►
Description
Join us for Kubernetes Forums Seoul, Sydney, 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
Serverless Kubernetes: Container in Cloud Native Way - Wei Zhang, Alibaba
Nowadays more and more users run containers on the cloud. However, running containers on cloud is far from ideal. It still requires significant efforts on cluster setup, configuration, and servers management. A new paradigm "Serverless Container" came out recently, which eliminates the need of managing cluster and servers, aiming to run containers on cloud more easily and agilely. In this topic, we will present the landscape of serverless container ecosystem, and how we build the serverless container service at Alibaba Cloud, which is easy to scale and has a flexible resource model, also provides the native Kubernetes API. Users will learn how to deploy container, ingress, and service discovery in cloud native way. We will also unveil its innovative cloud scale architecture, and talk about how it helps our customers to focus only on the application rather than server management.
To learn more click here: https://sched.co/FuKQ
A
The
some
of
our
practices
in
some
of
those
containers
and
its
progress
and
development
I'm
also
going
to
talk
about
a
demo
and
show
you
a
demo
if
time
allows
I'm
going
to
talk
more
about
questions
and
answers.
Thank
you.
So
much
for
stereo
time
to
attend
the
last
day
of
the
kuba
is
calm
conference,
and
this
room
is
also
very
hard
to
find.
So
thank
you.
So
much
for.
B
A
B
B
B
B
A
B
B
A
The
new
format
for
function
deployment,
application
deployment,
as
we
can
see
here
from
the
chart,
the
container
compared
with
the
VM.
The
innovation
is
like
the
m2
physical
machine,
so
the
technique,
not
a
technological
evolution,
will
not
stop.
The
innovative
technologies
are
not
from
the
scratch
from
0
to
1
or
0
to
2.
It
is
like
a
combination
of
different
innovations.
For
example,
the
docker
is
based
on
a
combination
of
basic
technologies
in
the
newest
community.
They
are
getting
mature,
they
have
based
on
say,
group
and
other
primitives
to
form
some
new
innovative
technologies.
B
A
B
A
Has
a
consistent
deployment
and
execution
has
a
faster
delivery
and
agility
for
surveillance.
The
server
the
users
do
not
need
to
consider
concern
about
server,
so
they
don't
have
to
consider
the
maintenance
and
operation
anymore
surveillance
plus
container
for
IT,
DevOps
or
maintenance.
This
is
actually
a
very
new
innovation.
A
B
A
We
have,
as
your
HC
I
launched
in
2017
at
the
very
beginning.
It
didn't
support,
kinetics
or
virtual
kubernetes.
I
will
talk
about
this
more
and
AWS
pocket
in
space
down
easy
as
orchestration.
It
was
planned
to
launch
in
2018
for
supporting
the
kubernetes
and
a
surveillance
computing,
and
now
we
have
our
CCI
launched
in
February
this
year.
We
also
have
Alibaba
cloud
service
kubernetes
from
early
cloud.
Now
we
are
planning
for
commercialization.
Now
it's
free
for
public
tests.
B
A
Let's
compare
surveillance,
kubernetes
and
the
traditional
combination
of
clusters
for
traditional
kubernetes
cluster
snow
weather
for
the
users.
You
can
deploy
the
classrooms
in
the
couch
in
a
very
faster
way.
So
this
is
a
very
classical
kubernetes
structure
because
they
have
the
mask
and
a
worker
for
the
mask.
They
have
different
types
of
deployment.
For
example,
the
user
can
choose
their
exclusive
master
version,
they
can
follow
up
in
estes.
The
master
is
for
the
user
clusters,
the
user
can
totally
control
the
capacity
and
clusters
and
another
wise
a
trusted.
She
trusts
she
trusteeship
trusteeship.
A
Those
of
them
have
similar
structure.
They
need
to
have
the
work
note.
The
work
no
should
be
a
virtual
machine
or
a
physical
machine
or
bare
metal
server.
The
port
should
be
the
work
nodes,
so
this
is
a
very
classic
pastor
for
epidemics
for
surveillance
kubernetes.
We
can
see
here.
It
also
has
the
container
scheduling
orchestration,
and
this
is
actually
outside
of
the
clusters.
The
users
cannot
see
it
they
don't
have
to
care
about.
Where
is
the
master?
A
B
B
B
A
They
don't
have
to
deploy
the
work
notes
and
their
payment
is
actually
very
different
from
traditional
one
in
your
service
kubernetes,
their
part
is
actually,
for
example,
you
deploy
a
linux.
When
you
started,
you
will
pay
on
the
demand
and
after
you
terminate
the
usage,
you
will
stop
image,
however,
for
the
classic
one
is
different.
Your
payment
is
actually
based
on
your
notes,
starting
from
the
deployment
of
the
cluster.
You
take
ten
work
notes
on
that
moment.
You
start
to
pay
Monique,
no
matter
whether
you
have
deployed
port
or
applications.
The
payment
is
permanent,
cannot.
A
If
you
eliminate
all
the
ports
in
the
classes,
you
then
can
stop
and
payment,
so
the
payment
model
and
the
payment
templates
are
quite
different,
so
the
traditional
way
is
based
on
the
node
for
payment.
The
users
need
to
be
very
cautious
about
the
scale.
The
size
of
the
worker
nodes,
for
example,
for
CAG
or
different
sizes
and
scales
have
different
fees,
but
for
the
service
kubernetes,
you
don't
have
to
choose
the
size
of
the
notes.
You
have
to
choose
one
size
by
the
size
of
the
port
and
there's
something
you
need
to
choose.
A
You
also
need
to
set
the
CPU
in
a
language
for
the
port,
for
example,
the
CPU
can
be
one,
the
language
can
be
2
G
or
the
memory
can
D
to
G,
and
then
the
payment
is
based
on
the
choices
of
the
memory
and
CPU.
Many
users
have
ignored
one
part
in
the
traditional
based
on
that
node
payment
model.
Your
node
capacity
should
be
over
than
the
port
community.
If
your
application
only
needs
to
use
1c
G.
B
A
Don't
have
to
create
a
work
node
based
on
that
you
have
to
create
a
work
node.
More
than
that,
you
need
to
prepare
an
idle
resources
for
preparation.
It
is
for
the
port
and
note
have
the
same
resources
will
cost.
The
port
will
be
overloaded,
so
this
is
a
very
huge
difference,
but
for
service
capacity
you
don't
have
a
redundant
resources,
all
the
resources
or
part
use
you
can
just
give
it
to
them.
You
don't
have
to
give
them
redundant
resources
so
for
different
departments,
it
will
have
different
cost
for
you.
A
When
you
compare
it,
the
users
will
be
very
cautious
about
the
differences
for
the
traditional
community
or
kubernetes.
You
need
to
have
some
idle
resources
that
redundant
others
redundant
resources.
You
never
will
use
it.
For
example,
the
15%
of
the
Castros
actually
used
an
85%
of
the
computed
results.
It
actually
wasted.
B
A
In
this
queue,
Banaras
and
surveillance,
Adam
note
we
also
have
six
types
of
surveillance.
For
example,
we
call
it
service
alone
compared
with
the
kubernetes
and
service
life
is
quite
different
in
a
service.
It
is
a
very
complete
cluster.
The
users
cannot
see
any
workload,
however,
for
those
existing
users
they
already
have
the
capacity
clustered.
They
already
have
a
lot
of
notes,
and
at
this
time
they
want
to
use
the
server
less
flexibility
function.
How
can
they
do
they
don't
need
to
create
a
new
service
cluster?
A
In
theory,
you
can
use
the
same
VPC
for
traditional
and
the
new
one.
The
pod
is
connected
in
another
way.
Well,
in
this
way
for
the
multiple
clusters,
you
need
to
manage
the
multiple
clusters.
However,
if
you
use
service
add-on,
you
can
use
your
own
plus
a
virtual
node,
and
you
don't
have
to
create
a
new
cluster
based
on
your
code
capacity.
You
can
add
a
virtual
node
at
you,
but
it's
flexibility
in
the
scalability,
for
example.
A
Here
we
have
three
notes:
they
have
limited
computing
capacity
if
your
service
is
like
e-commerce
in
a
double
11-day
the
shopping.
First
of
all,
you
need
to
use
more
ports,
you
need
to
scale
out,
and
you
don't
want
to
buy
too
many
nodes
or
the
easiest
or
node.
Then
you
can
use
the
virtual
nodes
to
satisfy
your
sudden
huge
demands.
This
is
actually
a
real
case
in
the
double
11-day.
I
was
shopping.
First
of
all,
just
finishes
the
release
ago.
We
have
some
customers
using
this
kind
of
way
to
add
the
virtual
notes.
A
B
A
A
For
it
here
this
is
virtual
kubernetes.
We
have
a
lot
of
providers,
for
example,
I
saw
baggage
and
other
providers
and
also
have
some
dynamic
called
resources
in
this
one
is
the
ECI
flexible
continual
instance
elastic
container
instance
ECI.
So
from
the
virtual
note,
the
easier
is
actually
the
same
with
the
real
note,
so
you
can
have
a
caste
or
IP
service.
You
can
have
an
easy
iPod,
so
in
the
real
instance,
you
can
also
send
any
information
around
the
virtual
node.
This
is
connected
with
the
real
note.
A
It
is
a
same
well,
overly
speaking
via
the
virtual
nodes.
You
can
actually
provide
some
unlimited
elasticity
and
flexibility
based
on
the
already
existing
notes.
So
now
I'm
going
to
talk
about
her
Kerber
medicines
herbalist
add-ons.
So
here
are
the
features
for
it.
The
first
one
you
don't
need
to
have
a
managing
service.
A
A
Kubernetes
is
very
simple
for
them.
They
don't
want
to
even
see
the
ECI
servers,
so
the
service
surveillance
and
the
kubernetes
can
satisfy
the
amounts
without
service.
They
don't
have
to
do
the
capacity
planning
they
don't
have
to
care
about
node.
They
don't
have
to
come
here
about
the
GPU
or
granularity.
They
don't
have
to
care
about
anything.
They
only
need
to
focus
on
their
own
applications.
We
don't
have
to
purge
any
complicated
management
tools
or
networks,
because
their
service
and
a
kubernetes
created
part
that
one
support
are
created.
B
C
B
C
C
B
C
Number
of
notes
and
the
capacity
of
the
note
they
start
the
upper
limit
of
the
cluster.
Therefore,
if
you
want
to
increase
the
upper
limit
of
the
cluster
computing,
then
you
need
to
increase
the
number
of
notes,
but
in
surveys
scenario,
because
there's
no
load,
so
the
logic
is
very,
very
easy.
You
just
need
you
do
the
deployment.
B
C
B
C
B
B
B
C
B
C
The
port
in
very
fine
granularity,
because
there
are
different
resources.
The
cost
of
all
people
will
be
different.
We
designed
in
this
way
because
we
want
to
improve
the
cost
performance,
because
previously
you
need
to
reserve
up
a
certain
amount
of
computing
resources,
but
in
some
ways
you
can
configure
only
pay
for
the
amount.
C
B
B
B
B
C
B
B
B
C
Viking
will
recent
shoot
to
see
the
changes?
If
you
want
to
establish
the
port,
it
will
call
API
to
create
a
portal
and
in
the
part
of
the
EC
is
there
is
a
logic
entity
and
there's
a
VM
and
they
can
have
a
Viking
agent
in
a
way
and
it's
very
similar
to
accommodate
the
it
can
call
your
port.
Of
course,
there
are
many
containers
and
in
vogue
called
a
container
and
also
have
a
doctor
mode,
and
it's
a
very
lightweight
bit.
C
C
B
C
B
C
C
B
B
C
I,
just
don't
have
to
manage
servers
and
the
advantage
is
electricity,
so
there
are
some
multimedia
processing
or
some
driving
adventure
driven
can
scenarios,
and
sometimes
the
functions
will
have
limitations.
For
example,
it
can
only
run
for
10
minutes
and
it
has
a
special
requirement
for
the
language,
but
in
the
container
you
can
operate
I'll
run
any
application.
C
B
C
B
B
B
B
B
C
Is
a
4G
and
Hg,
and
also
the
by
default,
the
number
of
three
and
the
you
can
change
it
according
to
add
amount.
This
is
a
traditional
cluster
and
they're
also
a
hosted
cluster.
The
difference
is
that
there
only
one
configuration
which
is
a
worker
mode,
there's
a
known
master
mode,
because
master
is
hosted
so
compared.