TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol.12, No.6, Jun
e
201
4, pp. 4444 ~ 4
4
5
0
DOI: 10.115
9
1
/telkomni
ka.
v
12i6.548
0
4444
Re
cei
v
ed
De
cem
ber 2
2
, 2013; Re
vi
sed
Febr
uary 15,
2014; Accept
ed March 1, 2
014
A Kind of New Real-time Scheduling Algorithm for
Embedded Linux
Li Yang*, Qing
y
a
n Zhu
Hun
an Co
lle
ge
of Information, Chan
gsh
a
, Ch
ina, 41
02
00
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: 4104
76
2@q
q
.
com
A
b
st
r
a
ct
T
o
solve th
e i
n
sufficiency
of real ti
me p
e
rfor
ma
nc
e
in th
e p
r
esent re
al ti
me sche
dul
in
g al
gorith
m
,
a new real time task classi
fi
cation
sche
d
u
l
i
ng
alg
o
rith
m
i
s
pro
pose
d
. A
ccordi
ng to
th
e class
i
ficati
on
of
arrival
of re
al t
i
me tasks, this
alg
o
rith
m
is d
i
vide
d i
n
to
peri
odic tasks
an
d
non-
peri
o
d
i
c tasks, an
d us
e
s
different improved real tim
e
s
c
hedul
ing algorithm
to sc
hedule different
ty
pe of real tim
e
tasks. Comparing
w
i
th the prese
n
t other real ti
me sch
ed
uli
n
g
algorit
h
m
s,
experi
m
e
n
ts sho
w
this algorith
m
has b
e
e
n
la
rgel
y
improve
d
for in
tegrated re
al ti
me p
e
rfor
ma
nc
e.
Ke
y
w
ords
:
embe
dde
d, rea
l
time sc
hed
ul
i
ng al
gorit
hm, c
l
assifie
d
sche
d
u
lin
g al
gor
ith
m
, perio
dic task, non-
peri
odic tasks
Copy
right
©
2014 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
Linux i
s
wi
del
y applied
in
embed
ded
sy
stem, real ti
me control a
nd oth
e
r
are
a
with th
e
advantag
es o
f
open sou
r
ce code a
nd the system
stability. But Linux belong
s
to general time-
sha
r
ing
ope
rating syste
m
, real-tim
e pe
rforma
nce
is poor,
so it is nee
ded to
improve fo
r the
appli
c
ation
of real
-time e
n
v
ironme
n
t [1]. Real
-tim
e
scheduli
ng alg
o
r
ithm is
one
of the impo
rtant
factors to influen
ce sy
ste
m
real-tim
e perform
an
ce [2], acco
rdin
g to the mecha
n
ism of sche
dulin
g
drive; re
al-ti
m
e sch
eduli
ng alg
o
rithm
can
be
divided into
thre
e cate
go
rie
s
[3]: Time-Driven
sched
uling a
l
gorithm
(TD), Priority-Dri
ven sc
hed
uli
ng algo
rithm
(PD), a
nd
Sharin
g-d
r
ive
n
sched
uling al
gorithm (S
D).
Priority-d
rive
n sch
eduli
n
g
algo
rithm
i
s
the
mo
st
popul
ar sche
duling
alg
o
rit
h
m; its
prin
ciple i
s
to
assign
each
task
a pri
o
rit
y
, while in ev
ery task sch
e
duling to
sele
ct the hig
h
e
s
t
prio
rity task to pe
rform. It
has t
w
o type
s of
stat
ic a
n
d
dynami
c
p
r
i
o
rity sche
duli
ng alg
o
rithm
s
, the
typical repre
s
entative al
g
o
rithm
s
a
r
e
RM a
nd EDF sche
duling
algo
rithms [
4
], and the
r
e
are
some im
prov
ed algo
rithm
s
on this ba
sis,
such as
NSRL [5], LLF [6]
etc.
Curre
n
tly in embed
ded Li
nux real
-time
sch
edul
i
ng algorith
m
is
adde
d to improve the
real
-time pe
rforma
nce of the em
be
dde
d system, but
in sch
edulin
g t
he hybrid real-time ta
sk,
this
method
exist
s
a
lot
of def
ects,
wh
en
m
any sc
h
eduli
ng al
gorith
m
s sche
dule
the
hybri
d
ta
sk
set
inclu
d
ing
pe
ri
odic tasks an
d ap
erio
dic tasks,
onl
y si
mply re
gard t
he a
peri
odi
c
tasks
as spe
c
ial
perio
dic ta
sks, make the real-time p
e
rf
orma
nce dr
o
pped
sub
s
tan
t
ially. In fact
according to
the
nature
of diff
erent ta
sk, real-time t
a
sks
can
be
divided into
different type
s [
7
], for exam
ple,
according
to the a
rrival of t
he task, the t
a
sk
can
be di
vided into
periodic ta
sks a
nd no
n-peri
o
dic
tasks, pe
riodi
c task refers t
o
rea
c
h
and
reque
st
the o
p
e
ration
of task a
c
cordi
ng a
certai
n pe
rio
d
,
aperi
odi
c tasks
refe
rs to t
he task of
ra
ndom a
rrival
system. Thi
s
pape
r ad
opts a ne
w real-ti
m
e
task cla
s
sification sche
duling algo
ri
thm,
st
atic
and dyn
a
mi
c sch
eduli
n
g
algo
rithms are
respe
c
tively use
d
in the perio
dic ta
sks and ap
erio
dic tasks, to maximally improve real-t
ime
perfo
rman
ce of
system.
2.
Classific
a
tio
n
Schedulin
g Algorithm
The gen
eral f
r
ame
w
o
r
k of
the cla
ssification
sched
ulin
g algorithm i
s
shown in Figure 1,
the dyna
mic
sched
uling
al
gorithm
is u
s
ed to
sch
edu
le a
pe
riodi
c
real
-time ta
sk, and
the
stat
ic
sched
uling al
gorithm i
s
used to sched
ul
e perio
dic
rea
l
-time task.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Kind of New Re
al-tim
e Sched
uling Al
gorithm
for Em
bedded Lin
u
x (Li Ya
ng)
4445
Figure 1. The
Idea of Classification Sch
e
duling Algo
rithm
In the prop
o
s
ed dyn
a
mic overmod
u
lat
i
on me
thod,
the dynamic conditio
n
is defined
whe
n
the torque e
rro
r exceed
s 5% of rated torq
ue. As the dyna
mic conditio
n
is en
cou
n
tered,
the origi
nal stator flux error statu
s
,
ψ
+
is modifie
d
b
a
se
d on info
rmation of flux position,
θ
ψ
to
prod
uce the
approp
riate fl
ux error statu
s
ψ
-
. In thi
s
way, the a
c
tive voltage ve
ct
or that
produ
ce
s
the large
s
t tangential flux comp
one
nt
is switched an
d held on, to cre
a
te the largest in
cre
a
se
in
load an
gle an
d hen
ce ra
pid
dynamic torq
ue.
The dist
ributi
on step
s of the cla
ssifie
d
sche
duling al
g
o
rithm are as
follows:
(1) To d
o
sch
edula
b
ility analysis to the
new a
rri
val re
al-time task, if it is a periodic task
to operate
ste
p
(2), othe
rwi
s
e turn to
ste
p
(3).
(2) To d
o
schedul
ability analysi
s
of dynamic
sched
uling algo
rith
m to the new arrival
aperi
odi
c rea
l
-time task, if it meets the sch
edu
lin
g con
d
ition, then the real-ti
m
e task
can
be
sched
ulable,
otherwise it is not sch
edul
a
b
le.
(3) To
do
st
atic
sched
uling alg
o
rithm
of sch
edula
b
ility analysi
s
to th
e ne
w a
rriva
l
perio
dic
real
-time task, if
it
meets
the sched
uling condition,
th
e
n
the re
al-ti
m
e task can
be
sched
ulable,
otherwise it is not sch
edul
a
b
le.
2.1. Task Mo
del
In the
environment
with
sin
g
le processor,
a coll
ection
12
{
,
,
...,
|
(
1
)
}
n
Tt
t
t
i
n
inclu
d
ing n re
al time tasks i
s
define
d
,the rule
s are:
(1) In the
c
o
llec
tion, the time attribute
of
ape
riodi
c real-time
ta
sk can be expre
s
sed
in
a
tr
ia
d
gr
ou
p
(,
,
)
ii
i
i
tA
C
D
. Among them,
i
A
represents
the arriv
a
l time
of task
s
,
i
C
rep
r
e
s
ent
s th
e worst exe
c
ution time
of
tasks,
i
D
rep
r
e
s
ents th
e
relat
i
ve dea
dline
of tasks, the
spe
c
if
ied t
a
sk
s mu
st
be co
mp
leted befo
r
e absolute de
adline
ii
i
dA
D
.
(2)
The time
attribute of p
e
riodi
c
real
-time ta
sk in th
e coll
ectio
n
can be
expre
s
sed
with
a
triad
(,
,
)
ii
i
i
tA
C
P
. Among
them,
i
P
represents the task
cycle.
(3) Ea
ch ta
sk is indep
end
e
n
t with others.
(4) T
he ta
sk
has
pre
e
mpti
on; the tasks with hi
gh
er
priority can
sei
z
e the CP
U co
ntrol of
lower priority tas
k
.
2.2. Relate
d Defini
tion
Definition
1. Processor utiliz
ation: the proportion bet
ween
processor time and
the cost
of real-tim
e tasks in th
e e
x
ecution, that
is,
the pro
p
o
r
tion bet
wee
n
the wo
rst ex
ecutio
n time
and
1
…
n
An
alysis o
f
Sch
e
du
lab
ility
Ap
eriod
i
c task
Period
ic task
D
y
namic schedu
ling
algorithm
Static schedu
lin
g algor
ithm
Que
u
e o
f
task
s
Tasks
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 6, June 20
14: 4444 – 4
450
4446
activity time. The
CPU
p
r
ocesso
r
utili
zation
of re
a
l
-time tasks i
s
/
ii
CD
(
i
t
r
e
p
r
es
en
ts
th
e
aperi
odi
c tasks) or
/
ii
CP
(
i
t
is
the periodic
task).
The processor utili
zation
of real
-time t
a
sk
se
t i
s
sum of the
processor
utilization of all
real
-time ta
sks. Th
e p
r
o
c
e
s
sor utili
zation
of impl
em
ent
n real-tim
e ta
sks i
s
shown
in exp
r
e
ssio
n
s
(1) a
nd (2
):
1
n
i
i
i
C
U
D
(U is
an aperiodic
tas
k
)
(1)
1
n
i
i
i
C
U
D
(U is
a periodic
task)
(2)
Definition 2.
Laxity Time: laxity time
i
L
is t
a
sk a
b
solute
deadli
ne mi
n
u
s th
e
re
st of
task
executio
n time
i
E
, and minu
s the curre
n
t time
t
,
su
ch a
s
ex
pre
ssi
on (
3
):
ii
i
Ld
E
t
(3)
Definition 3.
Wo
rst Case
Execution Ti
me: the wo
rst execution ti
me
i
C
is
refers
to
the
real
-time ta
sks d
u
rin
g
the
i
m
pleme
n
tatio
n
of a
maxim
u
m of th
e Ti
me, incl
udin
g
sche
dulin
g t
a
sk
of spen
ding T
i
me.
Define 4. Im
portant d
egre
e
(Impo
r
tan
c
e): it is
a po
si
tive integer, indicating that
this task
comp
ared to anothe
r Impo
rtance.
3. Analy
s
is o
f
Commonly
Used Sc
hed
u
ling Algorithm
(1) Rate Mo
n
o
tonic (RM)
sched
uling
al
gorithm
[8
],
it
is a
ki
nd of typical static prio
rity
sched
uling al
gorithm, a
c
co
rding to the l
ength of
task cycle to dete
r
mine
sched
u
ling prio
rity, the
smalle
r ta
sk cycle
s
i
s
corre
s
p
ondin
g
to hig
her prio
rity. Advantage
s: si
mple reali
z
a
t
ion
mech
ani
sm a
nd lo
w
sched
uling. Di
sa
dvantage
s: o
n
ly
t
a
s
k
cy
cle i
s
t
a
ke
n a
s
a
p
r
iorit
y
d
e
ci
sio
n
condition, it is easy to ignor
e those i
m
portant longer
cycl
e ta
sk; CP
U utilization rate is
low;
resources
utilization i
s
i
n
suffici
ent, the processing
capability fo
r emergency
real
-time i
s
not
st
ron
g
.
(2) Th
e Ea
rli
e
st
Dea
d
line
First (E
DF
) sche
dulin
g
algorith
m
[9]
is
a
kind
o
f
typical
pree
mptive d
y
namic
prio
rit
y
scheduli
n
g
algo
rithm, a
c
cordi
ng to
the de
adlin
e
of ea
ch ta
sk to
assign p
r
io
rity, the task with the earlie
st deadli
ne h
a
s the hi
ghe
st prio
rity. The advantag
e
is:
resou
r
ce allo
cation i
s
flexi
b
le to gu
ara
n
tee the ta
sk with ea
rlie
st deadli
ne to
impleme
n
t wi
th
prio
rity
and CPU utilizati
on
is
high.
Disadvanta
g
e
s
: syste
m
o
v
erhea
d is b
i
g, whi
c
h
ca
n't
guarantee th
e executio
n o
f
inaccessibl
e
deadlin
e ta
sk; in overlo
ad
situation the
r
e may ha
pp
en
domino p
hen
omena m
a
ke real
-time pe
rforma
nce fall sha
r
ply.
(3)
Lea
st
Lax
ity First
(LL
F
) sche
dulin
g al
gorithm
[10], i
t
is
also a
ki
n
d
of
com
m
onl
y use
d
dynamic
prio
rity sche
duling
algorithm, it i
s
the
imp
r
ove
m
ent to EDF
algorith
m
, according to th
e
non-de
scendi
ng o
r
de
r
of task’s lei
s
ure
time to di
st
ri
bute p
r
io
rity, the tasks
with shorte
r l
e
isure
time has hi
g
her p
r
io
rity. The provision i
s
la
xity time
of runni
ng ta
sks remain
s
unchan
ged,
and
laxity time of tasks in
re
ady
que
ue
de
cre
a
se
s
with
tim
e
; the p
r
io
rity is al
so
dynam
ically
cha
nge
d.
The advanta
ges a
r
e: saving CPU
re
so
urces a
nd
en
suri
ng that e
m
erg
e
n
c
y task implem
enta
t
ion
in prio
rity. Disadvantag
es: the prio
rity wa
iting
for tasks may becom
e
higher
with the de
cre
a
se of
free time, p
r
e
e
mption th
e
currently ru
nni
ng ta
sks,
ma
y result in fre
quent
switchi
ng ph
eno
me
non
betwe
en tasks (i.e. bump
)
, whi
c
h ma
ke
s the
system p
e
rform
a
n
c
e d
r
opp
ed sub
s
tantially.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Kind of New Re
al-tim
e Sched
uling Al
gorithm
for Em
bedded Lin
u
x (Li Ya
ng)
4447
4. Analy
s
is o
f
Improv
ed Scheduling Al
gorithm
4.1. D
y
namic
Scheduling Algorithm
Acco
rdi
ng to
the an
alysi
s
of above
ge
n
e
ral
algo
rithm
s
, to th
e a
periodic re
al-tim
e tasks,
a kin
d
of im
proved
Dead
line an
d L
a
xity (DAL
) first scheduli
n
g
algo
rithm, it con
s
id
ers t
h
e
influen
ce of the dea
dline
and idle tim
e
on pr
i
o
rity, and in
crea
se
s the p
r
eem
pting thre
sh
o
l
d
sched
uling al
gorithm to re
duce bump.
4.1.1. Analy
s
is of DAL Scheduling Alg
o
rithm
The algo
rithm
combi
nes th
e advantag
es of
EDF algori
t
hm and LLF
algorith
m
, as
well a
s
the prio
rity size of
i
P
, acco
rdi
ng to the expression (4) th
e
dynamic cha
nge
s are ta
ke
n place:
*(
1
)
*
ii
i
Pd
L
(4)
In the exp
r
e
ssi
on,
is the
bala
n
ce fa
ctor of
real
-time ta
sks, th
e si
ze
de
cid
e
s th
e
influen
ce
deg
ree
of the
pri
o
rity by the
d
eadlin
e an
d i
d
le time.
Wh
en
is 0
it is tendin
g
to E
D
F
algorith
m
, if it is 1, and ten
d
ing to LLF al
gorithm.
The ide
a
of the preemptin
g thre
shol
d sche
duling
alg
o
rithm is [1
1]: using the p
r
iority
(P
,
)
ir
i
P
to control the
implementati
on of tasks, hereinto
i
P
is the basi
c
pri
o
rity,
()
ri
i
r
i
PP
P
n
is the preemption threshol
d of ta
sks, whi
c
h permit
s the fulfillment
of preem
ptive task pri
o
rity an
d
basi
c
pri
o
rity at the same time, is greate
r
t
han the ba
sic ta
sk p
r
iori
ty and preem
ption thre
shol
d
w
i
th
th
e
task
p
r
e
e
mp
te
d
.
Th
is
a
l
g
o
r
i
th
m
a
b
s
o
r
bs
the
i
r
r
e
sp
ec
tive
a
d
va
n
t
a
g
e
s
o
f
pr
e
e
m
p
t
io
n
and
non-preempti
v
e sched
ulin
g algo
rithm,
redu
ce
s
the
numbe
r of p
r
eem
ption, a
nd redu
ce
s t
h
e
system
overh
ead by the
jo
lt. The co
mpl
e
te algo
rithm
pro
c
e
s
s of d
e
scribi
ng ta
sk sch
eduli
ng
is
s
h
ow
n
in
F
i
gu
r
e
2
.
Fig
u
re 2. Ta
sk S
c
he
duling Fl
o
w
Ch
art of Algorithm
4.1.2. Anal
y
s
is of Algorithm Schedulabilit
y
For an in
dep
ende
nt a peri
odic real
-time
task set, the minimum un
boun
d of pro
c
e
s
sor
utilization rati
o is 1. T
h
eref
ore,
as long as
the processor
utiliz
ation of
a real
-time
task set
1
U
,
then they ca
n be sch
edul
ed. Whe
n
a
new a
p
e
r
iodi
c re
al-time ta
sk
(,
,
)
ii
i
i
tA
C
D
arrives, if it
meets exp
r
e
s
sion (5):
1
i
n
i
C
U
D
(5)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 6, June 20
14: 4444 – 4
450
4448
So it ca
n be
sche
dule
d
, am
ong the
m
,
n
U
is
processor utilization ra
te
of current real-t
ime
tas
k
s
set.
4.2. Static Scheduling Al
gorithm
Acco
rdi
ng to
the analysi
s
of above ge
neral
algo
rith
ms, a ki
nd of
improved
Rate And
Importan
c
e (RAI) sche
duli
ng algo
rithm
is used to
so
lve periodi
c real-time ta
sk,
con
s
ide
r
s th
e
proportion of cycle and i
m
porta
nt degree of two fact
ors, improves impl
ement
ation probability
rate of import
ant tasks.
4.2.1. Analy
s
is of RAI Sc
heduling Alg
o
rithm
In RAI algori
t
hm, the prio
rity of a task i
s
ap
pointed
before
ope
rat
i
on, whi
c
h i
s
deci
ded
by perio
d an
d impo
rtant d
egre
e
, wh
ere
the impo
rt
an
ce d
egree i
s
set by the u
s
er a
c
cording
to
deman
d. Re
gulation i
s
: the task
with
sho
r
t cy
cle
has
high
er
prio
rity; the task with
hig
her
importa
nt deg
ree. Thi
s
algo
rithm ca
n be
divided into the followi
ng several ways:
(1)
Re
al-time
tasks
with same cy
cle a
nd differe
nt importa
nt deg
ree, a
c
cordin
g to the
importa
nt deg
ree exe
c
ute i
n
desce
ndin
g
orde
r.
(2) Real
-time tasks
with sa
me
importa
nt
degree
an
d di
fferent cycle,
accordan
ce
with the
RM sche
dulin
g algorith
m
, the tasks
with sho
r
ter
cycle
impleme
n
t in prio
rity.
(3) Real
-time
tasks
with
sa
me imp
o
rta
n
t
degree
and
cycle, a
c
cordin
g to th
e
way
of FIFO
to do sched
ul
ing.
(4) Re
al-time
tasks
with
di
fferent imp
o
rt
ant
deg
re
e a
nd
cycle,
reg
a
rdin
g the i
m
portant
degree a
s
preemptive co
n
d
itions, to wa
rra
nty pr
iority impleme
n
tation of importa
nt tasks.
4.2.2. Anal
y
s
is of Algorithm Schedulabilit
y
For n in
depe
ndent pe
riodi
c re
al-time ta
sks se
t with
each othe
r, the minimu
m up bou
nd
of the processor utilization ratio i
s
1
*(
2
1
)
n
n
. Therefore, as l
ong as
the processor
utilization U
of re
al-time
task
set i
s
no
t more
tha
n
t
he mi
nimum
up b
oun
d, th
en they
can
be
sched
ulin
g.
Whe
n
a ne
w perio
dic
real
-time task
(,
,
)
ii
i
i
tA
C
P
arrived, if it meets the formula (6):
1
*(
2
1
)
i
n
n
i
C
Un
P
(6)
Then it i
s
sch
edule
d
, and
h
e
reinto,
n
U
is th
e processo
r a
m
ortiza
tion of
cu
rre
nt real-t
ime
s
e
t.
5. Test And Analy
s
is of the Performa
nce
In ord
e
r to te
st the real-ti
m
e pe
rform
a
nce
of
this al
gorithm, in th
e experi
m
ent
platform
w
i
th AMD
Sempr
o
n 2.01GH
z
s
i
ngle pr
oc
es
sor
an
d 1.75GB memor
y
, to hybr
id
r
eal-
t
ime tasks
set incl
udin
g
100 tasks, re
spe
c
tively ad
opts this al
go
rithm, EDF sche
duling
alg
o
rithm, and
RM
sched
uling
a
l
gorithm, the
test is exe
c
uted
with
t
he pe
rforma
nce
evaluati
on indi
cato
rs of
sched
ule del
ays and d
ead
line miss rate
. In the tes
t,
the su
bmissio
n
orde
r of task is same, an
d
the differen
c
e
of task su
bm
issi
on time is
1s.
5.1. Scheduling Dela
y
Sched
uling
Laten
cy (SL) represe
n
ts
the ti
me sp
ent for real
-time tasks from the
happ
ening
of
sched
uling
chan
ce to
be
sched
uling; it
is a
n
imp
o
rt
ant indi
cato
r
to mea
s
ure t
h
e
real
-time pe
rf
orma
nce. Wh
ile the sch
e
d
u
ling del
ay is
lowe
r, the bet
ter is t
he
pe
rforma
nce of real
time.
From Ta
ble
1, the algorit
hm in this p
aper i
s
less
than EDF scheduli
ng algo
rithm no
matter the
la
rge
s
t, the
smallest, a
nd
the average
sched
uling al
gorithm, but they
are
sli
g
htly
above the RM sch
eduli
n
g
algorithm. It indicates t
he
sched
uling d
e
lay perfo
rm
ance of algori
t
hm
in this pap
er is betwe
en
EDF sche
duli
ng algo
rithm
and RM
sch
edulin
g algo
ri
thm, sch
eduli
n
g
co
st is slig
htly higher tha
n
RM sch
e
d
u
ling alg
o
rith
m, and far l
o
we
r than E
D
F sch
edulin
g
algorith
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Kind of New Re
al-tim
e Sched
uling Al
gorithm
for Em
bedded Lin
u
x (Li Ya
ng)
4449
Table 1. Co
m
pari
s
on
s of T
e
st Re
sult Ab
out Sched
ulin
g Delay
Method
Scheduling Delay /
s
Max
i
mum
Scheduling Delay /
s
Max
i
mize
Scheduling Delay /
s
Average
EDF scheduling algorithm
Classification alg
o
rithm
RM scheduling algorithm
6
4
3
19
15
14
13.92
11.31
10.68
5.2. Deadline
Miss Rate
Dea
d
line Miss Rate (DM
R
)
represents
the
rati
o b
e
tween q
uantity
of real
-time t
a
sks that
doe
s not fulfill deadlin
e in
the system i
n
a ce
rtai
n pe
riod of time a
nd total num
ber of real-ti
m
e
tasks; it is an
other im
po
rta
n
t indicator to
mea
s
ure the
real
-time p
e
rforman
c
e.
De
adline
miss
rate
is lower, the real-time p
e
rf
orma
nce is b
e
tter.
Figure 3 pre
s
ent
s th
e con
d
ition th
at deadlin
e miss
rate varie
s
wi
th the incre
a
s
e qu
antity of real-tim
e ta
sks, where three kin
d
s of al
gorithm
s are in
the same
co
ndition, the t
r
an
sverse
axis p
r
e
s
ent
s real-time ta
sk quantity, the vertical ax
is
pre
s
ent
s de
a
d
line mi
ss
ra
te. As can
b
e
se
en fro
m
the gra
ph, th
e dea
dline m
i
ss
rate of th
e
algorith
m
in this pa
per va
ries
with the
incre
a
se
of load, the gro
w
th rate is f
a
ster tha
n
EDF
sched
uling a
l
gorithm,
a
n
d
far slo
w
er than
RM
scheduli
ng al
g
o
rithm. Th
e
perfo
rman
ce
of
deadli
ne mi
ss rate of the
algorith
m
in t
h
is p
ape
r
i
s
b
e
twee
n
EDF
sched
uling al
gorithm and
RM
sched
uling al
gorithm.
Fig
u
re 3. Comparis
on
of Test
Re
sults of De
adline Mi
ss
Rate
6. Concludin
g
Remark
s
This p
ape
r p
u
ts forward a
kind of cl
assi
fica
tion sch
e
duling al
gorit
hm, usin
g two kind
s of
improve
d
sch
edulin
g algo
ri
thm, and ma
ke a reg
u
lati
on
that different types of real
-time tasks ca
n
flexibly choose suitable
scheduli
ng algo
rithm for
sch
edulin
g. In addition, sch
ed
ulability analysi
s
in the algo
rithm is
simple
; sch
edula
b
ili
ty analys
is o
f
real-time ta
sk
of usin
g
any sched
uli
ng
algorith
m
is indep
ende
n
t
with real
-ti
m
e ta
sks of
anothe
r ki
nd of sch
e
d
u
ling al
gorith
m
.
Sched
uling d
e
lay and dea
dline miss rat
e
as pe
rform
ance indicato
rs, to hybrid real-time ta
sks se
t
inclu
d
ing 10
0 tasks, the
algorithm in
this pape
r
make
s a
co
mpari
s
o
n
an
d EDF and
RM
sched
uling al
gorithm. Th
e experim
ental
results
sho
w
that the algorithm in this p
aper
ha
s thei
r
respe
c
tive ad
vantage
s in EDF and
RM
sch
eduli
ng a
l
gorithm, inte
grated
real
-ti
m
e perfo
rma
n
ce
has a g
r
eat d
eal of improv
ement.
Referen
ces
[1]
Hon
g
Ji
n, Ho
n
gan
W
ang, Qi
a
ng W
a
ng. A r
e
al-time
sc
he
dul
ing
al
gorithm
b
a
sed
on
pri
o
rit
y
ta
bl
e a
nd
its
implem
entati
o
n
.
Journal of Sof
t
w
a
re
. 2004;15
(3): 360-3
70.
[2]
Shen
gh
ui Liu, Song
Ma.
R
e
s
earch an
d
a
p
p
l
icatio
n of r
eal-ti
m
e sche
d
u
lin
g
mecha
n
ism
ba
sed o
n
Lin
u
x
kernel.
Co
mput
er eng
ine
e
ri
ng
and a
p
p
licati
o
n
. 2008; 44(
6): 121-1
23.
[3]
Chu
n
-Hsi
ung
L
an, Ha
i-Min
g
C
hen, Me
i-Hsi
u
Che
n
, Chi
h
-W
ei C
h
iu, C
h
e
n
g
-
Jui T
s
ai. Green Prod
uctio
n
Sched
uli
ng/Pl
a
nni
ng
w
i
th Par
a
lle
l-Mac
h
in
e La
yo
ut
for Multiple Orders a
nd Divers
i
fie
d
Due Dat
e
s
.
JCIT.
2012; 7(7): 43-50.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 6, June 20
14: 4444 – 4
450
4450
[4]
W
e
ixiao, Z
h
i
b
ao F
e
n
g
. Res
earch
and
im
plem
entatio
n
of adva
n
ce
d ED
F
sched
uli
ng a
l
gor
ithm.
Co
mp
uter eng
i
neer
ing
. 2
009;
35(1
8
): 231-
23
3.
[5]
Liu
CL, L
a
y
l
a
n
d
J. Sche
dul
in
g al
gorit
hms for mu
lti
pro
g
r
a
mming
in
a
hard r
eal-tim
e
envir
onme
n
t.
Journ
a
l of AC
M
. 2003, 20(
1) : 46-61.
[6]
Sun Yua
n
, Z
hao
Xia
obi
ng,
Yang Guos
he
ng. An
Optima
l Reserv
ation-
base
d
F
eed
ba
ck Schedu
lin
g
Algorit
hm for Vehic
u
lar Ap
plic
at
ion Sp
ecific
Operatin
g S
y
st
ems
. IJACT
. 2
012; 4(1): 3
59-
377.
[7]
Ji
w
e
n
Do
ng,
Yang Z
h
an
g. Improvem
ent a
nd a
ppl
ic
ati
o
n
of embe
dd
ed
real
-time
op
e
r
ating s
y
ste
m
sched
uli
ng a
l
g
o
rithm.
Co
mp
u
t
er appl
icatio
ns
. 2009; 29(
9): 2516-
251
9.
[8]
Yunfu T
an, Jie
Liu, Guo
h
u
a
Liu. A ki
nd of i
m
pr
ove
d
rea
l
-ti
m
e sche
dul
in
g
alg
o
rithm i
n
L
i
nu
x
and
its
app
licati
on.
Com
p
uter science
. 2008; 35(
10): 256-
258.
[9]
Hon
g
Jin,
Hon
gan W
a
ng. An
integr
ated
de
sign m
e
tho
d
o
f
task priorit
y
.
Journ
a
l of S
o
ftw
are
. 2
0
03;
14(3): 37
6- 38
2.
[10]
Dan
a
M, Pasc
ale M, L
aure
n
t G. Anal
y
s
is o
f
dead
lin
e assi
gnme
n
t metho
d
s in d
i
strib
u
ted re
al-ti
m
e
s
y
stems.
Co
mputer Co
mmun
i
catio
n
. 200
4; 27(15): 14
12-
14
23.
[11]
T
e
rrasa A, Gar
c
ia Fomes A,
Bo
tti VJ. F
l
exi
b
le re
al-time
Li
nu
x: a fle
x
i
b
l
e
hard re
al-tim
e envir
onme
n
t
.
Real-T
i
m
e Systems.
20
04; 22
(2): 151-1
73.
Evaluation Warning : The document was created with Spire.PDF for Python.