►
From YouTube: Fleek Network - Decentralized Content and Application Delivery Network - Mahmoud Shehata
Description
This talk was given at IPFS Camp 2022 in Lisbon, Portugal.
A
So
what
is
fleet
Network?
It's
a
decentralized
permissionless,
Content
delivery
Network.
So
anybody
could
join
the
network
and
there's
no
like
Central
server
or
Central
Authority.
That
controls
it.
It
scales
with
demand
and
it
serves
web3
contacts.
So
it's
agnostic,
so
that's
a
good
property
of
ipld,
but
it's
also
agnostic
to
the
underlying
storage
layer.
A
A
So
who
is
this,
for
everyone
needs
a
CDN,
but
we
desperately
need
one
for
web3
at
Fleek
we
serve
websites
and
storage
and-
and
we
would
be
the
largest
user
of
the
CDM
but
beside
us.
There's
video
platforms
like
live
peer,
there's
marketplaces,
nft
marketplaces
and
then
there's
D5
front
ends
and
there's
a
lot
of
users
that
we
think
are
going
to
use
the
CDN
and
be
yeah,
be
a
bit
big.
So
why
we're
doing
this
simply
to
accelerate
all
web
3
content?
A
So
I
like
to
view
it
as
fleeknetwork.
You
know
pulling
information
or
probably
storage
and
caching
it
around
the
world
from
all
these
storage
providers
and
we're
starting
with
falcoin.
A
So
let
me
give
you
a
distribution
of
the
number
of
requests
between
you
know,
storage
and
websites.
We
do
80
20,
so
it's
80
of
all
our
requests
are
actually
storage
based
and
20
are
for
websites
and
globally.
81
of
all
requests
are
actually
video
based
and
then
everything
else
is
like
websites
and
gaming
Etc
and
according
to
Cisco,
and
then
Cisco
also
says
that
71
of
all
internet
traffic
actually
goes
through
cdns.
A
A
A
You
know-
and
you
know,
mask
it
as
https,
basically
and
like
we're,
trying
to
obfuscate
the
web
traffic.
We're
trying
to
encrypt
the
web
traffic
and
also
use
Quick,
which
we
do
between
our
node
communication,
but
recently,
we've
also
seen
that
quick.
You
know
version
one
is
being
censored
at
multiple
countries
in
the
world
so
but
we
need
to
all
be
cognizant
of
that
anyways.
A
So
here
is
how
you
know
the
breakdown
is
of
flick,
Network
compared
to
a
traditional
CDN
where
cache
nodes
you
know,
operate
and
communicate
with
one
another
afforded
by
this
content
addressable
medium
and
they
can
scale
their
in
their
region
based
on
the
demand
by
not
having
to
hop
to
a
central
server
to
get
content,
and
they
can
also
stream.
A
So
it's
trustless,
it's
tamper-proof
and
that's
how
all
the
data
transmission
between
all
these
different
actors
communicate
on
and
that's
the
medium
so
cache
nodes
set
up
their
own
DNS
and
TLS
they're,
also
responsible
for
data
replication,
so
they're
responsible
for
saying,
okay
I
should
replicate
this
data
amongst
multiple
peers
and
they're,
also
responsible
for
routing
set
data
and
peer
Discovery,
obviously
like
we
know
and
then
cache
accountability.
So
if
you
serve
a
cache,
you
should
account
for
it.
A
So
you
can
get
incentivized
and
then
obviously
index
providing
which
is
very
important
because
you
can
announce
to
an
indexer
when
they
have
cash
and
you
can
re-announce
again
when
you
don't
have
that
cash
and
that's
kind
of
the
edge
node
and
then
in
front
of
it
there's
a
proxy.
And
then
we
have
an
indexer.
A
So
Gateway
knows
open
sessions
between
themselves
and
the
cache
nodes
and
that's
when
they
serve
requests
and
that's
how
cache
accountability
happens.
They
also
have
they
communicate
with
network
indexers
and
that's
how
we're
going
to
get
into
content
routing
in
a
bit,
but
that's
how
they
have
an
index
and
they
use
a
cherry
picker
based
on
the
geolocation
of
the
request
and
that's
how
they
can
communicate
to
the
these
Edge
nodes
for
gatewayners.
A
A
So
we
use
the
default
content
routing
which
is
like
you
know,
using
the
the
DHD
using
gossip,
but
we
also
have
combined
or
composed
consistent
hashing
index
providing
and
Bloom
filters,
which
is
something
we're
getting
into
right
now
and
with
consistent
hashing.
A
That's
what
I
talked
about
when
you
can
replicate
your
content
amongst
other
peers
and
for
index,
providing
that's
where
it
provides
you
the
fast
lookups
like
in
the
milliseconds,
so
we
can
know
exactly
where
the
content
is,
and
that
happens
on
the
gateways,
as
well
as
the
cache
nodes
themselves
and
then
Bloom
filters
are
a
really
interesting
topic.
We're
like
diving
into
now,
which
is
a
cache
node,
can
broadcast
its
cache
content
to
the
set
of
its
peers
and
then
that's
how
they
can
locate
their
data
amongst
a
local
DHT.
A
A
A
A
That's
where
consensus
on
the
Quorum
on
on
these
batches
happen
and
then
that's
how
the
signatures
get
aggregated
and
eventually
we
know
that,
like
the
availability
of
these
transactions
on
other
nodes,
so
this
is
why
we
originally
start
with
hot
stuff,
and
then
we
found
some
bottlenecks
there
and
that's
recently.
We
got
into
researching
about
narwhal
and
bull
shark
as
our
consensus.
So
again,
I'm
gonna
get
back
to
this
diagram.
A
So,
while
Falcon,
it
has
a
great
concept
which
is
a
retrieval
Market
or
retrieval
clients
and
I'm.
Pretty
sure.
A
lot
of
you
guys
know
already
about
this.
It's
very
competitive
and
it's
very
cheap
to
store
on
filecoin
there's,
also
a
consensus
lab,
which
is
an
amazing
group
of
people
who
are
doing
great
stuff
in
consensus,
and
that's
where
you
know
I'm
digging
into
our
consensus
mechanism
and
yeah,
there's
also
ipfs
and
liquid
TV
and
there's
many
more
and
obviously
an
amazing
community
and
yeah.
They
store
a
lot
of
bytes
of
data,
that's
huge!
A
So
by
the
end
of
the
year,
we
want
to
have
a
large
test.
Bat
running
where
a
thousand
nodes
of
cache
nodes,
basically
running
a
million
requests
per
node
and
a
bill
unit
request
by
the
end
of
the
year
for
the
entire
network,
and
you
know
we
could
run
it
ourselves,
but
we
might
need
like
the
community
to
run
it
with
us
and
then
obviously
like
the
the
metric
which
a
lot
of
people
talk
about,
which
is
time
to
first
bite.
A
You
know
we
do
want
to
Target
the
North
Star,
which
is
15
to
20
milliseconds,
but
our
goal
really
is
to
figure
out
how
to
speed
up
content
routing,
which
is
very
important,
I.
Think
if
we
can
massively
speak
content
routing
the
time
to
First
byte
will
always
like
be
good
enough
or
CDN
level,
and
then
we
have
missions.
So
I
have
some
time
to
explain
to
you.
A
We
just
finished
it
with
the
alpha
and
content
routing
and
then
end
of
the
year,
we're
integrating
filecoin
and
then
our
consensus,
which
is
narwhal
and
then
we're
gonna,
get
into
q,
n
minus
one
or
just
Seven
Stars,
and
next
year
is
going
to
be
governance
and
just
tap
into
all
these.
You
know
retrieval
markets
with
at
Falcon
and
other
storage
protocols
and
yeah.
A
We
started
it
last
year
with
ipfs
fan
and
then
it's
coming
back
full
circle
now
and
we
want
to
explore
creating
a
modular
compute
over
data
framework
that
you
know,
Cod
and
baklao
is
doing
more
likes
cluster
based,
which
is
amazing,
and
we're
getting
inspired
a
lot
from
that
project.
But
we're
also
using
this
framework
to
create
a
stateless
sort
of
serverless
framework,
and
then
you
can
tap
it
with
apis
into
a
statefulness
with
crdts
and
databases
and
then
obviously
for
a
CDN.