7. multimedia operating system – classification of real
TRANSCRIPT
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
1 / 5
2C
onte
nts
7.1
Ove
rvie
w
7.2
Rea
l Tim
e Sy
stem
s
7.3
Res
ourc
e M
anag
emen
t–
Dim
ensi
ons
in R
esou
rce
Des
ign
–R
eser
vatio
n S
trat
egie
s
7.4
Proc
ess
Man
agem
ent
–C
lass
ifica
tion
of R
eal-T
ime
Sch
edul
ing
Str
ateg
ies
–S
ched
ulab
ility
Tes
ts a
nd O
ptim
al S
ched
uler
s
–P
reem
ptiv
e vs
. Non
-Pre
empt
ive
Tas
k S
ched
ulin
g
7.5
Prot
otyp
e Sy
stem
s–
Rea
l-Tim
e M
ach
–Y
AR
TO
S (
Yet
Ano
ther
Rea
l-Tim
e O
pera
ting
Sys
tem
)
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
2 / 5
27.
1 O
verv
iew
Ope
ratin
g Sy
stem
Asp
ects
for
Mul
timed
ia P
roce
ssin
g:–
Mos
t con
vent
iona
l ope
ratin
g sy
stem
s do
offe
r on
ly li
ttle
or n
o su
ppor
t for
intim
epr
oces
sing
of c
ontin
uous
med
ia. T
his
conc
erns
all
func
tions
of a
n op
erat
ing
syst
em li
ke p
roce
ss, m
emor
y, fi
le o
r de
vice
man
agem
ent.
Res
ourc
e M
anag
emen
t:–
How
to a
chie
ve a
coo
rdin
ated
pro
cess
ing
of a
ll op
erat
ing
syst
em fu
nctio
ns in
orde
r to
ach
ieve
an
end-
to-e
nd Q
ualit
y of
Ser
vice
(de
lay,
cap
acity
, los
s ra
te,
jitte
r, ..
.)?
·A
n ab
stra
ct c
ontin
uous
res
ourc
e m
odel
·A
com
mon
res
ourc
e m
anag
emen
t pro
cedu
re
Proc
ess
Man
agem
ent:
–H
ow to
sch
edul
e pr
oces
ses
perm
ittin
g ea
ch to
term
inat
e ac
cord
ing
to is
tde
adlin
e?·
Rat
e-M
onot
onic
Sch
edul
ing
·E
arlie
st D
eadl
ine
Firs
t
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
3 / 5
27.
1 O
verv
iew
/ R
esou
rce
Man
agem
ent T
asks
1. A
dmis
sion
con
trol
–Is
ther
e en
ough
rem
aini
ng c
apac
ity to
han
dle
the
addi
tiona
l dat
a st
ream
?
2. Q
oS c
alcu
lati
on
–W
hich
cha
ract
eris
tics
(e.g
. in
term
s of
thro
ughp
ut a
nd d
elay
) ar
e av
aila
ble
for
the
new
str
eam
?
3. R
esou
rce
rese
rvat
ion
–R
eser
ves
the
reso
urce
s w
hich
are
req
uire
d to
mee
t the
dea
dlin
es.
4. Q
oS e
nfor
cem
ent
–P
rovi
sion
of s
ervi
ce gu
aran
ted
by a
ppro
pria
te s
ched
ulin
g, e
.g. b
y re
orde
ring
(ser
ving
task
s w
ith s
hort
dea
dlin
e ea
rlier
to a
task
with
less
str
ict b
ound
s).
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
4 / 5
27.
2 R
eal T
ime
Syst
ems
Rea
l Tim
e:
–A
rea
l-tim
e ta
sk is
a p
roce
ss w
hich
del
iver
s is
t pro
cess
ing
resu
lt in
a g
iven
tim
eor
acc
ordi
ng to
a g
iven
dea
dlin
e. A
dea
dlin
e is
giv
en e
.g. i
n th
e or
der
of
ms
for
inte
ract
ive
voic
e or
vid
eo d
ata
or
days
for
text
doc
umen
ts.
Dea
dlin
e:
–A
dea
dlin
e re
pres
ents
the
late
st a
ccep
tabl
e tim
e fo
r th
e pr
esen
tatio
n of
the
proc
essi
ng r
esul
t of a
task
. Dea
dlin
e ar
e ca
lled
hard
if fa
ilure
s ar
e m
issi
on-c
ritic
alor
thre
aten
ing
hum
an b
eing
s. T
hey
are
calle
d so
ft de
adlin
es if
they
can
not
exac
tly b
e de
term
ined
or
a vi
olat
ion
is le
ss c
ritic
al.
Fiel
ds o
f ap
plic
atio
n:
–C
ontr
ol s
yste
ms
for
man
ufac
turin
g pr
oces
ses,
mili
tary
sys
tem
s,te
leco
mm
unic
atio
n sy
stem
s, a
ircra
fts, a
utom
obile
s, n
ucle
ar p
ower
pla
nts
orin
tera
ctiv
e m
ultim
edia
sys
tem
s.
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
5 / 5
27.
2 R
eal T
ime
Syst
ems
Proc
essi
ng R
equi
rem
ents
:–
Pre
dict
able
fast
res
pons
e to
tim
e-cr
itica
l eve
nts,
–ac
cura
te ti
min
g in
form
atio
n,–
high
deg
ree
of s
ched
ulab
ility
, i.e
. res
ourc
e ca
paci
ty is
not
was
ted
(
clev
er s
ched
ulin
g; b
ut fi
ndin
g op
timal
sch
edul
es is
ofte
n a
NP
-com
plet
e ta
sk),
and
–st
abili
ty u
nder
tran
sien
t ove
rload
, i.e
. usi
ng b
uffe
ring
to c
ope
with
bur
sty
syst
ems.
Asp
ects
spe
cifi
c to
Mul
timed
ia S
yste
ms:
–In
-tim
e pr
oces
sing
, tra
nsm
issi
on a
nd p
rese
ntat
ion
of c
ontin
uous
med
ia d
ata
like
audi
o or
vid
eo.
Req
uire
men
ts a
re k
now
n as
Qua
lity
of S
ervi
ce (
QoS
) pa
ram
eter
s, e
.g.:
–th
roug
hput
–lo
cal o
r gl
obal
(i.e
. end
-to-
end)
del
ay,
–(d
elay
) jit
ter,
or
–re
liabi
lity.
The
se v
alue
s ar
e us
ually
spe
cifi
ed b
y gi
ving
an
aver
age
valu
e, a
wor
st c
ase
valu
e,pe
ak r
ates
, dis
trib
utio
n fu
nctio
n or
mom
ents
of
its d
istr
ibut
ion
func
tion.
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
6 / 5
27.
2 R
eal T
ime
Syst
ems
real
-tim
e sy
stem
s
soft
rea
l-tim
e sy
stem
sha
rd r
eal-
time
syst
ems
high
ava
ilabi
lty
high
inte
grity
fail
safe
fail
opea
rtio
n
tele
phon
esw
itchi
ngon
-lin
eba
nkin
gra
ilway
sign
alin
gfl
ight
cont
rol
soft
rea
l-tim
e sy
stem
- co
nseq
uenc
es o
f a
syst
em f
ailu
rear
e of
the
sam
e m
agni
tude
as
utili
tyhi
gh a
vaila
bilit
y-
dow
n-tim
e is
min
imal
high
inte
grity
- co
nsite
ncy
of d
ata
mus
t sur
vive
any
syst
em f
ailu
re a
nd m
alic
ious
atte
mpt
to a
lter
the
data
hard
rea
l-tim
e sy
stem
- co
nseq
uenc
es o
f fa
ilure
are
cata
stro
phic
fail-
safe
sys
tem
- pr
obab
ility
to d
etec
t any
fai
lure
is
clos
e to
1-
syst
em c
an b
e st
oppe
dfa
il-op
erat
iona
l sys
tem
- m
inim
al s
ervi
ce e
ven
in c
ase
of
failu
re-
syst
em c
anno
t be
stop
ped
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
7 / 5
27.
3 R
esou
rce
Man
agem
ent
Eve
n so
phis
ticat
ed c
ompr
essi
on te
chni
ques
can
not c
ompe
nsat
e fo
r th
e re
sour
ce
capa
city
nec
essa
ry f
or a
udio
and
vid
eo tr
ansm
issi
on a
s w
ell a
s pr
oces
sing
in c
urre
nt(i
nter
activ
e) m
ultim
edia
sys
tem
s.
Har
dwar
e R
esou
rces
in Y
ear
x
Res
ourc
e R
equi
rem
ents 19
8019
9020
00
Net
wor
k Fi
le A
cces
s
“Kill
er A
pplic
atio
n“
Inte
ract
ive
Vid
eo
Hig
h-Q
ualit
y A
udio
insu
ffic
ient
scar
ce
abun
dant
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
8 / 5
27.
3 R
esou
rce
Man
agem
ent /
Dim
ensi
ons
in R
esou
rce
Des
ign
Act
ive
vs. p
assi
ve r
esou
rces
dep
endi
ng o
n its
aut
onom
ous
proc
essi
ng
capa
bili
ties
–ac
tive
reso
urce
s: C
PU
, net
wor
k ad
apte
r ca
rd, e
tc.
–pa
ssiv
e re
sour
ces:
file
sys
tem
, mai
n m
emor
y, e
tc.
Shar
ed v
s. e
xclu
sive
res
ourc
e us
age
activ
e re
sour
ce a
re u
sual
ly a
lloca
ted
excl
usiv
ely
whe
reas
pas
sive
one
s ca
n be
sha
red
by m
ultip
le ta
sks
Sing
le v
s. m
ulti
ple
reso
urce
occ
uren
ces
PCs
usua
lly c
onta
in o
nly
a si
ngle
CPU
whe
reas
e.g
. a S
PA
RC
20
wor
ksta
tion
cont
ains
two
CPU
s
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
9 / 5
27.
3 R
esou
rce
Man
agem
ent /
Pro
cedu
re
2. S
ched
ulab
ility
3. Q
oS C
alcu
latio
n
Res
ourc
e M
anag
er1.
Req
uest
by
a ne
w ta
sk
CPU
Res
ourc
es
. . I/O
6. A
dd T
ask
Que
ue
5. C
alcu
late
Sch
edul
e
7. S
ched
ule
Tas
k
4. R
eser
vatio
n
4.
Dis
patc
her
8. A
ssig
nR
esou
rces
Ste
ps 1
to 5
: pre
pari
ng ta
sk p
roce
ssin
gSt
eps
6 to
8: t
ask
proc
essi
ng
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
10 /
527.
3 R
esou
rce
Man
agem
ent /
Res
erva
tion
Stra
tegi
es
prin
cipl
e
timel
ines
of
proc
essi
ngno
gua
rant
ee
reso
urce
util
izat
ion
over
book
ing
poss
ible
pote
ntia
lly h
igh
base
for
sch
edul
abili
ty te
st
aver
age
test
dete
ct a
nd s
olve
con
flic
ts
optim
istic
pess
imis
tic
avoi
d co
nflic
ts
max
imum
for
pea
k ra
te lo
ad
no o
verb
ooki
ng
“gua
rant
ee“
The
air
line
exam
ple:
Nor
thw
est
risky
, lot
s of
ove
rboo
king
,so
lve
conf
licts
by
findi
ng a
cust
omer
s w
ho le
aves
the
airc
raft
( by
pay
ing
som
ethi
ngin
cas
h or
-be
tter-
in v
ouch
er)
Luf
than
save
ry c
autio
us a
irlin
e(n
o ov
erbo
okin
g, n
o-sh
ows)
,lo
w a
ctua
l loa
d,hi
gh p
rices
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
11 /
527.
3 R
esou
rce
Man
agem
ent /
Abs
trac
t Con
tinuo
us M
edia
Mod
ellin
g: W
orkl
oad
Dat
a st
ream
s co
nsis
t of
peri
odic
ally
arr
ivin
g L
ogic
al D
ata
Uni
ts (
calle
d M
essa
ges)
and
are
desc
ribe
d by
the
“Lin
ear
Bou
nded
Arr
ival
Pro
cess
(L
BA
P)
Mod
el“.
A d
ata
stre
am is
a tr
ipel
(M
, R, B
), w
here
–M
is th
e m
axim
um m
essa
ge s
ize,
–R
is th
e m
axim
um m
essa
ge r
ate
(i.e.
the
num
ber
of m
essa
ge p
er ti
me
unit)
, and
–B
is th
e m
axim
um b
urst
ines
s or
allo
wed
wor
kahe
ad.
The
mod
el is
nam
ed li
near
bou
nd a
rriv
al p
roce
ss b
ecau
se it
ass
umes
that
the
num
ber
of m
essa
ge a
rriv
als
N in
a g
iven
tim
e in
terv
al
is b
ound
ed b
yN
()
= R
⋅ +
B
(R
⋅
mes
sage
arr
ival
s in
ti
me
units
)
Res
ourc
e
Mes
sage
s M
essa
ges
Inte
rfac
es
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
12 /
527.
3 R
esou
rce
Man
agem
ent /
Abs
trac
t Con
tinuo
us M
edia
Mod
ellin
g: E
xam
ple
Aud
io s
ampl
ing:
44.
1 kH
z
1
6 bi
t per
sam
ple
(
i.e. 1
6 x
4410
0 =
705
.600
bit/
sec)
R =
75
fram
es/s
ec
Ía
fram
e co
ntai
ns 7
05.6
00/7
5 =
9.4
08 b
it
ÍM
= 9
.408
(he
re: “
fram
e“
“m
essa
ge“)
Num
ber
N M o
f m
essa
ges
(of s
ize
M)
in t
time
units
is g
iven
by:
=
+
⋅⋅=
B)
is
size
burs
t
max
imum
(i
f
burs
ts)
tu
due
s v
aria
tion
no (i
f
)(
BM
R
MR
tN
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
13 /
527.
3 R
esou
rce
Man
agem
ent /
Abs
trac
t Con
tinuo
us M
edia
Mod
ellin
g: W
orka
head
The
wor
kahe
ad w
(t)
of a
n L
BA
P at
tim
e t d
escr
ibes
how
man
y m
essa
ges
have
arri
ved
that
are
not
pro
cess
ed y
et. I
t is
defi
ned
by
w(t
) =
max
{0,
N(
[t0,
t] )
- R
⋅ |t
- t 0
|}.
The
logi
cal a
rriv
e tim
e l(
mi)
of m
essa
ge m
i is
the
time
at w
hich
a m
essa
ge is
effe
ctiv
ely
bein
g sc
hedu
led.
It i
s de
fine
d by
l(m
i)
= a
i + w
(ai)
/ R
(= a
ctua
l arr
ival
tim
e +
del
ay d
ue to
wor
kahe
ad)
l(m
i+1)
= m
ax {
a i+
1, l(
mi)
+ 1
/ R
}
whe
re a
i is
the
actu
al a
rriv
ing
time
of m
essa
ge m
i.
t
w(t
)
12
34
56
78
910
123
a 1l(
m1)
a 2l(
m2)a 3
a 4 l(m
3)l(
m4)
a i =
act
ual a
rriv
al ti
me
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
14 /
52
Res
ourc
e
Mes
sage
s M
essa
ges
Inte
rfac
es
Que
ue
Serv
er
min
im. l
og. d
elay
= U
max
im. l
og. d
elay
= D
7.3
Res
ourc
e M
anag
emen
t / A
bstr
act C
ontin
uous
Med
ia M
odel
ling:
Res
ourc
es
The
logi
cal d
elay
d(m
) of
mes
sage
s m
bet
wee
n tw
o in
terf
aces
I1
and
I2 is
defi
ned
by d
(m)
= l2
(m
) -
l1(m
).
The
buf
fer
requ
irem
ents
of
reso
urce
for
a g
iven
dat
a st
ream
are
def
ined
by
buf
= B
+ R
(D-U
) w
here
:
B
= n
umbe
r of
mes
sage
s w
hich
arr
ive
unex
pect
edly
due
to b
urst
ines
s,
D
= m
axim
um lo
gica
l del
ay b
etw
een
inpu
t and
out
put i
nter
face
s,
U
= m
inim
um (
unbu
ffer
ed)
actu
al d
elay
bet
wee
n th
e sa
me
inte
rfac
es, a
nd
R(D
-U)
= n
umbe
r of
msg
. whi
ch m
ay b
e bu
ild u
p du
e to
the
vari
atio
n pr
oces
sing
tim
e
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
15 /
527.
4 Pr
oces
s M
anag
emen
t / B
asic
s
Proc
ess
Man
agem
ent d
eals
wit
h th
e as
sign
men
t of
the
CPU
to p
roce
sses
/task
s.
A p
roce
ss m
ay b
e in
one
of
five
bas
ic s
tate
s:–
initi
al, i
.e. i
t is
crea
ted,
but
not
in s
ched
ule;
pro
cess
is id
le
–re
ady,
i.e.
it is
wai
ting
for
CP
U a
ssig
nmen
t,
–ru
nnin
g, i.
e. it
is r
unni
ng o
n th
e C
PU
,
–w
aitin
g, i.
e. it
is w
aitin
g fo
r an
ext
erna
l eve
nt, o
r
–fin
ishe
d.
The
sch
edul
er c
hoos
es th
e ne
xt p
roce
ss to
bec
ome
runn
ing
acco
rdin
g to
a g
iven
sche
dule
. The
sch
edul
e de
term
ines
the
orde
r of
CPU
ass
ignm
ent t
o pr
oces
ses.
Goa
ls o
f tr
adit
iona
l sch
edul
ing
–op
timal
thro
ughp
ut, o
ptim
al r
esou
rce
utili
zatio
n, o
r fa
ir qu
euei
ng
Goa
ls o
f re
al-t
ime
sche
dulin
g–
exec
ute
max
imum
num
ber
of p
roce
sses
in ti
me,
i.e.
acc
ordi
ng
to th
eir
dead
lines
–m
inim
ize
dead
line
viol
atio
ns
inru
fire in
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
16 /
527.
4 Pr
oces
s M
anag
emen
t / C
lass
ific
atio
n of
Rea
l-T
ime
Sche
dulin
g St
rate
gies
Sche
dulin
g st
rate
gies
can
be
dist
ingu
ishe
d by
–st
atic
vs.
dyn
amic
sch
edul
e ca
lcul
atio
n,(s
tatic
= c
alcu
latio
n of
sch
edul
e in
adv
ance
dyna
mic
= r
e-ca
lcul
atio
n w
hene
ver
a ne
w ta
sk a
rriv
es)
–ce
ntra
l vs.
dis
trib
uted
sch
edul
e ca
lcul
atio
n, a
nd–
pree
mpt
ive
vs. n
on-p
reem
ptiv
e ta
sk p
roce
ssin
g
(pr
eem
ptiv
e =
a ta
sk m
ay b
e in
terr
upte
d by
any
task
with
hig
her
prio
rity)
The
y sc
hedu
le–
task
s w
ith p
erio
dic
or a
perio
dic
proc
essi
ng r
equi
rem
ents
, and
–in
depe
nden
t tas
ks o
r ta
sks
with
pre
cede
ncy
cons
trai
ns.
The
y ar
e ap
plie
d to
eith
er–
unip
roce
ssor
sys
tem
s,–
mul
tipro
cess
or s
yste
ms
(neg
lect
ing
com
mun
icat
ion
dela
y), o
r–
mul
ticom
pute
r sy
stem
s (t
akin
g co
mm
unic
taio
n de
lay
into
acc
ount
).
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
17 /
527.
4 Pr
oces
s M
anag
emen
t / S
ched
ulab
ility
Tes
ts a
nd O
ptim
al S
ched
uler
s
The
test
to d
eter
min
e w
ethe
r a
sche
dule
exi
sts
for
a gi
ven
task
set
is c
alle
d a
sche
dula
bilit
y te
st. T
here
are
thre
e ki
nds
of te
st: s
uffi
cien
t, ex
act,
and
nece
ssar
y
ones
: –S
uffic
ient
test
s, i.
e. if
the
test
is p
ositi
ve, t
he
task
set
is s
ched
ulab
le. A
neg
ativ
e
r
esul
t is
poss
ible
eve
n if
the
task
set
is s
ched
ulab
le (
“cau
tious
“ te
st).
–N
eces
sary
test
s, i.
e. if
the
test
is n
egat
ive,
the
task
set
is n
ot s
ched
ulab
le. A
posi
tive
resu
lt do
es n
ot g
uara
ntee
a ta
sk s
et´s
sch
edul
abili
ty (
“opt
imis
tic“
test
).
–E
xact
test
s re
turn
a p
ositi
ve r
esul
t if t
he ta
sks
set i
s sc
hedu
labl
e.
Mos
t exa
ct s
ched
ulab
ility
test
s be
long
to th
e cl
ass
of N
P-co
mpl
ete
prob
lem
s.
A s
ched
uler
is c
alle
d op
tim
al if
it a
lway
s fi
nds
a sc
hedu
le f
or ta
sks
sets
sat
isfy
ing
an e
xact
sch
edul
abili
ty te
st.
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
18 /
52
A ta
sk is
cha
ract
eriz
ed b
y is
t tim
ing
cons
trai
ns a
nd it
s re
sour
ce r
equi
rem
ents
.
Mos
t tas
ks in
mul
timed
ia s
yste
ms
are
peri
odic
and
hav
e no
pre
cede
nce
cons
trai
nts.
Mod
el f
or a
task
s tim
ing
cons
trai
nt: (
s i, e i,
d i, p i)
whe
re:
–s i s
tart
ing
poin
t, i.e
. rea
dy ti
me
for
first
per
iod
–e i p
roce
ssin
g tim
e fo
r pe
riod
pi
–d i d
eadl
ine
for
perio
d pi
(re
lativ
e to
ist p
erio
d´s
read
y tim
e)
–p
perio
d of
task
T
–r
rate
of T
(w
ith p
= r
-1)
If a
task
set
con
sist
ing
of p
erio
dic
task
s (T
1, ..
., T n)
is s
ched
ulab
le [T i =
(s i,
e i, d i,
p i)]
then
the
proc
esso
r ut
iliza
tion
is g
iven
by
(whe
re e
i/pi =
rea
ltive
pro
cess
or
utili
zatio
n by
task
Ti)
s i ↓
perio
d 1
perio
d 2
perio
d 3
perio
d 4
p id i
e i
∑ =
=n i
ii peU
1
7.4
Proc
ess
Man
agem
ent /
Mod
el f
or R
eal-
Tim
e T
asks
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
19 /
527.
4 Pr
oces
s M
anag
emen
t / P
reem
ptiv
e vs
. Non
-Pre
empt
ive
Tas
k Sc
hedu
ling
The
re a
re ta
sks
sets
that
hav
e va
lid p
reem
ptiv
e sc
hedu
les
but n
o no
n-pr
eem
ptiv
e
ones
. If
the
cost
for
pre
empt
ion
is n
egle
cted
, pre
empt
ive
sche
dulin
g is
alw
ays
bette
r or
equ
al th
an n
onpr
eem
ptiv
e sc
hedu
ling.
aab
cd
ef
b1
2
21
11
11
11
Dea
dlin
e vi
olat
ion
Hig
h-R
ated
Tas
k T
1
Low
-Rat
ed T
ask
T2
Non
-Pre
empt
ive
Sche
dule
Pree
mpt
ive
Sche
dule
d ad b
d cd d
d fd e
d 1d 2
Dea
dlin
es
cd
ef
ab
cd
ef
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
20 /
527.
4 Pr
oces
s M
anag
emen
t / R
ate
Mon
oton
ic A
lgor
ithm
(R
M)
- B
asic
s
Cla
ssif
icat
on
–st
atic
, pre
empt
ive
algo
rithm
for
perio
dic
task
s
Ass
umpt
ions
–al
l tim
e-cr
itica
l tas
ks h
ave
perio
dic
com
putin
g de
man
ds
–ta
sks
are
mut
ually
inde
pend
ent (
i.e. n
o pr
eced
ence
con
stra
ints
)
–a
task
s de
adlin
e eq
uals
ist p
erio
d (d
i = p
i)
–a
task
s m
axim
um c
ompu
ting
time
is c
onst
ant a
nd a
-prio
ri kn
own
–co
ntex
t sw
itche
s ar
e co
nsid
ered
tim
eles
s
Prin
cipl
es
–sh
orte
st p
erio
d ÌÍ
hig
hest
prio
rity
(i.e.
task
s ar
e or
dere
d by
dec
reas
ing
perio
d)
–pr
iorit
ies
are
reca
lcul
ated
if a
new
task
is a
dded
to th
e ta
sk s
et o
r a
task
is d
elet
ed
fr
om th
e ta
sk s
et [s
ched
ule
calc
ulat
ion
only
onc
e fo
r a
give
n ta
sk s
et]
RM
is o
ptim
al a
mon
g st
atic
sch
edul
ing
algo
rith
m, i
.e. i
f a ta
sk s
et is
sch
edul
able
by a
ny s
tatic
alg
orith
m th
en th
ere
exis
ts a
feas
ible
RM
sch
edul
e.
Í p
reem
ptio
n is
ass
umed
to c
ome
with
out
c
ost (
at le
ast w
ithou
t tim
e co
st)
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
21 /
527.
4 Pr
oces
s M
anag
emen
t / R
ate
Mon
oton
ic A
lgor
ithm
- E
xam
ple
T1
rece
ives
pri
ority
ove
r T
2 Í
T1
pree
mpt
s T
2 to
mee
t ist
dea
dlin
es
Perf
orm
ance
of
RM
dep
ends
on
the
arri
val p
atte
rn:
wor
st c
ase
(“cr
itica
l ins
tant
“): e
very
task
with
hig
her
prio
rity
arriv
es a
t the
sam
e tim
ein
stan
t
Dea
dlin
es
Hig
h-R
ated
Tas
k T
1
Low
-Rat
ed T
ask
T2
RM
Sch
edul
e1
ab
12
cd
1
ab
cd
12
peri
od o
f T
2
peri
od o
f T
1
d ad b,
d1
d cd d,
d2
pree
mpt
ion
of T
1T
1 is
res
umed
(pre
empt
ion
wou
ld n
ot h
ave
been
nec
essa
ry in
this
exa
mpl
e in
ord
er to
mee
t the
dea
dlin
es)
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
22 /
527.
4 Pr
oces
s M
anag
emen
t / E
arlie
st D
eadl
ine
Firs
t (E
DF)
Cla
ssif
icat
ion
–dy
nam
ic (
i.e. r
e-ca
lcul
atio
n w
hene
ver
a ne
w ta
sk a
rriv
es),
pre
empt
ive
algo
rithm
f
or p
erio
dic
task
sPr
inci
ples
–e
arlie
st d
eadl
ine
<=
> h
ighe
st p
riorit
y–
prio
ritie
s ar
e re
-cal
cula
ted
each
tim
e a
task
bec
omes
rea
dy (
even
for
an u
ncha
nged
task
set)
–ca
lcul
atio
n ha
s w
orst
cas
e co
mpl
exity
of O
(n2 )
1a
b2
cd
ab
cd
12
d ad b,
d1
d cd d,
d2
Dea
dlin
es
Hig
h-R
ated
Tas
k T 1
Low
-Rat
ed T
ask
T 2
ED
F S
ched
ule
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
23 /
527.
4 Pr
oces
s M
anag
emen
t / C
ompa
riso
n of
ED
F an
d R
M
RM
sch
edul
es p
oten
tially
req
uire
mor
e co
ntex
t sw
itche
s, i.
e. m
ore
(and
nec
essa
ry)
pree
mpt
ions
, tha
n E
DF.
The
hig
her
num
ber
of p
reem
ptio
ns f
or R
M h
as to
be
com
pare
d w
ith th
e ad
ditio
nal c
ost
for
ED
F (d
ue to
rec
alcu
latio
n of
sch
edul
es).
1a
b2
cd
ab
cd
12
d ad b,
d1
d cd d,
d2
Dea
dlin
es
Hig
h-R
ated
Tas
k T
1
Low
-Rat
ed T
ask
T2
ED
F Sc
hedu
le
1a
b1
2c
d1
RM
Sch
edul
e
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
24 /
527.
4 Pr
oces
s M
anag
emen
t / C
ompa
riso
n of
ED
F an
d R
M
ED
F is
bet
ter
than
RM
: If
RM
can
sch
edul
e a
task
set
the
n th
e sa
me
is v
alid
for
ED
F b
ut n
ot v
ice
vers
a.
Rat
e m
onot
onic
vs.
ED
F: p
roce
ssor
util
izat
ion
CA
3A
2C
7
C7
B5
CB
B
12
34 4
56 6
7H
igh
Rat
e
Low
Rat
e
ED
F
Rat
e M
onot
onic
d 1d 2
d 3d 4
d 6d 5
d Ad B
Dea
dlin
es
8
d 7
d C
AB
C
A2
43
AB
56
C8 8
Dea
dlin
e V
iola
tions
d A n
ot m
etd C
not
met
1 1A
B
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
25 /
527.
4 Pr
oces
s M
anag
emen
t / A
chie
vabl
e Pr
oces
sor
Util
izat
ion
with
RM
Min
imum
uti
liza
tion
for
all
sets
of
task
s (I
1, ..
., I n
) w
hich
are
sch
edul
able
and
whi
ch “
fully
util
ize“
the
proc
esso
r.
A ta
sk s
et fu
lly u
tiliz
es a
pro
cess
or if
–ta
sk s
et c
an b
e sc
hedu
led,
and
–if
a si
ngle
task
is in
crea
sed
in p
roce
ssin
g tim
e by
ε
> 0
then
the
assi
gnm
ent
beco
mes
infe
asib
le.
Exa
mpl
e:
Giv
en a
task
set
(I 1
, I2,
I3)
wit
h pe
riod
pi a
nd p
roce
ssin
g tim
e e i
for
eac
h ta
sk:
p1
= 3
, e1
= 1
; p
2 =
4, e
2 =
1;
p3
= 5
, e3
= 1
It c
an b
e sh
own
that
in th
is c
ase
[
x] =
sm
alle
st in
tege
r ≥
x
12
12
15
121
232
3m
ax,3
=⋅
−⋅
−=
⋅−
⋅−
=pp
epp
ep
e
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
26 /
527.
4 Pr
oces
s M
anag
emen
t / A
chie
vabl
e Pr
oces
sor
Util
izat
ion
with
RM
: The
orem
A s
et o
f n
inde
pend
ent a
nd p
erio
dic
task
s (T
1, ..
., T
n) c
an b
e sc
hedu
led
if
As
a co
nseq
uenc
e: A
low
er b
ound
for
pro
cess
or u
tiliz
atio
n is
:–
ln 2
if R
M is
app
lied
–1
(low
er a
nd u
pper
bou
nd)
if E
DF
is a
pplie
d (a
nd d
eadl
ine
= e
nd o
f per
iod)
() 1
2...
1
22
11−
⋅≤
++
+n
npe
pe
pe
nn
For
n →
thi
s ex
pres
sion
con
verg
es to
ln 2
≈ 0
,693∞
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
27 /
527.
4 Pr
oces
s M
anag
emen
t / A
chie
vabl
e Pr
oces
sor
Util
izat
ion
with
ED
F
For
ED
F a
muc
h be
tter
(i.e
. per
fect
) ut
iliza
tion
is p
ossi
ble:
Mix
ed s
chem
e:
Supp
ose
we
have
n p
erio
dic
task
s.
Prio
ritie
s gi
ven
acco
rdin
g to
RM
are
I1,
I2,
...,
I n (I
1 sh
orte
st p
erio
d, I
n =
long
est p
erio
d),
i.e. h
ighe
st p
rior
ity
for
task
1.
Com
prom
ise:
–S
ched
ule
I 1, I 2
, ...,
I k w
ith R
M (
1 ≤
k ≤
n)
–S
ched
ule
I k+1,
I k+
2, ..
., I n
with
ED
F w
hen
the
proc
esso
r is
not
occu
pied
by
I 1, I 2
, ...,
I k
1...
22
11≤
++
+nn pe
pe
pe
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
28 /
527.
4 Pr
oces
s M
anag
emen
t / A
chie
vabl
e Pr
oces
sor
Util
izat
ion
with
ED
F
060
A1
A2
A3
A4
A9
A14
A20
B1
B2
B3
B4
B5
C1
C2
C3
A1B
1 C1
A2B
2C2 A3
B3 A
4C3
A5 B
4
A6 B
5C
4
A7
C8
B6A
8 C5
A9
B7C
6 A10
B8
A11 C
7B9 A
12C
8B
10
A13
A14 B
11
C9 A
15B
12A16 C
10A
17B13 C
11A18
A19
C12
B14
B15 A
20
empt
yre
mai
n
slot
s
60 of
out
13
i.e.
, 6047
51
41
31 oc
cupa
tion
=+
+
C1´
C2´
C3´
empt
yre
mai
n
slot
s
60 of
out
1on
ly
i.e
. ,
6059
52
71
31 oc
cupa
tion
=+
+
A1 B
1
C1A
2 C1´
B2
A3
A4
A5
B3
C2´
C3´
C3´
B46
78
910
1112
1314
1516
1718
1920
Tas
k 1
(p1
= 3,
e1
= 1
)
Tas
k 2
(p2 =
4, e
2 =
1)
Tas
k 3
(p3
= 5,
e3
= 1
)
Tas
k 3
with
e3
= 2
RM
Mix
ed
stra
tegy
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
29 /
527.
4 Pr
oces
s M
anag
emen
t / D
eadl
ine
Mon
oton
e A
lgor
ithm
Shor
test
dea
dlin
e fi
rst:
–st
atic
alg
orith
m w
hich
giv
es h
ighe
r pr
iorit
y to
a ta
sk w
ith s
hort
er d
eadl
ine,
i.e.
T1 >
T2
if d
1 <
d2
If d
eadl
ine
= p
erio
d Í
RM
alg
orith
m.
Thi
s is
sim
ilar
but
not
iden
tical
to E
DF.
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
30 /
527.
4 Pr
oces
s M
anag
emen
t / S
hort
est (
Rem
aini
ng)
Proc
essi
ng T
ime
The
task
wit
h
–sh
orte
st p
roce
ss ti
me
(
if no
n-pr
eem
ptiv
e)
–sh
orte
st r
emai
nin
g tim
e (if
pre
empt
ive
h
as h
ighe
st p
riorit
y.
Thi
s st
rate
gy s
erve
s th
e m
axim
um n
umbe
r of
cus
tom
ers.
In a
n ov
erlo
ad s
ituat
ion
and
if a
ll ta
sks
have
the
sam
e de
adlin
e th
en th
is
stra
tegy
min
imiz
es th
e nu
mbe
r of
dea
dlin
e vi
olat
ions
.
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
31 /
527.
4 Pr
oces
s M
anag
emen
t / P
reem
ptiv
e vs
. Non
-pre
empt
ive
“Pre
empt
ive“
is m
ore
com
plic
ated
but
:
–in
crea
ses
the
feas
abili
ty o
f sch
edul
ing
(in s
ome
case
s a
pree
mpt
ive
sche
dule
exis
ts b
ut n
o no
n-pr
eem
ptiv
e sc
hedu
le)
–re
duce
s th
e am
ount
of p
riorit
y in
vers
ion
(i.e.
situ
atio
ns w
here
low
er p
riorit
y jo
bs
are
exec
uted
whi
le h
ighe
r pr
iorit
y jo
bs a
re w
aitin
g).
A ta
sk s
et w
ith p
roce
ssin
g tim
es e
i and
wit
h re
ques
t per
iods
pi i
s sc
hedu
labl
e if
:
In g
ener
al th
e re
sulti
ng s
ched
ule
is a
pre
empt
ive
one.
stra
tegy
)
ED
Fby
e.g.
ad
just
ed,
y dy
nam
ical
l
bem
ay
pr
iori
ties
(i
f 1
assi
gned
)
fixe
d
are
pr
iori
ties
task
(i
f
693
,02
ln
≤
≈≤
∑∑
iiii pepe
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
32 /
527.
4 Pr
oces
s M
anag
emen
t / P
reem
ptiv
e vs
. Non
-pre
empt
ive
Sche
dula
bili
ty f
or n
on-p
reem
ptiv
e st
rate
gies
is m
ore
com
plic
ated
:
Tas
k I k
can
be
sche
dule
d (w
orst
cas
e co
nsid
erat
ion)
if it
s de
adlin
e d k
sat
isfi
es
the
follo
win
g:
here
In
is th
e hi
ghes
t pri
ority
task
I1
is th
e lo
wes
t pri
ority
task
(*)
=
ow
n ex
ecut
ion
time
(**)
=
wai
ting
tim
e du
e to
job
foun
d in
ser
vice
at a
rriv
al ti
me
(wor
st c
ase:
max
imum
exe
cutio
n tim
e of
all
jobs
of
othe
r pr
iori
ty c
lass
es)
(***
) =
exe
cuti
on ti
me
of a
ll jo
bs o
f hi
gher
pri
ority
whi
ch a
re p
rese
nt a
t arr
ival
tim
e or
whi
ch a
rriv
e du
ring
the
wai
ting
time
x
job
of p
rior
ity
k ar
rive
shi
gher
pri
o jo
bsar
rive
and
are
ser
ved“m
y se
rvic
e tim
e“
xd
k
!
:tR
equi
rem
en≥
∑ +=
≠
+
−
⋅+
+≥
n kj
j
jj
ji
ki
kk
p
ed
ee
ed
1
1m
ax(*
)(*
*)
(***
)
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
33 /
527.
4 Pr
oces
s M
anag
emen
t / E
xam
ple
for
Ach
ieva
ble
Proc
esso
r U
tiliz
atio
n
Giv
en a
task
set
(I 1
, I2,
I3)
wit
h pe
riod
pi a
nd p
roce
ssin
g tim
e e i
for
eac
h ta
sk:
p 1 =
3, e
1 =
1 ;
p2
= 4
, e2
= 1
; p
3 =
5, e
3,m
ax =
2
Con
ject
ure:
Mix
ed s
chem
e is
nea
rer
to th
e be
tter
side
(na
mel
y E
DF)
than
to th
e “p
oo
r“
side
(na
mel
y R
M).
308.2
to
in
crea
sing
whe
n %
100
15
308,2
41
31
n(
ED
F)U
tiliz
atio
78
%0,
783
51
41
31
(RM
)n
Util
izat
io
%98
983
,052
41
31 sc
hem
e)
(mix
edn
Util
izat
io
* 33
=≈
=+
+
≈=
++
≈=
++
ee
I 1 sc
hedu
led
with
RM
,I 2
and
I 3 sc
hedu
led
with
ED
F
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
34 /
527.
4 Pr
oces
s M
anag
emen
t / A
chie
vabl
e Pr
oces
sor
Util
izat
ion
(her
e R
M)
Util
izat
ion:
The
act
ual u
tiliz
atio
n is
slig
htly
bet
ter
than
the
theo
retic
al m
inim
um:
78,0)1
2(331
=−
=th
eore
mU
I 1I 1
I 1I 1 I 2
I 2I 2
I 3I 3
I 1I 2
I 3I 1
I 2I 3
I 1I 2
Tas
k I 1
RM
sch
edul
e
Tas
k I 3
Tas
k I 2
t
%78
783
,051
41
31
33
22
11≈
=+
+=
++
=pe
pe
peU
actu
al
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
35 /
527.
4 Pr
oces
s M
anag
emen
t / L
east
Lax
ity F
irst
(L
LF)
Tas
k w
ith s
hort
est r
emai
ning
Lax
ity is
sch
edul
ed f
irst
.
Def
init
ion:
laxi
ty l k
(t)
of ta
sk T
in p
erio
d k
at ti
me
t
The
rem
aini
ng ti
me
from
t to
the
dead
line
d of
task
T in
per
iod
k th
at is
not
use
d
for
proc
essi
ng ta
sk T
.
lk(
t) =
(s
+ (
k-1)
p +
d)
-
(
t + e
rem
(t))
=
(D
eadl
ine
in p
erio
d k)
-
(ac
tual
tim
e +
rem
aini
ng p
roce
ssin
g tim
e)
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
36 /
527.
4 Pr
oces
s M
anag
emen
t / L
east
Lax
ity F
irst
(L
LF)
LL
F ha
s no
adv
anta
ge o
ver
the
ED
F fo
r un
ipro
cess
or s
yste
ms.
If
task
s ha
vesi
mil
ar la
xity
val
ues
cont
ext s
witc
hes
can
occu
r fr
eque
ntly
(ca
lcul
atio
nov
erha
ead
com
pare
d to
ED
F). L
Lf
mig
ht b
e su
itabl
e on
ly in
mul
tipro
cess
orsy
stem
s, i.
e. w
hen
seve
ral r
esou
rces
are
sch
edul
ed s
imul
tane
ousl
y.
s
e
s +
(k-
1)p
t 1t
e rem(t
)
t 2
l k(t 1)
l k(t 2) s +
(k-
1) p
+d
Dea
dlin
e in
per
iod
k
s +
k ⋅
p
peri
od k
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
37 /
527.
4 Pr
oces
s M
anag
emen
t / L
east
Lax
ity F
irst
- E
xam
ple
for
Tw
o T
asks
Giv
en a
min
imum
CPU
tim
e gr
anul
arity
of
2 Ì
if n
o gr
anul
arity
ass
umed
Í
task
pre
empt
ion
as o
ften
as
poss
ible
Í
“pr
oces
sor
shar
ing“
Lax
ities
do
not c
hang
e du
ring
exe
cutio
n tim
e of
a ta
sk b
ut b
ecom
e sm
alle
r whe
n a
task
is n
otse
rved
. If
the
laxi
ty o
f a
wai
ting
task
bec
omes
sm
alle
r th
an th
at o
f th
e ru
nnin
g ta
sk th
en th
e ru
nnin
gta
sk is
inte
rrup
ted.
I 1I 2
I 1I 2
I 1I 2
I 1t
L2(
I 1) =
20
L2(
I 2) =
14
I 2 in
terr
upts
I1
L8(
I 1) =
14
L8(
I 2) =
14
I 1 in
terr
upts
I2
L10
(I1)
= 1
4L
10(I
2) =
12
I 2 in
terr
upts
I1
I 2 fi
nish
edI 1
fini
shed
...
LL
F Sc
hedu
le
Tas
k I 2
Tas
k I 1
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
38 /
527.
4 Pr
oces
s M
anag
emen
t / S
ched
ulin
g A
peri
odic
(bu
t Ind
epen
dent
) T
asks
Mul
tim
edia
app
licat
ions
con
sist
s of
bot
h–
perio
dic
and
–ap
erio
dic
task
s,
like
mos
t rea
l-ti
me
syst
ems.
Peri
od ta
sks
are
used
for
tran
spor
t and
pro
cess
ing
of c
ontin
uous
med
ia d
ata
whe
reas
con
trol
or
man
agem
ent t
asks
are
ape
riod
ic. O
ften
it is
rea
sona
ble
to g
ive
to p
erio
dic
task
s pr
iori
ty o
ver
aper
iodi
c ta
sks.
Prob
lem
:–
How
to s
ched
ule
a ta
sk s
et w
hich
is c
ompo
sed
of p
erio
dic
as w
ell a
s ap
erio
dic
task
s?
Idea
: –O
ne s
peci
al p
erio
dic
task
(ca
lled
serv
er)
is p
ollin
g fo
r re
ady
aper
iodi
c ta
sks
to b
epr
oces
sed.
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
39 /
527.
4 Pr
oces
s M
anag
emen
t / S
ched
ulin
g A
peri
odic
(bu
t Ind
epen
dent
) T
asks
Prob
lem
:
–A
perio
dic
task
s ca
n on
ly b
e pr
oces
sed
whe
n se
rver
task
is s
ched
uled
, i.e
. the
y
mis
s th
eir
dead
line
if it
is e
arlie
r th
an th
e ne
xt ti
me
the
serv
er is
sch
edul
ed.
Solu
tion
:
So-c
alle
d ba
ndw
idth
pre
serv
ing
algo
rith
m li
ke
–P
riorit
y E
xcha
nge,
–D
efer
rab
le S
erve
r, o
r
–S
pora
dic
Ser
ver.
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
40 /
527.
4 Pr
oces
s M
anag
emen
t / B
andw
idth
Pre
serv
ing
Alg
orith
ms
Prio
rity
Exc
hang
e:
One
per
iodi
c ta
sk T
s ser
ves
all a
peri
odic
task
s. T
he s
erve
r ex
chan
ges
prio
rity
with
alo
wer
pri
or p
erio
dic
task
Ti i
f no
ape
riod
ic ta
sk is
rea
dy. T
he s
erve
r re
ceiv
e its
initi
alpr
iori
ty f
or th
e ne
xt p
erio
d.
Cas
e 1:
Ts s
erve
s >
0 a
peri
odic
task
s
Cas
e 2:
Ts ha
s no
task
to s
erve
, cha
nges
pri
ority
with
Ti a
nd is
rea
ctiv
ated
at t
he
tim
e T
i was
sch
edul
ed.
T1
Ts
T1
Tn
Ti
......
T1´
Ts´
T1´
Tn´
Ti´
......
peri
od k
peri
od k
+1
T1
Ti
T1
Tn
Ts
......
T1´
Ts´
T1´
Tn´
Ti´
......
peri
od k
peri
od k
+1
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
41 /
527.
4 Pr
oces
s M
anag
emen
t / B
andw
idth
Pre
serv
ing
Alg
orith
ms
Def
erra
ble
Serv
er:
One
per
iodi
c ta
sk (
calle
d de
ferr
able
ser
ver)
ser
ves
all a
peri
odic
task
s. I
t def
ers
its p
roce
ssin
g ti
me
if n
o ap
erio
dic
task
is r
eady
but
ret
ains
ist p
rior
ity. A
s so
on
as a
n ap
erio
dic
task
req
uest
occ
urs,
it e
ither
–(im
med
iatly
) pr
eem
pts
the
runn
ing
task
s (t
hat h
as a
low
er p
riorit
y), o
r
–re
sum
es p
roce
ssin
g af
ter
the
curr
ent t
ask
term
inat
es.
Ts
T1
T2T
3T
4...
Tn
Inte
rrup
T4 or
wai
tun
til T
4 is
fin
ishe
d
noth
ing
to s
erve
aper
iodi
c ta
sk a
rriv
al
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
42 /
527.
4 Pr
oces
s M
anag
emen
t / B
andw
idth
Pre
serv
ing
Alg
orith
ms
Spor
adic
ser
ver:
It c
ombi
nes
the
adva
ntag
es o
f th
e Pr
iori
ty E
xcha
nge
and
the
Def
erra
ble
Serv
er
algo
rith
m. A
spo
radi
c se
rver
exc
hang
es it
s pr
iori
ty w
hen
no a
peri
odic
task
is
read
y. A
ny s
pare
CPU
cap
acit
y, i.
e. ti
me
not u
sed
by p
erio
dic
task
s, is
tran
sfor
med
into
a ti
cket
, tha
t is
give
n to
a s
pora
dic
serv
er, w
hich
then
repl
enis
hes
its in
itial
pri
ority
. Thi
s w
ay, t
he s
pora
dic
serv
er is
allo
wed
to u
se
any
idle
tim
e of
the
CPU
.
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
43 /
527.
4 Pr
oces
s M
anag
emen
t / S
ched
ulin
g T
asks
with
Blo
ckin
g
The
blo
ckin
g of
a ta
sk is
cau
sed
by m
utua
l exc
lusi
on w
hen
tryi
ng to
acc
ess
acr
itica
l sec
tion
curr
ently
occ
upie
d by
ano
ther
task
.Pr
iori
ty I
nver
sion
Eff
ect:
Ass
ume
a hi
gh-p
rior
ity ta
sk T
h w
ants
to e
nter
a c
ritic
al s
ectio
n cu
rren
tly o
ccup
ied
by a
low
-pri
ority
task
Tl.
Th
is b
lock
ed u
ntil
Tl l
eave
s th
e cr
itica
l sec
tion.
Unt
il th
en, n
ot o
nly
Tl b
ut a
ll m
ediu
m-p
rior
ity ta
sks
Tm
hav
ing
high
er p
rior
ity th
an T
l and
low
er p
rior
ityth
an T
h w
ill b
e pr
oces
sed
prio
r to
Th.
Thi
s ph
enom
enon
is c
alle
d pr
iori
ty in
vers
ion.
The
rul
e: A
low
pri
ority
task
mus
t wai
t, i.e
. can
not b
egin
exe
cutio
n, if
a h
ighe
r pr
iori
ty
task
is r
unni
ng.
criti
cal s
ectio
n
proc
.bl
ocke
d un
til
unit
(s)
sign
al(s
)pr
oces
sing
Tas
k T
h
Tas
k T
l
Tas
k T
m
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
44 /
527.
4 Pr
oces
s M
anag
emen
t / S
ched
ulin
g T
asks
with
Cri
tical
Sec
tions
Seve
ral a
lgor
ithm
s ha
ve b
een
deve
lope
d in
ord
er to
avo
id p
rior
ity in
vers
ion,
e.g
.
–P
riorit
y In
herit
ance
, or
–th
e C
eilin
g P
roto
col.
Prio
rity
Inh
erita
nce:
–A
low
-prio
rity
task
Tl i
nhe
rits
the
prio
rity
of a
hig
h-pr
iorit
y ta
sk T h if
it ca
uses
the
bloc
king
of T
h. T
l ret
ains
ist i
nitia
l prio
rity
whe
n le
avin
g th
e cr
itica
l sec
tion.
Thu
s: –R
eady
jobs
bet
wee
n “lo
w“
and
“hig
h“ a
re b
lock
ed.
Dis
adva
ntag
e:
– L
ower
util
izat
ion
of s
erve
r, p
ossi
ble
dead
line
viol
atio
ns.
Adv
anta
ge:
–S
eque
nce
of s
ched
ules
is p
rese
rved
.
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
45 /
527.
4 Pr
oces
s M
anag
emen
t / S
ched
ulin
g T
asks
with
Cri
tical
Sec
tions
Mor
e Pr
oble
ms
wit
h “s
tand
ard“
prio
rity
inhe
ritan
ce:
–T
asks
can
be
bloc
ked
in e
ach
criti
cal s
ectio
n,
–de
adlo
cks
can
occu
r, a
s w
ell a
s
–tr
ansi
tive
bloc
king
is p
ossi
ble,
i.e.
T3
is b
lock
ed b
y T 2
that
is b
lock
ed b
y T 1
Prin
cipl
e (c
eili
ng p
roto
col)
:
Eac
h ta
sk h
as tw
o pr
iori
ties:
–a
stat
ic o
ne a
ssig
ned
by th
e sc
hedu
ler,
e.g
. acc
ordi
ng to
RM
str
ateg
y
–a
dyna
mic
one
, i.e
. the
max
imum
of t
he
stat
ic p
riorit
y an
d th
e hi
ghes
t prio
rity
inhe
rited
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
46 /
527.
4 Pr
oces
s M
anag
emen
t / S
ched
ulin
g T
asks
with
Cri
tical
Sec
tions
Eac
h se
map
hore
has
a m
axim
um o
r ce
iling
pri
ority
val
ue o
f al
l tas
ks th
at
actu
ally
use
it. A
sem
apho
re o
nly
can
be lo
cked
by
a ta
sk w
ith a
hig
her
dyna
mic
pri
ority
than
any
cur
rent
ly lo
cked
sem
apho
re. (
com
para
ble
to
“hie
rarc
hy o
f re
sour
ces“
app
roac
h to
dea
dloc
k av
oida
nce)
Ben
efits
:
–A
task
is o
nly
bloc
ked
once
, no
dead
lock
s an
d tr
ansi
tive
bloc
king
.
Pric
e: –R
estr
ictiv
e lo
ckin
g po
licy
(bru
te fo
rce
met
hod)
Í
long
er b
lock
ing
dela
ys fo
r
othe
r ta
sks.
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
47 /
527.
5 Pr
otot
ype
Syst
ems
/ Rea
l-T
ime
Mac
h
It is
a d
istr
ibut
ed o
pera
ting
syst
em f
or r
eal-
time
appl
icat
ions
bas
ed o
n th
eA
RT
S (A
dvan
ced
Rea
l-T
ime
Syst
em),
bot
h de
velo
ped
at C
arne
gie
Mel
lon
Uni
vers
ity a
t Pit
tsbu
rgh,
PA
, USA
.
RT
Mac
h:–
supp
orts
mul
ti-pr
oces
sor
–cl
uste
rs n
≥ 1
proc
esso
rs in
to a
pro
cess
or s
et
–as
sign
s a
sepa
rate
run
que
ue a
nd s
ched
ulin
g st
rate
gies
to e
ach
proc
esso
r se
t
–en
able
s an
app
licat
ion
to s
elec
t the
cur
rent
sch
edul
ing
stra
tegy
(at
run
-tim
e)
RT
Mac
h ke
rnel
proc
esso
r se
t
proc
esso
r
sche
dule
r
stra
tegy
run
queu
e
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
48 /
52
Gen
eric
Sch
edul
er
Thr
ead
Dis
patc
hing
Mgm
t.
Proc
esso
r Se
t Mgm
t.
RM
RM
/DS
RM
/DS
FP RR
7.5
Prot
otyp
e Sy
stem
s / T
ask
Sche
dulin
g in
RT
Mac
h
RT
Mac
h of
fers
thre
e cl
asse
s of
task
s (c
alle
d R
T T
hrea
ds):
–pe
riodi
c or
ape
riodi
c (s
pora
dic)
task
s w
ith h
ard
dead
lines
, or
–ta
sks
with
sof
t dea
dlin
es.
Tas
ks a
re s
ched
uled
by
one
of th
e fo
llow
ing
stra
tegi
es (
dyna
mic
ally
cha
ngab
le):
–ra
te m
onot
onic
(R
M),
with
def
erra
ble
serv
er (
RM
/DS
), o
r sp
orad
ic s
erve
r (R
M/S
S),
–fix
ed p
riorit
y (F
P),
as
wel
l as
–ro
und
robi
n (R
R).
Thr
ead
disp
atch
ing
man
agem
ent c
ontr
ols
idle
thre
ads
and
aper
iodi
c se
rver
(D
S o
r S
S).
Pro
cess
orse
t man
agem
ent p
erfo
rms
cont
ext s
witc
hing
, thr
ead
pree
mpt
ion,
or
proc
esso
r as
sign
men
t.
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
49 /
527.
5 Pr
otot
ype
Syst
ems
/ The
QoS
Tic
ket M
odel
The
QoS
Tic
ket M
odel
com
bine
s re
sour
ce r
eser
vati
on a
nd a
dapt
atio
n.
It h
as b
een
impl
emen
ted
on R
T-M
ach
3.0
supp
rotin
g so
-cal
led
Q-T
hrea
ds.
Qos
Tic
ket a
llow
s us
ers
to s
peci
fy to
lera
nce
rang
es f
or p
erio
d an
d co
mpu
tatio
n tim
e. T
icke
t is
issu
ed f
or e
ach
sess
ion
com
pris
ing
seve
ral t
hrea
ds. Q
oS p
aram
eter
s ar
e ad
apte
d dy
nam
ical
ly
base
d on
the
curr
ent r
esou
rce
cons
umpt
ion.
2. C
alcu
latio
n of
R
esou
rce
Allo
catio
n
QoS
Man
ager
1. Q
oS R
eque
st
4. I
ssue
Tic
ket
Mul
timed
ia S
essi
on
QoS
Tic
ket
3. R
esou
rce
Res
erva
tion
RT
Mac
h ke
rnel
5. C
onsu
mpt
ion
Inf
orm
atio
n
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
50 /
527.
5 Pr
otot
ype
Syst
ems
/ Com
pari
son
of R
T-T
hrea
ds a
nd Q
-Thr
eads
invo
catio
n tim
e
thre
ad a
ttrib
utes
invo
catio
n pe
riod
guar
ante
e of
exe
cutio
n
purp
ose
peri
odic
RT
-Thr
ead
Q-T
hrea
dus
er-d
efin
ed e
ntry
poi
nt is
cal
led
peri
odic
ally
fixe
d in
voca
tion
peri
odra
nges
for
per
iod
and
com
puta
tion
time
fixe
d (c
an b
e re
-spe
cifi
ed)
dyna
mic
, with
in r
ange
,ba
sed
on Q
oS c
ontr
ol p
olic
y
none
(po
ssib
le w
ith C
PUre
serv
atio
n)gu
aran
teed
, with
in th
e av
aila
ble
com
puta
tion
time
real
-tim
e pr
oces
sing
cont
inuo
us-m
edia
pro
cess
ing
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
51 /
527.
5 Pr
otot
ype
Syst
ems
/ YA
RT
OS
(Yet
Ano
ther
Rea
l-T
ime
Ope
ratin
g Sy
stem
)
Mic
ro k
erne
l im
plem
ente
d in
C o
n IB
M P
S/2
(Int
el 8
0386
pro
cess
or)
at
Uni
vers
ity o
f N
orth
Car
olin
a at
Cha
pel H
ill, N
C, U
SA.
YA
RT
OS
supp
orts
1.ta
sks,
i.e.
inde
pend
ent t
hrea
ds o
f co
ntro
l inv
oked
by
repe
ated
ly
occu
ring
eve
nts,
and
2. r
esou
rces
whi
ch s
ynch
roni
ze c
oncu
rren
t acc
ess
to s
hare
d da
ta.
Its
sche
dule
r gu
aran
tee
that
1. e
ach
task
invo
catio
n co
mpl
etes
pro
cess
ing
befo
re is
t dea
dlin
e, a
nd
2. n
o sh
ared
res
ourc
e is
acc
esse
d si
mul
tane
ousl
y by
mor
e th
an o
ne ta
sk.
7. M
ultim
edia
Ope
ratin
g Sy
stem
Multimedia Kommunikation SS‘99
52 /
527.
5 Pr
otot
ype
Syst
ems
/ YA
RT
OS
(Yet
Ano
ther
Rea
l-T
ime
Ope
ratin
g Sy
stem
)
It o
ffer
s tw
o se
para
te n
otio
ns o
f de
adlin
es f
or ta
sks,
1. O
ne f
or in
itial
aqu
isiti
on o
f th
e pr
oces
sor,
and
2. O
ne f
or e
xecu
tion
of o
pera
tions
on
reso
urce
s (t
o av
oid
prio
rity
inve
rsio
n).
It p
erfo
rms
on-l
ine
sche
dulin
g of
spo
radi
c ta
sks
exte
ndin
g th
e E
arlie
st D
eadl
ine
Firs
t alg
orith
m w
ith
sync
hron
ized
acc
ess
to s
hare
d re
sour
ces
on a
uni
proc
esso
r
syst
em.
YA
RT
OS
e.g.
ser
ves
for
a w
orks
tati
on-b
ased
aud
io a
nd v
ideo
con
fere
ncin
g
syst
em (
over
a 1
6 M
bit T
oken
Rin
g L
AN
).