Research Article Bandwidth Management in Wireless Home Networks for IPTV Solutions

Transcription

1 Digial Mulimedia Broadcasing Volume 2013, Aricle ID , 7 pages hp://dx.doi.org/ /2013/ Research Aricle Bandwidh Managemen in Wireless Home Neworks for IPTV Soluions Tamás Jursonovics 1 and Sándor Imre 2 1 Deusche Telekom AG Producs & Innovaion, T-Online-Allee 1, Darmsad, Germany 2 Deparmen of Telecommunicaions, Budapes Universiy of Technology, Budapes, Hungary Correspondence should be addressed o Tamás Jursonovics; jursonovics@gmail.com Received 29 Ocober 2012; Acceped 8 January 2013 Academic Edior: Massimiliano Laddomada Copyrigh 2013 T. Jursonovics and S. Imre. This is an open access aricle disribued under he Creaive Commons Aribuion License, which permis unresriced use, disribuion, and reproducion in any medium, provided he original work is properly cied. The opimal allocaion of he reransmission bandwidh is essenial for IPTV service providers o ensure maximal service qualiy. This paper highlighs he relevance of he wireless ranspor in oday s IPTV soluion and discusses how his new media affecs he exising broadcas echnologies. A new Markovian channel model is developed o address he opimizaion issues of he reransmission hroughpu, and a new mehod is presened which is evaluaed by empirical measuremens followed by mahemaical analysis. 1. Preface The erminology of Open IPTV Forum (OIPF) specificaion release 2 [1] has been chosen o describe he inerne proocol elevision IPTV feaures in his paper because we experience a wide diversiy of erms in several journals which may confuse he reader. We believe ha he OIPF erms are sraighforward, and hey can be easily inerpreed on any IPTV soluions alhough our work is independen from he sandard iself. 2. Moivaion The elecommunicaion indusry is ardily changing. The emerging marke of he new generaion over-he-op (OTT) services from Google, Microsof, Apple, or Amazon had pu a big pressure on operaors o move away from he radiional elecom model and assess hreas and opporuniies from OTT players. There is a big race for he cusomers oday, and legacy indusry has o exend is porfolio wih various value adding services like riple-play, rich communicaion, or mobile paymen. This paper evaluaes one specific opic of his compeiion, he video broadcasing services. We have observed he rapid evoluion of he IPTV services in he las decade. The high-definiion broadcas go popular since is inroducion in 2004, and he accessibiliy of 3D conen is growing year by year. The consumer elecronic devices become inegraed par of our life, cusomers access digial conen from se-op boxes (STBs) o ables and mobile devices. From he IPTV service provider poin of view, he demand of high-qualiy services emerged; however, he infrasrucure of he access nework remained he same. The main echnology of elecommunicaion operaors providing inerne remained he years old wised copper pairs. On one hand, he IPTV Service Providers are moivaed by he maximizaion of heir cusomer reach, bu in many cases digial subscriber line xdsl offers inadequae bandwidh for high qualiy services [1]. access nework providers need o find a soluion which enables hem o uilize heir curren infrasrucure. The answers may include he implemenaion of a more advanced encoding algorihm (H.264, H.265) which resuls in having he same qualiy on smaller hroughpu [2] or he inroducion of a hybrid service which replaces he mos bandwidh consuming scheduled conen ranspor wih digial video broadcasing DVB-X echnology [3], he usage of progressive download, or, as we poin ou in our work, he implemenaion of a more effecive bandwidh allocaion in he access nework which would ensure a more efficien ranspor.

2 2 Digial Mulimedia Broadcasing Bandwidh Throughpu Opporuniy-a Prio 3 Inerne Prio 2 IPTV SD HD sream SD SD sream AV daa B sream B AV daa B cri Opporuniy-b Prio 1-voice P overload P cri,sa P cri,dyn P LOSS Figure 1: Bandwidh allocaion in he access nework. On he oher hand, he service porabiliy allows consumers o access he conen no only on STBs, bu also from various hand-held devices. Today he prime wireless access echnology wihin he consumer domain is he wireless local area nework WLAN connecion, herefore service plaform providers have o adap heir IPTV soluion o suppor he specific requiremens of his wireless communicaion channel. Mos of he curren research papers [4 6] ake only one of he above menioned wo aspecs ino consideraion. The effec of he wireless channel is usually addressed on he media access conrol layer, and he research of he IPTV disribuion focuses only on he qualiy of he fixed nework infrasrucure. Our research does address he above described overlap of he IPTV delivery over WLAN in a limied resource environmen of xdsl echnology. We presen he concep and main problems of bandwidh allocaion in Secion 3. Our new heoreical model is inroduced by Secion 4. We develop a more effecive bandwidh allocaion algorihm in Secion 5, which we evaluae and validae. Finally, Secion 6 concludes our work and shows some poenial furher applicaion areas. 3. Bandwidh Allocaion Le us begin he discussion of bandwidh managemen by inroducing a ypical riple-play bandwidh allocaion scheme on Figure 1. Access nework providers usually dedicae a reserved bandwidh in he access nework for voice communicaion and share he remaining hroughpu beween IPTV and inerne services wih a prioriy for he former one. The acual hroughpu of IPTV service depends on he user s configuraion. In mos of he cases, a oken-based sream managemen allows he cusomers o simulaneously receive muliple sreams for live viewing or recording purposes (1SD+1HD or 3SD+0HD). One oken allocaes bandwidh for he AV (audio-video) daa ranspor and reserves a dedicaed bandwidh for he reransmission service. We discuss he problem and radeoff of his bandwidh allocaion, herefore, in he following paragraphs, we are going o describe i in deails. The balance beween he assigned bandwidh for AV daa and he reserved bandwidh for service is crucial for achieving he maximal qualiy in IPTV soluions. On one hand he sream bandwidh as consan in ime (because of he widely applied is considered consan birae CBR video Dynamic BW allocaion Saic BW allocaion Figure 2: Saic versus dynamic bandwidh allocaion. encoding) he more hroughpu is assigned for he AV daa and he beer sream qualiy can be achieved by he increase of he encoding birae bu he less opporuniy is given for error correcion. A smooh sharp sream may be disurbed by blocking or full frame ouages due o he insufficien hroughpu. On he oher hands reserving high bandwidh for error correcion degrades he overall sream qualiy due o he low encoding birae. Based on differen nework insallaions, he raio of bandwidh o AV daa is usually uned beween 10 and 25%, bu a subopimal value may significanly reduce he hroughpu and, qualiy of an IPTV Soluion. 4. Proposed Model for Bandwidh Allocaion IPTV Soluions in WLAN Home Neworks The main concep and benefis of our research are showed by Figure 2. The solid line represens he heoreical hroughpu of he AV daa sream a various packe loss probabiliies in case of saic bandwidh allocaion. The funcion is consan ill P overload loss probabiliy, where he loss rae is so high ha he reransmission raffic fully occupies is reserved bandwidh. Above his, rae reransmission does no have enough bandwidh o recover all he packe losses; herefore, he acual hroughpu of AV daa sream is decreasing (no ransmied packes), and cusomers experience qualiy deerioraion. We also declare a criical bandwidh value (B cri ) for AV daa ransmission. [Below his value, i does no makesenseoprovideiptvservicedueohemassivelosses. A dynamic bandwidh allocaion (dash-do line on Figure 2) increases he hroughpu of he AV sream a low packe loss raes o achieve a beer qualiy for he IPTVservice(opporuniy-a).Secondly,ahigherlossraes, he hroughpu of AV sream decreased o avoid double packe delivery (opporuniy-b) which enables he operaor o expand he value of he criical packe loss rae (P cri,dyn ) and provides a lower qualiy bu error-free IPTV service in a wors environmen. Inhispaperwepresenonlyoneparofouroverall research, he opimal selecion of he reransmission hroughpu. Our mehod inroduced by he upcoming secions predics he loss aribues of he wireless ransmission and defines he opimal value of he hroughpu considering he overall loss parameers wih he aim of

3 Digial Mulimedia Broadcasing 3 B S n TX B n S n 1 + Z n Z n Channel d d R n RX C n Figure 3: Channel model. minimizing packe losses. Firs, we describe he channel and he bandwidh models The Channel Model. Considering he requiremens above and he aribues of a WLAN ranspor, we decided o inroduce a discree-ime channel and a Markov model for he mahemaical descripion of he UNIT-17 inerface (AV daa sreams in he Access and residenial Neworks). We consider no only he packe arrival, bu also he reransmission raffic as a discree-ime sochasic process, and we prove ha i is a homogeneous Markov chain. Le S n = {0,foralln} represen he number of sen packes a he ime n on Figure 3, TXhemulicasconen delivery funcion, Z n ={0;1,foralln} an addiive whie noise in he wireless communicaion channel, d he ransmission delay, and RX he open IPTV erminal funcion (OITF). The received packes are expressed wih R n ={0, for successful packe ransmission;1,forpackeloss}. We assume ha he receiver deecs a packe loss (by checking he sequence numbers of packes) and requess he reransmission B n ={1, for a packe reransmission;0oherwise} of every los packes only once from he B Fas Channel Change/Reransmission server. We also assume ha he reransmission reques communicaion is proeced by an error-free proocol, like TCP. By his definiion, we obain d + R n B n =R n =Z n d. (1) The B n signal ravels hrough he same wireless channel; herefore, i is also effeced by he same Z n noise, and i may be also los. We model his effec by expressing he received correcion signal C n = {0, 1} wih hree operaors, an inverer ( 1), a muliplier ( ), and a channel ransmission delay (d). These funcions enable us o assign he value of 1 for C n only in he case when he reransmission signal is no effeced by he channel noise (no los), and he value of 0 oherwise. The final received and correced signal R n ={0,forsuccessful packe ransmission, 1 for packe loss (unsuccessful reransmission), and 2 for successful packe reransmission} is R n =R n d +C n =S n 2d +Z n 2d +B n d Z 1 n d (2) =S n 2d +Z n 2d +Z n 2d Z 1 n d. Le us observe ha he firs erm of he addiion equals o 0 by definiion and he las erm equals o he sampling of he Z n whie noise wih is own delayed signal. The auocorrelaion of he whie noise is zero for all nonzero ime shifs [7]; herefore, R n can be described as a sequence of an independen random variable which saisfies he Markov def propery. X n =R n is a homogeneous Markov chain. Le X n =0if he nh packe is received correcly; X n =1if he nh packe is los and has no been reransmied; finally X n =2if he nh packe is successfully reransmied afer loss. We analyzed and described a hree-sae Markov model in our previous publicaion; herefore, we simply lis mos imporan properies. For a deailed discussion including he resoluion of (3) (6) and for he meaning of he probabiliies, please refer o our former paper [8]. The ransiion marix 1 p 01 p 02 p 01 p 02 [ p 10 1 p 10 p 12 p 12 ]. (3) [ p 20 p 21 1 p 20 p 21 ] The seady-sae packe loss rae P LOSS,seady =(p 02 p 21 +p 01 p 20 +p 01 p 21 ) (p 01 p 20 +p 01 p 21 +p 02 p 21 +p 10 p 02 +p 12 p 01 +p 12 p 02 +p 21 p 10 +p 20 p 10 +p 20 p 12 ) 1. The seady-sae packe reransmission rae P,seady =(p 10 p 02 +p 12 p 01 +p 12 p 02 ) (p 01 p 20 +p 01 p 21 +p 02 p 21 +p 10 p 02 +p 12 p 01 +p 12 p 02 +p 21 p 10 +p 20 p 10 +p 20 p 12 ) 1. And he probabiliy of loss burs wih lengh l (4) (5) P,burs (l) = (1 p 20 p 21 ) l 1 (p 20 +p 21 ). (6) 4.2. The Model of Bandwidh Limiaion. The previous secion inroduced how he wireless channel affecs he packe ransmission, and now we are going o analyze he he bandwidh allocaion in IPTV soluions. We inroduce hree planes of he IPTV packe ransmission. The ransmier plane represens he provider s nework, receiver plane represens he OITF, and playou plan represens he conen presenaion wihin he OITF. The evens of he packe loss usually show a bursiness in wireless communicaion [9 11]. Therefore, we invesigae an inraburs loss on Figure 4 afer he firs k consecuive packe losses (b) he receiver requess hem for reransmission (c) which packes are delivered wihin he allocaed bandwidh for packe ransmission (d) o he presenaion device (e). For a successful reransmission, all reransmied packe should

4 4 Digial Mulimedia Broadcasing Bandwidh AV daa 12 k (c) Reransmied 1 k (a) Sen packes (d) Re. received 1 2 k AV daa 12 k (b) Los packes AV daa Dre D playo Transmier Receiver (e) Played ou 12 k Playou T re T playo Bandwidh (c) Reransmied 1 k n AV 1 k n daa (a) Sen packes (d) Re. received 1 k AV 1 k n daa AV daa (b) Los packes D re, k n Transmier Receiver (e) Played ou 1 k n Playou T T re, n re, k Figure 4: Inraburs reransmission. D playo Figure 5: Inerburs reransmission. T playo, n arrive earlier han heir presenaion ime (T re T playo ). Expressing his condiion wih he duraions D pkg + RTT +k B AV D B pkg < (k 1) D pkg +D playo, (7) where D pkg is he average ransmission ime for a packe, RTT is he round-rip ime, B AV and B are he allocaed bandwidhs on he communicaion channel, and D playo is he packe playou buffer in he OITF. The formula of he maximal number of consecuive packes which can be successfully reransmied is defined as k,inra,max ( B D playo 2D pkg RTT )= B AV D pkg ((1/ (B /B AV )) 1). (8) The acual hroughpu of he AV sream may vary by insallaions; herefore, we expressed his value as a raio of B and B AV. Second, we highligh he barrier of inerburs behavior on Figure 5. Afer a loss of long burs, he reransmission bandwidhisoccupiedbyherafficofhereransmied packes even if here is no oher packe loss a he ime in he video sream. This means ha a loss even blocks he reransmission channel. We are ineresed in he following quesion: assuming a k<k,inra,max longbursofloss,afer how many packes (n)cananewlossbursoccurwhichwould be also successfully reransmied (e.g., wha is he minimal disance (n k) beween wo loss burss if he firs burs lass def for k packes?). Now, T re,k = T playo,k,andt re,n T playo,n. Figure5 shows ha D pkg + RTT + (k+1) B AV D B pkg < (n 1) D pkg +D playo, (9) where n>k.expressingn n,iner,min (k, B AV B )=(k+1) B AV B + RTT D playo D pkg +2. (10) Number of packes B \B AV n,iner,min D playo = 1000 ms k,inra,max D playo = 1000 ms n,iner,min D playo = 800 ms k,inra,max D playo = 800 ms n,iner,min D playo =600ms k,inra,max D playo =600ms n,iner,min D playo =400ms k,inra,max D playo =400ms n,iner,min D playo = 200 ms k,inra,max D playo = 200 ms Figure 6: Inra- and inerburs limiaion. We presen he above declared funcions (8) (10) on Figure 6. This graph shows ha a small values of B AV /B (0% 15%), he effec of he iner-burs blocking is greaer. For example, a 5% he maximal consecuive burs lengh is 10 packes, and he reransmission channel is blocked by hisrafficfor30packes.above20%,hiseffecbecomes insignifican. Wealsosaehawihgraerplayoubuffer(D playo ), isableocorreclargerlossburssahepriceofagreaerendo-end delay.

5 Digial Mulimedia Broadcasing 5 5. The Opimal Reransmission In his secion, we inroduce a new mehod for reransmission bandwidh allocaion based on our models wih he aim of achieving a beer video qualiy. The mehod is realized in a es environmen, and our hypohesis is measured and proved. Tradiional algorihms reques all los packes; herefore, hey have o implemen a nework layer raffic shaping o fi he acual reransmission hroughpu ino he allocaed bandwidh. This is usually done by packe queuing which increases he overall packe reransmission ime; herefore, he probabiliy of a reransmission packe arrives lae (afer is playou ime) is grea. Several papers addressed his problem [12], inroducinga selecivereransmissionproocol byevaluaingherafficonheapplicaionlevel,andassigning prioriy and imporance for each packe reransmissions. The main advanages of our mehod are he minimal addiional delay, he low resource needs, and he consideraion of he wireless channel. Our mehod assess he mechanism on he nework layer, skips (forbids) he reransmission requess of a los packe according o he above described inra- and inerburs channel blocking, and akes he special properies of he wireless channel ino consideraion. Our algorihm consiss of hree seps. (1) The packe arrival process is coninuously moniored for packe loss. (2) k,inra,max and n,iner,min are calculaed. (3) A los packe is requesed reransmission only if he inra- and iner-burs channel blocking do no forbid he reransmission; oherwise, he reransmission reques is skipped Empirical Evaluaion. To evaluae our model and mehods, we followed he OIPF sysem archiecure [13] and implemened he following OIPF funcions in a es environmen (sourcecodesareavailableon[14]). (1) Mulicas conen delivery funcion, ser, c++ applicaion generaes UDP/RTP mulicas raffic and implemens a simple conrol proocol hosed by an x86 Linux server conneced o he core nework of Deusche Telekom (DT). (2) server, re, c++ applicaion sores he mulicas raffic in a circular buffer and implemens a simple reransmission reques proocol hosed by he same x86 Linux server. (3) Uni-17 inerface, par a was realized by he ADSL2+ access nework of DT and was provided by a DLink ADSL modem. (4) Uni-17 inerface, par b was realized by a b WLAN nework and was provided by a Cisco 1200 series wireless access poin (AP) conneced o he ADSL modem. (5) OITF: cli, c++ applicaion implemens a simple mulicas receiver and conrols funcions of he mulicas Throughpu (kbps) Time (s) Reserved bandwidh all skip Figure 7: Reransmission hroughpu. conen delivery funcion and server. Hosed on a x86 Linux lapop conneced o he Cisco AP. Figure 7 shows he hroughpu of he reransmission sream in wo cases: radiional reransmission ( all: all los packes were requesed for reransmission) and our new reransmission mehod ( skip: reransmission requess may be skipped based on he acual parameers of he channel). I can be clearly seen ha our algorihm kep he reransmission hroughpu under is dedicaed bandwidh which ensured he in-ime delivery of he reransmission packes however we inenionally skipped hose reransmission requess which in ime delivery would no be ensured due o he channel blocking (inra- and iner-burs effec). The main benefi of our mehod is showed by Figure 8.We compared he oal packe loss rae in he above menioned wocases,andwefoundhawihhesmarskippingof packe reransmission requess, we were able o achieve a beer (smaller) loss rae hen reransmiing all of he packes. Our mehod avoided he effec of lae reransmission. IF a packe is requesed for reransmission wihou ensuring he necessary ranspor bandwidh, hen i may delay furher reransmission requess which may arrive o lae afer heir playou ime. This causes an inefficien reransmission bandwidh uilizaion which increases he overall packe loss rae (on he playou plane) The Effec of he Inraburs Limiaion. Le us analyze our resuls heoreically as well. In his and in he upcoming secion, we characerize he Uni-17 inerface wih he ransiion marix of our Markov model and he design aribues of he access nework. Applying he inra- and iner-burs limiaions on our model, we derive he probabiliy of he reransmission skip caused by our algorihm. Finally, we express he overall packe loss rae which is a key indicaor for he qualiy of he video ransmission.

6 6 Digial Mulimedia Broadcasing Overall packe loss rae all skip Time (s) Figure 8: Overall packe loss rae. The inra-burs limiaions have a significan shor-ime effec if he disance of he burs losses is grea (p 00 1).Ourquesionisabouheprobabiliyofreransmission packe skip. Firs, we calculae he maximal number of consecuive reransmission requess k,inra,max = D playo 2D pkg RTT D pkg ((B AV /B ) 1). (11) The probabiliy of l long reransmission reques is given by he Markov model (6). We calculae he probabiliy of m packe skips if he reransmission burs is greaer han k R,burs,max,whichis P skip (m) = (p 20 +p 21 )(1 p 20 p 21 ) k,inra,max 1+m. (12) The overall probabiliy of a packe skip is given by P skip (i) P skip,inra = lim i i i=1 (p = lim 20 +p 21 )(1 p 20 P 21 ) k,inra,max 1+i i i i=1 = (p 20 +p 21 )(1 p 20 p 21 ) k,inra,max 1 (1 p lim 20 p 21 ) i i. i i=1 (13) Le us observe ha he las sum can be expressed as a specialformofhepolylogarihm(alsoknownasjonquière s funcion) z k Li s (z) s=1 = k s s=1 = k=1 k=1 z k k, (14) for every 1 z<1. (1 p 20 P 21 saisfies his crieria. Therefore using he well-known formula of Li 1 (z) = ln(1 z), he equaion can be expressed in a closed form P skip,inra = (p 20 +p 21 )(1 p 20 p 21 ) k,inra,max 1 ( 1) ln (1 (1 p 20 p 21 )) = (p 20 +p 21 )(1 p 20 p 21 ) k,inra,max 1 ( 1) ln (p 20 +p 21 ). (15) The overall packe loss can be expressed as a sum of he probabiliy of packe skip (15) and he seady sae probabiliy of packe loss (4) The Effec of he Inerburs Limiaion. We ask he same quesion as in he previous secion, wha is he probabiliy of packe skip? Le us assume ha he firs burs is small enough o be reransmied (k < k,inra,max ). The probabiliy of reransmission burs is of l size is given by our Markov model (6). The probabiliy of k reransmission burs followed by n k good ransmission burs and a second reransmission f (k, n) = (p 20 +p 21 )(1 p 20 p 21 ) k 1 p 20 (p 01 +p 02 ) (1 p 01 p 02 ) n k 1 p 02. (16) The firs packe of he second reransmission burs is skipped if n<n,iner,min (k). Fromhis,wecancalculae he probabiliy of one packe skip for k P skip,iner,k = n,iner,min (k) i=k f (k, i). (17) For he overal packe skip probabiliy, we have o summarize (17)foreveryk<k R,burs,max P skip,iner = k,inra,max j=1 n,iner,min (j) i=j f(j,i). (18) The overall packe loss can be expressed as a sum of he probabiliy of packe skip (18) and he seady-sae probabiliy of packe loss (4). 6. Conclusion In his paper, we highlighed he relevance of he wireless ranspor in oday s IPTV soluions, and we poined ou ha i s and he access nework s combined effec on he bandwidh managemen is no discussed deeply by publicaions. Our general research projec arges his specific area by inroducing several opimizaion mehods from which we presened one, he opimizaion of he packe reransmission on he previous pages. We creaed a new discree-ime channel model o describe he effec of he burs losses on he IPTV service qualiy and

7 Digial Mulimedia Broadcasing 7 he role of he packe reransmission. We proved ha i is a Markov Chain, and as par of our resuls, we expressed and evaluaed is mos imporan quaniaive parameers. Using our model, we also inroduced an algorihm for reransmission opimizaion in IPTV soluions over WLAN home neworks. As a furher evaluaion of our resuls, we creaed a esbed in alignmen wih he OPIF sysem archiecure, and we performed he empirical analysis of our channel model and mehods. We showed ha our concep improved he overall packe loss characerisics. Furhermore, we enclosed a mahemaical analysis of our algorihm, and we derived he heoreical packe loss probabiliies o suppor our measuremens. In he nex research phases, we are going o invesigae, inroduce, and leverage our heoreical resuls of hroughpu managemen in he adapive birae sreaming echnologies for IPTV soluions, and we are going o evaluae our channel model in he media access conrol layer of he wireless access echnologies. neworks, Compuer Communicaions,vol.31,no.11,pp , [13] Open IPTV Forum, Funcional Archiecure [V2. 1] [ ], hp:// [14] hp:// jursonovics/. References [1] Open IPTV Forum, Release 2 Specificaions, hp:// [2] A. Abramowski, Towards H. 265 video coding sandard, in Proceedings of he Phoonics Applicaions in Asronomy, Communicaions, Indusry, and High-Energy Physics Experimens, vol of Proceedings of he SPIE, [3] H. Beno, Digial Television: Saellie, Cable, Terresrial, Ipv, Mobile Tv in he Dvb Framework, Focal Press, [4] M.M.Hassani,S.V.Jalali,andA.Akbari, EvaluaingTCPflows behavior by a mahemaical model in WLAN, JournalofBasic and Applied Scienific Research,vol.2,no.3,pp ,2012. [5] A. C. Begen, Error conrol for IPTV over xdsl neworks, in Proceedings of he 5h IEEE Consumer Communicaions and Neworking Conference (CCNC 08), pp , January [6]H.Bobarshad,M.VanDerSchaar,andM.R.Shikh-Bahaei, A low-complexiy analyical modeling for cross-layer adapive error proecion in video over WLAN, IEEE Transacions on Mulimedia,vol.12,no.5,pp ,2010. [7] A. Papoulis, Probabiliy, Random Variables and Sohasic Processes, McGraw-Hill, [8] T. Jursonovics and S. Imre, Analysis of a new Markov model for packe loss characerizaion in IPTV Soluions, Infocommunicaions Journal,vol.2,pp.28 33,2011. [9] T. Jursonovics and Zs. Buyka, The implemenaion and analysis of ineracive mulimedia mobile services, in Proceedings of he 11h Microcoll Conference,, pp , Budapes, Hungary, [10] T. Jursonovics, Zs. Buyka, and S. Imre, Mulimedia ransmission over mobile neworks, in Inernaional Conference on Sofware, Telecommunicaions and Compuer Neworks (SofCOM 04),pp ,Dubrovnik,Croaia,Ancona,Ialy,2004. [11] A. Nafaa, T. Taleb, and L. Murphy, Forward error correcion sraegies for media sreaming over wireless neworks, IEEE Communicaions Magazine,vol.46,no.1,pp.72 79,2008. [12] Á. Huszák and S. Imre, Source conrolled semi-reliable mulimedia sreaming using selecive reransmission in DCCP/IP

8 Roaing Machinery Engineering The Scienific World Journal Disribued Sensor Neworks Sensors Conrol Science and Engineering Advances in Civil Engineering Submi your manuscrips a Elecrical and Compuer Engineering Roboics VLSI Design Advances in OpoElecronics Navigaion and Observaion Chemical Engineering Acive and Passive Elecronic Componens Anennas and Propagaion Aerospace Engineering Modelling & Simulaion in Engineering Shock and Vibraion Advances in Acousics and Vibraion