►
Description
Speakers: Rich Hammel, Director of Advanced Manufacturing at Brocade and Vivek Ganesan, Principal Architect at Impetus Technologies
Almost 10 years ago in a hotel room in Asia his first parser was born. That parser and its offspring have supported the development of world-class networking products at Brocade. This discussion will include how big data will change manufacturing, the essential ingredients for success in greenfield big data projects, and what it's like to be obsessed with quality.
A
It's
unbelievable
or
AV
from
impetus
technologies
we're
one
of
the
sponsors
here
in
this
conference,
where
a
big
data
solutions,
integrator,
that's
doing
a
lot
of
work
around
Cassandra
and
Hadoop,
and
one
of
our
early
white
papers
in
2009
on
Hadoop
configuration
actually
attracted
the
attention
of
our
pioneer
pioneer
rich
over
here
and
it's
my
pleasure.
Introduce
rich,
rich
Hamel
is
a
director
of
advanced
manufacturing.
B
And
on,
and
so
first
of
all,
I'd
like
to
thank
datastax,
you
know
this.
This
conference
has
been
tremendous
I've
learned
a
lot
of
things
here,
actionable
things
that
can
bring
back
to
my
company,
so
I
do
want
to
thank
them.
Tremendous
venue,
you
know
walking
from
these
different
talks
from
one
building,
the
other.
With
this
view,
I
actually
live
in
San
Francisco
and
seeing
these
all
these
sites
is
just
breathtaking
over
and
over
again,
so
do
want
to
thank
them.
B
I
also
appreciate
the
opportunity
to
talk
to
you
about
something
that
I'm
very
passionate
about.
So
that's
fantastic
too
I'd
like
to
start
off
by
saying
that
I
am
not
in
IT
I'm
an
engineer.
You
know,
I
think
the
last
programming
language
I
took
what's
Fortran,
so
I
think
that
dates
me
a
little
bit
but
I
do
want
to.
Thank
you
guys.
You
know
a
lot
of
people
in
the
audience
here
have
developed
technologies
and
tools
that
can
be
used
across
all
kinds
of
industries,
and
it
you
know
these
tools.
B
B
And
let's
see
here?
Okay,
so
this
is
not
working.
So,
let's
see
if
we
can
do
this,
okay,
so
here's
a
list
of
brocade
products
that
we
work
on
when
I
started
at
brocade
in
2003
it
was
just
the
blue,
the
blue
circles
there.
At
the
time
we
were
just
a
fibre
channel
company
working
on
storage
networks
and
basically
we
had
one
product
line
with
three
products
in
it.
The
one
in
the
center
is
what
we
call
a
director.
B
It's
a
chassis
based
product
where
a
number
of
blades
go
insert
it
into
that
chassis,
and
then
we
also
had
a
few
pizza
boxes,
which
are
one
you
devices
looks
kind
of
like
a
VCR
or
a
piece
of
audio
equipment
goes
into
the
you
know,
top
of
rack
applications.
So
that's
where
I
started
at
brocade
since
then,
we've
expanded
our
portfolio
quite
a
bit.
We
now
develop
embedded
switches
that
go
into
servers
from
some
of
our
OEMs
IBM
Dell
HP.
B
All
those
companies
have
embedded
switches
that
go
into
their
servers,
so
that
was
our
first
foray
into
kind
of
expanding
our
product
portfolio,
then
kind
of
the
big
thing.
The
big
thing
that
changed
this
was
in
around
two
thousand
eight
timeframe.
We
bought
a
company
called
foundry
foundry
networks
that
was
into
ethernet
the
ethernet
space
land
products
and
from
them
all
the
rest.
These
products
came
on
board
and
I'll
talk
a
little
bit
about
some
of
the
problems
we
saw
when
we
did
that.
B
But
you
know
you
could
see
up
here
on
the
top
left,
some
of
the
low
the
lower
complexity
lower-cost
products
for
the
campus
enterprise.
There's
wireless
there's
pizza
box
products
in
the
center.
Here
we
have
some
of
the
more
complex
products
for
service
providers
and
data
centers.
So
these
could
be
anything
from
a
closet-sized
chassis
and
again,
you
know
going
into
some
of
the
internet
exchanges
and
some
of
the
bigger
data
centers
in
our
customer
base.
B
So
one
of
the
other
things
I
want
to
mention
that
did
not
make
this
slide,
but
that
is
actually
over
to
the
right
is
some
of
our
new
products,
our
Ethernet
fabrics
and
those
things
are
very,
very
interesting
brand
new
technology,
their
self-healing
fabrics
that
scaled
a
big
big
data
applications,
so
very,
very
new
product,
I'm,
hoping
in
the
next
year
or
two
that
when
we
come
to
conferences
like
this
more
and
more
people,
we
were
talking
about
using
these
types
of
products
in
their
installations.
So
let's.
C
B
That's
it
on
products,
so
a
little
bit
about
manufacturing
and
I
know
it's
not
the
you
know
the
most
sexy
topic
to
be
talking
about
here,
but
you
know,
manufacturing
is
big.
There's
a
lot
of
things
going
on
right
now
for
those
of
you
that
followed
the
president
and
presidential
election,
a
lot
of
talk
about
manufacturing
jobs
and
what
it
takes
to
bring
those
jobs
back
to
the
United
States.
B
Let's
see
so
with
that
I
think
there's
a
huge
opportunity,
I
think.
Actually
some
of
the
tools
that
you
guys
are
developing
can
be
used
to
gain
a
significant
advantage
in
in
manufacturing
and
I'll
talk
a
little
bit
more
about
that
in
a
second
you'll,
see
some
examples
of
that.
I
don't
know
if
you
knew
this,
but
you
know
manufacturing
ninety
percent
of
the
patents
that
are
generated
in
the
US
or
related
to
manufacturing.
B
Seventy
percent
of
the
private
rd
in
the
US
is
related
to
manufacturing
and
fifty
percent
of
the
exports
from
the
United
States
are
related
to
manufacturing.
So
manufacturing
is
still
big,
just
electronic
manufacturing
alone.
So
if
you
look
at
cell
phones,
you
look
at
computers.
You
look
at
some
of
the
things
we're
doing
on
the
network.
The
servers,
if
you
look
at
all
of
that
in
2013,
that
market
is
expected
to
be
about
1.3
trillion
dollars.
So
you
know
anything
we
can
do
to
help
out
this
industry.
B
Let's
see
a
few
other
things
time
to
market
is
shrinking.
Things
are
getting
more
competitive,
especially
in
the
networking
space.
So
we
got
in
pretty
early,
but
after
that
Cisco
started
focusing
you
know
they
were
buying
video.
They
were
doing
all
this
other
stuff,
but
you
know,
after
that
they
realized
they
really
need
to
focus
on
their
core
networking
space,
so
they're
focused
they're,
they're,
big
company,
but
they're
focused
in
addition,
companies
like
HP,
IBM
and
Dell
separately,
bought
networking
companies
so
a
lot
of
people
and
then
there's,
of
course
Huawei
in
China.
B
So
there
are
a
lot
of
companies
that
are
focusing
in
this
marketplace.
So
time
to
market
is
a
key
differentiator
and
we
had
brocade
feel
like
we
have
an
advantage
there.
You
know
Cisco's
gigantic,
we're
much
smaller
company.
We
can
react
a
lot
quicker
than
they
can
so
I
think
we
have
been
advantaged
there
and
I
actually
showed
a
graph
here.
It's
really
nice
to
see
this.
It
was
interesting
looking
at
our
time
to
market
over
time
at
brocade.
So
this
is
from
the
time
of
concept
and
I
didn't
give
any
y-axis
here.
B
For
you
know
the
people
told
me:
no,
you
can't
you
can't
divulge
that,
but
I
did
least
it
shows
scale
and
it
started
2001
with
our
first
fibre
channel
products,
and
you
can
see
a
really
nice
decline
in
that
time
to
market
during
that
time
frame.
And
then
you
can
see
that
little
lump
up.
That's
when
we
bought
foundry.
B
So
we
took
a
hit
there,
but
you
then
you
can
see
the
same
trends
going
down
so
we're
very,
very
focused
on
decreasing
that
time
to
market,
because
we
feel
we
feel
it
gives
us
considerable
competitive
advantage.
Another
thing
that's
going
on
teams
are
much
more
distributed.
We
have
contract
manufacturers,
we
have
development
teams
in
China
in
India
in
the
US
all
all
over
the
world,
and
we
need
to
be
able
to
collaborate
on
this
data
both
within
our
company
across
geographic
boundaries,
as
well
as
across
other
companies.
B
So
that's
another
one
and
then
finally,
you
know
just
in
general
data
is
more
important.
I,
don't
know
how
many
you
follow
the
news,
but
you
know
the
president
recently
designated
1
billion
dollars
to
to
invest
into
his
concept
of
smart
factories,
which
means
all
of
them
firm.
All
the
tools
in
a
factory
are
all
integrated
together
and
we
collect
that.
B
Looking
looking
at
common
ways
to
collect
that
data
store
that
data
and
analyze
that
data
once
again
for
competitive
advantage
for
the
US,
specifically
around
manufacturing,
so
a
lot
of
stuff's
going
on
in
manufacturing,
so
I
briefly
want
to
talk
a
little
bit
about
our
manufacturing
process.
We
make
printed
circuit
boards,
an
example
is
down
on
the
right
hand,
side
I
wanted
to
show
some
pictures
of
what
these
factories
look
like.
It's
quite
interesting.
You
know
you
look
down
the
line
and
you
can
see
machines
after
machine
after
machine.
B
So
all
the
again,
all
these
machines
generate
data
and
that
data
we
can
take
advantage
of
so
quickly.
Take
you
through
the
manufacturing
process,
I'm
not
going
to
stop
it
at
every
single
stage,
but
I
wanted
to
give
you
the
gist
of
it.
So
we
started
off.
We
start
off
with
a
prayer,
a
bear
printed
circuit
board.
B
It's
the
green
thing
that
you
see
inside
of
a
computer
inside
of
a
server
at
that
point,
there's
a
paste
printing
process,
so
solder
solder,
paste
starts
off
and,
and
it's
screen
printed
much
like
a
t-shirt
is
screen
printed.
You
screen
print
that
solder
paste
onto
the
board,
place
components
on
to
that
and
then
reflow.
It
basically
heat
it
up
melt
that
solder.
So
we
get
a
nice
clean,
solid
solder
joint.
At
that
point,
it's
inspected.
We
do
x-ray
inspection
to
look
for
voids.
B
We
also
do
in
circuit
test
to
basically
a
bed
of
nails
test
that
comes
down
and
tests
all
the
different
components
on
a
board.
Some
cases
we
have
up
to
you
know
50
or
60,000
components
on
one
board,
so
we
really
need
to
to
understand
whether
some
of
those
components
are
bad
or
not.
That
would
be
the
first
place.
We
do
that
and
again
I'm
skipping
over
some
of
these,
but
and
then
at
the
very
end
we
have
what
we
call
functional
tests.
B
We
start
off
with
board
functional
tests,
which
is
a
benchtop
room
temperature
test.
After
that
we
perform
stress
testing.
We
put
it
in
a
chamber,
big
big
ovens,
we
cycle
temperature,
we
run
it
constantly
at
45
c
we
voltage
margins
up
and
down
we
power
cycle,
the
heck
out
of
it.
We
run
traffic
on
all
ports
at
the
same
time.
So
we're
really
trying
to
hit
these
things
hard
in
the
factory
so
that
our
customers
don't
see
the
kind
of
failures
we
would
see
early
on.
B
So
again,
a
picture
of
the
finished
product
is
down
here.
This
is
a
brocade
encryption
switch
and
you
can
see
on
the
left-hand
side
that
that's
where
all
the
s
FPS
or
the
rj45
s,
in
this
case
s
FPS
or
plugged
in
in
the
center.
The
grey
boxes,
those
are
all
the
different
pieces
of
silicon
we
use
and
then
on
the
right-hand
side,
the
connectors
to
the
backplane.
So
this
is
goes
into
a
big
chassis.
B
Ok,
let's
see
so
a
little
bit
about
statistical
variation,
and
this
is
very,
very
important,
especially
with
complex
supply
chains.
So
I
just
shrunk
it
down.
What
I
just
showed
you
here
and
I
showed
a
printed
circuit
board
kind
of
in
line
with
that,
and
this
is
the
the
the
variation
that's
associated
with
that
printed
circuit
board
as
all
those
different
machines
that
are
using
that
we're
using
to
build
our
products
as
they
vary
over
time.
B
You
get
statistical
variation
at
each
process,
and
so
consequently,
you
get
boards
out
of
that
process
that
have
has
have
variations
in
it
and
when
we
do
that
when
we
install
those
boards
into
a
system,
the
same
thing
happens:
we
get
statistical
variation
around
that
around
that
that
system.
Now
it
gets
more
complicated
than
that.
If
you
look
at
our
remember
the
complex
supply
chain,
we're
talking,
so
we
have
components
that
go
into
those
boards.
They
have
their
own
processes,
they
have
their
own
statistical
variation,
build
that
out
even
further.
B
We
have
material
suppliers
to
those
components,
they
have
their
processes,
they
have
their
statistical
variations,
so
you
can
see
when
all
this
statistical
variation
adds
up.
We
get
this
graph
down
in
the
center.
So
we
have
this
nice
bell
curve
for
the
good
material,
and
then
we
get
all
these
little
defective
sub
sub
populations
of
product
is.
Is
you
can
imagine
if
these
things
are
not
lining
up
appropriately?
You
get
stuff
out
out
into
out
to
the
left
here.
B
That
is
really
not
suitable
to
build
a
long-term,
reliable
product,
and
so
so
basically,
what
do
we
need
to
protect
our
customer
from
all
this?
We
need
end-to-end
traceability.
We
need
to
be
able
to
trace
material
from
from
all
the
way
up
top
through
components
through
boards
through
systems
through
back
all
the
way
to
our
customers.
So
if
we
see
a
problem
with
the
material,
we
should
be
able
to
know
what
customers
have
those
products.
B
Okay,
so
all
this
statistical
variation
you
seen
already
represents
risk
risk
to
us
risk
to
our
our
suppliers
or
our
customers.
I
showed
this
slide
that
I
showed
on
the
previous
slide.
I
want
to
spend
a
little
bit
of
time
on
this
graph.
This
is
a
standard
reliability
graph
that
we
use
to
assess
failures
when
failures
are
happening.
The
x
axis
is
time,
and
the
y
axis
is
failure
rate,
so
normally
during
the
course
of
a
product
you're
sitting
here
at
a
useful
life
with
a
with
a
standard
failure
rate.
B
But
what
happens
is
early
on
you
get
what
we
call
infant
mortality
failures.
These
are
defects
that
are
introduced
in
manufacturing.
Those
fail,
they're
a
lot
faster
and
they
fail
a
lot
more.
So
I
had
this
up
to
4x
the
failure
rate
of
the
kind
of
a
steady
state
failure
rate.
So
these
are
the
ones
we're
focused
on
focusing
on
these
infant
mortality
failures
and
then,
as
a
you
know,
you
see
that
useful
life
and
towards
the
end,
you
start
getting
wear
out
things
get
age
things
change
a
little
bit.
You
start
getting
failures.
B
Provided
some
examples
of
some
of
these
infant
mortalities,
which
are
very
very
interesting,
and
these
are
real
problems
that
we
saw
both
at
brocade
and
some
of
my
previous,
my
previous
lives
at
other
companies,
so
I'll
take
you
through
each
one
of
these.
The
first
one
was
titanium
contamination.
This
was
a
voltage,
regular
regulator
that
we
had
on
a
board
and
that
supplier
received
a
defective
material
from
their
supplier
with
titanium
in
it.
B
And
basically,
what
happened
is
that
titanium
presented
itself
inside
the
park
and
these
parts
were
failing
in
our
factory,
didn't
fail
at
our
suppliers,
outgoing
test,
it
failed
in
our
factory.
Luckily,
these
were
failures
that
we
caught
feel
fairly
quickly,
so
no
risk
to
our
supplier
contained
in
the
factory.
That's
good.
B
Next,
one
printed
circuit
board
barrel
cracking
so
printed
circuit
board
the
green
thing
that
goes
inside
your
your
your
servers
and
such
that
actually
has
a
very,
very
complicated
manufacturing
process,
and
in
this
case
what
happened
was
one
of
our
suppliers
did
not
bake
this
part
long
enough
before
going
to
their
next
step.
So
there
was
a
little
bit
of
water
vapor.
That
was
that
that
was
a
resident
in
the
in
the
laminate
of
this
material.
B
So
what
what
happened
was
when
we
got
this
material
and
we
ran
it
through
our
reflow
profile
as
we're
soldering,
the
part
under
the
board
that
water,
vapor,
expanded
and
especially
boards
are
susceptible
to
expansion,
the
z-axis
and
what
it
did.
Is
it
cracked
these
barrels,
these
copper
barrels?
This
one
was
a
little
bit
tougher.
This
one
failed
later
on
in
our
process.
We
still
contain
it
in
the
factory,
but
is
it
still
was
an
issue?
And
then
this
last
one
is
very
very
interesting.
I
mentioned
the
time-to-market
considerations.
B
We
have
well
in
this
case
a
lot
of
times
in
order
to
achieve
that
that
that
that
time
to
market
we
Co
develop
products
with
some
of
our
suppliers.
So
in
this
case
we
were
designing
a
brand
new
router
and
at
the
same
time
one
of
our
suppliers
was
was
designing
one
of
their
components
to
go
in
that
router.
So
you
could
imagine
all
the
balls
in
the
air
when
they're
doing
development
we're
doing
development
all
the
things
that
could
go
wrong
and
in
this
case
something
did
go
wrong.
B
The
outgoing
test
process
at
our
supplier.
Basically,
they
forgot
to
initialize
one
of
the
power
supplies
and,
in
some
certain
conditions
that
power
supply
applied,
minus
seven
volts
to
these
devices
and
what
that
did
is
it
cooked
the
device
but
I?
Guess
it
didn't
fully
cook
it?
It
was
just
a
you
know:
it's
just
a
braised,
I
guess
and,
and
what
that
did
was
it
created
these
pinhole
defects
in
the
poly
silicon?
And
this
would
this
one
was
a
real,
tough
one.
B
Our
first
failure
on
this
was
30
hours
into
a
48
hour
test
and
unfortunately,
by
the
time
we
root
cause
this
we
had
shipped
some
out
to
our
customer.
So
what
did
we
do?
We
did
analysis
to
figure
out
how
many
of
these
would
fail
after
year,
one-
and
we
actually
did
proactively,
went
to
our
customers
and
said,
look
you're
not
having
failures
right
now,
but
we
feel
that
these
components
here
will
fail
at
a
rate.
That's
not
acceptable
to
you.
B
There
were
some
suppliers.
That
said,
we
want
to
swap
out
the
gear
and
we're
fine
with
that
there
were
other
suppliers.
That
said,
no
we're
not
worried
about
it,
you're
fine,
but
we
really
appreciate
you
telling
us
that,
so
you
know
I
think
in
both
cases
they
both
appreciate
us
going
to
them
proactively
and
saying:
look
we
think
you're
going
to
see
some
failures,
so
you
know
this
kind
of
runs.
B
Okay,
on
the
importance
of
tests
at
brocade,
we
test
the
heck
of
our
products.
We
really
do
and
we
spent
a
lot
of
time
doing
that
and
I
think
we
want
to
prevent
our
customers
from
seeing
these
types
of
failures,
actually
I
actually
remember
being
in
at
one
of
our
contract
manufacturers.
One
time-
and
it
was
my
product-
I-
was
bringing
it
to
market
and
it
had
passed
a
test
process
and
I
said:
let's
test
it
again,
I
think
I
was
like
no
we're
not
going
to
do
it,
no
you're
crazy,
don't
do
it.
B
B
One
of
the
things
about
that
testing
is,
it
generates
a
lot
of
log
data
and
and
we
need
to
analyze
that
log
data,
so
one
of
the
best
ways
to
do
that
is
time.
Series
evaluation,
there's
a
lot
of
time.
Aspects
of
this
analysis
that
are
important
and
I
gave
some
examples
here.
This
is
actually
a
real-world
project
that
we
a
real
product
that
we
worked
on,
and
this
shows
some
of
that
stress
testing.
B
You
can
see
the
temperature
going
from
minus
10
to
60,
so
it's
actually
stressing
solder
joints
and
stressing
performance
quite
well
and
the
red
triangles
their
show
where
things
were
failing.
Ninety
to
ninety-five
percent
of
our
failures
are
intermittent.
They
happen
under
certain
corner
cases,
and
this
type
of
visualization
really
helps
us
understand
the
conditions
under
which
things
fail.
You
know
if
it's
only
failing
at
hot.
That
gives
us
idea
on
how
to
do
root,
cause
analysis.
So
that's
one
aspect
of
it.
B
The
other
one
down
below
this
is
a
standard,
viable
plot
that
reliability,
engineers
use
and
they
use
it
too.
So,
for
example,
on
that
on
the
one
I
showed
you
with
the
with
the
pinhole
defects.
It
failed
30
hours
at
40
our
tests.
We
can
then
take
that
time
to
failure,
put
it
into
this
Bible
plot,
with
an
acceleration
model
and
figure
out
how
many
you're
going
to
fail
in
the
field.
So
this
would
actually
help
us
determine
whether
we
need
to
do
a
field
action
or
not.
B
So
I
want
to
talk
a
little
bit
about
how
these
files,
all
these,
these
log
files
that
were
talking
about
with
all
this
rich
information
is
brought
back.
You
know
to
to
our
systems
for
analysis
and
this
this
system
was
designed
around
two
thousand
six.
When
we
had
one
product,
it
was
very,
very
easy.
We
had
one
software
train,
so
it
was
very
easy
to
control
the
output
of
those
formats.
We
have
worldwide
manufacturing
sites
in
various
in
Asia
and
America
and
amia.
B
Those
files
are
stored
on
unser
vers
at
each
one
of
the
manufacturing
locations.
It's
our
sink
back
once
a
day
to
our
data
center,
and
then
we
have
this
parser
that
we
used
that
we
developed
this
written
in
Python
and
it
uses
an
interface
back.
So
what
I
did
as
I
went
to
the
software
teams
and
I
say:
look
I
need
something
to
Kiev
of
to
parse
this
data
out
of
your
files,
and
we
ended
up
with
hashtags.
You
know
we
have
things
like
ER
our
pound,
to
designate
an
error
and
few
other
things
there.
B
So
we
define
this
back.
It
was
great
with
one
software
team.
Very
very
easy
to
do
so
that
would
convert
the
raw
log
files
into
XML,
which
would
then
get
updated
into
our
data
warehouse
for
analysis.
So
this
was
a
good
days.
This
was
before
we
bought
foundry.
This
was
all
San.
Everything
was
unique.
Everything
was
consistent
and
life
was
good.
B
So
what
happened?
You
know
2006.
We
had
one.
This
is
a
few
years
after
after
foundry
we
had
seven
product
logs
seven
product
lines,
then
it's
seven
different
problem
during
seven
different
OS
teams
so
trying
to
maintain
that
interface
back
around
seven
teams
was
very,
very
difficult.
In
addition,
you
saw
you
know
I.
No,
it's
funny
the
the
development
teams
always
say
you
don't
have
big
data,
you
don't
have
big
data,
but
I
would
argue
that
data
we
have
is
complex
enough
to
to
to
use
some
of
the
big
data
tools.
B
So,
let's
see
so
when
that
happened.
This
is
that
this
was
these
were
the
consequences.
So,
first
of
all
our
parser
broke,
we
didn't
have
all
the
hashtags
available
anymore
because
of
the
size
of
the
data
the
rsync
processes
took
forever.
Some
of
the
other
downstream
processes
also
took
for
a
very
long
time.
B
In
addition,
we
had
to
buy
new
Nass
servers
to
store
all
the
new
data
that
way
that
we
had
and
because
that
parser
broke,
what
we
started
doing
is
each
individual
user
would
sync
files
from
here
and
develop
their
own
scripts
to
parse
them.
So
you
can
see
I
mean
this
is
a
mess.
You
may
have
some
some
files
going
in
one
direction,
some
files
going
another.
B
B
It
has
alerts
for
new
errors,
which
is
really
good
if
it
finds
an
error
that
it's
never
seen
before
and
give
us
an
alert,
we
can
parse
an
analyzed
six
gig
in
less
than
a
minute
storage
scalable,
because
the
files
we're
looking
at
rapid
access
to
files
and
full
search
capability
reporting
I
won't
go
to
that
into
detail.
The
system
needs
low,
latency
flexibility
and
monitoring.
So
you
can
see
the
approaches
we
use
job
and
storm
Cassandra
and
solar,
and
then
we're
still
actually
deciding
on
the
on
the
reporting
side
timeline
real
quick.
B
We
started
our
consulting
engagement
with
impetus
in
October.
It
looks
like
about
a
year
project,
October
12,
through
October
13.
We
developed
a
proof-of-concept
parser
and
a
reference
architecture,
functional
spec,
and
then
we
expect
parser
and
storage
to
be
online
in
August
reporting
in
October.
So
we're
still
in
the
middle
of
the
process,
I
think
we're
going
to
have
our
first
version
of
that
of
that
parser
next
week.
B
So,
let's
see
I
did
want
to
talk
about
this
because
because
we're
preparing
our
organization
to
take
all
these
tools,
so
one
of
the
things
we
did
is
establish
metrics
up
front.
We
have
this
aggregate
yield
metric
that
rolls
out.
This
is
used
to
establish
bonuses
at
our
company
and
it
rolls
all
the
way
up
to
executive
management,
all
the
way
down
to
the
individual
contributor.
So
it's
a
pretty
powerful
metric
that
gets
everyone
aligned
on
what
we're
doing.
We
also
established
processes
to
support
those
metrics.
So
this
is
an
example
of
yield.
B
Metric
I
won't
go
into
this
because
I
think
I'm
running
a
little
bit
out
of
time.
So
if
anyone
has
any
questions
on
this,
they
can
ask
me
later.
This
is
the
this.
Is
the
concept
of
the
parser
a
flexible
parsing
engine
with
adapters
for
each
types
of
log
files,
those
adapters
defined
configuration
environmental
information,
time,
temperature,
etc
tests
specific
tests
that
are
run
and
error
information
a
little
bit
about
ROI
I
want
to
talk
about
that
briefly.
I
spend
a
lot
of
time
talking
my
team,
how
you
spending
your
time?
What
are
you
doing?
B
The
green
represents
the
stuff
that
we're
talking
about
here.
We
estimate
by
automating
that
we
can
get
ten
to
twenty
percent
efficiency
gains
per
person.
We
can
also
contain
work
non
conforming
material
in
the
factory,
so
that's
kind
of
the
harder
or
ROI
the
soft
collaboration
between
teams,
both
in
the
factory.
We
get
log
files
from
the
field
too,
so
we
need
to
be
able
to
collaborate
or
correlate
that
data
where
we
also
get
data
out
of
development.
B
So
if
we
can
actually
have
a
holistic
view
of
everything,
that's
happening
in
all
those
locations,
that's
great,
let's
say
I
won't
go
in
the
rest
of
that
detail.
Other
possibilities
that
exist
we
have,
as
I
mentioned
here,
all
the
machines
we
use
to
manufacture
our
product.
Those
all
generate
images
and
log
files.
If
we
can
capture
all
those,
then
we
have
a
much
more
holistic
view
of
what's
happening
in
the
manufacturing
arena
and
then
finally,
keys
to
success.
B
I
wanted
to
include
a
few
of
these
technology
is
easy
all
right,
but
people
are
people
are
hard.
It's
really
easy.
In
technology
we
have
equations,
we
have
other
things,
but
people
don't
necessarily
have
equations.
So
that's
the
tough
part,
so
relationships
are
important.
I
heard
someone
say
you
know:
bring
the
IT
team
out
for
drinks,
I'm
from
new
orleans
foods
important
to
me,
so
we
all
went
to
dinner.
That
night
we
found
out
about
each
other's
families.
We
found
out
about
our
kids
everything
else.
Those
relationships
are
very
very
important
to
when
it.
B
When
issues
happen,
you
can
fall
back
on
those
relationships
and
it's
not
that
big
of
a
deal
metrics
and
processes
already
talked
about
clear,
are
NRS,
but
one
of
the
key
successes
of
our
team
is
we
had
people
that
can
straddle
both
areas.
Our
last
con
call
people
were
asking
development
teams
were
asking
me
questions
about
my
business
and
and
our
IT
team
was
asked
for
answering
all
my
questions.
B
So
that's
that's
a
key
success
sign
for
a
highly
functioning
team
context
and
not
just
content,
and
then
this
last
one
is
big
look
hard
for
win-win
opportunities.
There
are
other
organizations
that
need
different
things
on
a
project
like
this.
It
may
mean
more
risk
for
me
to
develop
this
project,
but
it
gives
the
other
organizations
the
teeth,
the
information
they
need
to
make
this
thing
successful.
So
that's
another
big
one.
Okay,
so
that's
it
for
me.
C
So
as
thanks
rich
and
as
rich
mention,
you
know
they
started
the
consulting
engagement
with
impetus
about
a
about
nine
months
ago
and
we
were
able
to
push
this
project
through.
It's
almost
near
production
I'm
not
going
to
spend
a
lot
of
time
on
the
architecture
I'm
just
going
to.
Basically,
this
is
a
fantastic
use
case
of
real-time
analytics
on
machine
data,
and
those
of
you
are
interested
feel
free
to
come
talk
to
me
later,
so
it
uses
storm
Cassandra
as
well
as
solar,
and
this
is
sort
of
the
high-level
hope
of
you.
C
Don't
don't
worry
that
I
think
the
slides
are
going
to
be
available
later
on
and
we
actually
did
a
detailed
technology
comparison.
So
it
wasn't
that
we
just
picked
Cassandra
out
of
the
blue.
Several
competing
technologies
were
evaluated
and
Cassandra
came
out
on
top
because
of
several
reasons,
some
of
which
you
can
see
out
here,
and
this
is
about
impetus.
So
we
were
one
of
the
early
drivers
for
big
data
adoption.
We
had
a
big
data
SI
and
a
consulting
company
offering
analytic
solutions
as
well
as
big
data
platforms.