►
Description
Speakers: Matt Oswalt https://twitter.com/mierdin + David Gee https://twitter.com/vtep42 (Juniper)
Slides: https://gitpitch.com/nre-learning/inog10-nre/master?grs=github&t=black
Suggest a talk https://inog.net/talk
A
I
think
we're
good
computers
at
heart,
all
right
everybody,
my
name
is
Matt.
I
was
well
here
with
my
good
friend,
David
G
colleague
of
a
friend
before
that
we've
we've
done
some
shall
we
say
very
long,
hackathons,
together
on
trains
to
Tokyo
or
whatever
that
trip
was
so
we've
been
passionate
about
network
foundation
for
a
while,
and
we
we
both.
We
both
done
a
lot
of
work
in
this
space.
Coming
from,
like
the
network
operator
perspective
and
then
moving
into
software
development
and
then
trying
to
sort
of
blend,
the
two
skill
sets.
B
C
A
Looking
some
of
the
trends
in
the
industry
and
sort
of
probably
not
I'm,
not
just
within
networking
and
started
thinking
about
how
else
we
can
talk
about
this
conversation,
it
makes
it
a
little
more
inclusive,
rather
than
sort
of
constraining
the
automation
umbrella
to
a
small
subset
of
tasks
and
individuals.
I
think
there's,
there's
some
value
in
in
opening
this
discussion
back
up,
and
so
what
we're
gonna
be
talking
about
today
is
a
topic
called
a
network,
reliability
engineering.
A
A
At
site,
reliability,
engineering,
so
we'll
watch!
You
talk
about
the
relationship
there,
because
there's
modest
I
think
are
important
to
talk
about
before
I
get
started
in
that,
though,
like
I
said,
the
background
for
this
particular
topic
is
me
sitting
at
home
and
having
nothing
better
new
but
watch
Twitter
and
argument
people,
because
that's
one
that
I
started.
A
And
and
I
think
it's
important
to
talk
about
the
goal
versus
the
symptoms.
A
lot
of
what
I'm
talking
about
today
is
about
the
ultimate
goal
for
what
we
do,
especially
in
a
lot
of
automation,
context,
but
at
the
ultimate
goal.
I
think
that
we're
going
to
talk
about
today,
we'll
make
a
case
for
this
is
better
reliability.
I
love!
A
This
is
the
one
that,
if
needs,
were
because
we
sort
of
do
this
to
ourselves.
We
assume
that
Facebook
and
Google
are
the
only
ones
that
have
a
scale.
Sorry
to
do,
automation,
I,
think,
that's
sort
of
put
just
gonna
put
us
in
a
bind
from
the
get-go,
because
it
makes
that
assumption
that
automation
is
about
speed
or
scale
and
I.
Think
if
you.
A
You're
doing
with
respect
to
building
automation,
workflows,
it's
not
really
what
it's
about
might
be
identified
product.
Certainly,
if
you
get
to
that
levels,
you
kind
of
have
to
automate.
That's
definitely
true,
but
there's
a
big
nervous
being
cause
an
effector,
so
not
true,
just
to
be
mystify
for.
A
A
In
a
form
that
the
machines
can
understand
and
I
think
if
you
break
it
down
into
that
level
of
that
level
of
just
simplicity,
you
start
to
understand,
like
you,
don't
need
to
boil
the
ocean,
nothing
needs
to
be
done.
You
know
overnight,
and
it's
not
really
something
that
is
difficult
to
understand.
It's.
B
A
What
are
the
principles,
the
native
reliability,
engineering,
unique
and
valuable?
What
is
it
that
we're
actually
trying
to
accomplish
here
in
kakuzu
role
or
hopefully,
by
the
end
of
this
talk?
You'll
agree
this
and
I
I
can't
claim
to
mention
of
it
by
the
way
as
as
we'll
discuss,
there's
actually
companies,
probably
in
this
room
that
actually
are
hiring
for
this
particular
job
title.
So
it's
attacking
and
because.
A
B
A
Are
useful
concerns
but
I
want
to
sort
of
break
that
down
and
make
it
a
little
more
obvious.
If
you
go
through
the
sorry
book
that
Google
published
by
the
way,
you
can
buy
a
horse
on
Amazon
or
whatever,
but
you
can
also
read
it
for
free
online,
and
so
you
can
go
through
that
and
in
addition
to
the
book
there
have
been
a
lot
of
videos
published
by
folks.
A
C
A
A
A
Implements
DevOps
right
sra
is
a
specific
implementation
of
those
principles.
It's
not
a
separate
thing,
so,
if
you
can
fix,
if
you
think
of
DevOps
as
the
set
of
principles,
sra
is
a
specific
individual
that
follows
those
principles
and
similarly,
for
us
and
nra,
is
no
different.
It's
just
a
different
coat
of
paint
because
the
reality
of
that
network
engineers
job
is
we
all
know,
is
nothing
like.
You
know
what
the
system
in
this
job
there's
just
a
different
set
of
tools
and
end
processes.
Should
there
be
a
little
more
alignment.
A
Thing
it's
a
different
implementation
of
the
same
core
principles,
and
so
a
lot
of
the
principles
that
I'm
going
to
talk
to
you
about
today
are
influenced
exactly
by
that
original
model.
They're
not
really
separate,
you
know
ideals
or
around
what
it
is
that
defines
DevOps,
and
so
order
of
operations
is
important
when
we
talk
about
automation,
I,
think
back
to
that
first,
myth
where
I
talked
about
automation
is
about
doing
things
faster
hands
like
this
longer.
This
is
the
Nikes
masters
utterly
terrify
you
it
did
for
me,
yeah
I
mean
that's
horrifying.
A
C
A
B
A
For
instance,
copper
dollars
in
their
red
coat.
You
know
they
have
necessarily
like
you
know,
edit.
The
web
server
files
live.
They
go
through
sort
of
CSEE
process.
Is
that
CSV
process
slower
than
just
editing
the
final
spot?
Sure
it's
a
slower
rate
organizationally?
They
actually
move
much
quicker,
because
what
happen
is
when
you're
doing
things
in
this
sort
of
regimented
way
is
you're
not
moving
backwards.
There's
no
fire
fighting,
so
you
don't
have
to
usual
Springs
forward
and
then
run
back
and
settings
on
fire
me
for
fist.
A
A
Fantastic,
a
hum
I
have
another
customer
signing
for
easily
the
most
my
career,
and
this
is
like
zero.
So
the
point
here
is
that
is
that
is
it
in
a
week
week.
We
are
always
in
this
mode
if
one
see
more
functionality
and
radicals
or
not,
and
and
frankly,
you
know
whether
we're
talking
about
management
and
I'll
add
only
one
little
longer,
because
this
is
the
one.
Have
you
moving
like
this
Sdn
for
a
world
in
the
Navajo
Nation
world,
there's
one
that
isn't
on
this
find
it.
It
comes
up
often,
and
it's
just
infuriating.
A
B
A
A
That
possible
sort
of
an
AMA
networking
perspective
his
through
nays.
Similarly,
kubernetes
is
a
platform
for
deploying
applications,
so
you
don't
have
to
worry
about
things
like
scaling
and
policies.
My
password
is
built
into
the
framework.
It's
really
good
doing
what
it
does.
There's
some
sort
of
gap
I
follow
the
sorta
8020
rule
for
this
stuff.
My
boss
might
calls
it
at
the
automation
last
mile,
I
just
ended
up
with
the
term
composition.
A
A
A
Sort
of
loves
to
that's
have
your
head
spin.
You
know
I
need
to
see
member
services
in
this
offers.
Those
versions
of
that
developers
love
the
fact
that
this
office,
all
those
services,
little
micro
services,
that
they
write.
They
write
a
desert
city
business
line
and
then
use
an
event,
that's
getting
together
and,
of
course,
if
you
that
is
gonna
happen
at
least
forever.
A
So
a
obvious
knows
this,
there's
value
in
that,
so,
if
composition
super
important
and
both
as
the
glowing
eyes
wide,
open
and
airy,
knowing
that
you
won't
or
shouldn't
automate
everything,
the
other
thing
is
some
is
our
comes
sort
of
when
you
think
about
what
is
the
whole
point
of
all
this?
You
know
what
is
the
ultimate
desired
outcome
for
the
work
that
I'm
doing
and
I
haven't
to
use
this
example.
This
is
the
AWS.
There
is
health
dashboard,
you
simply
just
google
AWS
health
dashboard
and
what.
A
Me
get
to
see
oh
yeah,
the
check
mark
which
are
by
the
way
stored
in
Amazon
s3.
So
if
s3
is
down,
this
actually
happened,
it
was
hilarious,
home
and
so
yeah
it
was.
It
was
usual,
but
I
I
think
this
little
people
are
going
to
expect
for
all
infrastructure.
This
is
this:
like
I,
didn't
even
know
about
this
anymore.
People's
reading
abilities
services,
this
service
miss
Holly
internally,
but
in
contrast
from
an
engineering
perspective,
we
talked
about
that.
You
know
there's
very
little.
There's
compute
and
storage
and
networking
behind
all
of
these
things.
Right.
A
That's
up,
isn't
running
them.
Now,
it's
running
on
literal
hardware.
Yes,
the
abstraction
is
offered
to
customers.
Is
this
really
customer
and
outcome
driven
thing?
Don't
care
about
how
well
this
one
is
doing
how
old
everyone
is
good.
Yet
this
is
sort
of
the
respectively
provide
more
guys
our
customers
on
a
daily
basis,
without
really
thinking
about
and
I,
think
getting
that
sort
of
service
is
driven
mentality
like
how
hunger
people
actually
want
out
of
the
network,
and
they
don't
necessarily
want
switch
wanna,
be
working
that
sort
of
table
states.
A
They
want
something
higher
level
than
that.
The
third
of
things,
trade
off,
don't
go
from
Purcell
time,
and
you
really
already
look
if
you're
on
air
alleges.
This
is
actually
a
way
of
planning
for
failure
and
errors
of
going
in
because
trying
to
go
for
a
hundred
setup
times,
actually
prohibitive
and
again.
A
Trying
to
make
things
100
or
settle
time
we're
sacrificing
up
to
preserve
the
great
addition.
So
if
your
outcomes
with
you
know
what
it
is
that
your
organization
is
best
out
of
the
network,
you
can
actually
make
trails
I,
actually
don't
need
to
make
this
particular
system
super
reliable,
because
my
husband
is
the
best
spent
over
here.
Some.
A
Not
super
critical
but
I
try
to
automate
everything,
but
there's
that
one
workflow
that
either
breaks
all
the
time
or
makes
our
life
miserable
or
whatever,
and
you
focus
on
that,
and
only
only
automate
that
workflow,
at
least
for
starters
and
then
it
out
to
8020
rule
it's
important
to
go
in
my
eyes
wide
open
with
this
as
well.
You
know
I,
didn't
simple
what
you
can
like.
There's
no
there's
some!
You
know
automation
isn't
about
like
writing
everything
from
scratch.
A
A
A
A
C
Gonna
guess
I'm
gonna
say
the
five
minute
in
a
moment
right
so
I'll
speed
up
talking
I
always
am
I.
Go
about
20
min.
Quite
a
lot.
I
speak
at
this
kind
of
pace.
You
say
slow
down,
so
I'm
gonna
jump
in
time.
Hopefully,
when
you
see
this,
you
see
you
oughta
make
the
coffee.
What
the
hell
are
you
on
about?
You,
crazy
fool
and
the
whole
point
is
it's
a
mental
frame,
the
amount
of
times
I
go
into
a
customer
meeting
or
I
go
to
a
meter
and
the
conversation
is
solely
focused
on.
C
Let's
talk
me,
the
network
as
far
as
much
meaning
is
this
thing
generally
handle
quick
agreement
and
I'll
read
out,
obviously
for
the
camera.
Here's
an
agreement
you
reckon
this
is
a
good
mental
I'm
in
cancer.
You've
been
really
cruel.
If
none
of
you
agree
with
this
I'm
gonna
burn
this
and
thrown
in
a
bit
fair
enough.
All
right,
we'll
come
back
to
that,
we'll
go
on
Google,
it's
better.
We
take
you
laughter,
so
automation
I,
try
and
describe
this
in
a
slightly
different
way.
So
we
talk
about
Catherine,
pears,
great.
C
C
Pet,
the
networks
are
pet,
their
processes
of
pets
and
the
mental
visualization,
without
get
my
words
out
visualization.
For
this
imagine
going
into
a
Cathy
art,
you've
got
loads
of
cows
and
there's
a
dog
and
a
koala
bear
and
squirrel,
of
course,
milking
that's
great,
and
what
we
try
and
do
with
automation
really
is
provide
you
a
cow
with
battery
slots.
You
can
use
a
cup
of
technology
and
the
last
20
percent
is
a
customization
bit
and
the
way
we
achieve
this
is
through
the
through
the
UNIX
havi
1978
circa
era
built
something
good.
C
We
use
it
having
lots
of
outputs,
put
it
in
machine
which
takes
me
down
to
the
devil.
If
that's
composition,
the
last
flow
is
data
flows,
they
have
to
come
from
somewhere.
Data
is
a
currency
for
this
kind
of
thing.
So
right
now
we
see
data
driven
systems,
test
driven
design
and
all
these
kind
of
things.
Without
the
data,
without
good
lots
to
compose
and
without
a
high
level.
Thinking
I
want
to
say,
realism.
We
need
to
get
a
lot
more
real
about
this.
C
The
reason
they
wonder
product
there
will
probably
never
be
under
product
with
things
like
AWS.
We've
got
a
myriad
of
services
that
make
perfect
sense.
Some
of
the
network,
we've
kind
of
gone,
hey,
we've
got
a
device,
it's
a
first
solar
system
with
30,000
features
and
the
total
combination
of
those
features.
If
you
can
do
it,
is
about
five
hundred
and
million
I
think
it
might.
My
mass
isn't
too
shabby
on
that
one.
We.
C
C
We've
got
a
demo
here,
I'm
really
on
3G
I'm,
not
sure
it's
going
to
work.
I'm
gonna
be
very,
very
hopeful.
I'm
gonna!
Try
it
anyway.
Before
we
do.
This
I
want
to
cover
off
IAC,
so
people
in
the
room
dealing
with
the
network
as
if
it's
a
chunk
of
code,
config
files
Wow
for
the
camera,
these,
like
three
or
four
five.
C
Is
just
like
the
end
of
the
day,
you
need
a
bit
more
time.
Okay,
automated
testing.
Are
you
testing
for
syntax
and
semantics
on
the
things
you
generate?
Are
you
pushing
it
and
then
testing
in
the
network
level
handle
for
the
first
one
testing
it
config
files?
Okay,
and
then
you
know
what
put
my
blindfold
on.
C
Work
beep
right
who's
doing
this,
we've
got
two
fingers
at
Eli's.
Okay,
we've
got
a
good
mix,
I'm
really
pleased
about
this.
So
thank
you
for
being
honest.
If
we're
not
going
to
solve
the
world
challenges
by
this,
always
going
to
try
and
do
is
use
the
concepts
of
IEC
and
trying
somehow
inject
testing
writing
to
the
core
of
those
daily
process.
It's
making
part
of
what
you
do.
C
We
all
love
means
all
right
code.
What
does
this
meaning
infrastructure
as
code?
Those
really
hug
me
in
a
whole
opposite.
We
have
a
system
we
humans
have
built.
We
have
a
config
system
for
the
system
that
we
built
most
of
the
time
of
the
automation
we
take
hands
of
all.
We
take
some
templates.
We
take
some
variables
and
we
regenerate
a
human
version
of
the
state
machine
needs
to
be.
This
is
where
we
are
today,
especially.
The
network
has
gone
through
different
phases
like
open
flow.
C
C
Maybe
if
we
can
generate
these
things,
if
we
can
test
out
the
network
level
and
the
search
at
the
network
level,
you
know
what
we're
going
to
be
a
lot
more
alive
or
going
forward,
and
it's
going
to
go
really
really
quickly
through
these
tests
can't
fail
if
you're
doing
well,
then
I'm
I'm,
like
software
and
I
I,
am
terrible
to
writing
tests
I
think
I,
don't
really
like
them.
I
do
the
things
that
I
think
you're
gonna
fail.
C
I
mean
the
network
weirdly
with
the
automation,
I
test,
everything
every
single
time,
I
generate
something
if
I
use
it
very
well.
If
I
build
some
sort
of
config
files,
if
I
make
an
API
call,
I
don't
want
to
just
verify
the
informations
gone
down
to
there.
Controlling
I
want
to
make
sure
actually
the
network
service
the
thing
that
we're
trying
to
build
this
up
and
running
how
many
times
you
go.
Hey
we've
got
a
switch.
We've
got
a
winner.
We
need
to
test
everything.
C
You
know
if
you're
going
to
drop
one
service
on
that
switch,
make
sure
the
service
is
running.
That's
where
your
focal
point
is
going
to
be.
You
know
ever
gonna
have
a
system
which
takes
every
single
feature
and
every
single
combination
so
just
be
a
little
bit
more
focused
on
this.
The
solar
map
are
all
over
Twitter,
the
last
few
days.
Of
course,
an
ethnic
argument
I
think
it's
safe
to
say:
we've
got
to
backtrack
a
couple
of
times
and
I
think
that's
spread.
This
is
a
validation
of
the
validation
right
here.
C
We
share
these
thoughts.
People
come
back,
Amanda
I
know
how
about
some
exchanges
over
the
past
as
well,
and
this
basically
says
hey,
you
know
what
I'd
rather
validate
over
the
ability
to
push
convict
as
smoothly.
We
have
to
get
better
at
this
kind
of
stuff.
This
is
a
machine
control
theory.
This
is
open-loop
versus
closely
right
compliance
validation.
We
all
work
in
these.
In
these
areas,
I
mean
I've
worked
in.
C
You
know
kind
of
defense,
the
type
of
summaries
banking,
snorri's
insurance
scenarios
and
the
amount
of
times
an
engineer
goes
I'm
just
going
to
go.
Make
this
really
quick
change?
She'll
be
ok,
and
this
gives
out
ticket
closed
good
job
now
out
of
compliance,
and
if
anybody
ever
came
in
and
did
an
audit
would
be
shot
so
compliance
validation,
we
can
obviously
between
everything
is
code.
We
can
test
for
things
that
we
know
what
should
be
in
and
what
shouldn't
be.
We
think
okay.
C
Well,
what
if
I'm
the
kind
of
person
I'm
just
going
to
ignore
then
I
push
it
down
any
way.
We
can
take
it
back
to
a
whole
other
level.
We
can
go
down
to
the
network
and
grab
that
data
and
retest
think
well,
okay,
what
about
I'll
be
messing
around
there?
Is
a
human
only
been
doing
this
all
automatically,
so
we
stop
talking
about
event-driven
infrastructure,
and
this
is
where
I
think
I'm
going
to
run
out
of
time.
Unfortunately,
if
I'm
gonna.
C
C
Use
computers
in
anger
right
we're
not
really
good
at
this
kind
of
thing,
and
so
we
take
signals
a
signal,
consumer
system,
it's
an
emitter,
so
something's
happened,
doesn't
necessarily
mean
it's
an
event
and
the
event
is
an
occurrence
of
something
we're
interested
in
which
one
more
signals
tell
us
about
any
ideas
that
we
use
rules
to
identify
that
combination
of
signals
of
which
become
events
which
turns
into
business
logic.
The
business
logic
triggers
our
work
for
love,
a
workflow
being
an
engineering
flowchart.
The
star
point
action.
C
This
decision
points
we
tie
it
together
into
actions
and
outputs
actions
and
alcohols
can
happen
at
any
point
and
to
do
it
right.
We
feel
your
background.
We
close
the
loop
is
no
surprise
right
now,
I'm
going
to
speak
with
an
evil
vendor
Hammond
right
now,
I'm
very
sorry:
there's
lots
of
telemetry
up
going
into
network
operating
systems,
lots
of
GPP
type
encoding
for
telemetry.
We
need
data.
If
they'd
is
a
currency,
we
need
to
be
able
to
gather
that
data
metadata
and
trend
it.
So
EDI
is
nothing
more
than
we
have
imports.
C
We
have
rules,
we
have
outputs.
We
can
do
really
cool
things
with
these
things,
so
we
can
tree
off
stuff.
We
have
an
outage,
we
can
say
hey.
This
is
an
event.
Problems
happened,
go
off
to
the
network
gather
out
a
log.
So
when
the
human
comes
in
at
9:00
a.m.
great
all
the
informations
there
we
save
loads
of
time,
especially
with
automation.
People
are
always
hesitant
to
let
a
machine
loose
on
the
network
with
Auto
remediation.
C
C
C
C
C
C
Sting
and
interest
in
this
occurred
sort
of
thing
fails.
I
want
to
do
some
triage
I
want
to
create
a
ticket
for
the
customer
and
then
also
want
to
send
in
chat
ops
message
off
to
the
Encore
team.
It's
all
very,
very
straightforward.
Now
my
laptop
is
is
absolutely
dying
on
its
feet
and
I.
Don't
the
oscillations
here
which
happened
earlier
on
what
I'm
hoping
to
see
is
when
the
alarm
happens
in
a
second?
Is
this
going
to
work?
Hey
it
worked.
How
about
that?
So
actually
what
we
have
now
these?
C
If
you
can
see
this
the
work
for
they
did
a
database
tip,
it
did
the
alert,
the
on
shift
team
and
it
also
created
a
ticket
which
means,
if
I
go
down
to
slack.
We
can
see
the
slack
message
from
my
friendly
neighborhood
robot
called
Johnny
Five.
Any
80s
kids
in
the
room
will
appreciate
this.
Hey
we've
got
this
circuit
called
blah
blah
blah
and
ask
for
mr.
IP
engineer.
Obviously
that
being
me
just
to
give
you
an
idea
there
we
go.
C
And
what
we
also
have
where's
that
gone
sorry
I've
got
so
many
screens
open
I
should
also
have
a
ticket
on
Zendesk
and
there
it
is
so
the
tickets
been
created.
The
customer
sees
of
that
now.
I
appreciate,
you
know,
I
didn't
really
know
where
I
was
given
that
at
the
time
I
thought
it
was
kind
of
cool,
I'm
autumn
Utley.
So
what
happens?
Signals
in
database
they're,
two
types
of
information
go
around
and
then
the
idea
is
that
somebody
on
call
can
pick
it
up,
knowing
that
the
first
bit
has
been
handled.
C
This
is
really
really
simple.
This
is
a
very,
very
basic
kind
of
triage
exercise.
We
can
do
a
lot
more
complex
things
now.
It
doesn't
necessarily
matter
what
songs
you
pick
go
into.
Unix
philosophy,
pick
the
small
ones
and
do
the
job
well
consider
what
you're
trying
to
do
don't
bottle
the
ocean.
Think
of
it
and.