►
From YouTube: The Future of Open Source Observability
Description
The dynamic, distributed nature of today’s cloud computing and microservices environments presents a new set of observability challenges that can be best addressed by an open source, community-driven approach. Sign up for free → https://trynewrelic.com/datanerddays
Gartner predicts that by 2025, 70% of new cloud-native application monitoring will use open source instrumentation rather than vendor-specific agents for improved interoperability. Also the Cloud Native Computing Foundation (CNCF), End User Technology Radar shows that the most commonly adopted observability tools are all open source.
Join our panel of open source experts to learn about the current role and importance of open source in observability and how projects like OpenTelemetry and Pixie are shaping the future of open source observability.
Check out the rest of the Data Nerd Days 2.0 playlist → https://www.youtube.com/playlist?list=PLmhYj7Jl81JHVUZvGmDthWZ0AqqkMsP_p
Subscribe to our YouTube Channel → https://www.youtube.com/c/NewRelicInc?sub_confirmation=1
Session Track: Lunch Panel; Topic: Open Source; Presenters:
Priyanka Sharma, CNCF
Bartlomiej Plotka, RedHat
Zain Asgar, Pixie & New Relic
Michelle Nguyen, Pixie
Melissa Klein, New Relic
Jack Berg, New Relic;
#DataNerdDays #Panel #OpenSource
A
Just
so
you
know,
if
you
have
any
questions
as
we
go
along,
please
feel
free
to
put
them
in
the
chat
at
the
end
of
the
session.
We
are
going
to
go
one
by
one
and
try
to
go
through
as
many
as
we
can
alright.
So,
let's
get
started
so
I've
already
told
you
all
who
I
am
I'd
like
to
get
the
panelists
to
do
the
same.
How
about?
Let's
start
with
you
barte.
B
Sure,
hello,
everyone,
my
name
is
bartek
plotka.
I
live
in
london
and
I
work
for
work
on
open
source
observability
at
red
hat,
I'm
the
maintainer
of
two
big,
bigger
cncf
projects
from
meteos
and
tunnels,
and
I'm
also
active
in
this
cloud
native
computing
foundation
as
the
cncf
ambassador
and
tac
observability
tech
lead.
B
I'm
also
writing
a
book
with
aurelie
called
efficient,
go
how
I
got
involved
in
open
source.
Well,
I
would,
I
would
be
honest.
I
had
an
amazing
mentor
that
was
patient
enough
to
review
my
pr
to
apache
mesos
six
years
ago,
so
yeah,
that's
it.
This
is
how
I
felt
in
love
and
yes,
mentorship
is
very
crucial
in
open
source.
A
C
Sorry
about
that
my
name
is
jack.
Berg
I've
been
writing
software
and
have
been
a
consumer
of
open
source
for
quite
some
time
now.
I've
worked
in
a
variety
of
languages
but
have
most
experience
working
in
service
oriented
architectures
with
java.
I
work
for
new
relic.
I
joined
here
about
a
year
ago
to
help
with
open
source
java
initiatives,
and
I
currently
spend
most
of
my
time
working
on
the
open,
telemetry
java
project,
so
yeah
hello,.
A
D
Great
thanks:
yes,
I'm
melissa
klein.
I
am
the
open
source
program
manager
at
new
relic.
I
run
our
open
source
program
office.
I've
been
doing
open
source
program
work
for
a
lot
of
years,
but
I
started
my
career
as
a
java
developer,
working
with
open
source
frameworks
and
things
in
that
development,
and
so
I
learned
to
love
open
source
as
developer
and
moved
over
to
the
program
management
space.
A
E
Hi
everyone
I'm
michelle
nguyen-
and
I
am
a
principal
software
engineer
at
new
relic-
and
I
am
one
of
the
core
maintainers
of
pixi
and
pixi
is
kind
of
just
my
first
beginning
in
my
journey
in
open
source
and
I'm
just
super
excited
to
be
able
to
go,
and
you
know
just
meet
everybody
in
the
community
and
and
learn
about
open
source.
Through
this
way.
A
Oh,
that's
awesome,
it's
so
cool
to
hear
that
michelle,
yes,
pixi
is
a
cncf
sandbox
project
and
I
really
hope
you're
finding
the
community
welcoming
and
meeting
cool
folks
and
all
of
us
here
who
are
who've,
been
here
a
little
bit
longer.
I
know
that
if
you
ever
need
anything
just
holler
and
you
will
be
supported
with
that
said,
let's
get
to
the
last
panelist.
Last
but
not
least,
I
must
say,
zayn
tell
us
about
yourself.
F
F
A
Yes,
absolutely
zane
is
actually
on
our
board
of
the
cncf,
so
I
get
to
talk
to
him
quite
frequently
about
pixie
and
all
things
cloud.
Natives,
it's
quite
fun!
Thank
you,
everybody
for
introducing
yourselves,
as
I'm
sure
the
attendees
you
you
are
noticing.
This
is
quite
an
experienced
group
over
here
with
all
kinds
of
different
perspectives.
A
A
Observability
is
no
exception,
even
so.
Cncf
releases,
these
end
user
tech,
radars,
which
are
technology,
thought
leadership
pieces
by
end
users
who
utilize
these
technologies
and
they
pick
a
topic
and
they
recently
publish
one
on
observability,
because
it's
important
enough,
it's
more
than
important
enough
and
there's
a
wide
plethora
of
solutions
and
tools
available
and
there's
some
proprietary.
There's
some
open
source,
there's
all
kinds
of
like
mixes,
and
I'm
curious
to
hear
from
you
folks,
on.
A
A
B
It
makes
all
sense
and
for
me
I
feel
it's
very
critical
that
we
have
lots
of
observability.
You
know
in
open
source
and
to
me
this
is
because
observability
is
key
to
troubleshooting
and
reliability
work
or
on
anything
else
on
any
other
software.
So
if
we
have
less
access
to
those
tools,
it
essentially
reduces
innovation
in
other
fields,
right
because
people
spend
time
recreating
those
tools
and
creating
everything
from
scratch.
B
F
Sure
so
one
thing
I
wanted
to
add
is
you
know,
part
of
the
benefit
of
having
things
like
open
telemetry
is
that,
as
you
said,
priyanka
a
lot
of
a
lot
of
business
logic
is
a
lot
of
companies.
Basically
work
on
implementing
the
business
blue
logic
around
a
bunch
of
different
open
source
projects
and
having
something
like
open.
Telemetry
allows
you
to
instrument
the
core
infrastructure
in
in
a
flexible,
open
manner,
so
that
you
don't
have
to
recreate
that.
As
bartok
said
earlier,.
D
Definitely
I
think,
when
you're
talking
about
observability,
which
is
you
know
core,
how
do
you
make
sure
everything
is
up
and
running
and
and
performing
as
it
should
and
you're
talking
about
a
variety
of
architectures
and
languages
and
frameworks?
You
know
to
bring
that
all
together
in
a
standardized
way
is,
is
like
one
of
the
core
benefits
of
having
open
source
having
the
standard
way
to
talk
to
each
other,
no
matter
what
language
or
architecture
or
framework
you're
using.
E
No
yeah,
I
think
everybody
kind
of
summarized
my
thoughts
already,
but
I
mean
I
do
want
to
add.
It
is
really
just
like
very
special
to
have
this
community,
where
it's
like
sure,
every
at
the
end
of
the
day,
everybody's
application
is
different,
but
you
kind
of
run
into
the
same
issues
so
being
able
to
share
that
knowledge
across
just
like
so
many
different
teams
and
so
many
different
companies.
That
is
just
what
is
so
special
about
it.
A
A
You
know
you'll
be
surprised
how
sometimes,
maybe
an
automotive
company
can
support
amusing
music
streaming
apps
engineers
based
on
their
past
experience,
so
it
sound
just
based
on
this
first
question
that
I
had
for
you
folks,
it's
very
clear.
The
criticality
of
open
source
observability
and
melissa
mentioned
how
standardization
that
is
based
on
open
source
is
going
to
go
a
really
long
way
in
helping
everybody
involved.
A
C
Sure,
standards
and
technology
in
engineering
are
really
useful
in
general
because,
as
bartek
pointed
out
in
the
first
question,
you
know
what
they
allow
you
to
do
is
spend
less
time
on
integration,
problems
and
more
time
building
useful
products
when
a
standard
exists
and
when
it's
thoughtful
and
well
adopted
it's
the
obvious
choice.
It's
not
something
you
need
to
think
about.
C
You
know
an
example
that
comes
to
mind
is
hundreds
of
years
ago.
There
were
competing
systems
for
weighing
and
measuring
things.
Nowadays,
there's
there's
no
decision,
we
all
use
metric
units
and
even
for
the
folks
that
think
they're
using
imperial
units
out
there,
the
imperial
system
is
actually
based.
C
Fundamentally,
the
definitions
are
defined
based
on
the
metric
system,
so
one
inch
is
defined
as
exactly
2.54
centimeters,
so
imperial
units
is
just
metric
units
in
disguise,
and
I
I
think
open
telemetry
has
the
potential
to
to
to
become
that
standard
for
observability,
and
if
that
happens,
then
there
would
be
a
lot
less
duplicated
effort
which
would
allow
the
overall
quality
of
instrumentation
to
increase
and
for
application
authors.
What
that
means
is
that
it
would
be
easier
to
obtain
observability
data.
C
A
B
Definitely
I
I,
I
love
those
kind
of
directions
we
are
taking
here
in
the
cncf,
and
I
fully
agree
with
everything
jack
said,
and
I
just
want
to
add
that
we
already
see
the
fruits
of
like
open
standards
and
and
the
key
part
is,
you
know
the
the
stability
of
the
standard.
So
people
can
trust
that
this
will
not
change,
so
they
can
build
upon
that.
B
So
we've
seen
that
with
openmetrics,
which
you
know,
which
is
based
on
like
very
strong
prometheus
exposition
format,
established
like
2015,
and
that's
why
we
see
you
know
huge
ecosystem
of
exporters,
and
you
know
literally
for
everything,
ebpf
containers,
linux,
minecraft
servers
even
planned
health
for
gardening
right,
crazy
stuff.
Almost
every
cloud
application
is
expected
to
have
some
kind
of
open
metric
support,
so
it's
pretty
pretty
wild.
I
want
to
also
focus
on
this
open
world
word
right,
it's
very
important
for
the
standard
to
be
open,
so
it's
like
in
the
middle
ground.
B
You
know
between
all
all
companies,
bigger
and
smaller
and
open
source
is
a
great
place
for
that,
because
you
know
it
just
builds
this
trust
that
no
one
can
you
know.
Maybe
destroy
or
impact
this
protocol
in
a
you
know,
maybe
in
you
know
some
kind
of
inevitable
way,
so
open
standard
with
open
governments,
governance
is
is
gives
everyone
opportunity
to
contribute
for
overall
value
of
the
of
the
standard,
and
so
that's
great.
A
You
know,
if
I
put
my
you
know
first
time
hearing
about
this
hat
on,
which
is
not
the
case,
but
I'm
gonna
just
do
that.
I
heard
you
folks
completely.
A
line
on
like
hey
standards
are
essential
and
then
we
talked
about
open
telemetry
and
we
also
talked
about
open
metrics,
so
I'm
kind
of
sensing
a
little
bit
of
a
standard
sprawl
here
you
know
back
in
the
day
when
I
was
working
on
open
tracing,
we
used
to
reference.
The
xkcd
comic,
which
says
you
know
there
is.
A
We
must
have
standards,
and
then
somebody
releases
the
15th
standard,
and
so
I
would
love
to
hear.
First,
you
folks
take
on
open
trees,
open
telemetry,
open
metrics,
like
what
are
we
even
supposed
to
focus
on
and
then
I'd
like
to
open
it
to
the
rest
of
the
panelists
for
their
take
on
the
standard
sprawl
out
there.
B
Yeah
happy
happy
to
start
this
time,
so
I
don't
think
a
large
amount
of
options.
No,
you
know,
is
essentially
it's
necessarily
a
bad
thing.
It's
key
and
very
important
thing
to
encourage
friendly
innovations
in
this
space,
and
actually
one
of
the
most
important
missions
in
the
cncf
is
like
no
kings
making
rule
where
you
know
you.
B
We
are
not
here
to
to
to
choose
winners:
it's
rather
helping
communities
yeah
to
to
help
and
grow
and
help
those
who
bring
the
the
most
value
right-
and
I
think
the
most
important
part
here
is
that
we
don't
have
30
standards,
but
two
that's
already
a
huge
huge
win
and
there
is
a
truth
that
those
two
standards
are
not.
You
know,
especially
open
telemetry
metric
part
and
open
metrics
are
right
now
not
compatible,
but
we
are
working
together
and
we
are
learning
from
each
other
and
each
of
those
parties
brings
some
value.
B
A
C
Yeah
sure
so,
yeah
there's
there's
definitely
some
sprawl,
but
I
think
most
of
the
the
work
that's
being
done
today
is
is,
is
being
done.
You
know
from
a
good
faith
effort
to
build
on
the
work
that's
come
before
and
you
know
you
know,
incorporate
the
learnings
and
and
hopefully
produce
something
better.
One
thing
I
want
to
point
out
is
that
I
think
open
telemetry
actually
in
to
some
extent,
has
the
opportunity
to
actively
reduce
sprawl.
C
One
of
the
important
components
in
the
ecosystem
is
the
open,
telemetry
collector,
and
so,
rather
than
simply
introducing
a
new
standard,
open
telemetry
has
introduced
this
collector
component,
which
acts
as
an
adapter
for
compatibility
with
other
standards.
So
it
accepts
telemetry
data
in
in
lots
of
different
protocols
like
zipkin
and
jager
and
prometheus
and
more
and
maps
them
to
a
common
representation.
C
So
you
can
do
useful
things
in
a
centralized
location
and
so
to
me,
this
is
like
you
know,
imagine
if
there
was
a
new
cell
phone
charging
or
data
standard
that
came
out
to
replace
usbc
and
instead
of
just
forcing
you
to
buy
all
new
data
charging
bricks
and
cables.
There
was
like
an
adapter
that
allowed
your
old
cables
and
charging
bricks
to
work
with
it.
That's
what
I
think
of
as
the
collector
and
so
so
yeah.
A
E
Yeah
yeah
definitely
I
mean
I
think
I
found
that
being
able
to
have
these
standards
makes
it
so
easy
to
just
provide
use
to
people
from
pixie
immediately.
So,
for
example,
a
lot
of
people
ask
oh,
it's
pixie:
does
it
work
with
graffana?
Does
it
work
with
prometheus
and
all
these
other
different,
open
source
tools
and
by
being
able
to
provide
you
know
using
these
connectors?
E
A
F
C
Of
the
core
open
telemetry
data
types
which
are
metrics,
traces
and
logs
to
start,
there's
the
sdks
and
the
apis-
and
these
are
essentially
libraries
written
in
a
variety
of
languages
and
the
apis
are
what
library
and
framework
maintainers
will
use
to
instrument
their
code
and
the
sdks
are
what
application
owners
use
to
configure
what
happens
with
the
telemetry
data,
that's
generated
so
on
this
front.
Tracing
is
the
furthest
along.
C
Several
languages
have
released
stable
versions
and
metrics
is,
is,
is
next,
so
the
metric
spec
has
stabilized
for
open
telemetry
and
several
languages
are
expected
to
release
stable
versions
in
q1
of
2022
logs
the
last
data
type.
It's
not
exactly
clear
when
we
can
expect
stable
versions
of
apis
and
sdks,
but
I
think
a
safe
bet
is
the
second
half
of
2022.
C
Closely
related
to
instrumentation
are
the
open,
telemetry
semantic
inventions,
and
what
these
are
is
these
provide
describe
standards
for
how
to
shape
telemetry
data
for
common
operations?
So
in
a
nutshell,
if
you're
instrumenting
a
http
server,
that's
written
in
java,
it
should
emit
telemetry
data
that
is
similar
in
shape
to
an
http
server.
That's
written
in
python
and
the
semantic
inventions
are
what
allows
this
to
happen,
and
so
there's
been
progress
on
this
front,
especially
in
the
area
of
http
servers
and
messaging
systems.
C
But
everything
is
still
experimental
for
now
and
then
you
know
I
talked
about
the
open
telemetry
collector
before
I
just
want
to
provide
like
a
quick
update
on
that.
So
if
you
haven't
heard
of
it
the
collector
it's
it's
really
important
in
open
telemetry,
it's
essentially
a
highly
configurable
and
extensible
data
pipelining
tool.
C
So
there's
a
lot
of
moving
parts
and
there's
been
tremendous
progress,
and
you
know
I'd
say
that
up
to
this
point,
a
lot
of
the
work
has
been
foundational
in
nature,
we're
defining
the
core
data
types
and
the
common
apis
and
sdks,
and
once
metrics
and
logs
stabilize
I
anticipate
open
telemetry
to
get
a
lot
more
interesting,
as
we
start
to
see
folks
build
upon
those
those
foundational
pieces.
So
that's
a
that's
a
quick
update
on
the
state
of
hotel.
You
know
it's
a
work
in
progress,
but
you
know
there
there's
definitely
progress.
A
B
Sure
I
will
take
only
like
a
couple
of
seconds
so
matrix
things
are
much
easier
because
open
metrics
is
just
a
protocol
network
protocol.
Let's
say
for
exposing
your
metrics
and
it's
version.
One
zero
gaa
with
exemplars
with
you
know,
compatibility
with
promoters,
text
exposition
for
majority
of
things,
and
we
are
actually
looking
on
twosera
soon
with
you
know,
the
high
resolution
sparsed
histogram
feature,
and
we
also
have
pending
rfc
to
iatf
to
make
it
like
official
standard.
We
in
this
kind
of
you
know
weird
font
from
90s.
A
Very
cool,
very
cool:
well,
you
know.
So
this
is
great
stuff
but,
as
I
think
jack
said
very
correctly,
everything's
a
work
in
progress,
work
in
progress
and
that's
the
sort
of
this
you
know
the
state
of
being
when
you
work
in
software
and
definitely
open
source
software.
So
zayn,
I'm
curious.
What
advice
would
you
give
someone
who's,
trying
to
figure
out
how
they
should
operationalize
observability
in
their
organizations?
A
F
Oh
sorry
about
that,
I
think
that's
a
pretty
pretty
complex
question
to
answer,
and
it
really
just
depends
on
the
type
of
organization
and
what
you're
interested
in
doing
so.
I
think
there
are
a
lot
of
people
who
can
operationalize
open
source
software,
well,
install
it
and
maintain
it
themselves,
and
I
think
that's
a
fine
path
to
go
down.
There's
also,
obviously,
all
sorts
of
supported,
open
source
software
you
can.
You
can
buy
right
if
you're
not
actually
interested
in
maintaining
maintaining
the
software
yourself.
A
F
Sure
so
I
can,
I
can
take
a
quick
pass
at
it.
So,
as
you
know,
for
those
that
are
they're
not
completely
aware
of
this
pxe
works
on
is
an
observability
system
that
uses
evp
to
collect
data
without
adding
manual
instrumentation,
and
then
we
can
export
data.
In
you
know
various
formats,
including
things
like
open
telemetry,
so
part
of
the
the
interesting
thing
about
evpf,
is
that
you
can
basically
instrument
things
using
the
linux
kernel.
F
So
ebpf
is
this
sandbox
that
runs
inside
of
the
linux
kernel
that
allows
you
to
basically
run
code
that
will
essentially
insert
code
in
the
linux
kernel
and
make
sure
that
it
doesn't
break
break
your
your
machine
right.
So
it's
like
a
kernel
module,
but
a
lot
a
lot
safer.
So
it
allows
you
to
do
things
like
probe
your
applications.
F
Look
at
network
traffic
and
things
like
that
without
having
to
install
like
a
language,
specific
or
application
specific
agent,
so
part
of
what
we
do
with
pixi
is
we
enable
you
know,
lots
of
cool
things
like
oh,
we
can
help.
You
take
a
look
at
all
the
traffic,
even
if
you
don't
have
any
collectors
installed.
F
In
addition,
we
can
do
things
like
dynamically
instrument,
your
your
applications
and
also
provide
you
know
flame
grass
for
your
applications.
You
can
debug
performance
problems
and
all
this
is
done
using
evp
app,
which
is
which
is
pretty
cool
right.
It's
a
it's!
A
very
new
emerging
technology.
That's
being
used
all
the
way
from
network
traffic
to
telling
you
how
much
cpu
like
individual
functions
are
taking.
So
I
think
it's
pretty
pretty
revolutionary
in
that
way.
F
So
I
mean
I,
I
actually
think
that
there
will
be
many
solutions
right.
Observability
is
a
pretty
complex
space,
so
doing
things
like
open,
telemetry
and
making
sure
all
the
core
infrastructure
is
well
instrumented.
I
think
goes
a
long
way
and
you
know
using
things
like
ebpf
for
parts
of
the
application
stack
that
are
not
well
instrumented
or
doing
more
dynamic,
dynamic
instrumentation.
For
for
what
I
call
like
production
and
debugging
and
production,
I
think
those
will
be
the
new,
interesting
cases
with
vpf.
A
B
Excited
by
new
ideas
and
and
one
of
one
of
those
is
like
continuous
profiling.
So
we
talk
about
you
know
this:
hey.
We
have
three
observability
signals,
maybe
that's
the
fourth
one
like
continuous
profiling
right,
which
allows
you
to
have
like
a
profile
of
your
application,
usage
of
memory
or
cpu
or
anything
else.
B
You
know
in
the
past
and
you
can
kind
of
snapshot
that
and
merge
and-
and
you
know,
kind
of
find
the
bottleneck
and
even
if
your
application
boom,
so
I
went
out
of
the
memory,
you
can
take
a
look
back
retrospectively
and
check
why
what
was
actually
leaking?
Let's
say
in
the
memory,
so
that's
something
we
are
we
kind
of
it's
already
in
open
source.
B
There
is
kind
of
we
hope
in
the
cncf
soon
as
well,
but
this
is
super
amazing
stuff.
Of
course,
observability
correlations
so
being
able
to
link
different
observability
signals
together
on
ui
and
things
like
exemplars,
and
you
know,
open,
telemetry,
context,
propagation
and
kind
of
creating
locks
from
or
metrics
from
traces
all
those
kind
of
connections
between
signals
yeah.
I
could
speak
a
lot.
A
Now
that
is
so
cool
I.
This
is
why
I
love
this
space.
You
know
observability,
there
is
so
much
to
be
done
and
we
and
the
crew
that
works
on
it
all
you
folks,
there's
always
innovation
in
your
minds.
There's
like.
Oh,
we
figured
this
out,
let's
bring
ebpf
and
let's
do
continuous
profiling,
and
so
I
you
know
this
is
a
space
app.
As
I
told
you,
I
got
interested
in
infrastructure
because
of
observability.
I
think
it's
super
dynamic
and
folks
who
are
listening.
A
You
know
check
all
these
projects
out
check
out
what
they
do,
because
if
nothing
else,
you
will
find
some
really
interesting
cool
tech
to
look
at
speaking
of
interesting
cool,
new
tech.
To
look
at.
Let's
talk
a
little
bit
about
the
project,
pixie
michelle.
I
would
love
to
hear
from
you
about.
Like
you
know,
we
all
we've
all
discussed
how
okay,
it's
a
cncf
sandbox
project,
but
tell
us
more
about
what
pixi
does
and
then
why
a
new
relic
decided
to
make
it
open
source
and
contributed
to
cncf.
F
E
But
it's
just
like
using
technology
such
as
ebpf
we're
able
to
provide
instant
visibility
into
your
cluster.
Without
you
know,
any
complicated
deployments.
All
you
have
to
do
is
just
run
one
simple
command,
which
is
our
cli
px
deploy
and
like
immediately
you
just
get
tons
of
data
about
your
system
from
like
network
traffic
to
error
rates,
so
yeah
pxe
is
just
being
able
to
provide
that
experience
and
ship
all
this
data
out
to
like
other
interesting
places
or
viewable
through
our
ui
yeah.
E
A
E
Yeah
for
me,
I
think
it's
been
really
great.
I
think
there
are,
of
course,
like
small
things
that
we
realized
when
we
decided
to
open
source
pixie,
and
we
saw
just
like
oh
no.
We
have
a
lot
of
embarrassing
code
here
that
now,
suddenly
all
these
people
are
going
to
see-
and
I
mean,
of
course
that's
a
part
of
it,
but
the
other
part
is
just
being
able
to
now
fully
interact
with
this
community
and
share
with
people.
A
Is
yes,
I
can
totally
see
that
I'm
so
glad
it's
working
out
actually
melissa.
You
must
have
been
involved
in
a
large
capacity,
as
the
decisions
were
being
made
on
open
sourcing,
etc.
I'd
love
to
hear
what
your
experience
was
like
where,
where
you
were
most
worried
and
how,
how
how
has
it
all
turned
out.
A
F
F
We've
implemented
that
using
ebpf,
but
it
has
pretty
limited
support
on
compiled
languages
right
now
and
we're
just
about
to
expand
that,
to
you
know,
languages
like
java
and
hopefully
node.js
and
python
soon.
So
I
think
that
feature
set
of
just
getting
more
deep
application.
Visibility
is
something
that
we're
gonna
constantly
be
working
on.
Another
area
that
we've
been
pushing
a
lot
on
is
just
around
this
idea
of
like
data
science
for
infrastructure.
F
One
of
the
things
I
believe
about
believe
in
is
like
observability
in
some
ways
should
be
just
like
a
regular
part
of
every
developer's
toolkit,
and
not
just
you
know,
but
maybe
it'll
evolve
into
something
else,
but
not
just
about
observing,
but
in
terms
of
like
actually
taking
action.
So
one
of
the
talks
that
we
gave
at
kubecon
was
around
data
science
for
infrastructure
and
how
to
use
information
from
pixie
duh
to
dynamically
change
your
kubernetes
cluster.
So
those
are
types
of
things
that
we're
interested
in
experimenting
on
next.
A
F
Yeah
so
pixie
has
this
concept
of
pixel
scripts,
which
are
basically
you
know,
pandas,
compliant
or
mostly
compliant
scripts
and
pandas
is
like
a
data.
Processing
library
in
you
know
that
a
lot
of
data
science
people
use
in
python.
So
we
have
this
concept
of
pixel
scripts
that
actually
can
both
instrument
and
process
data
and
we're
always
looking
for
people
to
contribute
new,
pixel
scripts
or
different
use.
F
Cases
like
here's,
how
you
best,
monitor
a
kafka,
kafka
system
or
a
service
of
toxic
kafka,
and
we
really
hope
to
build
out
like
a
repository
of
script.
So
that's
like
one
area-
that's
that's!
You
know
right
for
contribution.
We're
also
always
looking
for
new
support
for
for
protocols.
Someone
just
you
know,
implemented
a
thrift
protocol
and
submitted
it
to
pixie.
So
things
like
that
are
you
know
very
good
for
the
project
overall,
for
it
to
become
more
more
usable
by
the
community.
A
C
Yeah,
so
you
know,
I
think
I
think
the
most
important
thing
is
for
folks
just
to
get
out
there
and
start
using
open
telemetry.
You
know
so
with
all
this
work,
that's
been
on
the
foundational
pieces
like
the
api
and
sdk
development.
You
know
there's
there
has
maybe
been
some
neglect
of
things
like
documentation
and
we
just
need
more.
A
Yes,
I
I
always
tell
people,
you
know,
take
a
pr
first
approach,
just
because
you
never
know
when
you
might
have
really
helped
a
project
out,
and
you
are
just
just
by
taking
that
mindset.
You're
going
to
like
break
your
own
barriers
of
how
much
involvement
you
can
have.
I
want
to
ask
a
question
related
to
open
telemetry
metrics.
A
C
Yeah
so
they're
different,
so
you
know
it
depends
on
exactly
what
you
mean,
but
open
telemetry
is
defining
apis
that
allow
you
to
instrument
code
with
metrics
right,
but
open.
Telemetry
also
has
this
thing
called
the
open
telemetry
protocol,
which
is
how
to
transmit
open
telemetry
data
over
the
wire
and
so
that
that
protocol
is
what
is
what
you
might
refer
to
as,
like
you
know,
is
it
the
same
thing
as
prometheus
and
no
it's
not
prometheus.
C
You
know
the
standard
way
to
do
it
is
to
expose
a
metrics
endpoint,
which
displays
like
your
metrics
data
in
a
in
a
particular
format.
Otlp
or
open
telemetry
protocol
is
not
like
that.
That's
a
primarily
a
binary
protocol,
that's
based
on
http
and
grpc
that
defines
protobuf
messages
that
describe
how
to
represent
traces,
metrics
and
logs,
and
so
no
not
quite
the
same
thing.
But
with
that
said,
open,
telemetry
metrics,
one
of
the
core
design
principles
was
full
prometheus
compatibility,
and
so
that's
not
going
anywhere.
C
And
you
know,
as
I
mentioned,
they
get
mapped
to
a
common
representation
where
you
can
do
things
like
normalization,
or
you
know,
data
enrichment
before
exporting
to
wherever
back
whatever
back
end.
You
ultimately
want
to
store
those
in
which
can
also
be
prometheus,
because
prometheus
is
like
the
format
that
applications
expose
metrics
in,
but
also
you
know
a
database
in
in
querying
tool
right.
So
you
can.
You
can
scrape
metrics
with
the
collector
and
then
send
those
to
prometheus.
C
A
A
When
I
was
working
on
open
tracing
I
was
I
was
at
a
company
called
lightstep
and
we
were
often
thinking
about
just
how
do
we
provide
insights
and
information
to
the
people
who
look
at
this
data
to
know,
for
example,
customer
a
is
having
a
terrible
experience
right
now,
while
everybody
else
might
be
just
fine,
but
customer
a
is
like
the
largest
customer.
I
have
so
I
better
get
on
it
or
hey.
The
ceo
is
looking
through
the
application
and
it's
the
latency
is
high.
A
B
If
I
understand
the
question
correctly
is
around
like
kind
of
smart
processing
of
this
data
after
you,
you
gather
them
right,
yes
and
and
that's
a
powerful
notion
and-
and
really
this
is
where
you
know
gathering
all
the
events
from
your
application
is
is
important,
because
then
you
can
know
that
you
know
decide
which
of
those
are
important,
more
important,
less
important
and
that's
kind
of
the
that.
Definitely
that
would
be.
B
You
know
more
solutions
around
this,
like
maybe
even
ai
or
some
kind
of
you
know,
pattern
matching
where
you
see
that
the
certain
data
is
more
useful
than
other
parts
of
data
and
what
to
bring
to
the
user,
and
I
think
it's
very
important
on
the
kind
of
open
telemetry,
especially
on
tracing
side,
because
there's
lots
of
data
right
and
usually
it's
auto
instrumented,
usually
idd
machines,
sort.
This
data
out
and
bring
you
know,
only
reduce
the
amount
of
what's
human
human
readable
already
right.
B
B
You
know,
which
is
there
are
so
many
kind
of
techniques
and
the
best
one
I
seen
was
like
a
tail-based
sampling.
Where
actually
you
decide,
if
you
have
you,
you
kind
of
collect
some
data
around
when
you
know
what
happened
to
the
request,
maybe
it
was
too
slow,
so
you
only
gather
two
slow
requests.
Things
like
that,
and
I
think
we
are
slowly
stepping
up
into
big
ai
kind
of
systems,
but
it's
it's.
It's
kind
of
in
future.
A
It's
more
futuristic
got
it,
and
you
know
I
think
this
conversation
is
like
it's.
I
think,
in
a
spot
at
which
you
know
the
conversation
about
observability
is
also
starting
to
feel
a
little
bit
like
a
conversation
around
apm
or
application
performance
monitoring,
and
I
I'm
curious,
especially
from
the
new
relic
folks,
to
hear
about
just
how
do
you
differentiate
between
observability
and
apm.
D
Sure
I
can,
I
can
give
my
high
level
thoughts
yeah,
so
so
for
observability
we're
looking
at.
Let's,
actually,
let
me
start
with
apm,
so
apm
is
going
to
collect
your
data.
It's
going
to
monitor
your
applications
right!
That's
that's
the
core
purpose
of
apm
from
observability
we
wanted.
We
wouldn't
actually
be
able
to
do
something
with
that
data.
We
want
to
collect
that
data
and
see
where
you
know
where
those
spikes
are.
D
Where
are
those
those
signals
that
are
showing
us
that
there
is
something
going
on
with
our
system
that
we
need
to
take
action
on
and
when
we're
monitoring
all
of
these
systems
and
getting
all
this
great
data
from
apm
and
and
more
and
more
systems
we're
going
to
have
more
data
that
we
need
to
sift
through
and
be
able
to
be
able
to
see?
Which
ones
are
really
something
that
needs
action
or
that
something
that
may
be
indicating
that
an
event
will
happen
that
we
can
perhaps
prevent.
So.
D
D
F
A
E
It's
like
contributing
is
just
like
you
know,
going
in
and
writing
code,
but
there's
so
many
other
ways
that
you
could
contribute.
So
it's
like
with
you
know,
putting
in
feature
requests
finding
bugs,
I
think,
even
just
like.
I
think
jack
mentioned
this
earlier,
but
it's
just
like
honestly
a
lot
of
open
source
projects.
They
need
help
with
documentation.
I
think
things
are
moving
so
so
quickly.
A
100
agree
with
that,
you
know
in
cncf,
as
you
know,
we
have
114
actually,
as
of
today
115
projects
because
dapper
just
joined
and
that's
what
every
maintainer
contributor
is
telling
me
that,
please
don't
wait
to.
You
know,
send
that
pr
that
will
revolutionize
this
project
with
your
code
no
start
at.
Like
you
know,
we
have
like
the
tagging
of
good
first
issues
where
it's
simple
stuff
or
it's
like
go
write,
go
help.
You
know
not
even
write
the
documentation,
go,
read
it
and
fix
typos.
A
That
can
be
even
a
great
great
way
to
get
started.
I
think
there
is
a
huge
barrier
that
people
face
bef
until
they
submit
a
pr,
and
so
I
encourage
anybody
who's
even
like
moderately
interested
to
go
just
do
the
smallest
possible
thing
you
can
and
know
that
it
will
be
appreciated.
It
won't
be
seen
as
oh,
you
fixed
a
typo
it'll,
be
like
oh
my
gosh,
thank
god.
Somebody
noticed
that.
So
just
please
keep
that
in
mind.
A
If
you're
listening,
if
you're
watching
this,
that
anything
you
do
is
going
to
be,
support
is
going
to
be
in
support
of
the
greater
mission,
and
I
would
also
say
that
that
includes
non
non-github,
touching
contribution
as
well.
There
are
so
like
you
know.
Just
two.
Three
weeks
ago,
we
concluded
the
flagship
conferences
we
host
kubecon
cloud
native
cons,
which
are
the
largest
open
source
conferences
in
the
world.
A
I
see
all
the
folks
planning
that
event
all
the
folks
who
come
figure
out
who
come
in
like
man
boots
who
help
people
set
up
for
the
experience,
they're
all
contributing.
So
I
think
there
is
a
myth
out
there
that
the
only
way
to
be
part
of
open
source
is
to
be
a
programmer,
and
that
too,
like
a
wizard
of
one,
no
every
little
piece
counts.
Every
little
thing
you
do
counts
and
I
think
people
just
knowing
that
is
probably
really
good.
C
I
I
one
one
thing
I
would
add
is
that
you
know
just
if
there's
a
project
you're
interested
in
or
if
there's
a
project,
you're
you're
using
clone
it
bring
it
down
locally
run,
it
run
the
build.
You
know,
make
it
something,
that's
more
concrete
and
not
just
this.
This
kind
of
fuzzy
hand,
wavy
thing
that
you
depend
on
I
I
swear.
The
code
is
way
more
approachable
than
you
might
think
so,
yeah
just
give
it
a
try.
A
F
We
don't
we
don't
know
yet,
in
terms
of
exactly
I
mean
pixie
is
a
pretty
much
kind
of
a
larger
mono
project.
Right,
like
all
of
our
uis
and
stuff
were
built
in.
It
is
possible
that
over
the
next
couple
of
years
will
probably
break
apart
some
parts
of
pixie
and
then
contribute
them
to
separate
projects
depending
on
how
things
go,
and
then
there
may
be
other
projects
like
open,
telemetry
collectors
and
stuff
that
we
might
contribute.
A
F
A
A
A
It's
a
great
place
to
join
in
get
involved,
find
a
community
and
also
advance
your
career,
so
participating
seems
great
what
about
the
challenges
of
using
open
source
in
your
companies
in
your
organizations-
and
I
know
zayn-
you
had
briefly
touched
on
it
as
for
some
folks,
it
like
may
not
be
worth
it,
but
but
where?
How
should
people
determine
today?
I'm
good
this
thing,
I'm
gonna,
you
know
get
all
in
and
use
this
open
source
library
directly.
This
other
thing
I'ma
buy
a
managed
service.
B
B
However,
I
I
you
know,
I
talk
with
contributor,
I
talk
with
vendors
and
you
know
open
source
communities
and
I
feel
stability
is
very
important,
so
I'm
I
think
we
can
do
a
bit
more
work
around.
You
know
making
sure
the
license
you
know
stays
the
same,
that
we
understand
other
licenses
and
you
know
the
rationales
for
that
and
how
this
impact
other
open
source
projects,
like
you
know,
recent
moves
into
a
gpl
or
sspl
or
those.
I
wish
from
the
developer
perspective
that
I
don't
need
to
care
about
that
right.
A
D
You
have
that
comfort
level
that
that
you
will.
There
will
be
some
stability
in
that
project,
because
there
are
policies
and
rules
set
up
to
keep
that
keep
that
going
along
with
the
vendor
neutrality,
as
zane
mentioned
earlier.
So
when
you're
looking
at
deploying
the
open
source
yourself
to
have
picking
up
open
source
from
foundations
is
a
great
great
way
to
go.
If
you're
looking
for
stability.
C
So
my
employer
new
relic
is
is
good
about
this,
and
but
I
have
worked
for
employers
that
make
you
dump
through
a
lot
of
hoops
just
to
contribute
bug
fixes
back
to
the
projects
that
you
depend
on
and
so
in
general.
I
think
companies
should
like
take
an
extra
effort
to
reduce
the
red
tape
and
and
playing
and
try
to
play
a
more
active
role
in
in
maintaining
the
open
source
projects
that
they
depend
on.
So
you
know
you
know,
try
to
fight
this.
C
You
know
many
projects
maintained
by
the
few
and
make
it
more
make
it
more
balanced.
And
on
that
note,
I
actually
have
to
run.
I
have
to
get
another.
I
have
to
be
involved
in
another
nerd
days.
Talk
I'm
going
to
be
instrumenting
an
application
in
java
with
open
telemetry.
So
if
you're
interested
in
that
come
hang
out
with
me,
but
it's
been
great
talking
with
you
all.
A
B
Yes,
I
can,
I
have
like
you
know
one
maybe
tricky
question,
but
I'm
super
curious.
What
is
the
secret
sauce
in
especially
in
pixie
team
around
portability,
because
you
know
when
you
play
with
epf
ebpf
is
very
powerful.
It's
very
you
know
it
allows
you
to
do
really
anything
inject
your
code
in
in
you
know,
kernels.
B
So,
like
the
thing
that
controls
everything
and
the
problem
is
like
it's
kind
of
sometimes
fragile,
I
was
you
know,
I
managed
to
set
up
some
key
probes
or
trace
points
on
my
machine,
but
you
know
when
I
shown
that
to
my
colleague
she
wasn't
able
to
run
it
because
you
know
different
kernel
versions.
So
how
big
she
you
know
helps
I
mean
how
pixie
kind
of
sold
that
I
think
it
has
to
be
hard
work
here.
F
Yeah
I
mean
there's,
there's
a
bunch
of
stuff
that
happens
behind
the
scenes.
For
that
I
mean
we
have
to
package
up
certain
linux
kernel
headers,
we
always
make
sure
compatibility
with
some,
usually
an
older
kernel.
Things
usually
tend
to
work
as
you
move
forward.
They
just
don't
work
on.
You
know
if
you,
if
you
build
for
like
5.10,
it's
probably
not
going
to
work
on.
You
know
an
older
linux
kernel,
so
we
each.
F
I
think
we
target
like
4.12
right
now
as
our
as
our
base
kernel,
that
we
target
there's
also
a
bunch
of
stuff
to
figure
out
what
the
offsets
are,
especially
for
you,
probes,
yeah.
If
you're
curious,
you
can
probably
go
take
a
look
at
the
code
inside
of
like
source
sterling
ppf,
and
I
can't
even
point
you
to
point
you
do
it
if
you're
interested.
F
B
E
Like
for
me,
I
mentioned
like
pixie
is
my
first
like
kind
of
foray,
and
so
the
open
source
so
just
being
able
to
also
hear
like
other
perspectives
from
like
bartex
who's
been
just
like
the
community
for
such
a
long
time
has
been
really
great.
So
I
hope,
just
like
everybody
got
a
good
take
away
of
just
you
know
what
is
so
special
about
having
observability
be
open
source
and
just
how
it
benefits
everybody
and
how
they
can
get
involved.
D
I
think
what
we
learned
in
this
is
that
yes,
observability
is
all
over
open
source,
and
so
it's
a
great
time
to
jump
in
and
get
involved,
because
things
are
evolving.
Things
are
changing,
it's
a
great
time
to
to
start
taking
a
look
start
installing
looking
at
documentation
and
contributing
back,
even
if
it's
just
feedback
on
how
it
works
for
your
system,
absolutely.
F
B
Yeah,
I
would
just
really
would
like
to
finish
with,
like
some
call
for
action
right.
If
you
are
looking
into
you
know
how
you
can
grow
your
career
help
just
just
really
start
contributing
like
close
your
eyes
and
and
put
this
issue,
maybe
something
silly
at
the
start.
Don't
worry.
We
all
make
mistakes,
we
never,
you
know,
accumulate
anyone.
Everyone
is
friendly
in
cncf,
so
far,
I've
seen.
So
please
really
do
this.
B
Just
do
this
right
and
you
know
search
for
mentorships
and
I
also
want
to
encourage
project
maintainers
to
really
grow
in
this
mentoring
effort
as
well.
I
know
it
takes
time.
I
know
technically,
you
know.
No
one
pays
you
for
that,
but
it
it
like.
We,
we
hired
so
many
people
from
that.
We
got
new
maintainers
that
that's
that's.
I
think
the
most
valuable
thing
I
would
love
to
finish
with.
A
Absolutely
yes,
and
I
would
like
to
echo
all
of
the
panelists
in
that
open
source,
espa
and
especially
the
cloud
native
ecosystem.
We
are
all
team
cloud
native
and
our
strength
is
our
diversity
and
we
welcome
each
and
every
one
of
you.
I
think
observability
is
seeing
so
much
innovation
and
that
a
lot
of
it
is
coming
from
open
source
and
that's
going
to
continue
being
the
case
so
come
on
in
get
involved
and
enjoy
the
ride
with
us.
With
that
said,
thank
you
so
much
everybody.