<s>
Chaos	B-Application
engineering	I-Application
is	O
the	O
discipline	O
of	O
experimenting	O
on	O
a	O
system	O
in	O
order	O
to	O
build	O
confidence	O
in	O
the	O
system	O
's	O
capability	O
to	O
withstand	O
turbulent	O
conditions	O
in	O
production	O
.	O
</s>
<s>
In	O
software	O
development	O
,	O
a	O
given	O
software	O
system	O
's	O
ability	O
to	O
tolerate	B-General_Concept
failures	I-General_Concept
while	O
still	O
ensuring	O
adequate	O
quality	O
of	O
service	O
—	O
often	O
generalized	O
as	O
resiliency	O
—	O
is	O
typically	O
specified	O
as	O
a	O
requirement	O
.	O
</s>
<s>
Chaos	B-Application
engineering	I-Application
is	O
a	O
technique	O
to	O
meet	O
the	O
resilience	B-General_Concept
requirement	O
.	O
</s>
<s>
Chaos	B-Application
engineering	I-Application
can	O
be	O
used	O
to	O
achieve	O
resilience	B-General_Concept
against	O
infrastructure	O
failures	O
,	O
network	O
failures	O
,	O
and	O
application	O
failures	O
.	O
</s>
<s>
Operational	O
readiness	O
can	O
be	O
evaluated	O
using	O
chaos	B-Application
engineering	I-Application
simulations	O
supported	O
by	O
Kubernetes	B-General_Concept
infrastructure	O
in	O
big	B-Application
data	I-Application
.	O
</s>
<s>
Gautam	O
Siwach	O
et	O
al	O
,	O
performed	O
evaluation	O
of	O
inducing	O
chaos	O
to	O
a	O
Kubernetes	B-General_Concept
environment	O
which	O
terminates	O
random	O
pods	O
with	O
data	O
from	O
edge	O
devices	O
in	O
data	O
centers	O
while	O
processing	O
analytics	O
on	O
big	B-Application
data	I-Application
network	O
and	O
infer	O
the	O
recovery	O
time	O
of	O
pods	O
to	O
calculate	O
an	O
estimated	O
response	O
time	O
.	O
</s>
<s>
While	O
working	O
to	O
improve	O
website	O
reliability	O
at	O
Amazon	B-Application
,	O
Jesse	O
Robbins	O
created	O
"	O
GameDay	B-Application
"	O
,	O
an	O
initiative	O
that	O
increases	O
reliability	O
by	O
purposefully	O
creating	O
major	O
failures	O
on	O
a	O
regular	O
basis	O
.	O
</s>
<s>
Robbins	O
has	O
said	O
GameDay	B-Application
was	O
inspired	O
by	O
firefighter	O
training	O
and	O
research	O
in	O
other	O
fields	O
lessons	O
in	O
complex	O
systems	O
,	O
reliability	O
engineering	O
.	O
</s>
<s>
While	O
at	O
Google	B-Application
,	O
Kripa	O
Krishnan	O
created	O
a	O
similar	O
program	O
to	O
Amazon	B-Application
's	I-Application
GameDay	I-Application
called	O
"	O
DiRT	O
"	O
.	O
</s>
<s>
The	O
intent	O
was	O
to	O
move	O
from	O
a	O
development	O
model	O
that	O
assumed	O
no	O
breakdowns	O
to	O
a	O
model	O
where	O
breakdowns	O
were	O
considered	O
to	O
be	O
inevitable	O
,	O
driving	O
developers	O
to	O
consider	O
built-in	O
resilience	B-General_Concept
to	O
be	O
an	O
obligation	O
rather	O
than	O
an	O
option	O
:	O
</s>
<s>
Instead	O
,	O
we	O
discovered	O
that	O
we	O
could	O
align	O
our	O
teams	O
around	O
the	O
notion	O
of	O
infrastructure	O
resilience	B-General_Concept
by	O
isolating	O
the	O
problems	O
created	O
by	O
server	O
neutralization	O
and	O
pushing	O
them	O
to	O
the	O
extreme	O
.	O
</s>
<s>
We	O
have	O
created	O
Chaos	B-Application
Monkey	I-Application
,	O
a	O
program	O
that	O
randomly	O
chooses	O
a	O
server	O
and	O
disables	O
it	O
during	O
its	O
usual	O
hours	O
of	O
activity	O
.	O
</s>
<s>
Knowing	O
that	O
this	O
would	O
happen	O
frequently	O
has	O
created	O
a	O
strong	O
alignment	O
among	O
engineers	O
to	O
build	O
redundancy	B-Application
and	O
process	O
automation	O
to	O
survive	O
such	O
incidents	O
,	O
without	O
impacting	O
the	O
millions	O
of	O
Netflix	O
users	O
.	O
</s>
<s>
Chaos	B-Application
Monkey	I-Application
is	O
one	O
of	O
our	O
most	O
effective	O
tools	O
to	O
improve	O
the	O
quality	O
of	O
our	O
services.	O
"	O
</s>
<s>
The	O
concept	O
of	O
chaos	B-Application
engineering	I-Application
is	O
close	O
to	O
the	O
one	O
of	O
Phoenix	O
Servers	O
,	O
first	O
introduced	O
by	O
Martin	O
Fowler	O
in	O
2012	O
.	O
</s>
<s>
Chaos	B-Application
Monkey	I-Application
is	O
a	O
tool	O
invented	O
in	O
2011	O
by	O
Netflix	O
to	O
test	O
the	O
resilience	B-General_Concept
of	O
its	O
IT	O
infrastructure	O
.	O
</s>
<s>
Chaos	B-Application
Monkey	I-Application
is	O
now	O
part	O
of	O
a	O
larger	O
suite	O
of	O
tools	O
called	O
the	O
Simian	B-Application
Army	I-Application
designed	O
to	O
simulate	O
and	O
test	O
responses	O
to	O
various	O
system	O
failures	O
and	O
edge	O
cases	O
.	O
</s>
<s>
The	O
code	O
behind	O
Chaos	B-Application
Monkey	I-Application
was	O
released	O
by	O
Netflix	O
in	O
2012	O
under	O
an	O
Apache	O
2.0	O
license	O
.	O
</s>
<s>
The	O
name	O
"	O
Chaos	B-Application
Monkey	I-Application
"	O
is	O
explained	O
in	O
the	O
book	O
Chaos	B-Application
Monkeys	I-Application
by	O
Antonio	O
Garcia	O
Martinez	O
:	O
</s>
<s>
The	O
Simian	B-Application
Army	I-Application
is	O
a	O
suite	O
of	O
tools	O
developed	O
by	O
Netflix	O
to	O
test	O
the	O
reliability	O
,	O
security	O
,	O
or	O
resiliency	O
of	O
its	O
Amazon	B-Application
Web	I-Application
Services	I-Application
infrastructure	O
and	O
includes	O
the	O
following	O
tools	O
:	O
</s>
<s>
At	O
the	O
very	O
top	O
of	O
the	O
Simian	B-Application
Army	I-Application
hierarchy	O
,	O
Chaos	O
Kong	O
drops	O
a	O
full	O
AWS	B-Application
"	O
Region	O
"	O
.	O
</s>
<s>
Chaos	O
Gorilla	O
drops	O
a	O
full	O
Amazon	B-Application
"	O
Availability	O
Zone	O
"	O
(	O
one	O
or	O
more	O
entire	O
data	O
centers	O
serving	O
a	O
geographical	O
region	O
)	O
.	O
</s>
<s>
Proofdock	O
is	O
a	O
chaos	B-Application
engineering	I-Application
platform	O
that	O
focuses	O
on	O
and	O
leverages	O
the	O
Microsoft	B-Application
Azure	I-Application
platform	O
and	O
the	O
Azure	O
DevOps	O
services	O
.	O
</s>
<s>
To	O
prepare	O
for	O
the	O
loss	O
of	O
a	O
datacenter	O
,	O
Facebook	B-Application
regularly	O
tests	O
the	O
resistance	O
of	O
its	O
infrastructures	O
to	O
extreme	O
events	O
.	O
</s>
