►
From YouTube: CNCF SIG Storage 2020-05-13
Description
CNCF SIG Storage 2020-05-13
B
A
D
A
E
F
So
so
higher
and
Amy
hi,
everyone
actually
hope
everyone's
doing
well.
So
the
main
agenda
item
for
today
is
the
TI
kv.
F
D
D
B
Yeah,
the
screens
set
up
for
tech-heavy
graduation
review.
So
thank
you.
Everyone
for
having
us
here
and
today
I
will
present
the
graduation
review
for
the
Thai
Keely
project,
and
here
with
me,
we
have
the
main
tenders
of
the
Taibbi
project
and
the
the
the
mentor
dress
and
I
will
present
this
review
for
you.
So
first
thing:
first,
if
you
didn't
know,
Tech
hippie
before
Kentucky
is
distributed
from
the
action
on
key
value.
B
Database
originally
created
by
thing
kept
as
the
under
law
storage
engine
fatality,
be
with
the
mission
to
enable
and
empower
the
next
generation
of
databases
by
providing
a
reliable,
high-quality,
practical
storage
foundation,
and
in
case
you
didn't
know
there
Taiki
the
architecture.
This
is
a
brief
overview
and
we
have
a
lot
of
technical
contents
on
the
tech-heavy
Doric
website.
For
you
to
take
a
look,
we
have
deep
dives
and
we
have
a
lot
of
case
studies
and
we
also
have
some
of
those.
You
know
technical
details
covered
there.
B
So
a
brief
introduction
of
the
history
of
TV,
as
I
mentioned
earlier,
is
created
as
the
storage
layer
for
Taibbi
in
April
2015,
and
it
was
then
open
sourced
on
April,
the
1st
2016.
Yes,
you
hear
me
right
is
April
the
first
ending
October
2017,
the
Taiki
be
1.0
was
released
and
luckily
for
the
tech-heavy
project,
it
entered
sin
CF
at
the
sandbox
project
in
August
2018
and
a
may
2019,
which
is
around
the
same
time.
B
B
So
when
we
talk
about
its
community,
the
most
important
part
for
forests
and
for
the
tech-heavy
project
is
the
adoptions.
So
thank
you.
He
has
been
I
have,
but
the
adoption
of
Taibbi
has
been
growing
very
steadily
as
the
project
becomes
mature.
A
sister
joinings
since
EF
tech-heavy
has
double
the
production
adoption
now
better
tested
by
over
1,000
adopters,
all
across
the
regions
across
different
rate,
different
industries
compared
to
the
status
before
incubating
the
following
major
progresses
have
been
made.
B
B
You
can
see
the
ng
and
booked
my
show
show
key,
etc
and
also
the
u.s.
we
have
adopters
there,
but
we
upon
request.
We
do
not
share
their
name
yet,
and
industry
wise.
We
have
adopters
across
a
variety
of
industries,
including
production
deployments
in
some
mission-critical
and
intensive
industries
such
as
finance
and
payments
and
in
Beijing
bank
and
vng,
etc.
We
also
have
online
streaming
and
video
platform
adopters,
such
as
billy
billy,
heh,
Hulu
and
dailymotion,
and
so
the
logos
here
are
only
selected
community
adopters
from
outside
China.
Just
for
your
reference.
B
So
we
also
have
some
projects
beyond
on
top
of
Thai
TV,
first
of
all
tidy
B,
and
then
we
have
Reddy's
anti
TV
projects
such
as
Teddy's
tighten
tight
et
cetera
and
for
premises
metrics
in
Thai
TV.
We
have
type
parameters
and
type
TV
also
enable
enables
our
customers
and
our
adopters
by
providing
different
deployment
methods
such
as
dagger
and
the
kubernetes.
F
Just
a
just
a
quick
question
there,
the
the
TI
kv
deployments
without
without
Taibbi.
Do
you
happen
to
know
what
sort
of
use
cases
they're
there
it's
being
used
for?
Yes,.
B
F
B
Yeah,
so
for
the
technical
progresses
we
have
made
since
Taiki,
v,
3.0
and
incubation
I
want
to
highlight
the
three
aspects:
number
one
is
performance
and
then
functionality
and
security
so
for
performance.
The
Thai
TV
community
has
Titan,
which
is
a
rocks,
TV
plug-in
for
key
value
separation,
and
it
seems
to
reduce
the
write
amplification
in
rocks
TV
and
improves
the
range
query,
performances
and
the
next
one
is
unified
thread
pool.
This
is
a
feature
from
hackathon
events.
We
helped
and
it's
the
it's
the
top
winner
of
the
hexam.
B
So
it's
for
storage,
read/write,
requests
and
coprocessor
requests
and
reduces
the
impact
of
having
career
online
Twitter
queries
and
for
follow
rate.
It's
about
strong,
consistent
reach
from
a
ruffed
followers.
It
balances
the
hospital
pressure
and
improves
overall
throughput
functionality-wise.
We
have
elastic
scheduling
it.
Also
split
replicas
and
auto
skills,
note
based
on
real-time
workload
and
may
King,
tech-heavy,
really
Clapton
native.
So
it's
this
is.
This
feature
is
contributed
from
the
maintainer
and
his
name
is
dubbing
Lee
and
he
will
present
their
cases
their
case
study
later.
B
So
the
next
feature
functionality
is
about
fast
backup
and
restore,
because
this
is
very
important
to
the
adopters
it
allows
distributed
physical
bekommen,
restore
from
taking
me
for
large
clusters
with
tunable
speed.
You
can
see
the
RFC
as
in
the
link
so
for
security.
Now
ty
keavy
supports
transparent
data
encryption,
so
this
is
very
important
for
for
the
community
users
as
well,
so
the
next
slice,
the
next
slide
I,
will
cover
the
clear
native
alignments.
B
So
techie
B
has
been
one
of
the
few
key
value
storage
engines
in
the
cloud
domain
that
can
balance
both
performance
and
the
years
of
operation
with
kubernetes.
Data
storage
is
one
of
the
important
components
of
any
cloud
native
infrastructure
platform
and
either
end
users
need
a
range
of
choices
to
meet
their
needs.
So
this
these
features
of
ty
Kiwi
really
fits
in
the
cloud
native
landscape
and
because
we
also
enable
cloud
native
deployments
such
as
docker
compose
and
covenant
is
using
high
DB
operator.
B
Now
we,
where
we're
still
working
on
the
TV
first
they're
working
on
sorry,
we're
still
working
on
the
tech-heavy
operator
and
it
will
be
available
soon
called
the
integration
and
partnership
with
the
cloud
native
ecosystem.
We
work
very
closely
with
etcd,
with
we
work
very
closely
with
etcd
and
we
among
the
bug
fixes
we
have
provided
for
etcd.
We
also
contributed
the
learner
feature
since
the
EDD
and
we
also
maintained
the
rest
implementation
of
ERP,
C
and
primitives,
and
we
also
worked
very
closely
with
a
graph
enough
so
for
our
community
adopters.
B
If
they
do
not
go
with
Tai
TV,
they
also
have
other
opportunities
and
alternatives
they
can
take
number
one
is
they
can
use
a
single
node
database
and
they
can
scale
vertically
or
shard
manually.
There
are
some
disadvantages,
but
that's
also
the
way
they
came
they
can
choose
and
they
can
also
use
alternative
key
value
stores
such
as
foundation,
DB,
Cassandra,
H,
pins,
HBase,
MongoDB,
etc.
B
You
can
see
a
comparison
in
the
next
slide,
so
if
they
do
not
school
with
the
these
two
options,
they
can
also
use
another
juicy
core
database,
such
as
cockroach
TV
and
you
buy,
etc.
So
this
is
the
comparison
I
mentioned
earlier
and
you
can
see
the
different
features.
All
those
all
those
solutions
provide
I
will
not
cover
the
details
but
feel
free
to.
Let
us
know
which
one
you
know.
If
you
have
any
questions.
F
B
Let
me
go
back
to
this
slide,
so
this
slide
is
about
the
tech-heavy
architecture.
So
in
TV
we
use
regions
so
ever
had
the
name,
but
this
radio
is
actually
a
split
or
in
chart,
as
you
just
mentioned
array
of
data,
so
we
separated
data
into
small
chunks,
most
blades
or
shard
in
this
term
region,
the
different
regions,
and
then
we
mminton
three
replicas
of
each
region.
B
C
Penny
each
city
it
may
be
used
raft
and
consolidate
apart.
Each
city
is
only
designed
for
saving
the
metadata,
not
so
many
whose
data
in
some
our
user
cases,
for
example
in
Turku
or
in
Kingdom,
and
the
user
we
are
save,
may
be
more
than
pega
data.
Pegye
man
data
in
the
one
Tigra
caster
and
now
you
can
see-
did
Kendall
Oh.
So
many
data
so
I
think
that
maybe
that
Katy
friend
from
the
video
ad
and
the
tactic
is
for
the
data
side
desire,
I.
G
G
B
F
That's
that's
correct
as
my
understanding,
so
so
you
know
it's
easy
I
think
it's
designed
for
having
a
strongly
and
extremely
strongly
consistent
set
of
keys,
which
are
which
are
replicated
so
I
give
a
number
of
times.
But
but
you
know
a
Sukkah
said
there
is
no
sharding
to
kind
of
distribute
the
scale.
B
Yeah
any
other
questions
so
far.
B
H
Sure,
hello,
everyone
I'm
come
true
who
are
Jewish
a
Chinese
Nadia
sharing
the
night
company
I'm
going
to
share
with
you.
The
adoptions
of
tacky,
riad
cuckoo
Thank
You
Mia
are
many
used
to
store
our
user
reading
history
in
the
personalized
history.
Oh
the
reason
we
store
the
items
are
we
need
to
use
to
use
this
to
filter
out
feeds
or
notifications
that
people
already
had
earlier,
so
it
won't
stay,
they
won't
see
them
duplicating
feeds.
H
H
Ok,
move
on
ok,
and
here
are
some
highlights
that
we
think
are
very
important
to
us.
First,
the
data
volume
of
reading
history
system
are
kept
increasing
rapidly
a
year
ago
will
be
moved
our
data
to
take
away
the
same
system
had
one
point:
three
trillion
rows
of
data
and
this
almost
doubled
within
a
year
to
me
that
I
were
increasing
workload,
we
simply
add
more
servers
to
tackle
a
cluster
and
they
just
works.
Fine.
The
high
availability
of
Tigra
cluster
is
also
very
crucial
for
our
system.
H
We
have
experienced
the
several
server
hardware
issues
in
the
past
and
I'll
say
the
outage
had
almost
no
impact
to
our
production
system.
In
addition,
the
capability
of
deploying
a
single
tackle
cluster
smoke
cloud
providers
gives
us
a
stronger
guarantee
to
away
invaluable
data.
Those
are
very
important
to
us.
Thank
You
V
has
proven
to
be
a
great
foundation
for
our
applications
in
the
past,
so
we
definitely
want
to
expand
its
application
scenarios.
At
this
time
we
are
building
a
data
table
store
on
top
optically
to
better
support
huge
single
table
scenarios.
H
B
A
A
Shotty
is
also
a
very
cool
feature
for
us.
We
can
increase
Aronian
without
the
great
people.
Redeployment
last
one
is
about
the
no
SPF,
no
single
point
of
failure.
We
she
can't
keep
as
our
service
our
service,
always
on
that
and
we
have.
We
have
used
the
peculiar
production
for
about
two
years
we
have
more
than
ten
casters,
including
the
testing
and
the
pre-production
environment,
and
the
more
than
13
years
of
records
new
one
single
cluster.
A
We
are
working
with
a
tiger
with
Europe
who
and
the
chain-3
know,
there's
no
more
problems
in
our
production
environment,
including
on
how
to
support
the
magical
capacity
for
weight,
either
brown
so
500
in
years
to
120
our
records
in
the
near
future
and
how
to
maintain
their
take
away.
There
is
our
cooperate
systems
and,
of
course,
on
how
to
reduce
a
latency
and
how
to
reduce
the
rejection
hand
it
can
help.
B
You
Debbie,
so
this
is
the
case
where
you
are
asking
about
what
thank
you:
what
users
are
using
tech
heavy
separately
from
Taibbi.
So
this
is
the
case.
Study
is
also
published
on,
since
you
have
yeah
the
next
one
is
okay
and
we
also
have
other
techie
adopters
and
if
you
are
interested,
I
can
share
that
with
the
names
and
with
you
later
thank
you,
Alex,
and
so
the
next
case
study
comes
from
IDI
insertion
and
Indians
issuing
is
a
news.
I
Tempo
from
eating
sushi,
which
is
interest,
earned
well
news
data
campaign
channel
tacky.
We
had
been
widely
used
in
production
with
more
than
ten
casters
when
100
machines
and
100
years
of
data.
It
cares
Courtley,
storage
and
online
services
such
as
a
supporter
feature,
and
the
typical
model
would
be
order
over
own
kV
proxy,
on
top
over
Tegrity
for
camera,
binti,
which
be
nice
to
have
interfaces
they
used
the
modified
had
to
be
an
support
which
play
on
their
emotions
and
use
modified
type.
You
operator
to
deploy
Telecasters
on
kubernetes
a
recently.
I
B
You
mr.
Jane,
so
that
concludes
the
presentations
from
the
case
studies
of
and
from
our
maintenance.
So
the
last
slide
is
to
show
that
to
show
the
tip
see
that
and
to
show
the
stick,
that
we
have
completed
the
graduation
criteria.
We
have
all
the
books
check,
detect
and
we
humbly
presents
the
due
diligence
for
your
review.
Please
let
us
know
if
you
have
any
questions
or
any
comments.
We
would
like
to
help.
You
answer
that
and
the
link
here
points
to
the
thai
TV
due
diligence
for
since
F
graduation.
Thank
you.
F
B
B
B
C
You're
right
Kenny,
you
just
have
clogging
over
rocks
to
be
not
have
four
core
rocks
P
and,
as
you
can
see,
sometimes
we
can
use
rocks.
We
support
plug-in
mechanism,
so
sometimes
we
can
write
our
own
party
in
for
XP.
So
on
a
letter,
for
example
inside
hand,
we
use
the
panning
to
maybe
to
hope
the
compaction
and
to
hook
the
flash.
So
when
the
rocks
we
do
the
flash
or
to
the
compassion
the
work
or
a
pattern.
So
if
ya,
just
Aparri
and.
F
C
F
C
Maybe
inspired
by
a
paper
code
with
a
key,
it's
just
a
key
value
of
supervision.
You
can
say
you
weigh
in
rocks,
be
in
the
orange
area,
where
you
save
the
key
and
plus
the
value
in
the
SAT
file
directly.
So
the
value
is
the
value
is
large
and
the
write
amplification
is
very
bad
for
the
promise.
So
we
developed
a
tank
for
the
key
various
operation
and
move
the
value.
The
big
way
at
the
big
big
data,
our
value
came
to
a
block.
We
call
a
block
fire,
another
fire,
nothing.
You
know
a
certified.
F
C
F
E
You
no
I've
found
their
document
I'm,
not
sure
who
the
TOC
representative
would
be
sad.
I,
don't
know
if
here
in
charge
of
doing
the
due
diligence,
I
created
a
doc
or
gone
through
there,
doc,
just
like
the
one
we
did
for
Brooke,
so
I
don't
know
sod.
If
you're
on
the
call
you
are
the
person
that
will
be
following
up
with
that
or
someone
else.
I.
C
F
F
F
E
E
F
So
it's
a
Oh
hide
Erica.
So
it's
since
you
know
we
we
we
had
to
the
privy
representation
and
enough
was
that
was
great,
I
guess.
The
last
time
you
presented,
you
said
that
you
were
concentrating
different
foundations
that
you
may
want
to
apply
to
I
guess
the
ball
is
is
currently
in
your
court.
In
terms
of
you
know,
did
you
decide
that
you
wanted
to
to
proceed
with
the
CN
CF
yeah.
J
J
So
as
long
as
everything
checks
out
we'll
proceed
with
that
I
guess
we
had
a
question
about
order
of
operations
with
respect
to
like
there's
some
formalities
around
like
github
issues
and
talk,
pr's
and
and
the
cigs
recommendation.
So
I
has
to
get
our
legal
approval
with
my
first
next
step.
Be
opening
that
that
issue
with
the
cig.
F
So
I
guess
the
first.
The
first
decision
point
is
at
what
level
you
were
thinking
of,
of
applying,
whether
it's
it's
at
a
sandbox
level
or
at
a
or
at
an
incubation
level.
There
are.
There
are
a
so
there's,
there's
a
fairly
simple
workflow,
which,
with
a
couple
of
templates
that
can
be
followed,
one
for
sandbox
and
one
for
incubation.
But
just
a
word
of
warning.
The
sandbox
process
is
just
in
the
process
of
being
reviewed
this
this
so
the
I
guess
there
are
different
criteria.
J
J
F
J
I
guess
you
guys
have
a
hand
in
making
that
recommendation
as
well,
so
we
maybe
participate
in
a
conversation
there,
but
we're
definitely
eager
to
get
maybe
access
to
infrastructure
sooner
than
later.
So,
if
sandbox
helps
with
that,
maybe
there's
a
benefit
in
there.
You
know
I,
don't
know
if
we
have
to
like
I,
don't
know.
If
incubation
that's
when
we
start
the
due
diligence
document.
Is
that
true
and
then
sandbox
can
kind
of
delay
due
diligence.
F
J
J
F
Let's
that's
fantastic
news:
you
know,
I
thought
prefigures
was
a
fairly
exciting
project
and
we
don't
currently
have
a
streaming
storage
provider
in
as
a
CN
CF
project.
Yet
so
this
this
fills
a
nice
niche.