►
From YouTube: Sponsored Keynote: The Future of Observability in K8s - Nik Jain & Steve Ng, New Relic
Description
Join us for Kubernetes Forums Bengaluru and Delhi - learn more at kubecon.io
Don't miss KubeCon + CloudNativeCon 2020 events in Amsterdam March 30 - April 2, Shanghai July 28-30 and Boston November 17-20! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects
Sponsored Keynote: The Future of Observability in K8s - Nik Jain, Observability Advocate & Steve Ng, Solutions Consultant, New Relic
New Relic has a bi-modal strategy to deliver what it takes to bring meaning to the day-in-the-life of our platform practitioners, we have always been an open-source fan and big supporter for Kubecon. What that means from a technology standpoint is that we are vested. In this short keynote presentation, New Relic Technologists - Steve and Nik will showcase two core use-cases.
1. Use of OpenMetrics and OpenTracing technologies to enable telescopic and microscopic insights into your K8s super-clusters (multi-cluster setup)
2. Showcase how our open-source nerdpack - Compute optimise delivers cost-containment for K8s services.
In a nutshell, our aim to enable the audience on double bottom line attainment i.e. revenue goals and innovation atop cloud-native technologies like Kubernetes.
A
Hello,
Sydney,
hey
everyone,
so,
in
terms
of
you
know,
our
introduction
Steve
is
known
as
Steve
ops
at
New,
Relic
and
I'm
known
as
monitor
Jen,
and
that
that's
not
to
talk
ourselves
up.
It's
actually
the
relentless
work,
the
passion
and
you
know
the
focus
on
client
problems
that
we
pour
into
everything
we
do
on
a
daily
basis
and
then,
if
you
think
about
what's
happening
with
the
kubernetes
world,
it's
it's
kind
of
similar.
A
If,
if
you
forget
about
the
technology
for
a
second
and
you
think
about
what
the
clients
are
doing
with
standard
kubernetes
stack
they're
trying
to
achieve
three
goals
and
I
want
to
talk
about
that,
because
Steve
is
going
to
take
you
into
he's
gonna
catch
you
up
with
what
has
changed
in
terms
of
tracing
and
monitoring.
I'm
gonna
get
then
go
into
a
little
bit
of
cloud,
optimization
for
the
kubernetes
realm
and
thirdly,
I
want
to
talk
about.
You
know
purely
from
a
maturity
model
perspective.
A
What
can
you
do
best
when
you
are
handling
a
kubernetes
service,
but
before
I
do
that
I
want
to
switch
back
to
those
three
goals,
so
the
three
goals
are.
Firstly,
find
a
market
fit
for
their
products
and
services.
The
second
goal
is
to
ensure
they
find
profitability
when
they're
delivering
those
transactions,
and
the
third
thing
they
want
to
do
is
they
want
to
make
sure
those
transactions
are
served
elegantly
so
that
they
don't
see
a
leaky
bucket
effect.
So
the
talk
is
geared
towards
those
things
and
I
want
to
give
Steve
the
spotlight.
B
One
thing
that
Nick
and
I
do
a
lot
is
we
actually
spend
a
lot
of
time
in
customers,
environment,
really,
understanding
the
tech
stack,
but
also
seeing
what's
coming
in
the
future.
So
when
one
of
our
key
topics
is,
we
were
working
so
much
with
customers,
but
we
also
want
to
give
you
a
glimpse.
What's
coming,
what
we
see
in
the
next
year,
but
also
what's
going
on
in
the
open
source
wall
now,
if
you've
been
to
any
conference,
any
events
any
meetups
and
in
any
presentation
you
probably
heard
a
word
about
observer.
B
It's
a
really
big
word
right
now,
but
if
you
really
scheme,
you
know
really
scrape
away
the
marketing
terms
the
fluff
around
it.
But
what
it
really
means
is
we
want
to
understand
the
insert
the
internals
of
a
system
using
only
external
outputs.
In
fact,
what
it
really
really
means
is
whoever
builds
the
app
who
is
probably
have
better
knowledge
to
actually
tell
you
what
they
need
to
monitor
in
the
system.
In
fact,
you've
probably
heard
about
the
term.
B
That's
given
rise
of
observability
driven
development
where,
in
the
past,
monitoring
it's
always
afterthought,
but
when
it's
done
correctly
in
the
moment
when
you
develop
app
dev,
ops,
sre
the
process
and
the
responsibility
of
looking
up
an
environment,
it's
much
easy.
In
fact,
observability
is
defined
in
modern
control
theory.
B
Now,
if
you
ever
read
about
or
follow
this
particular
trend,
especially
open
standards
is
going
on
in
the
war
was
a
real
observability.
There
are
four
key
standards
that
maybe
you
heard
about
it
first
is
open.
Metrics,
open
metrics
is
an
open.
Standard
of
trim
are
transmitting
metrics
form
an
environment
into
a
platform.
So
karlie
is
part
of
CN
CF
Sandbach
projects.
In
fact,
I.
Don't
know
if
you
know,
if
you
use
Prometheus
right
now,
the
data
format
is
seen
from
Prometheus,
it's
actually
using
open
metrics.
B
So
if
you're
using
that
right
now,
you're
really
seeing
that
the
second
part
we
got
this
a
lot
of
question
during
our
whole
day
in
the
booth,
hey,
hey
guys,
you
know:
what
do
you
guys
think
about
open
tracing
I?
Do
anything
about
open
tracing,
so
open
tracing?
It's
a
vendor-neutral
tracing
API
framework
that
or
that
actually
came
from
the
principal
album,
a
concept
that
was
written
by
Google
in
the
Google
technical
white
paper
called
dapper
by
a
person
called
Ben's
Siegelman.
So
back,
then
there
are
two
open
source
tracing
FET
format
that
come
up.
B
What
is
called
Zipkin
and
the
other
ones
called
Jaeger.
So
as
time
goes
by
they're
now
considered
as
part
of
open
tracing
format
is
part
of
the
CNC
of
incubating
project.
Then
the
third
open
standard
you've
probably
heard
about
it-
is
called
open
census.
Now,
Google
did
a
lot
of
things
in
open
source.
World
contribute
a
lot
of
things,
as
one
of
them
is
also
essentially
called
open
census,
which
is
a
set
of
libraries
that
Google
used
to
collect,
metrics
and
distributed
traces.
B
But
there's
one
problem
right
now:
I,
don't
know
you
see
it
got
too
many
standards
now
at
the
middle
of
this
year,
moving
towards
next
year,
I'm
gladly
to
say,
and
also
as
a
industry,
we're
moving
towards
open,
telemetry,
so
open
telemetry.
It's
a
merger
between
Google's,
open
census
and
open
a
CNC
F
open
tracing,
not
only
here
in
New
Relic
was
sharing
our
contribution
because
we've
been
doing
tracing
for
a
really
long
time,
but
also
we're
making
our
platform
our
agents.
B
We
want
to
give
you
the
ability
to
agentless
and
agent-based
tracing
we're
gonna
support,
big
time
coming
with
the
coming
release,
our
agents
approximately
in
the
next
few
months.
Now
we
talked
about
about
concept,
tracing
tracing
tracing
tracing,
but
what
really
means?
Let
me
give
you
a
really
more
relatable
example.
So
some
of
you
here
maybe
came
came
from
Melbourne.
Some
of
you
here
came
from
different
countries.
I
say
tracing
is
like
just
like
Earth's
as
a
passenger,
maybe
I
go
to
Melbourne
to
Sydney
Sydney
to
San
Fran.
B
Maybe
I
want
to
go
Vegas,
I
transit.
Another
flight
me
as
a
passenger
I
know
where
I'm
coming,
where
I'm
going
so
the
whole
concept
of
tracings
understanding
the
original
trace
order,
the
way
to
the
end
user
journey.
We
didn't
tracing.
Here's
an
example
where
open
tracing
and
proprietary
tracing
can
live
together
in
harmony.
B
This
is
an
example
where
we're
supporting
our
ppens
epyon
tracing
format
as
an
example
with
the
UI,
but
also
New
Relic
tracing
together
in
within
the
same
UI
I
want
a
question
that
we
get
asked
hey,
you
know,
can
open
tracing
a
New,
Relic
format
work
together,
and
the
answer
is
yes.
If
you
need
more
details,
feel
free
to
have
a
chat
with
us
later
on
underneath
the
hood.
What
do
we
do
right
here?
This
is
secret
sauce
that
we've
been
doing
for
so
many
years.
How
do
we
interrupt?
B
How
do
how
those
traces
interoperability
actually
works?
If
you
have
different
tracing
format,
usually
you
have
what
we
call
a
proxy
layer
to
translate
one
trace
format
to
another
one.
Currently,
people
actually
use
this
thing
called
a
d3
header.
Don't
you
do
that
trace
propagation?
Then
you
usually
have
two
type
of
ID
trace,
ID
and
parent
ID
to
correlate
all
the
trace
different
trace
format,
but
one
thing
that
I'm
also
happy
to
say
we're.
B
Also
doing
our
active
con
tribution
is
roughly
next
year
you
go
to
start
to
see
a
new
different
type
of
format
called
w3c
tres
contacts
propagation,
not
only
we're
going
to
support
it
big
time,
but
if
you're,
using
that
in
your
existing
trace
format,
we
can
support
it
too,
on
a
platform
on
the
backend
we're
already
working
on
it
yeah.
But
how
a
really
differentiate
ourselves
and
provide
value
is
hey.
You
definitely
can
use
trace
open-source
tracing.
B
If
you
want
to
use
neural
X
agent,
that's
also
fine,
but
the
one
we
we
truly
provide
value
is
we
provide
locks,
metrics
events
but,
most
importantly,
kubernetes
metadata
directly
into
the
trace
correlation.
Here's
an
example
of
a
trace
that
we
can
see
that
took
1.25
two
seconds
immediately.
You
can
see
the
correlation
with
the
in
kubernetes
itself
directly.
We
did
that
same
trace.
A
So
the
core
objective
was
to
showcase.
How
can
you
find
correlation
and
causality
into
a
single
view,
and
that's
why
we
decided
to
you
know
at
least
share
with
you
what
is
happening
in
the
tracing
world
and
how
we
are
supporting,
together
with
technologies
that
are
built
with
collective
intelligence
and
with
the
concept
of
using
open
source
and
cloud
native
incubator
projects.
So
we
build
this
maturity
model.
This
doesn't
necessarily
apply
just
to
New
Relic
or
a
particular
vendor.
This
is
more
from
an
education
standpoint.
A
This
is
how
you
want
to
design
your
kubernetes
maturity
model,
so
it
starts
with
various
dimensions.
Just
like
how
kubernetes
is
set
up,
you
got
control
plane.
You
got
the
data
plane
in
the
data
plane,
you
have
worker
nodes,
you
then
have
pods
and
within
the
pods,
you
might
have
containers
that
support
your
application
services.
So
what
we
do
here
is
follow
the
same
dimensional
model.
We
start
with
an
open
source
plugin,
which
is
KSM
and
prometheus,
based
that's
going
to
bring
in
all
the
infrastructure,
goodness
compute,
storage
and
disk.
A
Once
you
deploy
that
kubernetes
integration,
you
can
then
progress
into
watching
natively.
What's
happening
with
your
scale,
down
activities,
what's
happening
with
the
pod
events,
where
are
you
seeing
invalid
image
problems
or
even
image
pull
backs,
because
there
are
problems
with
those
container
services?
You
can
then
get
into
the
stuff
that
Steve
talked
about,
which
is
get
down
to
distributed
traces
level
using
an
APM
correlation
layer
which
is
essentially
deploying
an
agent
as
a
sidecar
mount
a
docker
volume
or
even
bake
it
into
your
container
image.
A
But
where
you
really
get
start
to
see,
the
best
part
is
when
you
bring
in
Prometheus
data
together
with
other
tools
that
you
might
have
in
your
kit.
So
what
we
do
is
we
integrate
the
Prometheus,
open,
metrics
data
and
pepper
it
around
the
traces
around
the
three
dimensional
view
where
you
go
from
a
telescopic
performance
view
into
a
microscopic
code
level
performance
view.
And,
lastly,
because
you
know,
ultimately,
there
is
seed
funds.
We
see
money
riding
on
a
lot
of
startup
projects.
A
You
want
to
bring
in
logging,
as
one
of
your
key
proof
points
to
be
able
to
see
exactly
the
line
of
code
that
is
emitting
an
error
or
a
warning.
So
that's
the
last
part
we
have
kept
here.
If
you,
if
you
slowly
start
to
build
your
observability
strategy
for
your
kubernetes
service,
you
could
approach
it
in
this
way.
I
want
to
close
it
off
with
I
know.
We
ran
out
of
time.
A
I
want
to
close
it
off
with
a
really
important
aspect,
which
is
to
reclaim
money
left
on
the
table
in
terms
of
cost
containment.
You
could
execute
for
your
kubernetes
services
as
well,
so
I
call
this.
Keep
your
compute
lean
and
keep
the
planet
green.
Because,
apparently,
when
you
talk
about
claiming
your
optics
cause,
it
doesn't
motivate
people
enough.
So
let's
do
it
for
the
planet
slightly
different
spin
to
it
still
super
relevant,
regardless
of
the
platform's.
You
are
able
to
directly
view
your
k8
annotation,
together
with
your
cloud
annotation.
A
So,
regardless
of
the
cloud
platform
you
use
and
you're
able
to
dissect
your
data
in
a
way
that
you
could
say,
show
me
my
certain
cluster
node
group,
and
then
it
gives
you
a
view
into
what
are
your
optimization
candidates
so
literally
able
to
extract
this
report?
This
is
an
open
source
plugin
and
we
supported
on
developer
New
Relic
comm,
so
you
could
literally
repurpose
the
code
to
be
able
to
work
it
within
your
own
systems
and
what
it
gives
you
is.
A
We've
seen
six
figure,
ten
seven
figure
savings
for
some
of
the
larger
clients
who
are
cloud
native
and
and
they
repurpose
those
funds
to
be
able
to
do
faster
delivery
to
the
market,
do
more
automation
and
better
transaction
delivery.
So
I
think
we
want
to
provide
our
counterparts
more
time
for
their
part
of
the
talk
as
well.
If
you
found
the
the
content
intriguing
and
you
want
some
more
details,
we'll
be
happy
to
share
the
models
with
you.
Thanks
for
your
time.
Thank
you.