►
From YouTube: ESPA - Supermicro's solutions for the Filecoin ecosystem
Description
Vik Malyala (SVP) with Supermicro discusses solutions for the Filecoin ecosystem at the Enterprise Storage Provider Accelerator (ESPA).
ESPA is a program built entirely around the use of Filecoin’s Web3 infrastructure. The decentralisation of storage providers and the availability of globally diverse and auditable data is a solution to the misaligned incentives of the server based Web 2.0 storage providers of today. It also sees added benefits of a flexible open spot market for storage and affordable prices set by independent storage providers bidding for the right to house data. All of which are important considerations for enterprise needs and large datasets.
ESPA, an initiative sponsored by PiKNiK, Protocol Labs, Seagate and AMD benefits everyone from the participants taking in new knowledge and know-how to the end users who will have a more efficient, secure, and robust way to store their data.
At the end of the program, participants will be able to leverage the untapped storage space available on Filecoin by onboarding large datasets. That is 50 PiBs of storage, 1 PiB each, over the course of a year, where 5-10 percent will be verified data stored by participants.
Sign up for upcoming cohorts as a participant:
ESPA: https://www.web3espa.io/?utm_source=youtube&utm_medium=videodescription&utm_campaign=YT
Stay connected with us:
Twitter: https://twitter.com/web3ESPA
LinkedIn: https://www.linkedin.com/company/web3espa/
Learn more:
PiKNiK: https://www.piknik.com/?utm_source=youtube&utm_medium=videodescription&utm_campaign=YT
Protocol Labs: https://protocol.ai
A
So,
first
of
all,
you
know,
thank
you
very
much
for
having
me
personally,
I
am,
you,
know,
quite
excited
about
november
3
technology
as
such,
because
it's
the
game
changer.
As
you
know,
every
one
of
us
here
in
the
room
now-
and
I
am
vic
maliala-
I
am
the
president
for
supermicroamia
as
well
as
I
manage
the
worldwide
field.
Applications
engineering
for
supermicro
I've
been
in
the
company
for
over
13
years,
and
before
that
I
was
an
asic
designer
developing
multiple
generations
of
chipsets.
So
you
know
when
I
see
the
technologies
coming
together.
A
A
Optimized
server
building
blocks,
that's
how
it
started
off
which
basically
based
on
intel
x86
servers
at
that
time
and
over
a
period
of
time
we
have
evolved,
and
we
started
focusing
on
complete
solutions,
fully
integrated
rack
solutions
and
whatnot
intel
amd
based
and
it's
a
company
that
we
expanded
into
other
geos.
So
we
have
office
in
netherlands
and
we
have
in
japan,
korea
and
all
that
stuff.
A
A
What
you
can
see
here
is
that
supermicro
as
a
company,
we
work
very
closely
with
our
partners
whether
it
is
related
to
our
channel
partners
that
are
taking
our
solutions
and
enabling
the
customers
or
customers
that
are
taking
our
products
directly.
So
what
idc
recognizes
is
the
ones
that
goes
in
as
a
super
micro
branded.
So
that's
why
you
see
the
number
around
like
4.9
percent,
but
realistically
it's
between
eight
and
nine
percent,
with
about
1.2
million
units
that
we
ship
every
year
before
we
jump
into
the
crux
of
it.
A
A
If
you
take
a
look
at
our
product
line
from
supermicro,
we
have
been
investing
in
our
product
portfolio
to
bring
different
variants
of
technology
together,
so
at
a
very
basic
level,
we're
talking
about
compute,
storage
and
networking
and
how
we
bring
that
together.
So
when
you
take
a
look
at
the
platforms
that
are
going
into
the
data
center
side,
it
ends
up
being
very
high
in
efficiency,
and
you
need
to
have
something
that
is
highly
performant
depending
on
the
type
of
data
center
and
what
kind
of
workloads
that
we
need
to
be
working
on.
A
A
A
You
know
local
storage
and
we
at
at
the
same
time.
We
also
have
a
need
for
people.
You
know,
for
the
systems
to
be
supporting
both
front
and
rear.
I
o,
depending
on
which,
which
kind
of
data
centers
should
go
to
now
every
one
of
us
talking
about
5g
telco
and
you
know
smart
edge
devices
and
whatnot.
All
these
things
require
a
small
form
factor.
A
You
know
in
that
they
were
the
first
to
bring
pc
express
gen
4
into
the
market,
so
because
of
that,
certain
platforms
can
actually
take
a
much
better
advantage
of
it.
I
will
get
into
the
details
as
to
why
amd
that
I
am
talking
about
currently,
but
when
you
take
a
look
at
these
platforms,
what
we
started
with
is
the
platforms
that
actually
can
support
a
dual
socket
lots
and
lots
of
memory.
You
know
you
know
it's
about
32
dimms,
for
example.
A
A
The
ultra
platforms
is
a
classic
example
for
that
and
then,
when
you,
when
you
are
looking
at
a
single
socket
platforms,
you
can
still
have
all
the
benefits,
mainly
because
of
amd
processor,
which
is
the
epic
series
having
128
lanes
of
pc
extras.
So
this
gives
us
the
ability
to
have
you
know
more
networking
cards
or
more
gpus
or
more
storage,
directly
connect
connected
to
the
processor,
and
there
are
reasons
why
people
will
be
using
multi-node.
They
can
bring
lot
more
density
into
the
rack
so
which
is
the
reason
we
have
either
two?
A
U
with
four
nodes,
or
you
know,
four?
U
with
eight
nodes,
and
we
also
have
a
blade
which
supports
20
nodes
in
a
atu
form.
Factor
accelerators
is
something
that
I
mentioned.
One
of
the
things
that
we
look
at
is
how
we
bring
the
right
balance,
so
we
do
have
a
dual
socket
platform
that
can
support
up
to
eight
gpus,
but
on
the
far
less
is
far
left
is
what
you're
seeing
is
a
gpu
platform.
A
A
A
A
That
is
where
we
are
headed,
and
in
order
to
do
that,
we
need
to
figure
out
a
way
to
get
the
rightly
sized
building
block
now
that
actually
can
help
us
to
build
from
that
into
any
solution
that
we
are
looking
for,
depending
on
the
scale
that
we
are
going
for.
So
this
is
a
very
quick
slide
in
here
and
when
we
are
looking
at
a
compute,
the
compute
can
be
a
single
or
multi
socket.
As
I
mentioned,
and
a
storage
point
of
view.
A
At
one
end
of
the
spectrum,
we
are
talking
about
very
high
performance,
parallel
file
system,
supported
using
nvme,
all
flash
rs
and
other
end.
We
are
talking
about
high
capacity
storage,
just
using
the
top
loading.
You
know,
storage,
which
can
get
to
a
petabyte
type
of
storage
in
a
single
box
which
is
not
going
to
be
great
for
performance,
but
it's
going
to
help
with
capacity
in
terms
of
accelerated
platforms.
We
are
going
to
have
lot
more
gpus,
so
in
a
product
categorization
point
of
view:
compute
storage,
accelerated,
compute
and
the
edge
devices.
A
These
are
the
four
broad
categories
that
we
are
focusing
on
when
we,
when
we
start
applying
these
into
what
we
can
do
with
the
requirements
here.
Whenever
we
are
looking
at
a
standard
system
deployment
in
an
enterprise
data
center,
we
look
at
basic
criteria.
You
know
what
is
the
performance
that
is
required
and
what
is
the
balance
of
the
processor
to
the
gpu
to
the
networking
that
is
required.
A
So
in
some
cases
people
want
to
have
all
the
gpus
to
be
hanging
around
hanging
on
a
single
pcie
root,
complex
or
you
can
have
something
that
need
to
focus
on
that
need
to
be
connected
to
the
processor
using
the
pci
by
16.
So
you
have
unrestricted,
you
know
bandwidth,
and
once
you
have
this
balanced
system,
then
you're
talking
about
what
kind
of
software
that
need
to
run
on
it
and
how
do
we
ensure
that
it
is
qualified
when
you
deploy
systems
at
scale?
A
A
You
don't
want
to
be
spending
your
you
know
your
dollars
on
the
cooling
of
this
equipment,
but
you
wanted
to
have
as
efficient
systems
as
possible.
Form
factor
becomes
an
important
thing
most
of
the
times.
People
like
these
1u2u
pizza
boxes,
because
it's
easy
for
single
person
to
handle
it
versus
people
who
are
having
a
lot
more
power
and
ability
to
handle
these
systems
at
scale
can
use
multi-node
systems,
and
that's
where
the
density
will
come
into
picture
and
environmental
impact.
A
You
want
to
be
developing
these
systems
in
a
way
where
these
guys
will
run
for
a
long
period
of
time.
They
don't
break
easily
and
we
will
consume
less
power
and
we'll
be
able
to
reduce
the
pue
close
to
1.0
to
the
best
we
can,
and
these
are
all
the
ones
that
will
help
with
the
environmental
aspect
of
it
and
when
you
deploy
at
scale.
Another
thing
also
important
is
to
be
able
to
manage
the
systems
easily.
A
This
all
of
it
will
come
into
the
total
cost
of
ownership.
The
thing
about
filecoin
is,
if
you
take
a
look
at
all
the
requirements
for
that,
this
is
exactly
same
as
what
is
needed
for
the
systems
to
be
deployed
in
a
large
scale
in
data
centers.
So
this
is
the
reason
you
know
why
the
systems
as
they
are
using
a
data
center
level
when
it's
sized
right
for
this
application,
it's
going
to
provide
the
best
value
and
best
performance.
A
So
what
are
we
doing
here?
If
you
take
a
look
at
it
when
we
work
with
espa,
the
initial
requirements
were,
like
you
know,
hey.
We
have
all
these
different
hardware
that
we
are
working
and
we
have
in
a
like
a
building
block,
but
we
don't
know
necessarily
what
is
most
optimal.
So,
in
order
to
do
that,
we
have
to
work
together
and
with
the
file
coin,
when
we
are
implementing
on
our
systems,
we
need
to
see
the
different
phases
of
the
compute.
A
A
By
the
same
time,
when
you
have
a
lot
more
cores,
you
also
need
to
have
a
lot
more
memory
to
support
that
and
in
order
to
support
bigger
memory
footprint,
as
I
was
mentioning,
we
ideally
would
like
to
have
as
many
memory
slots
as
possible.
So
we
can
go
for
a
lower
capacity
memory,
which
is
cheaper
but
still
will
be
able
to
get
a
higher
overall
system
level
capacity,
and
you
would
like
to
have
nvme
support,
and
this
is
how
we
started
so.
A
A
single
socket
epic,
64
core
and
up
to
a
terabyte
of
memory
and
getting
three
of
the
3.84
terabyte
nvme
devices
and
networking
point
of
view
at
the
very
minimum
at
10.
Gig
is
needed
and
that's
what
we
put
it
together
as
we
go
into
the
pc2
and
c2
phases,
which
is
where
we
realized
that
you
know
having
a
gpu
is
important,
but
having
more
gpus
per
processor
will
give
us
the
ability
to
map
it
properly.
A
So
we
pick
this
particular
platform,
which
is
a
2u2
node,
the
details
that
are
going
to
come
in
the
next
slider
too,
so
where
each
of
these
processors
is
mapped
with
up
to
three
of
the
nvidia,
you
know
rtx
a5000,
gpus
and
which
requires
you
know.
You
know
we
are
able
to
support
two
of
the
7.68
terabyte
drives
and
with
more
networking
is
needed
on
that
when
we
are
going
with
25
gig
as
a
network
controller,
so
so
the
system,
you
know
if
you
were
to
look
into
that.
A
So
let
me
see
this
is
the
first
system
so
for
this
for
the
first
phase,
so
this
particular
one
why
it's
important
I
kind
of
mentioned
that,
first
of
all,
we
wanted
to
have
something
that
is
cost
effective.
We
don't
want
any
more
bells
and
whistles.
We
want
something
that
exactly
optimized
for
that
one.
So
we
need
to
have
a
right
mapping
of
a
cpu
to
its
memory.
That
is
the
first
one,
first
and
foremost
thing
which
we
were
able
to
get
in
this
particular
platform.
A
Second,
one
is
today
based
on
the
applications
we,
you
know,
we
are
using
three
of
the
nvme
devices,
but
the
system
can
actually
support
up
to
ten
so
depending
on
what
would
potentially
be
needed
in
future.
This
is
kind
of
future
proof,
but
for
now
three
drives
is
sufficient,
which
is
what
is
needed
for
the
scratch
space
and
built
the
the
10
10
gigabit
ethernet.
You
know
we
can
have
an
option
to
have
a
specific
card
if
you
want
by
using
an
add-on
card,
but
the
on-board
itself.
A
We
have
a
dual
10
gig,
which
will
be
more
than
sufficient
and
in
its
enterprise
centric
controllers,
which
will
actually
be
sufficient
for
this
workload
and
from
the
processor
point
of
view,
we
are
able
to
get
support
for
the
entire
spectrum
of
the
epic
processor
family
up
to
64
cores,
so
that
is
the
highest.
So
we
are
able
to
do
that
and
by
having
a
turin
per
channel,
we
are
able
to
also
reduce
the
cost
of
it.
A
A
The
second
one
we
are
talking
about
for
the
second
and
third
phase
is
for
the
pc2
and
c2
phases.
We
wanted
to
be
able
to
support
gpus,
so
this
particular
platform.
I
am
actually
quite
excited
about
mainly
because
the
design
of
this
platform
is
to
support
both
epic,
as
in
you
know,
for
example,
a
third
generation.
Epic,
is
called
milan.
We
support
that,
but
at
the
same
time,
this
platform
is
also
designed
to
support
the
third
generation
amd
ryzen
threadripper
pro,
which
gives
higher
frequency
per
processor
core.
A
So
for
the
you
know,
within
the
same
platform,
you
know,
depending
on
what
we
are
trying
to
accomplish,
we
will
be
able
to
use
the
epic
or
as
well
as
the
threadripper
pro,
so
it's
compatible
with
both
of
them
and
because
we
are
able
to
put
up
to
three
gpus.
We
would
like
to
be
able
to
have
as
much
networking
connectivity.
A
In
many
cases
you
know
people
tend
to
use
100
gig
because
it's
reasonable
in
terms
of
cost
and
this
particular
system.
The
way
we
design
is
it's
a
two
nodes,
one,
you
know
one
in
the
top
one,
the
bottom
and
the
power
is
shared,
so
you
can
think
about
when
you
have
redundancy
instead
of
having
each
node
having
its
own
two
power
supplies.
Now
we
are
sharing
the
same
power
across
two
nodes.
We
are
sharing
the
cooling
across
two
nodes
which
improves
the
efficiency
of
the
platform.
A
A
So
where
does
it
take
us?
If
you
take
a
look
at
if
you
take
a
look
at
the
development
efforts,
we
want
to
be
able
to
create.
You
know
a
solution
that
easy
for
people
to
adopt.
So
hardware
is
different
from
the
software.
In
the
sense
people
who
are
good
at
providing
hardware,
they
may
or
may
not
know
the
software
that
well
are
the
people
who
know
how
to
use
the
software.
They
may
not
know
the
hardware
aspect.
A
A
This
is
going
to
have
enterprise
centric
features,
so
one
can
actually
take
it
and
deploy
at
any
scale.
Second
thing
is:
the
technology
is
changing.
As
you
can
see,
every
year
and
a
half,
we
are
getting
new
processors.
The
networking
is
getting
faster
and
whatnot,
so
we
are
able
to
take
the
newer
generation
process
from
intel
and
amd
the
latest
gpus
from
nvidia,
and
we
are
able
to
continue
to
validate.
So
we
will.
Our
hope
is
to
improve
the
performance
per
dollar
and
performance
per
watt
generation
over
generation.
A
So
we're
going
to
invest
time
in
this
one.
The
third
one
is
the
web.
3
technologies
itself
is
evolving
quite
a
bit
and
the
workloads
are
going
to
change
a
lot
more.
So
the
goal
of
this
is
basically
to
continue
to
invest
our
time
and
effort
in
expanding
the
product
portfolio
that
that
is
going
to
be
optimized
for
the
new
and
emerging
workloads
in
you
know
web
free
space,
and
the
fourth
is
to
collaborate
with
espa
to
enable
more
storage
providers
in
this
in
this
system.
Why?
A
Because
the
whole
idea
of
doing
this
is
to
ensure
that
one
don't
have
to
go
through
this
overall
exercise
all
over
again,
we
wanted
to
make
that
option
easy.
We
wanted
to
take
the
burden
on
ourselves
to
run
these
different
workloads
on
this,
and
that
way,
one
whoever
wants
to
adopt
these
platforms
and
deploy
it
at
any
scale.