►
Description
Don’t miss out! Join us at our upcoming event: KubeCon + CloudNativeCon Europe 2023 in Amsterdam, The Netherlands from April 17-21. 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.
Lightning Talk: CNCF Batch Working Group Update - Alex Scammon, G-Research
This talk will present an update from the CNCF Batch System Initiative Working Group, a newly-created group set up to discuss batch scheduling conversation at the end-user level. It will focus on how the users and operators of today’s batch workloads currently interact with the many various cloud-native batch-related projects like Volcano, Armada, MCAD, Yunikorn, Slurm, HTCondor, etc.. Ideally, the hope is to provide some rough guidance and information for the CNCF community on these higher-level batch scheduling approaches since the landscape remains fairly opaque.
This presentation will discuss what the working group has worked on so far, what it's hoping to achieve, and (crucially!) how this working group is different (but closely related!) to the Kubernetes Batch working group.
A
And
group
update
hi
I'm
Alex
scammon
I'm,
one
of
the
co-chairs
for
the
cncf
batch
working
group.
You
can
find
me
at
stack
sacks
in
a
bunch
of
the
regular
places,
the
ones
that
I'm
interested
in
being
on
at
least
I
gave
a
brief
update
last
time
in
Valencia
about
the
same
topic.
In
that
conversation,
I
thought
that
it
was
practical
to
entertain
people
with
lots
of
pictures
of
very
cute
animals,
the
the
theme
being
that
these
are
batches
of
animals
and
in
English
we
have
collective
nouns
for
each
one
of
these
batches.
A
So
you
have
an
embarrassment
of
pandas,
a
prickle
of
hedgehogs,
a
waddle
of
penguins,
a
bed
of
sloths
and
a
flamboyance
of
flamingos,
and
that
was
how
I
how
I
kept
people
engaged
or
so
I
thought
I,
wouldn't
stoop
so
low
this
time.
But
at
the
end
of
that
conversation,
somebody
in
the
audience
asked
me
if
we
have
collective
nouns
for
kubernetes
resources.
A
So,
at
the
very
end
of
this
talk,
I'll
put
forward
a
humble
proposal
for
collective
nouns
for
things
like
pods
and
deployments
and
things
so
until
then,
I'm
going
to
talk
about
what
the
batch
system
initiative
working
group
actually
is
and
surprise.
Surprise,
it's
a
new
working
group
yay
another
place
to
come
and
talk
about
woes
or
thoughts
or
ideas
about
match,
scheduling
and
a
brief
history
of
this.
A
Earlier
in
the
year,
there's
a
group
called
tag
runtime,
which
a
number
of
people
in
the
audience
here
are
involved
with
Ricardo
and
Ricardo
and
Klaus
and
things
people
are
here
in
this
room
who
drive
that
conversation,
but
they
realized
they
were
having
conversations
around
back
scheduling
over
and
over
and
they
felt
that
there
was
a
place
for
a
discussion
around
batch
specifically,
and
so
we
started
to
talk
in
March
about
that
thing.
Batch
scheduling
as
its
own
separate
group.
A
Recently
in
September
of
this
year,
we
got
approved
as
a
bona
fide
cncf
working
group
thing.
So
just
so
you're
you're
aware,
if
you're
worried
at
all,
we
are
actually
approved
and
you
can
come
speak
with
us
as
a
as
a
real
cncf
body
and
what
we're?
What
we're
largely
trying
to
do
is
just
have
the
conversation
around.
A
What's
going
on
in
batch
scheduling
in
a
cloud
native
context
today
and
we're
trying
to
do
that,
I
want
to
call
here
is
like
an
application
level
level
which
is
different
from
kubernetes
native
level,
where
kubernetes
could
do
batch
scheduling
itself.
But
it
can't
right
now
so
there's
a
lot
of
Alternatives
have
been
created
or
have
been
worked
on
at
an
application.
Level
things
that
you
would
do
with
another
thing
to
allow
you
to
do
batch
scheduling
in
a
cloud
native
context.
In
kubernetes,
for
example,
there's.
A
Volcano
is
what
Klaus
founded
so
there's
a
whole
bunch
of
reactions
to
kubernetes
not
being
able
to
do
match
scheduling
today,
and
all
of
these
various
applications
have
grown
out
around
that
need.
So
we
we
sit
down
and
talk
about
that
and
all
of
our
shared
woes
and
pains,
but
not
just
the
project
maintainers.
A
And
so
it's
a
lower
level
conversation.
Whereas
for
us
in
the
cncf
batch
working
group,
we're
not
so
focused
on
the
lower
level
implementation
details,
but
we
we're
talking
about
the
higher
level
workings
of
of
the
tools
that
sit
above
kubernetes
or
on
top
of
or
with
kubernetes
I
hope
that
makes
sense
if
not
come
and
ask
me
again,
I'll
be
happy
to
try
and
explain
it.
A
So
what
we're
doing
currently
in
this
conversation
largely
we're
having
the
conversation
we
are
trying
to
bring
people
together
all
the
interested
parties,
so
everyone
in
this
room,
you're
here
today,
you're
interested
in
this
conversation
I,
can
tell
you,
wouldn't
have
shown
up
on
a
Monday
first
thing
to
all
the
way
in
Detroit.
If
you
weren't
interested
in
this,
so
everyone
here
should
actually
be
interested
in.
What
we
have
to
to
talk
about.
So
this
talk
right
now
is
is
mostly
a
welcome.
Please
come
and
chat
with
us,
so
that's
that's
number
one.
A
A
You
know
this
is
large
scale
stuff
and
we're
hoping
that
we
can
get
some
resources
to
put
to
put
something
together
for
people
to
come
and
mess
around.
So
that's
what
we're
currently
working
on
very
quickly.
If
you're
interested
in
this
conversation,
please
come
and
join
us.
We
meet
every
week,
I'm
sorry,
every
other
week
at
8am,
Pacific
Standard
time,
there's
a
zoom
channel
that
you
can
join.
If
you
join
the
Google
group,
you
should
magically
have
the
calendar,
invite
added
to
your
thing
and
you'll
be
forever
ours.
A
A
Our
emails
are
there
I'm
really
happy
to
talk,
and
then,
lastly,
as
I
promised
some
of
the
collective
nouns
that
I
just
tumbly
proposed,
could
work
for
us
at
a
pondering
of
PODS,
perhaps
a
cacophony
of
containers.
Anyone
a
catastrophe
of
clusters,
a
conundrum
of
custom
resources,
definitions,
positively,
a
delight
of
demon
sets
a
duchy
of
deployments.
A
Revolt
of
replica
sets
just
just
some
thoughts
if
you
have
any
others.
If
you
want
to
to
suggest
things,
please
Reach
Out
and
then
a
raft
of
otters
thanks
very
much.
That's
it
for
me.
Thank.
A
Do
you
think
about
that
yeah
I
was
trying
very
hard
to
not
say
kubernetes,
specifically
throughout
this
talk
and
just
Cloud
native
context.
In
practice
we
have
been
talking
a
lot
about
kubernetes
as
the
lower
level
substrate
for
orchestration,
but
I
I
think
that
we
would
like
the
conversation
to
be
broader
than
that.
Ideally,
you
know
we
will
be
hopefully
looking
at
other
cloud
context,
so.
C
A
I,
don't
think
we
graduate
I
think
it's.
We
were
spun
off
of
a
conversation
out
of
a
tag
or
out
of
a
say,
and
we
will
be
our
what
we
are
for
the
duration.
So
it's
not
like
we're
trying
to
get
to
some
other
magical
level.
We
are
a
working
group
that
wants
to
have
people
like
you
involved,
either
to
help
the
conversation
or
to
to
do
some
of
the
initiatives.
A
C
A
I
mentioned
the
kubernetes
batch
working
group,
and
so,
if
it's
specific
to
the
the
kubernetes
coding
that
Aldo
was
talking
about,
for
example,
the
the
discussion
is
happening
in
the
kubernetes
working
group
and
I
believe
there's
a
Sig
that
is
taking
on
the
the
developments
of
the
job
API
and
the
the
queue
project
and
things
like
that
for
us,
the
the
maintainers
of
those
projects
that
I
mentioned,
like
mcad
or
Armada
or
volcano
or
slurm
or
actually
Condor.
Perhaps
unicorn,
are
all
in
the
discussion
there.