A study of Adaptve Replcaton Technque n routng te-constraned essages (VoIP) n MANET SWATI SAXENA Research Scholar AIM & ACT Banasthal Vdyapth Banasthal, (Ra.) Dr. MADHAVI SINHA Asst. Prof. (HOD) Dept. of Coputer Scence & Engneerng BIT Mesra, Japur Extenson, Japur (Ra.) ABSTRACT Iposng the constrant of tely delvery on usual essages s referred as te-constraned essages. The utlty of such essages depends upon the te at whch they arrve at ther destnaton. Due to contenton aong randoly arrvng essages at each node/ln and ult-hops between the source and destnaton that a essage ust travel, t s dffcult to guarantee the tely delvery of such essages. In case of faulty lns or node, t s lely possble for essages to reach at requred destnaton n te. To handle such type of faults, the dea of sendng ultple copes or replcated copes s used. The probablty of successful essage delvery wthn te wll be ncreased through sendng ultple copes through dsont routes, as at least one copy wll be reached wthn te. But ths wll degrade the networ perforance due to congeston and other probles. Therefore, t s requre sendng only optu nuber of copes through dsont routes. Such copes are estated on the bass of networ load and essage deadlne. In ths paper, we present the ethod to estate the optu nuber of copes. The syste s desgned for developng replcaton strategy naed as Adaptve Fault Tolerance Replcaton Technque (AFTRT) for VoIP applcatons used for Moble Ad hoc Networ. The paraeters such as end-to-end latency, node delvery ndex and essage delvery probablty are deterne by the destnaton and send as feedbac to the source. Based on ths nforaton, source replcates the essage. The perforance study s conducted through sulaton usng NS2 sulator. Keywords: Deadlne, Fault Tolerance, Message Delvery Probablty, Node Delvery Index. 1. INTRODUCTION Tely delveres of nter-tas essages are the desred requste n real-te councaton. Usual essages, augented wth a new paraeter deadlne, are referred as te-constraned essages. If such essages arrve at ther destnaton after ther deadlne has expred, they ay cause syste falure. However, earler arrvals of such essages also degrade the syste perforance because ths requres bufferng overhead. To delver such essages tely, any approaches have been explored n the lterature. In ths paper we use the floodng approach n whch sae copes of the essages are transtted along dsont routes. Ths approach s naed as ultple-copy approach [1] or replcaton approach. (The ter ultple-copy and replcaton can be use nterchangeably). Sendng duplcate copy through dsont route wll ncrease the probablty of reachng at least one copy of the essage wthn te at ts destnaton. But t ay leads to congeston and ncreases the traffc load. It ay also possble that essages wth later deadlnes affect the essages wth shorter deadlne. So, t s benefcal to send only requred nuber of copes along dsont routes. Such nuber should be estated based on the current networ load and essage deadlne. Sendng out optal nuber of copes through dsont routes wll ncreases the probablty of reachng essages wthn ther deadlne and networ load wll rean under control. Aong nuerous applcatons of te-constraned essages over real-te councaton, we are usng voce applcatons runnng over the Internet. It s referred as Internet telephony or Voce over Internet Protocol (VoIP). VoIP s defned as the transport of voce as pacets n the IP based networ. The IP networ perfors ISSN : 0975-3397 Vol. 3 No. 6 June 2011 2273
varous functons such as collect callng, gateways nto the publc voce networs and assocated events [2]. In addton to the antenance of a relable voce councaton n a broadband envronent, the developent of the best cost-beneft relatonshp s ensured by the VoIP networ. Assurance for specfc qualty s provded by the QoS paraeter along wth the supportng provders and users to attan Servce Level Agreeent (SLA). Based on the codec used for the copresson technque, the preferred QoS and the channel capacty ndeed vary n the VoIP networ. In case of channel securty, the encrypton ethod can be changeable whch n turn changes the throughput and the copresson rate [3]. The paper s organzed as follows: In the next secton (secton 2), we gve the overvew of related wors fro the lterature about fault tolerant strateges opted by varous routng protocols (used for te-constraned essages explctly for VoIP data). In secton 3, we explore the proposed algorth. Based on the wor on secton 3, we frae out the proposed Adaptve Fault Tolerant Replcaton Technque (AFTRT) n secton 4. Secton 5 explores the sulaton paraeters used for perforance study of the proposed algorth and a coparatve study s gven. Secton 6 concludes our wor. 1.1 VoIP n MANET A oble ad hoc networ (MANET) s a oble, ulthop wreless networ that does not rely on any preexstng nfrastructure. Such networs are characterzed by dynac topologes due to uncontrolled node oblty, lted and varable shared wrless channel bandwdth and wreless devces controlled by battery power [4]. In [5], QoS requred for Real-te voce conversatons n IP networs s explored and the perforance of VoIP n wreless ad hoc networ s consdered. Unle the conventonal VoIP pleentaton on fxed nodes, VoIP s pleentng n MANET n the current scenaro that guarantee the user about conveyng the relablty of voce. Faults n MANET In order to exchange nforaton, a node requres ultple hops n the MANETs due to restrcted transsson power n the networ. Therefore, routng dscovery and antenance are the aor challenges to consder n MANETs [6]. Due to ssues le unpredctablty of envronent, unrelablty of wreless edu, resourceconstraned nodes and dynac topology, MANETs are susceptble to varous types of faults. These faults can be grouped as follows [7]: Transsson errors Due to hgh transsson contenton and congeston, the MANET systes suffer fro hgh transsson error rate. Therefore, t s portant to provde hgher relablty for broadcastng operaton under dynac MANET syste but t nvolves severe probles. In [8], a sple broadcast algorth s proposed to prove the delvery rato n an envronent that has hgh transsson error rate. Node or ln falures In ad hoc networs, pacet losses occur anly due to ln falures or node falures. Due to such falures, the effect on the networ perforance s based on varous factors. Aong such factors, one of the s the routng protocol. In [9], the behavor of routng protocol s studed n case of path falure and presented an proveent schee. Route breaages One of the portant causes of route falure s the node falure whch ay be because of power shortage. By deternng resdual power n a partcular node, the node falure can be predcted when the resdual power goes lower than threshold value [10]. Congested nodes or lns Congeston ay occur when certan nodes or lns ay over utlze. Ths ay lead to long delay or any pacet losses. In [11], a routng protocol for MANET s suggested that s aware and adaptve about networ congeston. 1.2 Proble defnton and Proposed soluton The proposed strategy s explored n [12] but t was desgned for wred networ and worhole swtchng technque was used. Such concepts are not sutable for current scenaro. The proposed strategy s developed for wreless networ. In [13], the contenton based routng n oble ad hoc networs wth ultple copes s proposed. It s based on spray-select-focus algorth where source sprays a few essage copes nto the networ. Neghbors of that ISSN : 0975-3397 Vol. 3 No. 6 June 2011 2274
node receve a copy towards the destnaton. The technque of selectng neghbors s not specfed n ths paper. In ost of the prevous wors based on replcaton technque assued that the replcaton nodes (.e. nodes responsble to replcate a essage) are statc. Unle statc nodes, the replcaton nodes should be adaptvely estates the nuber of copes based on the delvery probablty. Our obectve s to desgn a replcaton technque for VoIP data n MANET by estatng the nuber of copes to be replcated. Ths s estated by deternng the optu nuber of replcaton nodes. The proposed technque offers an adaptve fault tolerance ethod for replcaton of VoIP data pacets as the optu nuber of copes and the nubers of nodes responsble for replcaton process are changng accordng to the networ load n hop be hop anner. The proposed strategy s naed here as Adaptve Fault Tolerant Replcaton Technque (AFTRT). A essage wll be replcated at ether source node or at nteredate node. The source node replcates the ultple copes of the pacet and they are destnng to the nteredate nodes. The source antans a forward node counter (FNC) wth ntal value dependng upon the current networ load. Intally, the pacets are replcated and send to -hop neghborhood. Fro the neghbor nodes, the pacets focus to reach the destnaton. On reachng the destnaton node, the average delay taen by the pacet and the nuber of pacets successfully receved are send as feedbac to the source. If the nuber of pacets and the average delay are below soe threshold, then source ncreases the FNC sequentally untl the feedbac goes above the threshold. 2. Related Wors A store-and-forward based routng (e.g. ult-copy relayng) for Interttent Connected MANETs (ICMANs) s proposed n [14]. It replaces tradtonal routng technologes that were desgned for fully connected paths. However, such ult-copy relayng strateges do not adapt to the frequent varatons of the networ condtons as they rely solely on the source to control the relayng process. The proposed approach provdes an effcent way for pacet delvery. An adaptve codng control schee s proposed n [15], to prove the qualty of voce transport over wreless ad hoc networ. The basc dea used here s to adapt the voce codng bt rate to the avalable networ bandwdth so as to axze the voce qualty. The MOS standard s used here for appng between the pacet losses rates to the MOS value to control voce encodng bt rate. A robust and secure fraewor for voce transsson over ultpath MANET s proposed n [16]. The fraewor s robust not only aganst dynacally changng topology but also aganst adverse envronent. A securty analyss of the proposed schee and the perforance evaluaton of fraewor are provded. A fault tolerant ult path protocol to reduce the pacet loss due to route breaage s proposed n [17] whch uses a new route dscovery echans. Nodes deterne ultple dsont routes wth ore battery power and resdual energy, to every actve destnaton. In ths fault-tolerant echans, the receved sgnal strength s easured and based on ths value t can send warnng pacets to the prevous node. The AOMDV protocol s used as a base for the ultpath routng. In [18], a novel bo-nspred protocol naed as Eraoble (Epdeac based Relable and Adaptve Multcast for Moble ad hoc networs) s proposed that as to delver ultcast data relably wth nal networ overhead, even under adverse networ condtons. Wth an epdec-based ultcast ethod, t copes wth dynac and unpredctable topology changes due to oblty. Also the echans used doesn t requre antanng any tree-le or esh-le structure for ultcastng. Equpped wth any other dstngushed features, the protocol s capable to delver data relably n both sparse and dense networs. 3. Proposed Wor In ths secton, we descrbe the topology learnng, estaton of delay, end-to-end latency, Node Delvery Index, Message Delvery probablty. On the bass of these paraeters, an optal nuber of replcated copes wll be forulzed. 3.1 Topology Learnng Usng relable path based topology learnng technque, t s essental to enable the source to deterne the path towards the destnaton. Consderng the networ consstng of several nodes, the source learns the topology n order to deterne the possble effcent route to the sn by the followng technque: Intally the source sends a Topology Learnng request () essage to the nearest neghbors. The essage ncludes the nforaton related to the source d, destnaton d and wll also append the d of the nodes that forward the essage towards the destnaton. Thus as the essage oves towards the destnaton, t ncludes all the nforaton about the nodes nvolved n the transsson. On recepton of the essage at the sn, t responds by sendng a route reply () essage to the source through the reverse path. ISSN : 0975-3397 Vol. 3 No. 6 June 2011 2275
Based on the nuber of essages receved at the source, the source ncreents or decreents the forward node count (FNC). The source has a threshold value (threshold ) for the nuber of RREP essages and ntally the FNC s set to a value by the source. If nuber of < threshold, then FNC = FNC + 1 If nuber of > threshold, then FNC = FNC 1 (The FNC s the nuber of nodes that wll receve the replcated copes.) Fg. 1 shows the essage transsson fro the source to the sn through the nteredate nodes. Durng the transsson, request transtted across src-d---n-sn, src-a-f-l-sn and src-e-g---sn successfully reaches the sn. But the request transtted across src-b-h and src-c are lost n the networ. As the essage gets forwarded, the d of every node s appended n the essage. b h d n src a f l sn c e g Fg.1: Transsson of essage fro source to sn Fg. 2 shows the essage transsson fro the sn to the source. On recepton of the essage at the sn, t reples bac by sendng the essage to the source through the sae path used by the essage. The sn sends the essage through the path used for correspondng essage by usng the nforaton appended durng the transsson of the essage towards the sn. Out of the 5 essages sent by the source, the total nuber of responses receved s 3. The source on recevng the essage now copares t wth the total nuber of essages receved wth the threshold value set by t. If the threshold value set up by the source s 2 then snce the nuber of receved essage s greater than 2, the FNC s reduced by 1. ISSN : 0975-3397 Vol. 3 No. 6 June 2011 2276
b h d n src a f l sn c e g Fg.2: Transsson of essage fro sn to source Ths process contnues untl the nuber of receved essages at the source s greater than the threshold value threshold. The a of threshold s to pre-defne the axu nuber of reply receved before a fractonal part of deadlne. It s not possble to wat for as uch as reply those are recevng beyond deadlne. Its value should be set by source based on tas deadlne and essage deadlne. Here, we have used soe fxed value for the sae of splcty. 3.2 Estatng Delay We calculate the delay based on the probablty functon as follows. Intally we consder the delay n transsson to a node through one ln [routng path =1]. Let us consder two hops and n such that there exsts a ln l = (,n). Hence the transsson latency of the essage fro to n s characterzed by the rando varable defned as D and probablty dstrbuton functon (pdf) s defned as the P(D ) such that the probablty that ntroduces a delay of at ost n forwardng the essage to the next hop.e., p ( D ) ( ) = P.. (1) The te nterval between the essages enterng the node t n and acnowledge essage recepton fro the recever t ac s regarded as the transsson latency of a essageδ. But an extra te wll be consued for the recepton of the acnowledge essage and t s depcted by ρ. Hence transsson latency of the essage s gven by δ = t t ρ ac n.. (2) Eq (2) depcts the delay at only one hop for every essage, hence for the calculaton of the delay durng the transsson of the essage through several hops.e., throughout ts path fro source to destnaton. Therefore the total delay n the essage transsson s obtaned by the suaton of the delay nvolved n every hop ISSN : 0975-3397 Vol. 3 No. 6 June 2011 2277
durng the transsson fro the source to the destnaton. For every hop, a partcular value of ρ s obtaned, so overall an ncreasng sequence of value ρ, ρ 0 1,..., ρ s generated. Each of these values represents a saple of the rando varable D for the gven hop. If the dstrbuton of D was nown, fndng out the expected value and ts probablty or the probablty of any gven delay would be reduced to trval. But f the value of D s not nown then the saple ean and saple varance whch characterzes the dstrbuton should be used for calculaton. 3.3 Estatng End-to-end latency The end-to-end latency of a path rp (say) s also a rando varable, D rp, fored by the coposton of the delays of ts nteredate lns: = Drp D rp p Drp (, ) (, ) ( τ ) P( τ ) = D rp (3) (4) Eq (4) ndcates the probablty for end to end latency where τ represents the at ost value of end-to-end latency ntroduced n forwardng the essage. 3.4 Estatng Node Delvery Index () The data transsson s based on the paraeter Node Delvery Index (). It s the probablty that a node correctly delvers data to the destnaton. s useful n selectng routes based on ts value. Let denotes the delvery ndex of a node N. The value of ntally set to zero and updated whenever there s a essage transsson or ter expraton. Whenever a node N transts a data pacet to another node N, should be updated such that K = ( 1 ρ ) + λ, (5) Where s the of node N and 0 < λ < 1 s a constant eployed to eep partal eory of hstorc status. Clearly, s always between 0 and 1. So f N s the destnaton, = 1, snce the essage s already delvered to the destnaton successfully, otherwse, < 1. Also each node antans a ter t 1. If there s no essage transsson wthn an nterval of, then the ter t 1 expres. The ter expraton ndcates that the node couldn t transt any data durng. So should be updated as = ( 1 λ) (6) So fro (1) and (2), we arrve that, the node delvery ndex ( ) of node N s updated as = ( 1 λ ) + λ, f there s a data Transsson = ( 1 λ), f there s a Teout.. (7) 3.5 Estatng Message Delvery Probablty (MDP) The Message Delvery Probablty (MDP) s used to represent the aount of redundancy and to ndcate the portance of a gven essage. The MDP s defned to be the probablty that at least one copy of the essage ISSN : 0975-3397 Vol. 3 No. 6 June 2011 2278
s delvered to the destnaton wthn te n the networ. We assue that each essage copy carres a feld to denote the MDP. When a essage s generated, ts MDP s ntalzed to be zero. Let us consder a node N whch s transttng copes of a data essage. Let MDP denote the MDP of essage n the ready queue of essages n node N. The essage transtted to a node N r s assocated wth a MDP of = 1 and the MDP of the essage at node N s updated as MDP Nr = 1 (1 MDP ) (1- ) (1 ) (8) 1 (1 MDP ) (1- MDP = 1 MDP MDP =.. (9) The above process repeats at each te when essage s transtted to another node. In general, the ore tes a essage has been forwarded, the ore copes of the essage are created, thus ncreasng ts delvery ndex. As a result, t s assocated wth a larger MDP. Further, MDP s used to estate the nuber of copes to replcate. 4. Algorth of Proposed Adaptve Fault Tolerance Replcaton Technque (AFTRT) The algorth for our fault tolerant replcaton technque n the MANET consders the latency nvolved and the data delvery rato n the networ. The steps nvolved n the algorth are as follows: Step 1: The source (S) n need of servce sends a Topology Learnng Request () to all ts nearest neghbors (N). Ths essage s then forwarded by the neghbor node to ts neghbor and t contnues tll the destnaton (S ) s reached. ) S N 1.. N n S Step 2: In response to the request, the destnaton sends the Topology Learnng Reply () essage through the path used for correspondng essage by usng the nforaton appended durng the transsson of the essage towards the sn. S N 1.. N n S Step 3: Based on the nuber of the essages receved, the source ncreents or decreents the forward node count (FNC). The source antans a threshold value (threshold ) for the nuber of essages and ntally the FNC s set to a value by the source. If nuber of < threshold, FNC = FNC + 1 else FNC = FNC 1 p Step 4: For any ln l = (,n), probablty ( ) essage to the next hop s calculated as p ( ) = P ( D ) that ntroduces a delay of at ost n forwardng the where D s the rando varable whch characterzes the transsson latency of the essage fro to n ISSN : 0975-3397 Vol. 3 No. 6 June 2011 2279
Step 5: Transsson latency of the essage,δ s gven by δ = t t ρ ac n where t n s the te at whch the essage enters the node t ac s the te at whch the acnowledge essage s receved fro the recever ρ s the extra te consued for the recepton of the acnowledge essage Step 6: The end to end latency, Drp = of the path rp (say) s represented as Drp D(, ) (, ) rp Step 7: The probablty of end to end latency s represented as p D rp ( τ ) P( τ ) = D rp Step 8: The Node Delvery Index ( ) represents the probablty that a node correctly delvers data to the destnaton. If there s a data transsson = ( 1 λ ) + λ else = ( 1 λ) Step 9: The MDP represents the probablty that at least one copy of the essage s delvered to the destnaton by other nodes n the networ. The essage delvery probablty (MDP) of essage n the queue of node N, whch s transttng copes to node Nr s gven by MDP Nr = 1 (1 MDP ) = 1 (1 and the MDP of the essage at node N s updated as MDP = ) (1-1 (1 MDP ) (1- MDP = 1 MDP Step 10: The destnaton deternes the overall Node Delvery Index ( ) along the path by addng the values at every node. = n = 1 11] The calculated end to end latency, MDP and the values of the path are transtted as a feedbac to the source by the destnaton. Based on ths feedbac, FNC value s ncreented or decreented by the source accordngly. The source antans a threshold for the latency value threshold LATENCY and threshold for the as threshold If Drp > threshold LATENCY and Else FNC = FNC +f FNC = FNC f ) < threshold Where f s the scale factor used to ncreent or decreent. ) ISSN : 0975-3397 Vol. 3 No. 6 June 2011 2280
Then the source replcates the pacets accordng to the value of MDP and transts to FNC nuber of nodes. Nuber of replcated copes = MDP θ Where θ s the nuber of edate neghbors of node N. Thus, the algorth consders the latency and the node delvery ndex value to replcate the pacet n order to effcently enable data transsson through relable paths. 5. Sulaton In ths secton, we conduct the sulaton-based perforance evaluaton of the proposed AFTRT schee. 5.1 Defnng paraeters for sulaton odel The perforance of AFTRT schee s evaluated through NS2 [19], networ sulator. The channel capacty of oble hosts s set fxed as 2 Mbps. For wreless LAN, the dstrbuted coordnaton functon (DCF) of IEEE 802.11 s used here as MAC layer protocol. Moble nodes are assued to ove n a 1000 1000 regon for 100 seconds sulaton te. We perfored the analyss by varyng the networ sze as 25, 50, 75 and 100 nodes and the pause te for the oble node s assued to be 10 seconds. Nodes are randoly dstrbuted n squared area. For the sae of splcty, the speed of the oble node s set fxed as 5 /s. All the nodes are assued to have the sae transsson range of 250 eters. Each and every node ay behave le a source or an nteredate or a recevng node. The data traffc used here for sulaton s VoIP. The VoIP Codec used s GSM.AMR, VoIP odel used s one-to-one, VoIP Encoder used s Applcaton/VoIPEncoder and VoIP Decoder used s Applcaton/VoIPDecoderOptal. The pacet sze s consdered as 512 b and rate of transsson s taen as 250 b whch can changes to 500 b for another run of experent. The nuber of VoIP fraes per pacet used s 2 and nuber of flows consdered are 5. Paraeters defned n algorth are ntalzed as: λ=0.2, = 1, = 0. 5.2 Perforance etrcs used for sulaton The perforance of the AFTRT schee s studed regardng ts ablty to adapt the networ condtons and copared wth the Adaptve Mult-Copy Routng (AMR) [18] protocol. To show that the proposed wor s relable and ore pacets delvered wthn te, t s requred to choose soe paraeters. We use four etrcs as overhead, average end-to-end delay, Pacet delvery rato and drop. Overhead ay be defned as the total nuber of routng pacets noralzed by the total nuber of receved pacets. The average end-to-end delay s averaged over all survvng data pacets fro source to the destnatons. The pacet delvery rato s the rato of the nuber of pacets receved successfully and the total nuber of pacets sent. Drop s defned as the nuber of pacets drop or not receved wthn te. We perfor the analyss n two segents.e. these etrcs are studed for two protocols on the bass of varyng the nuber of nodes and on the bass of changng the rates. 5.3 Results A. Based on varyng the nuber of nodes Frst, we ll study the perforance of etrcs on varyng the nuber nodes. The dea behnd varyng the nuber of nodes s to study the perforance when networ sze ncreases gradually (not nfntely). Fg. 3 represents the pacet delvery rato for the two protocols. It s clear that AFTRT acheves good delvery rato as copared to AMR. Fg. 3: Nodes vs. Delvery rato ISSN : 0975-3397 Vol. 3 No. 6 June 2011 2281
Fg. 4: Nodes vs. Delay Fg. 4 shows that average end-to-end delay of the proposed AFTRT protocol s less than copared to the AMR protocol. Fg. 5: Nodes vs. Drop Fg. 5 shows the result of drop etrc, whch shows that AFTRT schee has lower drop rate than that of AMR schee. Fg. 6: Nodes vs. overhead Fg. 6 ensures us that the overhead s less for AFTRT than that of AMR. B. Based on varyng the Rate Now, n the second experent we study the behavor of etrcs when rate s vared as 250, 300, 350, 400, 450 and 500. Fg. 7 shows the delvery rato for the two protocols under varyng the rates. It s clear that AFTRT acheves good delvery rato as copared to AMR protocol agan. ISSN : 0975-3397 Vol. 3 No. 6 June 2011 2282
Fg. 7: Rate vs. Delvery rato Fg. 8: Rate vs. Delay Fro fg. 8, we can observe that the average end-to-end delay of the proposed AFTRT protocol s less copared to the AMR protocol. Fg. 9: Rate vs. Drop Fg. 9 represents the result of pacet drop. Agan AFTRT schee perfor superor to AMR schee regardng drop etrc. ISSN : 0975-3397 Vol. 3 No. 6 June 2011 2283
Fg. 10: Rate vs. overhead Fro fg. 10, we can ensure that the overhead s less for AFTRT as copared to AMR. 6. Concluson Tely delvery of te-constraned essages s the foreost requreent for real-te councaton applcatons. To ae routng as fault tolerant, we proposed a ultple-copy or replcaton strategy that s adaptve. In ths paper, we proposed Adaptve Fault Tolerance Replcaton Technque (AFTRT) to replcate the data pacets wth fault tolerance. Adaptve n the sense that t adapts current networ condtons to estate the nuber of replcate copes. Intally, the source n need of servce perfors topology leanng by ssung the topology learnng request (). After the recepton of the topology learnng reply () fro the destnaton, the source calculates the delay and node delvery ndex (). One the for a partcular par of ln s deterned, and then overall s calculated fro source to destnaton. The source antans the threshold for the latency and values. On estatng the latency and the overall value, the source copares t wth the threshold value and then the FNC value s ncreented or decreented accordngly. Ths process s contnued untl the threshold value s reached. The source also estates the essage delvery probablty (MDP). Then t replcates the pacets accordng to the value of MDP and transts to FNC nuber of nodes. 7. References [1] K.G. Shn, C. M. Krshna, Real-te Systes, Tata McGraw Hll Publcatons [2] Eduardo B. Fernandez, Juan C. Pelaez and Mara M. Larrondo-Petre, Securty Patterns for Voce over IP Networs, Vol 2, No.2, August 2007 [3] Edward Paul Gullen, Dego Alandro Chacon, VoIP Networs Perforance Analyss wth Encrypton Systes, World Acadey of Scence, Engneerng and Technology 2009. [4] Mahesh K. Mshra, Sar R. Das, Ad hoc on-deand ultpath dstance vector routng, Wreless Councatons and Moble Coputng, 2006 pp. 969-988, publshed onlne n Wley InterScence. [5] Hu Yao Zhang, Mare E. Balows, Garry A. Ence and John Hoer, An Extended AODV Protocol for VoIP Applcaton n Moble Ad Hoc Networ, Transactons on Electrcal, Engneerng, Electroncs and Councatons ECTI, vol 7, No. 2, August 2009. [6] Fuan Qn and Youyuan Lu, Multpath Based QoS Routng n MANET, Journal of Networs, vol. 4, no. 8, October 2009. [7] Stephen Mueller, Rose P. Tsang, Dpa Ghosal, Multpath Routng n Moble Ad hoc Networs: Issues and Challenges, Lecture Notes n Coputer Scence, vol. 2965, Aprl 2004, pages: 209-234. [8] R. Balarshna and U. Raeswara Rao, G. A. Raachandra, Relablty n MANET s Usng Enhanced Double Coverage Broadcastng (EDCB), Internatonal Journal of Advanced Networng and Applcatons, vol 01, Issue 03, pg: 147 153, 2009. [9] X. Hou and D. Tpper, Ipact of Falures on Routng n Moble Ad Hoc Networs Usng DSR, Proceedngs of Councaton Networs and Dstrbuted Systes Modelng and Sulaton Conference, January 2003, Orlando, FL. [10] M. A. Razzaque, Son Dobson and Paddy Nxon, Cross layer self routng: a self-anaged routng approach for MANETs, In proceedngs of the 4 th IEEE Internatonal Conference on Wreless and Moble Coputng, Networng and Councatons. IEEE Press. pp 284-290, 2008. [11] Duc A. Tran, Harsh Raghavendra, Routng wth Congeston Awareness and Adaptvty n Moble Ad hoc Networs, Journal of IEEE Transactons on Parallel and Dstrbuted Systes archve, vol 18, Issue 7 July 2007. [12] Swat Saxena, An Approach to Fault-tolerant Routng n te-constraned essages (usng ultple-copy), Natonal Conference on Advanceents ninforaton & Councaton Technology (NCAICT - 2008), organzed by Coputer Socety of Inda, Allahabad Chapter, March 15-16 2008. [13] Ms. E. Jenefa Jeba Joth, Dr. V. Kavtha and Ms. T. Kavtha, Contenton Based Routng n Moble Ad Hoc Networs wth Multple Copes, Internatonal Journal of Engneerng and Technology vol.2(2), pp.93-96, 2010 [14] Zhoqun L, Lngfen Sun, Eanuel C. Ifeachor, Adaptve Mult-Copy Routng for Interttently Connected Moble Ad Hoc Networs, In Proceedngs of GLOBECOM 2006. [15] Hongq Zhang, Jyng Zhao and Olver Yang, Adaptve Rate Control for VoIP n Wreless Ad Hoc Networs, In Proceedngs of GLOBECOM 2006. ISSN : 0975-3397 Vol. 3 No. 6 June 2011 2284
[16] Bnod Vadya, Meso K. Deno and Joel J.P.C. Rodrgues, Secure Fraewor for Voce Transsson Over Multpath Wreless Ad-Hoc Networ, Proceedngs of the 28 th IEEE conference on Global telecouncatons (GLOBECOM), pp. 4299-4304, 2009. [17] G. Rauar, K. Durasay, A Fault Tolerant Multpath Routng Protocol to Reduce Route Falures n Moble Ad Hoc Networs, European Journal of Scentfc Research, Vol.50, no. 3, pp. 399-408, 2011. [18] Öznur Özasap, Zülüf Genç, Ere Atsan, Epdec-based relable and adaptve ultcast for oble ad hoc networs, Journal of Coputer Networs 53(9): pp 1409-1430 (2009). [19] Networ Sulator, http://www.s.edu/nsna/ns AUTHOR S PROFILE: Swat Saxena Research Scholar AIM & ACT, Banasthal Vdyapth, Banasthal, (Ra.) Dr. Madhav Snha She receved her Ph.D. degree fro Banasthal Vdyapth, Banathal (Ra.) and currently worng as Assocate Professor and Head of Departent (Coputer Scence & Engneerng) at Brla Insttute of Technology, Japur (Ra.). She has 25 years of teachng experence and publshed any Natonal and Internatonal publcatons. Her research nterests are Artfcal Intellgence, soft and language coputng. ISSN : 0975-3397 Vol. 3 No. 6 June 2011 2285