Sie sind auf Seite 1von 7

Proceedings

of the

ed. J. D.

Tew,

1994

Winter

Simulation

S. Manivannan,

PARALLEL

DEVS:

Alex

D. A.

A PARALLEL,

ChungHen

Object

Conference

Sadowski,

and

F. Seila

HIERARCHICAL,

Bernard
Department

TX

present

Discrete
eling

a revision

Event

the

collisions

external

Such
with

revision
and

enables
collisions.

necessity

scheduled
The

latter

tent

formal

tie-breaking

is replaced

by

construct

that

wit h both

also
select

a well-defined
allows

revision

and

with

provides

are

to

difficult
and

parallel-execution

by

Discrete

Event

System

malism

was introduced

tended

to

enable

tion

models

gler

1976,

ful

in

model

Not

methodology

recognized

gent

(1993)

as:

reduction

for

reuse

existing

chical

modeling,
the

Rozenblit
mercial
den

and

last

and
simulation
Davis

of

such

and

Although

(Zeigler
1990,

vendors

aid

result,

model
of

The

the

hierar-

have

emerged

allel
to

lision

Kim

1994)

com-

closed

parallel

multiple

external

events).

The

is constructed
In

is tricky

it

either
to

intended

fol-

if the

is a modelers

model

of the
DEVS

modeling

coupled

generalization

re-

coupling

to make

handling

Parallel

modular

(Peg-

to the

sim-

716

reviews

motivation

DEVS

1993,

interest

the

provide

Praehofer

and

(simply

and

sure that

and explainable.

can be exploited

article

cusses

1990,

and

the

that

ei-

transitions

that

the

proper

within

happen

the

In this

ar-

collisions

and

specification

in-

of both

kinds

models.

As

an impor-

of

exhibits

increased

par-

in a parallel/distributed

simulation.

support
in

a generalized

tant

from

a model

the

to facilitate

collisions

interfering

can

properly.

handled

we emphasize

allelism

capability

support

are showing

Distributed

Sar-

none

environments

Janknski

modeling

time,

languages

research

1992).

and to modify

the

term

together.

languages,

are well

tended

time

We use the

generated

understand

collisions

mappings.

simulation

model

models

handled

collisions
ticle,

increasingly

developments.

of models,

simulation
years

is

development

validation.

few

for

simulator

predominant

advantages

model

of a database

major
over

the

simulation

the

in
and

abstract

capability

as

future
lists

verification

its

(Zei-

a DEVS

deviate

event
to

a logic-based

mutually

or between

behavior

easily

discrete

phenomena.

to as simultaneous

component

are not

that

collisions

and external

when

model

sponsibility

a power-

a framework

occur

may

propose

modeling

for

also

via

manner

provide

and

In

simula-

Such

(referred

collisions

ex-

1990).

Hierarchical
paradigm

it

later

event

modular
does
but

generation

(Zeigler,

being

only

for-

70s and

discrete

a hierarchical,

behavior

concepts

early

constructing

1984).

modeling

Specijication(DEVS)

in the

low

to represent

internal

DEVS

of these

behavior.

events.

the

life

at the same

to as coiiisions)

the

et al.

behaviors

introduced

semantics

transitions

coupling

case,

manifest

(1994)
the

collisions

can

as simugeographi-

Lubachevesky

as real

non-intuitive

transitions

INTRODUCTION

The

Sargent

between

latter

to

conceptualize

state

referred

and

methodologies,

prone

to study

to such

ther

1990,

modeling

simultaneous

benefits.

all

transition

to

interest

consuming

(Fujimoto

are

lead

increasing

time

1991).

foundation

consis-

received
more

simulations

Multiple

transitions

also

Preiss

Radiya

away

function.

and

all

The

conceptual

ca-

does

of simultaneously
in the

activated.

models.

Engineering

85721

distributed

As
in

the modeling

revision

as embodied

be simultaneously
modeler

The

events

DEVS

us to extend

for

events,

of

become

between

external

function

of the
the

distinguishes

has

lations

1989,

mod-

AZ

ulation

cally

modular

(DEVS)

ordinary

transition

separation

pability

hierarchical,

FORMALISM

of Arizona

Tucson,

Specification

The

formalism.

transition

and Computer

University

78758

of the

System

MODELING

P, Zeigler

of Electrical
The

ABSTRACT
We

MODULAR

Chow

Technology
Products
IBM Corp.

Austin,

A.

formalism
the

means

problems.

The

under

coupling,

construction
definition

the

DE VS formalism

behind
is then
for

the

thus
properties.

of an Parallel

and
The

introduced

properly

formalism

revision.

and

handling
is also

preserving
The

shown

disPar-

shown
the

colto

be

hierarchical,
construct

leads

DE VS abstract

simu-

Paralel

later

which

exploits

the

THE

correctly

implements

increased

parallelism.

ORIGINAL

DEVS

DE VS is the formalism
ify a hierarchically
event

model

tion
the

one

to

and

to spec-

by a simulaevolved

specializations

section

discusses

and

as a discrete

formalisms

introduce
This

a modeler
system

simulated

DEVS

purposes.

formalism

allows

can be later

Several

original

ferent

that

formalism

717

FORMALISM

decomposable

that

engine.

the

DEVS

reviews

the

the

reasons

from
for

dif-

original

behind

the

re-

visions.
The

original

DEVS

=<

a set of external

S:

a set of sequential

Y:

a set of output

tie.t

~ : S +

states.
transition

S : external
e)ls

elapsed

>

events.

Q = {(s,

e is the

A,ta

events.

S : internal

: Q x X

where

is a structure:

X, S, Y,6%nt,6ezt,

X:

b~nt : S +

model

time

: output

transition

E S,0

Figure

function.

last

time

via

DE VS

its

transition

occurs,

the

termined
new

function.

state

on the

before

as well.

and

The
1.

the

internal

curs.

S2 =

an internal
etc.

internal

and

on the

output

A coupled
=<

yl

state,

external

A(sl),

the

s4, the

x 1, to

state
in this

in

Fig-

etc.

are

event

oc-

time

~(s2),

before
input

elapsed

the

event,

time,

a new

of

select

DE VS

{Zi,j},

select

chical

for

Due

1984),

is called

out

a set of components.

(x, t),
ternal

respectively.

which

in turn

transitions

(A

using

DE VS

a coupled

model.).
concept
and

called
Basically,

simulation
sequentially

and

coupling
in hierar-

models

by

into

another

under

DE VS

atomic

by receiving

as yet

closure

simulator
processors

DE VS models

be constructed

constructed

of the
by

this
can

an atomic

abstract

ordinator-s

each

E,

regarded

atomic

is not

simulation

D:

subset

several
be

to

models
from

model

the

in .ti,

D,

non-empty

can

e4,

triggered

any

groups

that

In

i is not

which

fashion

for

i in D,

that

>,

model

model

>

constraints

j,

model.

carried
{L},

translation.

c E.

A coupled

(Zeigler

Jijtai

of D +

for

de-

is a structure:
D, {M;},

that

a composite

the

X.~~f,Y,~~f,

; subset

se/ect(.E)

state,

to the

of D U {self},

Z~,j:l(+Xj.

transition

determine

Si, K,6i,

: Xseff

y2 =

external
external

Xi,

: Yi -+ Y.elj

e4, xl).
model

=<

is a subset

Zi,,eif

the

generated

is subject

zself,j

new

and

determine

events

resulting

event,

structure

output

i in D,

event

illustrated

no

etc.

When

The

A state

elapsed

&nt(s2),

transition.

current

input

S5 = 6e$~(s4,

are

when
and

tie-breaker.

Mi

event.

ta(s3),

i-to-j

the

Ii

depends

input
the

time

the

is a function,

such

transitions

the

that

of i.

is a function,

each

applied

intluencees

j in Ii,

Zi,j

The

transition,

transition.

the

S3 =

state

event

determines

sequences

are the

by &nt

an external

state,

transitions.

the

DA

6eZt

state.

internal

internal

&n~(sl),

z 1, is received,
pends

any

when

external

ta(s2),

and

the

The

transition

ure

event
is de-

each

select

S,

transition

current

an output

current

the

to the

states,

no external

state

before

happen

on the

the

is determined

Right

can

occurs

state,

model

can generate

state

When

ta applied

of the

transition

among

of an internal

old state.

the model

transits

functions.

time

by the

to the

based

model

z in D U {self},

each

I{ is the
for

The

Model

M~ is a component.

transition.

for

advance

of a DEVS

and

state

function.

ta:S+RO_+W:

Transitions

function,

< e < ta(s)}

since

1: State

finally

(Zeigler

coupled
simulators
the
invoke

and

is
co-

simulation

messages,
schedule

1984),

models

outputs,
new

is

(*, t) and
ex-

internal

718

Chow

transitions

at future

In this
a modeler
breaker
The

times

given

DE VS structure
must

and

employ

the

of simultaneously

formalism

at both

the

e = O and

arises

state

of the

elapsed

both
In

an external

to

exist

represent

coupled

models,

In

function,

tion

solves

from

the

uled

set

ambiguity
internal

time):

others,

for

the

events)

time

and

one

simulation

languages,

a potential

ization

it

engenders

co-occurrence

of

eled,

particularly

nent

components.

may

events

of

not
in

among

rent

When

properly

sim-

of

it

function

(simultaneous
the

original

the

event

set

simulators

both

treat

do not

the

to

sup-

simultaneous

in sequence.

simultaneous

can

subsets)

both

allow

abstract

set

all

transition

Although

They

event

(set

of a subset

a component
the

input

set

external

event.

the

to

the

above

for

the

The

external

discussion,

newly

paral-

events

is not

several

properties

DEVS

formalism.

revised

havior

are

is not

on the

a necessary

The

tion

of

collision

a collision

modeling

price

formalism

function

that

to

must

prohibits

reduces

ces.

The

parallelism

concur-

tions

and

simultaneous

the
to

is a limitation

Parallelism:

simulation

events

behavior
Predefining

be-

capability

pay

for

paral-

lelism.

immi-

to parallelize

The

be controllable.

that

mod-

among

Handling:

must

the

being

influencing

scheduled

function

exploited.
From

reflect

system

in a parallel

it is appropriate

of simultaneously
greater

among

serial-

the serialization

interpreted

The

concept.

lelism

ap-

The

of parallelism

the

power

formalisms

set,

sent

desirable

tie-breaking

mutual

select

e = O; for

error.

the

Moreover,

exploitation

environment

this

of conventional

this

source

the possible
events.

approaches

the

E-DEVS

events

sched-

current

component,

of the

the occurrence

Collision
the

is

the

(those

at the

selected

reflecting

proach

dling

components

transitions

this

as active,

that

formalism
be

as a single

the

port

func-

to

of events.

regards

be a power

by a model

e = ta(s).

Although
serial

by choosing,

of imminent

to make

ulation
the

this

to

then

A select

e = O or e = ta(s)?

DEVS

of a set

either

at the same time of its scheduled


internal
transition
which
elapsed time should
be used by the external
transition

the

defined

interval:

ambiguity

is received

is similar

replaces.
be

events.

time

of which

event

or,

as a tie-

internal

a model

Zeigle~

limitation

ta functions.
co-ordinat

function

allows

transition.

when

select

scheduled

boundaries

e = ta(s),

of internal

by the

its serial

and

not

use serializa-

possible

among

the

events

concurren-

internal

must

transi-

be fully

ex-

ploited.

hanClosure:

achieve

The

pling

speedup.

formalism

and thereby,

must

be closed

support

under

hierarchical

cou-

construc-

tion.

2.1

The

Extended

DEVS

Hierarchical

Consistency:

struction
The

Extended

1992,

1993)

moving
new

DEVS(E-DEVS)
attempts

the

order

which

select
function

queues

riving

at the

removed

events.

same

because

the behavior

is then

the

and the
applied

O, z).

colliding
to the

This

means

events

state

ar-

can

isfactory

since

that

transition

sitions

still

has

between

the

different

to correctly

rectly

choose

if an interdependency

external

transitions,
model

function

the

the
joint

an

THE

exists
order
effect

lisions,

function

is still

with

unsat-

these

two

in

of the same

behavior.

formalism

was developed

in or-

requirements.

DEVS

i.e.

In

at the

or by

contrast,

plicitly

define

or more

called

confluent

over
the

solutions

the

collision

the

transition

function,

i14 =<

X,S,

Y,ilnt,

b.$t,

VS)

a modeler
by using
6COn.

6COn, A,ta,

of an

transition.

corThis

to dethrough

priority

DE VS( P-DE

behavior

receives

attempt

external

enables

key
col-

prescheduled

either

imposing

a colliding
here

the

handle

as its

implicitly

Parallel

proposed

that

a component

time

Previous

function

ism structure

when
same

behavior

transition

we find

is to properly

behavior

collision

select

internal

the

transition.

the

discussion,

requirements

events

internal
fine

above

these

More-

cannot
events.

the
tran-

function.

function
of

DEVS
these

on the

external

external

among

same

con-

behavior

constructs

the

PARALLEL

to meet

carried

to understand

order

display

Parallel

to address

if a col-

sneW = 6~~t (s),

a modeler

interrelationship

over,

The

the
the use of an order

must

der

Based

time

e=O.
However,

hierarchical

hierarchical

uniform

be

predefine

is always

external
new

re-

at a collision

transition

model

model

function

formalism
model

different

The

render

arrangement,

external

select

that

re-

place

coupled

of this

E-DEVS

internal

the

the

The

of a component

occurs,

out first

Because

the

by

in its

of simultaneously

to hold

time.

to be s = tieot(dint(s),
lision

to

execution

are used

(Wang

parallelism

However,

is added

the

external

event

exploit

function.

serializes

ceived

to

formalism

must

>

formalto exthe

so-

Parallel

719

DEVS

at the

event

ponents
tions

receiving

for

the

the

confluent

collide,

ployed

instead

sition

to

2, all the

in

the

external
function

internal

or external

the

transitions

new

formalism

is emtran-

state.

are the

In

same

except

comtransi-

transition

the

DEVS

for those

and

determine

state

original

However,

internal

of either

function

ure

time).

which

Fig-

as those

at t2

and

t5 where collisions
occur.
At these instances,
the
new states are defined to be 53 = 6cOn(s2, 22) and
d::

~....

s6

i-

1, :
:

Y:

1 1

I++***

:
:

;Y3

fl
tI

ys

ta(s3) Mta(s$

1s(s2)

r!6

X:

2: State

Transitions

and

a set of input

DEVS

Model

a set of sequential

Y:

a set of output

each

Ro+_w:

where

Q = {(s,

e is the

time
e)ls

elapsed

It

events

at the

e = ta(s).
collision
cision

transition

advance

eler.

last

Indeed,

where
when

time

of its
than

i in D U {self},
j in Ii,

the

state

Mi

at the

setting

complete

VS formalism

receiving
generated
tions
for

2.

structure

Xi,

of the

The

E-DE

Sn

Paralle!

internal
time

external

by these
there

translation.

over

all

events.
imminents

non-imminents
are no internal

to the

constraints

that

for

&intil

6e$ti,

6co~i,tai

> is a P-DEVS

is a subset
: Xself

Zi,~~(f

: Yi +

of D U {self},
+

i is not

in li,

j,

Y$eif,

self
for

and

external

CLOSURE

de-

mod-

The

model

itself

and

of external

is

input

couplings.

and

resultant

X,Y,

D, {Mi},

~ezt(sn-1>o,~n-1

serialization

defined

out

components

showing

(M=<
ta(s)

DE VS are illustrated
are carried

UNDER

COUPLING

to

be

of
{Ii},

closure

of
a

minimum{a~li
s E S and

well

coupled

{Zi,j}

>)

15,~t, b..~,

is

external

events

s =

external

transi-

1kfA4(s)

(...,

A,ta

E D},
ei.

(s~, el) , )!
= {ilu;
= is(s)},

U;61;M(.)
~i},

INF(s)=

{;I;

CONF(S)

= IMA4(s)

n INF(s),

done

defined

D.
ai = ta(s~)

is
coupled

model

Let

Also,

scheduled

it

X, S,Y, Lt,
where iE

where

the

resultant

xQ~

trigger

components

of

the

The

(those

coupled

specification

demonstration

constructing

can

transitions

to the

allowing
output

b~~~(s~, O, z~),

imminent

refers

at

this

to the

VS formalism

transitions
for

is subject

Si, ~,

Z.qelf,j

S=

at receptive
which

output

i in D,

a device

e = O or

leaves

6COn(s, Xb ) to

Xn is a desired

event

no

of i.

receives
behavior

to use, if any,
the

for-

control

model

&nt(s),

DEVS

a component
transition,

P-DE

1, S1 =

next

iniluencees

is a function,
i-to-j

=<

Ii

by Order(zb).

in Figure

is the

transition.

internal

if so desired,

semantics

Ii

z {,j:~+Xj.

serializing

serialization

n > 1, and

The

>

structure,

function.

the previous

when

by

n z

from
modeler

Rather

be recovered

{Z~,j}

is

events,

each

The

function,

since

the

times,

{L},

model

events.

each

each

< e < ta(s)},

behavior

of what

D, {M~},

for

Z~,j

c S,0

time

gives

collision

X,Y,

coupled

i in D,

Here
malisms.

=<

revised

events.

6CO~is the difference


the

col-

by inter-

transitions.

for

the

transition
function.
+ S: external transition
function,
of bags over elements in X,
= (s, e).

ta : S +

Xb, to

generated

M~ is a component.

states.

-+ S: confluent
Yb: output function.

A :S +

events

a set of components.

hnt : S + S: internal
c$.=i : Q x Xb
Xb is a set
6,0~(s, e,q5)
6~On : S x Xb

need the bag,

still

of the

Y: a set of output

events.

S:

confluent

a set of input

X:

We

external

structure
DN

D:

of a Parallel

x5).

nal

for
Figure

a.0~(s5,

simultaneous

The

j,,

:
:
:

lect

P-DE
(DN

P-DEVS
>),

by

model

where

VS.
=<

model

Chow

720

INT(S)

= If

EXT(S)

We partition
transition
ready

events

sition.

not

COZVI(S)

four

the

for

the

events

and

also

scheduled

for

at the

same

time.

~iV(s)

contains

~(s)

= (...,

IAVI(S),
(s~,e;)= (bez~;(s~,e;+ta(s), z:), O) for
0) for

(s(,

ei)

(si,

= {20,i(&(sO))/o

~~}.

i G EXT(S),

The

resultant

kinds

of

tions
of

internal

component

of

lNT(s)

EXT(s)
elapsed

participation

lation

by using

relative

tion

that

of the

event

are

transi-

time.

transitions

updated

of

by

base rather

resultant

model,

received

by

is an internal

and

there

the

coupled

transi-

we construct

c$ec~(s,e,z~)

= (...,

outside

nately,

it turns
produced

ulation

time

lIVF

from

the

with

three

this

the

dezt of the

resultant.

(~.cti(si,ei

+ e,z$),

O) for

tions

of the

identifies

routed
j.

incoming
to the

The

the

event

event

resultants

external
Finally,

bag,

bag,

x:,

external

transitions
we construct

Zb

of each

influenced

transition

of the
the

translated

x:

IN

F(s),

COiVJ?(S)
IiVT(s)
EXT(S)

C lJi~(lMM(s)U{selj})

]A4fiM(s) n INF(s),

= IMM(S)
= INF(S)

15COn(S,Zb) = (...,

the

hierarchical

all

resultant.

(s~,e~),

~i},

is

we

we come

up

hierarchical
that

internally

or ex-

into

single

time

can

one

of

the

transition

into

the

one

resultant

the three

a more

elements
only

generic

For
in

and

should

be

example,

CONF(S)

contains

components

hint,

6COn, and

cases of a more

(s, ta(s),

model,
transition
transition

6cOn to the

where

a modeler
transition
and

to

functions
function

~,

the

model

level,

with

three

deal

than

function.

transparent

Xb # ~.

among

atomic
rather

Z*

and

semantics
at the

we

tran-

6int is applied
4),

O ~ e < ta(s)

individually

generic

z*)

6eEi,

generic

1984).

distinctive
for

semantics

with

lIMLf(s)

easily.

where

functions

is advantageous

times,

components

# O.

de.~ to (s, e, x*)

Compo-

elapsed

(s, e, z~) = (s, ta(s),

(s, e, x*)

lM_Lf(s)

ta(s).

function

6(s, e, z*)(Zeigler

cases when

them

which

of the

are special

when

Combining

to

z:

applica-

are maintained

identifies

which

and

transition

of the

have

component

definition

function

ei

llbf Lf(s)

applied

three

composition.

identify

lNF(s)

they

of the
semantics

bag).

which

the

different

eler,

same

set

influencers

w operation

functions

INF(s)

strictly

Because

IA4M(S).

in

empty

three

for

by the

with

not

ei = ta(s)

and,

INF(S),
. ..).

are

to

cases

it
= {jl.j

transition

~ (the

to the

children.

Let
liVF(s)

child,

comprises

influenced

6COn of the

and

Zb, at sim-

influence

The

whether

the

three

el < ta(s).

see that

is

bag,

the

6cOn.

application

components
that

sition
The

the

we see that

From

C Zb A i E ~.,[~}.

= {Z,e~~,~(Z)lZ

here

events,
at the

confluent

.z(sel f, i),

for

tiCOn of

extra

additional

couplings,

origiFortu-

between

event

the

i.e.,
of the

transitions

model.

the

redefining

is achieved

Examining

with

(s:, e:) = (si, ei + e) otherwise,


and

of the

incoming

also

events

difference

groups

external

internal
external

is simply

includes

generated,

IMM(s)

i = ls.~f,

we must

group.

6CO~

(s[, e;)=

that

all

applied

(s~, e~)),..),

the

By

similar

functions,

O <

at

by the

incoming

ternally

model

that
its dint

ta(s).

(s)

make

boundaries

out
and

effect

to

the

resultant

nents

where

z;

nating

is no external

time.
Next,

to

receive

throughout
this

coupling,

about

ta(s).

the

under

scheduled

event

in simuthan

closure

are

consistency

~~(s),

can be removed
time

A i E ~o}kd

to define the 6Co~ of the resultant,


to the coupled
model when some

also

are

gathers

transitions

remainder,

merely

by assumption,

being

the

of ~lV(s)
an absolute

elapsed

Note

and

times

(The

internal

confluent

components

whose

of four

external

establish

children

to

comprises

transitions:
components,

components,

GONF(S)

A i E l.}.

transition

C ~~M(s)

E z A i E ~~.lf }.

and

the

i G UN(S),

E ~illitf(s)

z! = {ZO,~(&(sO))lo

To

and
x!

i e COiVI(s),

otherwise,

e, + ta(s))

consider
how
what happens

i E COIVI(s),

otherwise

ei + is(s))

X,b, , O) for

remain-

i G

( S8, X1b, , O) for

(sj, e~) = (&Oni

(s;, e;) = (~COni(Si,

{Z~,lt,i(z)lz

(s~, e~)),..),

= (&nti(si),

O)

and

where
(s~, ei)

z!),

transi-

the

A self

E ~fbfikf(s)

i e IiVT(s),

ei + ta(s),

receiving

internal

O) for

(s~, e;)=
(6.Z,;(Si,
for i E EXT(S),

(s1, el) = (si,

tran-

Then,

= {Z~,~e~f(~~(s~))li

ii~nt(s)

input
receiving

components

input

components.

components

an internal

tions
ing

at any

without

components

scheduled

contains

sets

the

transition

contains

but

into

contains

an internal

EXT(S)

input

(s{, e:) = (~inti(si),

components
IiVT(s)

make

events.

Iiufivr(s).

the

time.
to

Zeigler

where

MM(s) INF(S),

= INF(S)

and

lumping
However,

to the

can be merged
so that

the

modinto
same

Parallel

implementation

can

be reused

6COn(s, Zb) ~ c$(s,ta(s),

isolated

by all transitions.

Zb),

tl.z~(s,

e, Zb) = $(s, e,zb)

ti~n,(s)

~ 6(s, ta(s),

for

O < e < ta(s),

721

DEVS

and

existing

of the

hierarchical

Finally,

4),

work
To summarize,

here

is the

generic

transition

der
= (...,

(s~, e~)),..),

(s[, e:)=

(&(s,,

ta(s),

d), O) where

(s;, ej) = (&(si,

e; + e, z!),

(s:, ej)

fa(s),

= (&(si,

i E lALP(s),

O) where

i E EXT(S),

@), O) where

the

abstract

example,

many

{ZO,; (AO(SO))IO

Sargent

(1994)

E lMA4(s)

A i e Io}w

This

research

Grand

single

totype

on

(Chow

level

abstract

the

and

simulator

CM-5

Zeigler

has

massively

been

parallel

pro-

CM-5

that

functions
model

at

malism

see

must

be

given

time.

but

sometimes

the

serial

DE VS
state
than

first

its

of its
helps

uniquely

models

manner

over

the

unique
and
time.

intuitively

as-

it severely

the

of

Extended

simulation

parallelism,

of control

states

transitions

more

for-

because

define

internal

renders

and exposes

latitude

models

lim-

collision

behaviors.

quired

Parallel

to

that

function

allows

the

follow.

the
the

model

P-DE

can

support

effort

tions

and combine
into

more

intuitive

into

events

the

and

speeds

up simulation

cur

the

at

same

design
one.
correct

time

follow

atomic

level.

form.

could

For

be devel-

confluent

transition

of Radiya

in this

and

context.

supported

Application

ARPA
under

by NSF

HPCC

Grant

ASC-

Group

participation
grant

and

employed

the

MCA94P02

Planning

allows
efficient

Kim,

T.

so that

Lubachevesky,

only

when

many

for

a given

rollback-based

Transactions
tion,
Pegden,

C. D.,

and

H. 1993.

The

simulation
environment.
A.,

and

puter

Simulation,

external
model.

Rozenblit,

create

and

it

57(3).

this

events
The

aho
oc-

Sargent,

well

event

:A
sys-

Proceedings,

W.,

R.

1993.

simulation

DEVS-based
Lisp /CLOS.

3.

distributed

discrete
and

Simulation
1994.

event

SCS

A logic-based

modeling
on

execu-

89.

and

Modeling

simula-

and

Com-

4(l).
and

framework

simulation

for

language

G. Sargent.

Transactions

J.

tegrated

Yaddes

Dtstrzbuted
R.

can put

results,

TM

modeling

Systems,

of discrete

modelers

does

Simula-

Arena

in common

specification

ACM

tran-

environment

1989.

B.

Radiya,

func-

An

Dynamic

Preiss,

1992.

Conference

Event

tion.

external

1991.
A CM

Computer

hierarchical

Simulation

Discrete

Press.

Schwartz.

AZ.
simulation

This

Manual,
Korea.

simulation.

D. A. Davis.

multiformalism

level.

A.

and

based

Winter

Phoenix}

tion

Users
Taejon,

and

on Modeling

SIMAN/CINEMA

hier-

transition

Weiss,

simula-

33(10):30-53.

1(2).

bag con-

of several
simulation

A.

of

and

final.

event

DEVSIM++

B.,

analysis

not

discrete

KAIST,

of

Simulation

of the ACM,

Dept,

simulators

Proceedings

Systems,

Parallel

1994.
EE

The

on AI,

Autonomy

1990.

the

of external
Not

G.
Lab,

foundation

also supports

Conference

in High

CORE

1994.

formalism.

Communications

a deep

simulation,

DEVS

R. M.

tion.

An

B. P. Zeigler.

Annual

Fujimoto,

event

consistency

to a single

executions

a single

new

so as to ren-

formalism

is suggestive

Parallel

Praehofer,

This

environments.

VS formalism

more

transition
to

to the

more

is re-

behavior.

hierarchical

property

in sequential

cept for simultaneous

sitions

down

to be flattened

restructuring

The

confluent
construction

and

this

a modeler

collision

coupling

coupling

In addition,

particularly

the

of a collision

under

archical

additional

captures

semantics

Closure

DE VS formalism,

supply

function

of the

frame-

with

in concrete

constructs

logic-based

A. C., and
the

tem.

In the

a sound

expression

specifications
language

at NCSA

the Fiflh

for

DE VS

simulator.

at the same

approach

output

determine

modelers
out

transitions

this

the

original

abstract

carrying

external

serted

to

and

determined

The

in a non-intuitive

formalism

Though

transitions

uniquely

a way

nature

by

external

any

provides

model

was partially

with

Chow,
the

provides

to experiment

REFERENCES

of
state

add

parallelism

computer

1994).

CONCLUSIONS

The

event

Challenge

9318169
A

DEVS

developers

specification

The

level

the

ACKNOWLEDGMENTS

E Xb A i E l$.,~}.

{.z~.lf,,(z)lz

the

function.

coupling

to exploit

i E COiV1(s),

(s;, e:) = (si, ei + e) otherwise,


and
z:

language

for

at each

DE VS models.

Parallel

of discrete

oped

where

groups
possibilities

the

for

forms

func-

tion.
6(s, e,xb)

transition

to the

P. Janknski.
for

of

natural

Hierarchical
(panel).

1990.

knowledge-based
systems.
modeling
Wtnter

An

in-

modeling
Simulation,
for

Simulation

discrete
Con-

722

Chow

ference
WangJ

Proceedings,
Y.

H.

page

1992.

569,

Los Angeles,

Discrete-event

simulation

CA.
on

lation.
and

Formalism

Discrete

Applications,

Zeigler,

Theory

B. P. 1984.

crete

Event

Zeigler,

New

B. P. 1990.

Object-Oriented

CHUNGHEN

Corporation

since

M .S. degrees

in Electrical

Chiao-Tung
in

University
in

P.

and

His

McGill
Ph.

1969.

He

and

ulation,

He
D.

has

systems.

and

Simulation

the

foundational

veloped

first

environment

gent

Agents

Academic
been
NASA,

and

by

USAF,

and

sponsors

and

Motorola.

plinary

team

including

agencies

the

Army,

Siemens,

He is currently
to demonstrate

deDEVS

high

autonSim-

Models:

Intelli-

published

by

research

has

Zeiglers

federal

out-

in the
to

Systems

1990.
US

Simuthe

Object-oriented

Modular

including
as well

McDonnell
heading

an innovative

of

book

Concepts

applied

Endomorphic

supported

as one

given

are implemented

Boston,

Modelling

Event

was
award

book,

simauton-

A second

Discrete

1984),

Hierarchical,

Press,

trial

and

and

omy issues in the latest

high
of

field.

jour-

and

is regarded

in the

MIT,

Michigan

hundred

and

of
Phys.

from
of

two

publication
work

Eng.

Theory

976)

Press,

of Elec-

modelling

systems
book

works

B.S.

M. S.E.E.

in

based

simulation

with

object-

University

University
over

Modelling

simulation

include
modeling

and

at the

1962,
the

(Wiley,l

in earlier

interests

his

articles

(Academic

standing

D, de-

University

is a professor

published

His

Multifaceted

and

National

his Ph.
the

interface,

received

from

knowledge

lation

user

IBM

B.S.

system

Engineering

conference

omy

and

research

University,

and

joined

from

from

ZEIGLER

Tucson.

ulation

with

framework.

Computer

Arizona,

nal

Dis-

Press,

his

Engineering

graphical

BERNARD

in

has

received

environment,

programming

1964,

CHOW

in Taiwan,

1990.

simulation,

oriented

from

Academic

Engineering

programming

trical

Simulation

Models.

He

1991.

Electrical

Arizona

and

and

Press, London.

BIOGRAPHIES

ALEX

visual

Modelling

California.

AUTHOR

gree

Simula-

1976.

Academic

Modular

Diego,

SimuTheory

and

York,

Multifaceted

Simulation.

Hierarchical,
San

Parallel

Systems:

of Modelling

Wiley-Interscience,

Zeigler,

of

Dynamic

3:193218.

B. P. 1976.

tion.

for Massively

Event

NSF,

as indusDouglas,

a multidisciapproach

Zeigler

to massively

parallel

simulation

tern

within

NSFs

models

of large

HPCC

Grand

scale

ecosys-

Challenge

initiative.
He was elected as Fellow of the IEEE for
his innovative
work in discrete event modelling
theory.

computer,
Ph. D. dissertation,
massively
parallel
Dept.
of Electrical
and Computer
Engineering,
University
of Arizona, Tucson, AZ.
1992. Extending
Wang, Y. H., and B. P. Zeigler.
the DEVS

and

Das könnte Ihnen auch gefallen