Internati
o
nal
Journal of Ele
c
trical
and Computer
Engineering
(IJE
CE)
V
o
l.
6, N
o
. 4
,
A
ugu
st
2016
, pp
. 18
66
~
1
879
I
S
SN
: 208
8-8
7
0
8
,
D
O
I
:
10.115
91
/ij
ece.v6
i
4.1
014
4
1
866
Jo
urn
a
l
h
o
me
pa
ge
: h
ttp
://iaesjo
u
r
na
l.com/
o
n
lin
e/ind
e
x.ph
p
/
IJECE
Cloud Computing CPU Allocation
and Scheduling Algorithms
Using CloudSim Simulator
Gibet T
a
ni Hi
cham
, El Amr
a
ni
Ch
aker
Laborator
y
of
In
formatics S
y
s
t
ems and Telecom
m
unicati
ons (
L
I
S
T), Dep
a
rtement of Computer
Engineering
Faculty
of Scien
ces an
d
Technologies, A
bdelmalek Essaad
i Univ
ersity
Route
Ziaten
, B
.
P. 416 Tangier,
Morocco
Article Info
A
B
STRAC
T
Article histo
r
y:
Received Feb 11, 2016
Rev
i
sed
Ap
r
19
, 20
16
Accepte
d
May 3, 2016
Cloud Computing is an
emerging co
mputin
g model, wher
eas Cloud
providers and u
s
ers are lookin
g
forward to profit and enh
a
n
ce their IT
exploitation
.
In
this paper
,
we
desc
ribe and d
i
scuss the Cloud
Computing
basic
com
pute r
e
sources schedu
l
i
ng and
al
loc
a
tio
n algor
ithm
s
, in
addition
t
o
the working m
echanis
m
.
This
paper al
so presen
ts a number of
experiments
conducted based on CloudSim
simulation too
l
kit in
order
to
assess and
evalu
a
te
the
p
e
rform
ance
of
th
ese scheduling algor
ithms on Cloud
Computing like infrastructure. Furtherm
ore, we introduced and explain
e
d the
CloudSim simulator d
e
sign, ar
chitect
ure and
pr
oposed two n
e
w scheduling
algorithms to enhance th
e existent one
s and highlight the weakn
e
sses and/or
effec
tiven
es
s
of thes
e algori
t
hm
s
.
Keyword:
C
l
ou
d c
o
m
put
i
n
g
C
l
ou
d c
o
m
put
i
n
g
si
m
u
l
a
t
i
on
First com
e
first
ser
v
ed
Ro
und
rob
i
n
Sche
dul
i
n
g al
g
o
ri
t
h
m
s
Copyright ©
201
6 Institut
e
o
f
Ad
vanced
Engin
eer
ing and S
c
i
e
nce.
All rights re
se
rve
d
.
Co
rresp
ond
i
ng
Autho
r
:
Gi
bet
Ta
ni
Hi
cham
,
Lab
o
rat
o
ry
of
I
n
f
o
rm
at
i
c
s Sy
st
em
s and Tel
e
c
o
m
m
uni
cat
i
ons (L
IST
)
,
De
pa
rt
m
e
nt
of C
o
m
put
e
r
E
ngi
neer
i
ng,
Faculty of Scie
nces a
n
d Tec
h
nol
ogies
,
Abde
lm
alek Essaadi
Unive
r
sity,
Route Ziate
n
,
B.P.
416 Tangi
e
r, M
o
rocc
o.
Em
a
il: g
i
b
e
t.tan
i
.h
ich
a
m
@
g
m
ail.co
m
1.
INTRODUCTION
No
wa
day
s
, C
l
ou
d C
o
m
put
i
n
g i
s
o
n
e
of t
h
e
ne
west
para
di
gm
s i
n
t
h
e I
n
f
o
rm
at
i
on Tec
h
nol
ogy
(I
T)
wo
rl
d
.
Thi
s
n
e
w m
odel
of
com
put
i
ng c
onsi
s
t
s
o
n
o
f
f
e
ri
n
g
IT
res
o
urces
(Se
r
ve
rs
, St
o
r
age
,
Ne
t
w
o
r
k
,
Ap
pl
i
cat
i
ons
…
)
as se
r
v
i
ces,
o
n
dem
a
nd
and
o
v
e
r
a
n
e
t
w
o
r
k
[
1
]
.
C
l
ou
d C
o
m
put
i
n
g
IT
m
odel
foc
u
ses
p
r
im
arily o
n
th
e flex
i
b
ility
an
d
t
h
e celerity o
f
IT
reso
urces allo
catio
n
,
wh
ich
lib
erat
e th
e end
u
s
ers fro
m
co
n
c
ern
s
ab
ou
t
th
e
I
T
inf
r
a
stru
ctur
e an
d lo
catio
n
,
and
all
o
f
th
is pr
esen
ted
i
n
a
p
a
y-
as-
you-
go
m
a
n
n
e
r.
In
o
r
de
r t
o
i
n
su
re t
h
at
C
l
ou
d C
o
m
put
i
n
g
of
feri
ngs
a
n
d
cha
r
act
eri
s
t
i
c
s are at
t
h
e hei
g
ht
of
expect
at
i
o
ns i
n
t
h
i
s
ne
w m
odel
of c
o
m
put
i
n
g
,
pe
rf
orm
a
nce and al
l
o
cat
i
on
p
o
l
i
c
i
e
s eval
uat
i
on i
s
nec
e
ssary
bef
o
re
a
n
y
rea
l
wo
rl
d depl
oy
m
e
nt
. W
h
i
l
e
u
s
i
ng real
i
n
f
r
as
t
r
uct
u
res f
o
r t
e
st
i
ng a
nd as
ses
s
m
e
nt
i
s
expen
s
i
v
e
and t
i
m
e-cons
um
i
ng, t
hus i
t
i
s
not
al
way
s
pr
om
i
s
i
ng t
o
per
f
o
rm
experi
m
e
nt
al
, repeat
abl
e
an
d scal
abl
e
i
nvest
i
g
at
i
o
ns
on
real
w
o
rl
d
c
l
ass C
l
ou
d e
n
v
i
ro
nm
ent
s
.
A sol
u
t
i
on t
o
t
h
i
s
pr
o
b
l
e
m
i
s
t
h
e use o
f
si
m
u
l
a
t
i
on t
ool
s
i
n
pu
rp
ose
of
eval
uat
i
n
g an
d t
e
st
i
ng t
h
e
cl
ou
d-c
o
m
put
i
ng m
odel
i
n
a
cont
rol
l
e
d a
n
d scal
abl
e
en
vi
ro
nm
ent
,
t
h
eref
o
r
e ge
nerat
i
n
g s
p
eci
fi
c res
u
l
t
s
base
d
on s
p
ecific m
e
asurem
ents. In the sa
m
e
context, Clou
d
S
im is an
in
n
o
v
a
tiv
e and
co
m
p
reh
e
nsiv
e sim
u
l
a
tio
n
fram
e
wor
k
t
h
a
t
sup
p
o
r
t
s
m
odel
i
ng, si
m
u
l
a
t
i
on a
n
d ex
pe
ri
m
e
nt
at
i
on o
f
C
l
ou
d com
put
i
ng i
n
f
r
ast
r
uct
u
res a
n
d
ap
p
lication
serv
ices [2
].
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
C
l
ou
d C
o
m
put
i
n
g
C
P
U
Al
l
o
c
a
t
i
on
an
d
Sc
hed
u
l
i
n
g
Al
g
o
ri
t
h
ms
Usi
n
g C
l
o
u
dSi
m .
...
(
G
i
b
et
Ta
ni
Hi
ch
am)
1
867
2.
CLOU
D CO
MP
UTIN
G
SI
MUL
A
TIO
N
A C
l
ou
d C
o
m
put
i
ng of
fer
rang
es fr
om
pro
p
o
si
n
g
a
speci
fi
c IT i
n
fra
st
ruct
ure
t
o
depl
oy
i
n
g
com
p
l
i
cat
ed appl
i
cat
i
o
ns an
d so
ft
ware s
o
l
u
t
i
o
n
s
. B
y
st
udy
i
ng t
h
e C
l
o
ud C
o
m
put
i
n
g
servi
ce del
i
v
e
r
y
m
odel
ori
g
i
n
at
es t
h
e
chal
l
e
ng
e o
f
m
a
nagi
ng
h
u
n
d
r
eds
of
t
h
ous
an
ds
of
use
r
s a
n
d a
ppl
i
cat
i
o
ns
req
u
est
s
.
The
r
efo
r
e, a
Clo
u
d
Co
m
p
u
t
in
g
p
r
ov
id
er sho
u
l
d
co
n
s
i
d
er in
tellig
en
t infrastru
cture
d
e
p
l
oy
m
e
n
t
in
o
r
d
e
r to
estab
lish
a
Clo
u
d
Co
m
p
u
tin
g
o
ffer,
wh
ich
i
n
sures transp
aren
cy, scalab
ility, secu
rity and
fore
m
o
st celerity (Qo
S
) [3
].
C
l
ou
d C
o
m
put
i
ng assessm
ent
and e
v
al
ua
t
i
on i
s
m
a
ndat
o
ry
fo
r b
o
t
h
C
l
oud p
r
o
v
i
d
ers t
h
at
are
planning a specific service delivery
and
Clo
u
d
u
s
ers who
are in
tend
ing
to
sh
ift th
eir IT infrast
ru
ct
u
r
e,
pl
at
fo
rm
or s
o
ft
ware
i
n
t
o
t
h
e
C
l
ou
d
(I
nt
er
n
e
t
i
n
case
o
f
p
ubl
i
c
cl
ou
d
o
r
I
n
t
r
a
n
et
i
n
ca
se o
f
pri
v
at
e c
l
ou
d)
.
Desp
ite t
h
e
fact th
at u
s
i
n
g real in
frastru
c
tu
re fo
r testin
g and
ev
alu
a
ting
cloud
d
e
p
l
oy
m
e
n
t
can
g
i
ve th
e
i
nvest
i
g
at
ors a
real
w
o
rl
d a
p
p
r
oach t
o
m
a
ke cri
t
i
cal
dec
i
si
on a
b
o
u
t
m
ovi
ng
f
o
r
w
ar
d
wi
t
h
t
h
i
s
m
odel
o
f
com
puting, in
m
o
st cases it can
be
very e
xpensive:
In
frast
ruct
ures
,
platf
o
rm
s and
soft
ware
hi
gh
costs
Necessity to
test on
scalab
le e
nvi
ro
nm
ent
s
(
m
ore i
n
frast
ruc
t
ure)
M
a
nagem
e
nt
and
m
a
i
n
t
e
nanc
e ex
pen
s
es
In
ad
d
ition
to th
is critical f
acto
r
,
we can ad
d
th
e tim
e
co
n
s
u
m
p
tio
n
in
o
r
d
e
r t
o
test a sp
ecific
scenari
o
:
In
frast
ruct
ures
Installations
an
d c
o
n
f
ig
urati
o
n
s
Repeatable a
n
d va
riable tests
D
e
bug
g
i
n
g
and tr
oub
lesho
o
ting
A m
o
re v
i
ab
le altern
ativ
e is th
e u
s
e of si
m
u
la
tio
n
to
o
l
s. Th
ese too
l
s
o
p
e
n
u
p
th
e
po
ssib
ility o
f
ev
alu
a
ting
th
e
h
ypo
th
esis
(
a
pp
licatio
n
b
e
n
c
h
m
ar
k
i
n
g
study)
in
a co
n
t
r
o
ll
ed
en
v
i
ron
m
en
t w
h
er
e
o
n
e
can
easily
rep
r
o
d
u
ce resu
lts
[2]
.
Th
ere are
v
a
ri
ety o
f
sim
u
lat
o
r too
l
s for mo
d
e
lling
and
si
m
u
la
tio
n
of larg
e-scale Clo
u
d
co
m
p
u
ting
envi
ro
nm
ent
s
[4]
(fi
gu
re 1
)
.
Gene
ral
l
y
, we can desi
gna
te
b
e
tween
two
typ
e
s o
f
sim
u
lato
rs: graph
i
cal u
s
er
i
n
t
e
rface
(
G
U
I
) si
m
u
l
a
t
o
rs o
r
pr
o
g
ram
m
i
ng l
a
ng
ua
ge
base
d
sim
u
l
a
t
o
rs (l
i
k
e Java
f
o
r
exa
m
pl
e).
Fi
gu
re
1.
C
l
ou
d C
o
m
put
i
ng
S
i
m
u
l
a
t
i
on f
r
am
ewo
r
ks
3.
CLOU
D CO
MP
UTIN
G
SI
MUL
A
TO
RS EVAL
UATI
O
N
As de
scri
bed
on t
h
e p
r
e
v
i
o
u
s
sect
i
on, t
h
er
e are a di
versi
t
y
of C
l
ou
d C
o
m
put
i
ng si
m
u
l
a
t
o
rs
, eac
h
with s
p
ecific
characte
r
istics and oriente
d
for a s
p
ecifi
c
ob
ject
i
v
e
.
C
h
o
o
si
n
g
t
h
e
fi
ne
st
C
l
ou
d C
o
m
put
i
n
g
si
m
u
lato
r is a
ch
allen
g
i
ng
m
i
ssio
n
. To
th
e
best o
f
ou
r
kn
owledg
e, it was n
o
t
v
e
ry
d
i
fficu
lt to
spo
t
Clou
dSim
as th
e co
re
p
l
atfo
rm
for th
e
m
o
st u
s
ed
Clou
d sim
u
lato
rs
u
p
to
t
h
is m
o
men
t
. Clo
u
d
S
i
m
was estab
lish
e
d as an
ex
ten
s
i
o
n of the Gri
d
Sim
si
mu
lato
r in
o
r
d
e
r to
in
trod
u
c
e th
e Cloud
Co
m
p
u
ting
v
i
rt
u
a
lizatio
n
layer t
h
at was
n
o
t
presen
t on
th
e
orig
i
n
al
si
m
u
la
to
r.
C
l
ou
dSi
m
i
s
a pro
g
r
am
m
i
ng l
a
ng
ua
ge ba
sed si
m
u
l
a
t
o
r
and e
v
en t
h
ou
gh i
t
does n
o
t
sup
p
o
r
t
a
gra
p
hi
cal
use
r
i
n
t
e
rface
f
o
r
si
m
u
l
a
t
i
on, i
t
pr
op
oses
t
h
e
Clou
dAn
a
lyst (
w
hich
is an
ex
tensio
n of
Cloud
Si
m
)
f
o
r
i
nvest
i
g
at
ors
wh
o
pre
f
er
s u
s
i
ng a
use
r
-
fri
endl
y
i
n
t
e
r
f
ace
t
o
car
ry
o
u
t
t
h
ei
r
researc
h
es
. C
l
ou
dSi
m
pr
esent
s
itself to the cloud-c
o
m
puting
researc
h
ers as
a Java ba
sed fr
am
ewor
k
t
h
at
sup
p
o
rt
s
the main cha
r
acteristics of
C
l
ou
d C
o
m
put
i
ng (
I
aaS
) wi
t
h
vi
rt
ual
i
zat
i
on su
p
p
o
r
t
and
t
a
sk sche
dul
i
n
g (Paa
S an
d S
aaS) an
d o
p
e
n
up t
h
e
do
o
r
fo
r em
er
gi
n
g
, i
n
t
e
g
r
at
i
ng a
nd t
e
st
i
n
g ne
w al
go
rith
m
s
fo
r task
sch
e
d
u
ling
or n
e
w ch
aracteristics
devel
opm
ent
,
whi
c
h hel
p
e
d
o
n
del
i
v
eri
n
g
ne
w
si
m
u
l
a
t
o
rs.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE
Vo
l. 6
,
N
o
. 4
,
Au
gu
st 2
016
:
18
66
–
1
879
1
868
Ou
r c
hoi
ce f
o
r
C
l
ou
dSi
m
st
em
s from
i
t
openne
ss an
d cl
e
a
r l
o
gi
c w
h
i
c
h
i
s
defi
ci
ent
o
n
t
h
e
ot
he
r
sim
u
l
a
t
o
rs spe
c
i
f
i
cal
l
y
wi
t
h
GU
I base
d si
m
u
l
a
t
o
rs
whe
r
e we were
not
a
b
l
e
t
o
t
ackl
e
t
h
e C
l
oud i
n
f
r
ast
r
uct
u
re
layer to
b
e
tter test alg
o
r
ith
ms related
to
C
P
U allo
cation
an
d
i
n
tro
d
u
ce
n
e
w al
g
o
rith
ms. Clo
u
d
S
im
was th
e
accurate selection for
our re
search, which is focu
se
d on evaluating a
nd asses
s
ing the CPU sche
duling
al
go
ri
t
h
m
s
for
reso
u
r
ces al
l
o
c
a
t
i
on i
n
o
r
de
r
t
o
hel
p
C
l
o
ud
pr
o
v
i
d
er
s an
d
users
m
a
ke pr
eci
se deci
si
o
n
abo
u
t
C
l
ou
d C
o
m
put
i
ng m
odel
a
d
o
p
t
i
o
n
.
4.
CLOU
DS
IM
AR
CHITE
C
T
URE A
N
D
D
E
SIGN
Clo
u
d
S
im
is a
Jav
a
app
licatio
n
th
at was fo
un
d
e
d
on
GridSi
m
,
wh
ich
is a si
m
u
lato
r and
a to
o
l
k
it for
m
o
d
e
lin
g
and
si
m
u
latio
n
of en
tities in
p
a
rallel an
d d
i
stri
b
u
t
ed
co
m
p
u
tin
g
[5
],[6
]. Clou
dSim
[5
] was
d
e
signed
i
n
a l
a
y
e
re
d a
r
chi
t
ect
ure a
s
s
h
o
w
e
d
on
fi
gu
re
2.
At
t
h
e lowest layer, we
find t
h
e “
S
imJava”
(Disc
r
et
e Eve
n
t
Si
m
u
latio
n
)
wh
ich
im
p
l
e
m
en
ts th
e co
re
functio
n
a
lities n
eed
ed
b
y
th
e
h
i
gh
er lev
e
l o
f
si
m
u
la
tio
n
(Data Cen
t
er,
Ho
st, Vi
rtu
a
l
mach
in
e…). Ju
st abov
e th
e
Si
m
J
av
a we
fi
n
d
th
e “Gri
d
S
i
m
” to
o
l
k
it fo
r
m
o
d
e
lin
g
m
u
lt
ip
le Gri
d
in
frastru
ct
u
r
es, in
clud
ing
n
e
tworks an
d
asso
ciated
traffic. At
th
e n
e
x
t
layer, we find
th
e Clo
u
d
S
im
si
mu
latio
n
l
a
y
e
r, w
h
i
c
h
pr
o
v
i
d
es s
u
p
p
o
rt
f
o
r m
odel
i
ng a
nd si
m
u
l
a
t
i
on o
f
vi
rt
ual
i
zed C
l
o
u
d
-
base
d
dat
a
cent
e
r
envi
ronm
ents including
de
dicated m
a
nagement
interfaces
for VMs,
m
e
m
o
ry
, stora
g
e, and
ba
ndwidt
h.
This
l
a
y
e
r han
d
l
e
s t
h
e f
u
ndam
e
nt
al
i
ssues, s
u
ch a
s
pr
o
v
i
s
i
o
ni
n
g
of
ho
st
s t
o
VM
s, m
a
nagi
ng a
p
pl
i
cat
i
on e
x
ecu
t
i
o
n
,
an
d
m
o
n
ito
ri
ng
d
y
n
a
m
i
c
sys
t
e
m
sta
t
e [2
]. Th
e top
layer in
th
e Clo
u
d
S
i
m
si
m
u
la
tio
n
to
o
l
k
it is th
e “User
Code”
which i
s
the m
a
in inte
rface
for sim
u
lation s
p
ecifi
cations a
n
d cha
r
a
c
teristics configuration (num
ber
of
machines, a
p
plications, tas
k
s,
users
,
sc
he
duli
n
g policies and their
basic st
ructure).
Fi
gu
re
2.
C
l
ou
dSi
m
Archi
t
ect
ure
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
C
l
ou
d C
o
m
put
i
n
g
C
P
U
Al
l
o
c
a
t
i
on
an
d
Sc
hed
u
l
i
n
g
Al
g
o
ri
t
h
ms
Usi
n
g C
l
o
u
dSi
m .
...
(
G
i
b
et
Ta
ni
Hi
ch
am)
1
869
5.
C
P
U
SC
HEDU
LING A
L
GOR
I
THM
S
A clo
u
d
pro
v
i
d
e
r m
a
in
in
trest is to
in
crease p
r
o
f
its
by
achi
e
vi
n
g
hi
g
h
l
e
vel
s
of
use
r
s’
sat
i
s
fact
i
o
n
,
wh
ich
co
m
e
s with
prov
i
d
ing th
e end
u
s
er with
th
e b
e
st
experience.
He
nce
,
com
e
s the importance
of c
h
oosi
n
g
t
h
e fi
nest
sc
he
dul
i
n
g al
g
o
r
i
t
h
m
s
for
reso
u
r
c
e
s al
l
o
cat
i
on a
nd t
a
s
k
sc
hed
u
l
i
ng. T
h
e
key
pu
r
pose
of
sch
e
dul
i
n
g
algorithm
s
is
the appropriate
allocati
on o
f
t
a
sk o
r
a j
o
b t
o
t
h
e ap
pr
op
ri
at
e reso
urce
. T
h
eref
ore
,
i
t
goes
bac
k
always to the period necessa
ry to carry
out t
h
e exec
ution of a specific tas
k
in
ord
e
r to
evalu
a
te th
e q
u
a
li
ty an
d
per
f
o
r
m
a
nce of
t
h
e sc
he
dul
i
n
g
al
go
ri
t
h
m
s
[3]
,
[
7
]
.
5.
1.
First
Come
, F
i
rst Ser
v
ed
(F
CFS
)
Sc
heduli
n
g
Th
e sim
p
lest [8
]-[10
]
, algo
rit
h
m
fo
r
resources sch
e
du
lin
g is th
e “FCFS”
alg
o
rith
m
(It is also
called
FIFO=First In
First Ou
t). Th
i
s
alg
o
r
ith
m
is
b
a
sed
o
n
th
e a
rrival tim
e
of the res
o
urce re
que
st. To clarify the
p
e
rform
a
n
ce o
f
th
e “FC
FS” al
g
o
rith
m
,
let tak
e
th
e
fo
llowing Ex
am
p
l
e:
Tab
l
e 1
.
First Ex
am
p
l
e
Task
s
list
Task
O
r
der
Task
Na
m
e
Burst Ti
m
e
1 T
a
sk-
1
10
2 T
a
sk-
2
4
3 T
a
sk-
3
5
4 T
a
sk-
4
20
Th
e ex
ecu
tio
n
sch
e
d
u
l
e
will be as fo
llow:
Tabl
e 2. Fi
rst
Exam
pl
e
Gant
t
C
h
art
Task
-1
Task
-2
Task
-3
Task
-4
0
10
14
19
39
The
waiting time for each task
will be the following:
Tab
l
e 3
.
First Ex
am
p
l
e
Task
s W
a
itin
g
Tim
e
Task
O
r
der
Task
Na
m
e
Waiting Ti
m
e
1 T
a
sk-
1
0
2 T
a
sk-
2
10
3 T
a
sk-
3
14
4 T
a
sk-
4
19
Th
e av
erag
e waitin
g
tim
e will
b
e
calcu
lated
as fo
llo
w:
Aw
T = (
∑
Tn) / NT
Whe
r
eas:
AwT:
Av
erag
e
W
a
itin
g Tim
e
Tn
: Task
s
Wai
tin
g
tim
e fo
r ex
ecu
tion
NT: Num
b
er o
f
task
s
C
onse
q
uent
l
y
,
t
h
e ave
r
age
w
a
i
t
i
ng t
i
m
e
wi
ll
be:
10
,7
5.
No
w, let ch
ang
e
th
e task
arrival o
r
d
e
r t
o
match
th
e fo
llowing
:
Table
4.
Second E
x
am
ple Tasks list
Task
O
r
der
Task
Na
m
e
Burst Ti
m
e
1 T
a
sk-
2
4
2 T
a
sk-
3
5
3 T
a
sk-
1
10
4 T
a
sk-
4
20
Th
e ex
ecu
tio
n
sch
e
d
u
l
e
will be as fo
llow:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE
Vo
l. 6
,
N
o
. 4
,
Au
gu
st 2
016
:
18
66
–
1
879
1
870
Table
5.
Second E
x
am
ple Ga
ntt Chart
Task
-2
Task
-3
Task
-1
Task
-4
0
4
9
19
39
The
waiting time for each task
will be the following:
Tab
l
e 6
.
Second
Ex
am
p
l
e
Task
s Waitin
g
Ti
me
Task
O
r
der
Task
Na
m
e
Waiting Ti
m
e
1 T
a
sk-
2
0
2 T
a
sk-
3
4
3 T
a
sk-
1
9
4 T
a
sk-
4
19
Th
e av
erag
e waitin
g
ti
m
e
will
b
e
:
8.
From
the two exam
ple
s
above,
we
ca
n co
ncl
u
de t
h
at
t
a
sks o
r
de
r
ch
ang
e
d
m
a
ssiv
e
ly th
e av
erag
e
waitin
g time for task
ex
ecu
tion
.
Consequ
e
n
tly, th
e “FCFS” presen
ts a
weak
ness
o
n
t
a
sks a
nd
res
o
u
r
ces al
l
o
cat
i
on s
c
hed
u
l
i
n
g,
bec
a
use i
f
a
heavy
t
a
sk t
a
kes
on t
h
e l
ead
of t
h
e
que
ue
list, all th
e o
t
her sm
all task
s will h
a
v
e
to wait u
n
til th
e ex
ecu
tio
n end
fo
r t
h
e lead
i
n
g tasks.
Thi
s
basi
c sc
h
e
dul
i
n
g al
g
o
ri
t
h
m
gave bi
rt
h
t
o
a new al
g
o
r
i
t
h
m
cal
l
e
d “
Shorte
d
Job Firs
t
” or “
SJF
”
whi
c
h we
reco
m
m
e
nd on t
h
i
s
pape
r i
n
or
der
t
o
en
hance t
h
e C
l
ou
dSi
m
algo
ri
t
h
m
based
on t
h
e “FC
FS”
. The
co
n
c
ep
t [8
],[11], o
f
t
h
is n
e
w alg
o
rith
m
is th
e sam
e
as fo
r “FCFS”,
ho
wever, th
e “
SJ
F
” alg
o
rith
m
in
tro
d
u
ce a
test o
n
t
h
e
b
e
g
i
n
n
i
n
g
to
cho
o
s
e th
e task
with
th
e shortest execu
tio
n ti
m
e
in
o
r
d
e
r to tak
e
t
h
e lead of the
q
u
e
u
e
,
whe
r
eas t
h
i
s
or
deri
ng
o
f
t
a
s
k
s
red
u
ce
d e
x
t
r
e
m
el
y
t
h
e avera
g
e
wai
t
i
ng t
i
m
e (Sec
o
n
d
Exa
m
pl
e above
).
5.
2.
Round
Robin
(RR) Sche
duling
The “R
o
u
n
d
R
obi
n” al
go
ri
t
h
m
[11]
-[
1
3
]
,
w
a
s desi
g
n
ed
ba
sed o
n
t
h
e di
st
ri
b
u
t
i
on
of t
h
e
C
P
U t
i
m
e
a
m
o
n
g
th
e sched
u
l
ed
task
s.
On th
e sam
e
co
n
t
ex
t, all
th
e
task
s
g
e
t on
a
q
u
e
u
e
list wh
ereas each
task
g
e
t a
sm
al
l
uni
t
of C
P
U t
i
m
e
(Qua
n
t
um
, usual
l
y
10-
1
00 m
i
ll
i
s
econ
ds
). I
n
o
r
de
r
t
o
deepe
n
[
10]
t
h
e un
der
s
t
a
n
d
i
ng o
f
the “RR” algorith
m
,
let take the sam
e
example as
before:
Tabl
e
7. T
h
i
r
d
Exam
pl
e Tasks
Li
st
Task
O
r
der
Task
Na
m
e
Ti
m
e
for Execu
ti
on
1 T
a
sk-
1
10
2 T
a
sk-
2
4
3 T
a
sk-
3
5
4 T
a
sk-
4
20
If
we con
s
id
er th
e “RR” algorith
m
with
a static
C
P
U t
i
m
e
qua
nt
um
of
“4
”, t
h
e e
x
ec
ut
i
o
n sc
hed
u
l
e
will b
e
as fo
llow:
Tabl
e
8. T
h
i
r
d
Exam
pl
e Gant
t
C
h
art
Task
-1
Task
-2
Task
-3
Task
-4
Task
-1
Task
-3
Ta
sk
-4
Ta
sk
-1
Ta
sk
-4
Ta
sk
-4
Ta
sk
-4
0
4
8
12
16
20
21
25
27
31
35
39
The
waiting time for each task
will be the following:
Tab
l
e 9
.
Th
ird
Ex
am
p
l
e
Task
s W
a
itin
g
Tim
e
Task
O
r
der
Task
Na
m
e
Waiting Ti
m
e
1
T
a
sk-
1
0+(
16-
4)
+(
25-
20)
= 17
2 T
a
sk-
2
4
3
T
a
sk-
3
8+(
20-
12)
= 16
4
T
a
sk-
4
12+(
21-
16)
+(
27-
2
5
)
= 19
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
C
l
ou
d C
o
m
put
i
n
g
C
P
U
Al
l
o
c
a
t
i
on
an
d
Sc
hed
u
l
i
n
g
Al
g
o
ri
t
h
ms
Usi
n
g C
l
o
u
dSi
m .
...
(
G
i
b
et
Ta
ni
Hi
ch
am)
1
871
The ave
r
a
g
e w
a
i
t
i
ng t
i
m
e
wi
ll
be ((
17 + 4
+ 16 + 1
9
) /
4
)
=
14
.
O
n
C
l
ou
dSi
m
, The R
o
u
n
d
R
obi
n
b
a
sed
algo
rit
h
m
u
s
ed
fo
r tasks sch
e
du
lin
g acts as th
e
fo
llo
win
g
:
So
rti
n
g
th
e task
s sub
m
itted
t
o
th
e VM in
ascend
i
ng
way b
a
sed
on
th
e time n
eed
ed
for
ex
ecu
tion
o
f
each
task (B
urst Tim
e
)
Ex
tracting
a st
atic ti
me Qu
antu
m
wh
ich
is calcu
l
ated
b
a
sed
on
t
h
e nu
m
b
er
of
in
st
ru
ction
s
th
e
pro
cesso
r
can e
x
ecute
pe
r sec
o
nd:
TQ = (NP *
MI
PS) / 1000
Whe
r
eas:
TQ =>
Tim
e
Quant
u
m
NP =>
Num
b
er
of
Process
o
rs
MIPS => Milli
o
n
In
stru
ction
p
e
r Seco
nd
All task
s are
qu
eu
ed
on
th
e read
y list for ex
ecu
tio
n
Fo
r each
task on
th
e qu
eu
e list:
o
Th
e CPU allo
cates th
e tim
e q
u
a
n
t
u
m
fo
r t
h
e
task
ex
ecu
tion
o
If t
h
e task
is execu
ted
,
it is sen
t
to
t
h
e
fin
i
sh
q
u
e
u
e
list
o
Else th
e task
is sen
t
t
o
th
e wai
tin
g
list
C
onse
q
uent
l
y
, once t
h
e
num
ber of a t
a
sk a
n
d am
ount
of i
n
st
ruct
i
o
ns get
hi
g
h
er
, t
h
e t
i
m
e
qua
nt
um
b
eco
m
e
s sm
al
l
e
r an
d th
e
waitin
g
list
g
e
ts long
er,
h
e
n
c
e, a
h
i
g
h
av
erag
e
waitin
g
tim
e.
The R
o
un
d R
o
bi
n base
d al
g
o
r
i
t
h
m
we are pro
p
o
si
n
g
o
n
t
h
i
s
paper t
o
s
o
l
v
e t
h
e pr
o
b
l
e
m
rel
a
t
e
d t
o
sm
al
l
e
r t
i
m
e
quant
um
wor
k
s
wi
t
h
a
dy
nam
i
c t
i
m
e
qua
nt
u
m
[14]
an
d t
h
i
s
i
s
as
fol
l
o
wi
n
g
:
So
rti
n
g
th
e task
s sub
m
itted
t
o
th
e VM in
ascend
i
ng
way b
a
sed
on
th
e time n
eed
ed
for
ex
ecu
tion
o
f
each
task (B
urst Tim
e
)
Ex
tracting
a time Qu
an
tu
m
wh
ich
is calcu
lated
b
a
sed as t
h
e
Av
erag
e Execu
tio
n tim
e fo
r tasks:
TQ =
∑
Tn
/ N
T
Whe
r
eas:
TQ =>
Tim
e
Quant
u
m
Tn =>
Tim
e
needed for exec
uti
ng a
specific
task
(Burst Tim
e
)
N
T
=> Nu
m
b
e
r
o
f
ta
sk
s
All task
s are
qu
eu
ed
on
th
e read
y list for ex
ecu
tio
n
Fo
r each
task on
th
e qu
eu
e list
If t
h
e task
b
u
rst ti
m
e
is s
m
al
le
r th
an
t
h
e tim
e
q
u
a
n
t
u
m
o
Th
e Tim
e
Qu
an
tu
m
is set to
t
h
e B
u
rst Tim
e
o
f
th
e task
o
Th
e CPU allo
cates th
e tim
e q
u
a
n
t
u
m
fo
r t
h
e
task
ex
ecu
tion
o
Th
e task
is ex
ecu
t
ed
and
sen
t
to
th
e
fin
i
sh
list
Else
o
Th
e CPU allo
cates th
e tim
e q
u
a
n
t
u
m
fo
r t
h
e
task
ex
ecu
tion
o
Th
e task
is sent to
th
e
waiting list
If t
h
e
waitin
g list is n
o
t
em
p
t
y
o
Send
task
s
from
wait
in
g
list to
th
e read
y list
o
R
e
st
art
fr
om
t
h
e be
gi
n
n
i
n
g
Here is an illu
stratio
n
o
f
th
e alg
o
rith
m
step
s flo
w
:
First
, all the
proces
ses are s
o
rte
d
in
ascending
way base
d
o
n
th
e
bu
rst
ti
m
e
o
f
task
s
(th
e
tim
e n
eeded
for
execution
of ea
ch tas
k
) a
n
d se
nt to t
h
e
ready
que
ue list
nt
nu
m
b
er
of
tasks
i
co
un
te
r
While (RQ != NUL
L)
TQ =
∑
Tn
/
NT
// NT = To
tal
nu
m
b
er of task
s
// Tn = Tim
e
needed fo
r th
e t
a
sk
ex
ecu
tio
n
//RQ = Ready
Que
u
e
//TQ = Tim
e
Quant
u
m
Second
,
A
ssi
gn
TQ
t
o
(1
t
o
n)
task on
t
h
e
r
e
ad
y qu
eu
e list
fo
r i =
1 t
o
nt
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE
Vo
l. 6
,
N
o
. 4
,
Au
gu
st 2
016
:
18
66
–
1
879
1
872
{
if (Ti < TQ)
T
Q
Ti
Ti
TQ
Send task t
o
fi
nish list
i
i +1
else
Ti
TQ
Send task
t
o
waitin
g qu
eu
e list
end if
}
en
d fo
r
// Assign
TQ t
o
all th
e
av
ailab
l
e task
s.
Third
, Test if th
e
waitin
g
qu
eu
e list is em
p
t
y
if (waitin
g
list
!= em
p
t
y)
Send
task
s
from
wait
in
g
list to
read
y list
Go
to ste
p
1
else
Finish
end if
Let con
s
id
er the fo
llo
wi
n
g
task
s
b
u
rst tim
e:
Tabl
e
10
. F
o
ur
t
h
E
x
am
pl
e Tasks Li
st
Task
s
Arrival Ti
m
e
Burst Ti
m
e
T1
0 40
T2
0 90
T3
0 20
T4
0 83
The al
gorithm
we a
r
e
proposi
n
g arranges
tasks
i
n
ascen
d
i
ng
w
a
y b
a
sed on
th
ei
r
b
u
r
s
t time:
Table
11. Tas
k
s List Arra
nge
m
ent
Task
s
Arrival Ti
m
e
Burst Ti
m
e
T3
0 20
T1
0 40
T4
0 83
T2
0 90
The t
i
m
e quant
um
i
s
cal
culat
e
d base
d o
n
t
a
sks am
ount
and t
h
ei
r b
u
r
st
t
i
m
e
:
T
Q
= 58
,2
5.
Th
e
ex
ecu
tion
sch
e
d
u
l
e will
b
e
as fo
llow:
Tabl
e
12
. F
o
ur
t
h
E
x
am
pl
e Gant
t
C
h
art
Tim
e
Q
u
ant
u
m
of 58,
25
T3 T1
T4
T2
0
20
60
118,25
176,
5
The tasks T
3
and T
1
are exe
c
uted (B
urst T
i
m
e
is s
m
a
ller
th
an
th
e Tim
e
Qu
an
tu
m
)
. The alg
o
r
ith
m
send
s th
e rem
a
in
in
g task
s to
t
h
e
waitin
g
list:
Tabl
e
13
. F
o
ur
t
h
E
x
am
pl
e Wai
t
i
ng Li
st
Task
s
Burst
Ti
m
e
T4
24,
75
T2
31,
75
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
C
l
ou
d C
o
m
put
i
n
g
C
P
U
Al
l
o
c
a
t
i
on
an
d
Sc
hed
u
l
i
n
g
Al
g
o
ri
t
h
ms
Usi
n
g C
l
o
u
dSi
m .
...
(
G
i
b
et
Ta
ni
Hi
ch
am)
1
873
Dynam
i
cally, the tim
e
quant
um will be recalculated as desc
ribe
d before:
T
Q
= 28,
2
5
.
The execution
sch
e
d
u
l
e
will be as fo
llow:
Table
14. Fourth E
x
am
ple Gantt cha
r
t After
Tim
e
Quatum
Recalculation
Tim
e
Q
u
ant
u
m
of 28,
25
T4 T2
176,5
201,25
229,5
The
Task T
4
i
s
exec
uted (B
urst Tim
e
is s
m
alle
r th
an th
e
Ti
m
e
Qu
an
tum
)
. Th
e algorith
m
sen
d
s
t
h
e
rem
a
in
in
g
task to
th
e waiting
list:
Tabl
e
15
. F
o
ur
t
h
E
x
am
pl
e Wai
t
i
ng Li
st
Task
s
Burst
Ti
m
e
T2 3,
5
Dyn
a
m
i
call
y
, t
h
e ti
m
e
q
u
a
n
t
um will b
e
set t
o
th
e burst ti
me o
f
task
T2
: TQ = 3,
5
. T
h
e last execution
sch
e
d
u
l
e
will be as fo
llow:
Table
16. Fourth E
x
am
ple Last Ga
ntt chart
Tim
e
Q
u
ant
u
m
of 3,
5
T2
229,
5
233
Let calculate the a
v
era
g
e
wai
ting tim
e of ea
ch tas
k
:
Tab
l
e
1
7
. Fou
r
th
Ex
am
p
l
e Task
s
Waitin
g Ti
me
Task
s
Burst Ti
m
e
Waiting Ti
m
e
T3
20
0
T1
40
20
T4
83
(
20+40)
+(
176,
5– 1
18,
25)
= 118,
25
T2
90
(
20+40+58,
2
5
)
+
(
201,
25–
176,
5)
=
143
Th
erefo
r
e, Th
e av
erag
e
waitin
g
tim
e
fo
r all task
s is:
70,3
1
.
Using
th
e same set
tin
g
s
wi
th
Clo
u
d
Si
m
g
a
v
e
us th
e fo
llo
wi
n
g
resu
lts:
Tabl
e
18
. C
l
o
u
d
Si
m
Out
p
ut
Task
s
Arrival
Ti
m
e
Burst
Ti
m
e
Finish
Ti
m
e
T3
0 20
80
T1
0 40
140
T4
0 83
226
T2
0 90
233
Fro
m
th
is ou
tpu
t
,
we ex
tracted
th
e av
erag
e
waitin
g
tim
e as fo
llo
w:
Aw
T = (
∑
Aw
Tn) /
NC
AwTn = FTn
–
EnT
Whe
r
eas:
AwT:
Av
erag
e
W
a
itin
g Tim
e
AwTn
:
Cloud
let/task
(n)
Av
erag
e
W
a
iting
Ti
m
e
NC: Num
b
er of Cloudlets
En
T: Cloud
let/task
(n) Ex
ecu
t
i
o
n
Tim
e
(Bu
r
st Ti
m
e
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE
Vo
l. 6
,
N
o
. 4
,
Au
gu
st 2
016
:
18
66
–
1
879
1
874
FTn
:
Fin
i
sh Ti
me o
f
Cloud
let/task
ex
ecu
tion
NCI: Nu
m
b
er o
f
Clo
u
d
l
et/task
In
st
ru
ction
s
Co
n
s
equ
e
n
tly, Th
e av
erag
e
waitin
g
tim
e f
o
r all task
s is:
11
1,
5.
Th
e “RR” sch
e
du
ling
algo
rith
m
in
trodu
ced
a
new p
e
rcep
ti
o
n
o
f
sch
e
du
ling th
at d
i
d
no
t ex
ist for th
e “FCFS” alg
o
rithm an
d
wh
ich
is th
e
co
n
c
urren
t
ex
ecu
tio
n of task
s. By co
m
p
aring
th
e ex
am
p
l
es stated
b
e
fo
re,
we can
see th
at
th
e av
erag
e
waitin
g
t
i
m
e
i
s
hi
ghe
r
wi
t
h
t
h
e
“R
o
u
n
d
R
o
bi
n”
al
g
o
ri
t
h
m
used
o
n
C
l
ou
dSi
m
and
t
h
u
s
t
h
e
al
g
o
ri
t
h
m
we are
pr
o
posi
n
g
o
n
th
is
p
a
p
e
r is m
o
re in
terestin
g.
6.
CLOU
DS
IM
SCHE
DULI
N
G
POLI
CIES
Th
e
v
i
r
t
u
a
lizatio
n
techno
log
y
is on
e
of
th
e
f
und
amen
tal co
n
c
ep
ts of
Cloud Co
m
p
u
tin
g
i
n
fra
st
ruct
ures
.
C
l
oudSi
m
depl
oy
en
orm
o
u
s
l
y
t
h
e vi
rt
ual
i
zat
i
on t
echn
o
l
ogy
i
n
o
r
de
r t
o
sim
u
l
a
t
e
IaaS and
PaaS
pr
ovi
si
on
i
ng a
n
d t
o
use
i
t
as a base
f
o
r u
s
ers
’
a
ppl
i
c
at
i
ons e
x
ec
ut
i
o
n.
O
n
t
h
e sam
e
per
s
pect
i
v
e
c
o
m
e
s
t
h
e c
h
al
l
e
nge
of
de
pl
oy
i
n
g t
h
e
fi
nest
res
o
u
r
ces al
l
o
cat
i
o
n
an
d sc
he
dul
i
n
g al
g
o
r
i
t
h
m
.
For
exam
pl
e, o
n
e
(
1
)
Data cen
ter that co
n
s
ists
o
f
o
n
e
(1
)
ho
st with
two
(2
)
pr
o
cessi
ng
u
n
i
t
s
a
n
d
w
h
ere
t
h
e c
l
ou
d
user i
s
t
r
y
i
ng t
o
in
stan
tiate two (2)
v
i
rtu
a
l m
ach
in
es
with
t
w
o (2
)
pr
oces
sing
units eac
h. L
o
gically, there is
a se
pa
ration
betwee
n the two
virtual m
a
chines, but in
reality, each
virtual
m
achine is l
i
m
ited to t
h
e processi
ng
powe
r
of
fere
d by
t
h
e
phy
si
cal
host
,
t
h
eref
ore we
cann
o
t
i
n
st
ant
i
at
e bot
h vi
rt
ua
l
m
achi
n
e on t
h
e sam
e
host
at
t
h
e
sam
e
tim
e wi
t
hout
a
n
a
p
pr
o
p
ri
at
e sche
dul
i
n
g
al
go
ri
t
h
m
[2]
.
In re
fere
nce to this critical fact
or, C
l
o
u
d
S
i
m
prop
oses t
w
o l
e
vel
s
of re
s
o
u
r
ces al
l
o
cat
i
on
pol
i
c
i
e
s
b
a
sed
on
t
w
o basic sch
e
d
u
ling po
licies, wh
ich
are th
e tim
e
-
s
h
a
r
ed
and
sp
ac
e
-
s
h
ar
ed
a
llo
ca
tio
n
p
o
lic
ie
s
.
T
h
e
s
e
allo
catio
n
po
licies are im
p
l
emen
ted
d
u
ring th
e
v
i
rtual m
a
ch
in
es con
s
tructio
n
and
throu
gho
u
t
t
h
e applicatio
n
execut
i
o
n.
T
h
e
Space
-S
hare
d
p
o
l
i
c
y
and
Ti
m
e
-Share
d
p
o
l
i
c
i
e
s are de
pi
c
t
i
ons
of
t
h
e “
F
C
FS =
Fi
rst
C
o
m
e
Fir
s
t Ser
v
ed
” an
d “RR = R
o
un
d Rob
i
n
”
algor
ith
m
s
r
e
sp
ectiv
ely.
In order t
o
illustrate clearly th
e
c
o
ncept of each
alloca
tion policy [8],[15]
, we
propose
the
followi
ng
exam
ple:
O
n
e (1
) d
a
ta Cen
t
er
w
ith
On
e (1
)
ho
st
Th
e
ho
st h
a
s t
w
o (2
) pr
ocessin
g
un
its
Th
e
u
s
er in
stantiate two
(2
)
v
i
rtu
a
l m
ach
in
es
th
at req
u
i
re
o
n
e (1
)
pro
cessing
u
n
it each
Th
e
u
s
er th
en
t
r
y to
ex
ecu
t
e two (2
) task
s (Clo
ud
lets)
in e
ach
virtual m
achine
(
eac
h tas
k
requires
one
(1)
pr
ocessi
ng
u
n
i
t
f
o
r e
x
ec
ut
i
o
n
)
Figure 3 re
pre
s
ents a space-s
h
are
d
policy for both
virtual
m
achines and tasks.
While each virt
ual
mach
in
e requ
ires on
e
(1)
p
r
o
c
essin
g
un
it, each
v
i
rtu
a
l
m
ach
in
e
will reserv
e
o
n
e
(1
) of t
h
e two
(2
)
processin
g
u
n
its
o
f
th
e host, n
e
v
e
rth
e
less on
ly on
e (1
)
task
can
g
e
t execu
ted at a sp
ecific ti
m
e
an
d
th
e seco
nd on
e will
wait fo
r t
h
e
first task
to end
i
n
o
r
d
e
r to
g
e
t ex
ecu
ted
.
Figu
re
3.
S
p
ac
e Sha
r
e
d
P
o
licy
fo
r
VM
s a
n
d
Tasks
Fi
gu
re 4
pre
s
e
n
t
s
a space
-sha
red
pol
i
c
y
fo
r
vi
rt
ual
m
achi
n
es and t
i
m
e-shared
pol
i
c
y
fo
r
t
a
sks. Eac
h
virtual m
achine will reserve one (1) of the
two (2)
pr
ocessing units of t
h
e host,
and while each tasks needs
o
n
e
(1
) pro
cessin
g
un
it to
g
e
t
ex
ecu
t
ed, th
e
p
o
licy algo
rith
m
will give ea
ch tas
k
a
slice of the
processi
ng unit
ti
m
e
u
n
til b
o
t
h
task
s are ex
ecuted
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
C
l
ou
d C
o
m
put
i
n
g
C
P
U
Al
l
o
c
a
t
i
on
an
d
Sc
hed
u
l
i
n
g
Al
g
o
ri
t
h
ms
Usi
n
g C
l
o
u
dSi
m .
...
(
G
i
b
et
Ta
ni
Hi
ch
am)
1
875
F
i
g
u
r
e
4
.
Sp
a
c
e
S
h
ar
ed
Po
lic
y f
o
r
V
M
s
an
d
T
i
me
S
h
a
r
ed
fo
r T
a
sk
s
Fi
gu
re 5
pre
s
e
n
t
s
a t
i
m
e
-shar
e
d p
o
l
i
c
y
for
v
i
rt
ual
m
achi
n
e
s
and s
p
ace
-sh
a
red
pol
i
c
y
fo
r
t
a
sks. Eac
h
v
i
rtu
a
l m
ach
in
e g
e
ts a slice
o
f
th
e p
r
o
cessin
g
un
it ti
m
e
. Th
e first v
i
rtu
a
l
m
ach
in
e g
e
t to
h
o
l
d
th
e first
pr
ocessi
ng
u
n
i
t
of t
h
e
h
o
st
i
n
o
r
der t
o
e
x
e
c
ut
e t
h
e fi
rst
t
a
sk a
nd t
h
e sa
m
e
t
h
i
ng g
o
es
fo
r t
h
e sec
o
nd
vi
rt
ua
l
mach
in
e. As
a resu
lt, bo
th
first
task
s o
f
bo
t
h
v
i
rtu
a
l
m
ach
in
es g
e
t
to
run
si
m
u
ltan
e
ou
sly.
Th
e second
task
o
f
b
o
t
h
v
i
rtu
a
l m
ach
in
es will ho
l
d
u
n
til th
e
first
task
is ex
ecu
t
ed
for
b
o
t
h
VM
s.
Figure
5. Tim
e
Sha
r
e
d
Policy for
VMs a
n
d Space Sha
r
ed for Tas
k
s
Fi
gu
re 6
pres
ent
s
a t
i
m
e
-shared
pol
i
c
y
fo
r bot
h vi
rt
ual
m
achi
n
es an
d t
a
sks. T
h
e t
i
m
e
of bot
h
p
r
o
cessi
n
g
u
n
i
ts o
f
th
e
h
o
st will b
e
sh
ared si
m
u
ltan
e
o
u
s
l
y
b
y
th
e fou
r
task
s d
e
p
l
o
y
ed o
n
t
h
e two
v
i
rtu
a
l
machines.
Figu
re
6.
Tim
e
Sha
r
e
d
P
o
licy
fo
r
VM
s a
n
d
T
a
sks
Evaluation Warning : The document was created with Spire.PDF for Python.