►
From YouTube: Dragonfly: Make Image Distribution Efficiently and Safely in Cloud Native - Yuxing Liu & Tao Peng
Description
Don’t miss out! Join us at our upcoming event: KubeCon + CloudNativeCon Europe 2021 Virtual from May 4–7, 2021. Learn more at https://kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects.
Dragonfly: Make Image Distribution Efficiently and Safely in Cloud Native - Yuxing Liu, Alibaba Cloud & Tao Peng, Ant Group
With the increasing scale of cloud native services in industry, how to distribute images efficiently is a new challenge for enterprises. Dragonfly is an intelligent P2P based container image distribution system which provides a native image distribution solution. In this session, we will cover the general Introduction to introduce you how dragonfly makes image delivery fast in data center. In addition, we have developed a new container image service that aims to provide more efficient way to deliver container images and address some outstanding issues of the current OCI image spec. We will dive into the new image service and show how it supplements to Dragonfly’s goal of making container image distribution fast and secure. Audiences will also learn the latest updates and what is next step in community.
https://sched.co/ekGa
A
A
A
A
A
This
chart
shows
the
time
consumption
of
the
downloaded
downloading
files
using
native
word,
knitting
mode
and
the
jungle
play
mode.
The
knitting
mode
means
that
all
nodes
downloaded
directly
from
the
source
in
the
chart,
the
x-axis
means
the
number
of
the
concurrent
downloads
and
the
y
axis
represents
the
download
time.
A
A
B
A
A
Jungfla
has
undergone
large-scale
internal
verifications
and
decided
to
open
sources
in
the
community
in
it
in
2018
trunk
flight,
spotted
the
image
distribution
and
donated
to
the
sincere
food
foundation
as
an
image.
Distribution
too,
as
in
imports
this
year,
is
under
the
sincere
incubating
stage
in
the
future.
Jungle
still
has
many
directions
to
exploit
and
we
welcome
all
of
you
to
join
this
project.
A
Firstly,
let's
analyze
the
core
of
jump,
flight,
p2p
and
cdn
mechanism
mechanism,
junk
flare
has
two
components:
dfcat
and
zoom.
Node
devcat
is
a
client-side
component
that
needs
to
be
installed
on
each
node,
which
takes
a
jaw
of
peer
in
a
pdp
network,
including
download
files
from
other
peers
and
acting
as
a
uploader
to
support
other
peers
to
download
the
files
from
it.
A
Supernode
is
a
server
set
component.
It
mainly
provides
these
capabilities.
First,
it's
a
track
is
a
tracker
and
schedule.
A
scheduler
in
the
pdp
network
that
choose
appropriate
downloading
downloading
netpass
for
hps
for
hp
sector
is
also
a
cdns
server
that
catch
download
catch
the
downloaded
data
from
source
to
avoid
downloading
the
same
files
from
source
repeatedly
and
secondly,
and
necessary.
A
A
Looking
forward
to
obtaining
the
scheduling
results
that
that
is
which
peer
has
which
blocks
after
subnaut
receives
the
download
request.
It
will
first
check
whether
the
file
is
cached
locally,
if
not
download
it
from
the
source
and
catch
it
and
knock
late
locally.
Then
the
super
node
will
return.
The
scheduling
results
according
to
first
the
recorded
recorded
relationship
between
the
peers
and
blocks
and
to
the
scheduling,
organizer
algorithm.
A
A
After
they
have
get
gets
scheduling
results
it
will
download
the
corresponding
blocks
from
the
peers.
According
to
the
scheduling
results
after
downloading
a
blog
df-guest
will
do
this
since
the
first
one
report
to
the
supernode
that
it
has
downloaded
a
block
and
then
probably
provide
a
provider.
The
ability
to
download
this
block
for
other
peers,
then
repeat
the
steps
too
too
fail
and
you
almost
enter
or
not
have
downloaded
the
interface
and
continued
okay.
I
hope
that
through
the
above
explanation,
you
have
a
good
understanding
of
dragonfly's
core
principles.
A
A
Firstly,
we
need
to
know
when
we
run
the
image
pool
command.
What
will
happen?
According
to
the
ocr
distribution?
Spec
spell
classification,
the
process
of
putting
an
image
around
retrieving
two
components
manifest
and
one
or
more
layers
a
manifest
is
a
json
document
which
defines
aussie
images
such
as
layers
sets
and
digest.
A
Layer
is
the
binary
form
of
content
that
is
stored
in
the
registry
addressable
by
by
the
tesla.
Then
the
process
of
putting
an
image
is
as
follows.
The
first
step
is
to
retrieve
the
manifest
by
sending
a
get
request
and
the
client
should
verify
the
return.
The
manifesto
signature
before
fetching
layers
and
then
pull
layers
which
are
not
exist.
A
A
A
A
At
present,
junk
flag
has
a
good
practice
in
combination
with
docker
and
contain
d.
Here
is
the
key
part
of
their
configuration
of
image
of
the
image
registry
process.
In
fact,
as
a
process
component,
deft
email
can
almost
support
all
containers
engines
as
well
as
download
the
download
file
request
is
for
the
2df
demon.
The
following
downloaded
logic
will
be
handed
over
to
dragonfly
to
complete.
A
Okay,
at
this
point,
you
have
learned
about
the
main
artifactual
architecture,
design
and
the
process
design
of
the
junk
blog.
In
addition,
jump
flare
has
the
abilities
to
provide
more
advanced
functionality,
such
as
network
bandwidth,
limit
transmission,
encryption
and
so
on.
If
you
are
interested
interested
welcome
to
join
us,
okay,
that's
all
that's
all
I
have
to
share
next,
let's
welcome
pontal
to
share
with
us
about
nether's
image
service.
Thank
you.
B
Thank
you,
vichy
hello,
everyone.
I
am
pantha,
I'm
a
software
engineer
from
hunt
group.
I
mostly
do
work
on
container
one
time
in
recent
years
today,
I'm
going
to
introduce
the
work
that
we
have
been
doing
in
the
past
year,
trying
to
improve
dragonfly
with
better
ways
of
container
image
distribution.
B
B
B
B
B
B
The
third
one
is
that
deleted
files
and
directions
are
still
saved
at
newer
layers
and
need
to
be
downloaded
when
starting
a
new
container,
even
though
they
won't
be
used
by
containers
and
all
the
oci
community
are
very
aware
of
their
resolvex
and
started
to
talk
about
ocr
v2
in
each
spec.
For
some
time
here
is
a
link
to
a
shared
document,
a
brainstorm
that
then
is
happened
in
the
oci
community.
B
B
Stanker
tries
to
solve
the
deduplication
property,
with
storage,
with
storage
and
ar
themselves
capability,
crfs
hacks
the
targese
layout
format,
and
make
it
seekable
to
support
on
demand
loading,
fire
grain
and
oci
break
image
layer
in
belt.
Saving
each
file
or
subfile
ends
up
as
an
object
in
the
image
registry.
B
All
these
are
generals
works
and
solve
their
problems,
and
they
deem
both
important.
However,
each
of
them
suffer
from
different
problems,
for
example
the
p2p
image
distribution
solution,
while
fast
still
requires
downloading
an
entire
image
before
starting
a
container
the
file.
Only
five
layer,
local
cache
in
server
methods
reduce
the
problem
to
only
require
a
file
to
be
fully
downloaded
before
processing,
but
it
is
still
not
a
fourth
ultimate
building.
B
Snacker
united
are
storage.
Stem
short
capability
that
is
not
always
available.
Crfs
requires
a
few
smartpoints
for
every
image
layer
and
in
performance
overhead,
with
each
layer
which
is
slow.
If
there
are
many
layers
in
the
image,
valkyrie
and
oci
both
suffer
from
the
problem
of
too
many
objects
for
each
image,
which
is
a
great
button
on
the
image
registry.
B
B
B
B
The
demand
unfolds,
the
metadata
layer
and
exports
file
system
views
to
containers
for
fuse
over
here
fees
variables,
so
nana's
can
support
both
traditional
1c
container
and
virtualized
containers
like
neck
cutter
appearance
because
net
as
image
format
is
compatible
with
oci,
artifact,
spec
and
distribution.
Stick
it
can
be
stored
in
the
docker
registry,
as
well
as
an
oss
in
cloud
environment
users
can
choose
to
deploy
an
ap
distribution
network
like
dragonfly
between
registry
and
the
computer
mode
to
further
accelerate
image
data
distribution
in
large
scale
inside
another
schema.
B
B
B
B
Each
file's
data
is
cut
into
fixed
sensor
change.
Each
tank
has
its
own
hashtag
hashtags,
calculated
by
its
data
saved
in
a
shared
data
layer,
because
the
data
layer
is
separated
from
metadata.
Here,
let
us
can
represent
a
container
image
file
system
view
just
with
the
metadata
layer
and
the
data
layer
can
be
shared
among
different
metadata
layers
and
the
enter
data
check
level
so
that
they
are
shared
between
different
container
images.
B
The
immediate
benefit
is
that
we
can
launch
containers
much
faster
in
a
container
startup
startup
time
benchmark.
We
measure
that
the
conventional
container
startup
tab
can
be
proportional
to
the
size
of
container
images.
For
example,
a
bd
box
can
be
launched
very
quickly,
since
this
image
is
pretty
small.
B
B
B
B
Then
it
is
uncompressed
to
a
local
file
system.
After
that
we
cannot
know
if
it
is
changed,
changed
intentionally
or
unintentionally,
or
if
there
is
a
bit
flip
over
silent
data
corruption
in
the
double
disk.
This
spatulas,
not
some
downloading
applications
for
long
time,
because
disk,
linears
and
silent
data
encroaching
cannot
be
anticipated.