1
Design, Implemen a ion and Valida ion o a
Simula ion Tool o Ne wo ked Vi ual
En i onmen s
Juan Luis Fon , Jose Luis Se illano , Daniel Cascado-Caballe o, Gema Lopez-Mu˜
noz , Be hanu Regassa
Depa men o Compu e Technology and A chi ec u e, Uni e si y o Se ille, Spain
{juanlu, se i, danic, gemalopez, be hanu}@a c.us.es
Abs ac —The popula isa ion o Ne wo ked Vi ual En i on-
men (NVE) applica ions in se e al con ex , including social
in e ac ion and e-Heal h ields, makes hem an in e es ing subjec
o s udy. The p ope es ima ion o hei ne wo k equi emen s
a e key o ensu e a good use expe ience, an impo an ac o ha
de e mines he accep ance and success o he applica ions. The
s udies ocused on NVE wi h an impo an social componen a e
ela i ely sca ce. This pape p esen s he design, implemen a ion
and alida ion o a simula ion ool o assis in he s udy o
he equi emen s o a NVE applica ion, Open Wonde land.
This ool will acili a e he s udy o se e al pa ame e s ha
de ine he use expe ience. The s udy desc ibes he o e all
s uc u e, de elopmen amewo k and models ha de ine he
a ic gene a ion pa e ns and alida es he simula ion esul s
by compa ing i s aces wi h eal a ic om p e ious s udies.
I. INTRODUCTION
Open Wonde land (OWL) is one o he so-called Ne -
wo ked Vi ual En i onmen applica ions, which a e ocused
on p o iding a i ual wo ld expe ience based on a dis ibu ed
simula ion. The success o se e al o hese applica ion in
con ex s such as gaming o social in e ac ion ha e con ibu ed
o in oduce hem o la ge audiences. This popula isa ion
oge he wi h he p oli e a ion o he “pe suasi e sys ems”,
ocused on mo i a ing heal hy li es yle habi s [1], ha e led
o ou esea ch g oup o de elop a pe suasi e sys em called
“Vi ual Valley”, based on Open Wonde land and elying on
he i ual wo ld concep [2], [3].
OWL has se e al cha ac e is ics ha make i sui able as
basis o a esea ch pe suasi e sys em. Fi s , i is licensed
unde he GPL 2, i is also a 100% Ja a p ojec [4], which
ensu es a high deg ee o po abili y, and i s design allows
o any o ganisa ion o pa icula o deploy an ins ance in
hei own ne wo k in as uc u e. The ne wo k a ic de i ed
om Wonde land is mainly due o h ee sou ces. Fi s , objec
synch onisa ion which allows all use s o ha e a cohe en iew
o he i ual wo ld (including mo ing objec s like a a a s).
Second, messages in ended o suppo communica ions among
use s, including oice a ic ( he main sou ce o a ic) bu
also ex messages (cha ). And, inally, a ic due o he
execu ion o applica ions sha ed among di e en use s. The
la e is e y di icul o model, as i depends on he pa icula
applica ion. The e o e, his s udy will ocus on he i s wo
sou ces: objec synch onisa ion and oice a ic.
The use expe ience is a key ac o o he success o a
NVE such as OWL. Poo g aphic pe o mance, ne wo k lag
o inconsis encies ansla e in o a poo use expe ience ha
may lead he use o ejec he applica ion. Assu ing such a
good use expe ience equi es he s udy o he pa ame e s ha
ha e a di ec impac on i , iden i ying he h eshold below
which he applica ion becomes unusable. In his con ex , he
ne wo k a ic pe o mance is o g ea impo ance. A i ual
wo ld en i onmen equi es ce ain ne wo k esou ces o keep
i s clien s synch onised, sha ing a common and cohe en iew
o he same simula ed eali y. I is impo an o de e mine he
maximum h oughpu o OWL clien s and se e s o p ope ly
size he equi ed ne wo k esou ces and be able o p edic
he h oughpu e olu ion depending on he scena io. Ne wo k
simula o s play a use ul ole o s udy such equi emen s.
This pape aims o de ine, implemen and alida e a simu-
la ion ool o acili a e he s udy o he ne wo k equi emen s
o OWL clien s. The models used o implemen he a ic
gene a ion pa e ns ha e been de ined in p e ious wo k. The
pape is s uc u ed as ollows: Sec ion II lis s p e ious wo k in
his a ea, Sec ion III con ains he ne wo k modes p oposed o
OWL clien a ic in p e ious wo k. The simula o amewo k,
s uc u e and implemen a ion de ails a e lis ed in Sec ion IV
and he alida ion and esul s a e discussed in Sec ion V.
Finally, he pape closes wi h he conclusions and u u e wo k
in Sec ion VI, acknowledgmen s and bibliog aphy.
II. PREVIOUS WORK
Mos s udies abou Ne wo ked Vi ual En i onmen s (NVE)
ocus on mul i-playe online games, specially on Fi s Pe son
Shoo e s (FPS) [5]. Real-Time S a egy (RTS) [6] and Mas-
si ely Mul i-playe Online Role-Playing Games (MMORPG)
[7] a e also gen es subjec o academic esea ch, al hough in
a lesse ex en .
The quali y o he use expe ience o each gen e depends
on di e en aspec s, so a ious echniques a e deployed de-
pending o he equi emen s. Fo example, UDP anspo
p o ocol is used by hose ha gi e p e e ence o ime es ic ed
upda es o e ansmission eliabili y. On he o he hand, ap-
plica ions ha equi e a eliable da a s eam se ice ely on
TCP p o ocol. OWL ollows a hyb id app oach using TCP
connec ions o objec synch onisa ion a ic and UDP o
audio ansmission.
Simila wo ks dealing wi h equi emen s o NVEs a e
sca ce: Quali y o Se ice (QoS) is s udied in [8] and [9].
2
Bandwid h equi emen s o some popula mul i-playe online
games a e s udied in [10] moni o ing ne wo k a ic du ing
se e al sessions.
To he bes o ou knowledge, he e a e no published
s udies abou he ne wo king esou ces needed o suppo he
execu ion o Vi ual Wo lds based on Open Wonde land excep
o he p elimina y app oach made in [11] and u he s udy
pe o med in [12]. This pape ollows a mic o scale modelling
app oach simila o ha desc ibed in [13] and p oposes a se
o models o desc ibe he a ic gene a ion pa e ns o OWL
clien s.
III. NETWORK TRAFFIC MODELS FOR OWL CLIENTS
This sec ion p esen s a summa y o he modelling p ocess
esul s p esen ed in [12], a s udy o he a ic gene a ion pa -
e ns o he OWL clien s. These models ocused on modelling
he In e -A i al Time be ween consecu i e packe s and packe
size o wo o he main OWL a ic sou ces: TCP objec
synch onisa ion and UDP audio a ic. The i s is used o
synch onise i ual wo ld simula ion sha ed by all he clien s
so i s olume o a ic depends on he use ac i i y. The second
a ic is esponsible o he audio con e sa ions and so phone
calls.
On he one hand, he objec synch onisa ion p ocess gen-
e a es upda e eques s and p opaga ions depending on he
use ac i i y. Cons an mo emen ansla es in o a highe a e
o objec synch onisa ion packe s mee ing some es ic ions
o a oid an upda es o e low. Highly ac i e use s gene a e
synch onisa ion packe s ollowing a unca ed exponen ial dis-
ibu ion. The a e pa ame e λ, ha de e mines he exponen ial
beha iou , depends on he numbe o concu en use s wi hin
he session. Expe imen al esul s p o ed ha λdec eased wi h
he inc ease in concu en use s. Table I shows a e λ alues
calcula ed by Maximum Likelihood Es ima ion (MLE) o
a ious expe imen al sessions. This model is desc ibed by a
con inuous andom a iable in Equa ion 1. Inac i i y pe iods
ha e a conside able impac on he packe a e and a e o li le
in e es o he cu en s udy. The e a e ew possible packe
sizes whose p obabili ies can be desc ibed using a disc e e
andom a iable as shown in Equa ion 2.
TABLE I
RATE PARAMETER λVALUES FOR OBJECT SYNCHRONISATION IAT
Session λ
2-clien 4.168
3-clien 3.687
5-clien 3.146
F(x;λ) = 




1,x > 0.5
1−e−λx ,0.5≥x≥0
0,x < 0
(1)
P(x) =









0.9704 ,x= 239 by es
0.0195 ,x= 478 by es
0.0057 ,x= 536 by es
0.0044 ,x= 717 by es
(2)
On he o he hand, audio packe s a e gene a ed ollowing
pe iodic pa e ns ha can sligh ly a y depending on he
unde lying ope a ing sys em and ne wo k s ack, al hough he
o e all h oughpu is simila among di e en con igu a ions.
This pape ocuses on he Windows XP case, whose In e -
A i al alue beha iou is desc ibed in Equa ion 3. The as
majo i y o he audio packe s ha e a UDP payload size o
1292 by es, so his pa ame e is conside ed o be cons an .
P(X=k) =









0.1426 ,k= 0.0114 s
0.6405 ,k= 0.0221 s
0.0723 ,k= 0.0340 s
0.1446 ,k= 0.0442 s
(3)
(4)
IV. IMPLEMENTATION OF THE SIMULATOR
The de elopmen o he simula ion ool has no been un-
de aken om sc a ch. Ins ead, ns-3 disc e e-e en ne wo k
simula o has been chosen as s a ing poin . Ns-3 p o ides
a modula and well es ed simula ion amewo k, i is an
open sou ce communi y-d i en p ojec and p o ides a well
de ined building p ocess, comp ehensi e documen a ion and
he possibili y o coding in C++ o Py hon [14], [15]. Ns-
3 is de ised o be execu ed in GNU/Linux and Unix-like
en i onmen s, being also possible o build and un i in
Windows sys ems using CygWin. The e sion used in his
pape is ns-3.13, las s able elease a he ime o w i ing.
Ns-3 code oughly dis inguishes in e nally be ween models,
helpe s and simula ions. The models a e classes ha imple-
men he unc ionali y o he modelled eal sys ems, such as
ne wo k nodes, da a packe s, p o ocol s acks, ne wo k de ices
o packe gene a ing applica ions. On he o he hand, helpe s
a e auxilia y classes ha make he c ea ion and con igu a ion
o la ge amoun s o ela ed en i ies easie , such as de ini ion
o se e al in e connec ed nodes, he ini ialisa ion o la ge
amoun s o ne wo k decides wi h consecu i e IP add esses,
and so on. Finally, a simula ion is a p og am ha de ines,
con igu es and launches he simula ion expe imen based on
he models and helpe s.
The ns-3 code ha implemen s he OWL clien s is w i en
in C++. Acco ding o he ns-3 so wa e a chi ec u e, he Node
class ac s as a con aine en i y ha ep esen s a ne wo k node.
This en i y con ains objec s ha model he ne wo k de ices,
p o ocol s acks and links o o he en i ies such as physical
channels. Se e al applica ion models can be “ins alled” on op
o a node objec . These applica ions a e classes ha inhe i
om he gene ic Applica ion class. They gene a e and
consume ne wo k a ic ollowing he logic coded by he
p og amme . The simula o s uc u e is shown in Figu e 1.
3

























  

  







Fig. 1. Simula o a chi ec u e
The classes ha implemen he OWL clien -se e a chi ec-
u e a e lis ed below:
•OWLMo Clien gene a es TCP objec synch onisa ion
packe s wi h In e -A i al Times de ined by he dis ibu-
ion in Equa ion 1. The TCP payload size is de e mined
by Equa ion 2. I uses a TCP Socke objec o pass
packe s o he IP 4 s ack o i s con aining Node. I is
se up wi h he IP add ess and TCP po o i s associa ed
OWL se e .
•OWLMo Se e ecei es TCP objec synch onisa ion
packe s om se e al OWLMo Clien objec s and
dumps hem o disk in o pcap iles.
•OWLAudioClien gene a es audio packe s wi h In e -
A i al Times de e mined by he dis ibu ion in Equa ion
3. The UDP payload size is 1292 by es. I uses an UDP
Socke objec o pass packe s o he IP 4 s ack o i s
con aining Node. I is se up wi h he IP add ess and
TCP po o i s associa ed OWL se e .
•OWLAudioSe e ecei es UDP audio packe s om
se e al OWLAudioClien objec s and dumps hem o
disk in pcap iles.
The simula ed ne wo k in as uc u e is de ined and con-
igu ed in he owl_clien _sim.cc ile, loca ed in
$NS_DIR/sc a ch. The simula ion comp ises a Ca ie
Sense Mul iple Access (CSMA) bus opology ope a ing a
100 Mbps and in e connec ing an a bi a y numbe o nodes.
This opology is equi alen o he o iginal 100 Mbps E he ne
ne wo k used in he es ing sessions. All he simula ion nodes
sha e he same CSMA channel and a e equipped wi h CSMA
ne wo k de ices. Each node has also a IP 4 s ack and associ-
a ed IP add ess. All hese models a e p o ided by de aul wi h
he ns-3 simula ion and a e well es ed and documen ed.
The Applica ion child classes a e ins alled in he
simula ed nodes. In his opology he e is only a single
se e and an a bi a y numbe o clien s. Each clien node
has a OWLMo Clien and OWLAudioClien objec as-
socia ed while he only se e has a OWLMo Se e and
OWLAudioSe e ins ance ins alled.
The ns-3 building p ocess gene a es an execu able ile
owl_clien _sim, ha lacks G aphical Use In e ace
(GUI) and can be launched om command line using wa , he
build au oma ion ool used by ns-3. The simula ion execu able
accep s se e al pa ame e s ha a e lis ed below:
•nClien s: speci ies he numbe o OWL clien s pa ic-
ipa ing wi hin he simula ion. De aul alue 2.
•packe Ra e: a e pa ame e λ, de e mines he dis ibu-
ion o he IAT alues o objec synch onisa ion packe s.
Clien s wi hin sca cely popula ed sessions ha e λ alues
close o 5 while λ ends o 3 o o e c owded sessions.
De aul alue 4.168, wo concu en highly ac i e use s.
• Simula ion: se s in seconds he du a ion o he
simula ed OWL session. De aul alue 250 s.
•seed: speci ies he seed o ini ialise he andom numbe
gene a ions.
•logPa h and logName: speci y espec i ely he s o e
di ec o y and he oo name o he ou pu ace ile. The
logName s ing is appended wi h a ime s amp. De aul
alues a e “/ mp/” and “owl ace ”.
The simula o gene a es as ou pu a ne wo k ace cap u ed
in he OWL se e side, including bo h audio and objec
synch onisa ion a ic. This ace is s o ed as a ile in pcap
o ma .
V. SIMULATION RESULTS AND VALIDATION
The packe analyse Wi esha k .1.6.5 was used o s udy
he simula ed and expe imen al aces. Tsha k is he e minal-
based e sion o Wi esha k, p o iding he same unc ionali y
ega dless he GUI and acili a ing he sc ip ing and au oma-
ion asks. Command line ool capin os was used o ex ac
s a is ics om he pcap iles, Bash sc ip ing o a icula e he
au oma ion asks and he s a is ics language R o calcula ions
and plo s.
Packe s om eal aces missed 4 by es co esponding o
he F ame Check Sequence (FCS) o E he ne . Mos ope a ing
sys ems do no p o ide hese by es o packe analyse s. On he
con a y, simula ed aces include he E he ne FCS. These 4
ex a by es pe packe ha e been aken in o accoun in he
calcula ion o he h oughpu o eal a ic.
Expe imen al da a om [12] came om h ee cases, com-
p ising 2, 3 and 5 concu en OWL clien s espec i ely. The
s udy o such a educed se o clien s is no de e mined by
he se e esou ces, bu by he numbe o use s ha can
simul aneously in e ac while p o iding a p ope use expe i-
ence wi hou o e whelming he end use . All he clien s wi hin
hose sessions pe o med high ac i i y a es. Th ee equi alen
simula ion con igu a ions ha e been de ined o be ec ea ed by
simula ion. The goal is o compa e he simula o h oughpu
wi h p e ious expe imen al esul s. Ra e pa ame e alues om
Table I ha e been used o se he objec synch onisa ion packe
a e acco ding o he numbe o concu en use s speci ied o
each simula ion un.
Con igu a ions o he 2, 3 and 5-clien cases ha e been un
100 imes each one in o de o ob ain he sample mean ( ¯
X)
and sample s anda d de ia ion (σ¯
X) wi h a Con idence In e al
(CI) o 95%. Each un used a di e en seed o ini ialise he
andom numbe gene a o s.The du a ion o each simula ed ses-
sion was o 250 s. Packe sende / ecei e classes do no ha e
o es ablish ne wo k connec ions o load simula ion en i ies
which gene a e hea y da a a ic. Ins ead, hey s a gene a ing
a ic ollowing s a iona y pa e ns om eal OWL clien s.
Thus, due o he logic coded in he packe sende / ecei e
4
classes, he wa m-up pe iod is almos negligible and he clien s
each long un h oughpu alues in ea ly simula ion s ages,
so long simula ion imes a e no needed.
Resul s showed ha he simula ed objec synch onisa ion
h oughpu dec eases as he numbe o concu en clien s
inc eases. Resul s o simula o uns using λpa ame e s om
Table I a e shown in Table II. 100 simula o uns (samples)
ha e been pe o med o each con igu a ion (2, 3 and 5-
clien s), sample means ( ¯
X) dec ease as he numbe o clien s
inc eases. The sample de ia ion (σ¯
X) does no expe ience
signi ican a ia ions. Figu es ha e a 95% Con idence In e al
assuming a no mal dis ibu ion o he sample means alues.
Expe imen al esul s om he p e ious s udy a e shown in
Table III. Al hough he simula ion esul s show a sligh
o e es ima ion, eal and simula ed clien h oughpu ollow
a simila e olu ion linked o he numbe o concu en use s.
This co ela ion can be obse ed in Figu e 2. I should be no ed
ha he expe imen al esul s ha e a ela i ely high de ia ion
due o he a iable na u e o he a ic. Despi e he sligh
o e es ima ion, he simula o esul s do no di e la gely i
he expe imen al de ia ion is aken in o accoun . Simula o
means a e wi hin o e y close o he in e al de ined by he
de ia ion abou he expe imen al means.
TABLE II
SIMULATED OBJECT SYNCHRONISATION THROUGHPUT
N. Clien s Samples ¯
X(b/s)σ¯
X
2 100 1475.36 48.07
3 100 1352.62 49.99
5 100 1231.95 42.76
TABLE III
EXPERIMENTAL OBJECT SYNCHRONISATION THROUGHPUT
N. Clien s ¯
X(b/s)σ¯
X
2 1407.49 19.27
3 1282.78 141.62
5 1085.91 134.72
Clien s
by e/s
2 3 4 5
0 500 1000 1500 2000
Expe imen al a ic
Simula ed a ic
Fig. 2. Simula ed Vs expe imen al objec synch onisa ion h oughpu
Audio clien a ic gene a ion ollows a pe iodic pa e n
dic a ed by jVoiceB idge, he so wa e componen esponsible
o sending audio packe s ollowing a heo e ical pe iod o 20
ms. Expe imen al esul s showed a di e en gene a ion pa e n
while he o e all audio h oughpu emained cons an o e
ime and sligh ly below he expec ed maximum heo e ical
audio h oughpu [12]. Thus, he audio In e -A i al Time
model (Equa ion 3) was p oposed o be e desc ibe audio
a ic and gi e a mo e ealis ic es ima ion han he heo e ical
one only based on he used audio coding. A summa y o he
simula o esul s o audio a ic a e shown in Table IV. All
he samples ha e a simila mean alue ( ¯
X) independen ly o
he numbe o concu en clien s pa icipa ing in he session.
The sample a iance σ¯
Xand Con idence In e al do no
show signi ican changes be ween con igu a ions. The simu-
la ed audio h oughpu alues o e es ima e he expe imen al
esul s shown in Table V. Despi e his di e ence, simula ion
esul s s ill p o ide a mo e accu a e es ima ion han heo e ical
maximum audio h oughpu , which was es ima ed in 62.4 KB/s
o he audio coding used. Simula ed and eal h oughpu also
show a conside able deg ee o co ela ion ha can be obse ed
in Figu e 3.
TABLE IV
SIMULATED AUDIO THROUGHPUT
N. Clien s Samples ¯
X(b/s) σ¯
X
2 100 54336.05 310.94
3 100 54363.64 332.80
5 100 54356.50 333.34
TABLE V
EXPERIMENTAL AUDIO THROUGHPUT
N. Clien s ¯
X(b/s) σ¯
X
2 48187.14 433.24
3 45678.54 3990.91
5 47816.95 4216.89
Clien s
KB/s
2 3 4 5
0 10 20 30 40 50 60 70
Expe imen al a ic
Simula ed a ic
Theo e ical maximum
Fig. 3. Simula ed Vs expe imen al audio h oughpu
5
VI. CONCLUSION AND FUTURE WORK
This pape desc ibes he design, implemen a ion and ali-
da ion o a simula ion ool o s udying Open Wonde land, a
Ne wo ked Vi ual En i onmen (NVE). The ool ocuses on
he ou going a ic o OWL clien s. This a ic is di ided in o
wo ca ego ies: audio and objec synch onisa ion. Thei mic o
scale models we e de ined in p e ious wo k.
Ns-3 p ojec was used as simula ion amewo k o a oid
de eloping he ool om sc a ch. Some o he ns-3 de aul
models, such as hose o E he ne and IP 4 ha e been used
along wi h o he cus om classes ha implemen he a ic
gene a ion beha iou o he OWL clien s. The esul ing com-
mand line ool is con igu able h ough pa ame e s and gene a e
ne wo k aces in pcap o ma . This ea u e allows o s udy
he simula o aces using ools such as Wi esha k and Tsha k,
o au oma ise he ace gene a ion and he in e ope abili y wi h
o he ools.
The alida ion o he simula ion ool was pe o med de ining
se e al simula ion scena ios ha eplica ed he expe imen al
sessions om p e ious wo k. These con igu a ions desc ibed
se e al mul i-playe sessions whose clien s showed high a es
o use ac i i y. The simula ion inpu was he a e pa ame e λ,
which de e mines he packe gene a ion ela i e o he numbe
o concu en use s. Th oughpu was chosen as he ou pu
simula o pa ame e o s udy. Mul iple uns o he simula o
we e pe o med o de e mine he mean h oughpu o bo h
audio and objec synch onisa ion a ic, hei sample de ia ion
and a con idence in e al o 95%.
The compa ison be ween p e ious expe imen al and sim-
ula ion esul s showed ha hey ha e a conside able deg ee
o co ela ion. The simula o sligh ly o e es ima e he objec
synch onisa ion h oughpu bu i s esponse o he inc ease in
he numbe o use s is simila o he expe imen al esul s. The
simula o also o e es ima e he audio h oughpu bu i s ill
gi es a be e es ima ion han he heo e ical maximum audio
h oughpu .
The mos immedia e u u e wo k would be desc ibing mo e
accu a ely he ela ionship be ween he numbe o concu en
clien s wi hin an OWL session and he alue o he a e
pa ame e λ ha de e mines he packe gene a ion o hose
clien s. This would simpli y he speci ica ion o he inpu
pa ame e o he simula o . Thus, he ool would only need he
numbe o concu en use s o au oma ically in e he p ope
λ alue.
Fu he u u e wo k would ocus on he s udy o o he
ne wo k pa ame e s ha ha e a di ec impac on he use
expe ience, such as packe loss, packe delay o he in luence
o he unde lying ne wo k opology and echnology (Gigabi
E he ne , Wi i, DSL, e c.). The simula ion p esen ed in his
pape p o ides he g ound o a modula , expandable and
con igu able simula o ne wo k ha would p o e o g ea help
o he s udy o he pa ame e s ci ed abo e and could be also
expo ed o o he ields ha equi e simila ne wo k e alua ion
app oaches.
ACKNOWLEDGMENT
This wo k was pa ially suppo ed by he p ojec
PROCUR@ - IPT-2011-1038-900000, unded by he p og am
INNPACTO o he Spanish Minis y o Science and Inno a ion
and FEDER unds; p ojec Vulcano: TEC2009-10639-C04-02;
and by he Tele onica Chai “In elligence in Ne wo ks” o he
Uni e si y o Se ille, Spain.
REFERENCES
[1] B. Fogg, Pe suasi e echnology: using compu e s o change wha we
hink and do, se . Mo gan Kau mann se ies in in e ac i e echnologies.
Mo gan Kau mann Publishe s, 2003.
[2] S. Rome o, L. Fe nandez-Luque, J. Se illano, and L. Vognild, “Open
Sou ce Vi ual Wo lds and Low Cos Senso s o Physical Rehab o
Pa ien s wi h Ch onic Diseases,” Lec u e No es o he Ins i u e o Com-
pu e Sciences Social-In o ma ics and Telecommunica ions Enginee ing,
ol. 27, pp. 84–87, 2010.
[3] D. Cascado, S. Rome o, S. Ho s, A. B ase o, L. Fe nandez-Luque, and
J. Se illano, “Vi ual wo lds o enhance ambien -assis ed li ing,” in
Enginee ing in Medicine and Biology Socie y (EMBC), 2010 Annual
In e na ional Con e ence o he IEEE, 31 2010-sep . 4 2010, pp. 212
–215.
[4] O. W. P ojec , Open Wonde land P ojec , accessed Ma ch 15 h 2011.
[Online]. A ailable: h p://www.openwonde land.o g/
[5] S. Ra i, B. Ha i i, and S. Shi mohammadi, “A Su ey o Fi s -Pe son
Shoo e Gaming T a ic on he In e ne ,” IEEE In e ne Compu ing,
ol. 14, no. 5, pp. 60–69, Sep. 2010.
[6] M. Claypool, D. LaPoin , and J. Winslow, “Ne wo k analysis o coun e -
s ike and s a c a ,” in Pe o mance, Compu ing, and Communica ions
Con e ence, 2003. Con e ence P oceedings o he 2003 IEEE In e na-
ional. IEEE, 2003, pp. 261–268.
[7] T. F i sch and H. Ri e , “The e ec o la ency and ne wo k limi a ions
on MMORPGs: a ield s udy o e e ques 2,” ACM SIGCOMM wo kshop
on Ne wo k, 2005.
[8] D. G acanin, Y. Zhou, and L. DaSil a, “Quali y o se ice o ne wo ked
i ual en i onmen s,” Communica ions Magazine, IEEE, ol. 42, no. 4,
pp. 42–48, 2004.
[9] T. Hende son and S. Bha i, “Ne wo ked games: a QoS-sensi i e applica-
ion o QoS-insensi i e use s?” in P oceedings o he ACM SIGCOMM
wo kshop on Re isi ing IP QoS: Wha ha e we lea ned, why do we
ca e? ACM, 2003, pp. 141–147.
[10] E. Asensio, “Analyzing he ne wo k a ic equi emen s o mul iplaye
online games,” Ad anced Enginee ing Compu ing and Applica ions in
Sciences, 2008. ADVCOMP’08. The Second In e na ional Con e ence
on, pp. 229–234, 2008.
[11] J. Fon , D. Cascado, J. Se illano, G. Lopez, S. Rome o, and G. Jimenez,
“Ne wo k equi emen s e alua ion o a mul i-use i ual en i onmen ,”
in Pe o mance E alua ion o Compu e Telecommunica ion Sys ems
(SPECTS), 2011 In e na ional Symposium on, june 2011, pp. 90 –97.
[12] J. ”Fon , D. Cascado, J. Se illano, F. D´
ıaz-del R´
ıo, and G. Jim´
enez,
“Ne wo k a ic analysis and e alua ion o a mul i-use i ual en i on-
men ,” Jan. 2012, submi ed o SIMPAT.
[13] M. Bo ella, “Sou ce models o ne wo k game a ic,” Compu e Com-
munica ions, ol. 23, no. 4, pp. 403–410, Feb. 2000.
[14] J. L. Fon , P. I˜
nigo, M. Dom´
ınguez, J. L. Se illano, and C. Amaya,
“A chi ec u e, design and sou ce code compa ison o ns-2 and ns-3
ne wo k simula o s,” in P oceedings o he 2010 Sp ing Simula ion
Mul icon e ence, se . Sp ingSim ’10. New Yo k, NY, USA: ACM,
2010, pp. 109:1–109:8.
[15] J. Fon , P. I˜
nigo, M. Dom´
ınguez, and C. Se illano, J.L.and Amaya,
“Analysis o sou ce code me ics om ns-2 and ns-3 ne wo k simula-
o s,” Simula ion Modelling P ac ice and Theo y, ol. 19, no. 5, pp.
1330–1346, 2011.