74
Modeling Real-Time System Architectures Bran Selic Principal Engineer IBM Software Group – Rational Software [email protected]
Mod
elin
g Re
al-T
ime
Syst
em
Arch
itect
ures
Bran
Sel
icPr
inci
pal E
ngin
eer
IBM
Sof
twar
e G
roup
–R
atio
nal S
oftw
are
bsel
ic@
ca.ib
m.c
om
2
Lab
Bu
ildin
g
rin
g r
oad
A Pa
rabl
e
♦“A
rchi
tect
ural
” dec
ay, c
ause
d by
:�
Lack
of h
igh-
leve
l (ar
chite
ctur
al) v
iew
(“th
e fo
rest
vs
the
trees
”)�
Diff
icul
ties
in fo
rmal
ly e
nfor
cing
arc
hite
ctur
al d
ecis
ions
Lisl
e, Il
linoi
s:
3
The
Case
of S
oftw
are
♦M
ost s
oftw
are
syst
ems
are
cons
truct
ed b
y la
bor-i
nten
sive
m
icro
con
stru
ctio
n m
eans
: …
.one
line
of c
ode
(tree
) at a
tim
e
♦Ea
ch li
ne o
f cod
e re
pres
ents
mul
tiple
des
ign
deci
sion
s
…an
y on
e of
thos
e de
cisi
ons
may
resu
lt in
a c
atas
troph
ic
failu
re o
f the
ent
ire s
yste
m
♦Th
e lik
elih
ood
of fa
ilure
exc
eeds
the
likel
ihoo
d of
suc
cess
…w
hat c
an w
e do
?
4
Thes
is: W
hat W
e Ca
nDo
♦D
esig
n an
d im
plem
ento
ur s
yste
ms
at th
e ap
prop
riate
(h
igh)
leve
l of a
bstra
ctio
n
…us
ing
a sp
ecifi
catio
n la
ngua
ge/fo
rm th
at m
atch
es th
e pr
oble
m
♦Si
gnifi
cant
ly in
crea
se th
e le
vel o
f aut
omat
ion
in th
e en
tire
proc
ess
but p
artic
ular
ly d
urin
g im
plem
enta
tion
…to
pre
vent
arc
hite
ctur
al d
ecay
5
Pres
enta
tion
Ove
rvie
w♦
Mod
els
and
mod
elin
g in
sof
twar
e♦
Req
uire
men
ts fo
r arc
hite
ctur
al m
odel
ing
♦C
once
pts
for a
rchi
tect
ural
mod
elin
g♦
Arch
itect
ural
pat
tern
s fo
r em
bedd
ed s
oftw
are
6
Early
Sof
twar
e M
odel
s
“…bu
bble
s an
d ar
row
s, a
s op
pose
d to
pro
gram
s,
…ne
ver c
rash
”--
B. M
eyer
“UM
L: T
he P
ositi
ve S
pin”
Amer
ican
Pro
gram
mer
, 199
7
Mon
itor
PH Rais
ePH
Cont
rol
PH
PH re
ache
d X
enab
le disa
ble
Curr
ent P
H
star
t
stop
Inpu
t val
veco
ntro
l
7
Mon
itor
PH Rais
ePH
Cont
rol
PH
PH re
ache
d X
enab
le disa
ble
Curr
ent P
H
star
t
stop
Inpu
t val
veco
ntro
l
…an
d W
hy T
hey
Faile
d
main () {
BitVector typeFlags
(maxBits);
char buf
[1024];
cout
<<
msg;
while (cin
>>
buf) {
if ...
?Im
ple
men
tin
g
pro
gra
m
8
Engi
neer
ing
Mod
els
9
Wha
t Eng
inee
rs D
o♦
Befo
re th
ey b
uild
the
real
thin
g...
…th
ey fi
rst b
uild
mod
els…
and
then
lear
n fro
m th
em
�
➼ ➼➼➼�
10
�Pu
rpos
e:To
hel
p us
und
erst
and
a co
mpl
ex p
robl
em o
r sol
utio
nTo
com
mun
icat
e id
eas
abou
t a p
robl
em o
r sol
utio
nTo
driv
e im
plem
enta
tion
Engi
neer
ing
Mod
els
♦En
gine
erin
g m
odel
:A
redu
ced
repr
esen
tatio
n of
som
e sy
stem
Mod
elM
odel
ed s
yste
m
11
Char
acte
ristic
s of
Use
ful M
odel
s♦
Abst
ract
�Em
phas
ize
impo
rtant
asp
ects
whi
le h
idin
g/re
mov
ing
irrel
evan
t one
s
♦U
nder
stan
dabl
e�
Expr
esse
d in
a fo
rm th
at is
read
ily u
nder
stoo
d by
obs
erve
rs
♦Ac
cura
te�
Faith
fully
repr
esen
ts th
e m
odel
ed s
yste
m
♦Pr
edic
tive
�C
an b
e us
ed to
der
ive
corre
ct c
oncl
usio
ns a
bout
the
mod
eled
sys
tem
♦In
expe
nsiv
e�
Muc
h ch
eape
r to
cons
truct
and
stu
dy th
an th
e m
odel
ed s
yste
m
Mos
t sof
twar
e m
odel
s of
the
past
faile
d on
one
or
mor
e of
thes
e po
ints
!
12
SC_MODULE(producer)
{ sc_outmaster<int> out1;
sc_in<bool> start; // to kick-start the
producer
void generate_data ()
{ for(int
i =0; i <10; i++).{
out1 =i ; //this will invoke the slave;}
} SC_CTOR(producer)
{ SC_METHOD(generate_data);
sensitive << start;}};
SC_MODULE(consumer)
{ sc_inslave<int> in1;
int
sum; // declare as a module state variable
void accumulate (){
sum += in1;
cout<< “Sum = “ << sum <<
endl;}
SC_CTOR(consumer)
{ SC_SLAVE(accumulate, in1);
sum = 0; // initialize the
accumulator}};
SC_MODULE(top) // structural module
{ producer *A1;
consumer *B1;
sc_link_mp<int> link1;
SC_CTOR(top)
{ A1 = new producer(“A1”);
A1.out1(link1);
B1 = new consumer(“B1”);
B1.in1(link1);}};
Mod
els
of S
oftw
are
«sc_
slav
e»co
nsum
er«s
c_m
etho
d»pr
oduc
erst
art
out1
in1
«sc_
link_
mp»
link1
13
Mod
el E
volu
tion:
Add
ing
Deta
il
♦D
etai
l can
be
adde
d co
ntin
uous
ly u
ntil
the
spec
ifica
tion
is c
ompl
ete
«sc_
met
hod»
prod
ucer
star
tou
t1
NotS
tarte
d
Star
tedstar
t
prod
ucer
NotS
tarte
d
Star
ted
star
t/gen
erat
e_da
ta( )
prod
ucer
void
gen
erat
e_da
ta()
{for (
int i
=0; i
<10;
i++)
{o
ut1
= I;}
}
St1
St2
14
The
Rem
arka
ble
Thin
g Ab
out S
oftw
are
Softw
are
has
the
rare
pro
perty
that
it a
llow
s us
to d
irect
ly e
volv
e m
odel
s in
to fu
ll-fle
dged
im
plem
enta
tions
with
out c
hang
ing
the
engi
neer
ing
med
ium
, too
ls, o
r met
hods
!
⇒Th
is e
nsur
es p
erfe
ct a
ccur
acy
of s
oftw
are
mod
els;
th
e m
odel
and
the
mod
eled
sys
tem
are
iden
tical
The
mod
el is
the
impl
emen
tatio
n
15
Mod
el-D
riven
Sty
le o
f Dev
elop
men
t♦
An a
ppro
ach
to s
oftw
are
deve
lopm
ent i
n w
hich
the
focu
s an
d pr
imar
y ar
tifac
ts o
f dev
elop
men
t are
mod
els
(as
oppo
sed
to p
rogr
ams)
�W
orki
ng in
the
prob
lem
dom
ain
rath
er th
an th
e im
plem
enta
tion
(tech
nolo
gy) d
omai
n
�Pl
atfo
rm in
depe
nden
ce
♦Ba
sed
on g
reat
er u
se o
f aut
omat
ion
�Au
tom
atic
gen
erat
ion
of c
ompl
ete
prog
ram
s fro
m m
odel
s�
Exec
utab
le m
odel
s –
early
in th
e de
sign
pro
cess
16
Mod
elin
g ve
rsus
Pro
gram
min
g La
ngua
ges
♦C
over
diff
eren
t ran
ges
of a
bstra
ctio
n
Leve
l of
Abst
ract
ion
high
lowPr
ogra
mm
ing
Lang
uage
s(C
/C++
, Jav
a, …
)
Mod
elin
gLa
ngua
ges
(UM
L,…
)
∆ ∆∆∆ LO: :::da
ta la
yout
, ar
ithm
etic
alan
d lo
gica
lop
erat
ors,
etc.
∆ ∆∆∆ HI: :::st
atec
harts
,in
tera
ctio
ndi
agra
ms,
arch
itect
ural
stru
ctur
e, e
tc.
17
Prog
ram
min
gLa
ngua
ge
Cove
ring
the
Full
Rang
e of
Det
ail
♦“A
ctio
n” la
ngua
ges
(e.g
., Ja
va, C
++) f
or fi
ne-g
rain
det
ail
Leve
l of
Abst
ract
ion
high
lowPr
ogra
mm
ing
Lang
uage
s(C
/C++
, Jav
a, …
)
Mod
elin
gLa
ngua
ges
(UM
L,…
)
impl
emen
tatio
n le
vel
Fine
-gra
inlo
gic,
arith
met
icfo
rmul
ae,
etc.
18
Sum
mar
y: M
odel
s an
d So
ftwar
e M
odel
ing
♦M
odel
ing
is p
artic
ular
ly w
ell-s
uite
d fo
r sof
twar
e de
velo
pmen
t bec
ause
the
mod
els
can
be g
radu
ally
ev
olve
d in
to fu
ll-fle
dged
impl
emen
tatio
ns w
ith th
e ai
d of
co
mpu
ter-b
ased
aut
omat
ion
♦M
odel
-driv
en d
evel
opm
ent i
s an
app
roac
h to
sof
twar
e de
velo
pmen
t tha
t tak
es a
dvan
tage
of t
his
uniq
ue p
rope
rty�
Wor
king
clo
ser t
o th
e pr
oble
m d
omai
n
�Le
ads
to h
ighe
r pro
duct
ivity
and
relia
bilit
y�
Plat
form
inde
pend
ence
19
Pres
enta
tion
Ove
rvie
w♦
Mod
els
and
mod
elin
g in
sof
twar
e♦
Req
uire
men
ts fo
r arc
hite
ctur
al m
odel
ing
♦C
once
pts
for a
rchi
tect
ural
mod
elin
g♦
Arch
itect
ural
pat
tern
s fo
r em
bedd
ed s
oftw
are
20
(Run
-Tim
e) A
rchi
tect
ure
♦An
abs
tract
vie
w o
f a s
yste
m th
at id
entif
ies
only
the
impo
rtant
ele
men
ts a
nd re
latio
nshi
ps
♦W
e w
ill fo
cus
only
on
run-
time
arch
itect
ures
:
The
run-
time
orga
niza
tion
of s
igni
fican
t sof
twar
e
com
pone
nts
inte
ract
ing
thro
ugh
inte
rface
s, th
ose
com
pone
nts
bein
g co
mpo
sed
of s
ucce
ssiv
ely
smal
ler
com
pone
nts
and
inte
rface
s
21
Why
Arc
hite
ctur
e is
Impo
rtant
♦En
able
s co
mm
unic
atio
n be
twee
n st
akeh
olde
rs�
expo
ses
how
indi
vidu
al re
quire
men
ts a
re h
andl
ed
♦D
rives
sys
tem
con
stru
ctio
n�
deco
mpo
sitio
n in
to u
nits
of r
espo
nsib
ility
and
para
llel
deve
lopm
ent
♦D
eter
min
es a
sys
tem
’s c
apac
ity fo
r evo
lutio
nary
gro
wth
A
CB
Med
iato
r
XA
CB
XA
CB
Med
iato
r
22
Beh
avio
rS
ervi
ces
Lay
er
Ap
plic
atio
n L
ayer
Ter
min
alA
Ter
min
alB
Ch
ann
el1
Ch
ann
el2
Str
uct
ure
Exam
ple
Com
plex
Arc
hite
ctur
e Sp
ec♦
Exam
ple
tele
com
sys
tem
arc
hite
ctur
e
23
Par
t
com
posi
tion
(exi
sten
ce d
epen
denc
y)
Basi
c Ru
n-Ti
me
Arch
itect
ural
Pat
tern
s
♦C
onta
inm
ent:
aggr
egat
ion
(info
rmat
ion
hidi
ng)
Lay
er N
+1
Lay
er N
Co
nta
iner
Par
t
Co
nta
iner
Par
t
Par
tBP
artA
�Pe
er-to
-pee
r com
mun
icat
ion:
�La
yerin
g
24
Arch
itect
ural
Com
pone
nt D
esig
n
Sys
tem
2
Sys
tem
1L
ibra
ry
Ter
min
alA
Ter
min
alB
Ch
ann
el1
Ch
ann
el2
Ter
min
alA
Ter
min
alT
este
r
Ter
min
al
Ch
ann
el
Ter
min
alT
este
r
25
Ter
min
alA
Ter
min
alB
Ch
ann
el1
Ch
ann
el2
Ter
min
alA
Ter
min
alB
Ch
ann
el1
Refin
ing
Arch
itect
ures
(Reu
se)
26
Lab
Bu
ildin
g
Arch
itect
ural
Dec
ay
♦Th
e (u
sual
ly) g
radu
al d
iver
genc
e be
twee
n an
ar
chite
ctur
al m
odel
and
its
corre
spon
ding
pro
gram
im
plem
enta
tion
♦C
ause
d by
:
�M
isun
ders
tand
ings
of a
rchi
tect
ural
inte
nt
�D
esig
n di
sagr
eem
ents
�Im
plem
enta
tion
(cod
ing)
erro
rs
♦O
ften
occu
rs d
urin
g lo
w-le
vel m
aint
enan
ce w
ork
27
Prev
entin
g Ar
chite
ctur
al D
ecay
♦En
sure
vis
ibilit
y an
d en
forc
emen
t of a
rchi
tect
ural
inte
nt
♦Ac
hiev
ed b
y:
�R
equi
ring
that
all
desi
gn w
ork
to ta
ke p
lace
at t
he m
odel
leve
l
�Au
tom
atic
ally
gen
erat
ing
impl
emen
tatio
ns d
irect
ly fr
om m
odel
s
…i.e
., us
e m
odel
-driv
en d
evel
opm
ent
28
Sum
mar
y: R
equi
rem
ents
for A
rchi
tect
ural
Mod
elin
g
♦Th
e hi
ghes
t-lev
el s
peci
ficat
ions
of a
sof
twar
e sy
stem
�Ke
y to
suc
cess
ful s
yste
m d
efin
ition
, con
stru
ctio
n, a
nd e
volu
tion
�R
ecur
sive
not
ion:
sub
syst
ems
also
hav
e ar
chite
ctur
es
♦Ar
chite
ctur
al s
pecs
incl
ude
both
stru
ctur
e an
d be
havi
or♦
Stru
ctur
al a
spec
ts c
an b
e re
pres
ente
d by
com
bina
tions
of f
our
basi
c m
icro
-pat
tern
s♦
Sign
ifica
nt b
enef
its a
re o
btai
ned
if ar
chite
ctur
al m
odel
s ca
n be
obje
ct o
rient
ed♦
A ke
y re
quire
men
t is
the
abilit
y to
pre
serv
e an
arc
hite
ctur
e in
the
face
of e
volu
tion
�R
equi
res
visi
bilit
y an
d au
tom
ated
enf
orce
men
t�
Mod
els
and
MD
D s
eem
wel
l sui
ted
to th
is p
urpo
se
29
Pres
enta
tion
Ove
rvie
w♦
Mod
els
and
mod
elin
g in
sof
twar
e♦
Req
uire
men
ts fo
r arc
hite
ctur
al m
odel
ing
♦C
once
pts
for a
rchi
tect
ural
mod
elin
g♦
Arch
itect
ural
pat
tern
s fo
r em
bedd
ed s
oftw
are
30
Por
ts
Stru
ctur
ed O
bjec
ts: E
xter
nal S
truct
ure
♦C
ompl
ex o
bjec
ts w
ith m
ultip
le “f
aces
” �
Mul
tiple
inte
ract
ion
poin
ts: p
orts
�Ea
ch p
ort i
s de
dica
ted
to a
spe
cific
pur
pose
and
pre
sent
s th
e in
terfa
ce a
ppro
pria
te to
that
pur
pose
31
S1
S2
S3
S1
S2
S1
tran
siti
on
S1t
oS
2:{i
nt
x;x
= 0;
p2.
sen
d(s
1);
p3.
sen
d(s
2);
… };
Stru
ctur
ed O
bjec
ts: I
nter
nal B
ehav
ior
♦M
ay a
ct a
s co
ntai
ners
for b
ehav
ior (
and
othe
r thi
ngs)
32
call
ack
tim
e
nu
mb
er
call
ack
talk
tran
sfer
Cal
ler
Op
erat
or
Cal
lee
Prot
ocol
s: R
eusa
ble
Beha
vior
Pat
tern
s♦
Inte
ract
ion
cont
ract
s be
twee
n ca
psul
es�
e.g.
, ope
rato
r-ass
iste
d ca
ll
33
Op
erat
orA
ssis
ted
Cal
l
Alic
e
Ch
arlie
Bo
bca
ller
calle
e
op
erat
or
initi
al
conn
ecte
d
conn
ectin
g
prot
ocol
sta
te m
achi
ne
calle
rop
erat
orca
llee
sign
ifica
nt s
eque
nces
Dex
ter
Prot
ocol
Spe
cific
atio
ns♦
A co
llabo
ratio
n th
at m
ay b
e re
quire
d on
mul
tiple
oc
casi
ons
and
situ
atio
ns
34
sig
nal
sou
rce
call
calle
r
nu
mb
erca
ller
ack
calle
e
Inco
min
g s
ign
als
sig
nal
targ
et
call
calle
e
tran
sfer
calle
r
ack
calle
r
Ou
tgo
ing
sig
nal
s
Op
erat
orR
ole
initi
al
conn
ecte
d
conn
ectin
g
prot
ocol
sta
te m
achi
ne
calle
rop
erat
orca
llee
sign
ifica
nt s
eque
nces
Prot
ocol
Rol
es♦
Spec
ifies
one
par
ty in
a p
roto
col
35
Prot
ocol
Ref
inem
ent
♦U
sing
sta
ndar
d in
herit
ance
sig
nal
sou
rce
call
calle
r
nu
mb
erca
ller
ack
calle
e
Inco
min
g s
ign
als
sig
nal
targ
et
call
calle
e
tran
sfer
calle
r
ack
calle
r
Ou
tgo
ing
sig
nal
s
Op
erat
orR
ole
sig
nal
sou
rce
call
calle
r
nu
mb
erca
ller
ack
calle
e
Inco
min
g s
ign
als
sig
nal
targ
et
call
calle
e
tran
sfer
calle
r
ack
calle
r
Ou
tgo
ing
sig
nal
s
rep
lyca
ller
qu
ery
calle
rE
xten
ded
Op
erat
orR
ole
36
En
viro
nm
ent
c :
Cls
X
S1
S2
Ports
♦Bo
unda
ry o
bjec
ts th
at�
help
sep
arat
e di
ffere
nt (p
ossi
bly
conc
urre
nt) i
nter
actio
ns�
fully
isol
ate
an o
bjec
t’s in
tern
als
from
its
envi
ronm
ent
“The
re a
re v
ery
few
pro
blem
s in
com
pute
r sci
ence
that
ca
nnot
be
solv
ed b
y ad
ding
an
extra
leve
l of i
ndire
ctio
n”
37
Ports
and
Pro
toco
ls♦
Each
por
t rea
lizes
a s
ingl
e pr
otoc
ol ro
le�
corre
spon
ds to
the
“type
” of t
he p
ort t
hat c
an b
e us
ed fo
r sta
tic
type
che
ckin
g�
exte
nsio
n of
the
tradi
tiona
l obj
ect i
nter
face
con
cept
with
a
dyna
mic
asp
ect
38
Con
nect
ors
mod
el c
omm
unic
atio
n ch
anne
lsE
ach
conn
ecto
r su
ppor
ts a
sin
gle
prot
ocol
Sta
tic ty
ping
rul
es a
pply
(co
mpa
tible
pr
otoc
ols)
sen
der
: F
ax
rem
ote:
Fax
Pro
t
rece
iver
: F
axre
mot
e:F
axP
rot
Co
nn
ecto
r
Colla
bora
ting
Obj
ects
♦U
sing
con
nect
ors
39
Fax
Cal
l
Rel
aypo
rtre
ceiv
eCtr
l: C
ontr
olse
ndC
trl:
Con
trol
c : C
ontr
olc
: Con
trol
sen
der
:Fax
rem
ote:
Fax
Pro
t
rece
iver
:Fax
rem
ote:
Fax
Pro
t
Com
posi
tion:
Stru
ctur
al P
atte
rns
40
f1:F
axC
all
sen
der
:Fax
rece
iver
:Fax
f1 := create(FaxCall);
Stru
ctur
ed O
bjec
t Sem
antic
s♦
Run
-tim
e as
serti
on: t
he c
ompl
ete
inte
rnal
stru
ctur
e of
a
com
posi
te is
aut
omat
ical
ly c
reat
ed (r
ecur
sive
ly, i
f ne
cess
ary)
whe
n th
e ob
ject
is c
reat
ed
41
Bene
fits
of R
un-T
ime
Asse
rtion
♦Ar
chite
ctur
al e
nfor
cem
ent:
only
exp
licitl
y pr
escr
ibed
ar
chite
ctur
al s
truct
ures
can
be
inst
antia
ted
�it
is n
ot p
ossi
ble
to b
ypas
s (c
orru
pt) t
he a
rchi
tect
ure
by lo
w-
leve
l pro
gram
min
g
♦Si
mpl
ifica
tion:
low
-leve
l pro
gram
cod
e th
at d
ynam
ical
ly
crea
tes
(des
troys
) com
pone
nts
and
the
conn
ectio
ns
betw
een
them
is e
limin
ated
�in
som
e sy
stem
s th
is c
an b
e as
muc
h as
35%
of a
ll co
de
♦M
ajor
net
gai
n in
pro
duct
ivity
and
relia
bilit
y
42
sen
der
:Fax
c : C
ontr
ol
rece
iver
:Fax
c : C
ontr
ol
c : S
yste
mC
ontr
ol
Beha
vior
Por
ts rece
iveC
trl:
Con
trol
~se
nder
Ctr
l: C
ontr
ol~
Beh
avio
r P
ort
Impl
emen
tatio
nB
ehav
ior
Por
t
♦Po
rts d
irect
ly c
onne
cted
to th
e st
ate
mac
hine
initi
al
conn
ecte
d
conn
ectin
g
cont
aine
r st
ate
mac
hine
43
Arch
itect
ural
Mod
elin
g an
d UM
L♦
Thes
e fu
ndam
enta
l arc
hite
ctur
al m
odel
ing
conc
epts
hav
e be
en in
corp
orat
ed in
to th
e la
test
ver
sion
of t
he U
ML
stan
dard
(UM
L 2.
0)�
Cur
rent
ly u
nder
goin
g st
anda
rdiz
atio
n (d
ue A
pril
2004
)
�To
ol s
uppo
rt al
read
y av
aila
ble
from
var
ious
ven
dors
♦U
ML
as a
n ar
chite
ctur
al d
escr
iptio
n la
ngua
ge
44
Sum
mar
y: A
rchi
tect
ural
Mod
elin
g Co
ncep
ts
♦Si
gnifi
cant
con
verg
ence
am
ong
diffe
rent
AD
Ls o
n th
e ba
sic
conc
epts
nee
ded
for a
rchi
tect
ural
mod
elin
g�
ACM
E, U
ML-
RT,
SD
L, M
etaH
/AAD
L
�St
ruct
ured
obj
ects
, por
ts, c
onne
ctor
s, p
roto
cols
♦Th
e U
ML
2.0
stan
dard
incl
udes
all
the
basi
c co
ncep
ts
requ
ired
for m
odel
ing
softw
are
arch
itect
ure
45
Pres
enta
tion
Ove
rvie
w♦
Mod
els
and
mod
elin
g in
sof
twar
e♦
Req
uire
men
ts fo
r arc
hite
ctur
al m
odel
ing
♦C
once
pts
for a
rchi
tect
ural
mod
elin
g♦
Arch
itect
ural
pat
tern
s fo
r em
bedd
ed s
oftw
are
46
Desi
gn P
atte
rns
♦A
desi
gn p
atte
rnis
a p
rove
n ge
nera
lized
sol
utio
n to
a
gene
raliz
ed p
robl
em th
at c
an b
e us
ed to
der
ive
a sp
ecifi
c so
lutio
n to
a s
peci
fic p
robl
em
♦R
epre
sent
dis
tille
d re
usab
le e
xper
ienc
e
♦M
ajor
ben
efits
of u
sing
pat
tern
s:�
Sim
plify
and
spe
ed-u
p de
sign
�R
educ
e ris
k
�Fa
cilit
ate
com
mun
icat
ions
bet
wee
n de
sign
ers
47
line
card
NEn
d us
er
line
card
1
unre
liabl
e tra
nsm
issi
onm
ediu
m
SW
ITC
H
. . .
AB
prot
ocol
AB
sen
der
AB
rece
iver
End
user
End
user
AB
sen
der
AB
rece
iver
Exam
ple
Syst
em♦
A m
ulti-
line
pack
et s
witc
h th
at u
ses
the
alte
rnat
ing-
bit
prot
ocol
as
its li
nk p
roto
col
48
pack
etiz
erun
pack
erR
ecei
ver
Sen
der
Alte
rnat
ing
Bit P
roto
col (
1)♦
A si
mpl
e on
e-w
ay p
oint
-to-p
oint
pac
ket p
roto
col
dat
a(1)
ackA
pkt
Ad
ata(
1)
ack
ack
dat
a(2)
ackB
pkt
Bd
ata(
2)
ack
ack
AB
prot
ocol
…et
c.
49
Sen
der
SM
ackB
/^ac
kda
ta/^
pktA
ackA
/^ac
kda
ta/^
pktB
timeo
ut/^
pktB
timeo
ut/^
pktA
Acc
eptP
ktA
Wai
tAck
A
Acc
eptP
ktB
Wai
tAck
B
pktA
/^da
taac
k/^a
ckA
pktB
/^da
taac
k/^a
ckB
timeo
ut/^
ackB
timeo
ut/^
ackA
Rcv
dP
ktA
Wai
tPkt
B
Rcv
dP
ktB
Wai
tPkt
A
Rec
eive
r S
M
Alte
rnat
ing
Bit P
roto
col (
2)♦
Stat
e m
achi
ne s
peci
ficat
ion
50
Addi
tiona
l Con
side
ratio
ns♦
Supp
ort i
nfra
stru
ctur
e SW
ITC
H
AB
rece
iver
AB
sen
der
op
erat
or
inte
rfac
e
DB
inte
rfac
e
Sys
tem
op
erat
or
DB
ase
AB
lin
esm
anag
er
51
Cont
rol
The
set o
f (ad
ditio
nal)
mec
hani
sms
and
actio
ns re
quire
d to
brin
g a
syst
em in
to th
e de
sire
d op
erat
iona
l sta
te a
nd to
m
aint
ain
it in
that
sta
te in
the
face
of v
ario
us p
lann
ed a
nd
unpl
anne
d di
srup
tions
�Fo
r sof
twar
e sy
stem
s th
is in
clud
es:
�sy
stem
/com
pone
nt s
tart-
up a
nd s
hut-d
own
�fa
ilure
det
ectio
n/re
porti
ng/re
cove
ry�
syst
em a
dmin
istra
tion,
mai
nten
ance
, and
pro
visi
onin
g�
(on-
line)
sof
twar
e up
grad
e
52
Retro
fittin
g Co
ntro
l Beh
avio
r
Acce
ptPk
tA
Wai
tAck
A
Acce
ptPk
tB
Wai
tAck
B
Faile
d
Just
Crea
ted
Hard
war
eAu
dit
Get
tingD
ata
Read
yToG
o
Anal
ysin
gFa
ilure
53
Faile
d
Just
Crea
ted
Hard
war
eAu
dit
Get
tingD
ata
Read
yToG
o
Anal
ysin
gFa
ilure
Ope
ratio
nal
The
Cont
rol A
utom
aton
♦In
isol
atio
n, th
e sa
me
cont
rol b
ehav
ior a
ppea
rs m
uch
sim
pler
54
Cont
rol v
ersu
s Fu
nctio
n
♦C
ontro
l beh
avio
r is
ofte
n tre
ated
in a
n ad
hoc
man
ner,
sinc
e it
is n
ot p
art o
f the
prim
ary
syst
em fu
nctio
nalit
y
�ty
pica
lly re
trofit
ted
into
the
fram
ewor
k op
timiz
ed fo
r the
func
tiona
l beh
avio
r
�le
ads
to c
ontro
llabi
lity
and
stab
ility
prob
lem
s
♦H
owev
er, i
n hi
ghly
-dep
enda
ble
syst
ems
as m
uch
as
80%
of t
he s
yste
m c
ode
is d
edic
ated
to c
ontro
l beh
avio
r!
55
Som
e Im
porta
nt O
bser
vatio
ns
♦C
ontro
l pre
dica
tes
func
tion
�be
fore
a s
yste
m c
an p
erfo
rm it
s pr
imar
y fu
nctio
n, it
firs
t has
to
reac
h its
ope
ratio
nal s
tate
♦C
ontro
l beh
avio
r is
ofte
n in
depe
nden
t of f
unct
iona
l
beha
vior
�th
e pr
oces
s by
whi
ch a
sys
tem
reac
hes
its o
pera
tiona
l sta
te is
ofte
n th
e sa
me
rega
rdle
ss o
f the
spe
cific
func
tiona
lity
of th
e
com
pone
nt
56
The
Recu
rsiv
e Co
ntro
l Ar
chite
ctur
al P
atte
rn
57
Basi
c De
sign
Prin
cipl
es
♦Se
para
te c
ontro
l fro
m fu
nctio
n
�se
para
te c
ontro
l com
pone
nts
from
func
tiona
l com
pone
nts
�se
para
te c
ontro
l int
erfa
ces
from
func
tiona
l int
erfa
ces
�im
bed
func
tiona
l beh
avio
r with
in c
ontro
l beh
avio
r
♦C
entra
lize
cont
rol (
deci
sion
mak
ing)
�if
poss
ible
, foc
us c
ontro
l in
one
com
pone
nt
�pl
ace
cont
rol p
olic
ies
in th
e co
ntro
l com
pone
nts
and
cont
rol
mec
hani
sms
insi
de th
e co
ntro
lled
com
pone
nts
58
Co
ntr
olle
dC
om
po
nen
t 1
. . .
Co
ntr
olle
dC
om
po
nen
t N
Cont
rol
inte
rface
Func
tiona
l(s
ervi
ce)
inte
rface
Cen
tral
Co
ntr
olle
r
The
Basi
c St
ruct
ural
Pat
tern
♦Se
t of c
ompo
nent
s th
at n
eed
to b
e co
ntro
lled
in a
co
ordi
nate
d fa
shio
n
59
Cen
tral
Co
ntr
olle
r
Co
ntr
olle
dCo
mpo
nent
1. .
.C
on
tro
lled
Com
pone
ntN
Cen
tral
Co
ntr
olle
r
. . .
Co
ntr
olle
dCo
mpo
nent
1. .
.C
on
tro
lled
Com
pone
ntN
Cen
tral
Co
ntr
olle
r
Recu
rsiv
e Ap
plic
atio
n♦
Hie
rarc
hica
l con
trol
�sc
ales
up
to a
rbitr
ary
num
ber o
f lev
els
60
Co
mp
Set
Faile
d
Just
Crea
ted
Hard
war
eAu
dit
Get
tingD
ata
Read
yToG
o
Anal
ysin
gFa
ilure
Ope
ratio
nal
c1:C
om
p1
cN:C
om
pN
Real
izat
ion
with
Por
ts a
nd O
bjec
ts♦
Com
posi
te p
lays
role
of c
entra
lized
con
trolle
r
61
Expl
oitin
g In
herit
ance
♦Ab
stra
ct c
ontro
l cla
sses
can
cap
ture
com
mon
con
trol
beha
vior
and
stru
ctur
e♦
Diff
eren
t sub
clas
ses
capt
ure
func
tion-
spec
ific
beha
vior
Ab
stra
ctC
on
tro
llee
po
rts
con
tro
lPo
rt:
Ctr
lPro
toco
l
Sen
der
Rec
eive
r.
. .
62
Faile
d
Just
Crea
ted
Hard
war
eAu
dit
Get
tingD
ata
Read
yToG
o
Anal
ysin
gFa
ilure
Ope
ratio
nal
Expl
oitin
g Hi
erar
chic
al S
tate
s
Ab
stra
ctC
on
tro
llee
po
rts
con
tro
lPo
rt:
Ctr
lPro
toco
l
Sen
der
63
The
Run-
Tim
e La
yerin
g Ar
chite
ctur
al P
atte
rn
64
Sem
antic
s of
Lay
erin
g (1
)♦
A fu
ndam
enta
lly d
iffer
ent t
ype
of s
truct
ural
rela
tions
hip
Op
erat
ing
Sys
tem
AB
sen
der
op
erat
or
inte
rfac
e
�La
yerin
g is
diff
eren
t fro
m c
onta
inm
ent
�H
ighe
r-lay
ers
do n
ot c
onta
in lo
wer
laye
rs�
Form
ally
, the
low
er la
yers
“con
tain
” the
hig
her l
ayer
s (e
xist
ence
dep
ende
ncy)
but
they
do
not e
ncap
sula
te th
em
65
Har
dwar
e
Lin
k
Net
wo
rk
Lev
el 4
Lev
el 5
Lev
el 6
Lev
el 7
Ope
ratin
gSy
stem
Sem
antic
s of
Lay
erin
g (2
)♦
In c
ompl
ex s
yste
ms,
laye
ring
is a
com
plex
m
ultid
imen
sion
al re
latio
nshi
p�
e.g.
, 7-la
yer m
odel
of O
pen
Syst
em In
terc
onne
ctio
n (O
SI)
66
Inad
equa
te R
epre
sent
atio
ns o
f Lay
erin
g♦
Stai
rcas
e m
odel
�To
aste
r mod
el
Ope
ratin
g Sy
stem
Appl
icat
ion
Gen
eral
Ser
vice
s
Spec
ializ
ed S
ervi
ces
67
Laye
r N
Laye
r N +
1
Mor
e on
the
OSI
Mod
el♦
Two
dist
inct
kin
ds o
f int
erfa
ces:
pee
r and
SAP
SAP
Laye
r N +
1
Publ
ic o
r Priv
ate
Inte
rface
?
Publ
ic In
terfa
ce
68
CD
op
erat
or
inte
rfac
eA
B s
end
er
AB
Tim
ing
Ser
vice
Inte
rnal
im
plem
enta
tion
com
pone
nt
Exte
rnal
im
plem
enta
tion
com
pone
nt
Impl
emen
tatio
n Co
mpo
nent
s♦
Priv
ate
sub-
com
pone
nts
requ
ired
to re
aliz
e th
e fu
nctio
nalit
y of
fere
d by
com
pone
nt th
roug
h its
pub
lic
inte
rface
69
Impl
emen
tatio
nIn
terfa
ce
Usag
eIn
terfa
ceInte
rface
Typ
es fo
r Lay
erin
g♦
Nee
d to
diff
eren
tiate
two
inte
rface
type
s:
�U
sage
inte
rface
: im
plem
enta
tion-
inde
pend
ent i
nter
face
th
roug
h w
hich
a c
ompo
nent
pro
vide
s its
ser
vice
s (fu
nctio
n an
d co
ntro
l)
�Im
plem
enta
tion
inte
rface
(ser
vice
acc
ess
poin
t):
impl
emen
tatio
n-sp
ecifi
c in
terfa
ce th
roug
h w
hich
a c
ompo
nent
ac
cess
es a
n ex
tern
al s
ervi
ce
♦Fr
ont-e
nd/b
ack-
end
view
s:
70
Impl
emen
tatio
n In
terfa
ces
♦Im
plem
enta
tion
inte
rface
s ar
e pu
blic
inte
rface
s bu
t can
be
vie
wed
as
bein
g in
a d
iffer
ent “
plan
e” (d
imen
sion
) fro
m s
ervi
ce in
terfa
ces
CD
AB
sen
der
op
erat
or
inte
rfac
e
AB
Tim
ing
Ser
vice
90o
71
Upp
erLa
yer
Inte
rnal
Co
mp
Serv
ice
acce
sspo
int
Tim
ing
Ser
vice
Mod
elin
g La
yers
with
Por
ts a
nd O
bjec
ts♦
Impl
emen
tatio
n in
terfa
ces
are
mod
eled
by
impl
emen
tatio
n en
d po
rts th
at c
an b
e co
nnec
ted
dire
ctly
to
ser
vice
por
ts o
f oth
er o
bjec
ts
72
Sum
mar
y: A
rchi
tect
ural
Pat
tern
s♦
Des
ign
patte
rns
are
a cr
ucia
l too
l for
all
softw
are
desi
gn�
Des
ign
sim
plifi
catio
n�
Des
ign
relia
bilit
y
�C
omm
unic
atio
n ve
hicl
e
♦Am
ong
the
mos
t im
porta
nt a
re a
rchi
tect
ural
pat
tern
s♦
Two
extre
mel
y us
eful
top-
leve
l arc
hite
ctur
al p
atte
rns:
�R
ecur
sive
con
trol p
atte
rn (f
or re
al-ti
me
syst
ems)
�La
yerin
g
♦Th
ese
patte
rns
can
be s
peci
fied
dire
ctly
in U
ML
73
Conc
lusi
on a
nd S
umm
ary
♦W
e ha
ve c
ross
ed th
e th
resh
old
of a
new
gen
erat
ion
of s
oftw
are
spec
ifica
tion
tech
niqu
es: m
odel
-driv
en d
evel
opm
ent
�Bo
th la
ngua
ges
and
tool
s ha
ve re
ache
d in
dust
rial s
treng
th m
atur
ity�
Man
y la
rge
syst
ems
have
bee
n pr
oduc
ed u
sing
MD
D (>
4M
LoC
)
♦Th
ese
tech
niqu
es a
re b
ased
on
�Ex
tens
ive
use
of m
odel
s an
d m
odel
ing
�Ex
tens
ive
use
of c
ompu
ter-b
ased
aut
omat
ion,
par
ticul
arly
for
impl
emen
tatio
n
♦Th
e ab
ility
to s
peci
fy a
nd e
nfor
ce s
oftw
are
arch
itect
ures
is o
neof
th
e be
nefit
s of
MD
D♦
We
have
des
crib
ed th
e co
ncep
ts th
at a
re s
uita
ble
for a
rchi
tect
ural
m
odel
ing
and
dem
onst
rate
d th
eir a
pplic
atio
n to
exp
ress
ing
usef
ulde
sign
pat
tern
s fo
r lar
ge re
al-ti
me
and
embe
dded
app
licat
ions
74
Bibl
iogr
aphy
♦Ba
ss, L
., P.
Cle
men
ts, a
nd R
.Kaz
man
, Sof
twar
e
Arch
itect
ure
in P
ract
ice,
Addi
son-
Wes
ley,
199
8.
♦B.
Sel
ic, J
.Rum
baug
h, U
sing
UM
L to
Mod
el C
ompl
ex
Rea
l-Tim
e Sy
stem
s, R
atio
nalw
hite
pape
r: (h
ttp://
ww
w.ra
tiona
l.com
/site
wid
e/su
ppor
t/whi
tepa
pers
/dyn
amic
.jtm
pl)
♦B.
Sel
ic, G
. Gul
leks
on, P
. War
d, R
eal-T
ime
Obj
ect-
Orie
nted
Mod
elin
g,Jo
hn W
iley,
199
4.
