►
Description
Forget about managing a fleet of servers - developers can now focus their efforts on business logic without concern for infrastructure. It’s about serverless computing and deploying code as opposed to deploying VMs or containers. This all leads to a cloud-native event based future embracing functions, containers, and managed services.
Building, deploying, and administering serverless applications requires more than just a function scheduler and executor. This is especially true in the enterprise, where IT administrators want control and developers want ease of use and freedom. Dispatch balances these competing needs with features beyond just function execution.
For IT, Dispatch brings tenancy and access management to open source serverless. For developers, Dispatch supports a wide variety of languages and leverages the open service broker standard to allow functions to interact with a wide variety of managed services, such as databases.
Is Dispatch right for you and your environment? Are you trying to figure out serverless and how this fits with your cloud-native infrastructure?
In this webinar, Dispatch lead Berndt Jung and project founder Mark Peek from VMware gives an introduction to functions and serverless as well as demonstrating how to solve real use-cases with Dispatch.
A
See
and
we're
live
all
right,
welcome
everyone
and
well
hello,
everyone
and
welcome
to
today's
webinar.
We're
very,
very
excited
to
have
you
all
here.
This,
as
you
might
already
know,
is
a
webinar
series
where
we
were
introducing
and
explaining
some
of
the
interesting
open
source
projects
that
stem
from
VMware
and
the
topic
of
today
is
project
dispatch
and
serverless
functions.
If
you
have
any
questions
throughout
the
webinar,
please
add
them
to
the
Q&A.
You
can
find
at
the
bottom
of
your
screen
and
we
will
answer
them
throughout
the
presentation.
A
Our
speakers
today,
our
dispatch
project,
lead
parent
young
and
a
project
founder
mark
peek
from
vmware
mark
has
been
working
within
the
open
source
space
even
before
the
term
open
source
existed,
so
he's
a
veteran
within
the
open
source
space
for
sure.
So
I'm
really
excited
to
hear
what
Baron
and
Mark
is
going
to
talk
about
today
and,
as
you
all
know,
there's
an
interactive
webinar.
So
please
ask
as
many
questions
as
you
can
with
that
I'm
handing
it
over
to
mark
who
will
start
the
presentation.
B
This
has
arisen
and
the
reason
you
know
as
we
as
we're
looking
through
this
progression,
we're
seeing
that
over
time,
there's
a
decreasing
concern
and
control
over
the
infrastructure
implementation,
but
at
the
same
time
we're
focusing
more
on
what
is
the
business
logic
associated
with
how
you
deliver
your
applications
and
so
as
a
developer.
You
know
I
just
want
to
be
able
to
write
my
code
hit,
submit
and
have
it
up
and
running
and
not
have
to
really
deal
with
a
lot
of
the
infrastructure
concerns
around
how
the
other
than
machines,
actually
a
provision,
etc.
B
B
B
When
we're
looking
at
this,
it's
a
piece
of
code
that
that's
running
a
function
along
with
its
dependencies,
it's
triggered
by
something
that
happens
from
an
external
side
of
things.
So
it's
event-driven
either
some
event.
Http
requests
the
timer.
You
know
some
some
something
gets
in
queued
into
it.
It
causes
the
trigger
to
to
occur
and
that
function
then
to
be
run.
B
The
function
is
stateless
and
because
of
that,
it's
now
more
easily
auto-scale,
so
that
you,
as
the
events
are
coming
in,
you
can
now
scale
up
the
number
of
instances
that
those
functions
are
running
in
order
to
address
the
needs
of
the
business
and
handle
all
of
those
requests
coming
in.
You
know,
so
the
the
back-end
services
could
be
that
it
talks
to
databases
it
could
in
queue
something
on
on
to
a
message:
bus.
B
You
could
give
an
immediate
result,
there's
a
lot
of
different
ways
for
it
to
communicate
back
but
being
able
to
handle
bursts
of
load
coming
from
something
like
IOT
devices.
Mobile
devices
or
just
your
normal
website
makes
it
very
handy
to
be
able
to
use
when
we're
looking
at.
You
know
how
long
does
it
last
you
know.
Typically,
we
saw
with
with
virtual
machines
that
you
know
you.
You
have
a
virtual
machine
that
would
last
for
you
know
days
two
years
depending
depending
on
how
how
much
you
look
you
enjoyed
the
naming
it.
B
Another
interesting
one
is
just
stream
processing
that
you
might
have.
You
know,
job
queues
that
you
need
to
handle.
You
may
have
the
ETL
that
extract
translated,
load
types
of
workloads
be
able
to
have
those
going
into
a
service,
a
functions
as
a
service
in
order
to
either
scale
up
the
number
of
requests
coming
in
DevOps.
We're
seeing
it
in
terms
of
notifications
and
see
ICD
pipelines
and
then
just
general
IT
operations
is
how
how
can
you
do
provisioning
and
provisioning
and
other
automation
more
easily
using
functions.
B
So,
let's
talk
about,
you
know
just
a
couple
of
serverless.
The
use
case
be
these
are
more
focused
on
public
cloud,
but
I
think
that
they're
they're
very
interesting
Nordstrom
has
it
has
a
open
source
site
that
it's
called
Apollo
retail,
where
they
designed
a
retail
site
that
is
entirely
written
in
functions
and
using
back-end
services
of
the
public
cloud,
and
it's
very
interesting
to
be
able
to
take.
You
know
decompose
something
that
a
lot
of
businesses
have,
which
is
a
retail
site
and
decompose
it
down
into
you.
B
How
would
you
actually
structure
that
in
two
different
functions
and
then
tie
that
in
to
the
databases
and
and
other
attributes
that
are
required
to
store,
stayed
or
provide
data
to
that
retail
site?
Another
one
was
coca-cola-
has
talked
about
their
their
Coke
machines,
where,
when
you
swipe
your
credit
card
to
buy
a
coke
that
then
will
trigger
a
function
to
be
run.
That
will
do
the
Tran
action
processing.
Now.
B
The
nice
thing
about
this
for
for
coke
is
that
they
only
have
to
pay
for
the
compute
in
the
public
cloud
when
someone
is
actually
doing
a
trend
transaction,
and
so
it's
a
very
you
know
one
the
one
they
understand
what
costs
are,
but
in
the
talk
that
they
gave,
it
was
interesting
where
at
a
certain
point,
it
made
more
sense
for
them
to
not
use
public
cloud
functions
but
to
switch
to
their
own
infrastructure,
and
so
you
really
do
have
to
look
at.
Where
is
the
break-even
cost
for,
especially
in
public
cloud
around?
B
Some
of
the
things
that
have
been
going
on
in
the
community
is
to
understand
what
is
service
and
the
cloud
native
compute
Foundation,
which
is
the
foundation
that
has
kubernetes,
Prometheus
and
and
other
open
source
projects
associated
with
it.
You
know
we
we
went
to
them
and
said
we'd
like
to
talk
more
about
what
is
service
and
have
an
opinion
about
it.
So
a
bunch
of
us
got
together
and
we
wrote
a
white
paper
to
discuss
what
is
serverless.
What
are
the
aspects
of
it
and
then
what
should
we
be
doing
about
it?
B
We
ship
the
the
white
paper
earlier
earlier
this
year
as
a
1.0.
We
also
pulled
together
a
landscape
in
order
to
be
able
to
say
what
are
all
the
projects
there
they're
occurring
in
service
at
this
time
around
both
public
cloud
and
private
cloud
on
from
use
cases.
You
know
one
of
the
projects
which
Baron
will
be
talking
about
is
our
open
source
dispatch
project.
B
As
you
see
here
it
listed
in
the
landscape
and
at
the
end
of
the
writing,
the
white
paper,
we
decided
that
we
needed
to
go
out
and
look
at
other
further
harmonizations,
and
so
the
the
next
area
that
we
talked
about
started
talking
about
was
around
cloud
events,
and
so
we
created
this
CN
CF
sandbox
project
to
be
able
to
understand
what
should
we
do
around?
How
do
you
send
events
and
identify
events
going
possibly
between
clouds
and
so
at
coop
con
in
Copenhagen
in
May?
We
we
actually
do
the
fun
demo.
B
This
was
Austin
Collins
from
serverless.
Comm
gave
a
talk
about
the
cloud
events
and
also
about
how
we
did
this
interoperability
demo.
So
we
had
two
publishers.
We
had
AWS
publishing
and
also
a
juror
going
into
either
of
them
gateway
or
after
event,
break
where
it
would
then
generate
a
cloud
event.
Those
cloud
events
were
sent
to
a
variety
of
different
subscribers.
B
That
would
then
do
something
with
that
cloud
event
coming
in
the
the
way
that
we
structured.
This
was
to
create
the
cloud
event
as
a
link
to
an
image,
and
then
we
would
do
some
each
of
the
subscribers
would
do
something
with
that
image
and
then
post
post
out
to
Twitter.
It
was
great
to
be
able
to
see
you
know
just
the
the
breadth
of
the
teams
that
were
working
on
the
spec,
the
breadth
of
which
the
projects
were
able
to
line
behind
the
0.1
specification.
C
A
C
Okay
yeah,
so
my
name
is
Marek
I've
been
sort
leading
the
dispatch
project
for
the
past
year
or
so
since
before
was
dispatched
but
dispatches
our
take
on
functions
and
so
us
being
VMware.
We
have
a
sort
of
enterprise,
centric
viewpoint
on
things
and
so
we're
trying
to
sort
of
address
the
fads
use
cases
for
for
enterprise,
where
we
don't
think
that
the
existing
open
source
alternatives
for
baths
are
necessarily
addressing
enterprise
use
cases
which
include
you
know,
multi
tendency
and
integration
with
your
enterprise
directories
like
Active,
Directory
or
whatnot.
C
Also,
not,
they
don't
necessarily
have
a
secure
ingress
like
we.
If
you're,
if
you're
familiar
with
AWS
lambda
they've
got
this
event
gateway
service,
which
is
a
way
of
routing
HTTP
requests
trigger
functions,
and
this
is
also
something
that
Cognos
very
well.
So
we
chose
Kong
or
we've
chose
to
use
call
within
dispatched
gain.
This
functionality,
also
of
note
within
within
AWS
I,
think
I've
read
the
80%
of
their
functions
of
their
land
of
functions
are
actually
triggered
by
API
by
a
request
coming
through
the
API
gateway.
C
So
this
is
a
I
think
whether
or
not
we
remain
with
call
an
API
gateway
is
a
key
component
to
fast,
and
that's
why
we
chose
to
include
it.
Mark
I
already
talked
about
cloud
events,
so
we
have,
and
and
and
the
properties
of
functions
in
that
they
are
event-driven.
This
is
sort
of
one
of
the
first
things
that
we
came
up
with
when
we
were
thinking
about
where
we're
going
to
build
dispatch.
C
We
want
to
you,
know
and
essentially
enable
that
same
functionality,
but
on
cram,
which
means
that
we
need
to
be
able
to
import
events
from
external
services
such
as
an
s3
or
public
cloud
events,
but
we
also
want
to
go
beyond
that
and
be
able
to
integrate
with
internal
events.
So
you
know
stuff
like
be
Center
events,
so,
for
instance,
being
able
to
trigger
for
functions,
wanted,
BM
is
deployed
so
and
you'll
actually
see
a
demo
that
up
being
world
if
your
year-round
anyway.
C
C
You
can
bind
them
which
really
just
means
getting
credentials
for
them
and
do
that
do
this
programmatically,
so
you
can
use
external
services
from
dispatch
or
or
anything,
that's
integrated
with
it,
and
then,
of
course,
one
other
important
piece
is
you'll,
see
that
we've,
you
know
drawn
a
circle
around
fast
in
case
we
have.
We
have
this
open
phase.
Icon
dispatch
itself
is
not
a
fast,
so
dispatch
itself
does
not
do
the
function
execution.
C
We
leverage
other
existing
fast
services
for
that,
such
as
open
faz
such
as
RIF,
which
is
a
pivotal
project,
and
we
chose
to
do
this
initially
because
there
you
know,
even
even
at
the
onset
of
dispatch,
there
were
already
a
lot
of
thousands
out
there
and
they're.
You
know
they're
they're
responsible
for
scheduling
and
executing
functions,
and
we
didn't
think
we
could
provide
a
lot
of
value
in
doing
that
ourselves,
because
you
know
that's
something
that
someone
else
is
doing
quite
well.
C
So
what
does
our
architecture
look
like?
As
you
can
see?
You
know
this
is
kind
of
a
simplified
view.
The
control
plan
here
is
actually
a
series
of
microservices
they're
responsible
for
each
sort
of
component
that
we
talked
about
previously,
underneath
it
all
running
everything
on
kubernetes.
That
was
a
decision
made
early
on
that
were
you
know
we
chose
kubernetes
to
be
our
infrastructure.
We
consider
that
our
I
asked
that
allows
us
to
run
dispatch
on
a
laptop
he's,
a
mini
cube.
It
allows
us
to
run
it
on
BKE,
PKS,
eks,
basically
any
kubernetes
service.
C
C
C
Additionally,
we're
going
to
integrate
we're
going
to
use
the
service
broker
to
manage
a
database
that
provides
the
state,
so
functions
themselves
have
no
state.
We
rely
on
external
services
and
we
are
going
to
show
that
we're
going
to
use
an
external
database
hosted
in
sir
to
provide
the
state
for
our
little
notes
application.
So,
with
that
I'm
going
to
switch
a
few
demo
mode.
C
C
If
you
want
to
follow
along
now
or
later,
this
whole
demo
is
actually
checked
into
github
under
the
dispatch
framework
organization.
It
is
the
Notes
app
and
so
I'm,
basically
gonna
be
following
these
steps.
I've
already
deployed
the
UI
I've
also
already
deployed
the
Postgres
server
and
the
reason
I
did,
that
is.
It
takes
a
long
time
to
spin
up,
but
I'll
show
you
how
you
do
that.
So,
let's,
let's
maybe
even
start
with
that.
So
as
I
said,
we
need
a
post
press.
C
We
need
a
Postgres
database,
we're
going
to
use
the
service
catalog
or
open
Service
Worker
functionality
to
do
that
and
manage
that
with
dispatch.
So
we
can
see
the
dispatch
yet
service
classes,
so
I've
associate
as
ER
service
broker
with
dispatch,
and
so
this
gives
me
access
to
all
of
user
services
that
they
make
available
with
this
service
broker.
You
can
see
a
list
here
and
then
I
have
already
dispatch
get.
C
C
C
These
are
on
zero
properties
and
then
you
give
it
a
name
in
this
case
with
Postgres
and
that's
what
shows
up
when
you
look
at
instances,
and
so
provision
means
that
that
has
been
that
it
has
been
deployed
in
Ezzor
bound
means
that
we
have
also
gotten
a
secret.
We
have
gotten
credentials
for
a
database
and
so
and
then
those
credentials
are
then
stored
in
dispatch
secrets
and
which
can
be
seen
yet
secrets,
and
so
this
it
just
has
a
a
UUID
name,
but
this
is
the
Postgres
the
credentials
through
this
Postgres
database.
C
Art
storm
you
can
this
dispatch
secret?
Okay,
so
we've
got
our
database.
The
next
thing
to
do
is
to
create
an
image,
so
images
within
dispatch.
So,
let's
look
at
what
we
have
dispatch
get
face,
so
I'll
start
with
mace
Anakin's
base
images
are,
as
the
name
implies,
they're
really
just
sort
of
what's
called
the
operating
system
plus
the
runtime.
So
in
this
case
we
have
base
images
for
nodejs,
PowerShell,
Python
and
Java.
C
This
particular
demo
is
a
Python
demo,
and
so
we
have
this
notion
of
base
images
and
then
images,
images,
capture
the
runtime,
but
then
also
dependencies
that
you
might
need
for
your
function,
and
so
our
function
is
going
to
talk
to
Postgres
and
therefore
has
some
dependencies
on
Postgres
like
a
Postgres
client,
as
well
as
some
like,
a
lightweight
ORM
that
we're
going
to
use
in
our
function
as
well.
So
we
need
to
capture
those
into
an
image,
and
so
this
being
Python,
we
kept.
C
We
specify
those
with
a
request,
dot,
txt
or
requirements
dot
txt.
These
are
all
the
these
are
all
the
yeah.
These
are
the
specific
dependencies
that
we
need,
and
then
that
goes
into
the
image
definitions
of
the
images
or
image
animal.
This
is
so
we're
creating
an
image.
It's
based
on
the
Python
3
base
image,
we're
calling
it
Python
3
notes
server.
C
C
C
C
All
right:
well,
it's
definitely
taking
longer
than
it
should.
Let's
just
go
into
what
a
function
looks
like
and
hopefully,
by
that
time
the
image
will
be
ready,
yeah,
okay,
so
for
the
functions
we're
going
to
create
two
separate
functions,
our
notes,
application
is
going
to
be
able
to
add
notes,
and
then
it's
also
going
to
be
able
to
list
out
the
notes
and
we've
specified
two
separate
functions
for
that.
The
the
way
we
defined
functions
is
similar
to
imager
this
week.
I
use
a
camel
file
because
I
think
it's
easier.
C
C
I'll,
explain
that
a
little
bit
more
and
I
basically
pass
it
to
that
input
implicitly
passes
the
secrets
into
the
function
so
that,
if
you,
you
know,
we
need
to
connect
to
the
database,
and
you
can
do
that
binding
without
actually
part,
including
any
credentials
into
the
function
itself.
So
add
know
get
note,
look
very
similar
from
this
perspective.
C
C
Most
of
this
is
specific
to
the
function,
which
is
good
like
like
Marcus
Chang's,
but
you
really
kind
of
focus
just
on
the
business
logic
itself.
I
think
the
most
of
the
the
piece
to
really
point
out
is
this
handle
function,
so
the
handle
is
sort
of
the
end
point
into
the
function
and
all
functions
and
all
especially
Python
functions,
they're
going
to
have
a
func
function,
signature
that
looks
pretty
much
identical
to
this,
but
all
the
all
of
different
language
will
have
something
very
similar
in
that
it's
a
function
and
it
takes
two
arguments.
C
The
two
arguments
are
a
context
and
a
payload,
so
the
context
is
going
to
include
anything
sort
of
specific
to
the
framework,
and
this
is
where
we're
going
to
inject
secrets.
We
inject
service
findings.
We
also,
in
fact,
add
if
you're
coming
in
through,
for
instance,
the
a
gateway
we
pass
in
an
HTTP
context.
So
it'll
tell
you
like
hey
what
path
did
this
come
in
on?
C
What
was
the
request
or
what
was
the
HTTP
method
and
what
not
it's
just
sort
of
where
we
can
inject
information
through
the
framework
and
then
make
it
available
to
function?
Payload,
on
the
other
hand,
is
sort
of
the
explicit
input
to
the
function,
and
so,
for
instance,
when
we're
creating
a
note,
the
payload
is
going
to
include
the
content
of
that
note
and
it's
what
is
actually
sent.
It's
the
arguments
to
the
function
when,
when
you're
triggering
it,
so
you
can
see
and
this
particular
function.
It
really
doesn't
take
any
arguments.
C
C
The
add
note
is,
like
I
said
it's
very
similar,
except
in
this
case
we're
taking
the
contents
of
the
payload.
In
this
case
you
can
see
the
the
argument
is
going
to
be
X
within
the
payload
and
say
and
creating
the
note
and
then
this
node
dot
create
is
going
to
save
that
into
the
database.
Oh
I
guess
maybe
worth
worth
stating
you
can
see
these
service
bindings
here.
So
that
is.
That
is
our
connection,
that
those
are
Prudential
squee,
the
Postgres
database
that
we've
linked
okay.
C
C
You
can
see
they're
both
ready.
They
both
have
a
docker
image
URL,
which
is
essentially
what
we're
doing
when
we're
creating
a
function,
and
the
last
thing
we
need
to
do
is
now
expose
these
functions
to
externally,
expose
them
and
we're
going
to
do
that
through
the
API
gateway.
So
we're
going
to
create
two
API
endpoints
flowing
for
listing
one
for
both
adding
the
functions,
and
you
can
see
how
those
are
the
client
also.
C
So
we've
got
two
two
defined
here:
one,
it's
going
to
be
a
get
method.
The
URIs
can
be
slash,
notes,
Coors
allows
it
to
work
across
domains
which
is
important
for
us
in
this
demo
and
we're
we're
pointing
it
to
the
get
notes
function.
Similarly,
we've
got
a
API
that
we're
calling
post
notes
that
it's
linked
to
the
add
note
function,
takes
a
post
method
and
same
URI.
C
C
C
You
know
it's
complicated,
a
web
app
that
you
want
dispatch
function,
that
service
make
creating
these
sort
of
simple,
simple
web
apps
web
hooks.
Anything
that
you
know
require
sort
of
a.
What
would
traditionally
require
you
to
stand
up
like
an
internet
server
and
a
whole
web
app.
You
can
do
without
any
of
that
infrastructure.
All
your
all!
You
know
you
saw
you
saw
me
create
it.
C
So
that
concludes
my
demo
and
I
guess
we'll
kind
of
move
on
to
what
we've
got
in
dispatch
today,
so
I've
kind
of
tried
to
highlight
a
fair
amount
of
fair
amount
of
the
features,
but
I'll
go
through
more
or
less
everything.
As
I
already
said,
it
runs
on
any
Cooper
Nettie's
I've
also
mentioned
that
were
wrapping
fasces
and
we
support
multiple
different
phases.
C
We
support
a
broad
range
of
languages.
I've
talked
about
the
pong,
API
gateway.
Another
thing
worth
mentioning
is,
you
know
we
support,
let's
encrypt,
so
that
makes
a
deployment,
especially
public
deployment,
very
easy.
So
the
all
of
you
all
of
my
API
endpoints
I
exposed
were
HTTP
endpoints
and
their
real
certificates,
which
is
very
nice
in
a
production
employment
deployment.
C
You
can
hook
this
up
to
an
open
tracing
back
end,
so
you
can
actually
trace
functions
throughout
the
system,
which
is
very
important
for
debugging
I've
mentioned
that
we
have
multiple
IDPs
I
didn't
really
demonstrate
that
the
cluster
that
I'm
using
right
now
is
hooked
up
to
github.
So
I
use
my
github
credentials
to
log
in.
We
also
support
multiple
organizations
and
organizations.
Surrett
is
it
you
know
kind
of
a
hard
tendency.
C
So
then
you
know
I'm
just
looking
at
the
whole,
it's
sort
of
hard
tendency
construct
so
that
you
know
the
use
case
being.
If
you
deploy
dispatch
in
a
large
organization
like
VMware,
you've
actually
specified.
Hey
I
want
to
see
nab
you
to
have
this
organization.
I
want
CMU
and
this
organization,
and
they
can.
They
can
essentially
use
dispatch.
C
Without
interfering
with
each
other,
we
have
our
back
policy,
so
you
can
have
different
roles
for
your
administrator
dispatch,
as
suppose
to
your
developer.
Let's
say
you
want
to
give
your
administrators
the
ability
to
manage
the
images
based
images
where
you
want
to
developers
to
just
get
some
execute
functions.
You
can
do
that
within
dispatch
already
mentioned
that
it's
cloud
events
native
and
you
have
this
event-
driver
this
extensible
event.
C
Driver
infrastructure
which
I
haven't
I
haven't
demonstrated,
but
that's
kind
of
alluded
to
the
the
demo
that
that
mark
was
talking
about
that
from
Q
Khan,
where
we
can
pull
in
events
from
programs
sources
and
we
provide
the
drivers
for
that.
They're
also
very
easy
to
write
your
own
driver
and
then
I
demonstrated
part
of
the
open,
serviceworker
integration.
C
This
is
a
open
source
project.
We
take
any
contributions
that
we
can
we'd
love
you
to
become
our
community
and
help
out
we're
always
looking
for
people
within
within
VMware
and
outside
of
VMware.
If
they've
got
use
cases
for
server
lists
or
functions
or
dispatch,
specifically
we'd
love
to
hear
more
and
yes,
that's
it
for
contributing
here's.
A
list
of
links
associated
with
the
content
that
we
we
just
showed
most
important,
probably
being
dispatched
framework,
dot,
IO
and
I
guess
does
that
complete?
That
concludes
things.
So
we've
got
a
got
some
time
for
Q&A.
A
C
So
the
images
themselves
will
remain
as
indefinitely
or
as
long
as
that
function
itself
is
defined.
So
if
I
go
through
and
explicitly
delete
a
function
through
dispatch
that
will
delete
the
image,
it
will
also
deep
provision
that
container,
which
the
faz
is
managing
I'm.
With
regards
to
how
long
that
container
is
scheduled
by
the
faz
is
more
or
less
fast
specific,
so
some
faz
is
like
open.
Faz
will
always
have
at
least
one
container
running
for
that
function.
C
Definition,
whereas
something
like
riff,
which
scales
all
the
way
to
zero,
will
have
some
sort
of
timeout
where,
if
the
function
is
not
invoked
after
a
certain
period
of
time,
you'll
scale
to
zero.
As
soon
as
it
is
invoked
again,
there
will
be
a
little
bit
of
a
warm-up
time
to
be
provisional
that
container
and
serve
the
function
and.
A
I
think
what
you're
talking
about
there
is
is
really
important
to
iterate
on
the
the
warmup
time
that's
needed
if
a
container
is
not
running
and
you're
trying
to
hit
that
function.
Of
course,
there
will
be
a
little
waiting
time
before
that
function
is
ready,
but
but
usually
not
that
long
right,
yeah.
C
Like
I
said,
it
is
kind
of
fast
specific,
with
with
open
faz.
You
will
hardly
notice
any
because
that,
because
that
function
will
already
revision
where
it's
like
rib,
you
know
the
the
process
for
the
fur
provisioning.
That
function
is
it's
actually,
you
know
a
kubernetes
deployment
more
or
less,
and
so
it
takes
it
takes
a
little
bit
of
time
and
it
is
language
dependent
because
a
job
a
function.
B
This
is
also
one
of
the
areas
where
functions
make
it
easy
that
you
don't
even
have
to
think
about
how
that
docker
container
scheduled
it
gets
deep
revision,
etc.
That
you
know,
with
docker
I
need
to
actively
say
to
run
this
container
with
functions.
I
just
wait
for
the
events
to
come
in
and
the
functions
of
the
service
takes
care
of
the
container,
orchestration
and
provisioning
of
that.
It's.
C
True
I,
don't
have
you
don't
have
to
write
a
kubernetes
deployment
in
order
to
schedule
that
function
or
schedule
that
that
container
I've
been
using
kubernetes
for
quite
some
time,
I
couldn't
tell
you
all
of
the
fields
by
heart
that
you
work
that
are
required
in
a
kubernetes
deployment.
I'm
still
copying
pasting
examples,
and
so
this
is
one
of
them.
This
is
why
we're
trying
this
is.
This
is
why
a
project
like
dispatch
exists
and
by
server
list
exists
is
to
make
that
easier
for
developers.
A
So
I
want
to
talk
a
little
about
the
results
from
the
poll
here.
Thank
you
so
much
everyone
for
the
for
the
participation
in
the
poll,
so
the
the
second
question
we'll
start
with
the
second
question
here:
how
soon
will
you
be
deploying
on
premises,
server,
less
functions?
We
had
a
few
that
talked
about
deploying
them
within
three
months.
We
had
a
few
as
well
that
talked
about
within
half
a
year
and
then
the
large
majority
said
don't
know.
A
C
Yeah,
no
I'm,
not
surprised
at
all
the
the
technology
is
new.
The
vast
majority
of
folks
using
serverless
are
using
AWS
lambda
and
there
have
been
other
services
with
other
cloud
providers
that
you
know
they
haven't
had
the
same.
They
haven't
seen
that
much
uptick
yet
as
well
as
a
lot
of
unfriend
solutions
that
also
haven't
seen
a
ton
of
like
real
sort
of
usage.
I
think
it
is
very
new
technology.
It
requires
sort
of
eight,
it's
a
different
design
pattern.
C
B
We
have
been
kind
customers
and
they
definitely
see
the
value
in
server
list
that
they
they
understand
why
they
want
this
and
have
some
initial
use
cases
for
being
able
to
put
this
in
place
in
their
infrastructure.
But
you
know
to
Berens
point:
it
is
early
days
with
respect
to
on-prem
functions
as
a
service,
and
you
know
what
we
wanted
to
do
with
dispatch
was
provide
something
that
was
slightly
opinionated
in
terms
of
helping
to
ease
the
installation
in
providing
the
tools
necessary,
such
as
API
gateway
cloud
events,
etc.
C
Yeah
no
I
I
could
have
demoed
it.
It
is
we
we
managed
that
altar
dispatch,
so
I
I,
think
I
showed
the
the
yam
will
file
for
it.
If
I
had
just
changed
the
name
and
hit
just
about
to
create,
it
would
have
provisioned
and
been
created.
Both
the
new
database
and
the
binding
for
that
that
is,
it
is
all
dynamic.
It
is
not
provisioned
at
the
time
of
function,
run
like
it
is
something
that
you
do
provision
before
you're
going
to
execute
that
function,
particularly
because
you
know
it
takes
well.
C
B
But
also,
if
you
have
events
coming
in,
you
wouldn't
want
the
function
invocation
to
be
creating
the
database
right
right
and
typically,
what
we
see
is
that
you
will
bring
up
the
supporting
services
alongside
when
you're,
creating
the
Earth.
You
know
the
functions,
etc
in
order
to
now
deploy
your
entire
web
application
or
function
plus
services.
So
that's
typically
not
done
as
part
of
the
function
itself,
but
it
is
just
part
of
bringing
up
the
application.
A
C
So
we
and
we
should
so
we
leveraged
a
project
called,
but
it's
actually
a
kubernetes
project,
so
I
think
it's
an
incubator
still,
but
it's
the
Service
Catalog,
which
is
sort
of
a
you
know.
It's
an
interface
from
kubernetes
to
open
service
brokers
and
open
service
brokers,
all
implement
the
open
service
broker,
API
and
so
there's
really
no
limit
to
which
open
service
brokers
we
we
can
interface
with
I've.
C
A
Awesome
so
no
more
questions
there.
So,
let's
go
back
to
the
poll.
The
the
first
question
we
had
in
the
poll
were
was
have
you
ever
used
or
deployed
a
serve
release
function
and
we
got
about
a
50/50
percent
response
and
yes
or
and
and
no
and
we
had
one
person
respond,
don't
know,
that's
perfectly
fine,
because
you
might
have
used
one
without
knowing
it
so
yeah
that
question
maybe
should
have
been
phrased
a
little
differently.
A
So
I've
recently
held
two
workshops,
one
workshop
and
one
hackathon
regarding
functions
as
a
service,
and
they
it's
a
it's
a
huge
it's
a
huge
impact
when
it
well
not
impact
but
huge
interest.
Rather
in
learning
about
server
less
and
function
as
a
service.
Where
would
you
recommend
someone
starting
out
to
go
and
learn
more
about
this
I
mean.
A
C
You
know
if
you,
if
you
want
to
use
a
hosted
service,
you
know
the
public
clouds
have
have
faz
that
you
can
play
around
with
if
you're
specifically
interested
in
something
that
you
have
a
little
bit
more
control
over.
What
I
want
to
run.
Something
on
primer
on
your
laptop
I
would
suggest
open
path,
not
just
because
you
know
Alex
Alice's
is
part
of
VMware
now,
but
also
because
its
intent
is
to
be
simple
and
to
help
people
learn
about
service
and
fast,
and
it's
got
a
community
and
it's
a
great
place
to
start.
A
Yeah
I
agree,
they're
an
open
fast.
That's
what
we've
used
both
for
the
workshop
and
the
hackathon,
and
the
open
fast
also
has
a
workshop.
It's
all
on
github
and
there
you
can
follow
along.
It
takes
a
few
hours
and
it's
all
self-paced,
it's
very
simple
and
you
walk
away
with
having
a
full-on
open,
fast
implementation
on
your
laptop.
So
it's
it's
pretty
neat
and.
B
I'd
say
the
the
distinction
between
open
paths
and
dispatch
is
that
you
know
confounds
is
a
wonderful
project
for
being
able
to
run
these
functions
on
your
laptop
or
even
even
you
could
even
use
it
in
production,
but
what
dispatch
is
attempting
to
do
is
pull
together.
Some
other
building
blocks
and
also
looking
at
the
enterprise
use
cases
that
it
open
founds
is
not
addressing
well.
A
C
Mean
you
can
make
an
argument
that
an
HTTP
request
coming
through
an
API
gateway
is
an
event,
so
I
think
I,
don't
think.
There's
necessarily
a
distinction
between
the
two.
The
only
maybe
distinction
that
I
would
create.
That
I
would
make.
Is
that
HTTP
requests
in
general
are
synchronous,
and
so
the
whatever
interface
you've
got
with
the
API
gateway
needs
to.
C
You
know
both
trigger
the
function
and
then
also
wait
on
the
response
and
then
serve
that
back,
but
I
think
that
the
reason
that
most,
the
reason
that
most
events
are
triggered
through
the
API
gateway
is
because
a
lot
of
people
are
using
are
using
faz
using
lambda
for
the
use
case
of
building
web
apps
and
web
hooks
in
creating
art.
Creating
API
is
like
I
thought
that
is
either
that's.
C
My
I've
been
building
web
apps
for
a
long
time
like
that
is
my
primary
use
case
and
and
for
you
know,
creating
something
very
simple
like
the
the
Notes
app
or
like
slack
integration
or
whatnot.
It's
a
lot
easier
to
just
write
a
small
function,
exposing
their
link,
a
single
API
up
to
that
function
and
then
use
that
as
your
your
external,
publicly
accessible
endpoint
I
think
that
it's
a
it's
a
solid
use
case
and
I
think
that's
that's
reflected
and.
B
A
C
It's
definitely
not
a
VMware
version
of
these
projects.
I
mean
dispatch,
tries
to
be
something
different,
I
think
we
are
leveraging
all
so.
A
lot
of
one
of
the
ways
I
describe
dispatch
is
that
it
is
the
the
stack
of
technologies
that
you
would
have
to
supply
or
build
if
you
were
going
to
put
fast
in
your
enterprise,
environment
and
we've
tried
to
wrap
those
all
up
together
as
a
single
package
in
dispatch,
we
are
not
trying
to
you
know.
It
is
not
a
version
of
call
right.
A
C
A
C
A
There
we
go
awesome
well,
thank
you
both.
So
much
there's
been
really
really
interesting.
Thank
you.
You
all
the
attendees
for
all
the
questions
and
everything
that
you
posted
today
really
really
appreciate
it.
We
wanted
this
to
be
interactive
and
it
definitely
was
so
again.
Thank
you,
we'll
post
this
up
on
YouTube
as
soon
as
we're
done
here,
awesome
have
a
great
day.
Everyone.