TELKOM
NIKA
, Vol. 11, No. 10, Octobe
r 2013, pp. 6
068 ~ 6
073
ISSN: 2302-4
046
6068
Re
cei
v
ed Ma
rch 2
4
, 2013;
Re
vised July
12, 2013; Accepted July 2
3
,
2013
An Architecture to Implement Event-Driven Web
Monitoring Systems
Gao Ying*
1
, Mu Lei
2
, Hao
Zhonghu
3
, Zheng Weiy
a
n
g
4
1
School of Co
mputer Scie
nc
e and En
gi
neer
ing, Sout
h C
h
i
na Un
iversit
y
o
f
T
e
chnolo
g
y
G
uan
gzh
ou,
510
00
6, Chin
a
2
Shang
do
ng Pr
ovinci
al C
o
mm
unic
a
tions Pl
an
nin
g
an
d Desi
g
n
Institute Jina
n, 2500
31, Ch
i
n
a
3
School of Co
mputer Scie
nc
e and En
gi
neer
ing, Sout
h C
h
i
na Un
iversit
y
o
f
T
e
chnolo
g
y
G
uan
gzh
ou,
510
00
6, Chin
a
4
Information C
enter R&D D
e
p
a
rtment, Chin
a
Souther
n Air
l
i
nes Com
pan
y
Limite
d Guan
g
z
hou, 51
04
06,
Chin
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: gao
yi
ng@sc
ut.edu.cn, 13
9
531
96
967
@16
3
.com,
haoz
ho
ngh
ud
e
@
16
3.com, zheng
w
y
@
csa
ir.com
A
b
st
r
a
ct
Traditional m
o
nitoring
system
s
are bas
ed on C/S
mode beca
use
desktop software still has
inh
e
rent
adv
a
n
tages
tho
ugh
deskto
p
soft
w
a
re is re
pl
ac
ed
by w
eb
ap
plicati
ons
rap
i
dly i
n
ma
ny fi
elds.
Meanw
hi
le, tra
d
itio
nal
mo
nito
ring syste
m
s
use rel
a
tio
nal
datab
ase as
d
a
ta source, h
o
w
ever, relatio
n
a
l
datab
ase
lacks
the ab
ility to
process
influx
of quer
ies
per
secon
d
. In this
pap
er, w
e
w
ill
do a
n
i
n
-de
p
t
h
researc
h
to design an archit
ecture to im
plem
ent event-dr
i
ven web monit
o
ring system
s.
Ke
y
w
ords
: ev
ent-driv
en, Co
met, bay
eux, Esper
Copy
right
©
2013 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
The mo
st i
m
porta
nt re
q
u
irem
ent of
moni
tori
ng
system i
s
p
r
ocessin
g
ev
ents
(or
messag
es) in
real-time o
r
near re
al-ti
m
e. Key
conside
r
ation
s
fo
r monitori
ng
appli
c
ation
s
are
latency, throu
ghput an
d co
mplexity of th
e logic requi
red [1].
Low late
ncy-appli
c
ation
s
that rea
c
t in real
-t
ime to con
d
ition
s
that occur (f
ro
m a few
milliseconds t
o
a few seconds)
High
through
put-ap
p
licatio
ns th
at p
r
o
c
e
s
s la
rge
volu
mes of m
e
ssage
s
(bet
wee
n
10
00
to
1000
0 messa
ges p
e
r seco
nd)
Compl
e
x co
mputation
s
-a
pplication
s
th
at det
e
c
t pat
terns am
ong
events, filter events,
aggregate
ti
me o
r
l
ength
win
d
o
w
s of
events,
join
e
v
ent stream
s,
trigger ba
se
d on absen
ce o
f
events etc.
To implem
en
t a real-time
or nea
r re
al-time web
monitori
ng systems, the
s
e three
requi
rem
ents
must be met. Con
s
e
quently
, two proble
m
s app
ear:
1.1. Ho
w
to
Push Ne
w
D
a
ta fr
om Serv
ers to Bro
w
sers Instantl
y
?
It’s easy for d
e
sktop software to meet the re
q
u
ire
m
e
n
t of low latency be
cau
s
e
softwa
r
e
has an i
nhe
rent a
d
vanta
ge in
pu
shi
ng in
stant
mess
ages to c
lients
but diffic
u
lt for web
appli
c
ation
s
,
becau
se tradi
tional web
ap
plicatio
ns
are
re
stri
cted to
Req
u
e
s
t/Re
s
pon
se p
a
ttern
.
This limitatio
n of HT
TP protocol
l
ead
s
serve
r
s not b
e
able to
pu
sh data to
clie
nts. That’
s
why
most monito
ri
ng syste
m
s a
r
e ba
se
d on
C/S mode th
ough
web a
p
p
licatio
ns a
r
e
taking pla
c
e
of
desktop software
rapidly
in recent
years.
1.2. Ho
w
to
Process
L
a
rge Volum
es of
Mes
s
a
ges
and P
e
rform
Com
p
lex Co
nditi
o
n
Chec
king Efficiently
?
Usually, appl
ication
s
u
s
e
relational d
a
taba
se a
s
data so
urce.
However, relational
databa
se a
n
d
SQL are
de
signed fo
r app
lication
s
in
which m
o
st d
a
ta is fairly stat
ic and
co
mpl
e
x
queri
e
s
are l
e
ss fre
quent.
To ret
r
ieve d
a
ta from
a
da
tabase, an a
pplicati
on mu
st issue
a qu
ery.
Always, usi
n
g relation
al databa
se in monitorin
g
syst
ems, you ne
ed to periodi
cally poll for state
cha
nge
s. Thi
s
po
or d
e
si
g
n
puts
a lot
of strain
on b
o
th we
b serv
er an
d data
b
a
se,
co
sting
you
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 10, Octobe
r 2013 : 606
8 –
6073
6069
CPU
cycl
es,
band
width, a
nd a d
e
cre
a
se in pe
rforma
nce.
Datab
a
se trigg
e
rs ca
n be u
s
e
d
to
fire
in re
spo
n
se to datab
ase u
pdate eve
n
ts
but they
tend
to be sl
ow
an
d often cann
o
t
easily pe
rform
compl
e
x
con
d
ition
che
c
ki
ng a
nd i
m
ple
m
ent lo
gic to
re
act [2]. So
rel
a
tional
da
tabase d
o
e
s
not
apply to moni
toring sy
stem
in which influ
x
of data need be pro
c
e
ssed in a sh
ort perio
d.
This
pap
er i
s
going
to giv
e
an i
n
tro
d
u
c
tion to ho
w t
o
solve the
s
e
pro
b
lem
s
a
n
d
then
de
sig
n
an
architectu
re t
o
impleme
n
t event
-d
riven
web mo
nitori
ng syste
m
s.
2. Solutions and Primar
y
Architec
ture
2.1. COMET
To solve p
r
ob
lem n INTRO
D
UCTYIO
N, two metho
d
s
are wi
dely ap
plied [3]:
With the
h
e
lp
of pl
ug-i
n
s (su
c
h
as flash
,
applet
or ActiveX) install
e
d
in
browse
rs, se
rvers
ca
n
push ne
w dat
a instantly by so
ckets [4] [5].
Client
bro
w
sers send
Aja
x
[6] [7] requ
est to
se
rver
at reg
u
la
r int
e
rvals to
che
c
k whethe
r t
h
e
serve
r
ha
s go
t new data [8].
Ho
wever, th
ese t
w
o m
e
thods
have
their o
w
n
p
r
oble
m
s. T
h
e first m
e
th
od ha
s
comp
atibility issue be
cau
s
e not all the browse
rs
su
pport the plu
g
-in
s
. The b
e
st co
ndition
to
reali
z
e the
seco
nd meth
o
d
is the p
o
lli
ng interval
s
are e
qual to
the se
rver’
s
update inte
rv
als
,
otherwise, it not only ca
n’t meet the
real
-tim
e re
quire
ment
s b
u
t also
re
sul
t
s in waste
of
r
e
sour
ces
.
Re
cently, a web ap
pl
ication mo
d
e
l named
“Com
et”
in whi
c
h
a long-
lived
HTTP
conne
ction all
o
ws we
b se
rvers to pu
sh
data to bro
w
sers witho
u
t plug-in
s is
more
popul
ar. Spe
c
ific m
e
thod
s of implem
ent
ing Comet fal
l
into two
maj
o
r
categ
o
rie
s
:
stre
aming
a
n
d
long pollin
g [9].
2.2. Ev
ent-Dr
i
v
e
n Archite
c
ture
In a world (such a
s
monit
o
ring
system
) f
illed with events and in
stant com
m
u
n
icatio
n,
traditional
da
tabase i
s
n
o
t
approp
riate
.
Con
s
e
que
n
t
ly, event-dri
ven ar
chite
c
t
u
re (EDA
) was
prop
osed [1
0
]. In an eve
n
t
-driven
archi
t
ecture,
an
i
m
porta
nt thin
g ha
ppe
ns in
side
or out
si
de
your b
u
si
ne
ss, which di
sseminate
s
im
mediately to
all intereste
d
partie
s
. Th
e i
n
tere
sted
pa
rties
evaluate the
event, and
optionally t
a
ke
actio
n
. The event
-driven actio
n
may inclu
d
e
the
invocatio
n
of
a
se
rvice, t
he tri
gge
ring
of a
bu
sin
e
s
s p
r
o
c
e
ss.
A cla
s
sical E
D
A
works li
ke a
databa
se tu
rned u
p
si
de-d
o
wn. In
stead
of stori
ng t
he data
and
run
n
ing
qu
erie
s, it allo
ws
appli
c
ation
s
to store que
ri
es an
d ru
n the data
thro
ugh. With E
D
A, appli
c
ations
can tri
g
ger
action
s when
event con
d
itions
o
c
cu
r am
ong event
streams, p
e
rf
o
r
m compl
e
x condition
che
c
king
and impl
em
ent logi
c to
rea
c
t qui
ckly. So, EDA can b
e
use
d
to sol
v
e Problem
in
INTRODUCT
I
ON.
2.3. Primar
y
Architec
ture
On the
ba
sis
of Com
e
t and
EDA, a
syst
em
archite
c
tu
re to im
plem
ent event-driv
en web
monitori
ng sy
stem
s ca
n be
desig
ned p
r
i
m
arily as Fig
u
re 1.
Figure 1. Primary archite
c
ture
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
An Archite
c
tu
re to Im
plem
e
n
t Event-Dri
ven We
b Moni
toring System
s (G
ao Ying)
6070
This p
r
ima
r
y architectu
re
can be divide
d
into three pa
rts:
The monito
rin
g
appli
c
ation
s
for captu
r
ing
event strea
m
s.
The event-dri
ven archite
c
ture for filte
r
in
g,
analyzin
g
and p
r
o
c
e
ssi
ng the
event
strea
m
s from
the monitori
n
g
appli
c
ation
s
The we
b app
lication
s
for showi
ng re
al-ti
m
e inform
ati
on to use
r
s. This pa
rt incl
ude
s Com
e
t in
serve
r
an
d Comet in client
browse
r.
3. An Actual
Application and Detailed Solution
We h
a
ve al
re
ady implem
e
n
ted a
we
b m
onitorin
g
syst
em
by
the architecture said
above.
In this section, we will gi
ve a detailed solu
tion
by descri
b
ing
how to realize the monitori
ng
system. Firstl
y, we give the list of tools, method
s, platforms o
r
protocol
s we use
d
as follo
wing
:
MyEc
lipse6.5 with J
D
K1.6
Tomcat 6 (Co
m
et-Fri
endly
serve
r
)
Oracle (Yes.
Relational dat
abase is
st
ill used. We will
give the reason later)
Bay
eux
(Prot
o
col for
Com
e
t)
Espe
r (a light
weig
ht, embe
ddabl
e open
sou
r
ce imple
m
entation of EDA)
Some of the
m
are familia
r whil
e othe
rs are not. Secon
d
ly, we in
trodu
ce Baye
ux and
Es
per.
3.1. Ba
y
e
ux
Protocol
The Bayeux
Protocol defines
usi
n
g long pollin
g approach [11].
HTT
P
requests sent from
the client are hold on
on the
se
rver till the requests
can
be fully
answere
d. This
protocol
provide
s
a flexible, sal
e
a
b
le API ba
se
d aroun
d a p
ublish/su
bscri
be mo
del. In
our
monito
ri
ng
system,
we
make
serve
r
to be p
ubli
s
h
e
r a
nd
client
to be
su
bscribe
r
. Me
ssa
ges
en
cod
e
d
as
JSO
N
obje
c
ts are ro
uted via cha
nnel
s.
Server sid
e
impleme
n
tations ne
ed to
use n
o
n
-
bl
ocking
IO
(NIO) to g
u
a
r
antee
an
efficient im
plementation.
So we
cho
o
se tom
c
at
6
(
can b
e
set to
sup
port
NIO
)
as
serve
r
. S
o
me
Java librarie
s
need to be importe
d. The
y
are “C
o
m
et
d-api.ja
r”, “Comet
d-bayeu
x.jar” and “j
etty-
until.jar”. T
h
e
n
the
web
se
rver
can
pu
bl
ish topi
cs. F
o
r exam
ple, t
he follo
wing
cod
e
pu
sh
es a
JSO
N
obje
c
t named
”total” via channel
“/Mana
ge/cal” to Browsers:
ServletContext application=
this
.getServletContext();
Bayeux b=
(B
ayeux)application.getAttr
ibute(Bayeux.DOJ
OX_COMETD_BAYE
UX);
Cha
nnel c=b.getCh
ann
el(“/
Manag
e/cal
”
,true
);
// ……
c.publi
s
h
(
u
s
e
r
Id,total,”total”);
To impleme
n
t
Comet at client side, we us
e a Jav
a
Scri
pt library-“doj
o”. The
n
client
bro
w
sers
can
sub
s
crib
e in
tereste
d
topi
cs from th
e chann
els. Fo
r
example, the
followin
g
co
de
receives the
messag
e the serve
r
pu
sh
e
d
:
<script src=”d
ojo-rele
ase
-
1.
1.2\dojo\doj
o.js”>
djCo
nfig=”isDebug:fal
s
e,pa
rse
O
nL
oad:true”>
</
s
c
ript
>
// ……
dojo.re
qui
re(“dojox.com
etd
”);
dojox.com
etd
.
init(“cometd
”
);
dojox.com
etd
.
sub
s
cribe
(“/
Manag
e/ca
l
”
,
c
urre
ntPro
c
e
s
sor,”total
”);
3.2. Esper
Espe
r i
s
a
Java library th
at ca
n b
e
i
n
tegrate
d
into
multiple a
ppli
c
ation
s
devel
oped
by
Java. In Esp
e
r, event can
be map, PO
JO a
nd XML.
Espe
r offers an event p
a
ttern lang
uag
e
(EPL) to spe
c
ify expre
ssi
on-b
a
sed ev
ent pattern
matchin
g
. Th
is metho
d
of event pro
c
e
ssi
ng
matche
s exp
e
cted
seque
n
c
e
s
of pre
s
e
n
c
e or ab
se
n
c
e of event
s o
r
com
b
i
natio
n
s
of
events. E
P
L
can al
so be
use
d
to provide the wind
o
w
s, agg
re
gati
on, joining a
nd analysi
s
functio
n
s for
use
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 10, Octobe
r 2013 : 606
8 –
6073
6071
with
stre
ams
of events.
EPL ha
s be
en
d
e
sig
ned
for si
milarity with
t
he SQ
L q
u
e
r
y langu
age.
F
o
r
example, an
event (POJO) is defined a
s
:
publi
c
cla
s
s CardEvent implement
s Seri
aliza
b
le{
private String
from;
private String
to;
private Do
ubl
e amount;
}
Then, the followin
g
statem
ent can b
e
used to do som
e
relevant sta
t
istics:
sele
ct
c
o
unt(*) as
noOfCard,
s
um(amount) as
total,
avg(am
ount) as
average
from CardEv
ent.win:time(15 s
e
c
)
;
With the
hel
p of EPL, E
s
per implem
e
n
ts a
pe
rfe
c
t EDA.
When
larg
e influx
of event
strea
m
s
com
e
to Esper, T
he EPL state
m
ents p
r
ev
io
usly define
d
judge
whethe
r the event stream
matche
s th
e
EPL. This m
e
thod
of eve
n
t pro
c
e
s
si
ng
match
e
s exp
e
cted
sequ
en
ce
s of
pre
s
e
n
ce
or ab
sen
c
e
of events or
combi
nation
s
of ev
ents. Lastly, Esper
outputs PO
JOs. In additi
on,
Espe
r
can
al
so a
c
cess with
the
histo
r
ical
datab
ase, a
databa
se
em
bedd
ed i
s
used to
en
su
re t
he
con
s
i
s
ten
c
y of the data
with the extern
al datab
ase.
If nece
s
sary,
Espe
r
can
combine
the d
a
ta
strea
m
and hi
stori
c
al data t
o
do com
p
lex
processin
g
(That’s
why we use
d
Ora
c
l
e
).
3.3. Detailed
Solution
On the ba
sis
of Bayeux and Espe
r, A detail
ed sol
u
tion
can be d
e
si
g
ned a
s
Figu
re
2.
Figure 2. Det
a
iled solution
The fun
c
tions of compon
en
ts in the whol
e system:
Monitori
ng a
pplication
s
(such
as sca
n
n
i
ng eq
uipme
n
t
) are
compo
nents fo
r
cap
t
uring
events.
Adapters: Th
ey are respo
n
sibl
e for de
fining EPL st
atements to
filter, analyze
and
pro
c
e
ss the d
a
ta delivere
d
from monito
ri
ng appli
c
atio
ns.
Espe
r Li
sten
er: Th
ey are
desi
gne
d to
be a
tta
che
d
with the
stat
ement in
“Ad
apters”,
whe
n
the sta
t
ement evalu
a
tes to true,
“E
sp
er Li
sten
er”
will be tri
ggered an
d
output
P
O
JO
s.
Acce
sslaye
r: It is for comm
unicating with
the histori
c
al
databa
se.
Comet Li
sten
er: This
com
pone
nt is for listening to Espe
r com
p
o
nents a
nd re
ceiving
POJOS from
the Espe
r co
mpone
nts.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
An Archite
c
tu
re to Im
plem
e
n
t Event-Dri
ven We
b Moni
toring System
s (G
ao Ying)
6072
Comet Handl
e: A
thread
pool that con
t
ains several
thread
s that is named “Come
t
Han
d
le Th
re
ad”. Th
ese thread
s
re
ceive
and ha
ndle
asyn
chrono
u
s
re
que
sts from the
client browse
rs.
Web a
ppli
c
ati
on is for di
spl
a
ying data on
browse
rs.
The m
o
st i
m
portant i
nnov
ative point
of this
archite
c
t
u
re i
s
co
mbin
g EDA
and
Comet to
impleme
n
t event-d
riven we
b monitori
ng
system
s.
4. Architectu
r
e Analy
s
is
4.1. Adv
a
ntages
Good p
e
rfo
r
mance
We di
d expe
riments to te
st our mo
nitori
ng sy
stem. T
w
o time
gap
s we
re record
ed: the
time gap betwee
n
monitoring appli
c
atio
n and the ev
ents we
re p
r
oce
s
sed ove
r
by Esper (si
m
ply
recorded a
s
“process eve
n
t time gap”), the time gap betwee
n
Espe
r and di
splaying data
to
use
r
s(record
ed a
s
“rea
ct time gap”).
The expe
rim
ents h
a
ve b
een d
one fo
r 1000 time
s.
We
sele
cted rand
omly 10 re
cords an
d sh
ow
them in Table
1.
Table 1. 10 Reco
rd
s of 100
0 Times
No.
Process event time gap
React time gap
Time gap
1 25ms
119ms
144ms
2 17ms
121ms
138ms
3 20ms
110ms
130ms
4 15ms
107ms
122ms
5 17ms
119ms
136ms
6 19ms
105ms
124ms
7 17ms
104ms
121ms
8 24ms
110ms
134ms
9 25ms
103ms
128ms
10 27ms
117ms
144ms
Average
20.6ms
111.5ms
132.1ms
The average
time gap of the 10 record
s is
13
2.1m
s, and the ave
r
age time g
a
p
of the
1000 time
s is 131.5ms
whi
l
e 50 EPL sta
t
ements a
r
e
i
n
clu
ded in ou
r system a
nd
10 of them are
compl
e
x que
ries. It mean
s that the architecture we
desi
gne
d can
pro
c
e
ss
eve
n
ts an
d di
spl
a
y
new
data to
client
s in
sta
n
tly. Good p
e
rform
a
n
c
e i
s
the m
o
st i
m
porta
nt adv
antage
of this
architectu
re.
Good
comp
atibility
Espe
r i
s
light
weig
ht. It’s e
a
sy to
be i
n
tegrate
d
into
the exi
s
ting a
pplication
s
d
e
velope
d
by Java. Bay
eux ca
n be
u
s
ed to
any browsers
sup
p
o
rted Ajax, a
nd many
se
rvlet-su
ppo
rting
and
NIO-su
ppo
rting se
rvers.
No firewall res
t
ric
t
ions
The archite
c
t
u
re i
s
ba
sed
on stan
da
rd
H
TTP
po
rts (80
and 443
) and
HTTP protocol. So,
it can wo
rk b
ehind the fire
wall
s.
4.2. Problem and Analy
s
is
Scalability Problem
In this archite
c
ture, the we
b serve
r
nee
ds to kee
p
a long co
nne
cti
on with ea
ch client to
achi
eve serv
er-pu
s
h. Whe
n
the nu
mbe
r
of con
c
u
r
ren
t
clients i
s
sm
all, this archit
ecture
can
work
norm
a
lly. Ho
wever,
when
the numbe
r is very big,
it’s hard to load. In this
scena
rio, loa
d
ing
balan
ce a
nd
clu
s
ter te
chn
o
logy ca
n be
con
s
id
ere
d
.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 10, Octobe
r 2013 : 606
8 –
6073
6073
5. Summarize and Prosp
ects
This archite
c
t
u
re
uses Co
met and
EDA to
imple
m
e
n
t an
event-d
riven
web
m
onitorin
g
system. In fact, it can
be appli
ed to many
we
b
application
s
of high rea
l
-time dem
a
nd.
Con
s
e
quently
, the archite
c
t
u
re introdu
ce
d in this pap
e
r
is valuabl
e.
In write
r
s’ opi
nion, Comet i
s
o
n
ly an i
n
te
rim te
chn
o
log
y
to reali
z
e t
w
o-way
com
m
utation
between web
servers
and browsers.
HT
ML5 will
become stand
ard of all browsers i
n
the near
future. At that time, Comet in the archite
c
tu
re
we de
si
gned
can b
e
repla
c
ed by “Web
so
cket”.
Ackn
o
w
l
e
dg
ment
The Pap
e
r i
s
supp
orted
by Nation
a
l
Natural Science Fou
n
d
a
tion of Chi
na No.
5120
4071;
Scien
c
e
an
d Te
chn
o
lo
gy Plannin
g
Proj
ect
of Gu
an
gdon
g Pro
v
ince
(201
2B01
010
0019
). Mini
stry of Edu
c
ati
on Univer
sity-Indu
stry-Re
s
earch P
r
oje
c
t
of Gua
ngd
o
n
g
Province (20
10B090
400
5
35) an
d The
Funda
mental
Re
sea
r
ch Fu
nds fo
r the
Central Universities
(201
12M
006
8).
Referen
ces
[1]
Kim, Yan
g
S
o
k. Usin
g kn
o
w
led
ge
bas
e fo
r eve
n
t-drive
n
sche
dul
in
g of
w
e
b mo
nitori
ng s
y
stems
.
Lecture N
o
tes i
n
Co
mp
uter Scienc
e
. 200
9; 5692: 16
9-1
80.
[2]
Shans
ha
n Li, Jian Z
h
o
u
.
T
r
ansmitter Station Remot
e
Monitor S
y
stem
Based on Br
o
w
ser /Serve
r
Structure.
T
E
LKOMNIKA Indones
ian J
ourn
a
l of Electrica
l
Engi
neer
in
g
. 2013; 11(
3): 159
4-15
99.
[3]
Liji
ng
Z
han
g, W
e
i Xio
ng, Xu
ehu
i Xi
an.
R
e
s
earch
on
W
eb-
base
d
R
eal-tim
e Mo
nitori
ng S
y
stem
on
SVG
and C
o
met.
T
E
LKOMNIKA Indon
esia
n Jour
nal
of Electric
al
Engin
eeri
n
g
. 2
012; 10(
5): 114
2-11
46.
[4]
Pohj
a, Mikko. Server pus
h fo
r
w
e
b
ap
pl
icati
ons via i
n
stant
messagi
ng.
J
ourn
a
l of W
e
b
Engin
eeri
n
g
.
201
0; 9(3): 227
-242.
[5]
CHEN Ha
ng,
Z
H
AO F
ang. Impleme
n
tatio
n
of
w
e
b insta
n
t
message s
y
s
t
em based
on
server push
techno
lo
g
y
and
XMPP.
Co
mp
uter Engi
ne
erin
g and D
e
si
gn
. 201
0; 31(5): 83
-88.
[6]
Z
hou
Xu
an,
W
ang L
i
-fan
g, JIANG Z
e
-ju
n
. Desi
gn
an
d Implem
entat
ion
of AJA
X
-
base
d
Instant
Messag
i
ng S
y
s
t
em.
Science T
e
chno
logy
an
d Engi
ne
erin
g
. 200
9; 9(2): 446
-449.
[7
]
L
I
Xi
an
-j
un
, L
I
U
Bo
, YU D
an, MA Sh
i
-
lo
ng
. Mix
e
d C
/
S and
B/S a
r
ch
i
t
e
c
tu
re
pa
tte
rn
b
a
se
d
o
n
AJAX
.
Journ
a
l of Com
puter App l
i
cati
ons. 200
9; 19(
4): 1135-
11
38.
[8]
Yuan
yu
a
n
Li
ao
, Z
hen
yu Z
h
an
g, and
Yan
q
i
n
g
Yang. W
e
b Ap
plicati
ons B
a
se
d on
Aja
x
T
e
ch
nol
og
y
a
n
d
Its Framew
ork.
Communic
a
tio
n
s in Co
mputer
and Infor
m
ati
o
n Scienc
e
. 201
2; 288(1): 3
20-
326.
[9]
XU
Xiu-h
ua,
W
E
N Bi-lon
g, LIU Da
n-Jia
ng,
BI
Shuo-b
en. A
realiz
atio
n me
thod
of inst
ant messa
gin
g
base
d
on
w
e
b.
Co
mp
uter Engi
neer
ing a
nd D
e
sig
n
. 200
3; 24(7): 40-4
2
.
[10]
Lever
agi
ng
Ev
ent Driv
en
Arc
h
itectures
http://
onepi
xel
ahe
a
d
.com/201
0/08
/
11/lever
agi
ng-
event-dr
i
ven-
architectures/.
[11]
T
he Ba
y
e
u
x
S
pecific
ation
http://svn
.Cometd
.
com/trunk/ba
y
eu
x/ba
ye
u
x
.ht
m
l.
Evaluation Warning : The document was created with Spire.PDF for Python.