►
Description
Talking during Kubernetes Community Day on January 30, 2021.
https://community.cncf.io/san-salvador/
A
B
Good
morning.,
the
invitation
to
this
event
from
cv
hernández
here
in
el
salvador.
Today's
title
is
more
in
learning.
Pipelines
concludes
closing.
It
will
be
a
small
demo
of
the
potential
that
all
this
technology
has
a
little
bit
about
me.
I
am
a
systems
engineer.
I
am
from
Guatemala
I
teach
at
various
universities.
They
end
up
in
guatemala.
C
B
B
A
B
B
Machine
learning
models
so,
to
start
a
little
here
like
a
little
joke
here
and
talk
about
the
match
in
learning
that
is
taking
over
us.
It
goes
there
and
the
machines
are
going
to
eliminate
us
so
right
now
we
are
in
an
era
of
a
lot
of
artificial
intelligence
and-
and
we
are
using
a
lot
of
what
is
continuous.
B
B
B
So
what
we
do
is
that
we
package
the
software
and
it
allows
us
to
run
it
anywhere
once
we
have
this
container
engine
currently,,
because
there
is
a
transition
in
what
is
agreed
Before,
to
be
good,
in
fact,.
The
new
version
will
no
longer
support
docker,,
but
rather
continue,,
but
in
the
case
of
docker,
at
a
low
level,
use
continue
and
to
do
this
abstraction
of
libraries
and
the
creation
part,
they
already
contain,.
So
indirectly
we
are
too.
using
only
that
now
he
is
going
to
use
it
directly.
B
Like
this
congress
on
q
berna
later
gubern,
then
let's
think
that
if
we
have
an
application
that
uses
various
contents,
It
is
designed
under
a
microservices
architecture,,
yes
or
no,.
So
when
it
already
sounds
like
a
large
amount
of
content,
well,,
we
already
use
what
is
called
an
orchestrator,.
So
what
this
orchestrator
will
allow
us
is
to
be
able
to
manage
that
amount
of
containers
and
manage
the
network
being
sometimes,
and
all
that
then
werner
comes
to
be
this
solution
when
we
no
longer
only
have
applications
of
a
single
contains.
C
B
Massively,
it
is
obviously
governed
by
a
lot
of
the
box
philosophy
and
they
are
well
designed
to
implement
it
or,
let's
say,
implement
processes.
I
was
saying,
and
civil
and
well,
it
uses
declarative
object,
configuration,
that
is,
we
have
a
file
where
we
declare,
for
example,
large
amounts
of
objects,.
We
have
rulers
that
control,
that
it
is
exactly
that
amount
of
objects,
and
if
one
is
deleted,,
then
replace
that
to
have
that
state
that
we
We
define
or
declare
and
that
it.
B
Object
that
we
want
the
horns,
something
that
is
very
important
when
we
are
in
companies.
We
do
technology
concept,
tests
a
piercing
or
we
can
call
it.
Then
these
concept
tests
allow
us
to
play
with
technology
and
determine
if
it
is.
Let's
say
this
technology
that
we
are
testing
is
adequate
to
provide
a
solution
to
a
problem
that
we
are
facing.
Then
he
to
have
this
context.
The.
B
Because
it
is
something
that
is
beginning
to
be
developed,
laws
comply
with
taxes.
He
talks
about
that.
We
have
these
the
computation
in
the
cloud
exactly
but
close
to
the
origin
of
where
the
data
is
so,
for
example,
we
can
right
now
how
visible
this
is,
but
obviously
our
phones
are
like
good
computers
to
a
microcomputer
logically,
and
we
can
develop
the
information
processing
no
longer
in
the
cloud,
but
rather
in
our
local
device.
B
So
right
now
the
law
is
to
comply
with
this
as
revolutionizing
and
many
of
these
microprocessor
architectures
were
being
used
at
rm
because
they
are
cheap,,
have
good
performance
and
allow
lower
costs.
So
this
is
computing
has
to
supply
a
lot
of
the
The
need
to
reduce
processing
costs
for
certain
types
of
specific
cases
is
in
progress,.
So
there
is
a
migration
to
this
use
of
deysi
cori
platforms
in
applications
that
we
may
have
of
next
computing,.
It
could
be
machine
learning,,
which
is
what
we
are
going
to
talk,
about.
B
of
data
in
memory
games
and
any
type
of
load
or
processing
that
we
want
to
migrate
from
an
intel
or
x86
32
64,
bit
architecture
to
an
air
microprocessor.
So
this
is
where
one
of
the
science
projects
of
the
richter
company
well
comes
to
charge.
Well,
you
include,
well,.
It
is
a
ruler,.
It
is
light,
small,
packaged
in
a
simple
binary
that
allows
it
to
run
on
rm
architectures,
and
that
is
already
with
it,.
It
already
allows
us
to
be
close
to
internet
testing,.
B
If
the
lights
computing,,
then
it
is
already
designed
to
run
in
production,,
then
how
You
can
see
that
the
architecture
of
rulers,
well,
it
has
a
server,.
It
has
an
agent,
the
people
who
have
packaged
everything
that
were
individual
binaries
in
one
and
in
the
part
of
the
server
also,.
Then
they
package
everything
in
a
100-
megabyte
binary,
and,
well,.
They
already
have
their
engine.
The
container
line
in
that
part
of
the
features
that
it
includes
later
is
a
single
binary,
as
it
mentions
it
can
support
vacancies.
B
Default,
it
is
culé,
live
it
is
it
and,
of
course,
here
it
is
handling
for
more
des
packages.
It
also
includes
an
integrated
English
controller,
which
is
graphical.
It
has
a
is,
as
a
component
of
the
jamb
that
I
have
been
able
to
create
how
to
install
my
libraries
as
and
define
it
in
a
file
that
integration
is
also
interesting.
B
B
Lot,,
so
it
is
something
to
observe,,
as
I
was
saying,
that
it
instructs
as
it
is
a
platform
for
and
that
it
is
prepared,
for
there
is
a
guy,
and
if
we
comply
with,
we
are
talking
about
microprocessors
that
do
not
have
so
much
power,.
However,
I
can
use
it
for
normal
cases
of
deploying
a
cluster
and
being
able
to
save
costs
either
for
a
test
for
something
in
production.
In
a
simple
end
of
easy
installation
and
maintenance
is
something
as
simple
as
I
can
create
a
cluster
using
raspberry
c
in
my
house.
B
B
That
you
want
to
test
or
show,
I,
don't
know
at
work,
a
proof
of
concept
of
the
option,.
So
here
comes
one
of
the
problems,,
maybe
not
a
problem,,
but
things
to
take
into
account
to
use
this
type
of
technology
is,
for
example,
a
case
of
use
is
a
woman
is
that,
for
example,.
There
are
devices
that
use
rm,
I,
don't
know,.
Let's
say
it
could
be
a
weather
station
or
something
like
that,
so
my
applications
that
are
going
to
run
have
to
be
in
rem
architecture.
C
B
B
Information
processing
and
are
starting
to
take
this
into
account
to
migrate
their
computing
and
processing
to
the
age,
contour
invap
to
reduce
costs.
So,
although
the
easy
computing
concept
is
talking
about
the
two
small
capacity
devices
doing
the
processing
where
the
data
is
and
not
the
new
one,
it
has
become
very
popular
that
they
say
that
there
are
already
to
be
fulfilled
in
the
cloud
and
some
of
us
are
growing
in.
B
C
B
C
A
B
Lower
costs
in
processing,
obviously
the
processing
does
it
with
less
energy
consumption
and
allows
certain
advantages.
You
get
a
better
price,
just
as
you
get
your
processing
of
what
you
are
trying
to
do
in
your
situation
in
the
company
at
work
So.
Here
we
can
see
an
example
that
is
very
simple,
after
it
is
installed,.
Basically
you
go
to
the
titles
site,.
Let's
say,,
it
recommends
a
minimum
4
GB
node,,
so
they
swipe.
B
In
fact,.
This
is
going
to
be
placed
in
a
repository
that
will
to
show
at
the
end.
So
with
this
you
can
run
all
the
installation
that
trias
basically
is.
This
installation
is
like
more
customized.
Let's
say
to
be
able
to
run
in
swimming
pool.
You
want
in
google
cloud
or
in
blue
with
a
node
of
at
least
4
gigabytes.
You
can
try
this
with
this
command
and
it
installs.
B
So.
It
is
not
like
a
methodology,
something
like
that,
more
like
a
culture,,
but
best
practices
have
emerged,.
People
have
written,
have
already
practiced,,
and
how
does
this
become
through
the
phil
methodology,
well,?
Something
is
very
attractive
to
be
implemented
by
a
company
So.
The
basic
development
cycle
of
the
build,
deploy,
test,,
release
that
comes
with
the
box,,
the
siding
parts
and
all
that,.
The
technology
is
Claudia
and
well,.
B
We
come
with
the
concept
of
pipeline
of
potera
pipeline
that
talks
about
there
being
information,
processing
and
elements
connected
in
series
that
it
may
be
that
the
output
of
one
becomes
the
input
of
information
of
another
to
process
and
do
something,
and
that
is
what
they
can
do
in
parallel,
serial
etc.,
and
that
is
how
these
pages
work.
I
liked.
This
concept
in
English
department,
because
talking
fully
about
those
elements
interconnected
in
series
and
that
the
output
of
one
becomes
thrown
from
erc
the
arcades.
Where
we
come
to
talk
about
what
it
is.
B
Basically,
we
could
say,
like
the
box
applied
to
match
in
learning
only
that
there
are
different
things
but
As
for
an
association,.
We
could
see
as
a
discipline
of
this
type,,
so
e-mail
obviously
comes
from
machine
learning,,
since
the
operational
administration
of
machine
LAN
models
that
will
provide
a
point-to-point
development
of
the
entire
process
that
has
machine
animal
construction
management
of
models.
That
is,
reproducible
that
you
can
test
the
models
and
that
the
models
can
also
evolve
to
improve
and
give
better
predictions
of
the
information
that
is
in
melo.
B
What
it
tries
to
do
is
automate
all
the
light
lines
of
consumption
to
data
to
the
deployment
of
machine
models,
so
that
system
users
can
use
it,.
Since
here
we
do
not
automate
the
deployment
of
normal
software,
such
as
web
applications
and
all
that,,
but
rather
automates.
Two
machine
learning
models,
the
entire
process
from
reading
data
model
generation
to
Publish
it
so
that
it
can
be
consumed
by
an
application
that
I
am
going
to
use,.
B
B
So
for
example,
differences
between
users
and
roles,
we
could
mention
the
box
developers
and
there
are
operators
we
can
mention.
Titles
like
the
star
is
the
box
ingénieur.
Knowing
that
you
like
it
is
a
culture
in
a
clown
title
in
general.
It
is
losing.
Are
the
titles
that
we
cut
in
the
markets
in
the
industry,
so
those
They
are
the
clients
and
the
users
are
the
stakeholders
that
are
involved
with
the
concept
of
opzz,
but
the
users
change
them
here.
The
users
are
data
centers.
They
are
data
engineers
or
lost
day
time
lapse.
B
So
we
can
also
focus
a
lot
on
the
holy
data
that
they
are,
the
that
are
present
in
the
m2
before
the
clients
change.
What
the
artifacts
go
as
I
mentioned
in
the
box,
because
the
super
general
would
be
a
front
end
of
a
web
application,
but
in
the
e-mail
it
is
the
specific
machine.
Learning
models
then,
and
It
makes
a
specific
application
to
automate
my
processes,,
but
for
the
machine
and
area
we
can
also
talk
about
differences
in
technologies.
A
B
Technologies,,
since
it
is
in
the
box,,
we
are
going
to
find
the
databases,
languages,
borders,
backend,,
UIES,,
python,,
mysql,,
etc.
However,.
When
we
talk
about
e-mail,
technologies,
well,,
we
find
machine
learning
models
and
data,.
We
are
going
to
find
payton
libraries
such
as
site
and
plan
del
sur
flow
and
other
types
of
software.
Libraries
for
data
processing
such
as
apache
spark
hadoop
and
cloud
services.
C
B
B
We
mentioned,
the
faster
we
publish,
we
reduce
errors,
already
in
production,,
a
lot
of
staff
and
trips.
Also
in
Melo.
What
you
are
trying
to
do-
is
automate
the
models
with
value
to
give
value
to
that
business,
that
is,
two
automates
can
be
fast
automate
the
entire
process,,
but
the
idea
is
to
deliver
machine,
learn
models
that
add
value.
B
B
Then
we
can
see
As
such
a
general
flow
of
what
the
machine
learning
part
is
and
this,
well,.
What
there
is
a
reference
is
a
great
site
where
they
can
learn,.
There
are
92
in
the
references,,
because
the
general
model
data
enters,,
which
can
change
the
model,
they
compile
it,,
then
they
pack
it
and
they
can
distribute
it
tasks
that
they
will
find
in
the
veil.
Without
data
management
exploration,.
A
B
Itil
extraction,
transformation,
alhaurín
clean
the
data
split,
the
data
partition
them
and
well.
There
are
a
lot
of
formats
like
parquets.
If
it
is,
I
saw
a
lot
of
formats
for
the
information,.
These
are
tasks
that
you
can
find
in
the
mail
ops,
you
can
find
them
in
the
part
of
the
machine,
learning,
Pérez,.
B
You
will
find
training
of
models,
evaluation
of
models,
tests,,
the
packaging
of
the
workflows,,
that
we
can
mention
the
training
and
the
prediction
and
well
here
we
are
going
to
talk
about
the
tree,
because
something
has
two
components
than
that
long
workflow.
If
there
is
something
city
and
well,
they
are
continuously
in
continuous
development.
It
is
something
basically
what
it
offers
them
is
a
tool
to
do
workflows
like
this
in
general,
but
it
has
been
using
a
lot
for
what
is
machine.
B
Is
that
each
step
that
you
execute
becomes
a
container
and
the
concept
of
graph
is
handled
graphs
directed
to
cyclics
the
tags
for
the
famous
dax?
And
it's
like
agnostic
to
any
cloud,
and
well
it's
native
for
rulers
and
with
content?
It's
basically
something
it
does
something
workflow
it's
done
for
klein
and
something
like
that.
It's
a
tool
to
do
continuous
deployment
for
then
something
like
that.
They
can
use
it
to
be
publishing
models,
the
machine
learning,
so
they
can
I
think
because
I
stop
charts
and
all
the
everything
that
carries.
B
They
can't
connect
it
to
the
deep
hop
repository,
keep
their
versions
of
their
model
today,
except
this
becomes
a
very
good
one
solution
that
includes
is
longer
language.
Something
super
light
live
to
make
a
pipeline
and
deployment.
Imagine
not
even
flags
these
with
a
good
match
for
m2
tools
that
could
regularly
replace
these
cases
may
well
apply
to
a
lot.
What
is
something
in
house
on
on
premise
band
in
hardware,
man,
or
maybe
the
business
may
not
be
very
big,
or
they
need
to
have
that
philosophy
to
be.
B
A
C
B
Are
going
to
connect
to
the
machine
and
to
be
able
to
run
away
to
be
able
to
recover
it?
We
are
going
to
enter
the
following
command
when
we
know
we
have
it
here
so
here,
I
have
a
hard
work,
glows
plan
that
would
produce
a
command
this
at
the
same
time
waits
for
you
as
an
interface
to
the
data
in
kiswa
by
allows
how.
B
C
B
We
can
even
see
the
result,
the
steps,
the
output,
the
one
from
the
container
well,.
What
it
did
was
that
it
basically
calculates
some
grade.
Point
averages
from
students,
since
I
teach
at
universities,
is
an
example
of
this,
and
well,.
It
downloads
information
from
a
bouquet
from
Google
Cloud
itself,.
It
transforms
me,
that
converts
the
model,
prepares
it,
and
in
the
next
step,,
which
is
the
diploma
process,
well,.
What
it
does
here
It
is
that
it
sends
to
call
something
civil
and
then
it
displays
annual
display,
and
here
we
are
going
to
do.
B
B
Service
the
posts,
then
that
layer
will
synchronize
it
and
send
it
to
be
called.
Obviously,
as
the
forest
is
wrong,
or
it
will
throw
little
red
hearts
indicating
that
it
could
not
and
publish
the
model
correctly,
so
we
are
going
to
open
and
start
the
little
red
hearts.
Then
I
can
already
do
a
roll
back
in
this
part.
Here,
for
example,
this
is,
it
could.
C
C
B
B
We
are
going
to
see
then
there
in
this
way,
because
they
are
already
connecting
something
with
obviously
here,
because
each
execution
is
the
pipeline,
they
go
twenty-one
seconds
ago.
So
this
step
plus
the
long
flame,
already
walk.
They
are
already
in
a
green
state.
Then
our
model
is
already
there
and
since
it
is
basically
a
beer,
it
is
the
support
they
consume
and
it
makes
me
the
prediction
that
given
some
marks,
that
are
partial
17
points
are
especially
seven
points
and
its
average
zone
to
the
final
exam,
which
is
25
points
33.
B
B
B
C
B
Piece,,
how
can
I
start
to
learn
more
about
two
twins,?
What
tools,
recommended
books,
well,
look,,
it's
so
real
and
average,
in
Beverly
Media,?
It
has
very
good
books,
some,.
There
are
some.
books,
the
machine
learning.
What
we
can
do
is
also
I'm
going
to
deal
with
this
question.
If
it
got
halfway,
we're
going
to
the
science
slack
of
El
Salvador,
maybe
if
they
can
stick
it
out
there
well
answering
there,
there
are
quite
a
few
books
on
the
machine.
B
B
B
Let's
see
for
a
walk
m
m
l,
ops,
heretic
point
that
is
a
very
good
site:
ml,
ops,
heretic,
point
great
site,
so
I
know
to
learn
by
myself.
Regarding
the
option,
I
would
say
that
it
is
the
easiest
site
to
learn
because
it
speaks
from
the
point
of
view
of
how
m
lops
explained
the
philosophy
that
they
deal
with,.
So
I
would
recommend
starting
here
and
suddenly
some
real
book
and
measuring
and
they.
C
B
Very
well,,
let's
see,
I,
repeat
here,
how
the
boxing
and
twins
interact
with
each
other
in
the
same
ecosystem,,
supposedly
when
they
are
looking
for
an
emelec,
well,.
They
are
looking
for
someone
like
a
data
engineer
with
machine
learning
capabilities
and
who
are
in
some
contact
with
the
box
back
how
they
are
like
hybrid
posts.
B
C
B
Here
also
hip
hop
dotcom,
pl,
/
mg
fight
and
a
lot
of
e-mail,
although
you
try,
if
that
is
the
answer,
and
the
two
of
them
will
be
alone
right
now
they
raised
their
last
puig
with
what
the
last
part
I
did.
So
there
is
already
a
good
part
here
online
and
let's
see
something,
if
something
and
workflows
do
you
recommend
it
for
another
workflow
apart
from
me,
what
I
do
recommend
it
for
other
workflows
it
has,
for
example,
to
do
as
chrome
processes.
B
What
happens
is
that
basically
something
workflow
is
not
only
for
machine
learning.
In
fact,
many
mention
that
workflows
is
a
generic
workflows,
machine,
yes,,
but
it
is
not,.
It
is
not
oriented
upwards,,
but
rather,.
It
is
a
generic
workflows
engine
that
can
be
used
for
anything,,
but
it
has
become
very
popular
in
the
parts
of
machines.
I
would
say
that
it
also
has
a
very
good
application
to
do
data
processing.
It
goes
for
what
is
so
generic,
so.
B
Well,
the
resources
that
skaters
can
see
the
tree
pages
tree
for
workflows,
sidi
tree
to
decide,
I
invite
you.
If
you
want
to
participate
in
cloud
black
and
guatemala.
These
are
the
links
of
our
myth.
It
is
we
are
a
friendly
community
of
salvador
good.
We
want
to
unite
the
central
americans
and
the
latinos
lashes
likes
is,
if
you
want
to
take
them
trincheta
and
count
quickly.
These
are
those
of
the
ice
city
sv
of
salvador
tree
that
three.
Then
you
can
access
To,
them,
Casey,,
it's
my
ib3
tree.
The
repository
had
already
shown
them.