►
Description
Don’t miss out! Join us at our next event: KubeCon + CloudNativeCon Europe 2022 in Valencia, Spain from May 17-20. Learn more at https://kubecon.io The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects.
CNCF TAG-Runtime: What's Next in Cloud Native Workloads? - Ricardo Aravena, Rakuten
What is the CNCF TAG-Runtime? How do we identify projects for CNCF admission? Where do we see the future of runtimes in the cloud native ecosystem? The session will cover: 1) Overview of the TAG-Runtime, how to join, and how to get involved. 2) Update of the related projects that have presented in our meetings. 3) Existing and potential working groups in the TAG and how to get involved. 4) How the TAG provides advise to the CNCF TOC. 5) Future trends for cloud native technologies in the TAG scope such as containers, VMs, WebAssembly and MLOPs.
A
How's,
everyone
doing,
I
hope,
everybody's
having
a
great
kubecon.
I'm
excited
to
be
here
to
talk
about
cncf
tag,
runtime
and
some
of
the
work
that
we've
been
doing
so
a
little
bit
about
our
charters.
We
are
here
primarily
to
help
the
adoption
of
different
types
of
workloads
and
whether
these
are
like
batch,
sensitive
workloads
or
late
latency,
sensitive
workloads,
but
all
of
them
in
the
context
of
cloud
native
environments.
A
A
A
Yeah,
some
of
the
things
that
we
do
primarily
three
things,
so
we
do
outreach
to
projects.
For
example,
we
go
to
github
repositories
or
we
see
what's
coming
into
the
cncf,
and
we
try
to
see
if
these
projects
are
interested
in
presenting
in
our
meetings
trying
to
engage
them
and
sometimes
what
we
do
is
like.
We.
We
talk
to
some
toc
members
and
see
if
the
some
of
these
projects
are
good
fits
for
like
the
cncf.
A
We
also
support
existing
projects,
navigate
the
whole
ecosystem
of
the
cncf
and
also
the
the
different
tags
and
the
cncfs
there's
there's
different
scopes
like
observability
there's
to
act,
delivery,
there's
the
tag
storage
so
like
the
different
different
tags
and
how
how
these
projects
fit
into
these
tags.
And
finally,
we
go
out
and
educate.
For
example,
this
talk
is
educate
the
community
on
what
the
tag
is
doing,
what
some
of
the
projects
are
about
and
how
people
can
get
involved.
A
A
So
we
have
these
different
scope
areas,
so
the
general
workload
or
orchestration,
and
then
you
know,
have
volcano
and
kubernetes
fit
in
there
and
we
have
the
vms
and
runtimes
so
cryo
container
d,
the
web
assembly
runtimes.
We
have
the
container
image
registries
like
quay
and
harbor,
and
other
projects
like
rootless
containers
are
there.
A
Another
part
of
the
scope
is
the
special
purpose
operating
systems
and
that's
operating
systems
meant
to
run
something
unique
and
in
this
case
it's
just
to
run
containers
so
like
flat
car.
It's
an
operating
system
that
allows
you
to
be
there
just
to
run
containers
and
then
there's
the
ai
edge
mlab
space
and
some
projects
there
include,
like
super
edge,
cube
edge,
there's
things
like
keyflow
and
ml
flow,
so
all
in
you
know,
scope
of
machine
learning
and
finally,
we
have
a
working
groups
and
we're
open
to
expanding
this.
A
A
But
essentially
what
they
want
to
do
here
is
have
a
model
where
you
have
these
capabilities
and
actors,
and
these
capabilities
and
actors
are
web
assembly
modules
and
the
capabilities
can
be
things
like
allow
something
to
connect
to
a
different
system.
For
example,
like
a
redis
data
database
or
like
a
mysql
database,
and
and
do
that
with
web
assembly
and
then
the
actors,
it's
a
model
that
makes
some
action
on
those
capabilities.
A
For
example,
if
it's
a
database
insert
some
data
into
the
database,
if
you're
interacting
for
for
with
a
device
at
the
edge,
you
would
actually
be
you
know,
reading
some
data
from
that
device.
So
those
are
some
other
examples.
So
a
lot
of
interesting
things
happening
here
so
exciting
to
see
what
happens
next
with
this
project.
A
So
what's
a
match
is
another
project
that
got
involved,
and
this
is
primarily
a
runtime
for
web
assembly.
So
if
you
build
your
webassembly
binary,
you
can
use
this
to
just
run
it
instead,
and
it
provides
some
capabilities
like
sandboxing
and
they're
trying
to
comply
with
the
the
wazi
specification
from
the
bytecode
alliance
and
use
cases,
are
you
know,
running
webassembly
at
edge
devices
or
iot
type
of
applications?
A
So
that's
in
native
and
quark
is
another
project.
It's
it's
a
another
runtime
in
oci
compliant,
it's
a
pretty
early
project,
but
essentially
what
they're
trying
to
do
is
you
have
this
runtimes
that
are
based
on
virtual
machines
like
kata
containers,
and
you
have
things
like
firecracker
from
amazon
and
basically
they're,
trying
to
trim
that
layer
down
the
vm
has
some
performance
implications.
We've
got.
A
We
have
to
bring
up
vm
layer,
so
it's
just
that
extra
layer
and
what
they
want
to
do
here
is
a
hypervisor
at
the
bottom,
from
scratch
and
in
called
q
visor,
and
that
basically
is
written
in
rust
and
then
on
top
of
that
you
have
a
custom
kernel.
So
the
main
purpose
of
this
is
having
something
higher
performance.
A
This
project
is
pretty
early
still
not
in
the
cncf,
but
you
know
still
in
progress,
so
another
initiative
that
we
got
the
folks
involved
with
is
qos
for
container
run
times.
So
right
now
you
have
things
like
the
cpu
scheduler
in
kubernetes,
where
you
can
specify
like
pin
a
cpu.
If
you
wanna,
you
know
the
full.
A
A
A
So
rootless
containers
is
another
project
from
the
entity.
Folks
and
container
d,
and
essentially
this
allows
you
to
run
a
container
as
a
root
user.
So
the
container
actually
thinks
it's
rude,
so
it
has
the
access
to
things
like
slash,
proc,
slash,
sales,
so
main
components
of
the
system,
so
you're,
fooling
the
container
that
that
it's
thinking
it's
rude,
but
essentially
on
the
host
you're
running
as
a
different
user.
So
the
use
cases
is
like
you
want
to
prevent.
A
A
Trial
is
another
project
that
it's
a
container
registry
but
they're
targeting
something
higher
performance
than
you
know
like
harbor
or
quay,
which
are
some
of
the
container
registries.
The
one
of
the
things
that
they're
doing
is
just
they're
writing
the
whole
thing
in
rust,
so
they're
what
they're
thinking
like?
Maybe
this
is
a
lower
memory,
footprint
and
they're
also
thinking
about
creating
a
p2p
mechanism,
so
you
distribute
your
container
images
across
africa
kubernetes
clusters
in
a
more
efficient
way.
A
A
So
this
actually
takes
the
hpa.
The
traditional
hpa
in
kubernetes
to
the
next
level-
and
this
project
is
currently
in
incubation,
que,
armata
or
carmeta
is
a
project
that
allows
you
to
manage
several
kubernetes
clusters
across
multiple
clouds
or
or
if
you
have
a
private
cloud
or
if
you
have
edge
clusters.
A
So,
for
example,
you
you
for
processing
a
lot
of
data.
You
need
to
reserve
like
a
hundred
cores
or
a
lot
of
memory
across
multiple
nodes,
so
this
project
allows
you
to
do
that,
so
it
integrates
with
things
like
tensorflow
spark
pytorch.
Some
others
are
really
popular.
Big
data
and
machine
learning
frameworks.
A
A
They
know
that
this
csv
or
csp
owns
the
infrastructure
underneath,
but
they
don't
want
the
the
cloud
provider
to
know
what's
being
run
there.
So
that's
why
it's
called
confidential.
So
there's
things
like
encryption
at
the
container
image
level.
There's
encryption
at
the
memory
level,
so
a
lot
of
different
layers
of
security
and
providing
that
that
confidentiality
in
whatever
you're
running.
A
So
they
created
different
components
out
of
the
main
components
in
kubernetes.
So
you
have
the
k3s
server,
which
includes
all
the
components
in
the
control
plane
and
it
also
instead
of
using
something
like
xcd.
It
uses
a
sql
lite
database
or
you
can
connect
to
something
like
mysql
and
then
you
have
a
k3s
agent
that
has
other
kubernetes
components
that
you
typically
have
on
a
kubernetes
node
in
you
know
like
the
cubelet,
the
networking
and
it
also
uses
container
d.
A
So
instead
of
just
using
your
typical
open
stack
to
manage
your
virtual
machines
or
using
something
like
aws
outposts
or
something
like
google
analyst
that
allows
you
to
manage
the
your
own
data
center.
You
can
use
kubernetes
for
that.
So
you
you
bring
up
your
your
vms
with
the
kubernetes
control
plane
and
you
know
it
interacts
with
a
cubelet.
A
A
So
we
looked
at
some
of
the
other
webassembly
projects
like
wasm
edge,
but
those
actually
not
necessarily
run
on
top
of
kubernetes.
So
this
makes
it
possible,
like
you,
you
build
your
web
assembly
module
and
then
you
can
run
it
on
a
kubernetes
node.
It
can
be
located
anywhere
like
in
the
edge
or
could
be
located
in
your
in
your
own
data
center
or
cloud
provider.
A
So
in
the
scope
of
special
purpose
operating
systems,
we
had
this
project
called
vortail
present
and
essentially
this
is
an
operating
system
defined
by
a
tumble
file.
So
you
can
create
your
own.
You
know
operating
system
for
your
specific
workload,
there's
no
ssh
login
or
there's
no
shell,
so
it's
very
custom
made.
They
also
have
a
library
of
different
operating
systems
and
a
similar
way.
You
have
like
a
docker
registry,
but
in
in
this
case
you
have
a
library
of
micro,
vms,
with
operating
systems.
A
So
now
for
the
machine,
learning
and
edge
and
ai
space,
so
we
had
the
tfx
project
come
in
and
present,
and
this
is
machine
learning,
end-to-end
management.
So
you
create
your
machine
learning
model.
You
manipulate
your
data
and
you
try
it
and
in
the
end,
you
actually
send
it
over
to
some
place
to
serve
that
model
to
serve
it
as
an
inference
model.
A
A
In
this
case
it
allows
you
to
track
your
models.
You
can
have
different
versions
of
machine
learning
models.
You
can
do
your
ci
cd,
so
maybe
you
have
a
model
version,
one
0.1
and
then
for
some
reason
that
actually
doesn't
work.
Then
you
can
revert
to
like
you
know,
zero
point
x
or
zero
point,
or
you
have
zero
point
two
and
then
you
wanna
refer
to
0.1
right.
So
you
can
manage
all
that
machine
learning
life
cycle
and
take
your
you
can
take
that
to
production
now
for
cube.
A
Dl,
which
is
another
project,
is
also
is
very
similar
to
amber
flow
and
tfx.
But
the
difference
here
is
that
you
can
run
that
on
top
of
kubernetes,
so
you
can
tweak
that
model
within
your
cluster
and
again,
you
can
make
something
optimal
from
your
machine
learning
model
and,
at
the
end
serve
that
in
a
production
environment
using
a
kubernetes
service,
a
typical
kubernetes
server
service.
This
project
is
currently
in
the
cncf
sandbox.
A
Akri
is
another
project
that
is
also
looking
at
the
edge
computing
space
and
essentially,
what
they're
trying
to
do
or
what
they're
working
on
is,
with
kubernetes
automatically
detect
devices
at
the
edge.
So
typical
use
cases
are
like
sensors,
like
temperature,
sensors
cameras
or
other
type
of
device
that
you
may
want
to
put
in
in
the
edge,
and
some
of
these
devices
can
come
and
go
depending
on.
A
You
know,
like
maybe
there's
something
breaks
like
a
cable
somebody
trips
over
a
cable
and
this
project
will
automatically
detect
that,
and
it
will
also
allow
you
to
have
backup
devices
so
like.
If
one
of
the
devices
fails,
then
it
automatically
gets
added
to
your
pool
devices
or,
if
somebody's
out
in
the
field
they
can
plug
in
the
device,
and
they
don't
need
to
do
any
anything
that
just
it
just
auto
configures.
A
A
At
the
edge
and
using
kubernetes
so
very
similar,
so
I
mean
this
is
something
that
the
cncf
is
trying
to
figure
out.
Some
of
these
projects
have
a
lot
of
overlap,
and
some
organizations
may
may
feel
that
they
want
to
use
a
specific
project
for
certain
reasons.
Maybe
they
the
configuration,
the
source
code
or
so
different
things
so
that
the
cncf
is
working
on
on
coming
up
with
like
ways
of
end
users
to
navigate
this
ecosystem.
So
you
can
see.
A
Super
edge
is
very
similar
to
something
like
you
edge
and
it's
currently
in
the
cncf
sandbox
then
open
yurt
is
another
project
that
is
also
again
very
similar
to
cube
edge
and
and
super
edge
and
yeah.
So
you
have
your
edge
component
and
then
you
have
a
centralized
component
that
can
run
in
the
cloud
or
your
own
data
center
and
just
to
manage
your
workloads.
A
So
we
did
have
some
other
projects
present
in
the
past
and
got
involved.
So
these
are
some
examples
like
we
talked
about
cube
edge,
quay
talos,
which
is
special
purpose
operating
system
for
running
containers,
so
lots
of
different
projects
so
encourage
you
to
take
a
look
if
you're
in
in
the
space-
and
we
also
have
some
upcoming
projects
in
clavera
containers.
A
A
So
if
you
have
any,
if
you
have
a
project
in
the
space
or
if
you
know
of
anything,
you
know
just
reach
out
and
and
let
us
know-
and
and
we
won
these
projects
in
the
space
to
get
involved
so
for
the
container
orchestrated
work
group.
So
we
have
this
single
work
group,
but
again,
like
I
said
we
we're
trying
to
expand
this
and
create
work
groups
for
other
areas.
A
One
example
could
be
like
something
at
the
edge
something
another
work
group
that
could
be
create,
or
working
group
could
be
created,
would
be
something
like
related
to
machine
learning,
so
lots
of
lots
of
opportunities
to
to
work
together
and
try
to
come
up
with
standards,
so
what's
happening
in
a
lot
of
these
open
source
projects.
Is
that
you
see
folks
in
different
projects
working
on
different
standards
or
ways
to
identify
something,
whether
in
machine
learning
or
whether
a
workload
in
the
edge?
A
So
for
the
container
orchestrated
working
group,
what
they're
trying
to
do
is
come
up
with
a
way
to
define
container
devices
or
device
specifications
for
regular
containers,
and
you
know-
and
the
challenge
here
is
that
you-
you
have
a
lot
of
different
disparate
definitions,
it's
very
fragmented,
so
the
team
is
working
to
try
to
get
that
together
and
again.
The
use
cases
are
are
run
times,
specific
things
like
deep
learning.
You
know
5g
and
in
targeting
specific
devices
like
gpus
or
or
specific
processors,.