to
store
and
retrieve
certain
data
was
built
into
the
application
and
higher
up
into
the
abbot
into
the
whole
application.
Stack
then
also.
Another
thing
that
happened
in
into
thousands
disks
became
really
really
cheap.
You
know
we
no
longer
felt
the
need
to
be
stingy
with
it
with
our
storage,
so
we
started
to
be
normalized
our
data.
You
know
the
thought
being
that,
if
we
simplify
the
model,
less
joins
were
needed
to
serve
a
request.
So
if
there's
less.
A
B
Now
I
also
talked
about
how
companies
you
know
sort
of
value
of
the
data
they
were
collecting
and
started
slicing
and
dicing.
This
massive
amount
of
data
well
to
be
able
to
do
that.
Your
techniques
in
analyzing
data
also
evolved
into
thousands.
You
know
technologies
like
mapreduce
that
led
to
products
like
Hadoop,
you
know
now
the
analysis
of
all
this
unstructured
data
from
different
sources
can
be
spread
out,
amongst
hundreds
of
notes,
allowing
companies
to
use
commodity
hardware
or
or
even
dispositive.
In
the
case
of
you,
know,
Netflix
disposable,
virtualized
servers
to
perform.
B
You
know
these
offline
computations,
so,
unlike
in
the
night,
is
when
we
were
scaling
machines
up.
You
know
in
the
2000s
we
achieved
with
our
hardware
and
started
using
commodity
hardware
and
and
and
the
products
we
were
using,
allowed
us
to
use
that
to
to
use
that,
because
of
the
availability
that
was
built
into
these
products
now
so
as
the
databases
and
the
challenges
run,
these
databases
database
applications
evolved
over
the
last
few
decades.
You
know
so
did
the
individuals
who
are
responsible
to
tackle
these
problems,
these
individuals
or
some
combo
you
know-
were.
B
B
B
At
Grails,
look
at
the
Django
you're
still
in
the
function
of
an
RT
BMS
in
the
persistence
layer,
so
dba's
you
know
mean
to
make
sure
the
physical
model
reflects
what
the
application
is
expecting
and
still
provide
the
SQL
to
create,
modify
and
query
the
data.
It
is
again
I
mean
even
though
a
lot
of
these
a
lot
of
these
application
frameworks.
I
can
do
that
for
you
it's
you
know.
B
My
experience
is
that
they're
less
than
ideal,
and
you
still
need
a
good
DBA
to
make
sure
things
are
things
are
ideal
at
the
physical
daily
deviations
are
also
data
product
experts.
You
know
they
need
to
know
they
need
to
know
either
Oracle
sequel,
server,
DB
PPD
to
or
my
sequel,
and
really
understand
how
the
individual
database
product
interprets
a
query
in
the
underlying
query.
A
B
A
A
B
In
your
database
server
that's
running
on
your
database
server,
you
know
that's
very
important,
because
you
know
you.
Gays
are
often
responsible
for
the
installation
of
the
database,
product
and
troubleshoot.
You
know:
troubleshoot
issues
of
bio
performance
find
oil
bottle
next
identify.
If
there's
enough
RAM.
B
B
B
B
They
also
became
responsible
for
choosing
how
it
for
have
that
particular
database
server
required,
and
you
don't
leave
enough
space
there
for
growth.
Another
thing
they
became
responsible
for
it
is,
and
it's
personal
at
least
favorite-
is
measuring
the
amount
of
ups
required
to
run
the
application
and
decide
between
iris
cutting
SATA,
and
you
know
a
fan.
B
A
B
So
it's
all
very
impressive
I
mean
it's
a
lot
of
responsibility
that
we
that
we
put
in
dba's
hands.
Well,
it's
good,
but
in
today's
big
data
reality,
unfortunately
not
enough-
or
at
least
it's
different.
What
we
need
today
are
BB
engineers,
so
so
what
is
that?
A
deep
attention
year
is
is
more
than
a
DBA,
because,
although
all
the
DBA
skills
are
necessary,
there's
also
some
new
skills
that
need
to
be
developed
and
we'll
cover
those
skills
in
a
little
bit.
B
B
The
engineer
needs
take
a
step
back,
look
at
the
forest
from
the
trees
and
understand
that
uses
or
her
decision
will
greatly
affect
multiple
layers
of
the
application
stack.
It's
even
more
responsibility
than
me
a
set
before,
so
what
are
the
skills
required
to
be
a
DB
engineer?
Let's,
let's
start
looking
at
that,
we
talked
earlier
about
how,
in
this
decade
the
collection
of
business
has
grown
astronomically.
So
the
DB
engineer
needs
to
figure
out
highest
world.
B
B
A
B
A
B
Are
you
willing
to
try
a
new,
no
sequel
design?
You
know?
Was
that
mean
to
the
dev
organization
and
your
existing
code?
You
know,
like
we
said,
there's
nowhere
joins
the
responsibility
for
joining
will
be
at
the
application
level
will
be
more
seeks
for
secondary
indexes.
So
you
know
in
trade-offs
and
the
advantages
in
the
trade-offs
of
going
no
sequel
need
to
be
understood.
B
Another
thing
is,
you
know,
how
will
your
existing
data
model
translate
to
distributed
data
model?
You
know
like
we
sit
in
a
look,
look
like
I
spoke
and
also
with
no
sequel.
You
know
if
you've
got
multiple
nodes
and
a
cluster,
multiple
shards,
you
can
join
between
them.
You
may
be
able
to
join
in
between
a
shard
because
it's
still
it's
still
possibly
in
ER
model,
but
in
addition.
A
A
B
B
Next,
if
you
can
go
down
photo
sequel
routes,
fusion-io's
will
then
need
to
implement
these
new.
You
know
noseda
models
using
one
of
you
database
products
out
there
I
mean
there's
I
mention
Apache
Cassandra
leaders,
MongoDB
or
you
know,
some
people
are
using
HBase,
which
one
to
use
really
depends
on
your
use
case.
I
mean
Cassandra
is
blazingly
fast
with
rights,
because
all
rights
are
immutable.
B
B
Every
and
it's
operationally
more
complex
to
run
the
cassandra
cluster,
so
the
engineer
decide
the
appropriate
technology
news
based
on
you
know:
half
the
company
works
at
you
know,
embedded
java
shop.
If
it's
a
java
shop-
and
you
know
Cassandra-
might
be
a
better
choice
because
it's
open
source,
it's
nice.
A
B
Java
code-
and
you
know
it's
not
difficult
to
patch
it-
and
roll
in
your
own
sticks-
hopefully
you're,
giving
back
to
the
community
as
well,
so
that's
also
a
factor
of
which
which
products
to
use-
and
if
you
engineer,
really
really
needs
to
understand
the
trade-offs
between
every
product,
so
we'll
have
to
play
with
these
different
products
and
understand
them.
Also
another
factor
which
product
to
use
is:
are
you
going
to
run
this
on
commodity.
B
Or
even
virtualize
your
database
here
in
netflix,
we
run
apache
cassandra
in
the
amazon
ec2
cloud,
not
on
physical
hardware.
However,
one
of
the
reasons
why
choose
cassandra
is
because
of
the
flexibility
it
allows
to
be
able
to
do
that
and
plus
it's
open
source.
Netflix
is
a
java
shop
and
we
patched
Cassandra.
B
B
B
A
B
B
B
You
know,
you're
introducing
another
hop
another
layer
in
your
architecture
and-
and
you
know
most
probably
increasing
latency
of
abilities
of
you-
know-
calls
between
different
layers
of
the
application,
so
I
mean
also,
you
know,
there's
there's
an
issue
of
consistency
as
well.
A
lot
of
these
distributed
database
systems
are
virtual,
consistent
systems.
I
mean
there
are
some
that
are
tunable,
but
if
you
really
want
to
get
you
know,
speed.
B
Afterwards,
you
know
you
might
not
get
what
you're
expecting,
because
it
hasn't
replicated
across
the
data
has
been
replicated
across
all
nodes.
So
you
know
there's
a
chance,
you
just
might
not
get
your
data,
retrieval
might
not
inconsistent,
and,
and
so
you
need
to
help
developers
and
application.
Architects
understand
that
and
build
build
mitigating
tools
into
the
application
stack
to
handle.
That.
B
Finally,
the
DB
engineer
also
needs
programming
skills
beyond
just
the
automation
and
scripting
the
performance
tbas.
So
we
talked
a
lot
about
you
know,
knowing
architecture,
you
know
recognizing
the
changes
that
need
to
happen
in
different
layers
of
application.
A
lot
of
times,
there's
going
to
be
resistance
to
change,
I
mean
old
versus
other.
So
there's
frequently.
B
Hey
you
better
brush
up
on
your
Java,
you
know,
there's
companies
that
are
Ruby
shops,
ticon
shops,
groovy
shops,
so
I
mean
you're
going
to
have
to
develop
prototypes
of
how
the
application
will
communicate.
With
with
the
persistence
with
this
new
persistence
layer
and
a
new
data
schema
also
I
mean
a
lot
of
times
prototypes
turned
into
production
code.
So
you
know
your
prototype.
Your
prototyping,
this
mainly
due
to
being
up
recipes
for
other
developers
and
your
company
to
model
their
implementation
on.
So
we're
still
developing
programming
skills
are
very
very
important.
B
A
B
Shops
and,
and
especially
of
companies
that
are
big
data,
that
big
data
focused
applications,
TVA
positions,
you
know
they're
being
outsourced
their
being
handled
remotely,
and
you
want
to
stay
cutting-edge
with
your
skills
and
really
remain
highly
marketable.
Then
you
really
need
to
kick
that
day
from
your.
You
know,
BB
administrator
title
and
and
be
a
DB
engineer.
Instead,
I
know
Netflix
I
don't
hire.
B
Tpa
is
anymore,
they
just
simply
don't
have
the
skill
set
I'm
looking
for-
and
you
know,
vb
engineers
are
where
it's
at
whether
it's
dev,
ops
or
or
in
an
application
and
application
development
team,
they're
becoming
really
really
important,
and
you
know
I
really
do
hope.
Dba
is
out
there.
You
know
continue.
B
A
B
A
A
B
So
it's
februari
six,
it's
an
our
offices
in
Los
Gatos
and
it's
after
six,
thirty
observing
food
beer
and
doing
some.
You
know
quickfire
presentation
of
our
open
source
projects
for
those
who
want
preview,
go
to
Netflix
github
com
and
we'll
also
have
demo
stations
actually
showing
off
this
software.
So
it's
going
to
be
tons
of
fun
if
you're
interested
in
in
open
source
software
and
see-
and
really
you
know,
look
at
a
company-
that's
that's
invested
highly
an
open
source
and
in
all
the
source,
their
own
stack.
Please,
you
know
come
out.
A
Okay,
great
and
we'll
start
to
take
questions
now
so
feel
free
to
ask
your
questions
in
the
Q&A
tab
in
WebEx,
so
while
you're
typing
those
and
I'm
getting
ready
to
ask
them
of
christos
a
couple
of
other
things
for
your
calendars.
So
we
have,
if
you're
on
the
East
Coast,
we
are
very
large
Cassandra
centric
event
in
March,
March
twentieth,
and
that
is
on
the
datastax
website.
A
lot
of
what
Christos
talked
about
today.
A
You
know
making
this
transition,
you
need
some
educational
materials
for
this
webinar
series
is
an
educational
webinar
series
to
make
sure
you
do
not
miss
the
next
installments
as
well,
but
also
planet,
Cassandra
org.
That's
the
Cassandra
community
website
that
has
a
lot
of
resources
in
it
as
well,
so
that
you
can
start
to
get
skilled
up
in
Cassandra
and
big
data.
So
Christos
we've
got
some
questions
coming
in
right
now,
oh
I'll
take
Arthur
zubarev.
A
He
says
I'd
like
to
see
more
practical
exam,
Cassandra
uses,
implementations
and
developer
experiences.
There
is
a
user's
page
on
planet
Cassandra,
which
has
some
use
cases
on
there.
We
are
adding
more
and
more
and
as
the
next
phase
of
the
college,
credit
webinar
series
were
actually
going
to
be
featuring
use
cases
from
different
companies
front
and
center.
So
thank
you
for
that
feedback.
Also.
B
B
A
Thank
thank
you
very
much
for
that.
Author
ping
me
Christian
at
date
respects
com.
I
will
get
you
a
discount
code
for
The
New
Yorker.
Then
we
are
featuring
a
lot
of
use
cases
at
that
event,
from
comcast
from
ebay
and
Blue
Mountains,
several
others
basements
and
great
use
cases
listo.
So
I'm
going
to
take
this
one
from
a
scene
dial.
We
are
an
Oracle
shop
and
are
very
used
to
sequel
and
ad
hoc
queries.
Other
similar
tools
for
Cassandra
is
query
the
database.
A
So,
yes,
there
is
a
tool
called
cql
Cassandra,
query
language
that
is
very
sequel
like
and
allows
you
to
easily
query
the
database.
If
you
are
familiar
with
Oracle,
you
know
of
a
tool
called
toad
and
there
is
actually
a
toad
for
cloud
databases.
That's
a
free
tool
available
from
quest
software
now
del
which
allow
you
to
use
standard
sequel,
to
call
cassandra
and
a
bunch
of
other
databases
as
well:
HBase,
hi,
fun,
bunch
of
others,
so
check
that
out.
B
Is
it
possible
to
use
Cassandra
without
involving
Java?
Well,
yes
or
no,
so
let
me
explain.
Arthur
cassandra
is
the
java
application
after
all,
so
the
server
itself
needs,
however,
clients
don't
actually
in
each
other.
I
believe.
There's
client
libraries
for
4c
sharp,
4c
fervor,
definitely
for
python,
so
there's
definitely
client
libraries
for
languages
other
than
Java.
A
All
of
those
client
drivers
are
available
on
planet
Cassandra
as
well
for
free
download.
There's
a
community
drivers
so
check
those
out.
Okay,
so
crease
toes
when
setting
up
a
new
production
Cassandra
cluster
in
AWS.
What
are
the
most
common
failures
or
config
issues
you
could
run
into,
and
how
do
you
mitigate
those
that
may
be
a
question
for
Jason
or
one
of
the
other
team,
but
wanted
to
put
it
on
your
radar?
Sure.
B
B
Level,
you
know
more
detailed
answer,
keeping
it
for
very
long
time
and
probably
requires
one
of
my
developers
to
dive
that
deep,
but
mainly
the
main
challenge
with
running
Cassandra
in
AWS
is
limited,
I,
oh
so
so,
if
you're
running
an
on
fs-d
ec2
instance,
you
know
there's
very,
very
limited,
io
available
team.
It's
actually
it's!
Actually,
if
there's
a
quick,
you
know
it's
topped
off
that.
Maybe
these
are
one
hundred
or
a
thousand
I
out,
and
so
one
of
the
major
challenges
I
want
to
call
it
failures.
B
But
one
of
the
major
challenges
in
configuring
Cassandra
cluster
production
designer
cluster
in
the
AWS
is
sizing.
So
you
know
if
I
was
running
my
own
hardware,
then
you
know
it's
I.
Can
it's
space
on
disk?
It
is
really
the
mitigating
factor
for
how
many
nodes
I'm
going
to
meet,
whereas
when
you're
in
a
cloud
you
also
have
to
either
you're
possibly
going
to
have
many
notes
that
have
very
little
data,
because
you
want
to
get
to
that
ratio
of
one
to
one
with
memory
and
data
within
your
within
every
note
of
your
Cassandra
cluster.
B
B
A
B
We
are
we're
moving
towards
data
snacks
enterprise,
especially
for
some
applications
that
require
what
is
coming
in
in
version
3.0,
you
know
enhance
security.
Netflix
is
a
public
company
or
subject
to
sox
compliance,
PCI
compliance,
so
you
know
we
need.
We
need
the
extra
security
layer
that
Dana
slacks
enterprise
through
point
O
is
moving
toward
and
plus
you
know,
cassandra
has
gotten
more
stable
now,
a
lot
more
within
the
past
year,
so
we'll
need
will
feel
the
need
to
patch
it
a
passionate
community
version
regularly.
B
So
now
we
can,
you
know,
upgrades
a
slower
and
kind
of
off
offset
that
responsibility
of
packaging
Cassandra
to
datastax
and
and
really
take
and
really
just
get
a
package
product
from
a
vendor
as
opposed
to
having
to
do
it
ourselves.
As
for
the
Hadoop
or
solar
integration,
solar
is
very
interesting
to
us.
B
You
know
if
anyone's
play
with
Cassandra
knows
you
know
querying
on
secondary
indexes,
although
a
cql
is
easier,
still
still
has
a
performance
impact
and
we're
hoping
we're
starting
to
dabble
solar
and
we're
hoping
that
the
solar
integration
will
now
allow
teams
that
want
to
do
more.
Ad
hoc
worrying
on
the
production
data
rely
easier.
B
B
A
Covered
it
pretty
well,
in
you
know
the
Apache
Cassandra,
you
know
open
source
version.
It
will
continue
to
add
security
measures,
but
in
in
date,
stacks
enterprise
30.
That
will
be
available
on
on
the
spot.
Now
februari
25th
we're
launching
we're
in
the
early
access
program
right
now.
It
has
an
advanced
security
measures,
that's
a
big,
so
water
family.
So
if
you
are
interested
in
that,
please
ping
me
and
I
can
connect
you
with
our
with
our
vp
of
products.
A
Gm,
masks
and
I'm
very
familiar
with
this
one
Jim
vp
of
engineering
is
dig,
would
let
go.
The
scapegoat
was
Cassandra
which
caused
uptime
issues
on
the
last
big
relief
pre-start.
I'm
going
to
point
this
one
right
at
you,
d
focus
on
droid
in
a
better
place
for
enterprise
workloads
where
stability
and
uptime
is
keys.
B
Absolutely
so
the
cater
with
you
which
I
wrote
their
gym
is
years
ago,
so
cassandra
has
evolved
tremendously,
a
you
know,
not
just
within
the
last
couple
of
years,
but
even
within
the
last
couple
of
months
and
last
six
to
eight
months,
Netflix
stores,
all
its
data
in
Cassandra
in
the
cloud
I'm
emphasizing
all,
except
for
financial
stuff
for
sox
compliance.
Everything
metadata
about
movies
are
customer
information.
Are
our.
B
A
B
You
know,
I,
don't
know
how
much
more
I
can
convince
people
it's
working
and-
and
if
you
look
at
some
outages,
some
amazon
outages
we've
had
just
you
know
within
the
past
12
months,
none
of
them
were
about
cassandra,
cassandra
hummed
along
you
know
perfectly
happily,
if,
if
you
know
anything
about
the
netflix
and
application
architecture
and
be
strides,
we
go
to
to
keep
things
up.
Please
do
read
on
attack
blog
about
chaos,
monkey
and
chaos
gorilla.
B
A
B
A
A
B
B
That
part
of
why
I've
got
no
DBAs
is
also
is
also
tightly
linked
to
Netflix
is
a
development
culture
of
freedom
and
responsibility.
We
try
and
not
have
operations
and
DBA
personnel.
We
have
to
DevOps.
We
give
a
lot
of
responsibility
to
our
developers
as
well
and
are
some
of
them.
I
would
call
TV
engineers.
B
A
B
A
Okay,
crees
those
Francisco-
you
talked
about
chaos.
Monkey
is
the
tool
that
those
chaos
monkey
open-source
can
someone
else,
pick
up
what
you're
doing
and
sort
of
replicate
those
same
failure
tests.
I'm!
Sorry,
do
you
repeat
the
question
your
chaos
monkey
casts.
If,
if
that
open
source,
can
someone
pick
that
up
and
serve
replicate
what
you're
doing
absolutely.
B
So
if
you
go
to
netflix.com
netflix
target
hub,
calm,
you'll
see
something
called
simian
army
and
chaos.
Monkey
incidence
is
part
of
that
and
it's
not
just
for
Cassandra
it's
for
killing
nodes
in
all
levels
of
your
of
your
physical
application
deployment.
So
yes,
it's
open
source
and
if
you're
in
the
bay.
A
Awesome
Thank
You
Sanjay
asks
using
cassandra
is
attached.
Are
there
any
use
cases
on
this
netflix
and
before
you
answer
I
just
like
to,
we
actually
have
a
new
success
story
on
datastax
top
arm,
which
is
Adobe's
marketing
cloud
and
they
use
Cassandra
as
that,
what's
called
that
edge
cash.
So
there's
some
information
there
christos
over
to
you
yeah.
B
You
could
definitely-
and
there
are
some
teams
that
are
using
it.
That
way
here
at
Netflix,
but
remember
the
whole
point
of
a
cache
is
is
to
increase,
read,
read
responses,
so
you
really
want
to
make
sure
you're
not
that
you're
getting
as
close
as
possible
or
or
at
that
one
to
one
ratio
of
theta
and
memory.
The
initial
as.
B
In
it
reads
from
disk,
and
then
it's
pains
in
system
memory,
you'll
get
great
response
times
and
and
using
Cassandra
as
as
a
cash
also
adds.
The
added
benefit
of
you
know
the
cab,
the
values
in
the
cash
are
actually
persistent
to
disk
in
case
of
failure,
as
opposed
to
like
EV
cash,
sorry
meant
catch,
which
you
know
is.
If
you
lose
the
cash
box,
it's
all
a
memory
movie.
You
lost
all
your
data.
A
Okay,
I
think
I'm
just
doing
a
quick
scan
and
we're
just
five
minutes
from
the
top
of
the
hour
yeah.
We
got
through
a
lot
of
questions
Christos.
Thank
you
very
much.
Indeed,
a
couple
of
people
were
asking
for
your
Twitter
handle
and
contact
information,
so
I
have
put
the
about
the
presenter
slide
back
up
there
at
chris
cole
and
follow
him
on
twitter.
I
am
at
sea
haske
chaska
on
twitter
and
you
know
we
try
our
best
to
be
responsive
to
the
community
and
get
questions
answered.
A
So
thanks
for
everyone
very
much
for
joining
us
today.
We
really
appreciate
it.
We
will
see
you
if
you
have
your
calendars
marked
for
valentine's
day,
we'll
do
a
little
bit
of
a
deep
dive
into
one
of
the
new
features
around
Cassandra
12,
which
is
called
be
nodes.
So
the
title
is
B
is
not
the
Valentine's
Day,
it's
avi
notes,
Patrick
McFadden
great
speaker
will
be
presenting
on
that.
We
hope
we
see
that.
Thank
you
very
much.