►
From YouTube: Autonomous Multi-Cloud serverless deployment and optimized management | Jakarta Tech Talks | Jan 19
Description
Paweł Skrzypek and Alicja Reniewicz Present:
The dynamic development of Cloud Computing with the introduction of novel Cloud computing models like serverless creates new challenges for Cloud deployment. This presentation describes how to implement Multi-Cloud native strategies using advanced an open source framework that allows for Cloud-agnostic Multi-Cloud deployment and optimized management of the serverless applications based on flexible monitoring, context aware maximization of the application owner’s utility of the deployed serverless components, and autonomic reconfiguration based on the application’s current execution context. All stages of the cloud deployment planning and designing process will be shown. Also, the key execution steps will be provided.
Get Social!
Website: jakarta.ee
Subscribe: https://www.youtube.com/user/EclipseFdn
Follow us on Twitter: https://twitter.com/JakartaEE
Like us on Facebook: https://www.facebook.com/JakartaEE/
Join us on LinkedIn: https://www.linkedin.com/groups/13597
Tag us: #JakartaEE
Want more great Eclipse Community webinars? Become a member of our Jakarta Tech Talks Meetup: https://www.meetup.com/jakartatechtalks
A
Hello,
everybody
we
are
live.
Welcome
to
another
jakarta
tech
talk.
Today's
topic
is
autonomous,
multi-cloud,
serverless
deployment
and
optimized
management.
We
are
lucky
enough
to
have
joining
us
today
from
ai
investments,
pavel
skirchip,
I'm
not
sure
if
I
got
that
right
or
not
pavel,
but
apologies.
A
B
So
we
will
try
to
to
have
around
the
25
minutes
for
the
slides
and
that
the
rest
of
the
time
will
be
live
demo.
If
you
have
any
questions,
please
ask
on
the
chat.
We
will
try
to
answer
them
simultaneously
during
the
presentation.
Also.
I
hope
that
we
will
be
able
to
have
some
q
and
a
session
at
at
the
end.
A
B
A
B
A
B
B
As
I
said,
it
was
originally
created
in
the
melodic
product
project.
It
was
the
project
on
research
and
development
project
under
the
horizon.
2020
and
current
development
of
the
melodic
project.
Melodic
platform
is
done
within
the
morphemic
project,
which
is
also
the
research
action
and
innovation
action
during
under
the
horizon.
2020
project.
B
So
why
to
use
melodic?
The
most
important
feature
is
that
melodic
is
probably
the
simplest
and
easy
way
to
use
the
multi-cloud
approach.
Application
is
modeled
in
the
camera
language
and
then
it
is
automatically
deployed
to
the
multi-cloud
environment
and
do
not
require
any
knowledge
about
the
particular
cloud
providers
we
can
deploy
to
the
aws
google
cloud
platform
azure
and
also
to
any
openstack
based
public
and
private.
B
B
The
probably
the
most
unique
feature
of
the
melodic
is
ability
to
automatically
optimize
the
usage
of
the
cloud
resources,
even
more
the
usage
of
the
multi-cloud
resources
so
based
on
our
knowledge,
it
is
the
only
platform
which
allows
for
the
automatic
optimization
of
the
multi-cloud
resources
in
the
runtimes,
so
the
application
is
deployed.
We
are
monitoring
them
and
are
able
to
to
add
or
remove
the
resources.
B
On
the
further
slides,
I
will
say
what
it
means
that
the
best
deployment
how
this
optimization
is
done
and
what
is
the
uniqueness
of
our
platform
and
also
what
are
the
topics
which
we're
currently
working
on,
but
first
step
is
to
model
the
application
and
we
have
developed
the
camera
language.
It's
a
acronym
for
the
cloud,
application,
modeling
and
execution
language.
B
B
Of
course
we
can
model
the
infrastructure,
but
we
are
not
specifying
the
exact
infrastructure
which
have
to
be
used,
but
we
can
define
the
requirements
which
are
used
by
the
optimization
process
to
find
the
best
possible
deployment
and
it
is
done
by
defining
the
user
requirements
so,
for
example,
what
should
be
the
maximum?
What
is
a
load?
The
maximum
response,
time
of
the
given
application
constrained
that,
for
example,
we
want
to
have
more
than
four
virtual
machines
for
the
given
component
and
utility
function,
which
is
a
very
unique
way
to
measure
how
good
deployment
is.
B
I
will
also
say
a
bit
more
about
that
later
and
also
very
important.
We
are
able
to
try
to
translate
this
camel
model
into
the
form
of
the
constraint
programming
model,
which
is
the
set
of
the
equation,
which
defines
the
optimization
problem,
and
this
problem
is
solved
by
the
platform
to
find
the
best
optimal
deployment
to
the
multi-cloud
and
environment.
B
So
what
is
this?
The
best
deployment?
There
is
no
easy
answer
because
for
the
different
application
it
could
means
different,
for
example,
for
the
e-commerce
application.
The
response
time
to
the
user
should
be
minimal
or
not
bigger
than
a
specified
one
for
the
big
data
application,
the
jobs
processing
time
should
be
minimal
or
no
longer,
and
so
on.
B
Also,
the
cost
should
be
as
slow
as
possible
and
also
other
functional
and
non-functional
requirements
should
be
defined,
and
this
deployment
this
the
quality
of
the
deployment,
is
assess
on
the
runtime
based
on
the
metric
collected
by
the
application.
Part
of
the
melodic
platform
is
a
novel
event
management
system
which
allows
for
the
multi-layer
metric
collections.
B
We
have
the
flexible
way
to
calculate
the
utility
for
the
particular
application,
so
we
can
define
a
utility
which
focus
on
the
business
value,
so
we
can
measure
the
cpu
memory,
but
we
also
can
measure
the
risk
average
response
time
to
the
user
or
average
time
of
the
job
processing
or
any
other
business
metrics
and
our
utilities.
So
deployment
could
be
based
on
that
metrics,
and
the
best
deployment
is
always
the
trade-off
between
the
cost
performance
availability.
B
Of
course,
if
we
remove
the
cost
from
that
utility,
then
we
can
use
the
best
possible
resources
and
as
many
of
them
as
possible,
it
will
have
the
best
performance,
but
the
cost
will
be
the
also
the
highest
yeah.
So
duty
usually
is
focused
on
the
optimizing
that
rate
of
between
the
cost
and
other
other
requirements,
and
this
deployment
this
optimization,
is
done
fully
automatically.
So
the
metric,
as
I
said,
are
collected
in
the
real
time
and
deliver
through
the
multi-layer
event.
B
Melodic
is
triggering
the
reconfiguration
process,
so
the
new
deployment
is
optimized
calculate
and
the
new
solution
is
found.
Then
melodic
automatically
reconfigure.
The
application
add
the
resources
to
the
certain
components
or
remove
the
resources,
and
it
is
done
fully
automatically
in
the
safe
way.
So,
for
example,
if
we
want
to
switch
component
from
the
one
virtual
machine
type
to
the
another,
then
the
first,
the
new
virtual
machine
is
created.
B
The
communication
is
switched
to
the
new
virtual
machine
and
then
the
the
old
virtual
machine,
with
the
previous
version
of
the
components,
is
deployed.
We
are
fully
supported
the
cloud
native
deployments,
so
the
application
is
due.
The
infrastructure
and
the
application
is
created
from
scratch,
based
on
the
on
the
model
defined
in
the
in
the
camel
language.
B
Here
is
a
very
brief
architecture
of
the
melodic.
You
could
go
to
our
webpage
to
get
more
more
documentation
and
more
information
about
that,
but
more
the
less.
It
looks
like
that.
We
are
fully
support
the
micro
service
architecture
with
the
control
plane
using
rest
protocol
and
monitoring
plane,
which
uses
the
active
mq
for
moving
the
the
monitoring,
metrics
and
values.
B
B
First
of
this
concept
is
polymorph
architecture,
so
we
want
to
be
able
to
choose
the
physical
architecture
of
the
application
so
to
choose.
If
we
want
to
use
the
given
application
and
deploy
deployed
to
the
virtual
machines
or
maybe
serverless
components,
or
maybe
some
hardware
accelerated
resources,
and
that
will
add
the
additional
layer
for
the
for
the
optimization.
B
So
the
goal
is
to
be
able
to
maximize
the
utility
of
the
application
by
choosing
not
only
the
number
of
the
resources,
but
also
the
technical
architecture
of
the
application,
and
the
second
concept
which
is
currently
developed
by
us.
Research
and
develop
is
the
proactive
adaptation
because,
as
I
said
on
the
previous
slides
currently
melodic
is
adapting
based
on
the
actual
current
values
of
the
metrics.
B
B
So
our
idea
for
the
proactive
adaptation
is
that
it
will
work
at
the
beginning
in
the
similar
way.
As
currently
melodic
is
working.
So
there
will
be
the
initial
deployment
of
the
application.
Then
the
metrics
will
be
collected,
but
the
optimization
will
be
based
on
the
forecasted
value
of
the
metrics.
So
we
are
anticipating
the
future.
Not
the
current
one
and
the
optimal
deployment
plan
will
be
taste
this
one
which
is
optimal
in
the
future,
not
now,
and
the
reconfiguration
will
be
based
on
on
that
yeah.
B
B
Allah
will
show
on
the
next
presentation,
but
I
want
to
say
to
show
two
to
two
use
case
applications.
One
of
them
will
be
shown
by
allah
life,
one
of
them
I
want
to
say,
because
it
is
used
by
the
ai
investments.
So
the
ai
investments
company
is
the
active
user
of
the
melodic
and
in
future
morphemic
platform,
and
it
saves
a
lot
of
money.
I
will
show
how
much
money
we
can
save
thanks
to
the
using
melodic
investments
is
the
company
which
is
working
on
the
machine,
learning
and
ai
models
for
the
portfolio.
B
B
So
this
especially
machine
learning
methods
requires
a
lot
of
the
computer
national
power
to
train
the
models,
especially
that
we
need
to
train
these
models
in
the
certain
amount
of
time
and
of
course
we
want
to
save
the
cost
and
minimize
the
the
budget.
So
one
of
the
example
of
our
business
core
is
to
train
50
predictions
models
in
the
one
hour,
using,
of
course,
as
many
more
resources
as
possible,
how
it
is
done
with
the
melodic
yeah.
We
are
starting
training,
so
this
this
one
guy.
B
It's
me,
I'm
not
sure
if
you
can
see
the
similarity
but
yeah,
that's
me
or
other
analyst.
In
the
investments,
we
are
starting
to
train
our
models
on
our
resources
on
our
own
premises,
resources
because
they
are
the
cheapest
one,
because
we
already
own
them
and
melodic
is
checking
how
long
it
takes
on
our
resources.
B
B
Melodic
is
checking
that
with
additional
resources
it
still
takes
two
hours,
so
add
additional
resources
and
with
additional
resources
we
are
able
to
finish
the
processing
in
one
hour.
So
after
the
finishing
the
training,
the
all
of
the
cloud
computing
resources
are
decommissioned,
so
melodic
is
able
to
optimize
it
in
the
runtime
and
with
very
good
results.
B
B
All
of
these
savings
are
documented,
so
there
are
a
detailed
description.
How
we
achieve
this
savings
using
the
multi-cloud,
optimization
and
yeah.
I
think
that
using
the
the
novel
features
of
the
melodic
platform,
which
are
developed
in
the
morphemic
project,
we
will
be
able
even
to
save
more
and
the
second,
maybe
let's
say
more
more
familiar
example-
is
genome
application.
B
It
is
the
application
which
processed
genome
data
and
compared
that,
with
the
reference
genome
to
find
the
notation.
It
is
the
very
typical
big
data
application
which
is
running
on
the
spark
and
it
works
exactly
in
the
same
way.
So
we
want
to
process
the
given
number
of
the
jobs
in
the
certain
amount
of
time,
and
we
are
starting
with
the
minimal
resources
based
on
the
speed
of
the
processing.
Melodic
is
calculated
how
long
it
takes
and
add
additional
crowd.
Resources
then
again
check
how
it,
how
much
time
it
takes
and
add
additional
resources.
B
So
everything
is
working
in
the
runtime.
It
is
a
very
unique
way
to
dynamically
optimize
the
resources
used
by
the
big
data
frameworks
and
big
data
application,
and
thanks
to
that,
melodic
allows
to
fulfill
the
business
goals,
for
example,
to
process
the
data
in
the
certain
amount
of
time,
and
also
to
save
a
lot
of
a
lot
of
money
really
lot
of
of
money.
So
that's
probably
the
most
the
best
business
reason
to
use
that
application,
and
now
we
will
switch
to
to
the
live
session
I'll.
B
C
C
We
will
monitor
application
metrics
and
observe
reconfiguration
process,
which
is
done
by
melodic.
For
reasons
of
optimization,
my
melodic
platform
is
installed
on
virtual
machine
on
aws
and
it
is
up
and
running.
I'm
logged
in
and
melodic
users,
I
managed
by
by
ldap,
and
we
have
three
possible
roles
of
users:
common
user.
He
can
perform
application
deployments,
administrators
admin
and
managers
of
users
accounts
and
also
has
all
privileges
from
common
user
and
the
last
type
of
user
technical
users.
C
C
C
C
A
C
In
the
meantime,
I
would
like
to
briefly
describe
application,
which
is
being
deployed
by
melodic
tao.
A
genome
is
a
big
data
application
which
performs
some
calculations
and
saves
results
in
aws
as
free
bucket.
So
we
needed
to
provide
developers
credentials
to
aws
and
genome's
performance
is
managed
by
spark
in
genome
application.
We
use
spark
as
platform
for
big
data
operations,
which
are
performed
parallel
on
many
virtual
machines
and
managed
by
one
machine
named
spark
master
sparkmaster
is
available
by
default
on
melodic
platform.
C
A
melodic
creates
proper
number
of
spark
workers,
as
virtual
machines
considered
our
requirements
from
camel
model
thanks
to
measurements
of
application.
Metrics
melodic
makes
a
decision
about
creating
additional
instances
with
workers
or
about
deleting
an
accessory
once
spark
divides.
All
calculations
named
tasks
between
available
workers
in
order
to
optimize
application,
performance
and
cost,
and
now
I
would
like
to
back
to
our
deployment
process
and
fetching
offers
is
the
first
step
of
deployment
process.
C
Us
and
the
next
step,
the
next
step
of
of
process
is
generating
constraint.
Problem
constraint
problem
is
generated
based
on
our
requirements
defined
in
common
model.
In
simple
process
view,
there
are
visualized
all
variables
from
constraint
problem
with
a
domain
values
for
genome.
We
have
worker
cardinality
a
worker
course
and
provider
for
component
spark
worker.
C
So
we
can
see
that
we
would
like
to
have
from
1
to
maximum
10
workers
instances
and
the
last
element
here
metrics
with
data
type
and
values
they
describe.
Current
performance
of
application,
thanks
to
them,
melodic
can
make
can
make
the
main
main
the
most
important
task,
which
is
a
cost
optimization
after,
of
course,
a
decision
about
triggering
the
reconfiguration
process,
which
means
creating
new
additional
instances
or
deleting
not
fully
used
once.
C
And
when
a
constraint
problem
is
generated,
it
is
time
for
designing
logic
finds
here
the
best,
the
most
profitable
solution
for
the
problem
defined
by
us
and
when
resulting
is
completed.
We
can
observe
information
about
calculated
solution,
a
utility
value
for
each
variables,
in
that
case,
one
as
worker
cardinality
for
worker
course
and
provider
for
spark
workers
from
zero
index,
and
then
next
step
is
process
and
the
next
step
in
our
process
is
deploying
here.
Melodic
performs
operations
based
on
calculated
solution.
C
B
B
Then
we
need
to
have
the
melodic
platform
itself
deployed
and
then
we
have
to
submit
our
comment
model
and
these
steps
are
already
done
and
then
starts
the
deployment
process
and
the
deployment
process
is
quite
complex
because
we
need
to
fetch
all
of
the
offers
from
the
cloud
providers
and
it
was
done
as
allah
has
presented
and
then
the
best
say
most
optimal.
Resources
needs
to
be
selected
and
then
deployed
yeah.
So,
as
we
have
said,
they
said
the
deployment
is
done
automatically.
B
So
this
is
done
on
the
on
the
all
clouds
in
the
in
the
very
similar
way,
yeah
so,
and
we
decide
to
to
show
that
real
and
and
live,
but,
as
you
can
see
of
all
sometimes
some
bad
luck
that
something
is
not
working,
but
let's
give
some
time
a
lot
to
check
that
and
also
yeah.
So
I
previously
said
the
melodic
is
fully
open
source
and
could
be
downloaded
under
this
link.
B
So
if
the
present
our
presentation,
our
platform
is
interesting
for
you,
then
you
can
go
to
that
to
our
website
download
and
try
to
use
by
by
yourself.
There
are
also
a
lot
of
the
resources:
videos
documentation,
some
tutorials,
how
to
use
the
melodic.
We
are
recording
the
presentation
and
they
are
available
then,
and
also,
as
I
already
mentioned,
but
yeah.
B
B
C
So
when
our
process
is
successfully
successfully
done,
we
can
see
we
can
see
here
information
about
about
that.
Okay
and
after
that,
we
can,
we
can
observe
and
the
result
of
it
also
on
a
spark
master
spark
master
ui,
as
I
mentioned
before,
spark
master
is
as
part
of
our
melodic
machine,
as
so,
we
can
check.
C
C
Each
application
has
own
settings
and
on
metrics
so
also
requires
own
own
dedicated
dashboard
and
also
in
case
of
of
dom
application.
We
can
see
information
we
can.
We
can
see
information
about
if
we
can
finish
on
time
or
not
thanks
to
these
traffic
lights
in
for
imitation,
and
also
we
have
also,
we
can
observe
how
many
minutes
left
from
the
time
period
under
time
left
value.
C
C
Indicate
that
the
expected
time
was
indicated
in
camel
model
and
in
case
of
genomic,
is
equals
to
16
minutes,
and
so,
for
example,
here
that
here
we
can
see
that
we
have
only
one
minute
of
performing
this
application
and
59
minutes
left
and
based
on
current
performance.
It
is
calculated
the
estimated
time
left
and
on
the
left
on
the
first
chart
we
can.
C
C
C
C
C
C
B
C
A
A
couple
more
minutes
here
to
answer
any
questions
if
anyone
has
any
last
minutes
but
yeah.
Otherwise,
thank
you,
everyone
for
joining
today
and
I'm
just
going
to
post
a
quick
link
to
the
survey
in
the
green
button
below
if
you,
if
people
watching,
would
not
mind
completing
that
afterwards,
it's
always
great
to
get
feedback
on
how
you
enjoyed
this
session,
not
seeing
any
more
questions
coming
in
here.
So
we'll
give
everybody
a
couple
of
minutes
back
in
their
day.