T
E
L
KO
M
NI
K
A
,
V
o
l.
1
4
,
N
o.
3
,
S
ept
em
ber
20
1
6
,
pp.
11
42
~
114
9
I
S
S
N
:
1
693
-
6
930
,
ac
c
r
edi
t
ed
A
b
y
D
IK
T
I,
D
e
c
r
e
e
N
o
:
58/
D
I
K
T
I
/
K
ep/
2013
D
O
I
:
10.
12928/
T
E
LK
O
M
N
I
K
A
.
v
1
4
i
3
.
3648
11
42
R
ec
ei
v
ed
M
ar
c
h
2
9
,
201
6
;
R
ev
i
s
ed
May
1
7
,
2
01
6
;
A
c
c
ept
e
d
J
une
5
,
201
6
F
lo
w
F
air
S
am
p
lin
g
Ba
sed
o
n
M
u
lt
is
t
ag
e
Blo
o
m
F
ilt
er
s
L
i
u
Y
u
a
n
z
h
e
n
*
,
H
u
a
n
g
S
h
u
r
o
n
g
,
L
i
u
J
i
a
n
z
h
a
o
Y
anc
heng
I
ns
t
i
t
ut
e of
T
ec
hno
l
ogy
,
Y
a
nc
heng
2
240
51,
J
i
a
ng
s
u,
C
hi
n
a
*
C
or
r
es
po
ndi
ng a
ut
hor
,
e
-
ma
i
l
:
l
i
uy
uanz
h
en_2
004
@
1
26.
c
om
A
b
st
r
act
N
et
w
or
k
t
r
a
f
f
i
c
di
s
t
r
i
but
i
on i
s
heav
y
-
t
a
i
l
e
d.
M
os
t
of
ne
t
w
or
k
f
l
ow
s
ar
e s
hor
t
and
c
ar
r
y
v
e
r
y
f
e
w
pac
k
et
s
,
and
t
he
num
b
er
of
l
a
r
ge
f
l
ow
s
i
s
s
m
al
l
.
T
r
adi
t
i
on
al
r
andom
s
am
pl
i
ng
t
end
s
t
o
s
a
m
pl
e
m
or
e
l
ar
g
e
f
l
ow
s
t
h
an
s
h
or
t
one
s
.
H
ow
e
v
er
,
m
any
app
l
i
c
at
i
ons
d
epe
nd
on
per
-
f
l
ow
t
r
af
f
i
c
ot
h
er
t
han
j
u
s
t
l
ar
g
e
f
l
ow
s
.
A
f
l
ow
f
ai
r
s
am
p
l
i
n
g
bas
ed
on
m
ul
t
i
s
t
ag
e
B
l
oom
f
i
l
t
er
s
i
s
pr
opo
s
ed.
T
h
e
t
ot
al
m
ea
s
ur
em
en
t
i
n
t
er
v
a
l
i
s
di
v
i
d
e
d
i
nt
o
n
c
h
i
l
d
t
i
m
e
i
nt
er
v
a
l
s
.
I
n
e
ac
h
c
hi
l
d
t
i
m
e
i
nt
er
v
al
,
em
pl
o
y
m
ul
t
i
s
t
age
B
l
oom
f
i
l
t
er
s
t
o
q
u
er
y
t
h
e
i
n
c
om
i
n
g
pac
k
et
’
s
f
l
ow
w
het
her
ex
i
s
t
s
i
n f
l
ow
i
nf
or
m
at
i
o
n t
abl
e or
not
.
I
f
ex
i
s
t
s
,
s
am
pl
e t
he p
ac
k
et
w
i
t
h s
t
a
t
i
c
s
am
pl
i
ng
r
at
e
w
h
i
c
h
i
s
i
nv
er
s
el
y
pr
opor
t
i
o
nal
t
o
t
he
e
s
t
i
m
at
i
o
n
f
l
ow
t
r
af
f
i
c
u
p
t
o
t
he
pr
ev
i
o
us
t
i
m
e
i
n
t
er
v
al
;
if
not
,
t
hat
i
s
i
t
’
s
a
new
f
l
ow
’
s
f
i
r
s
t
pa
c
k
e
t
,
c
r
eat
e
t
he
f
l
ow
i
n
f
or
m
at
i
on
,
i
ns
er
t
i
t
i
n
t
o
t
he
m
ul
t
i
s
t
ag
e
B
l
oom
f
i
l
t
er
s
an
d
s
am
pl
e
t
he
pac
k
et
w
i
t
h
100
%
pr
ob
abi
l
i
t
y
.
T
he
r
es
ul
t
s
s
how
t
h
at
t
he
pr
opo
s
ed
al
g
or
i
t
hm
i
s
ac
c
ur
at
e es
pec
i
al
l
y
f
or
s
hor
t
f
l
ow
s
an
d ea
s
y
t
o e
x
t
e
nd.
Ke
y
w
o
rd
s
:
n
e
tw
o
r
k
tr
a
ff
i
c
, s
h
o
r
t
fl
o
w
,
m
ul
t
i
s
t
ag
e b
l
oom
fi
l
te
r
s
,
f
l
ow
f
ai
r
s
am
pl
i
ng
C
o
p
y
r
i
g
h
t
©
20
16 U
n
i
ver
si
t
a
s A
h
mad
D
ah
l
an
.
A
l
l
r
i
g
h
t
s r
eser
ved
.
1
.
I
n
tr
o
d
u
c
ti
o
n
T
r
a
f
f
i
c
m
eas
ur
e
m
ent
i
s
es
s
ent
i
al
f
or
net
w
or
k
m
anage
m
ent
,
m
oni
t
or
i
ng
a
nd
s
c
he
dul
i
ng
,
and m
an
y
s
t
ud
i
es
ar
e c
ar
r
i
ed ou
t
f
or
di
f
f
er
ent
appl
i
c
at
i
ons
[
1
-
2
].
F
or
h
i
g
h s
pee
d n
et
w
or
k
,
r
out
er
s
c
annot
m
anage t
o gen
er
at
e c
om
pl
et
e dat
a f
or
ev
er
y
pac
k
et
.
T
he
y
ha
v
e t
o em
pl
o
y
r
es
t
r
i
c
t
ed
s
a
m
pl
i
n
g w
hi
c
h i
s
ef
f
ec
t
i
v
e
t
o r
educ
e dat
a an
d r
es
ou
r
c
e c
ons
u
m
pt
i
on.
I
n t
he f
i
e
l
d of
net
w
or
k
t
r
af
f
i
c
m
eas
ur
e
m
ent
,
ac
c
or
di
ng
t
o
t
he
s
am
pl
i
ng
s
t
r
at
eg
y
,
t
her
e
ar
e
m
ai
n
l
y
t
hr
ee
k
i
nds
of
s
a
m
pl
i
n
g
m
et
hods
:
s
y
s
t
e
m
at
i
c
s
a
m
pl
i
ng,
s
t
r
at
i
f
i
e
d r
andom
s
a
m
pl
i
ng and r
an
dom
s
a
m
pl
i
ng.
E
ar
l
y
t
r
af
f
i
c
s
a
m
pl
i
ng m
ai
n
l
y
c
onc
e
nt
r
at
e
d on
pac
k
et
s
am
pl
i
ng,
a
nd
l
at
er
on f
l
o
w
s
a
m
pl
i
n
g.
F
l
o
w
i
s
a c
ol
l
ec
t
i
on of
pac
k
et
s
w
i
t
h t
he s
am
e pr
oper
t
i
es
w
i
t
hi
n a t
i
m
e per
i
od.
T
her
e ar
e
m
an
y
m
et
hods
t
o
di
s
t
i
ngu
i
s
h
a
f
l
o
w
.
T
he
c
om
m
on
us
ed
i
s
t
he
f
i
v
e
t
up
l
e
m
et
hod.
T
he
f
i
v
e
t
up
l
e
r
ef
er
s
t
o s
our
c
e/
des
t
i
nat
i
on I
P
ad
dr
es
s
,
s
our
c
e/
des
t
i
n
at
i
on p
or
t
and pr
ot
oc
o
l
t
y
p
e.
I
f
any
pac
k
et
of
a
f
l
ow
i
s
s
am
pl
ed,
t
hen
t
he f
l
o
w
i
s
s
am
pl
ed.
N
et
w
or
k
t
r
af
f
i
c
i
s
s
el
f
-
s
i
m
i
l
ar
and
t
he
d
i
s
t
r
i
b
ut
i
on
i
s
h
eav
y
-
ta
i
l
ed.
M
os
t
of
net
w
or
k
f
l
ow
s
ar
e
s
hor
t
a
nd
c
ar
r
y
v
er
y
f
e
w
pac
k
et
s
,
and
t
he
num
ber
of
l
ar
ge
f
l
o
w
s
t
hat
c
ar
r
y
l
a
r
ge
am
ount
of
pac
k
et
s
i
s
s
m
al
l
.
A
s
t
ud
y
s
ho
w
s
9%
of
t
h
e
t
ot
a
l
f
l
o
w
s
oc
c
up
y
a
bou
t
9
0%
of
t
he
t
r
af
f
i
c
[
3
].
T
h
e
t
r
adi
t
i
on
al
s
am
pl
i
ng m
et
ho
ds
t
end
t
o s
am
pl
e m
or
e l
ar
ge f
l
o
w
s
t
han
s
hor
t
one
s
.
A
nd
s
om
e
al
g
or
i
t
hm
s
f
oc
us
i
ng
on
l
ar
g
e
f
l
ow
s
am
pl
i
ng
[
4
-
5
]
ar
e
p
r
opos
ed
t
o
s
ol
v
e
pr
o
bl
em
s
i
n
w
hi
c
h
om
i
t
l
ar
ge
f
l
o
w
w
i
l
l
br
i
ng
gr
eat
l
os
s
,
s
uc
h
as
t
r
af
f
i
c
ac
c
ount
i
ng
.
H
o
w
e
v
er
t
her
e
ar
e m
an
y
o
t
her
app
l
i
c
at
i
ons
w
hi
c
h
de
p
en
d
on p
er
-
f
l
o
w
i
nf
or
m
at
i
on r
at
her
t
han
l
ar
ge f
l
o
w
s
,
s
u
c
h as
at
t
ac
k
s
det
ec
t
i
on
[6
-
8]
.
T
he
pur
pos
e
of
t
r
af
f
i
c
m
eas
ur
e
m
ent
det
er
m
i
nes
w
ha
t
k
i
nd
of
s
a
m
pl
i
ng
m
et
hod
t
o be
ad
opt
e
d
[
9
].
N
o
m
at
t
er
w
h
at
k
i
nd of
s
am
pl
i
ng,
as
l
ong as
i
t
em
pl
o
y
s
1 i
n
N
pac
k
et
s
s
t
r
at
eg
y
,
t
he
s
a
m
pl
i
n
g r
es
ul
t
s
t
e
nd t
o
s
a
m
pl
e m
or
e pac
k
et
s
o
f
l
ar
ge f
l
o
w
s
an
d om
i
t
l
ot
s
of
s
hor
t
ones
.
S
am
pl
i
ng
t
ha
t
s
am
pl
es
m
or
e pac
k
et
s
f
r
o
m
s
hor
t
f
l
ow
s
at
ex
p
ens
e of
l
o
w
s
am
pl
i
n
g r
at
e
of
l
ar
ge
f
l
ow
s
,
i
s
c
al
l
ed
f
ai
r
s
am
pl
i
n
g.
I
n
t
he
f
i
el
d
of
f
a
i
r
s
a
m
pl
i
ng,
s
ev
er
al
r
es
ear
c
hes
h
av
e
been
c
ar
r
i
ed
out
[
10
-
12
].
A
s
k
et
c
h gui
de
s
a
m
pl
i
n
g a
l
gor
i
t
hm
i
s
pr
opo
s
ed
[
13
],
m
ak
e t
he pr
oba
bi
l
i
t
y
w
i
t
h
w
hi
c
h
an
i
nc
om
i
ng p
ac
k
et
i
s
s
am
pl
e
d a
dec
r
e
as
i
ng
s
am
pl
i
ng
f
unc
t
i
on
of
t
h
e s
i
z
e
of
t
he
f
l
o
w
t
h
e
pac
k
et
bel
ongs
t
o
, b
ut
t
h
e s
hor
t
f
l
o
w
es
t
i
m
at
i
on
er
r
or
i
s
gr
e
at
,
and
t
he
s
pac
e
ef
f
i
c
i
enc
y
i
s
l
o
w
.
Z
ha
ng J
,
et
al
.
,
[
14
]
pr
o
pos
e
s
p
ac
e
ef
f
i
c
i
ent
pac
k
et
f
ai
r
s
am
pl
i
ng,
t
r
af
f
i
c
i
s
as
s
u
m
ed
t
o
f
ol
l
o
w
Z
i
pf
di
s
t
r
i
b
ut
i
on
w
i
t
h par
am
et
er
z
,
and
d
es
i
gns
t
he
m
ul
t
i
r
es
ol
ut
i
o
n s
am
pl
i
n
g s
t
at
i
s
t
i
c
s
,
but
c
om
put
i
n
g
c
o
m
pl
ex
i
t
y
i
s
hi
g
her
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
K
A
I
S
S
N
:
1
693
-
6
930
F
l
ow
F
a
i
r
S
am
pl
i
ng
B
as
e
d
on
Mul
t
i
s
t
ag
e B
l
o
om F
i
l
t
er
s
(
Li
u Y
u
an
z
he
n
)
1143
I
n or
d
er
t
o
m
eet
t
he
nee
ds
of
t
hos
e a
pp
l
i
c
at
i
ons
t
ha
t
n
eed
per
-
f
l
o
w
i
nf
or
m
at
i
on,
a
nov
el
f
l
ow
f
ai
r
s
am
pl
i
ng m
et
hod bas
ed on m
ul
t
i
s
t
a
ge B
l
oom
f
i
l
t
er
s
(
MB
F
)
i
s
pr
opo
s
ed t
o c
apt
ur
e m
or
e
i
nf
or
m
at
i
on es
pec
i
al
l
y
f
r
om
s
hor
t
f
l
o
w
s
.
T
he
w
h
ol
e m
eas
ur
em
ent
i
nt
er
v
al
i
s
di
v
i
ded
i
nt
o n
um
ber
s
of
c
hi
l
d
t
i
m
e i
nt
er
v
a
l
s
.
D
ur
i
ng a c
hi
l
d
i
nt
er
v
a
l
,
s
am
pl
i
n
g r
at
e f
or
o
ne f
l
o
w
r
em
ai
ns
unc
han
ged
.
F
or
an
i
nc
om
i
ng p
ac
k
et
,
m
ul
t
i
s
t
age
B
l
oom
f
i
l
t
er
s
ar
e
em
pl
o
y
ed
t
o f
i
n
d t
he c
or
r
e
s
pond
i
ng
f
l
o
w
i
nf
or
m
at
i
on
t
abl
e,
an
d
s
am
pl
e
t
h
e
pac
k
et
bel
o
ngs
t
o
t
he
f
l
o
w
w
i
t
h
a
c
er
t
ai
n
s
am
pl
i
ng
r
at
e
.
T
he
s
a
m
pl
i
n
g r
at
e
i
s
i
n
v
er
s
el
y
p
r
opor
t
i
ona
l
t
o t
h
e es
t
i
m
at
i
on t
r
af
f
i
c
of
t
he f
l
o
w
u
p t
o t
h
e pr
e
v
i
ous
c
hi
l
d
in
t
e
r
v
a
l.
F
or
ne
w
ar
r
i
v
al
f
l
o
w
,
t
her
e i
s
n
o f
l
o
w
i
nf
or
m
at
i
on,
s
o a f
ul
l
s
am
pl
i
n
g r
at
e i
s
t
ak
en,
t
hat
i
s
s
a
m
pl
i
n
g
t
h
e
pac
k
et
w
i
t
h
1
00%
pr
oba
bi
l
i
t
y
.
T
he
a
l
gor
i
t
hm
i
s
s
i
m
pl
e
t
o
i
m
pl
em
ent
,
t
i
m
e
i
n
t
er
v
al
s
et
t
i
n
g i
s
f
l
ex
i
bl
e
,
an
d t
h
e r
es
ul
t
s
s
ho
w
t
ha
t
i
t
i
s
ac
c
ur
a
t
e es
pe
c
i
al
l
y
f
or
s
hor
t
f
l
o
w
s
.
T
he r
es
t
o
f
t
hi
s
paper
i
s
or
gan
i
z
ed as
f
ol
l
o
w
s
.
S
ec
t
i
on
2
pr
o
v
i
des
a br
i
ef
s
ur
v
e
y
of
B
lo
o
m
f
ilt
e
r
an
d m
ul
t
i
s
t
a
g
e B
l
o
om
f
i
l
t
er
s
.
S
ec
t
i
on
3
pr
es
ent
s
a
det
ai
l
ed
des
c
r
i
pt
i
on
of
t
he
pr
opos
e
d f
l
ow
f
ai
r
s
am
pl
i
ng.
S
ec
t
i
on
4
g
i
v
es
t
he
or
et
i
c
a
l
ana
l
y
s
i
s
and s
ec
t
i
on
5
gi
v
es
ex
per
i
m
ent
al
ana
l
y
s
i
s
.
W
e
c
onc
l
ud
e i
n S
ec
t
i
o
n
6
.
2
. B
lo
o
m
F
ilt
e
r
a
n
d
M
u
l
ti
s
ta
g
e
B
l
o
o
m
F
i
l
te
r
s
A
B
l
o
om
f
i
l
t
er
i
s
a s
pac
e
-
ef
f
i
c
i
ent
dat
a
s
t
r
uc
t
ur
e
and
i
s
us
ed
t
o
t
es
t
w
het
her
an
el
em
ent
i
s
a m
e
m
ber
of
a s
et
.
I
t
s
uppor
t
s
m
e
m
ber
quer
i
es
,
r
andom
s
t
or
age an
d has
c
ons
t
ant
h
as
h
l
ook
up t
i
m
e.
Y
ou c
an a
d
d an el
em
ent
and qu
er
y
f
or
an el
em
ent
w
i
t
h B
l
oo
m
f
i
l
t
er
.
S
i
nc
e
pr
opos
e
d
b
y
B
l
o
om
B
ur
t
on
i
n
197
0
[1
5
]
,
i
t
has
bee
n
us
ed
i
n
d
at
ab
as
e
ap
pl
i
c
at
i
ons
,
and
s
om
e
i
m
pr
ov
ed
a
l
gor
i
t
hm
s
ar
e
pr
opos
ed
[
1
6
-
1
7
]
.
I
n
r
ec
e
nt
y
ear
s
,
i
t
has
b
een
w
i
d
el
y
c
o
nc
er
ned
i
n
t
h
e
f
i
el
d of
ne
t
w
or
k
[
1
8
-
19
].
A
n
em
pt
y
B
l
oom
f
i
l
t
er
i
s
a
bi
t
ar
r
a
y
of
m
b
i
t
s
,
a
l
l
s
et
t
o
0.
T
her
e m
us
t
al
s
o be
k
di
f
f
er
ent
has
h f
unc
t
i
ons
def
i
n
ed,
w
i
t
h a uni
f
or
m
r
andom
di
s
t
r
i
but
i
on
.
E
a
c
h of
has
h f
unc
t
i
on
s
has
hes
s
o
m
e
s
et
el
em
ent
t
o
one
of
t
he
m
ar
r
a
y
pos
i
t
i
o
ns
,
an
d t
h
e
c
or
r
es
pond
i
ng
pos
i
t
i
o
ns
a
r
e s
et
t
o
1,
as
s
ho
w
n i
n
F
i
gur
e 1.
F
ig
ur
e
1
. S
t
an
dar
d
B
l
oom
f
i
l
t
er
I
f
an
el
em
ent
bel
o
ngs
t
o
a
s
et
,
t
hen
t
he
c
or
r
es
pon
di
ng
has
h
pos
i
t
i
o
ns
m
us
t
be
s
et
t
o
1.
F
al
s
e
n
egat
i
v
e
m
at
c
hes
ar
e
i
m
pos
s
i
bl
e.
H
o
w
e
v
er
,
du
e
t
o
t
h
e
s
am
e
pos
i
t
i
on
m
a
y
be
s
et
t
o
1
f
or
m
an
y
t
i
m
es
,
ot
her
e
l
em
ent
s
’
has
hi
ng
s
et
t
i
ng
m
a
y
l
ea
d
t
o
s
om
e
el
em
ent
does
n
ot
be
l
on
g
t
o
t
h
e
s
et
and
i
t
s
c
or
r
es
pon
di
n
g
h
as
h
pos
i
t
i
ons
ar
e
s
et
t
o
1,
t
hus
t
he
el
em
ent
i
s
er
r
oneo
us
l
y
as
s
um
ed
t
o bel
ong t
o t
he s
et
.
T
he pr
obab
i
l
i
t
y
i
s
c
al
l
ed f
al
s
e pos
i
t
i
v
e er
r
or
,
or
i
n s
hor
t
f
al
s
e pos
i
t
i
v
e.
F
al
s
e
pos
i
t
i
v
e m
at
c
hes
ar
e pos
s
i
bl
e.
T
he pr
oba
bi
l
i
t
y
of
f
al
s
e pos
i
t
i
v
e
w
i
l
l
be
an
al
y
z
e
d
be
l
o
w
.
A
s
s
um
e t
hat
a has
h f
unc
t
i
on s
el
ec
t
s
eac
h ar
r
a
y
p
os
i
t
i
o
n w
i
t
h e
qua
l
pr
ob
abi
l
i
t
y
.
Let
n
be
t
he
num
ber
of
t
he
s
et
el
e
m
ent
s
,
m
be
t
he
num
ber
of
bi
t
s
i
n
t
he
ar
r
a
y
,
and
k
b
e
t
he
num
ber
o
f
has
h
f
unc
t
i
ons
.
T
he
pr
ob
a
bi
l
i
t
y
t
hat
a
c
er
t
ai
n
b
i
t
i
s
s
et
t
o
1
b
y
a
c
er
t
a
i
n
has
h
f
unc
t
i
o
n
i
s
1/
m
.
T
he pr
obab
i
l
i
t
y
t
h
at
i
s
not
s
et
t
o
1 i
s
:
1
1
m
−
F
or
t
her
e ar
e
k
has
h f
unc
t
i
ons
,
s
o t
he pr
ob
abi
l
i
t
y
t
hat
t
he bi
t
i
s
not
s
et
t
o 1 b
y
an
y
of
t
he
k
f
unc
t
i
ons
dur
i
ng
one
i
ns
er
t
i
on
i
s
:
1
1
k
m
−
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
1
6
9
3
-
6
930
T
E
L
KO
M
NI
K
A
V
o
l.
1
4
,
N
o
.
3
,
S
ept
em
ber
201
6
:
11
42
–
1
149
1144
A
f
t
er
i
ns
er
t
n
e
l
em
ent
s
,
t
h
e pr
ob
ab
i
l
i
t
y
(
den
ot
e
d as
0
p
)
t
hat
c
er
t
a
i
n
pos
i
t
i
o
n s
t
i
l
l
i
s
s
et
t
o 0
i
s
:
0
1
1
kn
m
p
=
−
T
he f
al
s
e pos
i
t
i
v
e (
den
ot
e
d
as
er
r
p
)
is
:
(
)
0
1
1
11
k
kn
k
er
r
p
m
p
=
−
=
−−
(
)
/
1
k
kn
m
e
−
≈−
(
)
/
ln
1
kn
m
k
e
e
−
=
W
h
ile
(
)
/
ln
2
k
mn
=
, th
a
t
i
s
0
1/
2
p
=
,
er
r
p
t
ak
es
t
he m
i
ni
m
u
m
v
al
ue
:
/
ln
2
/
mi
n
11
1
0.6185
22
k
mn
mn
er
r
p
−
=
−=
=
Mul
t
i
s
t
ag
e B
l
oom
f
i
l
t
er
s
i
s
a
par
al
l
el
s
t
r
uc
t
ur
e of
s
ev
er
a
l
B
l
oom
f
i
l
t
er
s
,
as
s
ho
w
n i
n
F
i
gur
e
2.
E
ac
h
B
l
oom
f
i
l
t
er
em
pl
o
y
s
d
i
f
f
er
ent
has
h
f
u
nc
t
i
on
c
l
us
t
er
s
.
T
o
ad
d
a
n
el
em
ent
,
e
v
er
y
B
l
o
om
f
i
l
t
er
s
houl
d m
a
k
e an ad
d op
er
at
i
on.
T
o t
es
t
an
el
em
ent
,
o
nl
y
e
v
er
y
B
l
o
om
f
i
l
t
er
i
ndi
c
a
t
es
t
he e
l
e
m
ent
i
s
i
n t
h
e s
et
.
A
l
t
hou
gh t
he
i
nh
er
ent
f
al
s
e
pos
i
t
i
v
e er
r
or
of
B
l
oom
f
i
l
t
er
ex
i
s
t
s
,
t
he
di
f
f
er
ent
has
h f
unc
t
i
o
n c
l
us
t
er
s
c
an gr
eat
l
y
r
e
duc
e t
he
er
r
or
.
Let
p
i
s
t
he f
al
s
e
pos
i
t
i
v
e er
r
or
of
a
s
i
ngl
e
B
l
oom
f
i
l
t
er
,
n
i
s
t
he
num
ber
of
s
t
ages
,
as
s
um
e eac
h
B
l
o
om
f
i
l
t
er
has
e
qua
l
f
al
s
e pos
i
t
i
v
e
er
r
or
,
t
hen
t
he
f
al
s
e p
os
i
t
i
v
e er
r
or
of
a m
ul
t
i
s
t
ag
e B
l
o
o
m
f
i
l
t
er
s
w
i
t
h
n
s
t
ag
es
i
s
n
p
.
F
i
gur
e 2.
Mul
t
i
s
t
a
ge B
l
oom
f
i
l
t
er
s
3
.
F
l
o
w
F
a
i
r
S
a
m
p
l
i
n
g
M
e
t
h
o
d
Let
(
)
1
,
2...
=
k
fk
be t
he f
l
ow
dur
i
ng
t
he m
eas
ur
e
m
ent
i
nt
er
v
al
.
D
i
v
i
de t
he
w
h
ol
e
m
eas
ur
e
m
ent
i
nt
er
v
al
i
nt
o
n
c
hi
l
d
t
i
m
e i
nt
er
v
a
l
s
.
w
e
c
a
r
r
y
out
f
l
o
w
s
am
pl
i
n
g
v
i
a p
a
c
k
et
s
am
pl
i
ng,
t
hat
i
s
i
f
a
n
y
pac
k
et
of
a
f
l
o
w
i
s
s
am
pl
ed,
t
h
en
t
he
f
l
o
w
i
s
s
am
pl
ed
,
an
d
s
t
at
i
s
t
i
c
s
ar
e
b
as
ed
on
f
l
ow
s
not
p
ac
k
et
s
.
C
ol
l
ec
t
s
a
m
pl
ed pac
k
et
s
ev
er
y
c
hi
l
d
t
i
m
e i
nt
er
v
al
an
d
s
am
pl
i
ng
r
at
e
f
or
t
he
f
l
ow
i
s
unc
h
ang
ed
dur
i
ng
t
he
c
hi
l
d
t
i
m
e
i
nt
er
v
a
l
.
W
e
em
pl
o
y
f
l
o
w
i
nf
or
m
at
i
on
t
a
bl
e
t
o
m
ai
nt
a
i
n
s
a
m
pl
ed
f
l
o
w
i
nf
or
m
at
i
on.
Mul
t
i
s
t
ag
e
B
l
oom
f
i
l
t
er
s
ar
e
us
ed
t
o
add
or
qu
er
y
f
or
c
or
r
es
pondi
ng
f
l
ow
r
ec
or
d i
n t
h
e f
l
o
w
i
nf
or
m
at
i
on t
ab
l
e.
F
or
ne
w
ar
r
i
v
a
l
pac
k
et
,
l
oo
k
f
or
i
t
i
n
t
he
m
ul
t
i
s
t
ag
e
B
l
o
om
f
i
l
t
er
s
,
i
f
an
y
of
t
he
b
i
t
s
at
t
he
has
h pos
i
t
i
o
ns
i
s
0,
w
hi
c
h m
eans
t
he c
or
r
es
pondi
n
g f
l
o
w
r
ec
or
d d
oes
not
ex
i
s
t
,
t
he
n
a
dd
t
h
e
pac
k
et
i
nt
o
m
ul
t
i
s
t
age
B
l
oo
m
f
i
l
t
er
s
and
c
r
ea
t
e
t
he
f
l
o
w
i
nf
or
m
at
i
on
r
ec
or
d.
T
o
add
a
n
e
l
em
ent
,
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
K
A
I
S
S
N
:
1
693
-
6
930
F
l
ow
F
a
i
r
S
am
pl
i
ng
B
as
e
d
on
Mul
t
i
s
t
ag
e B
l
o
om F
i
l
t
er
s
(
Li
u Y
u
an
z
he
n
)
1145
a
ll
B
lo
o
m
f
ilt
e
r
s
’
c
or
r
es
pon
di
n
g pos
i
t
i
ons
s
hou
l
d
be s
e
t
t
o 1.
I
f
t
he r
ec
or
d ex
i
s
t
s
,
t
hat
i
s
al
l
B
l
oom
f
ilt
e
r
s
in
d
ic
a
t
e
t
h
e
c
or
r
es
po
ndi
ng
pos
i
t
i
o
ns
ar
e
s
et
t
o
1,
s
am
pl
e
t
he
pac
k
et
s
bel
o
ng
t
o
t
he
f
l
o
w
w
i
t
h s
t
at
i
c
s
am
pl
e r
at
e
t
k
p
and
upd
at
e t
he f
l
o
w
i
nf
or
m
at
i
on
,
as
s
ho
w
n i
n
F
i
gur
e 3
.
F
i
gur
e
3.
s
am
pl
i
n
g pr
oc
es
s
w
i
t
h
m
ul
t
i
s
t
age
B
l
oom
f
i
l
t
er
s
T
he s
a
m
pl
i
ng
r
at
e
t
k
p
i
s
unc
hang
ed
d
ur
i
n
g
t
h
e
c
hi
l
d
t
i
m
e
i
nt
er
v
al
.
I
t
i
s
i
nv
er
s
el
y
pr
opor
t
i
on
al
t
o t
he es
t
i
m
at
e
t
r
af
f
i
c
1
−
t
k
N
up t
o t
h
e pr
e
v
i
o
us
c
hi
l
d t
i
m
e i
nt
er
v
al
.
I
f
t
he f
l
ow
i
s
t
t
i
m
e
i
nt
er
v
a
l
ne
w
ar
r
i
v
a
l
f
l
o
w
,
t
h
en t
he
1
−
t
k
N
equa
l
s
t
o 0,
t
hen
s
am
pl
i
ng r
at
e
t
k
p
equal
s
t
o
1.
A
s
t
i
m
e
goes
b
y
,
t
h
e
c
ur
r
ent
es
t
i
m
at
e
t
r
af
f
i
c
1
−
t
k
N
gr
o
w
s
,
i
f
a
f
l
o
w
i
s
not
ov
er
,
t
he
s
am
pl
i
n
g
r
a
t
e
f
or
t
he
f
l
ow
dec
r
eas
es
w
i
t
h t
i
m
e
as
a s
t
ep f
unc
t
i
on.
S
am
pl
i
ng r
at
e
t
k
p
i
s a
s f
o
l
l
o
w
s:
(
1)
W
he
r
e
ε
i
s
a c
ons
t
ant
.
Let
i
k
N
be t
h
e
num
ber
of
pac
k
et
s
s
a
m
pl
ed
f
r
om
f
l
ow
k
f
dur
i
n
g c
hi
l
d
t
i
m
e i
nt
er
v
al
t
,
a
nd t
hen
t
he
es
t
i
m
at
e t
r
af
f
i
c
of
f
l
ow
k
f
af
t
er
t
i
m
e i
nt
er
v
al
t
is
:
1
/
=
=
∑
t
t
i
t
k
kk
i
N
Np
(2
)
F
l
o
w
f
ai
r
s
am
pl
i
n
g pr
oc
es
s
as
f
ol
l
o
w
s
:
di
v
i
d
e t
h
e m
eas
ur
em
ent
i
nt
o
n
c
h
ild
t
im
e
in
t
e
r
v
a
ls
;
in
i
t
ia
li
z
e
0
k
N
t
o
0
/
/
i
n
it
ia
li
z
e
t
r
a
f
f
ic
is
0
in
i
t
ia
li
z
e
a
l
l
1
k
p
t
o 1
/
/
s
am
pl
i
n
g r
at
e f
or
a
ne
w
f
l
o
w
dur
i
ng
a t
i
m
e i
nt
er
v
a
l
i
s
1
w
hi
l
e (
t
i
m
e i
nt
er
v
al
t
)
/
/
d
u
r
i
ng t
he c
h
i
l
d t
i
m
e i
nt
er
v
a
l
t
,
1...
=
tn
x
=
ar
r
i
v
al
pac
k
et
;
i
f
(
B
l
oom
f
i
l
t
er
s
t
age
_i
(
(
)
(
)
(
)
12
,
,
...,
k
h
x
h
x
h
x
==1
)
// th
a
t
i
s
x
ex
i
s
t
s
i
n f
l
o
w
i
nf
o
t
ab
l
e
s
am
pl
e t
he
p
ac
k
et
w
i
t
h s
t
at
i
c
r
at
e
t
k
p
upda
t
e f
l
o
w
i
nf
or
m
at
i
on
el
s
e
s
et
B
l
o
om
f
i
l
t
er
s
t
ag
e
_i
(
(
)
(
)
(
)
12
,
,
...,
k
h
x
h
x
h
x
=
1 /
/
i
ns
er
t
t
he e
l
em
ent
i
n
t
o M
B
F
1
2
1
1
1
2
ε
−
=
+
t
k
t
k
p
N
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
1
6
9
3
-
6
930
T
E
L
KO
M
NI
K
A
V
o
l.
1
4
,
N
o
.
3
,
S
ept
em
ber
201
6
:
11
42
–
1
149
1146
s
a
m
pl
e t
h
e p
ac
k
et
c
r
eat
e f
l
o
w
r
ec
or
d
if
t
im
e
in
t
e
r
v
a
l t
is
o
u
t
c
o
m
put
e
1
+
t
k
p
/
/
ac
c
or
di
n
g t
o
equ
at
i
on(
1)
c
o
m
put
e
t
k
N
/
/
ac
c
or
di
n
g t
o
equ
at
i
on(
2)
if
it
’s
t
im
e
t
o
c
le
a
r
M
u
lt
is
t
a
g
e
B
lo
o
m
f
ilt
e
r
s
c
l
ear
M
ul
t
i
s
t
a
ge
B
l
o
o
m
f
i
l
t
er
s
,
al
l
bi
t
s
s
et
t
o
0 /
/
e
v
er
y
t
hr
es
h
ol
d
t
i
m
e
T
/
/
t
hr
es
h
ol
d
=
∆
T
kt
,
w
h
er
e
k
i
s
an i
nt
e
ger
,
∆
t
i
s
t
he
dur
a
t
i
o
n of
c
hi
l
d t
i
m
e i
nt
er
v
al
F
or
t
he d
ur
at
i
on
t
i
m
e of
s
hor
t
f
l
o
w
i
s
s
hor
t
an
d l
ar
g
e f
l
o
w
i
s
l
on
g,
s
o s
am
pl
i
n
g ne
w
ar
r
i
v
a
l
f
l
o
w
f
ul
l
y
c
a
n get
m
or
e s
hor
t
f
l
ow
s
’
i
nf
or
m
at
i
on.
A
s
t
he num
ber
of
l
ar
ge f
l
ow
s
i
s
s
m
al
l
,
t
he
c
ons
um
ed r
es
our
c
es
ar
e l
i
m
i
t
ed dur
i
ng c
hi
l
d t
i
m
e i
nt
er
v
a
l
.
I
n or
der
t
o
r
educ
e t
he pr
es
s
ur
e o
f
r
es
our
c
es
c
aus
ed
b
y
s
am
pl
i
n
g,
w
e
c
an
d
i
v
i
d
e
t
h
e
m
eas
ur
em
ent
i
nt
er
v
al
i
n
t
o
s
ho
r
t
er
c
hi
l
d
t
i
m
e
i
nt
er
v
a
l
s
.
T
hen
t
he
r
es
our
c
es
c
ons
u
m
pt
i
on
f
or
l
ar
ge
f
l
o
w
s
dur
i
ng
t
he
f
i
r
s
t
ar
r
i
v
al
t
i
m
e
i
nt
er
v
a
l
i
s
r
educ
ed,
an
d dur
i
ng
ot
h
er
t
i
m
e i
nt
er
v
a
l
s
s
am
pl
e l
ar
g
e f
l
o
w
s
w
i
t
h c
er
t
a
i
n s
am
pl
i
n
g r
at
e.
A
t
t
h
e e
nd
of
t
he t
i
m
e i
nt
er
v
a
l
,
t
he f
l
o
w
r
ec
or
ds
w
h
i
c
h ha
v
e no
t
b
een
up
dat
e
d ar
e
as
s
u
m
ed i
n
ac
t
i
v
e
f
l
o
w
s
an
d
r
em
ov
ed f
r
om
t
he f
l
o
w
i
nf
or
m
at
i
on t
ab
l
e.
C
l
ear
u
p m
ul
t
i
s
t
ag
e
B
l
o
om
f
i
l
t
er
s
ev
er
y
t
hr
es
ho
l
d t
i
m
e
T
t
o r
educ
e
c
onf
l
i
c
t
of
h
as
h c
ol
l
i
s
i
on,
a
nd
at
t
h
e s
am
e t
i
m
e
i
t
w
i
l
l
br
eak
t
he f
l
ow
i
nf
or
m
at
i
on and br
i
ng er
r
or
es
pec
i
a
l
l
y
f
or
l
ar
ge f
l
ow
s
w
h
os
e dur
at
i
o
n t
i
m
e go
ac
r
os
s
s
ev
er
a
l
t
i
m
e i
nt
er
v
al
s
.
T
hr
es
hol
d
T
i
s
c
ons
t
ant
t
i
m
es
of
dur
at
i
on
∆
t
of
c
hi
l
d t
i
m
e
in
t
e
r
v
a
l:
=
∆
T
kt
(
3)
W
h
er
e
k
i
s
an i
n
t
eg
er
.
4.
T
h
eo
r
et
i
cal
A
n
al
ysi
s
4
.
1.
V
al
u
e E
st
i
m
at
i
o
n
T
he ent
i
r
e m
eas
ur
em
ent
t
i
m
e i
s
di
v
i
ded
i
nt
o
n
c
h
i
l
d t
i
m
e i
nt
er
v
al
s
.
Let
i
k
N
b
e t
he
pac
k
et
s
nu
m
ber
and
i
k
X
be by
t
es
num
ber
s
am
pl
ed f
r
o
m
f
l
ow
k
f
dur
i
ng c
hi
l
d t
i
m
e i
nt
er
v
al
t
.
T
hen t
he t
ot
a
l
pac
k
et
s
num
ber
es
t
i
m
at
i
on
v
al
u
e
k
N
is
:
1
/
=
=
∑
n
i
t
k
kk
i
N
Np
(
4)
T
he t
ot
al
b
y
t
es
num
ber
es
t
i
m
at
i
on v
a
l
u
e
k
X
is
:
1
/
=
=
∑
n
i
t
k
kk
i
X
Xp
(
5)
4
.
2
.
S
p
a
c
e
C
o
n
s
u
m
p
ti
o
n
T
he s
pac
e c
ons
um
pt
i
on
of
t
he pr
opos
e
d a
l
gor
i
t
hm
m
ai
nl
y
c
om
es
f
r
o
m
t
w
o
par
t
s
,
one
i
s
t
he m
ul
t
i
s
t
ag
e B
l
o
om
f
i
l
t
er
s
t
ak
i
ng up s
pac
e
,
an
d t
he s
ec
o
nd
i
s
f
l
o
w
i
nf
or
m
at
i
on t
ab
l
e
m
ai
nt
enanc
e.
A
B
l
oom
f
i
l
t
e
r
i
s
a
s
pac
e
-
ef
f
i
c
i
ent
has
h
dat
a
s
t
r
uc
t
ur
e
an
d
get
s
i
t
s
s
pac
e
ef
f
i
c
ac
y
at
t
he
c
os
t
of
f
al
s
e
pos
i
t
i
v
e
er
r
or
.
T
he
s
t
ages
of
m
ul
t
i
s
t
age
B
l
oom
f
i
l
t
er
s
c
an
be
det
er
m
i
ned
b
y
t
he ba
l
anc
e of
r
es
our
c
es
and f
al
s
e pos
i
t
i
v
e er
r
or
.
F
or
f
l
ow
i
nf
or
m
at
i
on t
ab
l
e m
ai
nt
en
anc
e,
t
he
w
or
s
t
c
as
e
i
s
t
hat
t
he s
pa
c
e c
ons
um
pt
i
on
i
s
c
ons
t
a
n
t
t
i
m
es
of
t
he num
ber
of
f
l
o
w
s
dur
i
n
g
t
h
e
t
ot
a
l
m
eas
ur
e
m
ent
t
i
m
e.
H
o
w
e
v
er
,
f
or
t
he
num
ber
of
s
hor
t
f
l
o
w
s
i
s
l
ar
ge
an
d t
h
e
dur
at
i
on t
i
m
e i
s
s
hor
t
,
at
t
he
end
of
t
he
c
h
i
l
d
t
i
m
e
i
nt
er
v
al
,
t
h
e
f
l
o
w
s
h
av
e
not
b
een
u
pda
t
ed
ar
e
r
em
ov
ed
f
r
o
m
f
l
ow
i
nf
or
m
at
i
on t
a
bl
e t
o
d
o f
ur
t
her
an
al
y
s
i
s
,
t
hus
r
el
eas
e m
uc
h s
pac
e,
s
o t
h
e
s
pac
e ne
ede
d
t
ur
ns
t
o c
ons
t
ant
t
i
m
es
of
f
l
o
w
s
dur
i
ng c
h
i
l
d t
i
m
e i
nt
er
v
al
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
K
A
I
S
S
N
:
1
693
-
6
930
F
l
ow
F
a
i
r
S
am
pl
i
ng
B
as
e
d
on
Mul
t
i
s
t
ag
e B
l
o
om F
i
l
t
er
s
(
Li
u Y
u
an
z
he
n
)
1147
4.
3
.
Er
r
o
r
A
n
a
l
y
s
i
s
T
he
er
r
or
i
s
c
o
m
pos
ed
of
s
am
pl
i
ng
er
r
or
,
f
al
s
e
pos
i
t
i
v
e
er
r
or
of
m
ul
t
i
s
t
age
B
l
oom
f
i
l
t
er
s
and
er
r
or
c
aus
e
d
b
y
r
egu
l
a
r
l
y
c
l
e
ar
up
m
ul
t
i
s
t
a
ge
B
l
o
o
m
f
i
l
t
er
s
.
T
he
f
al
s
e
pos
i
t
i
v
e
o
f
m
ul
t
i
s
t
age
B
l
o
om
f
i
l
t
er
s
i
s
ex
p
one
nt
i
al
l
y
dec
r
eas
ed
r
ef
er
t
o
s
i
ng
l
e
B
l
oom
f
i
l
t
er
.
C
l
e
ar
up
m
ul
t
i
s
t
a
ge
B
l
oom
f
i
l
t
er
s
r
egul
ar
l
y
i
s
t
o
r
ed
uc
e
has
h
c
ol
l
i
s
i
o
n
and
br
i
ngs
er
r
or
es
pec
i
al
l
y
f
or
f
l
ow
s
w
hos
e
dur
at
i
o
n
t
i
m
e
go
ac
r
os
s
t
he
t
i
m
e
i
nt
er
v
a
l
s
.
T
he
t
hr
es
ho
l
d
t
i
m
e
i
s
i
nt
eger
t
i
m
es
of
t
he
t
i
m
e
i
nt
er
v
a
l
,
a
nd
bet
t
er
l
o
ng
er
t
h
an s
h
or
t
f
l
o
w
dur
at
i
on
t
i
m
e,
w
h
i
c
h c
an
r
ed
uc
e
er
r
or
t
o
s
hor
t
f
l
o
w
s
.
C
om
par
at
i
v
e
l
y
,
t
he m
ai
n
er
r
or
i
s
s
am
pl
i
ng
er
r
or
.
S
am
pl
i
ng
i
s
an
ef
f
ec
t
i
v
e m
et
ho
d f
or
dat
a
r
educ
t
i
o
n,
ho
w
ev
er
,
s
am
pl
i
ng
i
t
s
el
f
c
a
n
n
ot
av
oi
d
i
n
h
er
ent
s
t
at
i
s
t
i
c
al
er
r
or
.
I
n
our
a
l
gor
i
t
hm
,
s
a
m
pl
i
n
g er
r
or
i
s
m
or
e f
or
l
ar
ge f
l
o
w
s
w
hos
e s
am
pl
i
ng
r
at
e i
s
i
n
v
er
s
el
y
pr
o
por
t
i
on
al
to
i
ts
tr
a
f
f
i
c
,
and
ha
v
e ac
c
ur
at
e i
nf
or
m
at
i
on
es
t
i
m
at
i
on f
or
s
hor
t
f
l
o
w
s
.
5
.
E
xp
er
i
m
en
t
al
A
n
al
ysi
s
In
th
i
s
s
ec
t
i
o
n,
w
e us
e t
he
ac
t
ual
dat
a of
t
h
e I
nt
er
net
t
o c
ar
r
y
o
ut
ex
p
er
i
m
ent
al
an
al
y
s
i
s
.
D
at
a c
om
es
f
r
o
m
t
he U
.
S
.
N
at
i
o
na
l
Lab
or
at
or
y
f
or
A
pp
l
i
c
at
i
o
n N
et
w
or
k
R
es
e
ar
c
h (
N
LA
N
R
)
p
u
b
lic
l
y
a
v
a
il
a
b
le
T
R
A
C
E
.
F
l
o
w
f
ai
r
s
am
pl
i
ng r
a
t
e c
hang
es
w
i
t
h t
i
m
e i
nt
er
v
a
l
.
F
i
gur
e 4
s
ho
w
s
s
am
pl
i
n
g r
at
e
c
o
m
par
i
s
on
of
our
f
l
o
w
f
ai
r
s
a
m
pl
i
ng
and
r
a
ndom
s
am
pl
i
ng
w
i
t
h
s
am
pl
i
ng
r
at
e
equa
l
s
t
o
0.
3
f
or
t
he
s
am
e
f
l
o
w
,
and
t
h
e
t
i
m
e
i
nt
er
v
al
i
s
1
m
ill
is
e
c
o
n
d
(
m
s
)
.
S
am
pl
i
ng
r
a
t
e
of
our
f
l
o
w
s
am
pl
i
ng
m
et
hod dec
r
eas
es
as
a s
t
e
p f
unc
t
i
on
ev
er
y
c
hi
l
d t
i
m
e i
nt
er
v
al
.
W
hi
l
e 1 out
N
r
and
om
s
a
m
pl
i
ng’
s
s
a
m
pl
i
n
g r
at
e
i
s
s
et
ar
t
i
f
i
c
i
a
l
or
ac
c
or
d
i
ng
t
o a
c
er
t
ai
n r
ul
e
and
r
em
ai
ns
s
t
at
i
c
.
F
i
gur
e 4.
S
am
pl
i
ng
r
at
e c
o
m
par
i
s
on
U
nder
t
he
s
am
e pac
k
et
s
am
pl
i
ng r
at
e
,
o
ur
f
l
o
w
f
ai
r
s
am
pl
i
ng m
et
hod
t
ur
ns
t
o s
am
pl
e
m
or
e
f
l
ow
.
F
i
gur
e 5 s
how
s
under
t
h
e s
am
e
pac
k
et
s
a
m
pl
i
ng r
at
e,
f
l
o
w
f
ai
r
s
a
m
pl
i
ng and r
an
dom
s
a
m
pl
i
n
g’
s
f
l
ow
r
at
i
o
s
am
pl
ed.
R
an
dom
s
a
m
pl
i
ng
s
a
m
pl
es
1
pac
k
et
f
r
o
m
N
,
f
or
net
w
or
k
t
r
af
f
i
c
di
s
t
r
i
b
ut
i
on i
s
h
ea
v
y
-
t
ai
l
ed
,
m
aj
or
i
t
y
of
pac
k
et
s
s
a
m
pl
ed ar
e
f
r
om
l
ar
ge f
l
ow
s
.
W
hi
l
e s
a
m
pl
i
ng
r
at
e i
s
c
l
os
e t
o
1,
a
l
l
f
l
o
w
i
nf
or
m
at
i
on c
an b
e ob
t
ai
ne
d i
n t
he
or
y
.
T
he
f
al
s
e
pos
i
t
i
v
e
er
r
or
of
m
ul
t
i
s
t
age
B
l
oom
f
i
l
t
er
s
dec
r
eas
es
as
t
he
nu
m
ber
of
s
t
ages
i
nc
r
eas
e.
F
i
gur
e 6 s
ho
w
s
t
he r
el
at
i
o
n of
t
he s
t
ages
num
ber
o
f
m
ul
t
i
s
t
age B
l
oom
f
i
l
t
er
s
(
MB
F
)
and
m
ax
i
m
u
m
o
f
f
al
s
e
pos
i
t
i
v
e
er
r
or
.
W
hi
l
e
s
t
a
ges
n
u
m
ber
equal
s
t
o
1,
m
ul
t
i
s
t
ag
e
B
l
oom
f
i
l
t
er
s
t
ur
ns
t
o a s
i
ngl
e s
t
ag
e B
l
oom
f
i
l
t
er
.
T
he
f
al
s
e pos
i
t
i
v
e r
at
e dec
r
eas
es
ex
pone
nt
i
a
l
l
y
w
i
t
h t
he
i
nc
r
eas
e
of
t
he s
t
ag
es
.
0
1
2
3
4
5
6
7
8
9
10
0.
1
0.
2
0.
3
0.
4
0.
5
0.
6
0.
7
0.
8
0.
9
1
t
im
e
(
m
s
)
s
am
pl
i
ng r
at
e
f
l
ow
f
ai
r
s
am
pl
i
ng
r
andom
s
am
pl
i
ng w
i
t
h s
am
pl
i
ng r
at
e=
0.
3
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
1
6
9
3
-
6
930
T
E
L
KO
M
NI
K
A
V
o
l.
1
4
,
N
o
.
3
,
S
ept
em
ber
201
6
:
11
42
–
1
149
1148
F
i
gur
e 5.
F
l
o
w
r
at
i
o s
am
pl
e
d c
om
par
i
s
on
F
i
gur
e 6.
R
el
a
t
i
o
n b
et
w
e
en
s
t
ages
of
m
ul
t
i
s
t
age
B
l
o
om
f
i
l
t
er
s
(
MB
F
)
and
m
ax
i
m
u
m
of
f
al
s
e pos
i
t
i
v
e
er
r
or
6
.
C
o
n
c
l
u
s
i
o
n
T
hi
s
paper
pr
opos
es
a f
ai
r
f
l
ow
s
am
pl
i
ng
al
gor
i
t
hm
bas
ed o
n m
ul
t
i
s
t
ag
e
B
l
oom
f
i
l
t
er
s
.
T
he t
ot
a
l
m
eas
ur
em
ent
i
nt
er
v
a
l
i
s
di
v
i
ded
i
nt
o n
c
hi
l
d t
i
m
e i
nt
er
v
a
l
s
.
D
ur
i
ng
t
h
e s
am
e t
i
m
e
i
nt
er
v
a
l
,
t
he s
am
pl
i
n
g r
at
e f
or
c
er
t
ai
n f
l
o
w
k
eep
s
t
at
i
c
.
T
he s
am
pl
i
ng
r
at
e
i
s
i
nv
er
s
el
y
pr
opor
t
i
on
al
t
o t
h
e f
l
ow
’
s
es
t
i
m
at
e t
r
af
f
i
c
up t
o t
he
pr
ev
i
ous
c
hi
l
d t
i
m
e i
nt
er
v
a
l
.
I
t
’s
s
im
p
le
t
o
ev
a
l
u
at
e t
he t
o
t
al
t
r
af
f
i
c
a
c
c
or
di
ng t
o s
am
pl
ed
pac
k
et
s
.
T
he num
ber
of
s
t
ages
of
m
ul
t
i
s
t
age
B
l
o
om
f
i
l
t
er
s
i
s
adj
us
t
abl
e
ac
c
or
di
ng
t
o
t
he
r
es
our
c
es
and
t
he
f
al
s
e
pos
i
t
i
v
e
er
r
o
r
de
m
and.
T
he
r
es
ul
t
s
s
ho
w
t
hat
t
h
e
pr
op
os
ed
al
gor
i
t
hm
c
an
s
a
m
pl
e
m
or
e
s
hor
t
f
l
ow
s
.
A
n
d
t
he
al
gor
i
t
hm
c
an
be a
ppl
i
e
d i
n m
an
y
ap
pl
i
c
at
i
ons
s
uc
h
as
at
t
ac
k
s
det
ec
t
i
on,
t
r
af
f
i
c
engi
ne
er
i
ng
.
R
ef
er
en
ces
[1
]
Z
hao Y
,
X
i
e
X
,
J
i
ang M
.
H
i
er
ar
c
h
i
c
al
R
eal
-
t
i
m
e N
et
w
or
k
T
r
af
f
i
c
C
l
as
s
i
f
i
c
at
i
on B
a
s
ed
on E
C
O
C
.
T
EL
KO
M
N
I
KA
I
ndone
s
i
a
n
J
ou
r
nal
o
f
E
l
e
c
t
r
i
c
al
E
ngi
neer
i
ng
.
2014
;
12(
2)
:
15
51
-
15
60.
[2
]
Z
hou
D
,
Li
u
W
,
Z
h
ou
W
,
et
a
l
.
R
es
ear
c
h
on
T
r
af
f
i
c
I
dent
i
f
i
c
at
i
on
B
a
s
e
d
o
n
M
ul
t
i
Lay
er
P
er
c
ep
t
r
on
.
T
el
k
om
ni
k
a
.
20
14
;
12(
1)
:
201
-
208.
[3
]
F
ang
W
,
P
et
er
s
on L
.
In
te
r
-
AS T
ra
f
f
i
c
Pa
t
t
e
rn
s
a
n
d
T
h
e
i
r
I
m
p
li
c
at
i
o
ns
.
P
r
oc
e
edi
ng
s
of
I
E
E
E
G
L
O
B
E
C
O
M
.
R
io
,
Bra
z
i
l
.
199
9
:
185
9
-
18
68.
[4
]
Z
hang
Z
,
W
a
n
g
B
,
Lan
J
.
I
d
e
nt
i
f
y
i
ng
e
l
eph
ant
f
l
ow
s
i
n
i
nt
er
net
bac
k
bo
ne
t
r
af
f
i
c
w
i
t
h
bl
oo
m
f
i
l
t
er
s
an
d
L
RU.
C
om
put
er
C
om
m
uni
c
at
i
o
ns
.
2
015
;
61:
70
-
78
.
[5
]
Z
hou
A
P
,
C
he
ng
G
,
G
uo
X
J
,
et
al
.
P
ar
al
l
el
de
t
ec
t
i
o
n
al
g
or
i
t
h
m
on
l
ong
d
ur
at
i
on
d
at
a
s
t
r
ea
m
i
n
g
.
J
our
n
al
o
f
C
om
m
uni
c
at
i
o
ns
.
2
015;
3
6(
11)
:
1
56
-
166
.
[6
]
P
ont
a
r
e
l
l
i
S
,
R
ev
i
r
i
ego P
,
M
aes
t
r
o
J
A
.
E
f
f
i
c
i
ent
F
l
ow
S
am
p
l
i
n
g
W
i
t
h B
a
c
k
-
A
n
not
at
ed C
u
c
k
o
o H
as
h
i
ng
.
I
E
E
E
C
om
m
uni
c
at
i
on
s
Le
t
t
er
s
.
2014;
18(
10)
:
1
695
-
1698
.
[7
]
B
ar
t
os
K
,
R
e
hak
M
.
I
F
S
:
I
n
t
el
l
i
gent
f
l
ow
s
am
pl
i
n
g f
or
n
et
w
or
k
s
ec
ur
i
t
y
–
an
ada
pt
i
v
e
appr
oa
c
h
.
I
nt
er
na
t
i
o
nal
J
our
nal
of
N
et
w
o
r
k
M
anagem
ent
.
2
015
;
25(
5)
:
26
3
-
282
.
[8
]
J
os
eph N
.
U
s
i
n
g S
ub
-
o
pt
i
m
al
K
al
m
an
F
i
l
t
er
i
ng f
or
A
n
om
al
y
D
et
ec
t
i
on i
n N
et
w
or
k
s
.
P
r
o
c
eed
i
ng o
f
I
nt
er
na
t
i
o
nal
C
onf
e
r
en
c
e
on E
l
ec
t
r
i
c
al
E
ng
i
ne
er
i
n
g,
C
o
m
pu
t
e
r
S
c
i
en
c
e a
nd
I
nf
or
m
at
i
c
s
(
E
E
C
S
I
201
4)
.
Y
ogy
ak
ar
t
a,
I
n
done
s
i
a.
20
14;
1:
408
-
414
.
[9
]
Z
hou A
P
,
C
heng G
,
G
uo X
J
.
H
i
gh
-
S
pe
ed net
w
or
k
t
r
af
f
i
c
m
ea
s
ur
em
e
nt
m
et
hod.
J
our
n
al
of
S
of
t
w
ar
e
.
2014;
25(
1)
:
1
35
-
153
.
[1
0
]
W
a
n
g
Y
X
,
C
hen
S
Q
,
Z
h
ang
Z
.
A
f
a
i
r
s
am
pl
i
ng
a
l
gor
i
t
hm
of
net
w
or
k
f
l
ow
bas
ed
o
n
d
y
n
am
i
c
c
ou
n
t
fi
l
te
r
.
I
E
E
E
4t
h
I
nt
er
nat
i
ona
l
C
onf
er
en
c
e
on S
of
t
w
ar
e E
ngi
neer
i
n
g
a
n
d
Se
rv
i
c
e
Sc
i
e
n
c
e
(I
C
SESS)
.
2013:
811
-
815.
[1
1
]
Yi
P
,
Q
i
a
n
K,
H
ua
ng
W
W
.
A
dapt
i
v
e
f
l
ow
s
am
pl
i
ng
a
l
gor
i
t
h
m
ba
s
ed
on
s
a
m
pl
e
d
p
ac
k
et
s
and
f
or
c
e
s
am
pl
i
ng t
hr
e
s
h
ol
d S
t
ow
ar
ds
anom
al
y
det
e
c
t
i
on.
J
our
n
al
of
E
l
ec
t
r
o
ni
c
s
&
I
nf
or
m
at
i
o
n T
e
c
hanol
og
y
.
2015;
37(
7)
:
16
06
-
16
11.
[1
2
]
D
u
ffi
e
l
d
N
.
F
ai
r
S
am
pl
i
ng
A
c
r
os
s
N
et
w
or
k
F
l
ow
M
eas
ur
em
ent
s
.
SI
G
M
ET
R
I
C
S’
1
2
.
Lond
o
n,
E
ngl
a
nd
,
U
K.
AC
M
.
2012:
367
-
3
78.
[1
3
]
Ku
m
a
r A,
X
u
J
.
S
k
et
c
h g
ui
d
ed s
am
pl
i
ng
-
U
s
i
ng on
-
l
i
ne e
s
t
i
m
at
e
s
of
f
l
ow
s
i
z
e f
or
adapt
i
v
e da
t
a
c
ol
l
ec
t
i
on
.
I
n
Pr
o
c
.
o
f
I
EEE
I
nf
oc
o
m
20
06.
W
a
s
h
i
ng
t
on
.
2
0
06:
1
-
11.
0
0.
1
0.
2
0.
3
0.
4
0.
5
0.
6
0.
7
0.
8
0.
9
1
0
0.
1
0.
2
0.
3
0.
4
0.
5
0.
6
0.
7
0.
8
0.
9
1
pac
k
et
s
am
pl
i
ng r
at
e
f
l
ow
s
r
at
i
o s
am
pl
ed
f
l
ow
f
ai
r
s
am
pl
i
ng
r
andom
s
am
pl
i
ng
1
2
3
4
5
6
7
8
9
0
0.
1
0.
2
0.
3
0.
4
0.
5
0.
6
0.
7
s
t
ages
of
M
B
F
m
ax
i
m
um
of
f
al
s
e pos
i
t
i
v
e er
r
or
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
K
A
I
S
S
N
:
1
693
-
6
930
F
l
ow
F
a
i
r
S
am
pl
i
ng
B
as
e
d
on
Mul
t
i
s
t
ag
e B
l
o
om F
i
l
t
er
s
(
Li
u Y
u
an
z
he
n
)
1149
[1
4
]
Z
hang
J
,
W
u
J
X
, N
i
u
X
N
. S
p
a
c
e
-
E
f
f
i
c
i
ent
f
ai
r
p
ac
k
et
s
am
p
l
i
ng al
gor
i
t
hm
.
J
o
ur
nal
of
s
of
t
w
ar
e
.
2
01
0;
21(
10)
:
264
2
-
26
55.
[1
5
]
B
ur
t
on
B
.
S
pac
e/
t
i
m
e
t
r
ade
of
f
s
i
n
has
h
c
odi
ng
w
i
t
h
al
l
ow
abl
e
er
r
or
s
.
C
om
m
uni
c
at
i
on
s
A
C
M
.
1970
;
13(
7)
:
422
-
4
26.
[1
6
]
P
ont
a
r
e
lli
S
,
R
e
v
ir
ie
g
o
P
,
M
ae
s
t
r
o
J
A
.
I
m
pr
ov
i
n
g
c
ou
nt
i
n
g
B
l
oom
f
i
l
t
er
per
f
or
m
anc
e
w
i
t
h
f
i
nger
pr
i
nt
s
.
I
nf
or
m
at
i
on P
r
o
c
e
s
s
i
ng L
et
t
er
s
.
2015
;
11
6(
4)
:
304
-
3
09.
[1
7
]
J
A
gui
l
ar
-
Sa
b
o
ri
t
,
P
T
r
anc
os
o,
V
M
unt
es
-
M
ul
er
o.
D
y
nam
i
c
C
ount
F
i
l
t
er
s
.
A
CM
S
I
G
M
O
D
Re
c
o
r
d
.
2006;
35(
1)
:
26
-
32.
[1
8
]
Bro
d
e
r
A,
M
i
t
z
e
n
m
ac
h
er
M
.
N
et
w
or
k
A
ppl
i
c
at
i
ons
of
B
l
o
o
m
F
i
l
t
er
s
:
A
S
ur
v
ey
.
I
nt
er
net
M
at
h
.
2003
;
(4
):
4
85
-
509
.
[1
9
]
Li
u
W
,
Q
u
W
,
G
o
ng
J
,
et
al
.
D
et
ec
t
i
on
of
S
u
per
p
oi
nt
s
U
s
i
ng
a
V
ec
t
or
B
l
o
om
F
i
l
t
er
.
I
E
E
E
T
r
a
n
s
ac
t
ions
on I
nf
or
m
at
i
o
n F
or
e
ns
i
c
s
&
S
e
c
ur
i
t
y
.
2
015;
11(
3)
:
51
4
-
52
7.
Evaluation Warning : The document was created with Spire.PDF for Python.