You are on page 1of 32

Uni

t-4
Topi
c-RTOS
1.
EmbeddedOper
ati
ngSy
stems

An embedded oper ati


ng sy stem i
s an  operati
ng sy stem 
for
 embedded
comput ersy st
ems.Thi stypeofoper at
ingsy st
em i st y
pical
l
ydesignedt o
beresour ce-
effici
entandr eli
able.Resour ceef fi
ciencycomesatt hecostof
l
osing somef uncti
onalit
yorgr anulari
tyt hatl argercomput eroperati
ng
systems pr ovide,includi
ng f uncti
ons whi ch may notbe used by t he
special
ized appl i
cations t
heyr un.Dependi ng on t he method used for
multit
asking,thi stypeofOS i sfrequent l
yconsi deredt obea real
-t
ime
operati
ngsy stem, or RTOS.
Thehar dwar er unni nganembeddedoper atingsystem canbev eryli
mi t
ed
i
nr esourcessuchas  RAM  and 
ROM,t heref oreembeddeddesi gnoft hese
operati
ng sy st ems may hav e a narrow scope t ai
l
ored to a speci fi
c
appli
cationinor dert oachievedesir
edoper ationundertheseconst r
aints.I
n
ordertot ake bet teradv antage oft he pr ocessing poweroft he CPU,
software dev eloper s may wr it
e cr it
ical code di r
ectl
y in assembl y.
This machineef fici
ent  
l
anguagecanpot ent ial
lyresul
tingai nsi nspeed
and determinism  att hecostofpor tabi
lit
yandmai ntai
nabil
i
ty.Of t
ent i
mes,
embedded oper ati
ng sy stems ar e wr i
tten ent i
relyin mor e por tabl
e
l
anguages, li
ke  C, howev er
.
Ani mpor
tantdiff
erencebet weenmostembeddedoper at
ingsy st
emsand
desktopoperati
ngsy stemsi st hattheappli
cati
on,i
ncludi
ngt heoperat
ing
system,i
susuallystaticall
ylinkedtogetheri
ntoasingleexecutabl
eimage.
Unli
keadeskt opoper atingsy stem,theembeddedoper at
ingsy st
em does
notloadandexecuteappl i
cations.Thismeansthatthesystem isonl
yable
torunasingl
eapplication.
EmbeddedSy
stem

 Ref
erst
otheuseofel
ect
roni
csandsof
twar
ewi
thi
napr
oductt
hati
s
desi
gnedt
oper
for
m adedi
cat
edf
unct
ion.

I
nmanycases,embeddedsy
stemsar
epar
tofal
argersy
stem or
pr
oduct.

 Anti
lockbraki
ngsyst
em
i
n a carwoul d be an
example.
Fi
g:Possi
bleOr
gani
zat
ionofanEmbeddedSy
stem

Char
act
eri
sti
csofEmbeddedOS

 Real
-t
imeoper
ati
on

 React
iveoper
ati
on

 Conf
igur
abi
l
ity

I
/Odev
icef
lexi
bil
i
ty

 St
reaml
i
nedpr
otect
ionmechani
sms

Di
rectuseofi
nter
rupt
s
Dev
elopi
nganEmbeddedOS
Twogener
alappr
oaches:

 Takeanexi
sti
ngOSandadapti
tfort
heembeddedappl
i
cat
ion

 Desi
gnandi
mpl
ementanOSi
ntendedsol
elyf
orembeddeduse

Adapt
inganExi
sti
ngOS

Anexisti
ngcommer ci
al OScanbeusedf
oranembeddedsy
stem byaddi
ng:
r ealti
mecapabili
ty
 streamli
ningoperati
on
 addingnecessaryfunct
ional
i
ty
Advantage:

•Fami
l
iari
nter
face

Di
sadv
ant
age:

•Notopt
imi
zedf
orr
eal
-t
imeandembeddedappl
i
cat
ions.

Pur
pose-
Bui
ltEmbeddedOS
Ty
pical
char
act
eri
sti
csi
ncl
ude:

f
astandl
i
ght
wei
ghtpr
ocessort
hreadswi
tch

 schedul
i
ngpol
i
cyi
sreal
timeanddi
spat
chermodul
eispar
tof
schedul
er

 smal
lsi
zer
espondst
oext
ernal
int
err
upt
squi
ckl
y

 mi
nimi
zesi
nter
val
sdur
ingwhi
chi
nter
rupt
sar
edi
sabl
ed

 pr
ovi
desf
ixedorv
ari
abl
e-si
zedpar
ti
ti
onsf
ormemor
ymanagement
 provi
desspeci
alsequent
ial
fil
est
hatcanaccumul
atedat
aataf
ast
rate

EmbeddedLi
nux
 Aversi
onofLi nuxr unni
nginanembeddedsy stem.
 Embeddeddev icest ypi
cal
lyr
equir
esuppor tforaspecif
icsetof
dev
ices,peri
pherals,andprot
ocols,dependingont hehar
dwar et
hat
i
spresentinagi vendev i
ceandtheintendedpur poseofthatdev
ice.

 AnembeddedLi
nuxdist
ri
but
ionisaver
sionofLi
nuxtobe
cust
omizedf
ort
hesizeandhardwar
econstr
aint
sofembedded
devi
ces.

Incl
udessoft
war
epackagesthatsuppor
tav
ari
etyofser
vicesand
appli
cat
ionsont
hosedevi
ces.

 AnembeddedLi
nuxker
nel
wil
lbef
arsmal
l
ert
hananor
dinar
yLi
nux
ker
nel
.
Cr
ossCompi
ler

 Akeydif
ferenti
atorbet
weendeskt op/ser
verandembeddedLi
nux
di
stri
but
ionsisthatdesktopandser v
ersoft
wareist
ypi
call
ycompi
led
onthepl
atform whereitwil
lexecute.

 EmbeddedLi
nuxdi
str
ibut
ionsar
eusual
l
ycompi
l
edononepl
atf
orm
butar
eint
endedt
obeexecutedonanot
her
.

t
hesof
twar
eusedf
ort
hispur
posei
sref
err
edt
oasacr
oss-
compi
l
er.

Adv
ant
agesofEmbeddedLi
nux
 Vendori
ndependence-
 Theplat
form provi
derisnotdependentonapar t
icularvendort
o
provi
deneededf eat
uresandmeetdeadl i
nesfordepl oy
ment
 Vari
edhardwaresuppor t-
 Li
nuxsupportforawi derangeofprocessorar
chitecturesand
peri
pher
aldevicesmakesi tsui
tabl
eforvir
tual
lyanyembedded
syst
em
lowcost-
 TheuseofLinuxmini
mi zescostfordevel
opmentandtrai
ning
 Opensource-
 TheuseofLinuxpr
ovidesalloftheadvantagesofopensource
sof
tware

Andr
oid

Ti
nyOS
 St
reaml
i
nest
oav
erymi
nimal
OSf
orembeddedsy
stems.

 Cor
eOSrequi
res400by
tesofcodeanddat
amemor
ycombi
ned
Notar
eal
-t
imeOS.

 Ther
eisnoker
nel
Ther
ear
enopr
ocesses.

 OSdoesn’
thav
eamemor
yal
l
ocat
ionsy
stem .

I
nter
ruptandex
cept
ionhandl
i
ngi
sdependentont
heper
ipher
al.

Iti
scomplet
elynonbl
ocking,
sot
her
ear
efewexpl
i
cit
synchr
oni
zati
onpri
mit
ives.

 Hasbecomeapopul
arappr
oacht
oimpl
ement
ingwi
rel
esssensor
net
wor
ksof
twar
e.

Fi
g:Ty
pical
Wir
elessSensorNet
wor
kTopol
ogy
Ti
nyOSGoal
s
Wit
hthet
inydi
stri
but
edsensorappl
i
cat
ioni
nmi
nd,
thef
oll
owi
nggoal
s
wer
esetf
orTinyOS:

 al
l
owhi
ghconcur
rency

 oper
atewi
thl
i
mit
edr
esour
cesadaptt
ohar
dwar
eev
olut
ion

 suppor
tawi
der
angeofappl
i
cat
ions

 suppor
tadi
ver
sesetofpl
atf
orms

 ber
obust

Ti
nyOSComponent
s
 Embeddedsof t
war esy st
emsbuil
twi
thTinyOSconsistofasetof
modules(cal
ledcomponent s),
eachofwhichper
formsasi mple
taskandwhichinterfacewit
heachotherandwit
hhar dwar
ein
l
imitedandwell-
definedways.

 Theonl
yot
hersof
twar
emodul
eist
heschedul
er.

 Becauset
her
eisnoker
nel
ther
eisnoact
ual
OS.
 Theappl
i
cat
ionar
eaofi
nter
esti
sthewi
rel
esssensornet
wor
k
(WSN)

Fi
g:Ti
mer
M component
Component
s–Tasks
 Asof
twar
ecomponenti
mpl
ement
soneormor
etasks

 Eacht
aski
nacomponenti
ssi
mil
art
oat
hreadi
nanor
dinar
y
OS

 Wi
thi
nacomponentt
asksar
eat
omi
c

 onceat
askhasst
art
edi
trunst
ocompl
eti
on

2.
MULTI
TASKI
NG
 Mul t
itaski
ng is a method bywhi ch multi
plet asks,also known as
processes,shar ecommonpr ocessi
ngresourcessuchasaCPU.Wi t
ha
mul t
itaski
ng OS,such asWi ndowsXP,y ou can simultaneouslyrun
mul t
ipleapplicati
ons.Mul
ti
taskingreferstot heabi li
tyoft heOS t o
quicklyswitchbetweeneachcomput ingtasktogi vetheimpressionthe
di
fferentappli
cat i
onsar
eexecutingmulti
pleacti
onssi mult
aneously.

 AsCPU cl ockspeedshaveincreasedst eadil


yov erti
me,notonlydo
appl
icati
onsrunfaster
,butOSscanswi tchbetweenappli
cati
onsmore
qui
ckly.Thisprovi
desbetteroverallperfor
mance.Manyact i
onscan
happenatonceonacomput er,
andi ndi
vidualappl
icati
onscanrunf
ast
er.
2.
1Si
ngl
eCor
e
 Wit
hasingl
eCPUcor e,onlyonetaskrunsatanypointinti
me,meaning
t
hattheCPUisact
ivel
yex ecut
inginst
ruct
ionsf
orthattask.Mul
ti
taski
ng
sol
vesthi
sproblem byschedul
ingwhi
chtaskmayr
unatanygi
vent
ime
andwhenanotherwaiti
ngtaskgetsat
urn.

Fi
g:Si
ngl
e-cor
esy
stemsschedul
etaskson1CPUt
omul
ti
task

2.
2Mul
ticor
e
 When r unning on a multicor e system,mul t
itasking OSs can t
rul
y
executemul t
ipl
etasksconcur r
ently.Themul t
iplecomput i
ngengines
workindependent l
yondiff
er enttasks.
 Forexampl e,onadual -
cor esy stem,fourappli
cations–suchaswor d
processing,e-mail
,Webbr owsing,andant i
vir
ussof tware– caneach
accessasepar at
eprocessorcor eatthesamet ime.Youcanmul t
it
ask
bychecki nge- mai
landt ypingal et
tersi
multaneousl y,t
husimprovi
ng
overal
lper f
ormanceforappl i
cations.

Fi
g:Dual
-cor
esy
stemsenabl
emul
ti
taski
ngoperati
ngsy
stemst
oexecut
etwot
asks
si
mul t
aneousl
y

 TheOSexecut esmulti
pleappli
cati
onsmor eeffi
cient
lybyspli
tt
ingthe
di
ff
erentapplicati
ons,orprocesses,betweentheseparat
eCPUcor es.
The computercan spr ead t
he wor k– each corei smanaging and
swi
tchingthroughhal fasmanyappl i
cati
onsasbef ore– anddeliver
bet
terover
allt hr
oughputandper formance.Ineff
ect,theappli
cati
ons
ar
erunni
ngi
npar
all
el.

3.MULTI
THREADI
NG
 Multi
threadi
ngext endstheideaofmultit
askingint
oappl i
cat
ions,soyou
can subdivide specif
ic operati
ons withi
n a single appl
icat
ion i
nto
i
ndividualt
hreads.Eachoft hethr
eadscanr uninparal
lel
.TheOSdi vi
des
processingtimenotonl yamongdi f
ferentappli
cati
ons,butalsoamong
eacht hr
eadwi t
hinanapplicati
on.

 Mult
it
hreadedappl
i
cati
ons,multi
plet
askscanpr
ogr
essi
npar
all
elwi
th
ot
herappli
cati
onst
hatarerunni
ngonthesyst
em.

Fi
g:Dual
-cor
esy
stem enabl
esmul
ti
thr
eadi
ng

Appli
cati
onsthatt
akeadvant
ageofmul
ti
thr
eadi
nghav
enumer
ous
benef
its,
incl
udi
ngthefol
l
owing:

 Bettersy st
em r eli
abili
ty.
 I
mpr ov edper formanceonmul ti
pr ocessorcomput ers.
 I
nmanyappl i
cations,youmakesy nchronouscal l
st oresour ces,
suchas
i
nstrument s.Thesei nstrumentcal lsoftentakeal ongtimet ocompl ete.
I
nasi ngle-thr
eadedappl icati
on,asy nchronouscal leff
ectivelybl
ocks,or
prevents,anyot hertaskwi thi
nt heappl i
cati
onf rom execut ingunt
ilthe
operationcompl etes.
 Multit
hr eadingpr eventsthisblocki ng.
 Whilet he sy nchr onouscal lrunson one t hread,ot herpar tsoft he
program t hatdo notdepend on t hi
s callrun on di ff
erentt hr
eads.
Execut i
onoft heappl icationpr ogr essesinsteadofst al
linguntilthe
synchr
onouscallcompl etes.Inthi
sway,amulti
thr
eadedappl
icat
ion
maximizestheeffi
ciencyoft heCPU becausei
tdoesnotidl
ei fany
thr
eadoftheappl
icationisreadyt
orun.

3.
REALTI
ME-OPERATI
NGSYSTEM

Realti
me  Operati
ngSy stems 
arev eryf
astandqui ckrespondentsystems.
Thesesystemsar eusedi nanenv i
ronmentwher ealargenumberofev ents
(gener
all
yext er
nal)mustbeaccept edandpr ocessedi nashorttime.Real
ti
mepr ocessingrequir
esquicktransacti
onandchar acter
izedbysuppl y
ing
i
mmedi ater esponse.Forexampl e,a measur ementf r
om a pet rol
eum
refi
ner
yindicati
ngt hattemper
atureisgettingtoohighandmi ghtdemand
forimmediateattenti
ontoavoi
danexpl osion.
I
nr ealti
me oper
ati
ng syst
em thereisal i
tt
le swapping ofpr ograms
between pri
maryand secondar
y  
memor y
.Mostoft het ime,processes
remaininprimarymemoryinor dertoprovi
dequi ckresponse,t
her efor
e,
memor ymanagementinrealti
mesy stem i
slessdemandi ngcompar edt o
othersyst
ems.
Thepr
imar
yfunct
ionsoft
her
eal
-t
imeoper
ati
ngsy
stem ar
eto:
 
1.Manage the pr
ocessorand ot
hersy
stem r
esour
ces t
o meett
he
r
equi
rement
sofanappli
cati
on.
2.Sy
nchr
oni
zewi
thandr
espondt
othesy
stem ev
ent
s.
3.Movethedat
aeffi
cient
lyamongpr
ocessesandt
oper
for
m coor
dinat
ion
amongthesepr
ocesses.
 
TheRealTi me sy stemsar eused i nt heenv i
ronment swher eal arge
numberofev ents(general
lyexternaltot he 
comput er system)i srequi
redt o
beacceptedandi stobepr ocessedi nt heform ofqui ckresponse.Such
syst
emshav et obet hemul ti
tasking.Sot hepri
mar yfunctionoft her eal
ti
meoper ati
ngsy st
em ist omanagecer t
ainsystem r esources,suchas
the 
CPU,memor y,andt ime.Eachr esourcemustbeshar edamongt he
competingpr ocessest oaccompl isht heov eral
lf unctionoft hesy stem
Apartfr
om thesepr i
mar yfuncti
onsoft hereal-
ti
meoper ati
ngsy stem there
arecert
ainsecondar yfuncti
onst hatar enotmandat orybutar eincludedt o
enhancetheper f
ormance:
1.Topr
  ovi
deanef
fi
cientmanagementof
 RAM.
2.Topr
ovi
deanexcl
usi
veaccesst
othecomput
err
esour
ces.
 
Thet er
m realti
meref
erst othet echni
queofupdat i
ng fi
l
eswitht
he
t
ransact
iondat
aimmedi
atel
yjustaf
tertheeventt
hati
trel
ateswit
h.
 
Fewmor
eexampl
esofr
eal
timepr
ocessi
ngar
e:
 
1.Ai
rl
inesr
eser
vat
ionsy
stem.
2.Ai
rtr
aff
iccont
rol
syst
em.
3.Sy
stemst
hatpr
ovi
dei
mmedi
ateupdat
ing.
4.Sy
stemst
hatpr
ovi
deupt
othemi
nut
e i
nfor
mat
ion 
onst
ockpr
ices.
5.Def
enseappl
i
cat
ionsy
stemsl
i
keasRADAR.
 
Real t
imeoper ati
ngsy st
emsmost lyuset hepreemptivepri
orit
yscheduli
ng.
Thesesuppor tmor ethanoneschedulingpol i
cyandof t
enallowt heuserto
setparamet ersassociatedwithsuchpol i
cies,suchast het i
me- sl
i
cei n
RoundRobi nscheduli
ngwher eeacht askint hetaskqueueisscheduledup
toamaxi mum t ime,setbyt hetime-sli
cepar ameter,inar oundr obin
manner.Hundr eds oft he pri
ori
tyl evels are commonl y avail
abl
ef or
scheduli
ng.Some speci f
ictasks can al so be indicat
ed to be non-
preempti
ve.
Fi
g:oper
ati
veenv
ironmentf
orRTOS

Atypical
real
-t
imeoperati
ngenvir
onmentisshowninFigur
e8.1.I
nthi
s
fi
gurewenotethatthecomputersyst
em hasanint
erf
acewhichis
embeddedwithini
tsenvi
ronment.

Char
act
eri
sti
csofr
eal
-t
imesy
stems

Theoper
ati
ngsy
stem achi
evest
hedesi
redext
entofr
egul
ati
onasf
oll
ows:

1.Senseanev ent:Thesystem monitorsi t


soperativ
eenv i
ronmentusing
somesensor s.Thesesensorskeepat abonsomemeasur ableenti
ty.
Dependingupont hecontextofuset hisent
itymaybeameasur eof
temperature,orastockpr i
cefluct
uat i
onorf l
uidleveli
nar eserv
oir
.
Thesemeasur ementsmaybeper iodic.Int
hatcaset hesystem would
acceptani nputperi
odicall
y.I
ncaset hemeasur ementofinputsis
takenatspeci f
iedti
mesofoper at
iont hentheOSmayschedul eits
i
nputatt hesespecifi
edt i
mesori tmaybei nt
erruptedtoacceptthe
i
nput.Thei nputmayev enbemeasur edonlywhenanunusual
dev iati
oni nt hev al
ueoft hemoni tor edent i
tyoccur s.Int hesecases,
thei nputwoul dcer t
ainlyresultinani nt errupt .Regar dlessoft hei nput
mode, thesy stem woul dhav eani nputf ollowi ngasensorr eadi ng
(whi chisanev ent).
2.Processt hedat a:Thenexti mpor tantt aski st opr ocesst hedat a
whi chhasbeenmostr ecent l
yacqui red.Thedat apr ocessi ngmaybe
aimedatchecki ngtheheal thoft hesy st em.Usual lyitist odet ermine
i
fsomeact ioni sneeded.
3.Deci deonanact i
on:Usual l
y,thepr ocessi ngst epsinv olvingar ri
v i
ng
atsomedeci sionsoncont r
ol setti
ngs.Fori nst ance, i
ft hest ock
pricescr osssomet hreshold, thenonehast odeci det obuyorsel lor
donot hing.Asanot herexampl e,theact i
onmaybet oopenav alvea
l
ittl
emor etoi ncreaseinflowi ncaser eser v oirleveldrops.
4.Takeacor r
ectiveact i
on:Incase, theset tingsneedt obeal ter ed, t
he
newset tingsar edeterminedandcont rol act uat orsarei niti
at ed.Not e
thatt heact ionsi nturnaffectt heenv ironment .Itisqui tepossi ble
thatasaconsequence, anewsetofev ent sgett r
igger ed.Al so, iti
s
possi blet hatthecor recti
vest epr equi resadr ast i
candani mmedi ate
step.Fori nstance, i
fanal armi stober aised, thenal ltheot hert asks
hav et obesuspendedorpr e-empt edandanal armr aised
i
mmedi ately.Real ti
mesy stemsqui teof tenr esor ttopr e-empt i
ont o
prev entacat astrophef r
om happeni ng.

Cl
assi
fi
cat
ionofReal
-t
imeSy
stems

Thecl assifi
cationofr eal-ti
mesy stemsi susual l
ybasedont hesev erityof
theconsequencesoff ail
ingt omeett imeconst raints.Thi scanbe
under st
oodasf ol l
ows. Supposeasy stem requi resar esponset oanev ent
i
nt imeper iodT.Nowweask: whathappensi fther esponsei snotr ecei ved
withinthest ipul
at edt imeper i
od?Thef ail
uret omeett het imeconst raint
mayr esulti ndifferentdegr eesofsev erityofconsequences.I nali
f e-crit
ical
orsaf et
ycr iti
calappl ication,thef ai
luremayr esul tinadi sastersuchas
l
ossofl i
fe.Acasei npoi ntist heshuttleCol umbi a'sacci dentinear ly
February22003.Recal l KalpanaChawl a,anaer onaut icsengi neeri
ngPh.D.
wasonboar d.Dur ingi tsdescent, about16mi nut esf rom landi ng,
the
spacecr afttemper aturer oset odanger ousl evelsr esul t
ingi nacat ast r
ophi c
endoft hemi ssion.
Clearly,
ther iseint emper atureasaspacedr aftent ersear th'satmospher e
i
sant ici
pat ed.Spacecr af t
shav eRTOSr egulatingt hecont roll
erstor espond
tosuchsi tuat i
onsf r
om dev eloping.Andy etthecool ingsy stem( s)
int his
casedi dnotof fertimelymitigati
on.Bot hint ermsofl ossofhumanl i
feand
thecostofmi ssionsuchaf ail
urehast hehi ghestsev erit
yofconsequences.
Wher easi nt hecaseofanonl inestockt r
adi ng,oragameshow, itmay
meanaf i
nanci allossorami ssedoppor tunity.Inthecaseofadr opped
packeti nav ideost r
eamingappl icati
onitwoul dsimpl ymeanagl itchanda
perhapsat empor arydropint hepicturequal it
y.Thet hreeexampl esofr eal
-
ti
mesy stem wehav egivenher ehav ediff
erentl evel
sofsev er
it
yint er
msof
ti
melyr esponse.Thef i
rstonehasl i
fe-
threateningi mpl i
cati
on;thesecond
caser eferstoami ssedoppor tunit
yandf inally,degradedpicturequalityin
viewi
ng.Associ atedwiththesear ethebr oadl yaccept edcategori
eshar d,
fi
rm andsof treal-ti
mesy stems.

Ar
chi
tect
ureofReal
-t
imeSy
stems:
Thebasi carchitectur
eofsuchsy st
emsi ssimple.
Asshowni nFigur e8.1,somesensor spr ovi
deinputfrom t
heoper ati
ve
environmentandacomput ati
ondet er mi
nest herequi
redcontrol
.Final
ly,
an
appropr i
ateact uatorisacti
vated.Howev er
,sincetheconsequenceof
fai
luret orespondt oeventscanbecat astr
ophic,i
tisimportantt
obui l
din
thefollowingt wof eaturesint
hesy st em.
(a)Itshouldbeaf aul
ttoler
antdesign.
(b)Theschedul ingpol i
cymustpr ovideforpre-emptiv
eaction.

Foraf aulttol
erantdesi gn,t
hest r
ategi
esmayi ncludemaj ori
tyvoti
ngoutof
thefaultysensors.Sy stemsl i
kesatell
it
eguidancesy st
em, usuall
yhave
back-up( orahot-stand-by)system tofall
backupon.Thi sisbecausethe
costoff ail
ureofami ssionissimplytoohigh.Designer
sofAi rbusA-320
hadpeggedt hefigureoff ai
lureprobabil
it
yatlowerthan10- 10forone-hour
peri
odi nf l
i
ght.
Fi
g:pr
ior
it
yst
ruct
uref
orRTOSt
asks

Asf ordesi gnofschedul ingpol icy,onef i


rsti dent i
fiest hecr iti
calf unct ions
andnotsocr i
ticalf unctionswi thi
nanoper at i
on.Theschedul i
ngal gorithm
ensurest hatt hecr iti
calf unctionsobt ai
nhi ghpr iori
tyi nterruptst oel i
cit
i
mmedi at eresponses.I nFi gure, wedepi ctt hepr i
or i
tyst ructuref orsucha
design..Yet ,itiswor t
hment ioningher et hatRTOSdesi gner shav et wo
basicdesi gnor i
ent ati
onst oconsi der.Onei st ot hi
nki nt ermsofev ent -
tr
iggeredoper ati
onsandt heot heri st othinkoft ime- t
rigger edoper ations.
Theseconsi derat i
onsal sodet ermi nei tsschedul i
ngpol i
cy.ar chitect ures
basedont heset woconsi derations.Theobser v
ationi sthatt ime-trigger ed
archit
ect uresobt aingr eaterpr edictabil
i
tyb utendupwast ingmor er esour ce
cyclesofoper ati
onduet omor ef requentpr e-empt ions.Ont heot herhand,
event-tr
igger edsy stem ar chitecturesseem t oscor eint er msoft hei rabi l
it
y
toadaptt oav arietyofoper atingscenar ios.Ev ent -
tri
gger edsy stemsar e
generallybet t
ersui tedforasy nchr onousi nputev ents.Thet i
me- trigger ed
systemsar ebet t
ersui t
edf orsy stemswi thper i
odici nput s.Fornow,l etus
exami nemi cro-ker nelswhi char eatt hehear tofRTOS, ev ent-tr
igger edor
ti
me- t
rigger ed.
Adv
ant
agesofRealTi
meOper
ati
ngSy
stem:
-
Ther
ear
esomeoft
hef
eat
uresofusi
ngRTOSt
hati
sdescr
ibedbel
ow

 Maxi mum Consumpt i


on:– RTOSgivemaxi mum consumpt i
onofthe
system andgivesusmor eoutputwhil
eusingall
theresourcesand
keepingalldevicesacti
ve.Thereisl
itt
leornodownt i
mei nthese
systems.Soi tcanbeal sousi
ngbyt heserver
sthatarehostedto
givemaximum out putofhostingcompanies.

 TaskShift
ing:– Ther
eisveryli
tt
letimeassi
gnedtoshift
ingt
asksin
thesesyst
ems.Forexampl einoldersyst
emsitt
akesabout10mi cr
o
secondsinshif
ti
ngonet asktoanotherandi
nlatestsy
stemsitt
akes
3mi cr
oseconds.

 FocusonAppl ication:– Thesetypeofoperat
ingsystem f
ocuson
appli
cati
onswhi char erunningandusual
lygivelessimport
anceto
otherappl
icat
ionr esidi
nginwai t
ingst
ageoflifecycl
e.Soless
appli
cati
onsort asksar emanagedandgi veexactresul
toncurrent
executi
onwork.

 Realt
imeoperati
ngsyst
em i
nembeddedsyst
em:– Duetosmal
l
si
zeofprogr
amsRTOScanal sobeusedi
nembeddedsyst
emsli
kei
n
tr
anspor
tandothers.

 Er
rorFree:– RTOSiser
rorf
reet
hatmeani
thasnochancesofer
ror
i
nperformingtasks.

 24-7sy
stems: –RTOScanbebestusedforanyappl
icat
ionswhi
ch
run24hoursand7daysbecausei
tdolesstaskshi
ft
ingandgive
maximum output
.

Di
sadv
ant
agesofRealTi
meOper
ati
ngSy
stem:
-

 Li
mi t
edTasks:– Thereareonlyli
mi t
edtasksrunatt
hesametime
andtheconcent
rati
onoft hesesystem areonfewappli
cat
iont
o
avoi
derror
sandot hertaskhavetowai t
.Somet i
mether
eisnoti
me
l
imitofhowmucht hewai ti
ngtaskshavet owait
.
 Useheav
ysy st
em r
esour
ces:– 
RTOSusedlotofsy
stem r
esour
ces
whi
chisnotasgoodandisal
soexpensi
ve.

 Lowmul ti
-t
asking:– Mult
itaski
ngisdonefewoft i
mesandt hisis
themaindisadvantageofRTOSbecauset hesesystem runsfew
tasksandstayfocusedont hem.Soiti
snotbestforsy stemswhich
uselotofmulti
-thr
eadingbecauseofpoorthreadprior
it
y.

Real
4. -Ti
meLi
nuxI
ntr
oduct
ion
Ifyouwant edtocont rolacamer aorar obotorascient i
fi
ci nst rumentf rom
a PC,i twoul d be nat uraltot hinkofusi ng Linuxso y ou coul dt ake
advantageoft hedev elopmentenv i
ronment,theXWi ndowSy st em,andal l
thenetworkingsuppor t
.Howev er,Li
nuxcannotr eli
ablyrunt heseki ndsof
hardreal-
ti
meappl i
cat ions.Asi mpleexperimentcani ll
ustratet heproblem.
Takeaspeakerandhooki tupt ooneoft hepinsfrom thepar allelport;then
runapr ogram thatt ogglest hepin.Ifyourprogram istheonl yoner unni ng,
thespeakerwi l
lpr oduceani ce,somewhatst eadyt one—notcompl et ely
steady,butnotbad.WhenLi nuxupdat esthef il
esy stem ev er ycoupl eof
seconds,youmi ghtnot i
ceasmal lchangeint one.I
fy oumov et hemouse
overacoupl eofwi ndowst hetonebecomesi rregul
ar.Ifyoust ar tNetscape,
youheari nt er
valsofsi l
enceasy ourprogram wai tsf orhi gherpr iority
processest orun.

Theprobl
em i
sthatLi
nux,likemostgener
alpur
poseoper
ati
ngsyst
ems,is
desi
gnedtoopti
mi av
ze  erage per
for
manceandtotr
ytogi
veever
yprocess
afai
rshar
eofcomputet i
me.

Thisisgreatforgener
al purposecomputing,butf
orreal
-t
imeprogramming,
precise t
iming and predictabl
e perf
ormance are moreimpor t
antthan
average perfor
mance.Forexampl e,ifa camer af i
ll
s a buff
erev er
y
mill
isecond,amoment arydel ayinthepr ocessreadi
ngthatbuf f
ercan
causedat al
oss.
I
fast eppermot ori
nal it
hographymachi
nemustbet ur
nedonandoffat
preci
sei nter
valsinordertomi nimi
zevi
brat
ionandtomoveawaf eri
nto
posit
ion at t he correct t
ime,a moment ar
y del
ay may cause an
unrecoverabl
ef ai
lur
e.

Consi
derwhatmi
ghthappeni
fthet
askt
hatcausesanemer
gencyshut
downofa
chemi
str
yexper
imentmustwai
ttor
ununt
ilNet
scaper
edr
awst
hewi
ndow.

I
tt ur
nsoutr edesi
gni
ngLi nuxtopr ovi
deguar ant
eedper formancewoul d
takeanenor mousamountofwor k.Andt akingonsuchaj obwoul ddefeat
ourori
ginalpurpose.I
nsteadofhavinganof ft
heshelfgeneralpur
poseOS,
wewoul dhav eacustom-madespeci alpurposeOSt hatwouldnotberiding
thewaveoft hemai nLi
nuxdev el
opmentef fort
.

Sowesl ippedasmal l
,simple,real
-ti
meoper atingsystem underneat
h Li
nux.
Linuxbecomesat askrunonl ywhent herei snor eal
-ti
met asktorun,and
wepr e-emptLinuxwhenev erar eal-
ti
met askneedst heprocessor.The
changesneededi nLinuxitselfareminimal.Linuxismost lyunawareofthe
real-
ti
me operati
ng sy st
em as i tgoes abouti ts busi
ness ofr unni
ng
processes,cat
chi
ngi nt
errupts,andcontrol
li
ngdev ices.Real-
ti
met askscan
runatquiteahighlevelofprecisi
on.

InourtestP120system,wecanschedul
etaskst
orunwi
thi
napr
eci
sionof
about20mi cr
oseconds.

Real
-t
imeLi
nuxi
sar
esear
chpr
ojectwi
tht
wogoal
s.

 Fir
st,wewantapr actical
,non-pr opr
ietarytoolwecanuset ocont r
ol
sci
ent i
fi
cinstrumentsandr obot s.
 Ourot hergoalist ouseRT- Li
nuxf orr esearchinr ealandnon-real
-
ti
meOSdesi gn.We' dli
ketobeabl etol earnsomet hingabouthowt o
makeoper at
ingsy st
emsef fi
cientandr el
iabl
e.Forexampl e,evena
non-real
-t
imeoper ati
ngsy stem shoul dbeabl et odet er
minewhether
i
tcanguar anteetimingneededf orit
sI/Odev i
ces.

We'real
sointer
estedi
nwhattypesofscheduli
ngdisci
pli
nesact
ual
l
yturn
outtobet hemostusefulf
orreal-
ti
meapplicati
ons.Foll
owi
ngthi
sdual
pur
pose,inthispaperwedi
scussbot hhow touseRT- Li
nuxandhow i
t
works.

Usi
ngRT-
Linux2.
0.RT.
1
Letusconsideranexample.Supposewewanttowrit
eanappli
cat
iont
hat
poll
sadevicefordatainreal
-t
imeandstorest
hisdat
ainafi
le.Themain
desi
gnphil
osophybehindRT-Li
nuxisthef
oll
owi
ng:

 Real-t
imepr ogramsshoul dbespl i
ti ntosmal l
,simpl epar ts,wi t
hhar d
real-
ti
meconst rai nts,andl ar gerpar tst hatdomor esophi sticated
processi ng.
 Foll
owi ngt hispr inciple,wespl itourappl i
cationi ntot wopar t
s.The
hard-r
eal -t
imepar twi llexecut easar eal-t
imet askandcopydat a
from thedev icei ntoaspeci al I
/Oi nt erfacecal led real-t
imef i
fo.
 Themai npar toft hepr ogr am wi l
lexecut easanor di naryLi nux
process.Thi spar twi l
lreaddat af rom t heot herendoft her eal-ti
me
fi
foanddi splayandst oret hedat ainaf i
le.
 Ther eal-ti
met askst r
uctur ewi llcont ainpoi nterst ocode,dat a,and
schedul i
ngi nformat ionf ort hist ask.Thet askst r
uct ureisdef i
nedi n
thef i
rsti ncl
udef il
e.Cur r
ent ly,RT- Linuxhasonl yonef air
lysi mple
schedul er.
I nthef uture,theschedul erswi l
lal sobel oadabl emodul es.Cur r
ent l
y,
the onl y way f or r eal-t
ime t asks t o communi cate wi th Li nux
processesi st hroughspeci al queuescal ledr eal-
timef i
fos.
 Real-t
imef if
oshav ebeendesi gnedsot hatt her eal-ti
met askwi l
l
neverbebl ockedwheni treadsorwr i
tesdat a.Fi gure1i l
lustratesr eal
-ti
mef ifos.

Fi
gur
e1:
.Real
-t
imef
if
os.
HowRT-
LinuxWor
ks

Thebasici deaist omakeLi nuxrunundert


hecontrolofareal
-t
imekernel
(SeeFigure2).Whent hereisreal-
ti
meworktobedone,t heRToperati
ng
system runsoneofi t
stasks.Whent her
eisnoreal-
ti
mewor ktobedone,
thereal
-timeker nelschedul
esLinuxt or
un.SoLinuxisthelowestpr
iori
ty
taskoftheRT- kernel
.

Fi
gur
e2

Thepr oblem wi thLi nuxdisabli


nginter
r uptsissolvedbysi mulati
ngthe
Li
nux i nterrupt-
rel
ated routi
nes int he r eal
-t
ime ker nel
.Forexampl e,
whenev ert heLinuxker nelinvokesthe cli
()- r
outi
net hatissupposedto
di
sablei nterrupt
s, sof
a  t
wareinter
ruptfl
agi sreseti
nstead.

Allinterr
uptsgetcaughtbytheRT-kernelandpassedtotheLi
nuxker
nel
accordingtothest
ateofthi
sflagandtheinter
ruptmask.

Therefor
e,thei
nter
ruptsarealwaysavai
lablefortheRT-ker
nel
,whi
l
estil
l
all
owingLinuxto“disabl
e”them.Intheexampl eabove,theLi
nuxker
nel
routi
newouldcal
lcl
i() 
tocl
earthesof
tint
erruptfl
ag.

I
faninter
ruptoccur
red,thereal
-t
imeexecutivewouldcat
chitanddecide
whatt
odo.I ft
heinter
ruptcausedareal-
timetasktoberun,t
heexecut
ive
woul
dsavet hestat
eofLi nuxandstartthereal
-t
imetaskimmediat
ely
.If
the i
nterr
uptjustneeded to be passed along to Linux,t
he real-
ti
me
executi
vewouldsetaflagshowi ngapendingint
errupt,thenr
esumeLi nux
executi
on wit
houtrunning the Li
nuxinterr
upthandler .When Linuxr e-
enables i
nter
rupt
s,the real -
ti
me executi
ve willpr ocess allpending
i
nterr
uptsandcausethecorrespondingLi
nuxhandlerstoexecute.

Thereal
-t
imekernelisit
sel
fnon-preemptabl
e,butsi nceit
srout
inesare
ver
ysmallandfast,t
hisdoesnotcausebigdelays.Testi
ngonaPent i
um
120showsthemaximum schedul
ingdelayt
obel essthan20ms.

Real-
ti
metasksrunatt hekernelpri
vil
egelevelinor
dertopr ov
idedirect
accesstot
hecomput erhardware.Theyhavef i
xedal
locat
ionsofmemor y
forcodeanddata—otherwise,wewoul dhavet oall
ow forunpredi
ctable
delay
swhenataskrequestsnewmemor yorpagesinacodepage.

Real-
ti
met askscannotuseLi nuxsy stem cal
lsordi r
ectl
ycal lrouti
nesor
accessordinarydat
astructur
esi ntheLinuxkernel
,asthiswoul dintroduce
thepossibi
lit
yofinconsi
stencies.Inourexampl eabove,thekernelrout i
ne
changingthequeuewoul dinvoke cl
i,
butthiswouldnotpr eventar eal-
ti
me
taskfr
om starti
ng.

Sowecannotal l
owthereal-
ti
met asktodi
r ectl
yaccessthequeue.Wedo,
however,needawayf orreal-
ti
met askstoexchangedat awit
hthekernel
andwithusertasks.I
nadat acoll
ecti
onappl i
cati
on,forexample,
wemight
needtosendt hedatacoll
ectedbyanRT- t
askov ert
henet wor
k,orwr
it
ei t
l
ocall
ytoaf i
le,whi
ledi
spl
ayingitonthescreen.

Real-
ti
mef ifosareusedt opassi nf
ormationbetweenr eal
-ti
mepr ocesses
andordinaryLinuxprocesses.RT-
fif
os,l
ikereal
-ti
met asks,ar
eneverpaged
out.Thiseliminatesthepr obl
em ofunpr edi
ctabl
edel aysduet opaging.
Andreal-t
imef i
fosaredesignedtoneverblockthereal-
timetask.

5.
RTOSKer
nel
RTOS Ker
nelprov
idesanAbstracti
onlay
ert hathi
desfr
om applicat
ion
sof
twaret
hehardwaredet
ail
soft hepr
ocessor/setofprocessor
supon
whi
chtheappl
i
cati
onsoft
wareshallr
un.

1.Thecent
ral
componentofmostOper
ati
ngsy
stemsi
scal
l
edKer
nel
.

2.Theker
nel
managessy
stem'
sresour
cesandt
hecommuni
cat
ion
3.Thekernelprovidesthemostbasici
nter
facebet
weent
hecomput
er
i
tsel
fandt
her estoftheoper
ati
ngsy
stem.

4.Theker
nel
isr
esponsi
blef
ort
hemanagementoft
hecent
ral
processor
.

5.Theker nelincl
udesthedispatchert
oall
ocatethecentralpr
ocessor
,to
determinethecauseofani nter
ruptandini
ti
ateit
spr ocessi
ng,andsome
provisi
onforcommuni cati
onamongt hevari
oussy stem andusertasks
current
lyactiv
ei nt
hesystem.

6.Ker
nel
ist
hecor
eofanoper
ati
ngsy
stem.

Basi
cfunct
ionsofRTOSker
nel

1.Taskmanagement

2.Tasksy
nchr
oni
zat
ion

•Av
oidpr
ior
it
yinv
ersi
on

3.Taskschedul
i
ng

4.I
nter
rupthandl
i
ng
5.Memor
ymanagement

•nov
irt
ual
memor
yforhar
dRTt
asks

6.Except
ionhandl
i
ng

Task
1.At ask(
alsocal
ledat
hread)i
sapr
ogr
am onacomput
erwhi
chcanbe
execut
edandrun.

2.A t
askisani ndependentthreadofexecuti
onthatcancompet
ewi
th
ot
herconcur
rentt
asksforprocessorexecut
iont
ime.

3.At
aski
sschedul
abl
e.

4.Thedesi
gnpr
ocessforareal
-t
imeappl
icati
oninv
olvesspli
tt
ingthework
t
obedoneintot
askswhicharer
esponsi
bleforaport
ionoftheproblem.

5.Eachtaski
sassi
gnedapr
ior
it
y,i
tsownsetofCPUr
egi
ster
s,andi
tsown
st
ackarea.

TaskManagement
1.Setofser
vicesusedt
oal
l
owapplicat
ionsof
twaredev
eloper
stodesi
gn
t
hei
rsoft
wareasanumberofsepar
atechunksofsof
twar
eeachhandl
inga
di
sti
nctt
opi
c,adi
sti
nctgoal
,andsomet
imesi
tsownr
eal
-t
imedeadl
i
ne.

2.Mai
nser
viceof
fer
edi
sTaskSchedul
i
ng

1.cont
rol
stheexecut
ionofappl
i
cat
ionsof
twar
etasks

2.canmaket
hem r
uni
nav
eryt
imel
yandr
esponsi
vef
ashi
on.

Pr
ior
it
y-BasedKer
nel
s
•Ther
ear
etwot
ypes

–Non-
preempt
ive

•Per
for
m “cooper
ati
vemul
ti
taski
ng”–Eacht askmustexpl
i
cit
lygi
veup
cont
roloftheCPU–Thismustbedonefrequentl
ytomai
ntai
ntheil
l
usi
on
ofconcur
rency

•Asynchronousevent
sar
esti
llhandl
edbyI
SRs–ISRscanmakeahigher
-
pr
ior
itytaskreadyt
orun–ButISRsal
waysr
etur
ntot
heint
err
upt
edt
ask

Fi
g:nonpr
eempt
ive

Adv
ant
agesofNon-
Preempt
iveKer
nel
s

•I
nter
ruptl
atencyi
sty
pical
l
ylow

•Canusenon-r
eentr
antfunct
ionswit
houtfearofcor
rupt
ionbyanother
t
ask–Becauseeachtaskcanruntocomplet
ionbef
oreitr
eli
nqui
shesthe
CPU–However,non-
reent
rantf
unct
ionsshoul
dnotbeal
l
owedt
ogi
veup
cont
rol
oft
heCPU

•Task-
responsei
snowgi
venbyt
het
imeoft
hel
ongestt
ask–muchl
ower
t
hanwithF/Bsyst
ems

•Lessneedt
oguar
dshar
eddat
athr
ought
heuseofsemaphor
es

–Howev
er,
thi
srul
eisnotabsol
ute.

– Shared I
/O dev
ices can st
il
lrequi
ret
he use ofmut
ualexcl
usi
on
semaphores

–At
askmi
ghtst
il
lneedexcl
usi
veaccesst
oapr
int
er

Di
sadv
ant
agesofNon-
Preempt
iveKer
nel
s

•Responsi
veness

–Ahi
gherpr
ior
it
ytaskmi
ghthav
etowai
tforal
ongt
ime

–Responset
imei
snondet
ermi
nist
ic

•Ver
yfewcommer
cial
ker
nel
sar
enon-
preempt
ive

Pr
eempt
iveKer
nel
s
•Thehi
ghest
-pr
ior
it
ytaskreadytorunisalwaysgi
vencont
roloftheCPU–
I
fanISR makesahi gher
-prior
itytaskready
,thehigher
-pri
ori
tytaski
s
r
esumed(i
nst
eadoftheint
erruptedt
ask).

•Mostcommer
cial
real
-t
imeker
nel
sar
epr
eempt
ive.
Fi
g:pr
eempt
iveker
nel

Adv
ant
agesofPr
eempt
iveKer
nel
s

•Execut
ionoft
hehi
ghest
-pr
ior
it
ytaski
sdet
ermi
nist
ic

•Task-
lev
elr
esponset
imei
smi
nimi
zed

Di
sadv
ant
agesofPr
eempt
iveKer
nel
s

•Shouldnotusenon-
reent
rantf
unct
ionsunl
essexcl
usi
veaccesst
othese
funct
ionsi
sensur
ed.

6.Realt
imeschedul
ing
Thet erm  schedul i
nganal y si
s in real-t
imecomput ingi ncludest heanal y
sis
andt estingoft he schedul er 
sy stem andt he 
algori
t hms  usedi nreal
-time
applications.In comput ersci ence, r
eal-t
imeschedul ingAnal ysisisthe
evaluation, t
estingandv erifi
cationoft he 
schedul i
ngsy stem  and
the algorithms usedinr eal -
ti
meoper ations.Forcriticaloper ati
ons,ar eal-
ti
mesy st em mustbet estedandv eri
fi
edf orperformance.I ncomput er
science, testi
ngandv eri
f i
cationi sal soknownas  model checki ng.
Ar eal
-t
imeschedul i
ngSy stem iscomposedoft heschedul er,cl
ockandt he
processinghardwareel ements.Inareal-
ti
mesy st
em, apr ocessort askhas
schedulabil
i
ty;tasksareaccept edbyareal-ti
mesy stem andcompl etedas
specif
iedbyt hetaskdeadlinedependingont hecharacteristicofthe
[
1]
scheduli
ngal gori
thm.  Modelingandev al
uationofar eal-
timeschedul ing
system concernisont heanalysisofthealgorit
hm capabilitytomeeta
processdeadl i
ne.Adeadl i
neisdefinedast hetimerequiredf orataskt obe
pr
ocessed.
Forexample,inareal-
timeschedulingalgor
it
hm adeadlinecoul dbesett o
fi
venano-seconds.Inacr i
ti
caloperati
onthetaskmustbepr ocessedint he
ti
mespeci f
iedbythedeadl i
ne(i.
e.fiv
enano-seconds).Ataski nar eal
-t
ime
system mustbecompl eted“neit
hertooearl
ynort oolat
e;.
.”
.Asy stem is
saidtobeunschedulablewhent askscannotmeett hespeci fi
eddeadlines
.At askcanbeclassif
iedaseitheraper i
odi
coraperiodi
cpr ocess.

Fi
g:r
eal–t
imesy
stem exampl
e

Real
-t
imesy
stemswi
l
lusual
l
yhav
ethef
oll
owi
ngchar
act
eri
sti
cs:

 Pri
ori
ty-
basedscheduler
 Guarant
eedmaxi mum timetoserviceint
errupt
s
 Abil
it
ytoensurethatprocessesareful
lyloadedint
omemor
yandst
ayi
n
memor y
 Consist
entandeffici
entmemor yal
locati
on
 Preemptablesystem cal
ls

Cl
assi
fi
cat
ions
Thecr iteri
aofar eal-ti
mecanbecl assifi
edas  hard,
 f
irm or soft.The
schedul ersettheal gori
thmsf orexecut i
ngt asksaccor di
ngt oaspeci fi
ed
[
4]
order.  Therear emul ti
plemat hemat ical modelstorepr esentaschedul i
ng
System, mostimpl ement at
ionsofr eal-ti
meschedul i
ngal gorithm are
model edf ortheimpl ementationofuni processorsormul ti
processors
configur at
ions.Themor echallengingschedul i
ngalgorithm isf oundin
multiprocessor s,i
tisnotalway sf easiblet oimplementauni processor
schedul ingalgori
thm inamul t
iprocessor .Theal gor
ithmsusedi n
schedul inganalysis“canbecl assifiedas  pre-
empt i
ve or non-pre-empti
ve"
.
Aschedul i
ngalgori
thm def i
neshowt asksar eprocessedbyt heschedul i
ng
system.I ngeneralterms, i
ntheal gori
thm forar eal-ti
meschedul ingsystem,
eacht askisassignedadescr ipt i
on,deadli
neandani dentifi
er(i
ndicati
ng
priori
ty).Theselectedschedul ingalgorithm determi neshowpr i
orit
iesare
assignedt oapar ti
culart
ask.Ar eal
-t
imeschedul i
ngal gorit
hm canbe
classifi
edasst ati
cordy nami c.Forast ati
cschedul er,taskprior
iti
esare
determi nedbeforethesy stem r uns.Ady nami cschedul erdeterminestask
priori
ti
esasi tr
uns.  
Tasksar eaccept edbyt hehar dwar eelementsi nareal-
ti
meschedul i
ngsy stem from thecomput i
ngenv ironmentandpr ocessedin
real-t
ime.Anout putsignal i
ndicat esthepr ocessingst atus.Ataskdeadl i
ne
i
ndi catestheti
mesett ocompl et ef
oreacht ask.

You might also like