or f gents etworks - Napierbill/bcs_agents/nikos.pdf · ed in Java: – P latform independence. •...
Transcript of or f gents etworks - Napierbill/bcs_agents/nikos.pdf · ed in Java: – P latform independence. •...
MA
RIA
N –
Mob
ileA
gent
sfo
rR
outin
gIn
Ad-
hoc
Net
wor
ks
Nik
os M
igas
PhD
stu
dent
2nd
Year
,Sc
hool
of C
ompu
ting,
Nap
ier U
nive
rsity
n.m
igas
@na
pier
.ac.
uk
Supe
rvis
ors:
W.B
ucha
nan,
K.M
cArtn
ey
Intro
duct
ion
•M
AR
IAN
is a
n on
-dem
and,
mul
ti-ag
ent,
met
ric-d
riven
rout
ing
prot
ocol
in
ad-h
oc n
etw
orks
.–
It ut
ilises
bot
h st
atic
and
mob
ile a
gent
s to
det
erm
ine
optim
al ro
utes
and
to
actu
ally
rout
e th
e da
ta.
•E
ach
mob
ile d
evic
e is
equ
ippe
d w
ith a
:–
Pro
xy a
gent
. •
resp
onsi
ble
for f
orw
ardi
ng n
etw
ork
traffi
c.•
Java
-bas
ed, m
ultit
hrea
ded
allo
win
g m
ultip
le c
onne
ctio
ns.
–B
ench
mar
king
mul
ti-A
gent
Sof
twar
e S
yste
m (B
AS
S).
•Ex
ecut
ing
stan
dard
per
form
ance
test
s (m
emor
y, C
PU
, pro
cess
ing
capa
bilit
ies,
et
c.).
•M
onito
ring
the
utilis
atio
n of
the
devi
ce.
•Pr
oduc
ing
a m
etric
of t
he p
erfo
rman
ce o
f the
dev
ices
in re
al-ti
me.
•Th
e pu
rpos
e of
this
pre
sent
atio
n is
to in
trodu
ce th
is n
ovel
rese
arch
ar
ea a
nd to
pre
sent
the
nove
l fin
ding
s co
ncer
ning
:–
Pro
xy c
apab
ilitie
s of
reso
urce
-con
stra
ined
dev
ices
.–
The
effe
ct th
e O
S a
nd th
e JV
M m
ay h
ave
in p
erfo
rman
ce.
–Th
e re
sour
ces
cons
umpt
ion
rate
s w
hile
rout
ing,
in te
rms
of:
•Ba
ttery
dis
char
ge ra
te, C
PU
util
isat
ion,
hea
p m
emor
y us
age,
and
tem
pera
ture
.
Res
earc
h H
ypot
hesi
s
–Is
it p
ossi
ble
to c
reat
e an
aut
omat
ical
ly c
onfig
urin
g sy
stem
w
hich
will
rout
e da
ta th
roug
h ad
-hoc
net
wor
ks in
an
optim
alw
ay?
For t
his
we
prop
ose:
–“A
fram
ewor
k us
ing
stat
ic a
nd m
obile
age
nts
that
can
be
used
fo
r rou
ting,
net
wor
k di
scov
ery,
and
aut
omat
ic n
etw
ork
reco
nfig
urat
ion
in a
d-ho
c ne
twor
ks, i
n an
effi
cien
t, ef
fect
ive,
an
d se
cure
way
.”
wire
less
dom
ain
Virt
ual
Inte
rnet
dom
ain
wire
less
dom
ain
Virt
ual
Inte
rnet
dom
ain
Pos
sibl
e be
nefit
s an
d ou
tcom
es (1
)
Max
imiz
e ne
twor
k pe
rfor
man
ce–
Mob
ile a
gent
s co
uld
filte
r hug
e am
ount
s of
dat
a lo
cally
to th
e de
vice
s,
gene
rate
d by
BA
SS
.–
Thus
, ben
efit
from
loca
l int
erac
tions
.–
Det
erm
inat
ion
of lo
w a
nd h
igh
band
wid
th ro
utes
.–
Low
ban
dwid
th ro
utes
will
be u
sed
for a
sync
hron
ous
netw
ork
traffi
c (e
-m
ails
, tex
t), w
hile
hig
h ba
ndw
idth
rout
es fo
r syn
chro
nous
net
wor
k tra
ffic
(real
-tim
e au
dio,
vid
eo).
Scal
abili
ty –M
As
are
inhe
rent
ly d
istri
bute
d, th
us th
ere
is a
sco
pe fo
r a to
tally
di
strib
uted
sys
tem
.–
The
wor
kloa
d w
ill be
dis
tribu
ted
even
ly:
»H
igh
perfo
rman
ce d
evic
es w
ill be
resp
onsi
ble
for c
ompl
ex c
alcu
latio
ns a
nd ta
sks.
»Lo
w p
erfo
rman
ce d
evic
es w
ill be
use
d as
an
alte
rnat
ive
whe
n hi
ghpe
rform
ance
dev
ices
ar
e un
avai
labl
e or
bus
y.
Dyn
amic
–M
obile
age
nts
are
reac
tive,
that
is, t
hey
sens
e th
e en
viro
nmen
t,an
d re
act i
n ch
ange
s of
the
envi
ronm
ent.
Rel
iabl
e en
d-to
-end
com
mun
icat
ions
–M
obile
age
nts
will
mig
rate
to o
ther
net
wor
k no
des
and
com
mun
icat
e lo
cally
.–
Mob
ile a
gent
s ca
n w
ait u
ntil
the
netw
ork
is a
vaila
ble,
and
mig
rate
th
en.
Pos
sibl
e be
nefit
s an
d ou
tcom
es (2
)
QoS
gua
rant
ees
–Th
e tra
nsm
ittin
g de
vice
cou
ld re
ques
t an
estim
ate
of th
e av
aila
ble
band
wid
th th
roug
h a
certa
in ro
ute.
Secu
rity –
Onl
y au
then
ticat
ed a
gent
s w
ill pa
rtici
pate
in th
e ro
utin
g pr
oces
s.–
A g
uard
sta
tic a
gent
will
prot
ect d
evic
e’s
reso
urce
s fro
m p
oten
tial
mal
icio
us a
gent
s.–
Bal
ance
bet
wee
n se
curit
y an
d pe
rform
ance
.R
econ
figur
abili
ty–
Mob
ile a
gent
s ca
n be
eas
ily d
ispa
tche
d to
inst
all u
pdat
es in
the
rout
ing
prot
ocol
.En
ergy
effi
cien
t–
Dev
ices
with
low
per
form
ance
cha
ract
eris
tics,
or l
ow b
atte
ry li
fe
will
not b
e ov
erut
ilised
by
the
rout
ing
prot
ocol
.»
Busy
dev
ices
will
not p
artic
ipat
e in
rout
ing,
as
long
as
othe
r dev
ices
can
do
the
sam
e jo
b.»
If ba
ttery
life
of a
dev
ice
is lo
w, t
he ro
utin
g pr
otoc
ol w
ill co
nsid
er th
at d
evic
e as
a
wea
k lin
k.
Bac
kgro
und
•W
irele
ss n
etw
orks
and
the
usag
e of
mob
ile d
evic
es a
re b
ecom
ing
incr
easi
ngly
pop
ular
, esp
ecia
lly in
cre
atin
g ad
-hoc
net
wor
ks–
The
grea
t ben
efit
of w
irele
ss n
etw
orks
is th
e ab
ility
of u
sers
to
com
mun
icat
e, c
oope
rate
, and
acc
ess
the
netw
orke
d se
rvic
es in
an
anyt
ime
and
anyw
here
fash
ion.
•W
irele
ss n
etw
orks
can
be
grou
ped
into
two
cate
gorie
s:–
Infra
stru
ctur
ed n
etw
orks
•Fi
xed
and
wire
d ga
tew
ays,
with
acc
ess
poin
ts•
A m
obile
uni
t con
nect
s to
, and
com
mun
icat
es w
ith, t
he n
eare
st a
cces
s po
int t
hat
is w
ithin
its
com
mun
icat
ion
radi
us.
–In
frast
ruct
ure-
less
net
wor
ks (t
ypic
ally
kno
wn
as a
d-ho
c ne
twor
ks)
•A
colle
ctio
n of
geo
grap
hica
lly d
istri
bute
d m
obile
nod
es th
at c
omm
unic
ate
with
one
an
othe
r ove
r a w
irele
ss m
ediu
m w
ithou
t the
nee
d of
fixe
d ne
twor
king
equ
ipm
ent,
such
as
rout
ers
and
switc
hes.
•M
obile
nod
es c
ould
act
as
gate
way
s to
oth
er n
etw
orks
.
Rou
ting
in A
d-ho
c ne
twor
ks
•R
outin
g is
the
mos
t im
porta
nt fu
nctio
n in
ad-
hoc
netw
orks
, ho
wev
er a
cha
lleng
ing
issu
e:–
Mob
ility
•e.
g. A
rout
e th
at is
con
side
red
as g
ood
now
, may
bec
ome
unav
aila
ble
or n
ot o
ptim
al s
ome
time
late
r–
Rou
ting
requ
ires
a lo
t of p
roce
ssin
g po
wer
•M
obile
nod
es ru
nnin
g on
bat
terie
s m
ay n
ot b
e ab
le to
cop
e
Mul
timed
ia
traffi
c?
Asy
nchr
onou
s tra
ffic?
AB
Sec
ure
traffi
c?
App
licat
ions
of A
d-ho
c ne
twor
ks
•M
ilita
ry •Ba
ttlef
ield
•Te
rror
ism
& R
escu
e O
pera
tions
•Ev
acua
tion
of a
bui
ldin
g on
fire
•H
ospi
tals
•R
etrie
ve p
atie
nt’s
info
rmat
ion
from
hos
pita
l’s d
atab
ase
whi
le in
surg
ery
•C
onfe
renc
e m
eetin
gs•
Shar
e in
form
atio
n qu
ickl
y•
Sche
dule
mee
tings
•N
etw
orki
ng w
hile
on
the
road
•In
ter-v
ehic
le c
omm
unic
atio
n
Mob
ile A
gent
s (1
)
•P
ropo
sed
as a
pro
mis
ing
solu
tion
for d
istri
bute
d co
mpu
ting
over
ope
n an
d he
tero
gene
ous
netw
orks
.•
Pro
pose
d to
ove
rcom
e ce
rtain
lim
itatio
ns o
f tra
ditio
nally
des
igne
d di
strib
uted
sys
tem
s su
ch a
s cl
ient
-ser
ver.
•A
mob
ile a
gent
can
be
defin
ed a
s a
softw
are
prog
ram
that
can
sus
pend
its
exe
cutio
n on
a h
ost c
ompu
ter,
trans
fer i
tsel
f to
anot
her a
gent
-en
able
d ho
st o
n th
e ne
twor
k, a
nd re
sum
e its
exe
cutio
n on
the
new
host
.•
The
key
feat
ures
of m
obile
age
nts
are:
•M
obilit
y; n
etw
orks
aw
aren
ess;
com
mun
icat
ion;
inte
lligen
ce; r
eact
ivity
; au
tono
mou
s; g
oal-o
rient
ed; t
empo
raril
y co
ntin
uous
; lea
rnin
g; fl
exib
le; a
nd
char
acte
r•
A m
obile
age
nt c
onsi
sts
of it
s co
de a
nd s
tate
, whi
ch c
arrie
s w
ith it
du
ring
the
self-
initi
ated
mig
ratio
n
Mob
ile a
gent
s v.
clie
nt/s
erve
r?
Res
pons
em
essa
ge
Get
m
essa
ge
Age
nt m
igra
tes
to
the
serv
er
Agen
t
Age
nt c
omm
unic
ates
with
the
serv
er, o
nth
e se
rver
Ser
ver
Trad
ition
alag
ent i
n a
clie
nt/
serv
er a
rchi
tect
ure
Mob
ile a
gent
sin
a c
lient
/se
rver
arc
hite
ctur
e
Net
wor
kN
etw
ork
Ser
ver
Clie
nt
Mob
ile A
gent
s S
yste
ms
(MA
S)
•M
obile
age
nt s
yste
ms
prov
ide
an e
nviro
nmen
t in
whi
ch m
obile
age
nts
can
exis
t an
d of
fer s
tand
ardi
zed
serv
ices
to a
gent
s do
ckin
g on
•Se
rvic
es in
clud
e: c
omm
unic
atio
n; m
igra
tion;
sec
urity
; cre
atio
n an
d te
rmin
atio
n of
ag
ents
•
Man
y m
obile
age
nt s
yste
ms
have
bee
n de
velo
ped
from
aca
dem
ic a
nd
indu
stria
l bac
kgro
und.
•A
ccor
ding
to a
com
preh
ensi
ve li
st o
n M
AS
(Frit
z H
ohl,
Stu
ttgar
t):–
Ther
e ex
ist a
ppro
xim
atel
y 70
mob
ile a
gent
s sy
stem
s (c
omm
erci
al &
educ
atio
nal)
.•
Mos
t MA
S a
re im
plem
ente
d in
Jav
a du
e to
pla
tform
inde
pend
ence
. •
Som
e w
ell-k
now
are
: MO
LE, T
eles
crip
t, A
glet
s W
orkb
ench
, ffM
AIN
, D’A
gent
s,
Trac
y, a
nd G
rass
hopp
er.
•H
owev
er, o
nly
few
sup
port
mob
ile d
evic
es:
–G
rass
hopp
er M
E(P
ocke
tPC
)»
The
mos
t com
preh
ensi
ve M
AS fo
r PD
As.
»W
ell-d
ocum
ente
d.–
Age
ntSp
ace
(Poc
ketP
C)
»In
suffi
cien
t doc
umen
tatio
n.–
Trac
y 2
(Poc
ketP
C)
»U
nder
dev
elop
men
t.
•N
o re
al s
uppo
rt fo
r oth
er O
pera
ting
Sys
tem
s (O
Ss)
:–
Fam
iliar L
inux
.–
Embe
dded
Lin
ux.
Mob
ile a
gent
app
licat
ions
•M
obile
age
nt p
arad
igm
has
bee
n pr
opos
ed fo
r a n
umbe
r of
appl
icat
ions
: –
Inte
rnet
-wid
e co
llabo
ratio
n sy
stem
s.–
Net
wor
k m
anag
emen
t.–
Mon
itorin
g sy
stem
s.
–In
form
atio
n re
triev
al.
–In
trusi
on d
etec
tion
syst
ems.
–E
-com
mer
ce.
•A
new
pot
entia
l app
licat
ion
that
mob
ile a
gent
s ar
e id
eal i
s w
irele
ss
netw
orks
–M
obile
age
nts
supp
ort a
sync
hron
ous
com
mun
icat
ion,
and
flexi
ble
quer
y pr
oces
sing
.•
Use
r tas
ks c
an b
e de
lega
ted
to m
obile
age
nts,
and
then
a m
obile
clie
nt c
an
disc
onne
ct.
•C
an re
duce
net
wor
k tra
ffic
unde
r som
e ce
rtain
circ
umst
ance
s.•
Mai
ntai
n lo
ad b
alan
cing
.
JVM
sup
port
forh
andh
elds
•M
ost m
obile
age
nt s
yste
ms
are
deve
lope
d in
Jav
a:–
Plat
form
inde
pend
ence
.•
Porta
bilit
y of
cod
e, w
hich
is th
e na
ture
of m
obile
age
nts.
•Th
eref
ore
a JV
M is
nec
essa
ry to
pro
vide
the:
–Ex
ecut
ion
envi
ronm
ent f
or M
AS
.
Fam
iliar
Linu
x
JRE
1.3
J9
Gra
ssho
pper
ME/
SE
Pock
et P
C
2003
CrE
me
J9
Gra
ssho
pper
ME
Pock
et P
C
2002
Jeod
eGra
ssho
pper
ME
J9C
rEm
e
Fam
iliar
Linu
xFa
mili
arLi
nux
JRE
1.3
J9
Gra
ssho
pper
ME/
SE
JRE
1.3
J9
Gra
ssho
pper
ME/
SE
Pock
et P
C
2003
CrE
me
J9
Gra
ssho
pper
ME
Pock
et P
C
2003
Pock
et P
C
2003
CrE
me
J9
Gra
ssho
pper
ME
Pock
et P
C
2002
Pock
et P
C
2002
Jeod
eGra
ssho
pper
ME
J9C
rEm
e
•IB
M J
9, In
sign
ia J
eode
, NS
Icom
CrE
me,
and
Bla
ckdo
wn
JRE
1.3
are
Sun
ce
rtifie
d JV
Ms.
•Je
ode
is o
nly
avai
labl
e fo
r Poc
ketP
C 2
002.
•J9
doe
s no
t offi
cial
ly s
uppo
rt Fa
milia
r Lin
ux.
•JR
E 1
.3 is
the
impl
emen
tatio
n of
the
full
Java
1.3
ver
sion
. The
rest
are
J2M
E
com
patib
le.
Res
ourc
e-co
nstra
ined
dev
ices
(PD
As)
:
•Li
mite
d ba
ndw
idth
–Av
erag
e w
hen
oper
atin
g in
a c
lient
-ser
ver m
ode.
•Ap
prox
imat
ely
4-5M
bits
/s.
–D
ram
atic
ally
redu
ced
whe
n ac
ting
as a
pro
xy.
•R
ange
s be
twee
n 3-
70K
bits
/s.
•W
orks
tatio
ns a
nd la
ptop
s ac
hiev
e up
to 1
0 tim
es m
ore.
•P
ower
con
sum
ptio
n (F
amilia
r Lin
ux,i
PA
Q).
–W
hen
IDLE
:•
Batte
ry la
sts
5 ho
urs
and
40 m
inut
es.
–W
hen
IDLE
, wire
less
on:
•Ba
ttery
last
s 3
hour
s an
d 20
min
utes
.–
Whe
n ro
utin
g, w
irele
ss o
n:•
Batte
ry la
sts
2 ho
urs.
•C
ompu
tatio
nal d
efic
ienc
y–
Smal
l-sca
le p
roce
ssin
g po
wer
, low
mem
ory
capa
city
, and
slo
w in
tern
al b
usse
s m
ay b
e ju
st a
few
of t
he re
ason
s th
at p
rohi
bit t
he e
xecu
tion
of c
ompl
exca
lcul
atio
ns.
–C
ompl
ex c
ompu
tatio
nal t
asks
suc
h as
the
4D b
ubbl
e so
rt of
30,
000
rand
om in
tege
rs:
•P
DA
requ
ires
15 ti
mes
mor
e tim
e th
an a
lapt
op.
•P
DA
requ
ires
40 ti
mes
mor
e tim
e th
an a
wor
ksta
tion.
Rou
ting
softw
are
vs. P
roxy
•R
outin
g so
ftwar
e:–
It re
quire
s co
mpl
ex a
nd p
ower
inte
nsiv
e ca
lcul
atio
ns.
•It
take
s in
to a
ccou
nt lo
w la
yers
of t
he O
SI m
odel
.•
Res
ourc
e-co
nstra
ined
dev
ices
may
not
be
able
to e
ffici
ently
exe
cute
th
em.
–It
requ
ires
high
impl
emen
tatio
n ef
fort.
•P
roxy
–It
base
s its
func
tiona
lity
on s
impl
e TP
C/IP
soc
ket i
mpl
emen
tatio
n.–
They
can
be
light
-wei
ghte
d.–
They
can
be
effic
ient
ly e
xecu
ted
on re
sour
ce-c
onst
rain
ed d
evic
es.
–P
roxi
es c
an b
e de
sign
ed a
s Ja
va-b
ased
sta
tiona
ry a
gent
s.
Age
nt d
iagr
am –
Ban
dwid
th a
nd R
esou
rces
mon
itorin
g
87
5
1
3
4
6
910 11
File
out
put 8
7
5
2
1
3
4
6
910 11
File
out
put
Libr
aria
n ag
ent
11
Dat
a ga
ther
ing
agen
t10
Band
wid
th c
alcu
latio
n ag
ent
9
TTC
P re
ceiv
er a
gent
8
TTC
P tra
nsm
itter
age
nt7
Prox
y ag
ent
6
Res
ourc
es m
onito
ring
agen
t5
Hea
p m
emor
y us
age
agen
t4
CP
U u
tilis
atio
n ag
ent
3
Batte
ry a
gent
2
Tem
pera
ture
age
nt1
•B
ench
mar
king
ban
dwid
th a
nd R
esou
rce
Con
sum
ptio
ns m
ulti-
agen
t S
oftw
are
Sys
tem
(RC
AS
S)
–is
des
igne
d w
ith p
urpo
se to
ben
chm
ark
band
wid
th p
roxy
-PD
As
in a
d-ho
c ne
twor
ks.
•It
also
has
the
capa
bilit
y of
mon
itorin
g re
sour
ces
cons
umpt
ion
of
the
PD
A in
term
s of
:–
Batte
ry d
isch
arge
rate
, CP
U u
tilis
atio
n, h
eap
mem
ory
usag
e, a
nd
tem
pera
ture
.
Ove
rvie
w o
f the
Ban
dwid
th a
nd R
esou
rces
mon
itorin
g fu
nctio
nalit
y
•TT
CP
rece
iver
list
ens
on a
fixe
d po
rt 10
01.
•P
roxy
age
nts
liste
ns o
n a
fixed
por
t 100
0, a
nd fo
rwar
ds in
com
ing
traffi
c to
the
rece
iver
at p
ort 1
001.
•TT
CP
tran
smitt
er s
ends
net
wor
k tra
ffic
to th
e pr
oxy
on p
ort 1
000.
•O
nce
the
prox
y es
tabl
ishe
s a
conn
ectio
n w
ith th
e tra
nsm
itter
and
rece
iver
:–
It fo
rwar
ds in
com
ing
netw
ork
traffi
c to
the
rece
iver
.–
It cr
eate
s th
e re
sour
ce m
onito
ring
agen
t, w
hich
in tu
rn c
reat
es:
•Te
mpe
ratu
re a
gent
*.•
Batte
ry a
gent
.•
CPU
util
isat
ion
agen
t.•
Hea
p m
emor
y ag
ent.
•Th
e tra
nsm
itter
and
rece
iver
cal
cula
te th
e ba
ndw
idth
thro
ugh
the
prox
y fo
r eac
h ite
ratio
n.•
The
data
gat
herin
g ag
ent c
olle
cts
the
resu
lts a
nd p
asse
s th
em to
the
libra
rian
agen
t.•
The
libra
rian
agen
ts s
tore
s th
e re
sults
in o
utpu
t tex
t file
s.
Pro
xy-b
ased
Ban
dwid
th o
fPD
As
01020304050607080
Fam
iliar L
inux
Poc
ketP
C20
03P
ocke
tPC
2002
01020304050607080
Fam
iliar L
inux
Jeod
e
CrE
me
JRE
1.3
J9Jeod
e
CrE
me
JRE
1.3
J9Jeod
e
CrE
me
JRE
1.3
J9
Poc
ketP
C20
03P
ocke
tPC
2002
Ban
dwid
thK
bits
/s•
The
band
wid
th o
ffere
d by
th
e pr
oxy
devi
ce h
ighl
y de
pend
s on
:–
The
JVM
use
d.–
The
OS
use
d.•
The
high
est b
andw
idth
is
achi
eved
by
Poc
ketP
C 2
003
–C
rEm
e•
The
low
est b
andw
idth
is
achi
eved
by
Jeod
e an
d J9
fo
r Poc
ketP
C 2
002.
Dis
char
ge ra
tes
forP
DA
sw
hile
act
ing
as p
roxi
es
0102030405060708090100
Dis
char
ge
rate
(s) Fa
mili
ar L
inux
Pock
etPC
200
3Po
cket
PC
200
2
0102030405060708090100
Dis
char
ge
rate
(s) Fa
mili
ar L
inux
Pock
etPC
200
3Po
cket
PC
200
2
•Th
e ba
ttery
dis
char
ge ra
te
rang
es fr
om 6
0 (s
) -70
(s).
•Fa
milia
r Lin
ux a
pplie
s be
tter
batte
ry c
onse
rvat
ion
mec
hani
sms:
–Ex
tend
s ba
ttery
life
by
30%
, com
pare
d to
P
PC
2003
.–
Exte
nds
batte
ry li
fe b
y 35
%, c
ompa
red
to
PP
C20
02.
•J9
ext
ends
the
batte
ry li
fe
by 1
5%, c
ompa
red
to
CrE
me
for P
PC20
03.
–H
owev
er, t
he b
andw
idth
is
sign
ifica
ntly
low
er
CP
U u
tilis
atio
n of
PD
As
whi
le a
ctin
g as
pro
xies
0102030405060708090
Fam
iliar L
inux
Pock
etPC
200
3Po
cket
PC
200
2
CPU
(%)
0102030405060708090
Fam
iliar L
inux
Pock
etPC
200
3Po
cket
PC
200
2
CPU
(%)
•C
PU u
tilis
atio
n w
hile
rout
ing
is ra
ngin
g be
twee
n 65
% -
85%
.–
JRE
is th
e he
avie
st J
VM
.–
The
light
est a
re J
eode
and
J9
.–
CrE
me
is h
eavy
as
wel
l re
quiri
ng 8
0% o
f the
CP
U.
•R
outin
g is
a C
PU
-inte
nsiv
e ta
sk.
•Th
e O
S d
oes
not p
lay
a si
gnifi
cant
role
in C
PU
ut
ilisat
ion.
Hea
p m
emor
y us
age
of P
DA
s w
hile
act
ing
as p
roxi
es
0
100
200
300
400
500
600
700
Fam
iliar L
inux
Pock
etPC
200
3Po
cket
PC 2
002
Hea
p (K
B)
0
100
200
300
400
500
600
700
Fam
iliar L
inux
Pock
etPC
200
3Po
cket
PC 2
002
Hea
p (K
B)
•H
eap
mem
ory
is th
e m
emor
y us
ed b
y Ja
va
obje
cts
to p
erfo
rm th
eir
task
s.•
Sign
ifica
nt d
iffer
ence
s ca
n be
obs
erve
d fro
m th
e gr
aph:
–C
rEm
e an
d JR
E 1
.3 a
re
the
mos
t hea
p-in
tens
ive
JVM
s.–
Oth
ers,
are
util
isin
g si
mila
r he
ap m
emor
y.•
Hea
p m
emor
y us
age
seem
s to
hav
e a
stro
ng li
nk
to b
andw
idth
:–
The
high
er th
e H
eap
mem
ory
usag
e th
e hi
gher
th
e ba
ndw
idth
.
Inte
rnal
tem
pera
ture
var
iatio
n of
PD
As
whi
le a
ctin
g as
pro
xies
•Te
mpe
ratu
re re
adin
gs w
ere
take
n fo
r eac
h ba
ttery
di
scha
rge
by 1
%.
•Th
e ov
eral
l var
iatio
n in
ba
ttery
has
bee
n ca
lcul
ated
.•
The
prox
y de
vice
s ru
nnin
g J9
exp
erie
nces
fast
er
incr
ease
s in
the
over
all
tem
pera
ture
.•
This
may
be
asso
ciat
ed to
th
e fa
ct th
at J
9 pr
ovid
es
high
er b
andw
idth
than
JR
E
1.3/
0
0.02
0.04
0.06
0.080.1
0.12
0.14
JRE
1.3
J9
Cel
sius
(C0 )
J9 JRE
1.3
J9 JRE
1.3
J9 JRE
1.3
J9 JRE
1.3
Ove
rvie
w -
Ban
dwid
th o
f C/S
and
Pro
xy-P
DA
s
•C
lient
-Ser
ver,
PD
A:
–4-
5M
bits
/s.
•C
lient
-Pro
xy-S
erve
r, P
DA
:–
3-70
Kbi
ts/s
.•
Clie
nt-P
roxy
-Ser
ver,
Lapt
op (m
ediu
m
pow
ered
dev
ice)
:–
200-
300
Kbi
ts/s
.•
Clie
nt-P
roxy
-Ser
ver,
Wor
ksta
tion
(hig
h po
wer
ed d
evic
e):
–40
0-50
0K
bits
/s.
0123456
PD
A: C
lient
-Se
rver
PDA:
Pro
xy
Ban
dwid
th (M
bits
/s)
Min
imum
Max
imum
200
300
400
500
600
Ban
dwid
th (K
bits
/s)
0
100
Wor
ksta
tion
Lapt
opP
DA
Ove
rvie
w-B
atte
ry li
fe o
fPD
As
•Fa
milia
r Lin
ux.
–ID
LE:
•5
hour
s an
d 25
min
utes
–W
irele
ss:
•3
hour
s an
d 12
min
utes
.–
Rou
ting:
•2
hour
s an
d 18
min
utes
.•
Poc
ketP
C20
03–
IDLE
:•
4 ho
urs
and
32 m
inut
es.
–W
irele
ss:
•2
hour
s an
d 03
min
utes
.–
Rou
ting:
•1
hour
and
50
min
utes
.•
Poc
ketP
C20
02.
–ID
LE •3
hour
s an
d 58
min
utes
.–
Wire
less
:•
1 ho
ur a
nd 5
8 m
inut
es.
–R
outin
g:•
1 ho
ur a
nd 4
5 m
inut
es.
00:0
0:00
01:1
2:00
02:2
4:00
03:3
6:00
04:4
8:00
06:0
0:00
Fam
iliar
PPC
2003
PPC
2002
IDLE
Wire
less
Rou
ting
hh/m
m/s
s
Ove
rvie
w -
Pro
xy re
sults
•Pr
oxy
effic
ienc
y de
pend
on:
–JV
M u
sed.
–O
S u
sed.
•R
esou
rces
con
sum
ptio
n on
pro
xy-P
DA
s is
hig
h in
term
s of
:–
Batte
ry li
fe, C
PU
util
isat
ion,
Hea
p m
emor
y us
age,
and
Tem
pera
ture
.•
Pock
etP
C 2
002
–Je
ode
and
J9 p
erfo
rm e
qual
ly fo
r Poc
ketP
C 2
002.
–C
rEm
e pr
ovid
es a
lmos
t 3 ti
mes
mor
e ba
ndw
idth
and
util
ises
3 ti
mes
mor
e he
ap m
emor
y th
an J
eode
an
d J9
.•
Pock
etP
C 2
003.
–J9
impr
oves
its
band
wid
th b
y 30
% c
ompa
red
to P
ocke
tPC
200
2.–
CrE
me
prov
ides
6 ti
mes
mor
e ba
ndw
idth
com
pare
d to
Poc
ketP
C a
nd u
tilis
es 4
tim
es m
ore
heap
m
emor
y.•
Fam
iliar
Lin
ux.
–J9
pro
vide
s 12
tim
es m
ore
band
wid
th c
ompa
red
to P
ocke
tPC
200
2 an
d 8
times
mor
e co
mpa
red
to
Poc
ketP
C 2
003.
–It
also
add
s 25
min
utes
bat
tery
life
.–
JRE
1.3
is q
uite
effi
cien
t for
suc
h a
heav
y JV
M.
–It
prov
ides
mod
erat
e ba
ndw
idth
and
doe
sn’t
over
-con
sum
e re
sour
ces.
•Th
e st
rong
er c
ombi
natio
n ba
ndw
idth
-wis
e, is
Poc
ketP
C 2
003
with
CrE
me.
•A
seco
nd a
ltern
ativ
e is
Fam
iliar
Lin
ux w
ith J
9, w
hich
ach
ieve
s m
oder
ate
band
wid
th k
eepi
ng
the
utili
satio
n of
reso
urce
s qu
it lo
w.
BA
SS
–B
ench
mar
king
mul
ti-A
gent
Sof
twar
e S
yste
m
•B
AS
S is
a J
ava-
base
d, li
ght-w
eigh
ted,
mul
ti-ag
ent s
oftw
are
syst
em th
at
is a
ble
to ru
n on
mob
ile d
evic
es.
•Th
e pu
rpos
e of
BA
SS
is to
:–
Det
erm
ine
the
fitne
ss o
f mob
ile d
evic
es b
y pe
rform
ing
stan
dard
test
s:•
Gro
up le
vel t
ests
:–
Ope
ratin
g S
yste
m v
ersi
on.
–JV
M v
ersi
on, a
nd s
o on
.•
Kern
el-le
vel t
ests
:–
CP
U te
sts:
bub
ble
sort
and
mer
ge te
st.
–M
emor
y an
d ha
rd-d
rive
test
s.
–M
onito
r util
isat
ion
of d
evic
es:
•N
etw
ork
leve
l tes
ts:
–Ba
ndw
idth
and
late
ncy.
–Er
ror p
acke
ts m
onito
ring.
–In
tern
et c
onne
ctiv
ity.
•U
tilis
atio
n m
onito
ring:
–O
vera
ll m
emor
y an
d C
PU
usa
ge.
–P
rodu
ce m
etric
s w
hich
can
late
r be
used
by
the
rout
ing
prot
ocol
to
dec
ide
on o
ptim
al ro
utes
.
Rou
ting
scen
ario
(1)
Hig
h ba
ndw
idth
Mod
erat
e ba
ndw
idth
Low
ban
dwid
th
Hig
h ba
ndw
idth
Mod
erat
e ba
ndw
idth
Low
ban
dwid
th
S
X Y
Z
W
D
PD
A
PD
A
Lapt
op
PD
A
Dev
ice
JRE
1.3
Fam
iliar L
inux
Z
CrE
me
Pock
etPC
2002
W
JRE
1.4
Win
dow
s XP
Y
CrE
me
Pock
et P
C 2
003
X
JVM
OS
Nod
e
D
XS
Y
Z W
J9Fa
milia
r Lin
uxZ
J9Po
cket
PC20
02W
Jeod
ePo
cket
PC20
02Y
CrE
me
Pock
et P
C 2
003
X
JVM
OS
Nod
e
•O
ptim
al ro
ute
from
S to
D:
–S
-> Y
->W
or Z
-> D
•O
ptim
al ro
ute
from
S to
D:
–S
-> X
-> Z
-> D
Rou
ting
scen
ario
(2)
low
low
low
high
Mem
ory
usag
e
low
low
high
Z WYXNod
e
mod
erat
ehi
ghlo
w
low
low
hi
gh
mod
erat
e
Bat
tery
lif
e
low
low
Tem
pera
ture
CPU
ut
ilisa
tion
J9Fa
milia
r Lin
uxZ
J9Po
cket
PC20
02W
Jeod
ePo
cket
PC20
02Y
CrE
me
Pock
et P
C 2
003
X
JVM
OS
Nod
e
S
X Y
Z
W
D
•O
ptim
al ro
ute
from
S to
D,
band
wid
th-w
ise:
–S
-> X
-> Z
-> D
–H
owev
er, X
has
:•
Low
bat
tery
.•
Hig
h C
PU u
tilis
atio
n.
•H
igh
mem
ory
usag
e.
•To
be
ener
gy e
ffici
ent,
and
light
-wei
ghte
d:–
Rou
te w
ill be
:•
S ->
Y ->
Z ->
D
Con
clus
ions
•W
e ha
ve p
rove
d th
e co
ncep
t of a
pro
xy-P
DA
and
show
ed th
at it
may
be
a go
od a
ltern
ativ
e to
pro
per r
outin
g so
ftwar
e du
e to
its:
–Lo
w im
plem
enta
tion
effo
rt.–
Ligh
t-wei
ghte
d na
ture
.•
We
have
por
ted
vario
us J
VM
s to
var
ious
OS
s fo
r iP
AQ
PD
As.
–Fa
milia
r Lin
ux:
•Bl
ackd
own
JRE
1.3
and
IBM
J9.
–Po
cket
PC
200
3:•
IBM
J9
and
NSI
com
CrE
me.
–Po
cket
PC
200
2.•
Insi
gnia
Jeo
de, I
BM
J9,
and
NS
Icom
CrE
me.
•W
e ha
ve p
rove
d th
at th
e O
S a
nd th
e JV
M u
sed
can
play
an
impo
rtant
role
in th
e ef
ficie
ncy
of th
e pr
oxy-
PD
A in
term
s of
ban
dwid
th, b
atte
ry d
isch
arge
rate
, CP
U u
tilis
atio
n, h
eap
mem
ory,
and
tem
pera
ture
.•
We
have
iden
tifie
d th
e be
st p
air a
s be
ing:
Poc
ketP
C 2
003
with
CrE
me
and
as a
n al
tern
ativ
e to
low
er u
tilis
atio
n of
PD
A’s
reso
urce
s w
ith m
oder
ate
band
wid
th b
eing
: Fam
iliar
Lin
ux w
ith
J9.
•W
e ha
ve s
how
n th
e im
porta
nce
of B
AS
S, w
hich
can
be
inco
rpor
ated
with
MA
RIA
N, i
n or
der
to a
ssis
t in
met
ric-d
riven
rout
ing
taki
ng in
to a
ccou
nt:
–Pe
rform
ance
cha
ract
eris
tics.
–C
urre
nt U
tilis
atio
n.•
Exte
nsiv
e re
sear
ch w
ill no
w b
e co
ncen
trate
d in
bui
ldin
g up
a s
cala
ble
mod
el fo
r dire
ct a
nd
indi
rect
com
mun
icat
ion
rang
e ne
twor
k di
scov
ery
and
indi
rect
. •
The
next
sta
ge in
clud
es s
imul
atin
g M
AR
IAN
usi
ng N
etw
ork
Sim
ulat
ion
(NS
2)
and
com
pare
its
perfo
rman
ce to
trad
ition
al ro
utin
g pr
otoc
ols
such
as
DS
R,
AO
DV
, and
so
on.
Pub
licat
ions
•IE
EE E
CBS
200
3–
Mob
ile A
gent
s fo
r Rou
ting,
Top
olog
y D
isco
very
, and
Aut
omat
ic N
etw
ork
Rec
onfig
urat
ion
in A
d-ho
c N
etw
orks
.•
IAD
IS W
WW
/Inte
rnet
200
3–
MA
RIA
N: A
fram
ewor
k us
ing
Mob
ile
Age
nts
for R
outin
g In
Ad-
hoc
Net
wor
ks.
•IE
EE E
CBS
200
4–
Mig
ratio
n of
Mob
ile A
gent
s in
Ad-
hoc,
Wire
less
Net
wor
ks•
Else
vier
Jou
rnal
of A
d-H
oc
Net
wor
ks (2
004)
–An
alys
is o
f an
Age
nt-b
ased
Met
ric-
Driv
en M
etho
d fo
r Ad-
hoc,
On-
Dem
and
Rou
ting
Que
stio
ns
Than
k yo
u !!!