►
From YouTube: OpenShift Commons .EDU SIG The Academic Environment and Cloud Native Technology (UMich)
Description
UMich Chris Kretler and Bob Killen on Academia and Cloud Native Technologies
B
A
Some
of
the
stuff
out
of
the
way
you
know
my
name
is
Bob
Dylan
I'm,
seeing
a
research
cloud
administrator
here
at
Michigan,
I
work
or
is
advanced
research,
computing
and
Technology
Services.
Our
group
essentially
handles
the
computational
research
needs
for
the
entire
universe,
whatever
they
may
be
classic
batch
system.
C
And
I'm
Kristin:
are
you
talking
a
little
bit
about
the
teaching
and
learning
applications
or
openshift
and
kubernetes
here
at
the
university
I?
Don't
have
as
catchy
of
a
title
as
mr.
Bobby
tables
you'll
just
have
to
remember
me:
a
secret
or
I
do
business
planning
for
our
central
IT,
and
so
over
the
last
couple
years.
That's
involved
standing
up
a
service
to
host
applications
that
want
to
be
containerized
and
kind
of.
C
Our
latest
thing
is
that
we're
starting
to
move
our
web
hosting
environment
into
containers,
and
so
when
Stephen
comes
by
and
he
starts
talking
about,
WordPress
I'm
gonna
be
very
interested
in
hearing
what
he
has
to
say.
But
if
you
want
to
get
in
touch
with
me,
my
email
address
is
listed.
That'll
slide
as
well.
A
B
A
About
that
okay
I'll
start
over
at
the
beginning
of
this
little
car,
then,
though,
there
are
really
sort
of
like
three
pillars
of
academia,
at
least
when
it
comes
to
the
IT
side
and
what
you
have
institutional
support,
which
is
your
central
IT
teaching
and
learning
which
Chris
will
be
able
to
talk
significantly
more
I.
Can
then
research,
although
research
could
technically
be
considered
optional
depending
on
the
educational
institution,
but.
C
A
B
A
That
they
tend
to
host
many
sort
of
off-the-shelf
applications
managing
those
off-the-shelf
applications.
A
lot
of
those
are
also
sort
of
central
services.
For
you
know,
logon
services,
Active
Directory,
at
printing,
there
are
critical
to
the
entire
institution
like
H
a
and
business
continuity
is
a
critical.
A
A
So
when
it
comes
to
like
I,
am
an
SSO,
this
tends
to
be
a
lot
more
complicated
than
people
think
and
tends
to
account
for
like
a
quarter
to
a
third
of
like
IT
infrastructure
and
resources.
This
will
make
a
little
bit
more
sense
when
I
rattle
off
some
stats
here
about
Michigan
have
19
schools
in
colleges,
45,000,
students,
undergrads
thousand
faculty
32,000
staff,
not
including
36,000
hospital
staff
and
also
doesn't
include
like
alumni
and
retirees
they
get
to
remain
to
retain
their
identity.
A
Some
campus
resources,
additionally,
there's
other
things
like
group,
Federation's,
so
visiting
faculty
sponsor
affiliates
all
those
other
fun
things.
Oh
managing
is
really
very
large
and
complicated
portion
of
all
this
analysis
actually
handled
under
sort
of
like
pure
services
to
us,
a
internet2
initiative
like
trust,
identity
and
research,
and
that
sort
of
manages
like
three
big
tools
that
you'll
find
and
just
about
every
institution
in
one
way,
shape
or
form.
That's
Shibboleth,
sort
of
the
edu
SSO
system,
grouper
and
co-manage
chip
is
honestly
kind
of
antiquated.
At
this
point,
there's
no.
B
A
A
So
then,
structure
app.
This
tends
to
be
I'm
hosting
a
lot
of
times.
It's
gonna
be
like
hosted
in
VMware.
There
might
be
one
or
you
managed
OS
offerings
like
well.
You
know
some
windows
flavor,
there's
very
little
to
no
adoption
of
virtual
networking
networking
policies
as
far
as
like
hosted
offerings
go.
These
tend
to
be
limited
to
like
Oracle
and
Emmas
SQL,
with
some
support
for
MySQL
and
Postgres.
A
A
So
what
would
adoption
of
sort
of
the
cloud
native
stuff
how
native
tools
enable
the
big
thing
is
like
adoption
of
containers,
craze
and
all
that
really
delivers
on
the
idea
of
like
build
once
build
once
run
anywhere
and
offers
a
like
normalized
support
platform
by
being
a
unified
API
for
no
on-prem
and
cloud
deployments?
One
of
the
things
that
I've
seen
to
shion's
is
like
when
going
to
the
cloud
institutions
tend
to.
B
B
A
Think
they
might
have
to
like
reteach
and
retool.
So
if,
like
we
select
AWS
in
a
couple
of
years,
you
know
we
might
have
to
go
to
Asia
or
something
like
that
means
we
have
to
throw
everything
out
and
we
learn
how
to
do
everything
in
Asia
and
by
building
towards
OpenShift
building
on
urban
at
ease
its
same
api
everywhere.
So
you
start
teaching
people
how
to
use
this
stuff.
It
makes
things
significantly
easier
to
someone
else
and
offer,
like
then
sort
of
P
packing
on
that.
We
also
have
like
improved
logging
and
metrics
collection.
A
Then
like
talking
about
like
spoke
to
long
ago
with
the
fluency
manage
logging
project,
we
can
actually
use
that
single
agent,
you
log
everything
and
it's
enough
to
multiple
systems
and
we've
actually
ditched
the
Splunk
agent
completely
here
because
like
for
us,
this
blink
agent
was
actually
like
producing
more
logs.
Then
they
cared
about
not
to
mention
like
it.
A
One
like
no
I
will
say
at
least
on
all
this
combined.
It's
like
since
many
institutions,
sort
of
the
first
go-around
with
adopting
option
of
lakes
or
cloud
native
methods
and
practices
so
building
immutable
systems,
building,
Packer
pipelines
and
pushing
stuff
out
there.
It's
a
great
time
for
them
to
actually
start.
C
I'm
gonna
talk
a
little
bit
about
how
we're
seeing
OpenShift
kubernetes
adoption
here
at
the
University
and
in
order
to
do
that,
I'm
going
to
provide
a
little
bit
of
historical
context
on
how
the
teaching
and
learning
space
was
covered
kind
of
technically
teaching
learning
space
for
many
years
was
dominated
by
the
running
of
the
learning
management
system
on
campus
and
while
that's
still
a
core
component
of
the
ecosystem
of
learning
of
teaching
and
learning.
It's
not
really
provided
on
campus
anymore.
C
C
You
know
monolithic,
Java,
applications
running
in
a
huge
stack
and
for
awhile.
The
learning
management
system
was
the
cutting
edge
of
technology
as
teaching
and
learning
was
provided
to
campus.
However,
I
think
what
we
started
to
see:
2009
2010
there
started
to
be
some
additional
tools
that
came
out
like
Piazza,
which
was
an
online
discussion
forum
that
was
pure
sass.
C
They
wanted
to
be
able
to
plug
into
the
learning
management
system,
I
think
over
the
course
of
2009
2010
2011
2012,
the
the
market
space
really
started
to
evolve
more
towards
SAS,
so
desire2learn
canvas
have
come
into
the
space
and
what
the
individual
universities
are
doing
is
much
less
focused
on
the
learning
management
system.
We
just
consume
SAS
based
learning
management
systems
now
and
we're
focusing
on
some
of
the
other
things.
C
So
in
the
case
of
great
cat
gradecraft,
which
is
a
gameful
learning
learning
management
system
that
runs
on
kubernetes,
that's
game,
polar
ting
is
a
new
way
of
looking
at
the
way
that
a
student
is
going
to
move
through
the
classroom.
So,
instead
of
having
a
set
number
of
assignments
that
they
go
through
prior
to
taking
a
quiz,
you
have
one
chance
to
take
a
quiz.
You
have
one
chance
to
take
a
test
in
grade
craft.
You
have
multiple
paths
to
get
through
course
material.
C
So
you
have
alternative
assignments,
kind
of
based
on
what
your
aptitude
and
based
on
what
your
interest
level
is
within
a
subject.
When
you
get
to
a
quiz,
you
have
a
couple
of
times
to
take
the
material.
The
ideas
based
on
eventual
mastery,
as
opposed
to
a
one
time
snapshot
of
where
you
are
in
regards
to
the
material
and
so
you
know
grade
craft,
is
one
of
the
LMS.
C
Is
that
it's
in
the
space
of
gameful
learning
grade
craft
and
the
multiple
paths
through
the
course
material
really
breaks
the
brain
of
your
traditional
LMS
like
a
canvas
like
a
Sakai
like
a
blackboard,
and
so
they
don't
integrate.
Well
too,
although
sometimes
you
can
integrate
content
for
a
particular
course
as
a
plugin
into
something
like
canvas
via
LTI,
and
then
your
students
would
just
go
to
gradecraft
for
all
of
their
course
material,
as
opposed
to
it
being
a
tool
within
the
learning
management
system.
C
So,
on
the
next
slide,
yeah
I
talked
a
little
bit
about
gameful
learning
and
the
other
example
I
mentioned
was
flipped
classroom.
So
when
I
came
to
school,
it
was
here
at
the
University
of
Michigan
and
I
was
expected
to
show
up
for
lectures
and
the
instructor
got
up
in
front
of
the
classroom
and
talked
for
45
minutes
about
economics
or
whatever
and
I
was
given
assignments
to
take
home
and
use.
C
So
Coursera
futurelearn
EDX
are
all
in
the
MOOC
space,
where
these
are
the
massive
open
online
courses
where
universities
have
the
ability
to
expand
their
scope
beyond
just
the
students
that
they
have
on
campus.
Do
anyone
who
can
log
in
to
the
MOOC
worldwide,
and
so
over
the
last
few
years
we've
seen
evolution
of
the
business
model
of
the
MOOC
from
completely
free
content.
C
To
now
you
have
some
certification
programs,
and
now
universities
are
getting
into
the
space
where
your
credentialing
and
operator
repro
programs
online
through
the
MOOC,
and
that's
something
that
we
here
at
the
university
are
starting
to
look
at.
There
are
some
risk
factors
there,
because
you
don't
want
to
take
away
from
the
educational
experience
and
the
value
that
you're
providing
to
your
brick-and-mortar
customers.
C
In
this
case,
which,
in
the
case
of
the
university,
are
the
students
who
are
on
campus,
if
you
can
get
the
same
educational
experience
for
a
fraction
of
the
cost
through
the
MOOC,
why
would
you
pay
for
the?
Why
would
you
pay
for
the
on-campus
experience,
and
so
you
know
they're
still
working
out
kind
of
the
business
model
for
how
that
makes
sense,
and
what
the
price
point
is
for
a
completely
credential
program,
but
in
the
meantime,
universities
are
kind
of
focusing
on
well.
C
How
do
we
add
value
to
this
MOOC
experience,
and
what
do
we
want
to
get
out
of
the
MOOC
experience,
in
addition
to
the
experience
that
we're
providing
to
students
and
so
kind
of
the
edge?
There
is
the
data
that
streams
out
in
the
MOOCs
in
the
form
of
data
analytics
that
are
consumed
in
a
couple
of
different
ways.
There's
also
some
research
that's
going
on
into
how
to
make
other
classes
other
than
the
stem
classes
available
and
viable
at
a
MOOC
level.
C
So
for
your
language,
based
courses,
how
do
you
do
peer
review
and
grading
there's
a
icon
on
that
slide?
There
called
M
right.
That's
one
of
the
big
projects
that's
going
on
here
at
the
University,
for
how
do
you
do
large-scale
peer
review
for
language
based
courses
either
for
introductory
online
courses
on
campus
or
at
the
MOOC
level,
and
that's
a
project,
that's
running
that
we
have
running
here
in
OpenShift,
but
there's
also
a
large
data
analytics
component
that
goes
into
that.
C
How
do
you
come
up
with
intelligent
algorithms
for
seeing
what's
within
a
language
based
assignment
and
responding
to
that?
There's
also
coaching
tools
that
are
in
use
in
some
of
these
MOOCs
as
well.
So,
in
addition
to
kind
of
administrative
tools
that
go
on
at
the
MOOCs,
there's
also
the
course
content
themselves,
and
this
is
something
that
you
see
if
you
ever
take
a
Python
or
a
data
analytics
course
in
Coursera,
you're,
probably
going
to
be
piped
into
a
Jupiter
notebook
and
the
Jupiter
notebooks.
Are
they
run
well
in
the
kubernetes
infrastructure?
C
I
know
here
within
our
School
of
Information
they're
running
a
lot
of
Jupiter
notebooks
Jupiter
hub
I'm
kubernetes,
and
a
lot
of
the
Python
assignments
are
then
through
a
parsing
text
tool.
That's
part
of
a
continuous
integration
pipeline,
though
less
slide
for
me
Bob,
so
data
analytics.
This
is
really
where
our
teaching
and
learning
group
here
on
campus
is
focused
in
addition
to
a
lot
of
researchers
and
I.
C
Don't
think
it's
not
unique
to
what's
going
on
at
the
University
as
you
have
your
campus
students
taking
classes
in
a
lot
of
different
formats
online
through
LMS,
like
canvas
through
a
MOOC
like
Coursera.
How
do
you
standardize
the
event
stream
of
your
students
accessing
course,
content
in
a
common
way,
so
that
you
can
consume
that
and
start
to
make
predictive
analysis
of
whether
your
students
going
to
be
successful
in
a
course
or
not?
C
If
you
pull
information
back
into
a
data
warehouse-
and
you
want
to
report
on
what
students
are
doing
in
order
to
make
predictive
analysis
in
order
to
learn
when
you
need
to
intervene
when
a
student's
not
doing
well,
you
need
to
speak
a
comments
in
taxes.
So
that's
what
the
caliper
standard
is
for.
So
when
you
have
a
common
framework
for
defining
the
educational
events,
then
that
feeds
back
into
a
data
warehouse.
The
university
is
part
of
a
consortium
called
unizin.
C
That's
standing
up,
something
called
the
unit
of
data
platform
which
feeds
in
university
specific
information
that
we
have
access
to,
but
it
also
has
de-identified
information
from
other
units
and
schools
that
our
researchers
can
use
for
the
basis
of
understanding.
You
know
those
factors
that
I
was
talking
about
earlier.
When
do
you
intervene
and
a
student?
How
do
you
predict
whether
a
student's
going
to
do
well
at
a
particular
class?
C
This
is
kind
of
the
edge
of
what
we're
trying
to
do
with
data
right
now
and
once
you
have
all
that
information
in
a
common
data
warehouse,
then
then
the
edge
is
building
the
tools
that
use
algorithms
to
analyze
the
data
and
make
those
predictions,
and
so
we
have
tools
that
are
running
an
open
shift
in
from
primarily
an
open
shift
on
campus.
That
are
where
the
open
shift
continuous
integration
pipeline
is
a
real
nice
value.
C
A
I'll
skip
over
all
that
cuz.
You
free
much
covered
that
I'll
dive
into
sort
of
my
bread-and-butter
area.
Actually,
first,
ball
is
my
mic
sounding
better
now
and
go
yes,
hopefully
anyway,
so
research.
This
is
this
is
where
I
work
every
day
and
it
is
for
the
the
wild
wild
west
of
university
IT
for
a
lot
of
reasons.
A
First,
for
us,
researchers
can
actually
like
own
their
resources
unless
it
happens
to
be
like
a
state
grant
the
grants
go
to
the
researcher,
so
they
can
own
their
hardware.
It
can
go
to
whatever
cloud
provider
they
want.
It
can
be
very
difficult
to
wrangle.
The
other
thing
is
like
some
grants.
These
days
will
come
with
credits
for
specific
cloud
providers,
and
it
also
means
when
they
leave
they
get
to
take
it
all
with
them.
A
Well,
it's
up
to
you
like
us,
as
a
research,
IT
support
staff,
sort
of
entice
the
risa,
the
researchers
to
use
our
centrally
managed
services.
That
might
be
a
bit
easier
for
them
to
consume,
and
we
can't
deliver
that
they're.
Just
gonna,
you
know
do
something
on
their
own.
The
other
thing
is,
a
lot
of
their
tools
can
be
costly,
written
in
just
about
anything.
A
It's
it's
very
sort
of
particular
about
oh,
but
it's
running
and
where
it
is
and
honestly
the
the
burden
of
making
it
work
tends
to
fall
to
the
undergrads
and
the
students,
their
researcher
just
wants
it
to
run,
and
the
students
are
the
ones
charged
with
making
it
work,
but
in
at
least
the
the
recent
computational
research
side,
there's
sort
of
three
big
fields
where
all
this
stuff
can
sort
of
fit
in.
We
have
our
data
science.
This
is
your
classic.
A
Like
new
spark
flanked
data
streaming,
it's
tend
to
be
what
people
think
of
when
they
hear
like
machine
learning
and
big
data
you
have
HTC
or
high
throughput
computing.
This
is
mostly
embarrass
lis,
embarrassingly.
Parallel
computing
good
examples
like
SETI
at
home,
and
then
you
have
each
PC,
which
is
sort
of
not
cloud
front
friendly
and
very
sort
of
batch
scheduler
and
require
a
specific
hardware.
A
Cool
thing
about
all
this
stuff
is
like
all
of
it
can
actually
run
on
top
of
like
use
of
cloud
native
methods
and
tooling.
It's
just
certain
things
are
better
than
others.
In
the
data
science
side
it
doesn't
work
extremely
well
both
on
Prem
and
in
the
cloud
and
adoption
like
sort
of
cloud
methods
can
be
better
because,
like
there's,
some
issues
like
spark
and
all
that
don't
handle
multi-tenancy
well,
so
we've
seen
a
growing
adoption
force
like
spitting
up
multiple
instances
of
it,
and
this
is
very
easy.
A
If
you
happen
to
see
or
attended
kind
of
you
note,
certain
did
a
great
demo
how
they're,
using
criminais,
Xin
Federation
to
use
condor
serve
their
big
scheduler
and
a
lot
of
this
stuff
is
like
and
they
she
was
like
very
job
focused
and
the
cool
thing
about.
It
works
like
anywhere
there's
available,
CPU,
there's
no
application,
checkpointing
or
anything
like
that.
It
really
works
equally
well
on
cloud
or
you
know,
in
a
cloud
or
on
Prem,
on
top
of
communities
on
per
metal.
It
really
doesn't
matter.
It's
just
this.
A
As
many
like
this,
as
many
courses,
we
can't
HPC
is
a
little
bit
more
of
a
challenge.
It's
it
tends
to
be
job
focused
like
HTC,
but
is
a
classic
multi-user
environment
where
everyone
has
their
own
UID
and
GID,
and
their
logging
and
doing
stuff
there's
also
more
specialty.
Hardware
like
InfiniBand
low,
low
latency,
like
hyper
horn,
shared
file
system
like
austere,
gpfs,
RDMA.
The
other
thing
is
like
systems
and
to
never
go
down
by
that.
A
I
mean
like
they
tend
to
go
down
once
or
twice
a
year
for
maintenance
and
jobs
themselves
can
actually
like
take
from
minutes
to
weeks,
lead
us
here.
Personally,
we
have
a
28-day
wall
time,
so
a
job
is
allowed
to
run
up
to
28
days
before
it
can
be.
You
know,
optionally
killed
these
sort
of
three
domains
that
like
at
least
it
when
it
comes
to
computational
research
like
what
your
researcher
would
normally
think
of
in
space.
A
Well,
they
don't
think
of
is
all
the
other
tooling
that
is
involved
in
just
running
those
things
are
running
them.
Well,
so,
on
the
like
HPC
side,
they're
going
to
be
running,
Postgres
they're
gonna
be
running
like
I'm,
going
to
be
usually
nginx
they're
also
gonna,
be
like
likely
running,
get
lab
or
something
that
to
keep
track
of
all
their
stuff.
A
They
might
be
running
squid,
which
is
commonly
used
both
in
the
HTC
nhbc
side,
just
to
help
cache
all
of
those
resources
that
they'll
be
using
they'll,
be
using
graph
on
to
visualize
their
stuff
elasticsearch
for
logging
and
the
data
science
side.
All
well
all
may
actually
use
like
Jupiter
hub
pretty
heavily
and
data
science
side
is
as
I
sort
of
like
talked
about
before
it's
already
like
adopted
this
stuff
really
really
well.
A
Argo
jenkins,
all
those
aren't
gonna,
be
run
on
any
one
of
those
systems,
but
they
need
some
place
to
run
along
like
long
side
them,
and
this
was
like
a
whole
suite
of
data
management
tools
and
a
whole
bunch
of
other
custom
things
that
are
written.
They
all
need
a
home
to
run,
and
that's
where
you
know,
adoption
of
containers
and
granese
all
this
stuff
can
really
be
beneficial,
and
I
started
to
talk
about
this
on
the
previous
slide,
but
like
what
does
adoption
of
this
stuff
really
enable?
A
And
it's
really
like
everything
from
the
institutional
support
side,
plus
one
whole
idea
of
one
api
to
rule
them
all
being
able
to
you
package
up
your
application.
Your
script,
your
generated
model
and
just
put
it
anywhere,
is
extremely
useful.
Both
from
a
institutions
support
side,
we
don't
have
to
set
up
anything
custom
for
them
to
run
their
stuff
from
the
researcher
side.
They
do
have
no
leave
the
institution
or
want
to
run
it
someplace
else.
They
can
pick
it
up
and
take
them
with
them.
It's
no
big
deal.
A
We
also
see
what
general
adoption
of
the
containers
know
is
much
better.
Full
engine
support
there
so
like
every
day,
there's
a
new
tool
in
this
space,
and
it
makes
it
it's
it's
possibly
keep
up
with
it
all
and
like.
Lastly,
the
the
big
thing
it
really
enables
is
the
whole
idea
of
a
fully
reproducible
research
pipeline.
It's
much
easier
to
just
hand,
someone
a
container
for
them
to
run,
and
it's
like.
A
Okay,
you
need
to
install
this
version
of
this
library
and
this
version
of
this
compile
it
and
then
pick
it
up,
and
hopefully
it
runs
on
platform.
If
not,
you
might
have
to
know
backrub.
Your
version
of
linux
grab
a
different
lab
version
of
like
GFC,
and
do
it
all
over
again
now
there
are
definitely
some
challenges
with
getting
the
outreach
and
adoption
of
sort
of
tools
and
these
methods.
A
One
thing
we
are
starting
to
see
a
little
bit
of
shift
away
from
from
papers,
and
you
know
give
me
a
paper
give
me
your
code
is
now
like
it.
Give
me
well
give
me
a
codes,
but
give
me
a
repo
give
me
your
container.
Give
me
a
way
to
reproduce
this
a
lot
easier
easily
on
the
conference
side,
but
a
lot
of
the
academic
conferences
are
catered
to.
Academics
and
other
academics.
A
Submissions
generally
require
a
paper
workshops
and
tutorials
are
everything
that
don't
necessarily
require
a
lot
like
my
submission
to
you.
Supercomputing
18
for
a
kubernetes
tutorial
had
an
8
page
submission,
which
you
know
at
least
in
the
like,
compared
to
cheap
con,
and
so
there
any
of
the
other,
like
I,
won't
say
commercial
conferences,
but
any
of
the
other
conferences.
This
is
crazy.
A
This
really
started
like
back
in
the
2000s,
when
hiring
of
devs
and
encouragement
of
situational
development
was
really
curtailed
and
they
went
don't
buy
more
things
off
the
shelf,
and
this
is
definitely
starting
to
turn
around,
but
there
is
still
a
long
way
to
go
and
adoption
of
containers
kubernetes.
The
surrounding
pooling
really
offers
us
many
advantages
of
all
as
providing
like
an
avenue
for
us
to
give
back
and
I
really
encourage
lake.
Well
you're.
A
B
Don't
see
any
questions
in
the
chat
and
but
I
don't
want
to
discourage
people
from
asking
them.
This
has
been
really
good.
I
saw
Bob.
Give
this
talk
at
another.
Another
community
meeting
for
kubernetes
and
I
really
thought
it
was
important
because
it
laid
out.
You
know
the
three
pillars
and
the
way
people
the
different
aspects
of
what's
going
on
at
edu
and
IT
and
I.
Think
it's
a
good
level
setting
so
that
everybody
can
kind
of
see
it
I'm
curious
if
this
is
similar
to
other
people's
experiences
and
at
their
universities.