►
From YouTube: ML on OpenShift SIG AIOps with ProphetStor
Description
Jeremy Wei and Brian Huang of Jeremy Wei (Prophetstor) – Demonstration of predicting computing resources usage for both nodes and containers, as well as predicting hardware issues.
A
A
Al
Gore
is
the
heart
to
that
to
improve
the
idea,
I
think
idea,
operation
by
providing
the
predictions
that
order
item
from
structure
underlying
components
and
viewed
at
the
correlation
with
players,
though
the
we
currently
deeply
walking
with
Seth
as
well,
and
they
contributed
that
takes
to
failure,
petition
technology
to
say
that
the
man
distribution
will
be
that
guarantees
on
the
tested
by
the
community
should
be
ready
to
bite
particular
Curie,
and
we
can.
A
B
B
It
include
air
engine
base,
patented
AI,
taking
knowledge
the
best
provided
for
ice
machine
learning
features
the
first
one
is
a
performance
application
and
a
resource
planning.
So
we
know
that
the
characteristic
of
content
operation
is
very
dynamic,
so
it
is
very
important.
You
understand
what
is
going
on
in
the
future
and
that
just
depends
on
the
column
entries
decide
when
to
at
the
pass
or
deploy
part
of
different
nouns
to
go
performance,
petition
and
risa
premium.
B
It's
very
important,
more
efficient
control
operation
and
we
also
focus
what
do
we
call
cross
layer
performs
issue
troubleshooting,
because
we
know
that
even
we,
the
user,
understand
that
how
how
containers
work,
there's
still
a
lot
of
situations
and
problems
are
the
problems
occurred
in
the
underlying
layers?
That
means
in
structure
our
component,
such
as
disk
hard
drive,
flash
drive
or
memory
or
even
CPA
cash,
and
we
use
the
machine
learning
to
predict
when
the
disease
is
going
to
fail
and
also
correlate
those.
B
How
will
the
issue
to
the
conform
of
the
issue
in
the
upper
layer
means
the
container
layer
or
virtualization
layer
or
even
the
most
hardened
layer?
I
mean
the
application
layer,
and
the
third
nationally
feature
would
provide
is
what
we
call
in
competition.
You
note
that,
since
we
can
predict
since
figured
I
predict,
the
hardware
issues
are
going
to
happen
in
the
future
in
the
future,
so
use
them
would
like
to
know
how
those
pass
ball.
B
B
Matrix
events
logs
and
also
a
smart
I
mean
this
is
sparta
because
record
I
predicted
this
video.
So
we
may
sound
smarter
how
to
pick
up
when
the
disk
is
going
to
fail.
A
personal
favorite
I
have
all
these
information
and
we
contact
a
figure
type
contact.
Our
performance
of
Knoblauch
be
patient,
based
on
active
in
your
network
and
Howard
issue
petition.
It
also
pays
on
did
neural
network
particular
time
serious,
the
neural
network
of
Alisa,
and
then
we
use
also
use
the
machine.
B
B
The
the
long
content
and
even
content-
and
we
correlate
those
long
and
on
lolly
anomalies
of
within
log
and
events
through
Nancy
pace,
to
help
Heidi
and
me
traders,
who
first
include
a
partner
who
calls
from
the
large
base
and
the
solution
that
the
possible
solutions
that
may
apply
to
the
issue
they
are
facing
and
the
actions
that
user
can
take
input,
cost
layer,
usually
troubleshooting,
but
we
in
figured
I
would
put
first
layer
aspect
as
our
core
value
of
L.
Yes
cause.
B
It
is
simpler
when
IBM
traitor
want
to
troubleshoot
an
issue
just
looking
at
the
application
performance
issue
is
lagging
now.
They
also
want
to
know
when
an
application
has
a
problem.
The
issue
of
a
table
cost
in
the
only
take
a
look
at
have
the
region
application.
They
also
need
to
know
the
underlying
objects,
such
as
the
containers
and
even
the
hardware
components.
B
B
B
It
is
based
on
petition
information,
not
just
the
color
matrix.
It
is
very
important
because
we
know
that
internal
operation
very
dynamic,
though,
if
it
hurts
you
know,
part,
is
going
down
with
may
be
in
the
short
run
to
the
shorter
feature
and
then
the
turn
feature,
and
so
with
the
schedule
and
autoscaler
and
better
this
type
and
the
decision
of
where
to
allocate,
pass
and
or
even
to
add
ball
the
vacation
step
to
pass,
or
even
add
more
now,
to
connect
this
question.
B
Okay,
sorry,
since
we
are
going
to
run
up
time
so
I'm
to
skip
some
of
the
slides-
and
this
is
matrix
of
data
sources
and
what
we
can
do
from
these
data
sources
for
possible
for
the
performance
matrix
and
this
part,
these
are
the
current
they
will
network.
We
generate
the
prophetís
petition
and
despair
of
petition
and
for
a
structured
data
include
a
lot
either
a
large
base
configuration
to
use
natural
language
processing.
B
The
I
will
speak
this
one
this
one-
and
this
is
the
disability-
start
studying
what
we
are
going
while
we're
doing
right
now
for
authorship
or
a
step
just
journey.
As
Chris
mentioned,
we
are
working
with
the
subcommittee
to
provide
the
Despero
petition
for
that,
because
we
know
that
when
the
OSD,
this
veil
is
large
for
participation
and
we
actually,
we
we
we
started
our.
We
started
our
container
study
even
reported
version,
one
of
the
negative
we
we
already
should
I
mean
apparently
400
years
ago,
and
also
we
enforce
been
burning.
B
B
A
B
With
IBM
I
came
traitor
along
to
troubleshoot
an
issue.
They
knew
to
look
back.
What's
going
on
in
there
I
think
Brahman.
Don't
they
realize
the
biomedical
technician
that
user
to
give
birth
back
with?
What's
going
on
in
the
IT
environment
and
the
fire
hose
layer
map
that
users
see
how
how
how
and
how
even
that
may
impact
the
order,
control
vendor
in
their
environment,
and
since
we
also
provide
the
protection
No.
B
This
competition-
and
this
is
a
screenshot
of
the
display,
a
petition
under
frontier
out
this
particular
reason
and
provide
very
high
accuracy
rate
for
a
disc
sander,
somehow
lower
literacy
rate
or
SAS.
This
notice
artist
doesn't
have
happened,
then
Smart
View
as
Socrates,
and
also
they'll
use
the
pass
ball,
sent
a
friend
of
pointing
failed
disk,
and
we
also
provide
the
competitiveness
time
based
on
the
performance
petition.
B
Okay,
that
in
summary,
we
used
our
figured
I
learned
that
performance
patterns
and
detect
anomaly
from
the
historical
metrics,
logs
and
dance
and
based
on
the
performance.
We
use
some
kubernetes
mechanism,
such
as
scheduler
and
horizontal
autoscaler,
that
was,
scanner,
provide
the
pattern,
operation,
automation
and
also
we
are
working.
B
For
and
now
filter
would
include,
we
will
put
more
focus
on
natural
language
processing
because
of
it
has
petered
that,
for
example,
we
tried
to
transform
the
anomaly
with
detective
performers,
maybe
in
honey,
some
serious
correlation
into
a
natural
language
and
then
use
a
natural
language
to
research,
a
solution
from
a
large
base
from
discussion
forum,
but
are
there
and
the
things
we?
We
have
a
lot
of
data
need
to
train
and
we
are
also
working
to
test
how
to.