►
From YouTube: NFV – What is Important and Why - Don Clarke
Description
Talking during Kubernetes Community Day on January 30, 2021.
https://community.cncf.io/san-salvador/
A
Colleagues
across
the
telco
industry
and
we
published
a
white
paper
which
was
now
quite
famous,
the
so-called
nfv
white
paper
published
in
october
2012,
and
this
announced
the
future
direction
for
telecom
networks
and
that
paper
introduced.
The
term
network
functions,
virtualization
or
nfv,
which
we
which
we
actually
ourselves
cooked
up,
and
then
we
founded
a
body
called
etsy
european
technology,
telecommunications
standards
institute
to
host
all
of
the
documentation
and
specification
work
that
we
wanted
to
do.
A
That
was
back
in
2013
in
2014.
We
wanted
to
get
the
open
source
communities
involved
in
what
we
were
doing.
So
we
founded
a
new
plan
open
source
community
called
open
platform
for
nfv
opnfv.
Probably
some
of
you
are
pretty
aware
of.
It
is
now
merging
with
another
organization
called
cntt
and
is
now
called
anuket.
So
opnfv
is,
is
now
transitioning
to
a
new,
a
new
form
back
in
2014.
You
guys
know
kubernetes
initial
release
from
google
and
that's
why?
A
I
guess
I'm
here,
because
kubernetes
then
became
a
new
way
to
implement
software
within
the
cloud
environments,
and
we
wanted
to
use
that
too.
The
first
commercial
nfv
deployments
took
place
in
2015.
Just
four
years
after
we
brought
out
our
white
paper
and
since
then,
we've
had
obviously
commercial
deployments
of
nfv
around
the
world,
but
at
cnnv
it's
been
doing
all
of
the
specification
work
for
the
industry
and
we've
had
now
three
releases
of
specifications
and
the
fourth
release
is
now
in
progress.
A
So
that's
really
the
basic
the
basic
concept
main
thing
is:
we
want
to
use
industry
standard
hardware
and
run
all
kinds
of
exciting
network
workloads
on
those,
but
there
are
some
points
to
remember
about
telecommunications
right
telecom.
Network
infrastructures
are
not
the
same
as
as
enterprise
it
environments.
A
It's
heavily
regulated
it's
massive
geographic
and
volumetric
scale,
and
there
can
be
no
downtime
for
software
updates
right.
So
when
someone's
dialing
9-1-1
on
their
phone,
you
can't
suddenly
have
that
call
cut
off
because
a
it
guy
somewhere
is
doing
some
maintenance.
So
that's
one
of
the
really
key
things
to
remember
about
telecommunications
network.
The
other
thing
about
telecoms
is
that
the
services
persist
for
many
many
years,
and
that
means
that
the
technology
itself
on
which
the
service
runs
is
changing,
but
the
service
itself
mustn't
change.
A
A
good
example,
of
course,
is
is:
is
telephony
making
a
call
today
is
the
same
as
it
was
50
years
ago
and
you
think
well,
you
know
the
technology
and
that
time
has
evolved,
but
the
service
hasn't
evolved.
Very
much
network
operators
are
unique.
Operations
are
unique
to
each
operator
and
jurisdiction,
so
the
requirements
for
telecoms
operators
in
europe
a
little
bit
different
from
the
us.
So
there
isn't
a
one
size
fits
all
for
the
telecommunications
environment
right.
So
one
operator
has
a
particular
set
of
requirements
and
time
scales.
A
A
Each
operator
must
be
in
full
control
of
its
destiny,
so
we
can't
have
a
community
that
decides
what
the
time
scales
are
and
tells
the
telecom
operator
what
that
time
scales
are
for
a
release,
saver
of
openstack
or
something
else
the
operator
says.
No.
I
want
that
release
on
this
day
and,
if
I
don't
get
it
on
that
day,
then
I'll
miss
some
of
my
regulatory
or
commercial
imperatives
so
operate.
Telecom
operators
must
be
in
control
of
their
own
destiny
and
that
really
comes
down
to
the
key
thing
in
telecoms
and
that's
standards.
A
When
people
hear
the
word
standards,
they
think
oh
that's
a
long
time
to
take.
It
takes
a
long
time
to
do
standards.
It's
very
slow
and
it's
not
very
innovative-
that's
probably
true
in
some
respects,
but
it's
absolutely
crucial
to
facilitate
innovation,
because
we
want
interoperability
between
all
of
the
people
that
contribute
technology
to
our
networks.
A
So
one
of
the
things
that
I
have
got
here,
I'm
not
going
to
go
into
the
detail
on
this
slide.
You
can
read
it:
how
comparing
standards
with
open
source?
What's
the
difference?
What's
the
pros
and
cons
standards
basically
is
about
specifications.
Open
source,
of
course,
is
about
code.
So
where
does
standards
not
an
open
source
meet?
Well.
First
thing
to
say:
is
that
standards
define
interoperable
frameworks?
A
These
are
the
frameworks
that
produce
the
network
architectures
that
we
want
to
deploy
and
that
one
we
want
that
to
support
software
diversity,
so
the
standard
is
issued
as
a
and
then
we
want
the
industry
to
deliver
to
us
the
the
software
solutions
that
support
the
support,
those
standards
open
source.
Of
course,
I
know
you
guys
that
know
about
open
source
during
kubernetes.
A
It's
not
an
either
or
they're,
both
very
important
in
our
industry
and
one
of
the
things
I'd
like
to
draw
your
attention
to
which
I
rail
knows,
I'm
passionate
about
is
diversity
so
in
in
all
things,
we
want
diversity.
We
want
all
kinds
of
viewpoints,
we've
got
all
kinds
of
cultures
to
bring
their
ideas,
and
so
one
of
the
things
that
I
I've
always
tried
to
do
in
my
career
is
is
form
collaborations
with
different
people
from
different
places
with
different
backgrounds.
A
So
I'm
basically
an
international
guy,
I'm
british,
but
I'm
living
in
the
united
states-
and
I
have
many
many
industry.
Colleagues
around
the
world
that
have
become
friends
over
time,
critical
capabilities
for
telecommunications,
so
this
list
is
something
which
we
all
hang
our
future
credibility
on.
Right,
designing
systems
that
are
secure
by
design
retrofitting
security
is
not
a
good
idea.
You
need
to
to
actually
design
security
right
in
from
the
very
beginning.
A
A
So
we
want
to
make
sure
that
if
anything
goes
wrong,
we
are
able
to
detect
that
quickly
and
remedy
it
automatically,
preferably
without
any
manual
intervention.
This
all
sounds
really
motherhood
and
apple
pie,
but
these
are
critical
things.
They're
not
easy
to
do,
they're
not
easy
to
implement,
and
then
for
telecommunications.
We
want
deterministic
performance,
we
don't
want
to
click
on
something
and
wait
five
seconds
absolutely
what
throughput
that
we
promised
our
customers.
A
We
want
the
latency
the
delay
from
going
one
into
another
end
of
the
network
and
the
timing
to
be
communicated
all
the
way
through.
To
the
end
point
and
there's
this
thing
called
jitter,
which
I
won't
go
into,
but
it's
basically
the
the
timing,
variations
that
happen
in
in
small
intervals
of
time,
and
then
this
is
something
that,
for
the
open
source
community
is
very
important.
How
do
we
make
sure
that
features
persist
from
release
to
release
if
we
fork?
The
software,
for
example,
into
a
proprietary
form?
A
All
sounds
pretty
obvious,
but
these
things
happen
and
then
we
want
seamless
software
updates.
I've
said
this
once
already
no
down
time
and
something
else
to
bear
in
mind
is
in
in
in
the
commercial
world
most
software's
license.
Of
course,
there
are
open
source
licenses
and
they
are
all
they're
all
present
in
many
of
the
software
systems
that
we
deploy.
We
must
make
sure
that
we
understand
what
licenses
are
applicable
and
that
we
actually
pay
for
those
resilience.
I'd
like
to
make
a
few
a
few
remarks
here
about
resilience.
A
First
of
all,
software
running
on
low-cost
standard
hardware
can
be
made
highly
resilient
right.
So
I
think
that
probably
pretty
obvious
for
you
guys,
you
can
deploy
your
kubernetes
cluster
on
one
system
and
and
have
a
backup
in
another
system,
and
if
one
system
goes
down,
you
can
switch
over
to
the
backup
so
and
once
you're
in
the
software
environment.
There's
lots
of
novel
things.
You
can
do
to
enhance
resilience.
A
Now
without
amazon,
web
services,
tetsuya
worked
on
a
system
in
japan
after
the
fukushima
disaster,
where
most
of
the
of
the
mobile
network
in
japan
was
taken
down
due
to
overload,
and
he
designed
a
resilient
system
based
on
nfv,
which
actually
migrated
the
core
network
within
30
minutes
to
to
another
data
center
100
miles
away.
So
we
can
have
a
very,
very
large
geographic
redundancy
using
software.
A
We
all
know
about
the
elimination
of
huawei
from
from
many
countries
systems,
because
they're
chinese,
that's
pretty
controversial.
I
actually
own
the
relationship
with
huawei
when
I
was
at
bt
and
they're
a
good
company,
but
nonetheless
that's
the
perception,
and
so
the
supply
chain
has
been
damaged
to
some
extent
by
perceptions
of
security
vulnerability.
A
There
is
a
shortage
of
people
in
the
telecoms
industry
with
hybrid
skills.
We
need
telecoms
people
that
can
write
software.
We
need
software
people
to
understand
telecoms
and
there's
a
big
hint
about
commun
future
career
opportunities.
For
those
of
you
that
are
thinking
about
your
career,
telecommunications
could
be
a
good
place
if
you've
got
some
really
good
software
skills,
but
you'll
need
to
get
some
telecom
knowledge
lack
of
diversity
of
running
code.
If
everybody's
using
the
same
code
and
it's
hacked
and
there's
a
and
there's
an
attack,
everybody
goes
down.
A
This
is
where
software
diversity
is
really
important
and
that's
one
of
the
advantages
of
proprietary
software
somebody's
code
is
different
from
someone
else's
code.
If
everybody
uses
the
same
open
source
release,
everybody's
code's,
the
same
the
complexity
of
integration,
that's
something
which
I'm
not
going
to
go
into
here.
Maybe
that's
a
a
a
topic
for
another
talk,
but
basically
bringing
all
these
systems
together.
Different
vendors
onto
a
common
platform
is
complex
and
and
there's
a
question
in
the
industry.
A
All
of
that
sounds
all
very
nice
and
there's
a
bunch
of
work
that
we've
done
in
etsy
nfv
that
looks
into
resilience
so
for
all
of
you
that
are
interested
in
reliability
and
resilience.
There's
some
really
good
work.
There
fault
management,
just
a
few
say
a
few
things
about
this,
and
I'm
conscious
of
time
fault,
management's,
really
really
important
for
network
operators
probably
number
one
after
security.
A
We
need
to
be
able
to
deploy
virtual
network
functions
from
different
vendors
onto
a
common
platform
and
thought
assure
them
so
gulf
good
fault
management
requires
ability
to
detect,
filter
and
correlate,
and
what
we
don't
want
to
do
is
overload
the
infrastructure
with
fault
management
telemetry.
So,
yes,
we
can
instrument
everything.
We
can
monitor
everything,
but
what
happens
to
all
of
that
data
when
it
gets
abstracted
up
to
two
operations?
A
So
there's
some
very
interesting
work.
That's
been
going
on
there
to
try
and
figure
out
how
we
can
use
nfv
to
to
assure
networks
and
have
some
novel
techniques
for
analysis
a
little
bit
more
about
resilience.
We
we
know
from
the
hardware
world
that
we
have
this
thing
called
the
bathtub
curve
at
the
beginning
of
the
product's
lifetime.
There
are
hardware
bugs
they
get
ironed
out
and
over
time
the
hardware
starts
to
deteriorate
might
be
the
battery
in
your
phone,
and
we
see
this
kind
of
end
of
life
wear
out.
A
It
may
be
surprising
to
know
that
software
actually
has
some
kinds
of
these
features
when
we're
doing
test
and
debug.
We've
got
a
lot
of
problems
with
bugs
and
the
software
may
crash,
and
that
means
it's
not
very
reliable.
As
we
go
time,
we
should
through
time
we
shoot
updates
we
we
fix
bugs,
but
in
fixing
the
bugs
we
introduce
another
bug.
So
you
have
this
kind
of
curve.
A
Where
you
you
do.
Software
release
the
software
crashes,
because
you
introduce
new
bug,
you
fix
that.
Then
you
do
another
release
and
then
you
get
another
book.
So
you
send
see
this
rather
interesting
curve,
and
so
this
is
one
of
the
things
which
we
we're
concerned
about
in
telecommunications.
We
want
predictability.
A
So
how
do
we
make
sure
that
when
we
do
software
updates,
we
don't
get
an
unreliable
network
and
there
are
some
ways
to
to
to
mitigate
that
and
one
of
the
things
this
thing
called
end
version
programming,
which
is
to
say
that
we
have
a
number
of
software
functions
that
are
normally
identical,
but
only
one
if
one
of
those
is
faulty
for
example,
or
has
been
hacked,
then
we
want
to
make
sure
that
the
others
are
the
ones
that
control
the
in.
Let's
say,
for
example,
it's
controlling
the
aircraft.
A
You
don't
want
your
aircraft
to
go
down
because
there's
a
bug
in
the
aviation
system,
so
a
number
of
different
things
that
can
be
introduced
there.
This
is
my
last
but
one
slide.
The
ecosystems
are
very,
very
complex
and
the
first
thing
that
happens
in
telecommunications
is
that
if
we
want
to
do
something
completely
new
like
for
example
5g,
then
we
get
together
with
a
bunch
of
other
people
in
the
industry,
and
we
have
a
discussion.
What
do
we
want
to
do
and
can
we
converge
all
of
our
requirements?
A
A
A
You
know
all
those
ieee
specs
that
you
know
and
love
they've
done
in
the
metro,
ethernet
forum
and
the
ieee,
so
those
are
really
important,
but
then,
when
it
comes
to
actually
implementation,
how
can
we
accelerate
that-
and
this
is
where
open
source
comes
in
really
handy,
because
open
source
communities
get
interested
in
how
they
can
make
very
quick
progress
on
code,
and
so
we
have
the
potential
for
innovation
in
the
open
source
communities.
I've
got
kubernetes
up
there,
but
also
open
source
can
be
used
to
validate
some
of
the
specs.
A
That's
all
great,
and
I'm
just
gonna
make
a
few
closing
remarks
so
that
we
can
hand
off
to
some
q
a
if
you
guys
have
got
questions
the
pandemic,
we're
all
suffering.
Aren't
we
all
at
home-
and
you
know
the
first.
The
keynote
was
fascinating
to
me
to
to
figure
out
how
you,
how
you
build
communities
and
maintain
communities
while
we're
all
locked
down
in
our
homes
in
in
and
in
our
respective
countries.
A
Anybody
who
can
write
code,
but
not
everybody,
can
write
useful
code
right.
So
what?
What
does
the
code
do?
Is
it
any
good
documents-
and
this
is
something
which
is
an
old
bugbear
for
me
right?
I've
written
a
lot
of
software
in
my
time
and
I've
had
to
try
and
figure
out
how
somebody
else's
software
worked,
and
when
I
opened
up
the
source
code,
there
was
no
comments,
nothing,
no
documentation,
I'd
literally
go
through
line
by
line
and
try
and
figure
out
what
this
software's
doing
and
so
documenting
your
code.
A
This
is
a
big
question,
and
can
you
develop
a
coding
skill
that
is
in
short
supply?
This
is
not
about
learning
python.
This
is
about
what
do
you
do
with
the
code
and
so
some
of
the
examples
where
I
think
we've
got
shortages
and
opportunities
for
those
of
you
that
are
looking
towards
your
future
career
telecom
needs
software
developers
to
understand
networks,
so
telecomma
needs
software
developers
who
understand
networks.
So
what
I
mean
by
understanding
networks,
that's
doing
some
basic
telecommunications
courses.
This
is
about
figuring
out
how
routers
work,
how
do
switches
work?
A
How
do
telecommunications
packet
processors
work,
software
developers
who
understand
security?
Architectures
are
in
very
high
demand
if
you've
got
some
security
background
and
you
understand
security,
architectures
and
you
can.
You
can
teach
yourself
a
lot
of
this
stuff,
but
hopefully
get
some
certifications
as
well.
A
So
that's
something
which
to
bear
in
mind
and
then
big,
rail
and
eyes.
As
raoul
said
very
beginning,
we've
been
we've
been
friends
very
long
time
and
I'm
interested
to
explore,
particularly
after
the
pandemic,
when
it's
becoming
clear
to
the
world
that
people
can
make
useful
contributions
anywhere
in
the
world
and
they're
every
bit
as
good
as
some
as
in
other
places.
A
In
november
we
were
going
to
go
to
madrid,
but
because
we
were
virtual,
we
were
able
to
invite
papers
from
all
over
the
world
without
the
requirement
for
the
person
submitting
the
paper
to
have
to
travel
to
the
conference.
Travel
is
expensive
attending
the
conference
expensive.
So,
for
the
first
time
we
had
a
phd
student
in
panama
who
contributed
to
our
conference
and
that
person
would
almost
certainly
not
have
been
able
to
submit
the
paper
because
they
couldn't
afford
to
go
to
madrid
to
present
it.
So
that's!
That's
it
from
me.
B
And
I'm
gonna
combine
a
couple.
Questions
have
to
do
with
nfv
with
nfp
and
also,
let
me
see
cnf
cloud
native
functions,
and
so
the
question
is
why
why
cnfs
and
on
the
same
question
is
more
about,
like
I'm
curious
to
the
impact
of
concernization
in
terms
of
nfv
performance
access
to
underlying
hardware
and
complex
networks,
architectures.
A
Yeah,
how
can
I
say
this
in
cloud
net
cloud
native
functions?
Cns
are
no
different
from
virtual
network
functions,
virtual
network
functions
or
vns
is
the
industry
standard
term
in
telecoms
that
hasn't
changed.
You
can't
go
into
a
telecoms
environment
and
start
talking
about
cnfs
and
say
that
they're
different
from
vns,
vns
and
cnfs
are
the
same.
A
But
as
we've
gone
through
time,
the
original
nfv
white
paper
in
2012
there
weren't
there
was
no
cloud
that
cloud
native
thing.
If
you're
a
cloud
person
you
lived
in
a
cloud
native
environment,
but
now
that
now
that
you're,
a
telecom
person
you're
living
in
a
network
environment,
we
only
understood
things
like
virtual
machines.
A
We
understood
about
virtual
machines,
and
so
in
our
estimation
you
created
a
virtual
network
function
within
a
virtual
machine,
and
everything
would
be
great,
but
then
over
time
we
realized
that's
very
inefficient,
because
now
we
can
use
cloud
native
techniques,
particularly
things
like
microservices,
and
so
how
do
we
now
implement
network
functions
in
a
cloud?
Environment
was
always
the
intention,
so
the
cloud
guys
took
out
took
our
concept
and
said:
well,
we
don't
want
vns.
We
want
cns.
A
There
is
a
a
a
document
that's
been
published
by
edc
nfv,
which
talks
about
how
vnf
management
can
operate
in
a
cloud
native
environment.
So
this
doesn't
change
the
underlying
concept
of
virtual
network
functions
at
all.
It's
simply
that
how
do
you
efficiently
implement
these
things
in
a
cloud
native
environment?
A
While
the
framework,
the
architectural
framework
of
the
system,
doesn't
change,
it's
simply
that
you're
implementing
it
using
cloud
native
techniques,
so
that's
a
useful
place
to
start
and
if
any
of
you've
got
any
questions
after
you
after
you
read
that
or
or
you
look
into
this
topic
a
bit
more
get
in
touch
with
me-
and
I
can
put
you
in
touch
with
some
of
the
experts
that
are
working
on
virtual
network
functions
in
cloud
native
environments
using
kubernetes.
Kubernetes
is
the
standard
way
to
do
these
things
now.
B
A
A
I
don't
so.
I
I
like
open
ram,
but
I
think
there
are
a
lot
of
challenges
for
open
round
to
deliver
the
full
openness
vision,
because,
first
of
all
there
the
specifications
for
openran
are
are
not
particularly
well
defined,
and
so
that
means
that
interoperability
between
the
components-
the
d?
U
the
r?
U
may
be
problematic
and
operators
must
have
interoperability.
A
The
idea
that
you
can
have
an
open
ecosystem
is
fantastic
and
I'm
very
much
involved
in
open
ecosystems,
but
each
player
in
the
ecosystem
needs
to
be
able
to
see
a
path
to
business,
and
I've
actually
put
a
link.
This
isn't
in
the
version
I
sent
url
at
the
bottom
yeah
we've
got
this
thing
called
the
telecom
ecosystem
group,
which
I'm
a
co-founder.
A
I
think
there
are
challenges
to
deliver
the
full
openness
vision
unless
the
operators,
in
particular
take
very
strong
action
to
define
the
interfaces
and
specifications
so
that
there
is
interoperability
between
all
of
these
entities,
not
just
the
network
layer,
the
control
layers.
I
hope
that
answers
your
question
in
a
very
vague
way.
Great.
A
B
A
A
A
A
A
A
That's
going
to
require
avatar,
3d
environments,
richer
experiences,
and
so
that
is
going
to
put
more
emphasis
on
telecommunications
and
also
on
security
right.
We
know.
Zoom,
for
example,
is
compromised
at
the
very
beginning,
some
very
bad
things
happened
in
zoom.
They
quickly
released,
secured
secured
encryption
and
waiting
rooms
and
things
like
that.
A
A
Having
said
that,
right,
the
infrastructure-
you
know,
the
you
know
the
wireless
network
or
the
or
the
cable
network-
that's
not
going
to
change
overnight
right,
I'm
talking
about
most
of
the
telecoms
execs
are
talking
about.
How
does
telecoms
move
up
the
value
stack
right,
yeah?
So
we're
not
talking
about
open
ram,
we're
talking
about
what
runs
over
open
rand.
That's
not
google,
it's
att,
so
they
want.
They
want
these
rich
experiences
to
be
owned
by
the
telecoms
guys,
not
necessarily
by
by
facebook,
instagram
and
and
tick
tock.
B
A
B
Skills
and
some
of
the
participants
probably
would
find
that
one
also
great
that's
helpful,
that's
great
good
playing
that
was
going
to
be
in
spanish.
Thank
you,
dan
for
sharing
your
knowledge
and
welcome
to
the
community.
Also,
you
need
to
know
that
we
have
the
community
and
you
can
present
there
as
well.
You
know
we
have
meetings
once
once
a
month
and
so
now
now
now
we
know
where
to
find
you.