►
From YouTube: 005 Istio Roadmap Session
Description
Lin Sun and Louis Ryan from the Istio technical oversight committee, lead an update on the development of the project and the roadmap for 2021.
A
A
A
A
The
isro
constituent
committee
was
composed
of
google
and
ibm
in
the
initial
phases.
We
opened
up
the
seats
for
red
hat
and
also
salesforce,
as
the
third
largest
contributing
company
super
excited
to
see.
The
steering
committee
opened
up
four
additional
community
seats
as
part
of
the
election
results
super
happy
to
see
solo.
I
o
is
also
on
the
israel
steering
committee
for
2020.
A
A
A
A
Our
user
loves
secret
discovery
service
to
distribute
key
and
search
without
restart
of
envoy,
and
we
have
enabled
sds
across
both
the
gateway
and
saikai
in
istio
by
default.
Auto
metro
ts
was
also
introduced
to
smooth
that
transition
from
permissive
to
strict
tas.
So
you
don't
need
to
create
a
destination
rule
as
you
transition
to
strict
mutual
tis.
A
A
We
spent
a
lot
of
effort
on
simplify
multi-cluster
configuration,
onboarding
services
from
the
virtual
machine
into
the
mesh
and
also
graduate
those
two
beta
as
well.
We
spend
a
lot
of
time
to
optimize
the
cycle
proxy
latency
and
also
introduce
the
cycle
resource
so
that
user
can
optimize
the
config
distribution
for
the
outside
class.
A
A
A
A
A
B
So
lin's
been
talking
a
bit
about
you
know
what
we've
been
seeing
in
the
project
so
far
the
progress
that
we've
made
on
a
number
of
important
dimensions,
and
also
about
the
feedback
that
we've
gotten
from
users
about
what
what
they
see
as
important
for
us
to
continue
to
work
on,
and
so
this
year
a
big
part
of
our
focus
will
be
on
what's
commonly
known
as
day
two
operations
right.
You
know
where
day
one
is
you're
able
to
install
the
project
and
get
it
up
and
running
day.
B
Two
is
really
you
know:
maintaining
upgrading
using
the
the
product
to
achieve
the
effects
that
you
want
to
achieve
and
being
able
to
generally
maintain
it
as
an
operator-
and
you
know
this
is
an
indication
of
where
we
are
kind
of
the
the
maturity
of
the
project.
You
know
we
now
have
a
lot
of
people
using
these
two
in
production,
so
we
want
to
make
sure
that
they
are
able
to.
B
You
know
continue
to
grow
and
evolve
with
istio
without
worrying
about
you
know,
breaking
when
there's
an
upgrade,
and
so
it's
very
much
in
the
mold
of
move
slowly
and
fix
things,
and
so
you
know
that's
that's,
definitely
a
focus
for
us
and
so
you'll
see
a
kind
of
shift
in
the
project
from
you
know
in
previous
iterations,
where
we
were
either
you
know
paying
down
technical
debt
or
you
know
rapidly,
building
up
the
set
of
features
based
on
what
we
were
hearing
from
user
feedback
about
what
you
know
the
major
gaps
of
the
system
were.
B
We
now
think
we
have
a
lot
of
the
functionality
and
a
very
stable
kind
of
core
infrastructure
that
people
need,
and
so
it's
you
know,
while
we
expect
users
to
keep
asking
for
new
features,
we
expect
the
rate
of
feature
introduction
into
istio
to
be
much
more
measured,
right
and,
and
our
focus
will
be
in
making
sure
that
if
there
are
new
features
coming
in
there,
there
are
clear
migration
paths.
B
There
are
clear,
upgrade
paths
and
and
support
paths
for
those
features
in
the
context
of
everything
else,
that's
being
provided
by
the
solution
and
then
a
major
you
know.
Part
of
this
is
really
you
know,
reducing
the
operational
overhead
right,
so
so
the
people
who
install
maintain
and
upgrade
istio
and
roll
it
out
in
production
either
at
small
companies
or
big
companies
right.
B
You
know,
there's
there's
a
there's
only
so
much
time
that
those
people
have
in
the
day
to
perform
their
function,
and
so
we
want
to
make
sure
you
know
that
they
can
kind
of
keep
keep
doing
all
the
other
jobs
and
not
you
know
worry
so
much
about
how
much
time
they
have
to
go
and
spend
on
you
know
maintaining
or
upgrading
or
debugging
issues
in
sdo
and
that
the
experience
you
know,
particularly
around
upgrade,
which
was
a
big
focus
of
last
year,
has
gotten
a
lot
better.
B
The
stability
of
the
releases
has
gotten
a
lot
better,
and
so
you
know
keep
driving
down.
You
know
the
effort
involved
in
performing
these
functions
and
provide
better,
and
you
know
tooling
that
helps
people.
You
know,
particularly
in
the
case
of
debugging,
assess
issues
in
the
project,
give
them
guidance
on
how
to
resolve
the
issue
right.
So
it's
not
that
there's
a
bug
and
you
have
to
go,
read
a
bunch
of
documentation.
B
It's
that
there's
a
tool
and
a
tool
will
tell
you
what
to
do
right,
so
much
more
directed
guidance
and
all
of
these
things
will
have
a
big
impact
on
the
kind
of
aggregate
experience
of
people
operating
istio
in
production.
So
that's
that's
a
major
part
of
our
focus
and
you
know
we
think
it's
going
to
be.
You
know
really
powerful
for
users,
powerful
the
project
and
and
support
that
existing
user
base,
which
is
is
already
quite
large.
So
you
know
in
the
vein
of
stability
and
maintainability
right.
You
know
we
have.
B
We
put
a
lot
of
work
into
the
upgrade
experience,
but
we
expect
to
do
a
lot
more
with
upgrade
over
the
course
of
2021..
We've
already
formed
a
new
upgraded
working
group
that
is
solely
focused
on
this
issue
and
is
is
driving
the
the
items
that
are
listed
here
below
you
know.
We've
already
introduced
a
new
mechanism
called
revision
based
upgrades
where
we
allow
you
to
have
more
than
one
version
of
istio
installed
at
the
time
and
incrementally
migrate
workloads.
B
You
know
whatever
granularity,
you
feel
comfortable
between
those
revisions
and
the
ability
to
move
forward
and
backward
quickly
right.
So
it
isn't
it's
kind
of
like
the
best
practice
of
you
know:
blue
green
roll
out
in
production,
but
you're
actually
able
to
do
it
with
the
istio
control
plane
and
so
for
people
who
have
really
big
production
systems
right.
B
We
think
that's
very
important
for
people
who
have
a
kind
of
more
limited
budget
or
you
know
who
want
to
be
able
to
constrain
the
amount
of
time
they
invest
in
performing
an
upgrade
or
or
doing
this
you
know,
maybe
maybe
you
can
only
budget
to
do
one
istio
upgrade
a
year,
then
we're
working
on
a
feature
which
would
allow
people
to
upgrade
from
you
know.
B
B
So
you
know
that
you
don't
have
to
go
through
an
incremental
upgrade
where
you
have
to
go
through
each
histo
minor
release
version
to
get
all
the
way
up
to
the
latest
right,
because
that
obviously
represents
a
decent
amount
of
work,
and
so
we
could
just
collapse
all
those
steps
into
what
we
now
term
a
skip
level
upgrade
and
people
who
are
doing
you
know
single
control
plane,
maybe
not
using
the
revision
based
upgrade
process.
You
know
that's
going
to
really
help
them
we're
going
to
introduce
tooling
to
check
your
istio
deployment.
B
You
know
both
before
and
after
the
upgrade
in
particular
checking
it
before
to
highlight
any.
You
know,
issues
that
we
might
see
maybe
you're
using
an
api
that
was
deprecated
a
long
time
ago
or
a
feature,
that's
still
unstable
and
you
know
and
we
track
the
stability
features
on
the
site,
and
I
encourage
people
to
check
into
those
things.
The
tooling
will
provide
guidance
and
saying
look
you're
doing
these
things.
You
know.
B
Maybe
you
want
to
try-
and
you
know
address
these
before
you
run
the
upgrade
and
then
also
after
the
upgrade,
maybe
provide
some
recommendations
as
well
and,
and
you
know,
we
continue
to
invest
in
making
sure
that
our
testing
coverage,
you
know,
captures
more
accurately
what
we're
seeing
from
users
in
production.
We
already
have
a
very
extensive
testing
harness
we
test
upgrades.
B
So
we
talked
a
little
bit
on
the
troubleshooting
side.
You
know
the
pre
and
post
script
create
checks,
but
there
will
be
other
things
that
will
do.
You
know
not
upgrade
or
maintenance
related
to
help.
B
People
with
troubleshooting,
including
you
know,
a
lot
of
investment
in
command
line
tools,
analyzing
usage
of
apis
or
features
we'll
also
look
at
producing
what
are
commonly
turned
events
where
the
control
plane
or
data
plane
can
produce
events
that
go
into
a
system
that
can
be
used
to
to
do
reporting
or
assessment
of
what
we
would
consider
critical
events
in
the
system.
B
You
know,
maybe
there's
there's
an
issue
with
an
api
that,
when
it's
rolled
out
is
causing
traffic
to
be
dropped
because
it
misconfigured
where
you
know
where
service
was
or
what
endpoint
the
tropic
was
supposed
to
be
going
to
api
validation
right.
We
do
a
lot
of
work.
You
know
almost
we're
tracking
the
features
and
the
stability
of
every
feature
captured
in
the
api
and
whether
that
api
is,
you
know
the
crds,
whether
it's
you
know
command
line
flags
things
in
config
maps,
environment
variables.
B
We
treat
all
of
those
things
as
api
and
so
we'll
track
the
usage
of
those
and
the
status
of
those
and
we'll
use
that
to
provide
reports
about
you
know
you
know,
are
you
using
you
know
if
you
want
to
be
able
to
answer
our
question?
Am
I
using
a
development
or
alpha
feature
in
production,
because
maybe
you
have
a
policy
about
wanting
to
use
beta
features
or
above
then
you
know
you'd
be
able
to
get
that
kind
of
information,
and
then
you
know
integrating.
B
Okay,
so
let's
talk
a
little
bit
about
the
feature
process,
so
we
we
did
a
lot
of
work
over
the
last
year
to.
B
Precisely
define
a
process
for
which
features
could
be,
you
know
brought
in
as
development
or
kind
of
early
testing
features
graduating
through
the
process
and
what
quality
bars
they
had
to
hit
to
progress
right
and
so
there's
detailed
checklist
about
these.
They
come
in
from
experimental,
going
all
the
way
up
until
stable
and
we
track
the
status
of
all
these
things
and
we
publish
the
status
of
all
these
things
on
the
site,
but
we
do
it
not
just
for
the
crds.
B
We
do
it
for
a
variety
of
other
things,
including
annotations
or
labels,
and
resources
for
things
that
are
not
sdcrds
and,
in
fact
many
other
features,
so
that
process
has
helped
us
a
lot.
So
now
you
know
you'll
see
that
you
know
if
a
feature
is
beta
or
stable
right
it.
It
has
all
the
things
you
would
expect
for
features
that
are
progressed
to
that
level,
whether
it's
for
documentation,
end-to-end,
testing
stability,
testing,
integration,
testing,
upgrade
testing
right.
B
B
B
Our
ability
to
provide
stable
extension
mechanisms
are
really
what's
going
to
drive
the
kind
of
last
mile
ability
for
people
to
adopt
istio
in
environments
right
where
you
know
if,
if
you're
a
a
company-
and
you
have
some
proprietary
policy
or
something
like
that,
you
know
it's
very
unlikely
that
istio
is
going
to
be
able
to
have
an
api
that
has
exact
or
high
fidelity
match
to
what
your
policy
is.
B
So
what
we've
done
is
try
to
enable
extensions
at
key
points
in
the
configuration
and
data
flows,
so
that,
if
you
want
to
be
able
to
integrate
into
a
policy
system
like
the
one
I
just
described
that
you
can
do
that
right
and
so
then
you
get
all
the
value
of
sdo
and
you
get
the
policy
or
whatever
other
value.
Add
that's
proprietary
to
your
organization,
integrated
into
that.
B
So
you
know:
we've
been
working
on
webassembly
for
about
two
years
now.
You
know,
we've
done
a
lot
of
work
to
get
webassembly
upstream
done
to
envoy,
and
that
happened.
In
late
last
year
we've
been
working
with
the
community
on
sdks
we've
been
working
on
improving
the
ability
to
deploy
and
manage
deployment
of
webassembly
extensions,
and
so
now
we're
more
focused
on
you
know:
stabilizing
those
kind
of
distribution
and
developer
experience
aspects
and
bringing
the
totality
that
feature
set
to
beta.
B
We've
also
tried
to
normalize
or
pick
you
know
key
extension
points
in
the
data
flow
for
which
there
are
already
extensions
built
into
envoy
and
give
people
a
more
stable
api
experience
for
leveraging
that,
then
you
know
the
somewhat
break
glass,
only
filter
api
that
we've
had
so
far.
So
recent
work
has
made
the
envoy
xd
filter
available
in
a
much
more
maintainable
way
than
it
was
historically,
and
this
is
actually
one
of
the
more
kind
of
useful
policy
extension
points
that
we
see.
People
interested
in
so
stabilizing
and
kind
of
normalizing.
B
But
when
you
go
into
either
you
know
big
cloud
environments
or
you
know
large
on-premise
deployments.
You
know
it's
very
often
that
there's
already
an
incumbent,
pki
rca
system
in
place.
You
know
citadel,
can
work
with
that.
You
know
at
google,
where
I
work,
we
have
a
managed
ca
solution,
but
other
big
cloud
providers
also
have
their
own
kind
of
ca
systems,
and
also
we
see
common.
B
You
know
big
industry
players
in
on-premise
solutions
like
venifier
things
like
that,
so
making
sure
that
you
know
we
can
integrate
into
those
ecosystems
and
work
with
you
know.
People's
requirements
around
pki
are
very
important.
B
You
know
which,
as
a
user,
you
should
care
about
we're,
also
doing
a
bunch
of
work
and
have
been
doing
a
bunch
of
work
internally
to
make
istio
as
a
project
itself
more
efficient,
to
make
it
easier
to
engage
with
the
product
to
make
the
ability
to
contribute
easier
right.
B
B
You
know,
we've
done
a
lot
of
work
in
the
testing
environments,
to
stabilize
them,
and
so
flakes
would
you
know
flakes
are
kind
of
a
major
tax
on
the
ability
to
you
know,
get
your
pr
through
a
system
and
we've
also
provided
tooling
to
help
people
understand
the
coverage
of
what
they're
doing,
but
from
a
featuring
a
coding
standpoint,
I
talked
a
little
bit
about
the
the
kind
of
feature
promotion
process
having
process
helps
actually
accelerate
promotion
of
features,
and
so
you
know
the
experience
we've
had
with
the
formalization
of
the
process
last
year
really
helped
the
developers
in
the
project.
B
We've
made
a
wide
variety
of
other
improvements,
and
these
have
actually
paid
real
benefits,
because
we
see
developer
satisfaction
going
up,
particularly
elimination
of
test
leaks,
has
made.
You
know
the
kind
of
regular
developers
in
istio
much
much
happier.
B
One
last
thing
that
we've
had
a
particular
focus
on
is
to
make
sure
that
anytime
we
produce
documents
that
is
recommending
to
a
user
to
perform
a
specific
set
of
tasks.
Whether
it's
install
upgrade
setting
up
multi-cluster
or
you
know,
getting
a
vm
working-
that
we
have
documentation
that
is
tested
right
so
that
the
documentation
is
an
accurate
representation
of
what
the
system
is
actually
capable
of
doing.
B
So
those
are
the
kind
of
core
focus
areas
like
like
any
other
big
project
right.
There
are
also
other
areas
that
are
not
necessarily
directly
correlated
to
the
primary
focus,
but
are
still
important
for
us
to
be
dealing
with
right,
because
maybe
there
are
things
outside
of
our
control
or
existing
ecosystem.
B
So
one
of
these
is
the
fact
that
kubernetes
itself
is
putting
a
lot
of
effort
into
having
better
networking
apis,
and
there
are
two
two
major
developments
going
on
in
kubernetes
that
it
you
know
it
would
be
reckless
of
us
not
to
be
paying
attention
to
and
to
be
helping
to
support.
Frankly,
so
one
is
the
kubernetes
gateway
apis.
B
So
this
is
a
new
set
of
apis
to
replace
ingress
in
kubernetes
and
in
a
lot
of
ways
these
apis
look
a
lot
like
the
istio
api
is
like.
I
think,
it's
fair
to
say
that
kubernetes
has
learned
some
things
from
istio
about
how
to
model
the
topology
of
networking,
how
to
model
routing
for
protocols
and
how
to
kind
of
compose
those
things
to
build
a
useful
system.
B
So
several
members
of
these
two
community
being
active
in
that
that
working
group
myself
included
and
then
you
have
the
kubernetes
multi-cluster
efforts
where
they've
you
know
done
some
things
to
enable
you
know:
traffic
to
span
multiple
clusters
with
the
uniform
mechanisms
around
transport
and
service
discovery,
and
and
so
that's
something
that
we're
also
paying
close
attention
to.
B
Also,
the
other
big
thing
in
istio,
ray's
envoy,
right
envoy,
is
our
common
network
substrate,
and
so
we
work
very
closely
with
the
envoy
community,
and
you
know
we
must
keep
pace
with
changes
in
envoy
right,
because
envoy
itself
doesn't
sit
still,
and
there
are
innovations
coming
in
envoy
that
are
beneficial
to
istio,
and
so
as
and
when
those
innovations
land
you
know
we
we
want
issue
to
be
able
to
take
advantage
of
them
to
deliver
a
better
experience,
two
key
ones
here:
delta
xds,
which
is
the
ability
to
ship
smaller
configuration
sets
to
the
sidecars
and
to
gateways
to
be
more
responsive
when
configurations
change,
because
you
don't
have
to
do
large
state
of
the
world
shifts
and
also
to
you
know
in
some
cases,
even
enable
on-demand
loading
right.
B
A
good
example.
Here's
virtual
hosts,
if
you
know
you
receive
traffic
for
a
host
and
the
route
isn't
loaded
yet
into
envoy,
you
could
go
and
ask
the
control
plane
to
give
you
the
route
and
so
for
really
large
messages.
These
kinds
of
things
can
be
important.
B
Another
thing
that's
coming
is
support
for
http
tunneling,
which
hopefully
would
give
us
the
ability
to
have
a
a
a
better
kind
of
networking
abstraction
in
istio
and
and
fixes
or
address
some
things
that
we've
had
to
work
around
with
how
we
do
tunneling
with
mtls.
Today,.
B
And
then
you
know,
there's
some
other
specific
things
use
cases
that
we
hear
about
being
very
important
staple
set
integration.
You
know
there
are
lots
of
pretty
widely
used.
You
know,
particularly
data
management
workflows
that
stateful
sets
would
be
quite
important
and
we
continue
to
invest
in
our
website.