►
From YouTube: Filecoin Dev Demo 04 - PoRep Benchmarking Tools
Description
How to set up and use Filecoin's benchmarking tools.
A
Falcone
proof
benchmark
tools
can
be
used
to
evaluate
running
times
for
Falcone's
replication,
algorithms
and
proofs
Falcone
uses
proofs
of
replication
or
PO
reps
to
check
and
prove
that
storage
miners
are
storing
their
clients.
Data
correctly
developers
and
researchers
can
use
these
benchmark
tools
to
understand,
study
and
compare
power
up
algorithms
and
parameter
choices.
In
this
demo,
we
will
show
how
to
install
and
use
some
of
the
benchmark
tools
to
measure
the
run
time
of
the
zigzag
poep
algorithm.
This
is
file
coins.
A
Current
poep
construction-
please
note
everything
you
see
in
this
demo
is
under
active
development
and
likely
to
change
the
proof,
implementations
and
benchmark
tools
live
in
the
rust
proofs
repo.
You
can
find
instructions
for
how
to
install
and
run
these
benchmarks
in
the
repo
readme.
First,
you
will
need
rust
and
rust
nightly
next
clone
the
repository.
Then
we
go
into
the
new
rust
proofs
directory
and
we
build
all
proof
examples.
A
A
Today
we
will
focus
on
zigzag,
which
is
the
file
coin
protocols.
Current
po
rep
construction.
This
benchmark
can
be
run
with
many
different
parameters
in
order
to
see
all
of
the
options,
as
well
as
the
default
parameter
values
we
can
pass
in
the
help
flag.
The
only
required
parameter
is
size.
It
specifies
the
size
and
kilobytes
of
the
data
to
replicate.
There
are
other
key
parameters
here
as
well,
such
as
the
number
of
layers
in
the
PO
rep
circuit,
which
defaults
to
10.
A
You
can
learn
more
about
the
zigzag,
poep
construction
and
what
each
of
these
parameters
means
by
visiting
the
rust-proof
repo
on
github
or
reading
the
papers.
Let's
start
by
running
the
zigzag
benchmark
on
1024
kilobytes
of
data
using
the
default
options
for
other
parameters.
You
can
see
the
program
starting
to
replicate
the
data
in
layers.
As
we
noted
earlier,
there
should
be
ten
layers
in
total.
Once
the
replication
is
completed,
the
program
will
generate
a
zigzag,
poep
proof.
A
A
A
A
Replication
time
for
one
gigabyte
of
data
was
four
thousand
five
hundred
and
nineteen
seconds
or
around
seventy-five
minutes.
While
the
time
to
generate
a
proof
was
two
hundred
and
seven
point
five
seconds
or
around
three
and
a
half
minutes,
we
can
run
benchmarks
with
different
parameters
on
different
hardware
setups
and
with
new
optimizations,
to
find
configurations
that
lead
to
faster
replication
and
proving
times
these
benchmarks
will
change
over
time.
As
our
proof,
implementations
continue
to
develop,
developers
and
researchers
can
use
these
tools
to
understand,
study
and
improve
performance
characteristics
of
file.