Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
12
,
No.
3
,
Decem
ber
201
8
, p
p.
1
0
94
~
1
1
05
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v1
2
.i
3
.pp
1
0
94
-
1
1
05
1094
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Integr
ation o
f Li
nu
x Containers i
n OpenS
tack:
An Intro
spection
Ash
ish
Li
nga
yat
1
, R
an
j
ana
R.
B
ad
re
2
,
An
il
Kuma
r
Gu
p
ta
3
1,2
Com
pute
r
Eng
ine
er
ing, MIT
A
ca
dem
y
of
Enginee
ring
,
Dehu
Phata
,
Al
andi
(D)
,
Pune,
412105,
Mah
aras
htra
,
Ind
ia
3
HP
C
Infra
structure
and
E
cos
y
st
e
m
s,
C
-
DA
C,
C
-
DA
C
Innova
ti
on
Park,
Pune,
411
008,
Maha
rashtr
a
,
In
dia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ma
r
28
, 201
8
Re
vised Ju
n
1
5
, 2018
Accepte
d
Se
p
25, 201
8
In
cl
oud
computing,
sharing
of
r
esourc
es
is
supported
using
h
ea
v
y
we
ight
e
d
tra
ditiona
l
virt
ua
li
z
at
ion
techniqu
es.
Such
te
chni
q
ues
invol
ve
h
y
p
erv
isors
to
emulat
e
h
ard
wa
re
for
creat
ing
virt
ual
m
a
chines.
The
inc
lu
sion
of
an
addi
ti
on
al
l
a
y
er
of
h
y
per
v
isor
ove
r
host
oper
a
ti
ng
s
y
st
em
depr
ec
i
at
es
t
h
e
per
form
anc
e
of
the
vir
tua
l
m
ac
hin
e.
R
ec
en
t
evol
ut
ion
is
a
li
ghtwe
ight
al
t
ern
ative
to
t
he
virtual
m
achine
ca
l
le
d
con
ta
in
ers
which
h
ave
g
ai
ne
d
popula
rity
among
deve
lop
er
s
and
administra
tors.
Cont
a
ine
r
Based
virt
ualiz
at
ion
h
a
s
prove
n
ver
y
e
ffic
i
ent
reg
ard
in
g
per
form
anc
e
,
and
m
an
y
industri
es
are
n
ow
m
igra
ti
ng
th
ei
r
v
irt
ua
li
z
ed
e
nvironment
to
r
un
on
L
inux
cont
a
ine
rs.
Con
t
ai
ner
s
use
host
oper
ating
s
y
st
e
m
s
ker
nel
and
i
solat
e
each
cont
a
ine
r
b
y
en
ca
psula
ti
ng
th
e
m
with
the
ir
r
e
q
uire
d
servi
ce
s
a
nd
pac
kag
es.
Li
nux
ker
n
el
is
ver
y
ben
efi
c
ia
l
in
implementi
n
g
cont
a
ine
rs,
w
hic
h
is
t
h
e
rea
son
for
the
e
xiste
nc
e
of
Li
n
ux
cont
ai
n
ers.
Li
nux
containe
r
s
uti
li
ze
l
ess
storage
spac
e
a
nd
consum
e
opt
imal
computat
io
nal
power,
givi
n
g
a
hike
i
n
per
fo
rm
anc
e
.
Ha
ving
the
m
int
egr
at
ed
in
to
the
cl
o
ud
surel
y
b
ene
fi
t
s
consum
er
and
c
loud
pro
vide
r.
Man
y
p
roje
c
ts
have
e
xte
nded
their
support
in
inc
orpora
ti
ng
co
nta
in
ers
in
the
c
loud.
In
thi
s
pap
er,
we
wil
l
disc
uss
var
ious
Li
nux
con
tainer
s
and
their
m
an
age
m
ent
tool
s
a
long
with
cl
oud
computing
software
,
Open
Stac
k,
in
cl
udin
g
proje
ct
s
und
ert
ak
en
b
y
Op
enSta
ck
for
int
egr
at
ing
conta
ine
rs i
n
th
e
c
lou
d.
Ke
yw
or
ds:
Cl
oud
c
om
pu
ti
ng
Con
ta
ine
r or
c
he
strat
ion
Linux c
onta
ine
rs
Op
e
n
sta
c
k
Copyright
©
201
8
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed.
Corres
pond
in
g
Aut
h
or
:
Ash
is
h
Li
ngay
at
,
Com
pu
te
r
E
ng
i
neer
i
ng, MIT
Acad
em
y of
E
ng
i
neer
i
ng,
Dehu Phata
,
A
l
and
i
(
D)
,
P
un
e
,
412105,
Maha
r
ashtra,
In
dia.
Em
a
il
:
ashishvijayl
ing
ay
at
@
gm
ail.co
m
1.
INTROD
U
CTION
Virtuali
zat
ion
te
chn
iq
ue
is
use
d
t
o
virtu
al
i
ze
native
hard
war
e
f
or
r
unni
ng
gu
est
opera
ti
ng
syst
em
known
a
s
virt
ua
l
m
achine.
Vi
rtuali
zat
ion
ha
s
bee
n
play
in
g
a
ver
y
si
gn
ifi
cant
r
ole
in
cl
oud
c
om
pu
ti
ng
wh
ic
h
resu
lt
s
in
its
extensi
ve
usa
ge
.
Ma
xim
u
m
ut
il
iz
at
ion
of
res
ources
is
carri
ed
out
by
sh
ar
ing
them
with
virtu
al
m
achine.
The
t
echn
i
qu
e
of
virtu
al
iz
at
ion
in
du
ce
s
an
over
he
ad
in
the
pe
rfor
m
ance
of
gu
est
op
erati
ng
s
yst
e
m
by
con
s
um
ing
storag
e
,
m
e
mo
ry
an
d
wasti
ng
CP
U
res
ources
[1
]
.
I
n
vi
rtuali
zat
ion
,
hy
perviso
r
is
us
ed
to
e
m
ulate
hard
w
are
for
operati
ng
syst
em
to
r
un
on
it
.
Vi
rtu
al
iz
ed
en
vir
on
m
ent
has
sepa
rate
ke
r
nels
f
or
ho
st
op
e
rati
ng
syst
e
m
and
guest
op
e
rati
ng
syst
e
m
.
The
hype
r
visor
is
a
n
a
bst
ract
la
ye
r
bet
ween
g
uest
op
erati
ng
syst
e
m
and
na
ti
ve
ha
rdwa
re.
Ha
ving
hype
r
visor
betwee
n
guest
op
e
rati
ng
syst
em
and
host
hard
ware
ad
ds
ov
e
r
head
to
t
he
perform
ance
of
virtu
al
m
ac
hin
e
a
nd
ultim
at
el
y
slow
s
the
work
i
ng
of
vi
rtual
m
achine
du
e
t
o
e
m
ulate
d
hard
war
e
.
Fig
ure
1
sh
ows
t
he
a
r
chite
ct
ur
e
of
hy
perviso
r
an
d
Linux
co
ntain
ers.
It
s
hows
us
the
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
In
te
gr
atio
n of
Linu
x
Co
nta
i
ne
rs in O
penSta
ck:
A
n In
tr
osp
ect
ion
(
As
hish
Lingay
at
)
1095
la
ye
rs
involve
d
in
im
ple
m
ent
ing
virt
ualiz
at
i
on.
T
his
li
m
it
a
ti
on
of
virt
ual
m
achine
ha
s
le
d
to
the
de
velo
pm
ent
of Lin
ux
c
onta
iners
.
Figure
1
.
Hy
pe
rv
is
or v
s
.
Li
nux
c
on
ta
ine
rs
Linux
c
on
ta
in
ers
are
li
ghtl
y
weig
hted
as
t
hey
co
nsum
e
l
ess
am
ou
nt
of
sp
ace
a
nd
ha
ve
sho
rter
dep
l
oym
ent
t
i
m
e
[2
]
.
They
are
easi
ly
manag
ea
ble
an
d
help
in
m
ax
i
m
izing
util
iz
at
ion
of
com
pu
ti
ng
resou
rces.
T
ho
ugh
Li
nux
c
on
t
ai
ner
s
us
e the
ho
st
o
pe
rati
ng
syst
e
m
k
ern
el
, t
her
e is isolat
io
n
bet
ween
t
he
m
b
y
us
in
g
va
rio
us
encapsulat
in
g
te
chn
iq
ues
.
In
Linux
co
ntainers
,
inter
op
e
r
abili
ty
and
portabil
it
y
hav
e
m
ade
them
pr
evalen
t
in
to
day’s
c
om
petit
ive
m
a
rk
et
.
Re
ce
ntly
,
Mi
cro
s
of
t
ha
s
al
so
sta
rte
d
bu
il
di
ng
W
i
ndows
con
ta
ine
rs
a
nd
hav
e
exte
nded
their
s
upport
in
integ
rati
ng
the
m
with
W
i
ndows
syst
em
.
In
the
cl
oud,
m
anag
em
ent
of
the
res
ources
is
ver
y
vital
in
eff
ic
ie
ntly
ru
nnin
g
serv
ic
es
or
app
li
cat
io
ns
[
3].
U
n
d
e
r
u
t
i
l
i
z
a
t
i
o
n
of
resour
ces
will
no
t
ben
e
fit
consum
er
nor
th
e
cl
oud
prov
i
der.
To
im
pr
ove
util
iz
at
ion
of
res
ources,
nee
d
of
m
anag
in
g
th
e
res
ources
by
us
in
g
s
che
duli
ng
te
c
hn
i
qu
e
s
is
ben
e
fici
al
[4
]
,
[5
]
.
Feat
ur
es
of
Lin
ux
c
onta
iners
ha
ve
le
d
t
o
the
inte
gr
at
io
n
of
Lin
ux
co
n
ta
iners
in
the
c
loud
that
sim
plifie
s
the
m
anag
e
m
ent
of
a
pp
li
cat
ion
s.
P
roje
ct
s
are
bein
g
unde
rtake
n
to
achie
ve
m
axim
u
m
integrati
on
of
con
ta
ine
rs
in
t
he
cl
oud.
T
he
sche
du
li
ng
te
c
hn
i
qu
e
s
for
c
onta
iners
in
the
cl
oud
will
help
i
n
pro
per
util
iz
at
i
on
of
res
ources
th
us
be
nef
it
in
g
t
he
cl
ou
d
pro
vid
e
r
a
nd
c
on
s
um
er
[6
]
.
The
r
e
is
a
c
onside
r
able
nu
m
ber
of
project
s
wo
r
king
in
co
ntainers
a
nd
cl
oud,
but
we
will
rev
ie
w
a
few
popula
r
ones
am
on
gs
t
th
e
m
.
Re
d
H
a
t
’
s
Op
e
nShift,
Am
azon
W
e
b
Ser
vice
s
has
its
A
m
a
zon
Ela
sti
c
Co
ntainer
Se
rv
ic
e
(A
m
azon
E
CS),
Mi
cro
s
of
t
Azure
use
s
Fa
bri
c,
G
oogle
Cloud
Plat
form
(G
CP)
util
iz
es
Go
ogle
Con
ta
ine
r
En
gin
e
(
GCE)
,
and
Op
e
nS
ta
c
k
has
Ma
gnum
,
Nov
a
D
ock
e
r
Hype
rv
is
or, Koll
a,
Zu
n,
Kata f
or
m
anag
in
g
c
onta
iners
i
n
the
cl
oud.
Op
e
nS
ta
c
k
,
an
op
e
n
source
cl
oud
softwa
re
,
is
widely
ado
pte
d
by
cl
oud
pro
vid
e
rs
due
to
it
s
featur
e
s
a
nd
ha
s
pro
ve
n
a
n
excell
ent
platfor
m
for
inte
grat
ing
Li
nux
c
on
ta
ine
rs.
Syst
e
m
con
ta
ine
rs
get
m
anag
ed
easi
l
y
in
Op
e
nS
ta
c
k
since
it
prov
i
des
I
nfrastr
uct
ur
e
as
a
Se
rv
ic
e
(I
aaS
).
O
pe
n
so
urce
pro
j
ect
s
are
ind
e
pende
ntly
integrate
d
int
o
O
penSta
ck
to
pro
vid
e
cl
oud
in
fr
ast
ru
ct
ure.
Th
ese
in
divi
du
al
pro
j
ect
s
get
dep
l
oyed
a
utono
m
ou
sly
f
or
f
ulfill
ing
assi
gned
w
orkloa
d,
s
uch
as
ci
nd
e
r
i
s
i
m
ple
m
ented
to
pro
vid
e
st
orage,
Neu
t
ron
for
n
et
work
i
ng.
Si
m
il
arly
,
Op
en
Stac
k
has
i
ntegr
at
e
d
op
e
n
s
ource
pro
j
ect
s
f
or
buil
ding
an
d
util
iz
ing
m
axim
u
m
ben
efit
s
of
Lin
ux
Co
ntainers
.
O
penSta
ck
Ma
gnum
pr
oject
off
ers
an
Applic
at
io
n
Pr
og
ram
m
ing
I
nterf
ace
(API
)
for
interact
in
g
with
Co
ntainer
Or
c
hestrati
on
En
g
ines
(COE
s)
[
7].
Op
e
nS
t
ack
Zu
n
is
use
d
f
or
m
anag
in
g
c
on
ta
ine
r
im
ages
by
util
iz
ing
res
ources
pro
vid
e
d
by
O
pe
nS
ta
ck
.
Op
e
nSt
ack
Ko
ll
a
is
a
platfor
m
fo
r
de
plo
y
ing
pr
oductio
n
read
y
co
ntain
ers.
Re
centl
y,
Op
e
nS
ta
c
k
Kat
a
proj
ect
ha
s
be
en
anno
un
ce
d
t
o
c
om
bin
e b
e
nef
it
s of
virtu
al
m
achine a
nd
c
onta
iners.
2.
CONTAI
NER
Con
ta
ine
rs
are
a
li
gh
twei
ght
al
te
rn
at
ive
to
virtu
al
iz
at
ion.
They
are
por
ta
ble
and
c
ons
um
e
le
ss
m
e
m
or
y
wh
ic
h
can
acco
unt
for
M
B
’
s
as
c
om
par
ed
to
vir
tual
m
achines
wh
e
re
they
c
onsu
m
e
m
e
m
or
y
in
G
B
’
s
[8
]
.
Th
ey
are
us
ed
to
store
e
ntire
app
li
cat
ion
hav
i
ng
al
l
of
it
s
dep
end
e
nc
ie
s,
li
br
aries
and
config
ur
at
io
n
f
il
es
req
uire
d
f
or
r
unni
ng
the
m
.
Con
ta
iners
are
isolat
ed
from
each
oth
er
by
us
ing
var
i
ous
pack
a
ging tec
hniq
ues [9
]
.
Fea
tures o
f
c
on
ta
i
ner
s
ensu
re in
t
eropera
bil
it
y and p
ort
abili
ty
o
f
appli
cat
ion
s
[9].
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
1
1
94
–
1
1
05
1096
”
Con
ta
inerizat
ion
is
no
t
hing
bu
t
the
pr
ocess
of
abstra
ct
ing
al
l
the
diff
ere
nc
es
in
op
e
rati
ng
syst
e
m
distrib
utions
and
thei
r
unde
rly
ing
inf
rastr
uc
ture
by
enca
psula
ti
ng
disc
rete
com
po
nen
ts
of
ap
plica
ti
on
log
ic
,
inclu
di
ng
the
a
pp
li
cat
io
n
platf
or
m
and
its
depend
e
ncies,
w
it
h
the
h
el
p of l
igh
t
weig
ht cont
ai
ner
s
[
10
]
.”
Ty
pes of co
nta
inerizat
ion
:
1)
Op
e
rati
ng
syst
e
m
con
ta
ine
rizat
ion
:
O
per
at
i
ng
syst
em
con
ta
inerizat
ion
pr
ov
i
des
us
e
r
-
s
pa
ce
isolat
ion
a
nd
util
iz
es
the
kernel
of
t
he
host
operati
ng
syst
e
m
.
They
are
si
m
il
ar
to
virtu
al
m
achine
but
not
preci
sel
y
virtu
al
m
achin
e.
F
or
creati
ng
op
e
rati
ng
sy
stem
con
ta
ine
r
s,
we
ca
n
us
e
te
ch
niques
s
uc
h
as
O
pe
nVZ,
LXC, L
in
ux VSer
ver
,
S
olaris
and m
any
m
ore.
2)
Applic
at
ion
c
onta
inerizat
io
n:
Op
e
rati
ng
syst
e
m
con
ta
iners
run
m
ulti
ple
p
ro
ces
ses
an
d
s
erv
ic
es
w
he
rea
s
app
li
cat
io
n
co
ntainers
a
re
s
pecific
to
a
n
app
li
cat
io
n.
Rocket,
D
ocker
are
e
xam
pl
es
that
pro
vide
app
li
cat
io
n
co
nt
ai
ner
s.
Adva
ntages
of
con
ta
ine
rizat
io
n:
1)
C
on
ta
ine
rs
a
re
li
gh
twei
ght
as
t
hey
occ
upy
le
sser
sto
ra
ge
s
pa
ce
than
virt
ual
m
achine.
2)
In
c
rease
d
CP
U
u
sa
ge o
ver
vir
tual
m
achine.
3)
Con
ta
ine
rs
a
re
highly
porta
ble.
4)
Du
e
to
t
he red
uced siz
e
of
c
onta
iners
, m
any n
um
ber
of co
nt
ai
ner
s ca
n be
r
unning
on a
sin
gle
host.
5)
CR
UD
operati
on
s
ca
n be
perf
or
m
ed
sm
oo
thly
o
n
them
.
Dr
a
wbacks
of
con
ta
ine
rizat
io
n:
1)
Com
plete
isolat
ion
betwee
n
c
on
ta
ine
rs
is
not
possible.
2)
They are
vuln
e
rab
le
with as
pe
ct
to
sec
ur
it
y.
3)
Con
ta
ine
rs
a
re
no
t t
he whole
r
eplace
m
ent to v
irtual
m
achin
e as they
us
e
host
op
e
rati
ng s
yst
e
m
k
ern
el
.
4)
Feat
ur
es
of c
onta
iners
are:
5)
Re
plica
ti
on
: I
de
ntica
l im
ages o
f
contai
ners c
an be inc
orp
orat
ed
co
ntaini
ng a co
m
plete
appli
cat
ion
.
6)
Test
ing
a
nd
is
olati
on
:
C
on
ta
i
ner
im
ages
are
isolat
ed
a
nd
pa
ckag
e
d
to
ca
r
ry
their
de
penden
ci
es
,
li
br
a
ri
es,
bin
a
ries
w
hich
are
re
quire
d
to
r
un
their
serv
ic
e
or
proces
s.
Is
olati
on
e
nsures
prop
e
r
r
unni
ng
an
d
abstracti
on
of
con
ta
ine
rs
in
a
ny
en
vir
onm
ent
m
aking
it
s
uitable
f
or
te
sti
ng
accu
ratel
y.
7)
Scal
abili
ty
:
Cr
eat
ion
a
nd
te
r
m
inati
on
of
c
onta
iners
are
ca
rr
i
ed
out
by
buil
ding
m
ulti
pl
e
co
ntainers
.
Th
e
or
c
hestrati
on t
oo
ls
u
se
d f
or
s
cal
ing
c
on
ta
i
ne
rs
a
re
Do
c
ker
Sw
arm
, Apac
he
Mesos,
Ku
be
rn
et
es.
8)
Perfo
rm
ance:
Con
ta
ine
rs
are
li
gh
twei
ght
a
nd
do
not
ha
ve
any
over
hea
d
la
ye
r
w
he
n
dep
l
oyed
on
host
op
e
rati
ng syst
e
m
. Th
is
con
t
rib
utes to
m
axi
m
um
p
erfor
m
ance as c
om
par
ed t
o
vi
rtual m
achine.
9)
High
Av
ai
la
bil
it
y:
High
Av
ai
la
bili
ty
of
m
any
con
ta
ine
rs
is
possible
in
cl
us
te
rs
beca
us
e
they
re
qu
i
re
l
ess
stora
ge
a
nd m
e
m
or
y and
has f
ast
er d
e
plo
ym
ent
rate.
2
.
1.
Li
nu
x C
on
t
ainer
(
L
XC
)
Linux
Co
ntain
ers
are
de
velo
ped
us
i
ng
C,
Pyt
hon,
S
hell,
and
Lua
la
nguag
e
t
hat
ai
m
for
offe
rin
g
env
i
ronm
ent
s
i
m
i
la
r
to
a
com
ple
te
virtu
al
m
achine.
It
is
fr
ee
softwa
r
e
li
censed
unde
r
GNU
LG
P
Lv
2.1+
li
cense
[
11
]
.
It
i
m
ple
m
ents
syst
e
m
-
le
vel
virtu
al
iz
at
ion
us
i
ng
cg
r
oups
[12],
na
m
espaces
an
d
acce
s
s
c
on
t
ro
l
thu
s
isolat
in
g
con
ta
ine
rs
f
rom
each
oth
er
[13].
Per
form
a
nce
of
c
on
ta
in
ers
is
increase
d
by
reducin
g
the
ov
e
r
head
of
usi
ng
m
ulti
ple
kernels
an
d
e
m
ula
te
d
hard
w
are.
P
rojects
de
velo
ped
by
li
nuxc
on
ta
ine
rs.org
ar
e
LXC,
LX
D, an
d
L
XCFS
.
a)
LXD
:
Fou
nd
e
d
by
Ca
nonica
l
Ltd.
has
c
ontrib
uto
r
s
f
ro
m
al
l
ov
e
r
the
gl
ob
e
.
L
XD
act
s
as
a
m
anag
er
for
syst
e
m
co
ntain
ers.
LX
D
is
de
plo
ye
d o
n
t
op
of LX
C
for i
m
pro
ving the
u
s
er e
xp
e
rience
[14
]
.
b)
LXCFS:
T
he
cod
e
of
L
XCF
S
is
wr
it
te
n
i
n
C
la
nguag
e
.
D
evel
op
e
d
t
o
ov
e
rc
om
e
the
disad
va
ntages
of
Linux
kernel
w
hile ha
nd
li
ng f
i
le
syst
e
m
f
or
c
on
ta
ine
rs [
15
]
.
c)
LXC:
It
pro
vide
s
an
inter
face
to
the
us
e
r
f
or
con
ta
ini
ng
fea
tures
of
Li
nux
kernel.
U
ses
A
PI
an
d
to
ols
f
or
m
anag
in
g
a
ppli
cat
ion
or syste
m
co
ntainers [
11
]
.
Fo
r
c
on
ta
ini
ng
process
, L
XC
us
es
fo
ll
ow
i
ng
featur
e
s
of
Lin
ux k
e
r
nel [16
]
:
a)
Kernel
nam
espaces
b)
SELin
ux pr
of
il
es
c)
Seccom
p
poli
ci
es
d)
Chroots
e)
Kernel ca
pab
il
it
ie
s
f)
Con
tr
ol
gro
up
s
(
c
groups
)
2
.
2
.
Do
c
ker
Do
c
ke
r
is
de
ve
lop
e
d
in
Go
program
m
ing
la
ngua
ge
[17]
a
nd
us
es
featu
res
of
Lin
ux
ke
r
nel.
Do
c
ker
is
a
co
ntainer
ba
sed
platfo
rm
desig
ne
d
to
e
ase
creati
on
a
nd
r
unning
of
an
a
ppli
cat
ion
.
It
is
ope
n
so
urc
e
platfo
rm
fo
r
buil
ding,
s
hippi
ng,
an
d
run
ning
distri
bute
d
s
yst
e
m
.
Do
cke
r
syst
e
m
con
sis
ts
of
D
ock
e
r
E
ng
i
ne,
portable
a
nd
li
gh
twei
gh
t;
a
runtim
e
of
pa
ckag
i
ng
to
ol;
and
D
oc
ker
H
ub.
Applic
at
ion
s
ca
n
be
qu
i
ckly
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
In
te
gr
atio
n of
Linu
x
Co
nta
i
ne
rs in O
penSta
ck:
A
n In
tr
osp
ect
ion
(
As
hish
Lingay
at
)
1097
com
bin
ed
a
nd
ease
t
he
ta
s
ks
of
dev
el
opm
ent,
an
d
qu
al
it
y
assur
anc
e
in
a
produ
ct
ion
e
nviro
nm
ent.
Fo
r
is
olati
on,
it
us
es
var
i
ou
s
is
olati
on
groups
of
Linux
s
uch
a
s
nam
espac
e,
cgroups
,
an
d
UFS.
Each c
onta
iner
run
s
within
th
at
nam
espace
a
nd
do
e
s
not
ha
ve
a
ny
acc
ess
outsi
de
it.
Con
tr
ol
gro
up
s
(cgrou
ps
)
prov
i
de
co
ntr
ol
ov
e
r
res
ources
.
Cgro
up
s
al
lo
ws
D
oc
ker
for
sh
a
ri
ng
avail
able
res
ources
f
or
gi
ving
eff
ic
ie
nt
m
ulti
-
te
na
nt
env
ir
o
nm
ent
on
the
ho
st.
Cgroups
ca
n
ha
ve
rese
rv
at
ion
or
lim
it
a
ti
on
to
res
ources
f
or
a
co
ntainer
.
UF
S
(
U
nion
F
il
e
Syst
e
m
),
or
Un
io
nFS
[18]
is
a
fi
le
syst
e
m
us
ed
for
m
a
king
D
ock
e
r
li
gh
t
weig
ht
and faste
r by c
reati
ng lay
ers.
Do
c
ke
r uses
U
nionFS
v
a
riant
s su
c
h
A
U
F
S
[
16
]
De
vice
Ma
pp
e
r.
Con
ta
ine
r
f
orm
at
co
m
bin
es
al
l
these
com
po
nen
ts
an
d
pack
a
ges
it
.
By
def
ault,
th
e
fo
rm
at
fo
r
con
ta
ine
r
w
rappin
g
is
li
bc
ont
ai
ner
.
D
oc
ker
con
ta
ine
rs
can
r
un
on
al
l
po
pu
la
r
Li
nux
di
stros
due
to
its
op
e
n
sta
nd
a
rds.
D
oc
ker
c
on
ta
ine
rs
can
r
un
o
n
virt
ual
m
achine,
na
ti
ve
hard
war
e
,
cl
oud
i
nfrastr
uctu
re
pr
ov
i
de
d
by
Goo
gle,
Mi
crosof
t,
Am
azon
W
e
b
Ser
vice
a
nd
oth
e
r
cl
ou
d
pro
vid
er
s.
Eac
h
D
oc
ker
c
onta
iner
will
hav
e
i
ts
root
file
syst
e
m
,
pro
cesses, m
e
m
or
y,
netw
orks,
na
m
espace,
to
ha
ve
str
ong
is
olati
on
.
Be
nef
it
s
of
us
i
ng Doc
ker co
nt
ai
ner
s:
a)
Adding
or r
em
ov
i
ng of c
onta
iners
ensu
res
t
he
scali
ng of se
r
vices.
b)
They are
buil
t
ver
y ea
sil
y an
d q
uickly an
d
t
he
ir d
e
plo
ym
ent tim
e is al
so
v
e
ry less.
c)
The
e
ff
ic
ie
ncy
of
Do
c
ker
c
onta
iners
in
re
ga
rd
s
to
res
ourc
e
util
iz
at
ion
is
ve
ry
hi
gh
due
to
li
m
i
ta
ti
on
and
oth
e
r
re
source
m
anag
em
ent strate
gies.
d)
The de
ns
it
y of
Do
c
ke
r
c
on
ta
in
er ca
n
inc
rease
to han
dle m
ore w
orkloa
ds
.
2
.
3
.
Rk
t
Rkt
(s
pelle
d
a
s
“ro
c
ket”)
de
velo
ped
by
C
or
e
OS
,
I
nc.
[20]
in
Decem
ber
20
14
is
op
en
s
ource
con
ta
ine
r
e
ng
i
ne
desi
gned
for
m
anag
ing
a
ppli
cat
ion
co
nta
iners.
T
he
wor
k
of
r
kt
is
by
de
fau
lt
li
cense
d
unde
r
Ap
ac
he
2.0
li
cense
unle
ss
s
pe
ci
fied
[21].
Feat
ur
es
of
rk
t
’s
a
re:
a)
Pod
-
native
b)
Secu
rity
c)
Com
po
sabili
ty
d)
O
pe
n
sta
ndar
ds an
d
c
om
patibil
ity
Figure
2
.
Perf
orm
ance co
m
par
iso
n betwee
n barem
et
a
l, v
irtual
m
achine & con
ta
ine
rs
Figure
2
s
hows
the
com
par
iso
n
be
twee
n
ba
re
m
et
al
,
KV
M,
Do
c
ke
r,
L
XC,
and
Rkt
wh
e
re
the
val
ues
wer
e
ob
ta
ine
d
by
M.
Ali
Ba
bar
a
nd
Be
n
Ra
m
se
y
[2
2]
on
pe
rfor
m
i
ng
ben
c
hm
arks
for
knowin
g
the
perform
ance
of
nati
ve
m
achine,
Li
nux
c
ont
ai
ner
s,
an
d
virt
ual
m
achine
(
KV
M
).
T
he
re
su
lt
s
sho
w
that
le
sser
the
tim
e
bette
r
is
the
CPU
perform
ance
wh
i
le
do
i
ng
be
nchm
ark
in
g
f
or
C
PU
.
T
he
m
e
m
or
y
be
nch
for
c
opy
is
sh
owin
g
t
he
m
ean
m
e
m
or
y
bandw
i
dth
in
MB
/s.
Her
e,
hi
gh
e
r
the
ba
ndwidth
higher
t
he
co
py
rate.
I
nput
-
Ou
t
pu
t
(
IO)
be
nch
m
ark
pe
rfor
m
ed
for
w
rite
op
erati
on
c
al
culat
ed
in
MB
/s
sh
ows
hi
gh
e
r
the
w
rite
rate
,
higher
t
he
pe
r
form
ance.
Data
is
sent
from
con
ta
ine
r
to
host
an
d
cal
c
ulate
d
in
GB/s.
In
respec
t
to
KV
M
,
the
co
ntaine
r
is
de
plo
ye
d
on
top
of
K
VM,
and
the
n
data
is
sent
to
host.
Higher
the
rate
of
tran
sfe
r
m
or
e
eff
ic
ie
nt
is
the
networ
k
perf
orm
ance.
Acco
r
di
ng
to
the
res
ults
sh
ow
n
in
above
Fig
ure
2,
one
can
say
that
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
1
1
94
–
1
1
05
1098
there
is
no
co
m
pet
it
or
for
ba
rem
et
a
l
bu
t
i
n
so
m
e
cases,
con
ta
ine
rs
gi
ve
equ
al
pe
rform
ance.
Discus
sion
on
con
ta
ine
rs
a
nd
virtu
al
m
achi
ne
sho
w
that
LXC
is
bette
r
than
r
kt,
Do
c
ke
r,
an
d
K
VM
resp
ect
ive
ly
in
CPU
perform
ance.
Wh
il
e
D
oc
ker
is
bette
r
than
r
kt,
LXC
a
nd
K
VM
re
ga
rd
i
ng
m
e
m
or
y
Ba
ndwidth,
sim
i
la
rl
y
for
IO
w
rite
op
e
r
at
ion
LXC
is
bette
r
than
r
kt
wh
il
e
ot
her
s
hav
e
sam
e
per
form
ance
o
bta
ined
with
m
e
m
or
y
band
width.
N
e
twork
pe
rfo
rm
ance
is
gi
ven
be
tt
er
b
y
rk
t
tha
n
L
XC,
D
oc
ke
r,
a
nd
KV
M
re
sp
ect
ively
.
One
can
see
perform
ance v
a
riat
ion
i
n
c
on
ta
ine
rs, b
ut
ov
e
rall
they
pe
rfor
m
b
et
te
r
th
an virtual m
achine.
Perfo
rm
ance
fo
r
sta
rtu
p
tim
e
and
file
co
py
per
f
orm
ance
sh
ows
t
hat
na
ti
ve
is
m
uch
bette
r
tha
n
Do
c
ke
r,
w
hile
Do
c
ke
r per
for
m
s v
ery well
than K
VM
[23].
T
a
b
l
e
1
sho
ws
the
per
f
orm
a
nce
com
par
iso
n
betwee
n
nat
ive,
K
VM,
an
d
D
ock
e
r
en
vi
ronm
ents
ob
ta
ine
d
by
W
e
s
Felt
er
et
al
.
[24]
s
how
t
hat
native
is
al
way
s
a
bette
r
opti
on
for
virt
ual
m
achine
(
KV
M
),
an
d
in
s
om
e
cases,
D
ock
e
r
pe
rform
s
equ
al
to
na
ti
ve.
Wh
il
e
cal
culat
ing
net
work
la
te
ncy,
one
can
see
t
hat
K
VM
perform
s
bette
r
tha
n
Do
c
ke
r
al
though
not
bette
r
tha
n
nat
ive.
T
he
res
ults
sho
w
pe
r
for
m
ance
be
nchm
ark
s
giv
e
n on
CP
U
perform
ance,
m
e
m
or
y
ba
ndwidth,
RAM,
ne
twork
la
te
ncy,
blo
c
k
I/O
,
a
nd
netw
ork ba
ndwidth.
The
reas
on
be
hind
the
decli
ne
in
perform
ance
of
the
virtua
l
m
achine
is
t
he
existe
nce
of
add
it
io
nal
la
ye
rs,
w
he
rea
s
D
oc
ker
has
fe
wer
la
ye
rs
an
d
directl
y
interact
s
with
the
kernel
of
the
host
m
ac
hin
e.
Ther
e
a
re
dif
fe
ren
ces
see
n
be
tween
di
ff
e
ren
t
con
ta
ine
r
for
m
at
s
du
e
to
their
arch
it
ect
ur
e
and
the
pur
po
se
of
their
dev
el
op
m
ent.
Table
1.
C
om
par
iso
n of nati
ve
, v
i
rt
ualiz
at
ion
a
nd Doc
ker c
on
ta
ine
r
3.
CONTAI
NER
ORCHE
STR
ATIO
N
EN
GI
NE
Con
ta
ine
rs
a
re
de
plo
ye
d
ve
r
y
den
sel
y,
where
it
bec
om
e
s
exten
sively
tric
ky
f
or
m
anag
in
g
s
uch
enorm
ou
s
nu
m
ber
of
con
ta
iners.
Or
c
hestr
at
ion
te
chn
i
qu
es
are
us
e
d
to
achieve
auto
m
at
ion
fo
r
ha
nd
li
ng
con
ta
ine
rs.
Co
ntainer
r
untim
e
API
c
an
pe
rfor
m
al
l
the
op
erati
on
s
on
a
c
on
ta
ine
r,
but
it
is
lim
i
te
d
to
a
sing
l
e
con
ta
ine
r.
Actual
prob
le
m
a
rises
wh
e
n
m
ulti
ple
con
ta
i
ne
rs
are
run
ning
on
var
io
us
ho
sts
.
T
o
ha
ve
the
eff
ic
ie
nt m
anag
em
ent o
f
c
onta
iner
a
nd to
ov
erco
m
e above ob
sta
cl
es
we u
se orche
strat
io
n
to
ols.
Com
plex
and
m
ul
ti
ple
con
ta
iner
ap
plica
ti
ons
de
plo
ye
d
in
a
cl
us
te
r
are
carried
out
by
exp
a
ndin
g
capab
il
it
ie
s o
f
Con
ta
ine
r
Or
c
hestrati
on Engi
ne
[13]. A
ll
m
a
nag
em
ent tasks p
erfo
rm
ed
on
co
ntainer
s f
r
om
i
ts
dev
el
op
m
ent,
dep
l
oym
ent,
a
nd
sc
he
du
li
ng
ti
ll
it
s
te
r
m
in
at
ion
is
autom
at
ed
us
in
g
c
onta
iner
orche
str
at
ion
too
ls.
Op
e
nS
ta
ck
sup
ports
only
Do
cke
r
Swa
rm
,
Ap
ache
Me
so
s,
an
d
K
ub
e
r
netes
as
fa
r
its
c
urren
t
ve
rsion
Pike
is
co
ncern
ed.
Ma
gnum
,
one
of
open
s
our
ce
pr
oj
ect
f
ro
m
O
pe
nS
ta
ck
prov
i
des
these
or
chestrati
on
t
oo
ls.
Fo
r
ac
hieving
autom
at
ion
and m
anag
in
g
c
onta
iners
d
i
ff
e
re
nt to
ols ar
e
a
va
il
able:
1)
Lin
u
x
Co
n
ta
in
e
rs
(L
X
C):
LXC is the
um
br
el
la
project
beh
i
nd L
XC, LX
D,
and L
XCFS.
2)
Apa
ch
e
A
u
ro
ra:
In
te
ll
igently
re
sche
du
le
s
fail
ed
c
on
ta
in
ers o
ver ot
her
r
unni
ng m
achines.
3)
Doc
k
er
E
n
g
in
e:
An
a
pp
li
cat
io
n
is s
ha
red an
d r
un acr
os
s
va
riou
s
Lin
ux syst
e
m
s thu
s
pro
vi
ding a
platf
orm
for
a
dm
inist
rat
or
s
and
de
velo
per
s
.
4)
Kontena:
They
h
a
ve been
buil
t t
o
be
d
e
pl
oyed
a
nd ru
n on a
ny in
fr
ast
r
uctu
re. A
pp
li
cat
ions are
de
velo
pe
d
us
in
g Konte
na Ser
vice.
5)
Weav
e
w
orks
:
Con
ta
ins
to
ols
for
m
anag
in
g
a
nd im
ple
m
entin
g m
ic
ro
serv
ic
es in cl
us
te
rin
g.
6)
Wer
c
ker
:
A
utono
m
ou
sly
us
es
p
ipeli
ni
ng c
oncept ’a
uto
m
at
e
d workflo
w
’
f
or
dep
l
oying m
ulti
-
ti
ered
clo
ud
native a
pp
li
cat
ion
s
.
7)
rk
t:
Co
ntaine
rs
m
anag
ed by u
sing C
omm
and
Line
In
te
rf
ace
(CL
I
).
Na
tive
K
VM
Do
ck
e
r
CPU
(PXZ
)
co
m
pr
e
s
si
ng
1
GB o
f
W
ik
ip
ed
ia
d
ata
7
6
.2
(
±
0
.93
)
5
9
.2
(
±
1
.88
)
less
er
2
2
%
7
3
.5
(
±
0
.64
)
Lesser 4%
Me
m
o
r
y
Ba
n
dw
i
d
t
h
(S
trea
m
)
Co
p
y
(GB/s)
4
1
.3
(
±
0
.06
)
4
0
.1
(
±
0
.21
)
less
er
3
%
4
1
.2
(
±
0
.08
)
Lesser 0%
RAM
(R
an
do
m
Access)
GUPS
0
.01
2
6
(
±
0
.00
0
2
9
)
0
.01
2
5
(
±
0
.00
0
3
2
)
less
er
1
%
0
.01
2
4
(
±
0
.00
0
4
4
)
Lesser 2%
Netwo
rk La
ten
cy
(
N
et
pe
rf
)
less
than
40
µ
s
less
than
70
µ
s
m
o
re
th
an
70
µ
s
Blo
ck
I
/O(f
io
)
(I
O
PS
I
np
ut
/
o
u
tp
u
t
o
p
eration
s
Read
I
O is
near to
8
0
0
0
0
I
OPS
write
is
abo
v
e 10
0
0
0
0
I
OPS
Read
I
O is
near to
4
0
0
0
0
IOPS
write
is
belo
w 60
0
0
0
I
OPS
Si
m
ila
r
to
Native
Si
m
ila
r
to
Nativ
e
Netwo
rk
Ba
nd
w
i
dt
h
(nu
ttcp
)
Tr
an
s
m
it
aou
n
d
2 cy
cles/b
y
te
Receiv
e abo
v
e 2 le
ss
than
2.5
cy
cles/b
y
te
Tr
an
s
m
it
between
2
-
2
.5
cy
cles/b
y
te
Receiv
e abo
v
e 3 c
y
cles/b
y
t
e
Ab
o
v
e 2.5
cy
cle/b
y
te
Receiv
e between
2
.5 to
3
cy
cles/b
y
te
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
In
te
gr
atio
n of
Linu
x
Co
nta
i
ne
rs in O
penSta
ck:
A
n In
tr
osp
ect
ion
(
As
hish
Lingay
at
)
1099
3.1.
D
ocker
S
w
arm
Do
c
ke
r
Sw
a
rm
is
an
op
en
sou
r
ce
cl
us
te
r
and
orchest
rati
on
m
anag
em
ent
too
l
[2
5].
D
ock
e
r
pro
vid
es
a
sta
nd
a
rd
API
f
or
c
omm
un
ic
ating
with
c
on
ta
i
ner
s
.
S
war
m
us
es”
swa
p,
plug
an
d
play
”
for
Do
c
ke
r
co
ntainer
s
.
Sw
a
p
ens
ur
es
that
Do
c
ker
co
ntain
ers
ar
e
hig
hly
portable
even
wh
e
n
r
unning.
D
oc
ker
Sw
arm
arch
it
ect
ur
e
consi
sts
of
m
a
ste
r
nodes
an
d
wo
r
ke
r
nodes
.
The
co
nf
i
gur
at
ion
of
Sw
a
r
m
cl
us
te
r
can
be
declarat
ivel
y
done
us
in
g
Y
A
M
L
fi
le
s
[26]
.
Keyw
ords use
d
in
Doc
ker S
war
m
:
a)
Nod
e:
Distribu
te
d
ac
ro
s
s
on
-
pr
e
m
ise
s o
r p
ub
li
c cloud.
An i
nst
ance of a
S
wa
rm
.
b)
S
wa
r
m:
Or
ches
trat
ing
se
r
vices
in
a cl
us
te
r
is
autom
at
ed.
c)
Ma
n
ag
e
r n
o
d
e:
The use
r
se
nds
d
ist
rib
uted
servic
e d
e
finiti
on
to wo
rk
e
r node
s.
T
hey can
the
m
se
lves act
as
worker
node
.
d)
Wo
rke
r N
o
d
e:
Re
cei
ves
an
d r
un
s
tasks
assi
gned
b
y t
he
m
ast
er no
de.
e)
Se
rvi
ce
:
Serv
ic
e is inte
nde
d for s
pecifyi
ng c
onta
iner
and
re
pl
ic
as.
f)
Task
:
Is
an
at
om
ic
u
nit o
f
a
s
erv
ic
e as
sig
ned to
wor
ker
node
.
Feat
ur
es
of
D
oc
ker S
war
m
:
a)
Inb
uilt
clustering usi
ng
Do
c
ke
r
E
ng
i
ne.
b)
Decent
rali
zed
desig
n.
c)
Declarat
ive
Se
rv
ic
e M
od
el
.
d)
Scal
ing
.
e)
Mult
i
-
te
nan
t
di
sco
ver
y
of ser
vi
ce, f
a
ult t
olera
nce.
f)
Roll
ing
update
s.
3.2.
Ku
berne
t
es (abbre
viate
d a
s
K
8s
)
”
Kubernete
s
is
an
op
e
n
sou
rce
too
l
for
de
plo
ym
ent
and
m
anag
em
ent
us
ing
aut
om
at
io
n
for
the
con
ta
ine
rized
a
pp
li
cat
io
n
[
27]
.”
K
ub
e
r
netes
was
devel
oped
by
Goo
gle
f
or
inter
nal
cl
us
te
r
m
anag
e
m
ent
was
pr
e
viously
known
as
B
org
(a.
k.
a.
Om
ega)
[
28]
.
It
was
in
J
une
2014
t
hat
G
oogle
a
nnou
nc
ed
it
s
open
sou
rce
cl
us
te
r
m
anag
e
m
ent
too
l
cal
l
ed
Kube
rn
et
es
in
Goo
gle
De
velo
per
F
orum
[
29
]
.
Lat
er
G
oogle
donate
d
it
to
Cl
oud
Nati
ve
Com
pu
ti
ng
Foundati
on.
It
prov
i
des
a
unif
orm
AP
I
f
or
m
a
nag
i
ng
co
ntain
ers
in
the
cl
us
t
er
of
native a
nd v
i
rtual
m
achine.
Kube
rn
et
es
us
e
s
its
keywor
ds
and
al
s
o
a
dd
s
ne
w
co
nce
pts.
C
om
po
ne
nts
of
Kube
rn
et
es a
re
:
a)
Nod
e:
Nati
ve
or V
M
, ru
nn
i
ng
on to
p wh
e
re c
on
ta
ine
rs
a
re sc
heduled
.
b)
Pod
s:
They are
set
o
f
contai
ne
rs
c
om
bin
ed
lo
gical
ly
.
c)
Se
rvi
ce
s:
Serv
ic
es are
pro
vid
e
d usin
g IP
a
ddr
ess and
netw
or
k protoc
ol.
d)
Re
p
li
ca
ti
o
n
Co
n
tro
ll
er:
Used
for ef
fici
ent m
a
nag
em
ent of th
e
cl
us
te
re
d
e
nv
iro
nm
ent.
e)
A
P
I se
rve
r:
Ku
bernetes m
ast
e
r node
us
es
A
P
I
as m
anag
em
ent hu
b.
f)
Co
n
tro
ll
e
r
Man
ag
er:
The
de
sir
ed
sta
te
is m
at
c
hed w
it
h cu
rr
e
nt stat
e on sc
al
ing
w
orkloa
ds
in a clu
ste
r.
g)
Sche
d
u
le
r:
It assign
s
workl
oad to a
n
a
ppr
opriat
e n
ode.
h)
Kub
el
et
:
Ma
n
a
ges p
od
s
that a
re ru
nn
i
ng on t
he host
, by rec
ei
vin
g co
nfi
gur
at
ion
from
A
PI ser
ver.
i)
Lab
el
s:
Key
-
va
lue p
ai
rs use
d
f
or searc
hing a
nd
updatin
g
c
onta
iners.
3.
3
.
A
pa
c
he
Mes
os
Ap
ac
he
Me
s
os
is
al
so
cal
le
d
as
the
kernel
of
distrib
uted
sy
stem
s,
as
it
is
m
or
e
op
e
n
t
ha
n
Do
c
ke
r
a
nd
Kube
rn
et
es
an
d
has
fine
gr
a
nu
la
rity
.
In
Me
so
s,
res
ource
m
anag
em
ent
a
nd
ta
sk
sc
heduling
is
i
m
plem
ented
ind
ivi
du
al
ly
.
Ap
ac
he
Me
s
os
pro
vid
es
an
a
bs
tract
io
n
for
com
pu
ti
ng
res
ources
i
n
la
r
ge
cl
us
te
rs
.
T
his
sh
a
re
d
pool
is
c
reated
by
c
om
bin
ing
CP
U
,
m
e
m
or
y,
an
d
st
or
a
ge.
Thes
e
s
har
e
d
pools
a
re
t
hen
al
locat
ed
as
pe
r
fi
ne
-
gr
ai
ned
resou
r
ce requirem
ents. Apach
e Me
s
os
is fau
lt
-
t
oler
ant and scala
bl
e fo
r
m
assivel
y sca
li
ng
appli
cat
ion
s
or ser
vices
[30
]
.
Me
so
s
pro
vid
e
s d
ist
ri
bu
te
d
sy
stem
s u
sin
g:
a)
Apa
ch
e
Au
ro
r
a
:
It
is
serv
ic
e
scheduler
f
or
lo
ng
-
r
unni
ng
c
onta
iners
a
nd
ha
s
the
capab
il
it
y
of
scal
ing
th
e
m
highly
.
Additi
onal
featu
res
of
Aurora
a
re
r
olli
ng
updates
,
a
quota
for
res
our
ces
an
d
ser
vice
r
e
gistrati
on.
b)
Ch
ron
o
s:
It is f
ault
-
tolera
nt sc
heduler
u
se
d f
or r
e
placi
ng cr
on jo
bs
t
o or
c
he
strat
e co
ntain
ers.
c)
Ma
r
at
h
o
n
:
It is
serv
ic
e
sche
dule
r bu
il
t o
n
t
op
of Mesos
and
Chronos.
3.
4
.
Micr
osof
t
Az
ure
Servic
e Fabri
c
It
is
a
platfo
rm
fo
r
distrib
uted
syst
em
s
us
e
d
for
en
ca
ps
ulati
n
g,
devel
op
in
g
a
nd
dep
l
oying
con
ta
ine
rs.
T
he
dep
loye
d
ser
vices
are
scal
able,
highly
m
a
nag
ea
ble
an
d
r
el
ia
ble
fo
r
the
producti
on
-
re
ady
env
i
ronm
ent.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
1
1
94
–
1
1
05
1100
Ca
pab
il
it
ie
s o
f
Ser
vice Fa
br
ic
are:
1)
Pr
ovisi
onin
g.
2)
Dep
l
oying.
3)
Mon
it
ori
ng.
4)
Upgradi
ng.
5)
Delet
ion
.
Ser
vices
pr
ov
i
ded b
y M
ic
r
osoft’
s Se
rv
ic
e
F
abr
ic
a
re
[31]:
1)
Azure
SQ
L
D
a
ta
base.
2)
Azure C
os
m
o
DB.
3)
Cortana
.
4)
Mi
cro
s
of
t P
ow
er BI
.
5)
Mi
cro
s
of
t
In
t
une.
6)
Azure E
ve
nts
Hub.
7)
Azure
IoT
Hub
.
8)
Dynam
ic
3
65.
9)
Sk
ype
for B
us
i
ness.
10)
More
ov
e
r, m
a
ny m
or
e.
Ser
vices
can
be
bu
il
t
us
in
g
Serv
ic
e
Fa
br
ic
Pr
og
ram
m
ing
m
od
el
s
[3
2],
AS
P
.
NET
c
ore
[3
3].
It
ha
s
the
capa
bili
ty
of
bu
il
di
ng
st
at
el
ess
or
sta
te
fu
l
ser
vices.
Mi
cro
ser
vices
can
orche
strat
e
and
a
uto
m
ate
in
Ser
vice Fa
br
ic
[34].
W
i
ndows
cont
ai
ner
ty
pes:
1)
W
i
ndows
Ser
ve
r
Co
ntainer
s
[35]:
It
util
iz
es
ke
rn
el
of
t
he
host
to
ac
hi
eve
isolat
io
n
by
abst
racti
ng
process
a
nd
nam
espace.
Li
m
it
a
ti
on
of
W
indow
ser
ver
con
ta
ine
r
is
th
e
requirem
ent
for
the
sam
e
ver
si
on of
kern
el
an
d co
nfi
gur
at
ion
du
e
to
th
e u
ti
li
zat
ion
of
ho
st
k
e
rn
el
.
2)
Hyp
er
-
V
Isol
at
io
n
[35
]:
M
ic
roserv
ic
es
r
un
in
a
virtu
al
m
achine
wh
ic
h
is
hig
hly
optim
iz
ed.
They
do
not
sh
are
a
kernel
of Host a
nd th
us
a
re inde
penden
t
of
kernel
ver
si
on and c
onfi
gurati
on.
3.
5
.
G
oogle
C
on
t
ainer
Eng
i
ne
-
Ku
bernet
es
Goo
gle
Co
ntainer
E
ngine
runs
K
uber
netes
1.8,
s
upporti
ng
high
a
vaila
bili
ty
,
m
ulti
-
maste
r
cl
us
te
r
wh
ic
h
im
pr
ove
s Service
Le
ve
l Objecti
ves
to 99.9
9%
[36
]
.
Kube
rn
et
es
E
ngine
runs
on
it
’s
ope
n
sou
rc
e
pro
j
ect
cal
le
d
as
Kube
rn
et
es
w
hich
was
pr
e
viously
Goo
gle’s
i
nter
nal cluster
m
anag
em
ent p
la
tfo
rm
, th
e
n
cal
le
d”
Borg”
[
37
]
.
Fr
om
G
m
ai
l
t
o
Y
ou
T
ube
to
Search,
ever
y
thing
at
G
oogl
e
runs
in
co
ntainers
[
37]
.
G
oogle
create
s
ov
e
r 2
bill
ion
c
on
ta
ine
rs
ea
ch
week.
Feat
ur
es
of
G
oogle C
onta
iner
Engine
are:
1)
Hardwa
re
Acc
el
erati
on
.
2)
Node A
uto
re
pa
ir.
3)
Au
t
os
cal
in
g
cl
us
te
rs
.
4)
Me
te
ring
a
nd s
cal
ing
.
5)
Extensi
bili
ty
.
6)
Node A
uto
up
gr
a
de.
7)
Secu
rity
an
d re
li
abili
ty
.
8)
Con
ta
ine
r Nat
ive
Netw
orkin
g (E
xclusi
ve wit
h Goo
gle Cl
ou
d
Plat
f
or
m
(
G
CP))
.
9)
Mon
it
ori
ng.
Be
nef
it
s
of
us
i
ng Ku
bernetes
En
gin
e:
1)
Id
e
ntit
y and ac
cess m
anag
em
ent.
2)
Op
e
nsource
po
rtabil
it
y.
3)
Au
t
o upg
rad
e
.
4)
Secu
rity
.
5)
Hybr
i
d Netw
orkin
g.
6)
Au
t
os
cal
in
g.
7)
Au
t
o repair
.
8)
Re
so
urce l
im
i
t
s.
9)
Pr
ivate
c
onta
in
er r
e
gistry.
10)
Do
c
ke
r
im
age su
pp
or
t.
11)
Straig
ht full
appli
cat
ion
s
uppo
rt.
12)
Mon
it
ori
ng a
nd lo
ggin
g.
13)
Fu
ll
y m
anag
ed
.
14)
Op
e
rati
ng syst
e
m
b
uilt
for co
nta
iners
.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
In
te
gr
atio
n of
Linu
x
Co
nta
i
ne
rs in O
penSta
ck:
A
n In
tr
osp
ect
ion
(
As
hish
Lingay
at
)
1101
3.
6
.
Am
az
on
ECS
(A
m
az
on Web
Services
El
as
tic
C
on
t
ai
ner Service
)
Am
azon
W
e
b
Ser
vices
ha
s
it
s
pro
pr
ie
ty
c
onta
iner
m
anag
e
m
ent
serv
ic
e
cal
le
d
as
Am
a
zon
Ela
sti
c
Con
ta
ine
r
Se
r
vi
ce
(A
m
azon
E
CS)
[
38]
.
To
i
nteract
with
its
co
ntainer
ecos
yst
e
m
,
i
t
prov
i
des
native
AP
I
for
con
ta
ine
rs.
It
has
ca
pa
bili
ti
e
s
of
m
anag
in
g
cl
us
te
r
en
vir
onm
ent
to
m
ake
Am
azon
ECS
faster,
scal
a
ble
to
op
e
rate
Do
c
ker co
ntainers
. Fo
r
acce
ssi
ng Am
azon
ECS
, the
re is
no n
ee
d f
or instal
li
ng any
so
ftwa
re.
The follo
wing
m
et
ho
ds ca
n b
e u
se
d
t
o
ac
ces
s A
m
azon
EC:
1)
A
WS
Ma
na
ge
m
ent Con
s
ole:
A web
-
ba
sed
interf
ace
for m
a
nag
i
ng A
m
azon
ECS
.
2)
A
W
S
Com
m
and
Line
T
o
o
l
(C
LI)
:
a
faster
a
nd
co
nvenie
nt
to
ol
than
us
in
g
m
anag
em
ent
co
nsole
.
3)
Am
azon
ECS
CLI:
this
interf
ace
com
es
un
de
r
com
m
and
lin
e
too
l, u
se
d
f
or
entirel
y
co
ntro
ll
in
g
Am
azon
ECS.
4)
A
WS
S
D
K:
Us
ed
for
acce
ssi
ng
Am
azon
EC
S th
rou
gh v
a
riou
s
pr
ogram
m
i
ng lan
gu
a
ges
.
Feat
ur
es
of
A
m
azon
ECS
ar
e:
a)
A
WS
Fa
rgat
e
Suppor
t
.
b)
Do
c
ke
r
S
uppor
t.
c)
Com
patible
w
it
h W
i
ndows C
on
ta
ine
r.
d)
Local
dev
el
opm
ent n
at
ive s
uppo
rt.
e)
Cl
us
te
r
c
on
tr
ol
.
a.
Sche
du
li
ng.
b.
Task plac
em
ent p
olicy
.
f)
Netw
ork
a
nd s
ecur
it
y.
g)
Loa
d balanci
ng.
h)
Mon
it
ori
ng a
nd lo
ggin
g.
The
ty
pes
of C
on
ta
ine
r
im
age suppo
rted:
a)
Do
c
ke
r.
b)
Kube
rn
et
es.
c)
Core
OS
.
4.
CLOUD
A
ND OPEN
ST
ACK
Virtuali
zat
ion
te
chn
iq
ues
are
us
e
d
in
the
cl
oud
t
o
ac
hieve
m
axi
m
u
m
uti
l
iz
at
ion
of
resour
ces
an
d
scal
ing
of
sh
a
red
res
ource
s.
Ther
e
are
va
r
iou
s
ve
rsions
for
the
def
i
niti
on
of
cl
oud
c
om
pu
ti
ng
gi
ve
n
by
diff
e
re
nt institu
ti
on
a
nd
orga
nizat
ion
. In
cl
oud
c
om
pu
ti
ng
, reso
urces
are
a
vaila
ble
to
us
e
ov
e
r
a
netw
ork
with
m
ini
m
al
eff
or
ts
in
m
anag
ing
t
hem
.
Cl
oud prov
i
des
foll
ow
i
ng serv
ic
es b
ase
d on it
s u
sa
ge:
1)
Infr
ast
ru
ct
ur
e a
s a S
e
rv
ic
e
(I
aa
S)
.
2)
Plat
fo
rm
as a Se
rv
ic
e
(P
aaS
).
3)
So
ft
war
e
as a
Ser
vice (S
aa
S).
T
a
b
l
e
2
giv
es
a
li
st
of
cl
oud
pro
vid
e
rs
al
ong
with
th
ei
r
c
onta
iner
s
er
vice
nam
e
wh
e
re
f
e
w
of
the
m
pro
vid
e m
or
e t
han one c
onta
iner
ser
vice.
Table
2.
Cl
oud
Prov
i
ders
Along
with
T
heir C
on
ta
ine
r
Se
r
vi
ces
Clo
u
d
Pr
o
v
id
er
Co
n
tain
er
Service
Alib
ab
a Clo
u
d
Alib
ab
a Clo
u
d
Co
n
tain
er
Service
A
m
azon
W
eb
Se
rvices
-
A
m
azon
E
lasti
c
Co
n
tain
er
Service
(E
CS)
[
3
8
]
-
A
m
azon
Far
g
at
e
Clo
u
d
Found
ry
W
arden
and
G
arde
n
[
3
9
]
Dell techn
o
lo
g
ies
REX
-
Ra
y
-
A Stora
g
e Or
ch
estrato
r
Go
o
g
le Clo
u
d
Platf
o
r
m
Go
o
g
le Kub
ernetes Eng
in
e
IBM
Clo
u
d
IBM
Clo
u
d
Co
n
tain
er
Service
Micr
o
so
f
t Azure
Micr
o
so
f
t Azure F
ab
ric
Se
rvice
Red
Hat
Op
en
Sh
if
t
Red
Hat
Op
en
Sh
if
t
Co
n
tain
er
Platf
o
rm
Orac
le
Clo
u
d
I
n
f
rastructu
re
Orac
le
Clo
u
d
I
n
f
rastructu
re
Co
n
tain
er
Service
Clas
sic
Rack
sp
ace
Carin
a [
4
0
]
Op
en
Stack
-
Magn
u
m
-
Zun
-
No
v
a Dock
er
H
y
p
ervis
er
-
Ko
lla
-
Kata
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
1
1
94
–
1
1
05
1102
4
.
1
.
O
penS
t
ac
k
Op
e
nS
ta
c
k
is
fr
ee
an
d
open
s
ource
cl
oud
operati
ng
syst
em
that
prov
i
des
I
nfrastr
uctu
re
as
a
Serv
ic
e
(I
aaS
)
by
virtua
li
zi
ng
com
pu
ta
ti
on
al
res
ourc
es.
O
penSta
ck
sta
rted
as
a
pro
j
ect
of
Ra
ck
sp
ace
Ho
sti
ng
an
d
NASA
j
oi
ntly
in
2010.
Op
e
nSt
ack
is
now
m
anag
ed
by
O
penSta
ck
f
oundat
ion
a
nd
s
upported
by
m
or
e
tha
n
500
c
om
pan
ie
s
since
Septem
ber
20
12.
Ma
ny
of
the
le
adin
g
s
of
t
war
e
in
dustrie
s
ha
ve
e
xt
end
e
d
t
heir
s
uppo
rt
in
de
velo
pi
ng
Op
e
nS
ta
c
k.
Op
e
nS
ta
c
k
co
m
bin
es
oth
er
op
e
n
s
ource
pro
j
ect
s
(
or
t
ools)
for
pr
ov
i
din
g
cl
oud
inf
ra
struct
ure
.
Fo
r
pro
vid
in
g
cor
e
cl
oud
c
om
pu
ti
ng
res
ources
,
it
us
es
six
op
e
n
sou
rce
too
ls
f
or
com
pu
ti
ng
,
st
or
a
ge,
netw
orkin
g,
i
m
age
storing,
identific
at
ion,
or
c
hestrati
on.
The
ser
vices
pro
vid
i
ng
the
to
ols
are
cal
le
d
Nova,
Sw
ift,
Ci
nder,
Neu
t
ron,
Gla
nc
e,
Keyst
one,
a
nd
Heat
resp
ec
ti
vely
.
Op
e
nS
t
ack
us
es
A
pp
li
cat
ion
Pro
gr
am
m
ing
In
te
r
face
(
API
)
f
or
inte
racti
ng
with
res
our
ces.
In
2014,
the
Op
e
nS
ta
c
k
com
m
un
it
y
decided
to
s
upport
con
ta
ine
rs
i
n
Op
e
nS
ta
c
k.
O
pe
nS
ta
ck
is
ca
pa
ble
of
handlin
g
syst
em
con
ta
iners
li
ke
LXC
(Lin
ux
Co
ntai
ner
s
)
and
Virt
uo
zz
o.
Fr
om
Liberty
release
of
O
pe
nS
ta
ck
orchest
rati
on
t
oo
ls
like
Do
c
ke
r
S
warm
,
Ku
be
rn
et
es
an
d
Ap
ac
he
Me
s
os
are
al
so
incl
uded
.
O
pe
nS
ta
c
k
Co
ntainer
s
T
e
a
m
fo
rm
ed
in
Ma
y
20
14
is
w
orkin
g
to
wards
the
creati
on
of
ne
w
too
ls
for
ha
nd
li
ng
c
on
ta
in
e
r
te
chnolo
gy.
Their
ai
m
an
d
ob
j
ect
ive
is
to
giv
e
us
e
r
-
fri
end
ly
exp
e
rience
for
m
anag
in
g
a
nd
creati
ng contai
ner
s
in
O
penSta
ck.
Figure
3. Com
par
is
on b
et
wee
n barem
et
al
, O
penSta
ck, an
d Do
c
ke
r
c
on
t
ai
ne
r
Figure
3
Jy
oti
Sh
et
ty
et
al
.
[
41
]
pro
vid
es
the
perform
ance
res
ults
obta
ined
usi
ng
Be
nc
hm
ark
s
on
CPU,
m
e
m
or
y,
data
rea
d/wr
it
e
and
A
pach
e
ben
c
hm
ark
.
T
he
res
ults
s
how
that
bar
em
et
al
is
a
bette
r
op
ti
on
than
D
oc
ker
a
nd
O
pe
nS
ta
c
k,
but
Do
c
ker
outpe
rfor
m
s
bare
m
et
al
in
Ap
a
che
be
nch
m
ark
.
The
reas
on
beh
i
nd
the
decli
ne
i
n
perform
ance
of
O
pe
nS
ta
ck
is
the
presenc
e
of
a
ddit
ion
al
la
ye
rs.
D
oc
ker
sh
ows
bette
r
r
esults
for
A
pache
be
nch
m
ark
agai
nst
bar
em
et
al
.
Do
c
ke
r
outpe
r
form
ing
bar
em
et
al
is
du
e
to
the
lowe
r
c
om
plexit
y
of
D
oc
ker
f
or
pro
vid
i
ng
the
s
erv
ic
es.
T
he
re
su
lt
s
sh
ow
that
Do
c
ker
is
a
be
tt
er
op
ti
on
tha
n
O
penSta
ck,
bu
t
it
la
cks
capa
bili
ties
of
cl
oud
c
om
pu
ti
ng
.
O
penSta
ck
carries
c
har
act
erist
ic
s
of
cl
oud
that
ar
e
no
t
pro
vid
e
d
by
Do
c
ke
r.
By
re
su
lt
s,
one
ca
n
unde
rstan
d
tha
t
there
is
en
orm
ou
s
scop
e
f
or
im
pr
ov
in
g
th
e
perform
ance
of
Op
e
nS
ta
c
k by
integ
rati
ng cont
ai
ner
s
with
th
e
m
.
Pr
oject
s
f
or
i
ntegr
at
in
g
c
onta
iners
in O
penSta
ck:
4
.
1
.
1.
Ope
nS
t
ack Koll
a
It
pro
vi
des
c
onta
iners
that
a
re
producti
on
read
y
al
ong
with
t
he
dep
l
oym
ent
too
ls
require
d
for
m
anag
in
g
te
am
.
Sub
-
pro
j
ect
s a
vaila
ble fo
r
K
ol
la
instal
la
ti
on
are:
1)
Ko
ll
a
-
A
ns
ible.
2)
Ko
ll
a
Kube
rn
e
te
s.
Ko
ll
a
prov
i
des
D
oc
ke
r
co
ntai
ner
s a
nd
An
si
bl
e p
la
yboo
ks
to
d
epl
oy in Ope
nS
ta
ck
on a ba
rem
et
a
l or
virtu
al
m
achin
e, and
kube
r
netes te
m
plate
s to
de
plo
y
kube
rnet
es in
Op
e
nS
t
ack a
nd m
eet
Ko
ll
a’s
m
issi
o
n.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
In
te
gr
atio
n of
Linu
x
Co
nta
i
ne
rs in O
penSta
ck:
A
n In
tr
osp
ect
ion
(
As
hish
Lingay
at
)
1103
4
.
1
.
2.
Ope
nS
t
ack M
agnum
Ma
gnum
us
es
Op
e
nS
ta
c
k
A
P
I
f
or
pro
vid
i
ng
con
ta
i
ner
or
c
he
strat
ion
e
ngin
es
li
ke
Sw
a
rm
,
kub
e
rn
et
es
and
A
pac
he
Me
so
s
[42].
It
us
es
Heat
ser
vice
of
O
penS
ta
ck
f
or
or
c
he
strat
ing
operat
ing
syst
em
Im
age
of
Do
c
ke
r
an
d
ku
bernetes
to
run
in
virtu
al
m
ac
hin
e
or
ba
re
m
et
al
in
cl
us
te
r.
Ma
gnum
pr
ov
i
des
the
sam
e
l
evel
of
abstracti
on
si
m
il
ar
to
No
va
run
ning
a
virt
ua
l
m
achine.
Pyt
hon
m
agn
um
cl
ie
nt
us
es
tw
o
bin
aries
,
RES
T
API
serv
e
r, an
d pyt
hon
c
onduct
or
to ru
n
the
proc
ess.
Pyt
hon
m
agn
um
c
li
ent
is
horizo
ntall
y
scal
able.
Ma
gnum
us
es
Heat
for
or
c
hestr
at
ing
Do
c
ke
r
a
nd
COE
for
a
utom
at
ing
cluster
config
ur
at
io
n.
The feat
ures
of Open
Stac
k
M
agnum
are:
1)
Cl
us
te
r
isolat
io
n
-
m
ulti
te
nan
cy
in
cl
ust
er.
2)
Av
ai
la
bili
ty
o
f
Ap
ac
he
Me
s
os, D
oc
ker S
warm
, k
8s
.
3)
Keyst
one inte
grat
ion
-
ch
oice
of
us
in
g VM o
r nati
ve.
4)
Neu
t
ron
i
ntegrat
ion
.
5)
Ci
nd
er
inte
gr
at
ion
.
6)
Th
e con
tainers in
clu
ster gi
ve l
ess
ico
n
ma
nag
e
me
nt
u
sing
stand
ar
d
API.
Figure
4
.
The
a
rch
it
ect
ure
of
Nova D
ock
e
r h
yper
visor [4
3]
4
.
1
.
3.
O
penS
ta
c
k No
va Do
cker h
yp
er
vis
or
Af
te
r
the
la
unc
h
of
Ha
va
na,
O
penSta
ck
has
intr
oduce
d
a
D
oc
ker
dri
ve
r
wh
i
ch
is
hype
r
-
vis
or
dri
ve
r
i
n
Op
e
nS
ta
c
k
Nova
c
om
pu
te
cal
le
d
O
penSta
ck
Nova
D
ock
e
r
hy
perviso
r
[41].
D
oc
ker
int
ern
al
API
inte
ract
s
with
t
he
N
o
v
a
’
s
HTTP
cl
i
ent
us
i
ng
U
nix
s
ock
et
.
Fi
gure
4
s
hows
Do
c
ke
r
is
m
anag
e
d
an
d
c
ontr
olled
us
in
g
HTTP
A
PI.
All
the
Do
cke
r
im
ages
get
store
d
in
i
m
age
serv
ic
e
of
Op
e
nSt
ack
cal
le
d
G
la
nce.
The
co
nt
ai
ner
s
run usin
g
t
he
D
ock
e
r virt
dr
i
ve
r
in
No
va by
us
in
g virt
A
PI
.
4
.
1
.
4.
O
penS
ta
c
k Z
un
Ma
nag
i
ng
c
on
t
ai
ner
s
in O
pe
nSt
ack
is
done
usi
ng
Z
un
(e
x.
H
ig
gin
s
).
It p
r
ov
i
des
O
pe
nS
t
ack
A
PI
f
or
dep
l
oying
an
d
m
anag
in
g
c
on
t
ai
ner
s
us
in
g
va
rio
us
c
on
ta
ine
r
te
chnolo
gies.
It
pro
vide
s
the
AP
I
for
m
anagi
ng
and
c
on
tr
olli
ng
dif
fer
e
nt
c
onta
iners
in
O
penS
ta
ck
[
44]
.
Com
po
ne
nts of
Zun:
1)
Zu
n API
:
De
plo
yi
ng contai
ne
rs.
2)
Zu
n
Com
pu
te
:
Dep
l
oying an
d m
anag
ing
c
on
ta
iners.
3)
Keyst
one:
Iden
ti
ty
m
anag
em
e
nt.
4)
Neu
t
ron:
Netw
orkin
g.
5)
Glance: St
or
i
ng c
on
ta
ine
rs
i
m
ages.
6)
Do
c
ke
r netw
ork plu
gin
.
4
.
1
.
5.
Ope
nS
t
ack
Kat
a
It
is
an
op
e
n
source
pr
oj
ect
w
hich
ai
m
s
fo
r
c
om
bin
ing
sec
uri
ty
adv
anta
ges
of
virt
ual
m
ac
hin
es
wit
h
the
s
peed
an
d
m
anag
eabil
it
y
of
co
ntaine
rs.
This
pro
j
ect
wi
ll
ensure
t
hat
li
gh
t
weig
ht
virtu
al
m
achines
will
be
bu
il
t
us
in
g
fe
a
tures
of
VMs
and
c
onta
iner
s
.
Feat
ur
e
s
su
c
h
as
w
orkloa
d
isolat
ion
a
nd
s
ecur
it
y
from
virtu
al
m
achine
an
d
pe
rfor
m
ance
li
ke
co
ntainers
ar
e
pro
posed
t
o
i
nteg
rate
into Kat
a.
O
penSta
ck
Kata
de
sig
n
m
akes
it
co
m
patible
w
it
h Op
e
n
C
on
ta
iner
I
niti
at
ive
(O
C
I), a
nd Co
ntainer
Ru
ntim
e I
nterf
ace
(CR
I)
f
or
kube
r
netes.
Evaluation Warning : The document was created with Spire.PDF for Python.