Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
8
, No
.
6
,
Decem
ber
201
8
, p
p.
4164
~
4174
IS
S
N:
20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v
8
i
6
.
pp
4164
-
41
74
4164
Journ
al h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
Reconfi
gu
ra
ble
High Pe
rform
ance
Secu
re
d
NoC De
sign
U
sing
Hierar
ch
ical
Agent
-
b
ase
d Monit
or
in
g System
Kend
aganna
Swamy
S
,
An
an
d
Jatti,
Um
a
B.
V
Dep
ar
t
m
ent
o
f
E
le
c
tri
c
al &
Com
m
unic
at
ion
Eng
g,
R.
V
C
ollege
of E
ng
i
neer
i
ng
,
India
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ma
r
3
, 2
01
8
Re
vised
Ju
l
18
,
201
8
Accepte
d
Aug
7
, 2
01
8
W
it
h
the
rap
id
i
ncr
ea
se
in
dema
nd
for
high
per
form
anc
e
computing,
the
r
e
is
al
so
a
signif
ic
an
t
growth
of
data
comm
unic
at
ion
tha
t
le
ads
to
levera
ge
th
e
signifi
c
anc
e
of
net
work
on
ch
ip
.
Thi
s
p
ape
r
pro
poses
a
re
conf
ig
ura
ble
fau
l
t
tol
er
ant
on
ch
ip
arc
hi
te
c
ture
with
hie
r
arc
hi
cal
agent
base
d
m
onit
or
ing
s
y
st
em
for
enha
nc
ing
th
e
per
form
anc
e
o
f
net
work
base
d
m
ult
iproc
essor
s
y
stem
on
chi
p
against
fau
l
t
y
l
inks
and
nodes.
The
se
distr
ib
ute
d
age
n
ts
provide
heal
t
h
y
stat
us
and
conges
ti
on
informati
o
n
of
the
net
work.
Thi
s
sta
tus
inf
orm
at
ion
is
used
for
furthe
r
pac
ke
t
routi
ng
i
n
the
net
work
with
the
he
lp
of
XY
routi
ng
al
gorit
hm
.
The
func
ti
on
al
i
t
y
of
Agent
is
enhance
d
not
onl
y
t
o
work
as
informati
on
prov
ide
r
but
a
lso t
o
t
ake
d
ec
ision
for packet
to
either
pass or
stop
to
the
pro
c
essin
g
el
ement
b
y
s
etting
the
fire
w
al
l
in
orde
r
to
prov
i
de
sec
uri
t
y
.
Propos
ed
design
provide
s
a
be
t
te
r
per
form
an
ce
and
area
opt
i
m
iz
at
ion
b
y
avoi
ding
d
ea
dlo
ck
and
li
ve
lock
as
compare
d
to
exi
st
ing
appr
oac
hes
ov
e
r
net
work de
sign
.
Ke
yw
or
d:
Congesti
on
Fault
i
de
ntific
at
ion
Fault
t
olera
nt
Netw
ork
f
ault
Netw
ork
-
on
-
c
hip
Rou
ti
ng
Copyright
©
201
8
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Kenda
ganna
S
wam
y S
,
Dep
ar
t
m
ent
of
Ele
ct
rical
&
I
nst
um
entat
ion
,
R.V
C
ollege
of Enginee
rin
g,
Ba
ng
al
or
e,
In
di
a
.
Em
a
il
:
ken
da
ga
nn
a
swam
ys@rv
ce
.edu.i
n
1.
INTROD
U
CTION
Since
la
st
50
y
ears,
as
a
res
ul
t
of
a
dvancem
ent
in
sem
ic
ondu
ct
or
te
c
hnol
og
y,
scal
ing
co
ntinu
e
s
f
ro
m
tod
ay
’s
16nm
featur
e
siz
e
to
1nm
featur
e
siz
e
exp
ect
e
d
in
2028
[1
]
.
T
his
e
na
bles
to
inte
gr
a
te
m
or
e
nu
m
be
r
of
IP
c
or
es
in
a
sing
le
syst
em
o
n chip
. W
it
h
the
g
r
owth
of
num
ber
o
f
c
ores,
com
m
un
ic
at
ion
dem
and
b
et
w
een the
processi
ng
c
or
es
increase
s.
T
his
m
ay
req
uir
e
high
c
omm
u
nicat
ion
ba
ndwidth
with
l
ow
la
te
ncy,
lo
w
powe
r
consum
ption
a
nd
hi
gh
scal
a
bili
ty
network.
The
co
nventi
on
al
bus
base
d
a
rch
it
ect
ure
will
no
t
m
eet
these
requirem
ents
and
this
le
ad
to
com
m
un
ic
ation
perf
or
m
ance
bo
tt
le
neck.
A
so
luti
on
f
or
suc
h
a
com
m
un
i
cat
io
n
bo
tt
le
nec
k
is
netw
ork
on
c
hi
p
to
im
pr
ov
e
the
pe
rfor
m
ance
f
or
m
any
cor
e
syst
em
s
[
2
]
.
As
c
om
pared
to
pr
e
vious
works
pr
ese
nted
i
n
[
3
]
,
[
4
]
,
N
oC
i
s
the
popula
r
i
nterc
onnecti
on
infr
ast
ru
ct
ur
e
for
m
any
cor
e
inter
com
m
un
ic
at
ion
beca
us
e
of
it
s
high
thr
ough
pu
t,
l
ow
la
te
nc
y,
scal
abili
ty
a
nd
reusabil
it
y.
No
Cs
a
re
com
po
s
ed
with
three
co
m
po
nen
ts
su
c
h
as
ro
uter
,
li
nks
and
net
wor
k
interface
(
NI).
Rou
te
r
s
are
th
e
switc
hing
el
e
m
ents
that are
res
ponsi
ble for
forwa
rd
i
ng the
data
pack
et
s
fro
m
one
router t
o
a
nothe
r on
e
.
Link
s
a
re
the
connecti
on
pa
r
ts
between
dif
fer
e
nt
n
odes
a
nd
they
are
usual
ly
bid
irect
ion
al
net
wor
k
interfaces
,
w
hi
ch
act
s
as
the
wr
a
pper
betwe
en
the
r
oute
r
a
nd
processi
ng
el
e
m
ents
(P
E).
Rou
te
rs
will
ta
ke
the
routin
g
de
ci
sio
n
based
on
t
he
routin
g
al
gorit
hm
.
In
No
C
ba
sed
m
ulti
ple
cor
e
syst
e
m
s,
th
e
ne
gative
a
s
pe
ct
s
of
te
chnolo
gy
sc
al
ing
m
ay
inc
rease
the
pro
bab
il
it
y
of
c
hi
p
de
fects
int
rod
uced
w
hich
m
ay
be
ei
t
her
i
n
op
e
rati
onal
or
in
m
anu
factu
ri
ng
phases
.
T
he
se
fa
ulty
No
C
syst
e
m
s
m
ay
hav
e
def
ect
s
i
n
processi
ng
el
em
ents
(P
E
)
or
r
ou
te
rs
or
inte
rcon
nec
ts.
D
ue
t
o
t
h
e
f
aulty
interco
nn
ect
s
an
d
router
s,
the
num
ber
of
r
ou
ti
ng
pat
hs
are
reduce
d,
wh
ic
h
res
ults
in
unbalance
d
tra
ff
i
c
distrib
utio
n
and
m
or
e
tra
ffi
c
congesti
on
[
5
]
.
T
he
la
ck
of
non
local
fau
lt
awa
ren
es
s
le
ads
to
perform
ance
degra
dation
i
n
No
C.
T
he
perform
ance
pa
ra
m
et
ers
are
bec
om
ing
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
Reco
nfigu
r
abl
e H
ig
h Perf
orma
nce
Sec
ur
e
d N
oC
Desig
n U
sing Hier
ar
c
hical
…
(
Ke
ndag
anna
Swamy
S)
4165
i
m
po
rtant
asp
e
ct
s
in
m
ulti
ple
cor
e
chi
p
desi
gn.
In
t
his
pro
po
s
ed
desig
n
the
rec
onfig
ur
a
ble
high
pe
rform
ance
secur
e
d
No
C
desig
n
us
in
g
hierar
c
hical
A
gen
t
base
d
m
on
it
or
i
ng
syst
e
m
pr
ovides
a
prom
isi
ng
so
l
ution
t
o
address
the
a
bove
issue
[
6
]
,
w
hich
is
su
it
able
f
or
la
r
ge
m
ul
ti
cor
e
sys
tem
s
with
hu
ndre
ds
of
proce
ssing
el
e
m
ents.
In
th
is
design,
age
nt
s
are
distribu
t
ed
hiera
rch
ic
al
ly
to
accu
m
ula
te
,
distribu
te
a
nd
m
anag
e
the
fau
lt
y
inf
or
m
at
ion
along
with
secu
r
it
y using ra
ndom
arb
it
er r
oute
r
wit
h
XY
r
ou
t
ing
al
gorithm
.
The
pr
e
vious
works
are
rela
te
d
to
t
he
hier
arch
ic
al
a
gen
t
s
f
ound
in
[
7
]
-
[
11
]
.
I
n
[
6]
a
nd
[
1
1
]
t
he
ov
e
rall
str
uctu
re
of
age
nt
ba
sed
m
anag
e
m
ent
syst
e
m
is
discu
ssed
without
a
ny
detai
le
d
desig
n.
The
hierar
c
hical
a
gen
ts
a
re
use
d
in
[
8
]
an
d
[
7
]
to
m
on
it
or
the
po
wer
consum
ption
i
n
N
oC
usi
ng
DVFS
(d
if
fer
e
ntial
vo
lt
age
an
d
fr
e
quency
scal
in
g)
t
echn
i
qu
e
.
In
[
1
0
]
an
a
ge
nt
ba
s
ed
m
anag
em
ent
m
et
ho
d
is
us
e
d
to
enh
a
nce
the
pe
rfor
m
ance
of
No
C
base
d
m
ul
ti
cor
e
sys
tem
on
chip
de
sign
agai
ns
t
the
fau
lt
s
or
f
ai
lures
resu
lt
ed
i
n
the
neig
hbor
no
de
s
in
ad
diti
on
t
o
their
own
c
om
po
nen
ts
a
nd
interco
nnect
io
n
li
nk
s
.
T
hese
agen
ts
inf
or
m
the
rou
te
rs
ab
ou
t
dif
f
eren
t
fau
lt
s
in
the
net
work
w
hich
helps
t
he
routin
g
pr
oces
s
to
be
m
or
e
s
cal
able
us
in
g
XY
r
ou
t
ing
al
gorithm
and
al
so
t
o
im
pro
ve
the
perf
or
m
ance.
H
ow
ever,
sti
ll
m
any
issues
nee
d
to
be
addresse
d.
Pr
e
vious
works
ar
e
lim
i
te
d
to
4x
4
a
gen
t
base
d
No
C,
no
n
rec
onfi
gurab
le
a
nd
non
se
cu
red
a
gen
ts
.
The
a
rb
it
ers
use
d
in
t
he
pre
vio
us
r
ou
te
r
are
no
t
se
rv
ic
i
ng
t
he
pa
ckets
e
qu
al
ly
i
n
al
l
directi
on
s
of
t
he
no
de
a
nd
it
serv
es
t
he
pa
cket
acco
r
ding
to
t
he
pri
ori
ty
wh
ic
h
m
ay
l
ead
to
i
ncr
ease
in
pac
ket
sta
ki
ng
i
n
one
dire
ct
ion.
The
a
ge
nt prov
ides
on
ly
the
c
ongestio
n
a
nd
healt
hy stat
us
of the
netw
ork.
In
the
pr
opose
d
desi
gn
al
l
these
lim
i
ta
ti
o
ns
are
ad
dr
e
ss
ed,
by
intr
oduc
ing
rec
onfig
urable
N
xN
hierar
c
hical
a
ge
nt
ba
sed
N
oC
with
ra
ndom
arb
it
er
r
ou
te
r
usi
ng
X
Y
routin
g
al
go
rithm
,
wh
ic
h
overc
om
e
s
the
pack
et
sta
cki
ng
by
ser
vicing
the
pack
et
ra
ndom
ly
,
wh
ic
h
avo
ids
lo
ss
of
pack
et
s
an
d
im
pr
ov
es
the
m
e
m
or
y
area.
T
he
a
ge
nt
functi
onal
it
y
is
furthe
r
e
nh
a
nced
to
wor
k
as
an
i
nfor
m
ation
pro
vid
e
r
a
nd
al
so
ta
ke
de
ci
sion
for
pac
kets
to
ei
ther
pass
or
stop
to
t
he
pr
ocessin
g
el
em
ent
by
set
ti
ng
the
fire
wall
w
hich
inte
rn
prov
i
des
secur
it
y.
Sect
ion
1.1
discu
s
ses
about
the
existi
ng
li
te
ratur
es
wh
e
re
di
f
fer
e
nt
te
chn
i
ques
are
discusse
d
f
or
detect
ion
sc
he
m
es
us
ed
in
powe
r
tra
ns
m
iss
ion
li
nes
f
ollo
wed
by
disc
ussi
on
of
resea
rc
h
pr
ob
le
m
s
in
Sect
ion
1.2
a
nd
pr
opose
d
s
olu
ti
on
i
n
1.3.
Sect
ion
2
discusse
s
a
bout
al
gorithm
i
mp
l
em
entat
ion
f
ollow
e
d
by
dis
cussion
of r
es
ult analy
s
is i
n
Sect
io
n 3. Fi
nally
, th
e
conclusi
ve
rem
ark
s a
re
pro
vid
e
d
in
Secti
on
4.
1.1.
Back
ground
This
sect
io
n
di
scusses
a
bout
the
e
xisti
ng
appr
oach
es
f
or
s
olv
in
g
t
he
identific
at
ion
pro
blem
s
of
netw
ork
relat
e
d
fa
ults
.
T
he
work
car
ried
out
by
Sa
nto
s
e
t
al
.
[12
]
has
pr
ese
nted
a
m
echan
ism
to
identify
m
axi
m
iz
d
i
m
p
eden
ce
fau
lt
s
usi
ng
discrete
w
avelet
transfo
r
m
.
Stud
y
towa
rd
ide
ntific
at
io
n
of
real
-
ti
m
e
fau
lt
s
has
al
s
o
bee
n
ca
rr
ie
d
out
by
Pign
at
i
et
al
.
[
13
]
ov
e
r
sim
ilar
dis
tri
bu
te
d
netw
ork
us
in
g
sta
te
-
base
d
est
i
m
ation
te
chn
iq
ue.
Si
m
il
ar
fo
rm
of
appr
oach
wa
s
al
so
i
m
ple
m
ented
by
Nikand
er
and
Ja
rv
e
nt
austa
[14
]
for
netw
ork
fau
lt
ide
ntific
a
ti
on
.
C
onside
ring
a
case
stu
dy
of
sp
a
cecra
f
t,
Ra
it
eri
an
d
Po
rti
nale
[
1
5
]
hav
e
use
d
Ba
ye
sia
n
netw
ork
f
or
id
entify
ing
a
nd
m
itigati
ng
fa
ults
occu
r
rin
g
ov
er
sp
acec
raf
t.
Re
search
to
wa
rd
s
e
xp
li
ci
t
analy
sis
of
beh
a
viou
r
of
a
pack
et
is
carr
ie
d
ou
t
by
W
a
ng
et
al
.
[1
6
]
us
in
g
a
un
i
qu
e
fo
rm
of
cl
assifi
cat
ion
te
chni
qu
e
.
Ado
ption
of
pr
ob
a
bili
ty
theory
has
be
e
n
us
e
d
for
devel
oping
a
fr
am
ewo
r
k
for
id
entify
in
g
fau
lt
s
over
s
ens
or
y
app
li
cat
io
n
as
witnesse
d
in
th
e wor
k of Ntal
a
m
piras
[
1
7
]
.
The
a
uthors
ha
ve
use
d
Hidd
en
Ma
r
kov
M
od
el
for
this
pur
pose.
Z
ha
ng
and
Zha
ng
[
18
]
ha
ve
us
e
d
gr
a
ph
-
base
d
appr
oach
f
or
de
velo
ping
a
fr
a
m
ewo
r
k
of
fa
ult
identific
at
ion
ta
ki
ng
t
he
case
stud
y
of
sat
el
li
te
netw
ork.
The
occura
nces
of
netw
ork
fau
lt
is
al
so
in
vestig
at
ed
over
an
opti
cal
netw
ork
by
Am
aral
et
al
.
[1
9
]
wh
e
re
the
aut
hors
ha
ve
us
e
d
sp
eci
fic
dev
ic
e
to
acc
om
plish
the
ta
sk.
Sim
i
la
r
stu
dy
towa
rd
s
opti
cal
net
work
has
bee
n
al
so
s
tud
ie
d
by
Z
hu et
al
.
[20
]
, w
he
re
a
m
at
he
m
at
i
cal
m
od
el
ing
ha
s
bee
n
util
iz
ed
f
or
de
vel
op
i
ng
tw
o
dim
ension
al
cod
i
ng
-
m
on
it
or
i
ng
syst
e
m
s.
A
doptio
n
of
Ba
ye
sia
n
netw
ork
is
agai
n
see
n
f
or
t
he
work
carrie
d
ou
t
by
Ca
i
et
al
.
[2
1
]
.
The
re
hav
e
been
al
s
o
stu
dies
towa
rd
s
de
velo
ping
fau
lt
tolera
nc
e
syst
e
m
in
ex
ist
ing
li
te
ratur
e.
C
on
siderin
g
the
ca
se
stu
dy
of
chi
p
s
witc
hing,
K
oh
le
r
et
al
.
[
2
2
]
hav
e
dev
el
oped
a
n
fa
ult
tol
eran
t
m
od
el
fo
r
im
p
rovin
g
Net
wor
k
-
on
-
c
hip
pe
rfor
m
ance.
Vall
et
al
.
[2
3
]
ha
ve
dev
el
ope
d
an
est
i
m
ation
te
ch
nique
of f
a
ults occ
urr
ing
i
n
se
nsory
netw
ork.
Yao
et
al
.
[
2
4
]
hav
e
li
nea
r
sta
te
feed
bac
k
m
echan
is
m
fo
r
de
velo
pin
g
a
c
on
t
ro
ll
e
r
syst
e
m
of
sign
ific
a
nt
fa
ul
ts
occ
urrin
g
ov
er
net
w
ork
ar
c
hitec
ture.
Eg
hb
al
et
al
.
[2
5
]
h
a
ve
ca
rr
ie
d
out anlay
sis
of
net
work
-
on
-
c
hip
a
rch
it
ect
ur
e
for
ov
e
rco
m
ing
va
rio
us
hard
war
e
r
el
at
ed
issues
on
c
hip
desig
n.
Re
n
et
al
.
[2
6
]
ha
ve
pr
ese
nted
a
n
adap
ti
ve
c
om
m
un
ic
at
ion
str
at
egy
to
over
com
e
fau
lt
s
fo
r
m
it
igati
ng
dead
l
ock
c
on
diti
on
.
Shuw
ai
li
et
al
.
[2
7
]
ha
ve
disc
us
se
d
a
bout
fa
ult
tolera
nce
m
echan
ism
for
ne
twork
f
un
ct
io
n
virtu
al
iz
at
ion
us
i
ng
cod
i
ng
-
based
a
ppr
oa
ch
.
Sim
ilarly
Pereira
et
al
.
[
28
]
a
nd
Wu
et
al
.
[
29
]
ha
ve
al
s
o
pr
ese
nt
ed
a
m
echan
is
m
of
fau
lt
tole
ran
ce
syst
e
m
fo
r
chi
p
a
nd
se
nsor
node
s
r
e
sp
ect
iv
el
y.
The
refor
e
,
it
can
be
see
n
that
ther
e
a
re
va
rio
us
reser
a
hers
w
ho
ha
ve
al
rea
dy
carried
out
stud
ie
s
to
ward
s
i
m
pr
ovin
g
the
pe
rfor
m
ance
of
fa
ult
tolerance
associat
ed
with
the
net
wor
k
syst
e
m
especial
ly
the
chip
-
base
d
arc
hitec
ture.
Eac
h
a
ppro
ac
h
ha
s
t
hei
r
own
un
i
qu
e
ness
as
well
as lim
it
ati
on. T
he ne
xt se
ct
ion
ou
tl
ines
the
pro
blem
s as
so
ci
at
ed wit
h
t
he
e
xisti
ng r
es
earch
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
201
8
:
4164
-
4174
4166
1.2.
Rese
arch
Pr
obl
em
The
si
gn
i
ficant
r
esea
rch p
robl
e
m
s ar
e as
fo
ll
ow
s:
a.
Existi
ng
resea
r
ch
to
wards
fa
ul
t
tolerance
do
esn’
t
em
ph
asi
z
e
on
the
scal
abili
ty
wh
il
e
ev
olv
in
g
up
with
fau
lt
tole
ran
t
prot
oco
l
over
ne
twork
d
esi
gn.
b.
None
of
the
e
xi
sti
ng
stu
dies
t
ow
a
r
ds
N
oC
ha
s
highli
ghte
d
any
desi
gn
iss
ues
with
it
s
pr
ocessin
g
el
em
e
nts
that o
ff
e
r
la
te
nt
f
aults i
n a
ny
netw
ork
a
rch
it
ect
ur
e.
c.
Althou
gh
exis
ti
n
g
stu
dies
ha
ve
work
e
d
on
fa
ult
ide
ntif
ic
at
ion
but
th
ere
are
le
s
s
num
ber
of
stu
di
es
towa
rd
s
classi
f
yi
ng
the
f
a
ults
existi
ng over
t
he netw
orks.
d.
The
m
echan
is
m
of
form
ulatin
g
the
decisi
on
in
en
surin
g
be
tt
er
perform
ance
of
fa
ult
tole
ran
ce
net
work
i
s
no
t cl
ea
rly
de
fined in a
ny e
xis
ti
ng
st
ud
ie
s.
Ther
e
f
or
e,
the
pro
blem
sta
teme
nt o
f
th
e pr
opos
e
d
stu
dy c
an
be
stat
ed
as
“Dev
el
op
i
ng a
cost ef
fecti
ve
m
od
el
ing
to
encapsule
com
pr
e
he
ns
ive
net
work
fau
lt
s
w
it
h
equ
ivale
nt
fo
cu
s
on
pac
ket
-
le
vel
co
ntr
olli
ng
m
echan
ism
in
chip
a
r
c
hitec
ture
is c
om
pu
ta
ti
on
al
ly
ch
al
le
nging
.”
1.3.
Prop
os
ed
S
olu
tion
The
pr
im
e
aim
of
the
pro
pose
d
syst
em
is
to
dev
el
op
a
sim
ple
an
d
novel
appr
oach
that
can
op
ti
m
iz
e
the
pe
rfo
rm
ance
of
the
net
w
ork
by
perfor
m
ing
inte
gr
at
e
d
op
e
rati
ons
over
the
netw
or
k.
With
an
ai
d
of
a
n
analy
ti
cal
m
od
el
ing
,
t
he
pr
opos
e
d
syst
em
p
erfor
m
s
a
series
of
operati
on
e.
g.
i)
ide
ntific
at
ion
of
fa
ul
ts,
ii
)
identific
at
ion
of
t
raffic
bott
le
neck
c
onditi
on
s
,
ii
i)
inc
orporati
ng
pace
y
-
le
vel
secu
rity
,
an
d
iv
)
ef
f
ect
ive
m
on
it
or
ing
of
the
ongoin
g
co
m
m
u
nicat
ion
.
The
pro
pose
d
syst
e
m
act
s
li
k
e
a
com
pli
m
entary
m
od
el
to
assist
s
the ro
uter fo
r f
or
m
ulati
ng
a
preci
se d
eci
si
on. T
he
sc
hem
a o
f
the
pro
pose
d sy
stem
is as shown i
n
Fi
g
ure
1.
R
A
M
(
1
6
x
1
6
B
y
t
e
s
)
B
y
p
a
s
s
R
e
g
i
s
t
e
r
S
e
s
s
i
o
n
M
o
n
i
t
o
r
i
n
g
C
o
n
f
i
g
.
R
e
g
i
s
t
e
r
C
o
n
t
r
o
l
P
a
c
k
e
t
B
y
p
a
s
s
/
n
e
g
l
e
c
t
s
p
e
c
i
f
i
c
t
a
r
g
e
t
f
r
o
m
p
a
c
k
e
t
C
o
u
n
t
U
p
/
D
o
w
n
Figure
1
.
Sc
he
m
a o
f Pr
opos
e
d
Me
th
od
The
pr
opos
e
d
schem
e
assist
s
in
ag
gr
e
gatin
g,
m
anag
in
g,
a
nd
distrib
utin
g
the
inf
or
m
at
ion
relat
ed
t
o
netw
ork
fau
lt
s
us
in
g
L
ocal
Fa
ult
Re
gister
(
L
FR)
w
hile
it
ta
kes
the
hel
p
of
Re
gional
Fa
ult
Re
gister
(RF
R)
f
or
perform
ing
updatin
g
op
e
ra
ti
on
on
it
s
ne
ighborin
g
node
s.
T
he
cel
l
agen
t
e
xch
a
ng
e
s
the
c
ongesti
on
inf
or
m
at
ion
bi
directi
onal
between
the
age
nt
s
by
us
in
g
sa
m
e
li
nk
in
side
the
de
dicat
ed
net
work
[
3
0].
The
congesti
on
in
f
or
m
at
ion
or
t
he
fau
lt
inf
or
m
at
ion
is
deter
m
ined
by
the
agen
ts
with
th
e
help
of
e
ncodin
g
an
d
deco
ding
proce
ss.
This cell
age
nt w
il
l pro
vide
the securit
y to
the
processi
ng elem
ent u
sing
con
fig
re
gist
er and
con
t
ro
l
pac
ket
sta
ge.
Config
reg
ist
er
is
us
e
d
for
sou
rce
port
co
nf
i
gurati
on
(
us
i
ng
lo
ok
up
ta
ble
co
nce
pt)
in
order
to
bl
ock
the
unwa
nted
an
d
unrelat
e
d
pac
kets
to
gi
ve
sec
ur
it
y
(li
ke
blo
c
king
t
he
websi
te
or
vir
us
pack
et
s
).
Co
nt
ro
l
pac
ket
will
get
the
auth
ori
zed
pac
ket
inf
or
m
at
ion
fro
m
the
con
fi
g
reg
ist
er
a
nd
de
ci
des
wh
et
her
pac
ke
t
m
us
t
be
pa
ss
ed
or
not
to
t
he
processin
g
el
e
m
ent.
In
ge
ner
al
,
pe
op
le
c
an
h
ack
t
he
se
cur
e
d
firew
al
l,
bu
t
in
the
pro
posed
desig
n,
s
om
e
of
the
po
rt
addr
esses
are
it
sel
f
blo
c
ked
i
n
the
hard
war
e
(
i.e
insid
e
the
chi
p)
,
w
hic
h
av
oi
ds
the
i
nt
ruder
by
hac
ki
ng
t
he
fi
rew
al
l.
The
cel
l
age
nt
will
ignore
or
bypass
so
m
e
of
the
pack
et
s
,
if
thos
e
pack
et
s
co
ntain
vid
e
o
or
aud
io
relat
e
d
data
us
ing
bypass
re
gister.
The
age
nts
wi
ll
al
so
m
on
it
or
the
m
axim
u
m
sessio
ns
pe
r
no
de
usi
ng
sessi
on
m
on
it
or
i
ng
sta
ge.
This
sessio
n
m
on
it
ori
ng
sta
ge
will
ta
ke
care
of
st
art
sessio
n
a
n
d
cl
ose
sessi
on
(lim
it
ed
to
0
-
31
ses
sio
ns
)
after
perform
i
ng
the
ta
s
k.
T
he
nex
t
sect
ion
outl
ines abo
ut the al
gorithm
u
sed
fo
r
this
pur
pose.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
Reco
nfigu
r
abl
e H
ig
h Perf
orma
nce
Sec
ur
e
d N
oC
Desig
n U
sing Hier
ar
c
hical
…
(
Ke
ndag
anna
Swamy
S)
4167
2.
ALGO
RITH
M
I
MPLEME
NTATIO
N
This
inf
or
m
at
i
on
relat
e
d
to
the
fau
lt
in
the
network
is
qu
it
e
us
efu
l
for
the
router
to
f
orm
ulate
an
eff
ect
ive
decis
ion
m
aki
ng
du
rin
g
r
outi
ng.
The
fau
lt
dete
ct
ion
ci
rc
uitry
in
the
age
nt
will
pro
vid
e
t
he
fau
lt
inf
or
m
at
ion
of
the
netw
ork.
A
N
oC
r
ou
te
r
is
assum
ed
to
con
ta
in
a
pr
i
ori
ty
enco
de
r,
ra
ndom
arb
it
er
and
t
he
cro
ss
ba
r
s
witc
h.
As
c
om
par
ed
to
te
c
hn
i
qu
e
in
existi
ng
sy
stem
[1
0
]
,
[31
]
,
[32],
in
t
his
pa
per
t
he
pr
opose
d
router
desig
n
buff
e
rs
are
a
voide
d
to
re
duc
e
the
hard
ware
ov
e
rh
ea
d
a
nd
to
im
pr
ov
e
t
he
pe
rfor
m
anc
e.
The
pr
i
or
it
y
enc
oder
wh
ic
h
sel
ec
ts
the
in
puts
a
ccordin
g
t
o
th
e
sel
ect
signa
l
ori
gin
at
es
fro
m
the
ra
ndom
arb
it
er
.
This
pro
posed
r
ou
te
r
will
serv
e
the
pac
ket
s
ra
ndom
ly
without
a
ny
loss
or
sta
c
king
of
the
pack
et
s.
W
it
h
ref
e
ren
ce
to
[7
],
[3
3]
fa
ult
detect
ion
ci
rcu
it
ry
is
ado
pte
d
in
the
N
oC
router
and
the
li
nks
a
re
us
e
d
to
dete
ct
the
per
m
anen
t f
aul
ts o
n
the
n
et
work with
an ac
ce
ptable
hard
ware o
verhea
d.
The
fau
lt
detect
ion
ci
rc
uitry
will
pro
vid
e
t
he
a
pprop
riat
e
sign
al
s
,
w
hic
h
giv
es
t
he
in
form
ation
of
fau
lt
awa
ren
es
s
relat
ed
to
ra
ndom
arb
it
er,
pr
iority
encoder
,
cro
ss
bar
s
witc
h
an
d
al
l
li
nk
s
in
fou
r
directi
on
of
each
r
ou
te
r.
I
n
ad
diti
on
to
this,
it
al
so
prov
i
des
in
form
at
ion
of
fa
ulti
ness
of
ot
her
com
po
ne
nts
s
uch
a
s
Pr
oc
essin
g
Ele
m
ent
(P
E)
or
cor
e
a
nd
Net
w
ork
I
nter
face
(
NI).
In
t
he
propose
d
desig
n,
al
l
the
li
nk
s
of
t
he
netw
ork
a
re
bid
irect
io
nal
an
d
if
any
per
m
an
ent
fa
ult
occ
urs
in
a
ny
one
di
recti
on,
the
n
t
he
entire
li
nk
w
il
l
be
treat
ed
as
fa
ulty
. A
ss
um
e a south direct
io
n r
ou
te
r
is fa
ulty
o
r u
navai
la
ble
for
r
outi
ng pr
oc
ess,
on
ly
if t
he
south
li
nk
or
s
outh
i
nput
pin
of
t
he
curre
nt
node
or
t
he
north
i
nput
pin
of
the
so
ut
h
neig
hbor
rou
te
r
is
fa
ulty
.
This
conditi
on is sta
te
d
in e
quat
io
n (
1)
us
i
ng f
a
ult detect
io
n
sta
ge
g
e
ner
at
e
d
si
gnal
.
S=
Link
s
or
In_
Port
S
Current_rou
te
r
or
In_P
or
t
N
S_Router
(1)
In
e
quat
ion
(1)
al
l
the
te
r
m
s
a
re
one
bit
sta
tu
s,
if
any
te
rm
i
s
equal
to
‘
1’
t
hen
res
pecti
ve
com
po
ne
nt
is
f
aulty
,
el
se
it
is
healt
hy
.
I
n
a
ny
r
oute
r
if
the
in
pu
t
pi
n
is
fa
ulty
then
it
can
be
m
od
el
ed
by
ass
um
i
ng
it
s
li
nk
is
fa
ulty
.
e
qu
at
io
n (
2)
is
basical
ly
u
se
d for all
the
f
our direct
io
ns
of th
e r
ou
te
r
.
n=
Link
n
or
In_
Port
n
Cur_Router
or
In_Port
(1
-
n)
n_r
outer
(2)
In
(2)
‘n’
ca
n
be
a
E,
W,
N
or
S
i.e
East
,
Wes
t,
North
or
S
outh
directi
ons
,
r
especti
vely
.
Li
nk
n
shows
the
s
ta
tus
of
c
urren
t r
ou
t
er i.e b
idirect
io
nal li
nk
in
the ‘n’
di
recti
on.
I
n_P
or
t
n
Cur_Router
giv
es the stat
us of
the input p
i
ns
of
pr
i
or
it
y
enc
oder
in
the
‘
n’
di
recti
on
of
the
current
router
and
I
n_Port
(1
-
n)
n_
router
giv
es
th
e
sta
tus
of
neighb
or
router
in
put
po
rt
to
w
hich
the
opposit
e
direc
ti
on
of
n
a
nd
pl
aced
in
the
n
directi
on
of
th
e
current
r
ou
te
r.
(
1
-
n)
ind
ic
at
es
the
opposit
e
directi
on
of
n,
w
hich
m
eans
N,
S
,
E
and
W
f
or
S,
N,
W
an
d
E
directi
on
s
,
res
pe
ct
ively
.
If
a
ny
one
of
the
com
ponen
t
inside
the
rout
er
is
fau
lt
y,
th
en
entire
route
r
is
co
ns
ide
red
as
fau
lt
y.
O
nc
e
the
router
is
fau
lt
y
it
is
no
t
avail
a
ble
to
do
it
s
task
(i.e
r
ou
ti
ng
the
pack
et
s
f
rom
inp
ut
to
it
s
c
orres
pondin
g
outp
ut
port).
On
e
bit
inf
or
m
at
ion
of
LFR
is
us
ed
to
ind
ic
at
e
the
fau
lt
iness
of
the
n
od
e
w
hic
h
is
la
beled
as
Node.
e
qu
at
io
n (
3)
de
te
rm
ines the st
at
us
of
t
he Nod
e/
Rou
te
r.
No
de
=
Prio
rit
y_en
c
oder
or
R
andom
_arb
it
er
o
r Cr
os
s
bar
_Swi
tc
h
(3)
In
e
quat
io
n
(3)
if
any
one
of
the
a
bove
te
rm
is
fau
lt
y,
then
e
ntire
node
is
c
on
si
der
e
d
as
fa
ulty
no
de.
I
n
m
ul
ti
ple
cor
e
ne
tworks
on
c
hip
,
th
e
proc
essi
ng
el
em
ents
ar
e
co
nn
ect
e
d
to
netw
ork
via
the
ne
tw
ork
int
erf
ace.
If
th
e
PE
is
not
wor
king
the
n
platfo
rm
le
vel
will
autom
at
icall
y
re
m
ap
that
pac
ket
into
s
om
e
oth
er
c
or
e
on
t
he
netw
ork
acc
ordin
g
to
healt
hy
sta
tus
inf
orm
at
ion
.
Eq
uati
on
(
4)
say
s
t
ha
t
if
PE
is
‘
1’
then
it
is
consi
der
e
d
a
s
fau
lt
y
or
it
s
ne
twork
inter
fa
ce
or
the
local
li
nk
is
con
ne
ct
ed
betwee
n
the
router
a
nd
PE
is
fau
lt
y
then
PE
beco
m
es unava
il
able.
PE =
PE
Local
or N
I or
Link
Local
(4)
Fault
inf
or
m
at
i
on
s
are
determ
ined
us
in
g
e
quat
ion
(
1)
t
o
(
4)
,
w
hich
is
us
ef
ul
f
or
routin
g
process
i
n
ord
er
t
o
i
m
pr
ove
the
pe
rfor
m
ance
by
av
oid
i
ng
dea
d
loc
k
a
nd
li
ve
lock
sit
uation.
This
fau
lt
i
nfo
rm
ation
is
cl
as
sifie
d
and
tra
nsfer
red
to
the
top
le
ve
l
of
the
syst
e
m
to
m
ap
the
pack
et
into
the
he
al
thy
no
de
w
hich
in
tu
rn
im
pro
ves
the
fa
ult
tolera
nt
capa
bili
ty
a
nd
t
he
c
os
t
of
routin
g
al
gorit
hm
[3
4].
S
uc
h
local
fa
ult
inf
or
m
at
ion
is
sto
red
i
n
the
LFR.
The
local
fa
ult
re
gis
te
r
is
8
bit
in
s
iz
e.
In
this
6
bi
ts
are
us
e
d
for
ind
ic
at
in
g
the
fau
lt
y
sta
tus
a
nd
2
bits
for
fu
t
ur
e
enh
a
ncem
ent.
Fu
rt
her
i
n
thes
e
six
bits,
t
he
LFR
us
es
f
our
bits
to
up
date
the
sta
tus
of
f
our
i
nput
pin
s
of
the
rou
te
r,
w
hich
help
s
neig
hbor
node
s
to
update
th
ei
r
local
fau
lt
r
egiste
rs
acco
r
di
ng
to
e
qu
at
io
n
(2
)
.
The
rem
ai
nin
g
2
bits
are
us
e
d
to
update
t
he
sta
tus
of
Node
a
nd
PE
.
As
su
m
e
that
center
no
de
is
the
curren
t
node
a
nd it
is
ha
ving
four n
ei
ghbo
rin
g nodes.
The
cu
rr
e
nt
node
updates
it
s
own
c
om
po
nen
t
fa
ult
inf
or
m
at
ion
in
L
FR
and
al
s
o
updates
th
e
neig
hbori
ng
fa
ult
inform
at
ion
with
the
help
of
RFR
.
T
he
RFR
is
8
bit
in
siz
e,
in
this
fo
ur
bit
is
us
ed
to
update
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
201
8
:
4164
-
4174
4168
the
neig
hbor
node
fa
ult
inf
or
m
at
ion
and
rem
ai
nin
g
four
bits
are
us
e
d
f
or
fu
t
ur
e
e
nh
a
nc
e
m
ent.
Eq
uation
s
(
1)
t
o
(4)
will
up
da
te
the
L
FR
of
al
l
the
no
des.
This
L
FR
will
help
to
update
the
RFR
of
the
entire
neig
hbori
ng
node
with
new
fa
ulty
inf
or
m
at
ion
.
Acc
ordin
gly
the
ce
nter
node
will
ha
ve
the
fa
ulty
inform
a
ti
on
of
N,
E,
W
an
d
S
sides
of
t
he
nodes.
If
a
ny
on
e
bit
of
L
F
R
of
the
north
side
node
is
e
qual
to
one,
the
n
the
current
no
de
R
FR
will
update
s
the
‘NN’
bit
to
on
e
an
d
the
rem
ai
nin
g
bits
will
be
e
qu
al
to
zer
o,
wh
ic
h
say
s
that
the
north
s
ide
no
de
is
un
healt
hy
an
d
t
he
pac
kets
sho
ul
d
not
be
sent
t
ow
a
r
ds
north
node
if
the
desti
nation
is t
op r
ig
ht
node
.
The
pro
pose
d
hierar
c
hical
ag
ent
structu
re
is
as
sh
own
in
F
ig
ure
2.
Eac
h
and
e
ver
y
cel
l,
cl
us
te
r
age
nt
gets
update
d
with
ne
w
fa
ul
t
info
rm
at
ion
of
it
s
own
cl
us
te
r
cel
l
and
th
e
neig
hbori
ng
cl
us
te
r
cel
ls
bid
irect
io
naly
[6
]
.
S
uc
h
fau
lt
inf
or
m
at
ion
is
sent
to
t
he
to
p
le
vel
of
the
sys
tem
,
wh
ic
h
hel
p
to
m
ap
the
pa
cket
to
healt
hy
no
de
by
sel
ect
ing
best
path
.
Cl
us
te
r
Se
par
at
i
on
M
odule
(C
SM)
hel
ps
the
pack
et
to
rea
ch
it
s
resp
ec
ti
ve
cl
ust
er
agen
t
by
con
si
der
i
ng
cl
ust
er
sel
ect
ion
bits
on
t
he
pa
cket.
Th
en
the
cl
us
te
r
will
r
ou
te
the
pack
et
i
nto
t
he
res
pecti
ve
cel
l
agen
t
acc
ordi
ng
to
t
he
r
outi
ng
inf
orm
ation
in
the
pac
ket.
Finall
y
age
nt
will
decide
w
hethe
r
the
pac
ket
ha
s
to
pass
or
s
t
op
i
nto
t
he
processin
g
el
em
ent
by
pro
vid
i
ng
the
sec
uri
ty
in
th
e
agen
t
.
C
l
u
s
t
e
r
A
g
e
n
t
C
e
l
l
A
g
e
n
t
C
e
l
l
A
g
e
n
t
C
e
l
l
A
g
e
n
t
C
l
u
s
t
e
r
A
g
e
n
t
C
e
l
l
A
g
e
n
t
C
e
l
l
A
g
e
n
t
C
e
l
l
A
g
e
n
t
C
l
u
s
t
e
r
S
e
p
a
r
a
t
i
o
n
M
o
d
u
l
e
T
w
o
c
l
u
s
t
e
r
a
g
e
n
t
s
P
l
a
t
f
o
r
m
L
e
v
e
l
A
p
p
l
i
c
a
t
i
o
n
L
e
v
e
l
Figure
2
.
Hiera
rch
ic
al
a
gen
ts
in
tw
o neig
hbor
clusters
A
4X4
a
gen
t b
ased N
oC
a
s
s
how
n
in
F
ig
ur
e
3
in
cl
udes p
r
oc
essing
el
em
en
t,
netw
ork
i
nterf
ace,
r
ou
t
e
r
and
a
gen
ts
(t
hese
age
nt
can
be
et
her
a
cel
l
agen
t
or
a
cl
us
te
r
agent
).
All
the
agen
ts
are
co
nnect
e
d
bid
irect
io
nally
and
one
bit
inf
or
m
at
ion
is
exch
a
nged
between
the
a
gent
s
to
up
date
t
he
RFR
.
Lat
er
these
agen
ts
a
re
co
nnect
ed
to
t
he
No
C
router
network
.
Pac
kets
from
the
app
li
cat
ion
le
vel
en
te
r
into
the
rou
te
r
via
the
agen
ts
in
order
to
c
heck
the
secur
it
y
aspects
w
hich
will
be
exp
la
ined
in
furthe
r
sect
ion
.
T
he
pro
pose
d
agen
t
ba
sed
m
on
it
ori
ng
syst
em
us
es
two
ty
pes
of
com
m
u
nicat
ion
:
nam
el
y
peer
to
peer
com
m
un
ic
at
io
n
(use
d
betwee
n
t
he
a
ge
nts)
an
d
base
li
ne
data
net
work
com
m
un
ic
at
ion
(
for
c
ontrol
li
ng
a
nd
routin
g
t
he
pack
et
s
i
n
t
he
netw
ork).
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
Reco
nfigu
r
abl
e H
ig
h Perf
orma
nce
Sec
ur
e
d N
oC
Desig
n U
sing Hier
ar
c
hical
…
(
Ke
ndag
anna
Swamy
S)
4169
Figure
3
.
A
4X
4
a
gen
t
base
d No
C
The
cl
ust
er
a
ge
nt
will
accum
ulate
the
crit
ic
al
fau
lt
inf
orm
at
ion
(
PE
or
w
ho
le
node
fail
ure)
occ
urred
inside
the
cl
ust
er
and
will
send
that
in
form
at
ion
to
hi
gh
e
r
le
vel.
Then
t
he
cl
us
te
r
will
receive
the
co
m
m
and
from
the
hig
he
r
le
vel
for
reconfi
gurati
on
or
rem
app
ing
of
the
pac
ket
or
ta
sk
m
igrati
on.
The
cl
us
te
r
se
pa
rati
on
m
od
ule
will
play
an
i
m
po
rt
ant
ro
le
to
se
par
at
e
the
rec
ei
ved
pack
et
a
nd
se
nd
it
to
the
res
pecti
ve
cl
us
te
r
accor
ding
to
t
he
n
th
bit
of
t
he
pack
et
.
If
t
he
n
th
bit
is
zero
th
en
the
rec
ei
ve
d
pack
et
belo
ngs
to
cl
us
te
r
_1
e
lse
it
belo
ngs
to
cl
ust
er_2.
If
t
he
nu
m
b
er
of
cl
ust
er
s
increase
s
the
n
the num
ber
of
cl
us
te
r
sel
ect
ion
co
ntr
ol
bits m
us
t
al
so
be
inc
reas
ed
in
or
der
t
o
s
egr
e
gates th
e
pa
cket to
desi
re
d
cl
ust
er.
The
X
Y
base
d
fau
lt
tolerant
r
ou
ti
ng
al
gorith
m
[3
3
]
,
[
34]
is
inco
rpor
at
e
d
in
the
propose
d
hierar
c
hica
l
ag
ent
base
d
m
anag
em
ent
m
e
thod.
T
his
no
de
is
dev
el
ope
d
based
on
ab
ove
disc
us
se
d
m
at
he
m
at
ic
a
l
equ
at
io
n.
This
r
ou
ti
ng
al
gorithm
is
low
cost,
ada
ptive
and
c
ongestio
n
awa
re
w
hich
is
su
it
able
fo
r
No
C
bas
ed
m
ulti
ple
cor
e syst
e
m
on
chip
.
F
or
exa
m
ple
con
sider
a
3x3
n
et
work in
w
hich
the
ce
nter
node
i
nclu
des
the
cl
us
te
r
ag
ent
.
In
s
uch
netw
or
k
to
p
le
ft n
ode
is source no
de
and
bott
om
r
igh
t i
s the d
est
in
at
ion
no
de.
T
he
so
urce n
ode
sh
oul
d
be
awa
re
of
th
e
sta
tus
of
E,
S,
ES
an
d
SE
l
abeled
li
nks
surr
ounded
by
de
sti
nation
no
de
[35]
,
[
36
]
.
T
he
fau
lt
y
sta
tuses
of
t
he
se
li
nk
s
are
not
up
dated
i
n
the
nei
ghbor
node
of
the
source
node
.
T
he
n
cl
us
te
r
a
ge
nt
will
pro
vid
e
this
in
form
ation
to
t
he
r
ou
ti
ng
al
gorithm
.
W
it
h
the
hel
p
of
this
inform
at
ion
,
the
r
ou
ti
ng
al
gorithm
will
colle
ct
ively
ga
ther
al
l
the
fau
lt
y
an
d
c
ongestio
n
in
form
at
ion
a
nd
r
eac
h
the
desi
red
l
oc
at
ion
in
t
he
s
hortest
path. Al
gorith
m
(
1)
s
hows
t
he
m
anag
em
ent
al
gorithm
u
sed
b
y t
he
H
ie
rar
c
hical
Agen
ts
.
Ag
e
nt
M
ana
ge
ment Alg
orit
hm
Inpu
t:
Faulty
, c
ongestio
n
a
nd
secu
rity
inform
at
ion
f
r
om
cell
ag
ents
(NA
) a
nd n
ei
ghbor
c
luster
a
gen
ts
(C
A)
for
each
age
nt
do
Wait
unti
l a ne
w
c
onge
sti
on
or f
a
ult i
nfor
m
at
ion
is re
cei
ved
;
I
f
(
a
node
f
ai
lure
or
N
I fai
ls
or
PE
fail
s
or
no c
ontrol
p
ac
ket is re
cei
ved f
ro
m
a NA
w
it
hin
the ti
m
e)
th
en
i
nfor
m
the top le
vel a
nd it
s ass
ociat
ed node a
ge
nts;
receive
the
pack
et
r
em
app
i
ng or tas
k real
locat
ion i
nf
or
m
at
ion
;
s
eg
regat
e the
fail
ed
P
E or
Node;
el
se if
(
ne
w fault i
nfo
r
m
at
ion
r
ecei
ve
d from
a CA)
t
hen
in
f
or
m
the n
e
w fa
ult an
d
c
ongest
ion
i
nfor
m
at
ion
to
n
ei
ghbori
ng clu
ste
r
a
gen
t
s and
as
so
ci
at
ed n
od
e
agen
ts;
el
se if
(
ne
w f
ault i
nfo
r
m
at
ion
f
r
om
the NA)
th
e
n
in
f
or
m
the n
e
w
c
ongestio
n
a
nd
fau
lt
in
form
at
i
on to neig
hbori
ng no
de
a
gen
ts
w
it
hin
th
e cluste
r
a
ge
nt;
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
201
8
:
4164
-
4174
4170
en
d if
if
(
de
sti
nation
a
ddre
ss of
XY == c
urre
nt add
ress
of XY
)
t
hen
c
he
ck
the
sec
ur
it
y
to d
eci
de wh
et
her pac
ket
has t
o
pa
ss int
o
the
d
est
inati
on
PE
or
no
t;
if
(sec
ur
it
y chec
k
=
=
0)
th
e
n
32
bit pac
ket
data
will
r
each
the
destinat
io
n Node;
el
se
pa
cket will
be disca
rd
;
en
d if
3.
RESU
LT
A
N
ALYSIS
T
o
a
na
l
y
z
e
t
he
im
po
r
t
a
nc
e
of
t
he
pr
op
os
e
d
hi
e
r
a
r
c
hi
c
a
l
s
ec
ur
e
d
a
ge
nt
ba
s
e
d
m
on
i
t
or
i
ng
s
y
s
te
m
on
ne
t
w
or
k
pe
r
f
or
m
a
nc
e
.
T
he
4
x
4
a
ge
nt
ba
s
e
d
N
oC
de
s
i
gn
us
i
ng
H
D
L
c
od
e
a
nd
s
i
m
ul
at
e
d
us
i
ng
X
i
l
i
nx
I
S
E
14
.
2
t
oo
l
wi
t
h M
od
e
l
S
im
6.
3f
re
s
pe
c
t
i
ve
ly
.
It
i
s
s
yn
t
he
s
i
z
e
d a
nd
i
m
pl
em
e
nt
e
d o
n
ve
r
t
e
x
5 F
P
G
A
(
X
C
5v
F
X
7
0
T
)
ki
t
.
T
he
pe
r
f
or
m
a
nc
e
of
p
r
o
po
s
e
d
m
e
t
ho
d
of
s
e
c
ur
e
d
a
ge
nt
ba
s
e
d
m
on
i
t
or
i
ng
s
y
s
t
em
i
s
a
na
l
y
z
e
d
a
nd
c
om
pa
r
e
d
w
i
t
h
t
he
e
xi
s
t
in
g
m
e
t
ho
ds
.
I
n
t
he
pr
op
os
e
d
de
s
i
gn
e
a
c
h
c
l
us
t
e
r
i
s
a
4X
4
s
ub
ne
t
w
or
k;
i
n
t
hi
s
t
he
c
e
nt
e
r
n
od
e
i
s
t
r
e
a
t
e
d
a
s
c
l
us
t
e
r
a
ge
nt
.
I
t
i
s
a
s
s
um
e
d
t
ha
t
6.
2
5%
of
f
a
ul
t
y
no
de
(
o
ne
n
od
e
i
s
f
a
ul
t
y
ou
t
of
1
6
n
od
e
s
)
a
n
d
20
.
83
%
of
f
a
u
l
t
y
l
i
nk
(
f
i
ve
f
a
ul
t
y
li
nk
s
o
u
t
of
2
4
l
i
nk
s
)
l
e
a
ds
t
o
27
%
of
s
y
s
t
em
f
au
l
t
i
s
a
s
de
pi
ct
e
d
i
n
F
ig
ur
e
4
(
a
)
.
Figure
4
(
a
).
A
fau
lt
y 4
x4
NoC
F
i
gu
r
e
4
(
b
).
A
ve
r
a
ge
pa
c
ke
t
l
a
t
e
nc
y
a
na
l
y
si
s
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
Reco
nfigu
r
abl
e H
ig
h Perf
orma
nce
Sec
ur
e
d N
oC
Desig
n U
sing Hier
ar
c
hical
…
(
Ke
ndag
anna
Swamy
S)
4171
S
im
i
l
a
r
a
s
s
um
pt
i
on
i
s
m
a
de
f
or
bo
t
h
w
i
t
h
a
nd
w
i
t
ho
ut
a
ge
nt
ba
s
e
d
ne
t
w
or
k
a
n
d
t
he
pe
r
f
or
m
a
nc
e
a
na
l
y
s
i
s i
s
a
s
sho
w
n
i
n
F
ig
ure
4
(
a
)
.
F
r
om
t
he
gr
a
ph
i
t
i
s
ob
s
e
r
ve
d t
ha
t
,
a
wi
t
ho
ut
a
ge
nt
ba
s
e
d X
Y
m
et
ho
d d
oe
s
no
t
ha
ve
a
ny
m
e
a
ns
t
o
r
e
l
ia
bl
y
s
e
nd
a
l
l
t
he
pa
c
ke
t
s
t
o
t
he
i
r
de
s
t
i
na
t
i
on
i
n
t
he
f
a
ul
t
y
s
it
ua
t
i
on
s
,
t
he
r
e
m
a
y
be
c
ha
nc
e
of
pa
c
ke
t
s
t
uc
k
i
n
t
he
f
a
ul
t
y
no
de
t
he
n
pa
c
ke
t
ha
s
t
o
be
r
e
s
e
nt
f
r
om
t
he
t
op
le
ve
l
,
w
hi
c
h
l
e
a
ds
t
o
pe
r
f
or
m
a
nc
e
de
gr
a
da
t
i
on
[
33
]
.
I
n
t
he
pr
o
po
s
e
d
de
s
i
gn
w
i
t
h
t
he
pr
i
or
kn
ow
l
e
dg
e
of
a
l
l
t
he
f
a
ul
t
y
l
i
nks
a
n
d
no
de
s
t
he
pa
c
k
e
t
w
i
l
l
r
e
a
c
h
t
he
he
a
l
t
hy
no
de
w
i
t
h
a
r
e
l
i
a
bl
e
t
im
e
.
T
he
pr
o
po
s
e
d
s
e
c
ur
e
d
hi
e
r
a
r
c
hi
c
a
l
a
g
e
nt
-
ba
s
e
d
s
y
s
t
em
l
ea
ds
t
o
hi
gh
e
r
pe
r
f
or
m
a
nc
e
an
d
s
a
t
ur
a
t
i
on
po
i
nt
s
a
s
c
om
pa
r
e
d
t
o
m
e
t
ho
d
i
nt
r
od
uc
e
d
i
n
[
10
]
.
A
s
de
pi
c
t
e
d
i
n
F
i
g
ure
4(
a
)
s
o
ur
c
e
n
od
e
S
s
e
nd
s
t
he
pa
c
ke
t
t
o
de
s
t
i
na
t
i
on
no
d
e
D
.
I
n
o
r
de
r
t
o
r
e
a
c
h
t
he
pa
c
ke
t
f
r
om
s
ou
r
c
e
no
de
t
o
de
s
t
in
a
t
i
on
no
de
t
h
e
r
e
a
r
e
t
w
o
pa
th
s
P
1
a
nd
P
2,
a
m
on
g
t
he
s
e
P
1
i
s
t
h
e
m
i
ni
m
al
pa
t
h
w
he
n
c
om
pa
r
e
d
t
o
P
2.
H
ow
e
ve
r
p
r
o
po
s
e
d
hi
e
r
a
r
c
hi
c
a
l
s
e
c
ur
e
d
a
ge
nt
w
i
l
l
s
e
l
ec
t
t
he
m
i
ni
m
a
l
pa
t
h
P
1
t
o
r
o
ut
e
t
he
pa
c
ke
t
us
i
ng
X
Y
r
o
ut
i
ng
a
l
go
r
i
t
hm
.
F
i
g
ur
e
5
w
a
v
e
f
or
m
s
ho
w
s
t
he
n
od
e
t
o
no
de
pa
c
ke
t
t
r
a
ns
f
e
r
pa
t
h
be
t
w
e
e
n
no
de
2
t
o
no
de
16
(
i
.
e
.
N
2
-
N
6
-
N
10
-
N
15
-
N
1
6)
a
s
hi
gh
l
i
gh
t
e
d
on
t
he
w
a
ve
f
or
m
.
I
t
i
s
s
y
nt
he
s
i
z
ed
a
n
d
im
pl
em
en
t
e
d
o
n
ve
r
t
e
x
5
F
P
G
A
(
X
C
5
vF
X
70
T
)
ki
t
.
Fig
ure
5
.
4X
4 sec
ur
e
d
a
ge
nt
base
d No
C
De
sign i
m
ple
m
entat
ion
w
a
ve
for
m
F
i
g
ur
e
6
(
a
)
a
n
d
F
i
g
ur
e
6
(
b
)
t
he
p
r
o
po
s
e
d
de
s
i
gn
t
hr
ou
gh
pu
t
i
s
c
om
pa
r
e
d
w
i
t
h
t
he
e
xi
s
t
i
ng
f
u
nc
t
i
on
a
l
di
a
gn
os
i
s
m
e
t
hod
w
i
t
h
no
r
m
al
a
nd
he
a
vy
l
oa
d
c
o
nd
i
t
i
on
u
n
de
r
t
he
un
i
f
o
r
m
t
r
a
f
f
i
c
.
A
c
c
or
di
ng
t
o
t
he
g
r
a
ph
,
t
he
pr
o
po
s
e
d
m
e
t
hod
ha
s
hi
g
h
t
h
r
ou
gh
pu
t
a
s
c
o
m
pa
r
e
d
t
o
t
he
m
e
t
ho
d
i
nt
r
o
d
uc
e
d
i
n
[
3
0]
.
T
o
a
na
l
y
z
e
t
he
a
r
e
a
ov
e
r
he
a
d
o
f
t
h
e
pr
o
po
s
e
d
de
s
i
gn
,
t
he
im
pl
em
e
nt
e
d
D
y
X
Y
[
35
]
i
s
t
he
ba
s
i
c
a
da
pt
i
ve
r
ou
t
i
ng
m
e
t
ho
d,
a
n
a
da
pt
i
ve
f
a
ul
t
t
ol
e
r
a
nt
t
e
c
hn
iq
ue
R
A
F
T
[
33
]
,
e
xi
s
ti
ng
A
g
e
nt
ba
s
e
d
f
a
ul
t
t
ol
e
r
a
nt
r
ou
t
i
ng
a
l
go
r
i
t
hm
[10
]
i
s
c
om
pa
r
e
d
w
i
t
h
t
he
pr
op
os
e
d
s
e
c
ur
e
d
a
ge
nt
ba
s
e
d
f
a
ul
t
t
ol
e
r
a
nt
r
ou
t
i
ng
a
l
go
r
i
t
hm
us
i
ng
H
D
L
c
o
de
w
i
t
h
a
20
1.
7
4
M
H
z
c
l
oc
k
s
pe
e
d.
F
i
g
ur
e
6
(
a
).
T
h
r
ou
gh
pu
t
w
i
t
h
no
r
m
a
l
l
oa
d
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
201
8
:
4164
-
4174
4172
F
i
g
ur
e
6
(
b
).
T
h
r
ou
gh
pu
t
w
i
t
h
he
a
vy
l
oa
d
T
a
bl
e
1
s
ho
w
s
t
he
a
r
e
a
ut
i
li
z
at
i
on
of
pr
o
po
s
e
d
f
i
ve
po
r
t
r
o
ut
e
r
c
om
pa
r
e
d
t
o
ot
he
r
e
xi
s
t
i
ng
m
e
t
ho
ds
.
I
n
a
dd
i
t
i
on
t
o
t
hi
s
,
t
he
t
a
bl
e
gi
ve
s
t
he
pr
o
po
s
e
d d
e
s
i
gn
a
r
e
a
o
v
e
r
he
a
d
a
s
c
om
pa
r
e
d
t
o o
t
he
r
m
e
t
ho
ds
.
B
a
s
e
d
o
n
t
he
ha
r
d
w
a
r
e
a
na
l
y
s
i
s
t
a
bl
e
t
he
pr
o
po
s
e
d
de
s
i
g
n
a
r
e
a
ut
i
l
i
z
at
io
n
a
n
d
ha
r
dw
a
r
e
ov
e
r
he
a
d
i
s
1.
4
%
i
m
pr
ov
e
d
a
s
c
om
pa
r
e
d
t
o
ex
i
s
t
i
ng
a
ge
nt
b
a
s
e
d
f
a
ul
t
t
ol
e
ra
nt
m
et
ho
d
[
1
0
]
.
I
t
i
s
w
or
t
h
t
o
m
e
nt
i
on
D
y
X
Y
m
et
ho
d
d
o
e
s
n’
t
ha
ve
r
e
l
i
a
bi
l
i
ty
t
o
t
r
a
ns
f
e
r
a
l
l
t
he
pa
c
ke
t
s
s
uc
c
e
s
s
f
ul
l
y
t
o
t
he
i
r
de
s
t
i
na
t
i
on
u
nd
e
r
t
he
f
a
ul
t
y
s
i
t
ua
ti
on
T
a
bl
e
1
.
D
e
vi
c
e
U
t
i
l
iz
a
ti
on
S
um
m
a
r
y
R
o
u
t
i
n
g
M
et
h
o
d
A
r
ea
u
t
il
i
za
t
i
o
n
(
Ga
t
e
c
o
u
n
t
)
fo
r
5
P
o
r
t
A
r
ea
O
v
e
r
h
e
a
d
C
o
m
p
ar
i
s
o
n
(
%
)
D
y
X
Y
[3
5
]
36350
1
2
.
7
R
A
F
T
[3
3
]
39355
4
.
1
A
g
e
n
t
b
a
se
d
R
o
u
t
in
g
[1
0
]
41574
NA
P
r
o
p
o
s
e
d
Sy
st
e
m
40922
1
.
4
4.
CONCL
US
I
O
N
In
t
hi
s
pa
pe
r
,
a
hi
e
r
a
r
c
hi
c
a
l
s
e
c
ur
e
d
a
ge
nt
ba
s
e
d
m
on
i
t
or
in
g
s
y
s
t
em
i
s
pr
op
os
e
d
f
or
f
a
ul
t
t
ol
e
r
a
nt
m
ul
ti
c
or
e
N
o
C
ba
s
e
d
s
y
s
t
e
m
on
c
hi
p.
T
he
hi
e
r
a
r
c
hi
c
a
l
l
y
di
s
t
r
i
bu
t
e
d
a
ge
nt
w
i
l
l
c
ol
l
e
c
t
,
m
a
na
ge
a
nd
di
s
t
r
i
bu
t
e
t
he
f
a
ul
t
a
nd
co
n
ge
s
t
i
on
i
nf
or
m
a
ti
on
of
t
he
ne
t
w
or
k
t
o
hi
g
h
e
r
l
e
ve
l
of
t
he
s
y
s
t
em
.
T
hi
s
f
au
l
t
i
nf
or
m
at
i
on
he
l
ps
a
pp
l
i
c
a
t
i
on
l
e
ve
l
t
o
r
ou
t
e
t
he
pa
c
ke
t
t
o
he
a
l
t
hy
no
de
,
w
hi
c
h
w
i
l
l
im
pr
ov
e
t
he
pe
r
f
or
m
a
nc
e
of
t
he
ne
t
w
or
k
by
a
vo
i
di
ng
t
he
p
a
c
ke
t
l
a
te
nc
y
ag
a
i
ns
t
f
a
ul
t
y
no
de
a
n
d
l
i
nk
s
.
I
n
a
d
di
t
i
on
t
o
t
hi
s
t
he
a
ge
nt
w
i
l
l
pr
ov
i
de
s
e
c
ur
i
t
y
t
o
t
he
P
E
i
n
or
de
r
t
o
bl
oc
k
t
he
u
nw
a
nt
e
d
a
n
d
u
nr
e
l
a
t
e
d
pa
c
ke
t
e
nt
e
r
i
ng
i
nt
o
t
he
P
E
w
hi
c
h
w
i
l
l
a
vo
i
d
t
he
l
i
ve
l
oc
k
s
i
t
ua
ti
on
o
f
t
he
hi
gh
p
r
i
or
i
t
y
pa
c
ke
t
w
hi
c
h
i
s
r
e
l
a
t
e
d
t
o
t
he
de
di
c
a
t
e
d
n
od
e
.
A
c
c
or
di
n
g
t
o
t
he
s
i
m
ul
at
i
on
a
n
d
s
y
nt
he
s
i
s
r
e
s
ul
t
,
t
he
pr
op
os
e
d
de
s
i
gn
w
i
l
l
e
nh
a
nc
e
t
he
n
e
t
w
or
k
pe
r
f
or
m
a
nc
e
w
i
t
h
an
im
pr
ov
e
d
h
a
r
dw
a
r
e
ov
e
r
he
a
d
by
us
i
ng
t
he
m
od
i
f
i
ed
r
ou
t
e
r
de
s
i
g
n
.
REFERE
NCE
S
[1]
Inte
rna
ti
ona
l Te
chnol
o
g
y
Ro
ad
m
ap
for
Sem
ic
o
nduct
ors,
2015
.
Avail
ab
le
:
htt
p
:/
/
ww
w.i
trs.
net
/
.
[2]
W
.
Dally
and
B.
Towle
s,
“
Route
pac
ket
s,
no
t
wire
s:
On
-
chi
p
int
e
rco
nnection
netw
orks,”
Proc.
Des.
Aut
om.
Conf
.
,
pp.
684
–
689
,
20
01
.
[3]
A.
Jantsch
and
H.
Te
nhun
en,
“
Network
on
Chi
p,
”
Klu
w
er
Ac
ad
emic
Publishers,
2003.
[4]
T.
C
.
Xu,
et
al
.
,
“
An
opti
m
iz
ed
n
et
work
-
on
ch
ip
desig
n
for
d
at
a
par
allel
FF
T,
”
P
roce
dia
Engi
ne
e
ring
,
vol
.
30
,
pp
.
313
–
318,
2012
[5]
Y.
Y.
Ch
en,
et
a
l.
,
“
Path
-
Diver
si
t
y
-
Aw
ar
e
Faul
t
-
Tol
er
ant
Rou
ti
n
g
Algori
thm
for
Network
-
on
-
Chi
p
S
y
stems
,
”
IEEE
Tr
ansacti
ons on Paralle
l
and
Dis
tribut
ed
Syste
ms
,
v
ol
/i
ss
ue:
28
(
3
)
,
2017
.
[6]
M.
Vali
n
at
a
j,
e
t
al.
,
“
Reliable
On
-
Chip
Net
work
Design
Us
ing
an
Agen
t
-
base
d
Man
a
gement
Method
,”
Proce
ed
ings
of
the
19th
Inte
rnati
onal
Confe
renc
e
Mix
ed
Design
of
Inte
grated
Circ
uit
s
and
Syste
ms
(
MIX
DES
)
,
pp
.
447
-
451,
2012
.
[7]
L.
Guang,
et
al
.
,
“
Hier
arc
h
i
ca
l
age
nt
m
onit
orin
g
design
appr
oac
h
towar
ds
self
-
awa
re
par
a
llel
sy
stems
-
on
-
chi
p
,
”
ACM
Tr
ans.
on
Embe
dded
Com
puti
ng
S
yste
ms
,
vol
/i
ss
ue:
9
(
3
)
,
2
010
.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
Reco
nfigu
r
abl
e H
ig
h Perf
orma
nce
Sec
ur
e
d N
oC
Desig
n U
sing Hier
ar
c
hical
…
(
Ke
ndag
anna
Swamy
S)
4173
[8]
L.
Guang,
e
t
al.
,
“
Hier
arc
hical
p
ower
m
onit
oring
on
NoC
-
a
ca
s
e
stud
y
for
hie
r
a
rch
ical
ag
ent
m
onit
oring
de
sign
appr
oac
h
,
”
Proc
.
28th
NORCHIP
Conf.
,
2010
.
[9]
P.
Rant
a
la,
e
t
al
.
,
“
Novel
ag
ent
-
b
ase
d
m
ana
geme
nt
for
fau
lt
-
to
le
r
anc
e
in
ne
twork
-
on
-
chi
p,
”
Proc
.
10th
Euromicro
Conf.
on
Digit
a
l Syste
m Design
(
DSD)
,
pp.
551
–
555,
2007
.
[10]
M.
Vali
nataj,
et
al.
,
“
Enha
nc
ed
Fault
-
Tol
er
ant
Network
-
on
-
Chi
p
Archi
te
c
ture
Us
ing
Hier
arc
hi
ca
l
Agents,”
16th
Inte
rnational
Sy
mpos
ium
on
Design
and
Diagnos
ti
cs
o
f
E
lectroni
c
Circu
it
s
&
Syst
ems
,
pp
.
141
-
14
6,
2013
.
[11]
A.
W
.
Yin
,
et
al
.
,
“
Hier
ar
chica
l
age
nt
m
onit
ori
ng
NoCs
:
a
design
m
et
hodolog
y
with
sca
la
bi
li
t
y
and
var
ia
bi
li
t
y
,
”
Proc.
26
th
NOR
CHIP
Conf.
,
pp
.
202
–
207,
2008
.
[12]
W
.
C.
San
tos,
et
a
l.
,
“
High
-
Im
peda
nc
e
Fau
lt
I
dent
ific
a
ti
on
on
Distribut
ion
N
et
works
,
”
IEEE
Tr
ansacti
ons
o
n
Powe
r Del
iv
er
y
,
vol/
issue: 32(1),
pp.
23
-
32
,
2017
.
[13]
M.
Pignat
i
,
e
t
al.
,
“
F
aul
t
Det
e
ct
ion
and
Faul
t
ed
Li
n
e
Ide
n
ti
fi
ca
t
ion
in
Ac
ti
v
e
Distribution
Networks
Us
ing
S
y
nchr
ophasors
-
Based
Real
-
Ti
m
e
State
Esti
m
a
tion,
”
I
EE
E
Tr
an
sacti
ons
on
Pow
er
Deli
v
ery
,
vol
/
issue:
32
(
1
),
pp
.
381
-
392,
2017
.
[14]
A.
Nikande
r
a
nd
P.
Järve
nt
a
usta,
“
Ide
nt
ifica
ti
o
n
of
High
-
I
m
peda
nce
Ea
rt
h
Fault
s
in
N
eut
ra
l
Isola
te
d
or
Com
pensa
te
d
M
V Net
works
,
”
I
E
EE
Tr
ansacti
ons
on
Pow
er
Del
ivery
,
vo
l
/i
ss
ue:
32
(
3
)
,
pp
.
1187
-
1
195,
2017
.
[15]
D.
C
.
R
ai
t
eri
an
d
L.
Port
ina
l
e,
“
D
y
namic
B
a
y
esi
an
Networks
for
Fault
De
te
c
ti
on
,
Ide
n
ti
fi
ca
t
ion,
an
d
Re
cove
r
y
in
Autonom
ous
Sp
ac
e
cra
f
t,
”
IEEE
Tr
ansacti
ons
on
Syste
ms
,
Man,
a
nd
Cybe
rnet
ic
s:
Syste
ms
,
vol
/
issue:
45
(
1
),
pp
.
13
-
24,
2015
.
[16]
H.
W
ang,
et
al.
,
“
Prac
tica
l
Network
-
W
ide
Packe
t
Beh
avi
o
r
Ide
n
ti
fi
ca
t
ion
b
y
AP
Cla
ss
i
fie
r,”
I
EE
E
/A
C
M
Tr
ansacti
ons on Networ
ki
ng
,
vol
/
issue:
25(5), pp.
2886
-
2899,
201
7.
[17]
S.
Ntal
ampira
s,
“
Fault
Ide
nti
fi
cation
in
Distributed
Sensor
Networks
Based
on
Univer
sal
Proba
bil
isti
c
Modeli
n
g,
”
IEE
E
Tr
ansacti
o
ns on
Neural
N
etwor
ks
and
Learning
Syst
ems
,
vol
/
issue:
26(9), pp.
1939
-
1949,
201
5.
[18]
X.
Zha
ng
and
Z
.
Zh
ang,
“
Li
nk
f
aul
t
ide
n
ti
fi
catio
n
using
dep
endent
fa
il
ure
in
w
ir
el
ess
comm
unicati
on
net
works
,
”
El
e
ct
ronics
Le
tt
e
rs
,
vol
/i
ss
ue:
52
(
2
)
,
pp
.
163
-
165
,
2016.
[19]
G.
C.
Am
ara
l
,
et
al.
,
“
Autom
at
ic
Fault
Det
ec
t
ion
in
W
DM
-
PON
W
it
h
Tuna
bl
e
Photon
Counti
ng
OTDR,”
Journa
l
of
Lightwave
Te
chnol
ogy
,
vo
l/
issue:
33
(
24
)
,
pp
.
5
025
-
5031,
2015
.
[20]
M.
Zhu,
et
al.
,
“
Optimal
fibe
r
li
nk
fau
lt
de
ci
si
on
for
opti
ca
l
2
D
codi
ng
-
m
onit
oring
sche
m
e
in
passive
opti
ca
l
net
works
,
”
IEEE
/OSA
Journal
of
Optic
al
Comm
unic
ati
ons
and
Ne
tworki
ng
,
vol/is
sue:
8
(
3
)
,
pp
.
137
-
147,
2016
.
[21]
B.
Cai
,
e
t
al.
,
“
Ba
y
esia
n
Netwo
rks
in
Fault
Diagnosis,
”
IEE
E
T
rans
act
ions
on
I
ndustrial
Informatic
s
,
vol
/i
ss
ue:
13
(
5
)
,
pp
.
2227
-
2240,
2017
.
[22]
A.
Kohler
,
et
al
.
,
“
Fault
Tolera
nt
Network
on
Chip
Sw
it
chi
ng
W
it
h
G
rac
efu
l
Perform
anc
e
De
gra
dation,
”
IE
E
E
Tr
ansacti
ons on Com
pute
r
-
Ai
ded
Design
of
Int
egrate
d
C
ircui
ts an
d
Syste
ms
,
vol/is
sue:
29
(
6
)
,
pp
.
8
83
-
896,
2010
.
[23]
E.
O
.
A.
Vall,
et
al.
,
“
Distribu
te
d
Fault
-
Tol
er
a
nce
for
Eve
n
t
Dete
c
ti
on
Us
ing
Hete
roge
n
eous
W
ire
le
ss
Senso
r
Networks,”
I
EEE
Tr
ansacti
ons
on
Mobile
Comp
uti
ng
,
vol
/
issue:
11
(
12
),
pp
.
1994
-
2007,
2012
.
[24]
J.
Yao,
et
al
.
,
“
NetSimplex:
Control
ler
Fault
Tol
er
anc
e
Arch
i
te
c
tur
e
in
Netw
orke
d
Control
S
y
stems
,
”
IEEE
Tr
ansacti
ons on Indus
trial
Infor
matic
s
,
vo
l
/i
ss
ue
:
9
(
1
)
,
pp.
346
-
3
56,
2013
.
[25]
A.
Eghbal,
e
t
al
.
,
“
Anal
y
ti
c
al Fau
lt
Tol
er
ance
As
sess
m
ent
and
Me
t
ric
s for
TSV
-
B
a
sed
3D Net
work
-
on
-
Chip,
”
IE
EE
Tr
ansacti
ons on Com
pute
rs
,
vol/i
ss
ue:
64
(12), pp. 3591
-
3604,
201
5.
[26]
P.
Ren,
et
al
.
,
“
A
Dea
dloc
k
-
Fre
e
and
Connectiv
ity
-
Guara
n
te
ed
Methodol
og
y
fo
r
Achie
v
ing
Fau
lt
-
Tolera
n
ce
in
On
-
C
hip
Networks,
”
IEEE
Tr
ansactions
on
Compute
rs
,
vol
/i
ss
ue:
65
(
2
)
,
pp
.
353
-
366
,
2016.
[27]
A.
Al
-
Shuw
ai
li,
et
al
.
,
“
Coded
Network
Function
Virtua
l
iz
a
ti
o
n:
Fault
T
ole
r
an
ce
vi
a
In
-
Netwo
rk
Coding,
”
IEEE
Wirel
ess Comm
unic
ati
ons
Letters
, v
ol
/i
ss
ue: 5(6),
pp.
644
-
647
,
20
16.
[28]
T.
F
.
Pere
ira,
e
t
al
.
,
“
Mec
hanis
m
s
to
Provide
Fault
Tol
era
n
c
e
to
a
N
et
work
-
on
-
Chip,
”
IE
EE
Latin
Am
eric
a
Tr
ansacti
ons
,
vo
l/
issue:
15
(
6
)
,
pp
.
1034
-
1042
,
20
17.
[29]
Y.
C.
W
u
and
C.
C.
Tu
an,
“
Fault
tol
er
ance
e
vent
s
orde
ring
b
y
agi
ng
l
ea
rn
i
ng
in
wire
l
ess
s
ensor
and
a
ct
ua
tor
net
works
,
”
IET
Comm
unic
ati
ons
,
vol
/i
ss
ue:
11
(
12
)
,
pp
.
1895
-
190
2,
2017
.
[30]
G
.
Schley
,
e
t
al
.
,
“
Multi
-
Lay
e
r
Diagnosis
for
F
ault
-
Tolera
nt
Netw
orks
-
on
-
Chip,
”
I
EE
E
Tr
ansacti
o
ns on
Computers
,
vol/
iss
ue:
66
(
5
)
,
2017.
[31]
G.
S
.
N
.
Ra,
e
t
al
.
,
“
D
y
namic
Ti
m
e
Sli
ce
Ca
lc
ul
at
ion
for
R
ound
Robin
Pr
oce
ss
Schedulin
g
Us
ing
NO
C
,
”
Inte
rnational
Jo
urnal
of El
e
ct
ri
c
al
and
Comput
er
Engi
n
ee
ring
(
IJE
CE)
,
v
ol
/
issue:
5
(
6
)
,
pp
.
1480
-
1
485,
2015
.
[32]
A
.
H
.
Brata,
et
al
.
,
“
Software
Deve
lopment
of
Autom
at
ic
D
at
a
Co
ll
e
ct
or
fo
r
Bus
Route
Planni
ng
S
y
st
em
,
”
Inte
rnational
Jo
urnal
of El
e
ct
ri
c
al
and
Comput
er
Engi
n
ee
ring
(
IJE
CE)
,
v
ol
/
issue:
5
(
1
)
,
pp
.
150
-
15
7,
2015
.
[33]
M.
Vali
nataj,
e
t
al.
,
“
A
rec
onfigura
ble
and
ad
a
pti
ve
routi
ng
m
et
hod
for
fau
lt
-
t
ole
ran
t
m
eshba
sed
net
works
-
on
-
chi
p,
” El
sevi
er,
I
nt.
J. E
l
ectronics
and
Comm
unications
(
AE
Ü)
,
vol
/i
ss
ue:
65
(
7
)
,
pp
.
630
–
640
,
2011
.
[34]
M.
Vali
na
ta
j
,
e
t
al.
,
“
Fault
-
awa
re
and
rec
onf
igurable
routi
ng
al
go
rit
hm
s
for
Networks
-
on
-
Chip,
”
I
ETE
Journal
of
Re
search
,
vo
l
/i
s
sue:
57
(
3
)
,
pp
.
2
15
–
223,
2011
.
[35]
M.
Li,
e
t
al
.
,
“
Dy
XY
-
a
p
roximit
y
conge
stion
-
aw
are
d
ea
dlo
ck
-
fre
e
d
y
namic
rout
i
ng
m
et
hod
for
N
et
work
on
Chip
,
”
Proc.
43th
Desig
n
Aut
omat
ion
C
onfe
renc
e
(
DAC)
,
pp.
849
–
852
,
2006.
[36]
Anala
M
.
R
.
,
e
t al.
,
“
Perform
ance
Anal
y
s
is of
Mesh
-
base
d
NoC’s
on
Routi
ng
Alg
orit
hm
s
,
”
Inte
rn
ati
onal
Journal of
El
e
ct
rica
l
and
C
omputer
Engi
n
e
ering
(
IJE
CE)
,
v
ol
/i
ss
ue:
8
(
5
)
,
20
18.
Evaluation Warning : The document was created with Spire.PDF for Python.