►
From YouTube: CI WG demo: Microsoft Azure for Research (***)
Description
(***)
Date: 04/28/17
Presenters: Vani Mandava & Jeff Prosise
Institution: Microsoft
West Big Data Hub
A
So
with
that
I'd
like
to
present
the
Bonnyman
Dava
and
her
colleague,
Jeff
Prosise
they'll,
be
talking
about
Microsoft
Tudor
for
research,
bringing
the
car
of
the
cloud
to
big
data.
Bonnie
is
the
director
of
the
data
science
for
research
effort
at
Microsoft
Research
in
Redmond.
Her
role
in
the
Microsoft
research
outreach
team
is
to
enable
academic
researchers
and
institutions
develop
technologies
that
fuel
data,
intensive
science,
typic
research,
using
advanced
techniques
in
distance
and
data
mining.
A
She
holds
patents
and
service
infrastructure
and
after
spending,
the
bulk
of
her
career
in
engineering
finds
it
fascinating
to
work
alongside
researchers
and
scientists
and
Jeff
Prosise
is
the
co-founder
of
intellect
and
a
software
developer,
who's
written
nine
books
and
hundreds
of
magazine
articles
trained
thousands
of
developers
at
Microsoft
and
spoken.
Some
of
the
largest
of
software
conferences
in
his
former
life
is
mechanical
and
aerospace.
A
Engineer
Jeff
worked
at
Oak,
Ridge,
National,
Lab
and
Laura
and
Livermore,
where,
among
other
things,
can
develop
software
and
could
buy
thermal
and
structural
finite
element
methods
to
model
optical
systems
for
high-power
laser
beams
scientists
have
laser
beams
and
love.
It
today
is
thrilled
and
somewhat
amazed.
There
was
a
few
button
clicks.
A
You
do
approach
all
he's
been
up,
makes
2d
cluster
almost
as
powerful
as
the
crazy
used
to
work
on
I
love
it
well
and
I
want
to
say
big,
thank
you
to
Bonnie
and
the
Microsoft
team,
because
they
are
providing
Microsoft
desert
credit
in
total
750,000
direct
credit
to
the
South
hub,
three
million
across
all
the
hubs
and
they'll
tell
you
a
little
bit
about
this
in
their
presentation
and
we
will
be
rolling
out
the
how
you
can
apply
for
those
credits
in
the
coming
week.
So
with
that
Bonnie
and
Chuck
all.
B
C
Absolutely
Jeff,
thanks
to
the
South
big
data
hub
for
this
opportunity
to
present
of
the
community,
we're
excited
for
the
partnership,
layer
and
dan,
and
we
look
forward
to
what's
going
to
come
out
this.
This
collaboration
that
we
have
the
Big
Data
Jeff
is
going
to
give
you
an
overview
about
Asha
and
how
we're
using
it
for
research
and
I'll
do
a
deeper
dive
on
the
Big
Data
aspects
of
Asha,
and
some
of
the
successes
that
we've
have
out
of
the
programs
that
we've
run
over
the
last
tree
is.
B
C
B
And
share
my
screen
perfect,
so
Carl
that's
coming
up
now,
and
can
you
just
confirm
that
you
can
see
a
big
blue
slide
that
says
Microsoft
Azure
overview?
Yes,
indeed,
fantastic!
Well,
thanks
for
giving
to
talk
to
you
today,
as
Vani
said,
I'm
going
to
spend
10
to
15
minutes
just
with
not
so
much
a
general
overview
of
Azure
but
an
overview
of
it.
What
it
is
with
specifically
the
intention
of
addressing
some
of
the
things
that
it
does
for
data
scientists,
some
of
the
services
that
it
offers
that
might
be
useful
to
researchers.
B
B
Let
me
just
start
by
asking
a
question
and
that
question
is
what
is
the
cloud
kind
of
a
silly
question,
especially
for
researchers
who
have
been
using
it
for
a
long
time,
but
I
often
begin
presentations
like
this
with
this
slide,
because
it's
an
interesting
way
to
ask
people
to
think
about
what
the
cloud
was
just
a
few
years
ago
when
I
was
first
introduced
of
the
cloud,
it
was
about
storage
and
it
was
about
infrastructure
like
virtual
machines.
It
was
about
storage
and
compute,
and
a
lot
of
people
still
think
of
it.
B
That
way,
but
the
cloud
is
so
much
more
today,
what's
really
happening
in
the
cloud.
These
days
is
software
as
a
service
where
very
high
level
services,
encapsulating
machine-learning
stream,
analytic
stream
processing,
and
things
like
that
are
basically
made
available
to
you
with
the
button
click
in
the
introduction.
A
moment
ago,
Lea
said
I
find
it
amazing,
sometimes
that
I
can
go
into
the
azure
portal
and
spin
up
an
HPC
cluster,
that's
very,
very
comparable
to
the
Krays
I
used
to
work
on
back
in
the
1980s.
B
That's
what
I
find
most
interesting
about
the
cloud
these
days,
the
fact
that
it's
making
incredibly
vast
compute
resources
available
to
us
that
would
have
been
next
to
impossible
to
get
ahold
of
several
years
ago
and
the
fact
the
building
software's
of
service
offerings.
On
top
of
that
that
make
it
easy
to
do
the
things
that
might
not
be
so
easy,
otherwise
get
access
to
HPC
cluster
spark
clusters,
and
things
like
that.
B
I
wanted
to
throw
just
a
few
numbers
in
there,
not
because
the
numbers
themselves
are
important,
but
just
to
demonstrate
the
scale
of
a
sure
as
a
cloud
platform
think
about
what
it
would
take.
If
you
were
going
to
build
out
the
infrastructure
for
something
that
would
process
777
trillion
storage
transactions
per
day.
That
number
has
gone
up
by
the
way
we
have
an
updated
is
slide
in
a
few
months.
That's
just
a
massive
number
of
transactions.
B
What's
interesting
about
the
cloud
these
days
is
that
it
offers
an
economy
of
scale
that
was
unthinkable
just
a
few
years
ago.
The
reason
as
you
can
do,
what
you
saw
in
the
previous
slide
is
because
there
are
currently
30
for
data
centers
online,
with
four
more
that
have
been
announced
and
in
some
cases
are
actually
under
construction
right
now,
if
you
have
tracked
the
development
of
Azure,
you
know
that,
just
a
year
ago
that
was
24
data
centers
Microsoft
is
literally
investing
billions
of
dollars
in
building
data
centers
around
the
world.
B
Most
of
these
are
public
data
centers,
but
some
are
not
in
certain
countries.
Local
laws
and
regulations
require
data
to
not
cross
the
boundaries
of
the
country.
So
essentially
we
stand
up
private
data
centers.
In
those
countries
there
are
gov
data
centers
to
which
are
only
available
took
of
customers,
but
with
this
kind
of
infrastructure
and
the
kind
of
throughput
that
you
saw
in
the
previous
slide,
the
big
message
is
that
we
can
deliver
an
economy
of
scale.
That
would
be
very,
very
hard
to
duplicate
yourself
in
our
role.
B
We
often
talk
to
CEOs
of
companies
who
are
trying
to
decide
whether
to
move
to
the
cloud
or
not,
they
often
say
well,
you
know
we
have
these
data
centers.
We
have
a
large
investment
in
them
and
stuff,
but
when
we
show
them
the
map,
we
talk
about.
The
investments
have
been
made
and
talk
about
the
economy
of
scale.
How,
with
the
cloud
you're
not
having
to
pay
to
keep
that
infrastructure
maintained
and
upgraded,
and
things
like
that
that
usually
paints
a
picture
that
makes
them
think
that
this
is
at
least
worth
considering.
B
Now,
Azure
itself
has
most
of
the
services
that
you
would
expect
from
a
cloud
platform.
You
can
spin
up
virtual
machines.
There
are
lots
of
different
types
of
VMs.
You
can
spin
up
some
of
them
cost
as
little
as
seven
cents
an
hour
to
run
some
of
them
cost
many
times
more.
You
choose,
for
example,
the
number
of
cores
the
types
of
course
the
amount
of
RAM,
the
number
of
disk,
whether
they're
SSD,
drives
or
what-have-you,
and
basically
can
create
these
virtual
machines
to
do
work
on
your
behalf.
B
If
you
want
to
network
those
virtual
machines
to
create
HPC
clusters
from
them,
that's
easy
to
do
to
often
just
a
button
click
because
of
something
called
as
your
deployment
templates,
which
is
one
of
the
mechanisms
that
Azure
gives
us
to
essentially
script.
The
creation
of
Azure
resources.
What's
really
interesting
to
the
research
community,
though,
is
what
you
see
in
the
center
of
this
slide.
It's
not
so
much
to
compute
resources
and
the
storage
resources
themselves.
It
is
the
data
services
that,
as
your
offers,
would
utilize
those
lower
level
infrastructural
services.
B
For
example,
HD
insight
is
Microsoft's
name
for
an
azure
service.
That
makes
it
very
very
easy
to
spin
up
clusters
of
virtual
machines
pre-installed
with
members
of
the
Apache,
a
data
ecosystem
like
Apache,
Hadoop
and
Apache
spark
and
other
things,
I'll
say
more
about
those
in
just
a
moment,
but
also
under
analytics
and
IOT
they're.
In
the
center
you
see
as
your
data
factory
as
revent
ABS,
Azure
machine
learning
stream,
analytics
I'll
be
talking
about
some
of
these
and
more,
but
as
a
researcher,
this
is
probably
the
part
of
azure.
B
B
Yes,
it's
running
on
Azure
virtual
machines,
but
with
the
future
of
the
cloud
being
software-as-a-service,
where
it's
not
just
about
managing
two
VMs
yourself,
but
clicking
a
button
and
creating
resources
that
already
have
software
pre-installed,
etc.
That's
where
things
are
going
now,
one
of
the
things
I
want
to
mention
just
because
it's
not
something
that
we
do.
A
very
good
job
of
communicating,
sometimes,
is
that
just
because
Microsoft
builds
and
sells
this
product
called
Windows
doesn't
mean
that
there's
a
strong
tie
between
a
sure
in
Windows.
B
In
fact,
when
you
go
in,
you
create
a
virtual
machine
and
Azure,
you
have
a
choice
in
most
cases,
whether
to
make
it
a
Windows
machine
or
a
Linux
machine,
and,
to
be
honest
with
you
much
of
the
effort
and
the
work
that's
going
into
Azure
today
is
not
around
making
it
a
better
platform
for
Windows,
but
it's
making
sure
that
it
runs
Linux
and
members
of
the
Linux
ecosystem.
Very
very
well.
So
we
like
people
to
know
that,
just
because
it's
Microsoft
we're
not
necessarily
tied
to
Windows.
B
We
are
just
as
happy,
if
not
happier,
to
host
your
virtual
machines
on
Linux,
as
we
are
with
Windows
now,
just
real
quick
here,
so
I
can
get
it
over
to
Vani.
So
she
can
talk
about
the
data
science
program.
I
just
wanted
to
make
you
aware
of
a
few
of
the
specific
features
and
services
of
a
sure
that
might
be
interesting
to
you,
as
a
researcher
for
one
measure
has
supported
virtual
machines
for
a
long
time.
B
If
you,
if
you're
interested
in
knowing
what
types
of
virtual
machines
you
can
create
and
cluster
together,
what
the
specs
are
just
to
a
search
on
Azure
VM
sizes,
the
most
recent
VM
size
or
VM
series
it's
been
introduced
is
our
in
series
of
virtual
machines,
of
which
there
are
two
types:
they're
NC
VMs
and
NV
VMs.
Both
of
those
are
outfitted
with,
in
most
cases,
multiple
nvidia
gpus.
B
The
idea
here
is
that
if
you
need
an
HPC
cluster
with
GPU
support
with
with
thousands
of
GPUs,
if
you
want-
because
you
can
connect
these
these
machines
together
with
our
DMA
and
InfiniBand,
then
you
can
create
a
cluster
as
large
or
small
as
you
want
with
those
GPUs.
In
fact,
Microsoft
researchers,
UK
branch
recently
worked
with
some
folks
at
the
alan
turing
institute
in
the
UK
who's,
a
single
MC
24
vm,
to
do
some
computations.
B
That
would
have
taken
a
lot
longer
on
that
researchers
laptop
in
c
24
by
the
way,
as
24
cores
has
4k
80
GPUs
to
under
24
gigabytes
of
RAM
and
other
things.
So
it's
a
pretty
beefy
machine
and
in
this
case
the
researcher
was
using
that
machine
standalone.
But
again
we
can
network
these
things
together
and
create
GPU
clusters
from
them.
Another
thing
that
might
be
interesting
to
you
as
a
researcher
HD
insight,
which
I
mentioned
a
moment
ago.
B
Many
of
the
universities
that
Vani
and
I
visit
use
Hadoop
and/or
SPARC,
and
it's
not
a
common
for
the
IT
departments
at
those
universities
to
build
and
maintain
large
and
very
expensive
compute
clusters,
then
typically
the
the
faculty,
the
postdocs,
the
grad
students
will
have
to
write
proposals
to
get
access
to
those
to
those
clusters
and,
if
you've,
ever
by
the
way,
built
a
Hadoop
machine
or
SPARC
machine
yourself,
much
less
a
cluster.
You
know
how
difficult
that
is.
Hd
insight.
B
You
literally
go
into
the
azure
portal
answer
a
few
questions
like
what
VM
sizes
do
you
want?
How
many
nodes
do
you
want
there
to
be
in
the
cluster?
What's
off
to?
Where
do
you
want
pre-installed,
you
click,
OK
and
typically
10
to
15
minutes
later
you
come
back
and
that
cluster
has
been
created
for
you,
Azure
machine
learning,
something
that
we're
particularly
proud
of,
because
machine
learning,
as
you
know,
is
figuring
very
prominently
into
data
science.
These
days,
if
you've
never
seen,
Azure
machine
learning
check
it
out
sometime.
B
It's
a
slightly
different
way,
thinking
about
machine
learning
where,
rather
than
code,
everything
from
scratch
yourself.
We
give
you
this
browser-based
tool
called
ml
studio
where
you
build
sophisticated
machine
learning
models
using
a
drag-and-drop
paradigm,
basically
dragging
modules
representing
algorithms,
representing
actions
like
various
operations
used
to
clean
the
data,
even
training,
the
model
itself,
dragging
modules
out
to
a
workspace
and
then
connecting
them
together
to
define
a
workflow.
What's
really
neat
about
Azure
machine
learning
and
one
thing
that
sets
it
apart
is
once
a
models
built
trained
and
evaluated
you're
happy
with
it.
B
A
couple
of
button
clicks
operationalize
that
model
by
deploying
it
to
the
cloud
as
a
web
service
with
HTTPS
reachable
endpoints.
That
means
I
can
then
take
that
model
and
I
can
build
software
around
it,
whether
it's
software
from
my
iPhone
software
for
my
browser
from
my
workstation
or
whatever
makes
it
very,
very
easy
to
operationalize
the
model.
Something
else
that's
getting
a
lot
of
tension
today
in
data
science
is
streaming
data
dealing
with
the
massive
amounts
of
data
coming
from
IOT
devices
and
other
places.
Just
very
briefly.
B
Something
you
might
not
have
heard
of
is
something
that
is
technically
not
part
of
azure,
but
in
fact
it
has
come
out
of
Microsoft
Research
is
hosted
on
Azure,
etc.
It's
called
Microsoft
cognitive
services.
It's
a
set
of
right
now,
24
api's,
it's
grown
by
three
recently
that
help
you
build
intelligent
software.
There
is,
for
example,
a
computer
vision
API
to
which
you
can
pass
images
by
URL
or
invite
streams.
It
will
analyze
the
images,
identify
objects
and
them
identify
colors
captioned,
the
images
score
them
for
adult
content
and
race
eNOS,
and
things
like
that.
B
Is
the
text
analytics
API
kind
of
interesting
and
you
feed
it
a
text
document
or
a
Twitter
stream,
or
something
like
that,
and
it
will
analyze
that
data
in
several
ways
it
will
tell
you
what
language
it
believes
that
was
written
in
it
will
detect
and
identify
topics
and
key
phrases,
and
perhaps
the
most
interesting
part
of
it
is.
It
will
do
a
sentiment
analysis.
B
It
will,
for
example,
return
to
you
upon
request
a
value
from
zero
to
one
indicating
the
sentiments
being
expressed,
one
being
hey
everyone's,
very
happy
who's
in
this
Twitter
stream,
zero
being
they're,
not
very
happy,
would
have
been,
would
have
been
very
interesting
to
take
the
Twitter
stream
with
hashtag
United
last
week
and
feed
it
into
this
API
just
to
see
how
it
turned
out.
And
finally,
one
of
the
things
we're
very
proud
of
about
a
sure
is
that
we
implement
in
it
now
in
a
service
called
the
Azure
container
service.
B
A
native
docker
stack
now
lots
of
cloud
services
support
docker.
Most
of
them
do
that
with
a
combination
of
open
source
and
proprietary
components,
we
have
literally
taken
dr.,
which
is
open-source
implemented.
A
native
stack,
which
means,
if
you
want
to
take
docker
eyes
or
containerized
applications
and
run
the
manager,
you
can
do
it
natively,
you
don't
have
to
do
anything
other
than
connect.
As
your
container
service,
your
doctor
images
don't
have
to
change.
You
can
still
pull
images
from
the
docker
from
docker
hub
or
from
other
docker
registries.
It
just
works.
B
We
actually
support
now
out-of-the-box
three
different
orchestration
services
for
it,
so
whether
you
prefer
kubernetes
or
d
cos
or
docker
swarm,
it's
literally
just
a
drop
down
list.
In
the
questions
we
ask
you
before
we
deploy
that
service.
For
you
one
last
thing
one
one
reason
that
may
be
interesting
is
that
one
thing
we
hear
from
our
customers
at
universities
and
research
institutions.
Time
and
time
again,
is
we
like
the
cloud,
but
we
don't
want
to
be
tied
into
any
one
vendor
or
cloud
platform,
one
of
the
great
things
about
containerized
workloads.
B
B
So,
if
docker
is
not
something,
that's
on
your
radar
or
the
container
theory
in
general,
it
might
be
something
you
want
to
look
at,
because
I
think
you're
going
to
see
that
it
not
only
plays
a
large
role
in
Azure
in
the
future,
but
a
large
role
in
the
research
community
as
well
and
with
that
I
am
going
to
turn
it
over
to
Vani
and
Vani.
I
apologize
I
meant
to
take
about
five
minutes
less,
but
I
realize
you
are
dialed
in
by
phone.
B
C
Thanks
Jeff
I
am
I
can
view
the
seven
also
identities
in
this
life
I'm
going
to
go
over
the
programs
that
we
have,
that
the
also
academic,
researchers
and
violas
are
considered
to
be
advanced.
The
stability
of
measure
of
the
good
disabled,
big,
big
community
and
I
think
it's
important
to
impress
upon
the
community
of
what
so
we
can
offer
in
terms
of
research
on
big
data
loss
loads.
C
So
the
data
grant
program
out
of
iping
give
academic
researchers
and
office
serving
four
steps
in
order
for
a
DBMS
once
we
design,
we
have
partnerships
with
the
National
Science
Foundation
and
a
fog
of
that
of
the
three
million
dollars
worth
of
projects
by
the
second
unrelated
across
the
for
big
data
hubs
that
we
are
talked
about.
They're
also
part
of
an
NSF
big
data,
solicitation
and
you're
excited
as
NSF
is
moving
away
from
traditional
hardware
based
grants
to
cloud
infrastructure.
C
C
They
are
rewarding
when
example
year.
We
provide
training
to
the
academic
community
and
how
to
use
them.
So
the
next
slide
has
a
representation
of
how
historically,
students
were
doing
dignitary
sources
of
light
from
about
three
years
ago,
from
the
University
of
Washington.
They
send
system
as
far
as
resources,
and
they
request
a
cloud
computing
grants
garage
and
it's
very
conversant.
The
Cubs
fix
it
and
there
are
a
lot
of
universities.
Festers
are
on
this
model.
C
The
students
are,
you
know,
using
lots
of
data
sharing
is
not
experiments
are
taking
a
very
long
time
fast
forward
about
three
years
into
the
program
and
over
a
thousand
translator.
We've
had
some
real
successful
out
of
the
article
research
program,
as
is
evident
in
the
case
study
setting
a
published
on
our
website.
C
That's
one
of
into
a
start-up
and
they're
doing
very,
very,
very
proud
of
this
project.
There's
been
a
few
more
such
successes,
both
in
academic
and
startup,
and
it's
across
multiple
domains
and
what
this
with
healthcare,
the
environmental
science.
There
is
one
really
interesting
one
where
the
professor
at
University
of
Rochester
is
in
Asia
to
help
improve
their
public
speaking
skills
by
just
using
a
camera
on
their
laptop
critical
video
than
fin
gets
analyzed,
and
you
know
using
Azure
machine
learning.
They
came
to
sentiment,
analysis
and
and
so
on.
C
So
why
are
we
focusing
so
much
on
all
these
cognitive
services?
And
all
this
intelligence
based
on
data,
we
believe
that
we
are
in
the
middle
of
this
industrial
revolution
and
which
is
known
as
the
fourth
Industrial
Revolution
there's
this
for
this
duration,
business
or
the
proliferation
of
small,
inexpensive,
IOT
devices
and
computers,
and
these
devices
are
as
if
this
critical
piece
is
in
use
and
all
these
machines
are
everywhere
and
constantly
creating
and
collecting
of
ton
of
data,
and
we
hope
they'll
democratize,
with
Industrial
Revolution
by
providing
the
building
blocks
to
empower
the
people.
C
So
if
it
is
to
do
that
means
that
there
is
a
big
digital
transformation
that
drives.
You
know
business
value
and
research,
in
fact,
through
two
degrees
of
these
intelligent
services
that
is
built
upon
the
cloud.
So
we
can
probably
skip
over
the
next
two
slides
for
lack
of
time
and
go
straight
to
slide
21
in
the
deck.
So,
there's
a
huge
transformation
by
the
digital
caption
which
we
are
taking
you
from
date
our
website,
and
we
believe
that
you
know
analytics,
can
go
all
the
way
from
just
descriptive,
which
is
20
years
ago.
C
You
had
a
time
judge
Chen
because
of
both
services.
This
is
some
visual
representation
of
the
data.
The
diagnostic
both
of
these
are
considered
to
be
hindsight,
type
of
analytics
to
predictive
analytics,
which
is
more
insight
of
based
analytics
to
prescriptive
analytics,
which
of
actually
going
to
you,
know,
predict
and
tell
you
what
is
going
to
happen.
C
That's
considered
the
bony
clearing
of
analytics
is
for
site,
based
of
analytics,
where
you
going
all
the
way
from
hindsight
to
foresight
through
the
digital
transformations
and
again,
there's
this
very
fun
graphic
that
I
don't
know
you
might
have
seen
people
that
Microsoft
present,
which
is
known
as
the
rice
equation.
We
expect
that
by
2020
all
these
devices
and
users
are
going
to
generate
40
data
bytes
of
data.
So
if
a
single
grain
of
rice
was
considered,
one
of
one
byte
of
data
or
data
byte
of
data
would
sending
specific
emotion.
C
That's
how
much
data
these
devices
are
going
to
generally
and
Microsoft
advanced
analytics
portfolio
has
the
capabilities
to
build
solutions
based
on
all
the
way
from
processing
data.
Open-Source
tool,
such
as
police
fog
and
our
based
analytics
to
the
cognitive
service
is
an
analytics
API
that
will
empower.
Everyone
comes
into
engineers
and
data
scientists
to
the
citizen,
scientist
to
the
developers
and,
finally,
the
business
decision-makers
and
Technology
decision
maker.
C
So
in
terms
of
our
collaboration
with
the
South
big
data
hub,
you
can
watch
out
for
a
broad
RSC
that
we
are
going
to
be
rolling
out
in
I
believe
the
main
newsletter
which
will
I
and
I
think
sometime
next
week
there,
in
addition
to
some
of
the
total
grants
that
they
are
providing
some
of
the
NSF,
you
know
funded
researchers.
We
are
also
opening
it
up
to
the
broader
subject
in
our
community
and
on
that
layer.
C
D
So
this
is
standing
halt.
I
have
a
kind
of
a
strange
question,
but
hopefully,
hopefully
it'll
loosen
other
people
up.
So
how?
How
can
the
South
PD
hub
be
most
helpful
in
helping
Microsoft
derive
value
from
the
donations
of
the
resources
that
you
all
have
put
forward?
What
what
you
want
to
see
come
out
of
your
contributions
to
you
know
the
Academy
I.
C
Think
engagement
is
key.
You
know
coming
to
events
like
this
and
being
aware
of
what
is
available
and
being
aware
of
how
it
is
helping
other
academic
community
that
is
actually
rate.
Their
research
forward
is
to
me
some
it's
something
that
might
work
what
Microsoft
has
successfully
done.
What
are
you
have
successfully
done
over
the
past
few
years
of
extracting
academic
communities
with
the
cloud
resources
and
a
significant
amount
of
youngest?
C
Yeah
sure
so
we
are
cool,
locating
our
next
training
event
and
the
files
big
data
hubs,
along
with
the
All
Hands
meeting,
which
is
on
June,
might
be
during
the
training
the
day
before
the
All
Hands
and
it's
an
all-day
training
and
it's
mostly
hands-on.
It
is
the
training
that
we
do
fairly
regularly
and
it's
been
very
useful
to
everyone
who
comes
to
these
sessions
because
or
they
get
a
chance
to
try
some
of
these
services.
We
give
$500
worth
of
azure
credit
that
is
valid
for
up
to
a
month
after
the
training.
C
It
can
go
any
long
way
you
can
try
to
even
you
know,
start
to
build
up
your
research
on
this
training
path.
So
the
topics
we
cover
are
around
Asia
storage,
some
of
the
big
data
you
know
using
spark
in
her
dalhwan
Azure,
using
containers
of
acidification
using
Azure
batch,
while
the
training
is
on
github
and
we
have
more.
Obviously,
we
have
more
content
and
we
can
cover
in
one
day,
but
it's
a
fairly
intensive
and
fun
way
of
learning
how
to
use
the
cloud.
C
B
And
one
thing
excuse
me:
we
can
invite
people
to
do
since
the
training,
as
Bonnie
said,
is
very
hands-on.
Just
ask
any
students
or
whomever
that
comes
to
bring
a
laptop
with
him.
They
don't
have
to
have
anything
installed
on
that
laptop
and
it
doesn't
have
to
be
a
Windows
laptop
either
it
can
be
a
MacBook
or
a
Linux
machine
if
they
want.
When
we
build
the
training
content,
we
work
very
hard
to
make
sure
that
it
will
run
on
any
platform,
just
as
a
sure
will.
C
Yeah
we're
very
sensitive
to
the
fact
that
we
have
to
be
hundred
percent
open-source
and
platform
agnostic
in
terms
of
teaching
people
how
to
use
the
cloud,
and
we
emphasize
those
aspects
and
pretty
much
you
know-
do
not
use
any
content
which
sometimes
some
of
the
newer
services
gets
get
rolled
out
of
two
windows
first
and
then
to
the
Knox
and
Mac.
And
we
make
sure
that
we're
using.
So
this
is
that
I'm
mature
enough
and
our
way
to
move
on
Mac
and
run
off.
D
Would
like
to
say
thank
you
very
much,
I'm
really
fascinated
by
both
of
the
presentations
I
thought
they
were
excellent
and
I'm
gonna
go
sign
up
for
Globus
and
I'm,
going
to
be
at
the
Microsoft
Azure
training
in
DC
and
I'm
eager
ly.
Looking
forward
to
it,
I
did
not
realize
you
all
had
as
many
api's
created
as
you
were
just
now
showing,
and
some
of
it
really
looks
interesting,
including
perhaps
for
classes.
So
thank
you
very
much.