A QoS balacig model for Mobile Ad hoc Networks HENG LUO, DAVID. I. LAURENSON, AI-HUANG GUO, JIANPING CHEN, PEIDE YANG, HUA JIANG Suzhou key laboratory of mobile etworkig ad applied techologies SuZhou Uiversity of Sciece ad Techology 1 # KeRui Road SuZhou, JiagSu, CHINA luoheg1981@163.com Abstract: -Quality of Service (QoS) provisio is required for MANETs i recet years to support the rapid growth of video i mobile traffic. I spite of umerous QoS routig s, litter literature, it is observed, discusses about the QoS balacig model with several QoS metrics cosidered simultaeously, which is realistic i real applicatios, due to various uits ad attributes of QoS metrics. A ew model, deoted by SAW-AHP, which is a combiatio of Simple Additive Weightig (SAW) ad Aalytic Hierarchical Process (AHP), is proposed i this paper to balace the competig QoS metrics. Despite of differet uits ad attributes of diverse QoS metrics, SAW-AHP is able to rak the alterative routig s reliably ad cosistetly by cosiderig the relative preferece of QoS metrics which is eglected by much literature. Key words: -Mobile ad hoc etwork; Routig Protocol; QoS; SAW-AHP; Mobile traffic; Simulatio 1 Itroductio A mobile ad hoc etwork (MANET) is a autoomous system of mobile odes that are free to move about arbitrarily. The possibility of establishig such etworks i places such as disaster relief sites ad coferece rooms, which are characterized by lack of prefixed ifrastructure, justifies the developmet of MANETs. QoS provisio was ot cosidered iitially i MANETs. However, with the rapid growig of video traffic which exceeds 50% of mobile traffic [1], QoS guaratee becomes ecessary, leadig to the existece of a collectio of QoS-aware routig s [2][3]. However, a umber of QoS metrics should be guarateed simultaeously i real applicatios ad some algorithms, it is observed, are able to support oly part of QoS metrics o the cost of others, resultig i the radom selectio of s i MANETs. I this paper, a ovel QoS metrics balacig model, deoted by SAW-AHP, which is a combiatio of the Simple Additive Weightig (SAW) [4] ad the Aalytic Hierarchical Process (AHP) [5], is proposed to balace QoS requiremets by cosiderig the relative importace of differet QoS metrics which is eglected by much literature. Extesive simulatios are performed to validate the efficiecy as well as reliability of the SAW-AHP model. For simulatio, versio 2.32 of the wellkow ope-source software NS-2 is used. This paper is orgaized as follows. Sectio 2 discusses the problems after simulatios. The third part provides the problem solutio method ad the fial sectio cocludes this paper. 2 Problem Formulatio The etwork performace of several mobile termials (MTs), which share a commo access poit to access the Iteret, as show i Fig.1, is studied as a example for the SAW-AHP model. The simulatio cofiguratios as well as results are itemized i Table 1 ad Table 2 respectively. Fig.1 Simulatio sceario E-ISSN: 2224-2864 372 Volume 13, 2014
Table 1. Parameter Simulatio time Simulatio parameters Descriptio 3000s Simulatio rus 50 Number of odes 32 Node mobility patter Trasmissio rage Routig Traffic load Topology PDR (%) Table 2. Simulatio results Delay (ms) Radom Way Poit Model Jitter (ms) 25m ad 2 streams 100m*100m Thrput (Mb/s) EC (J/pkt) 94.7 1.98 2.41 3.68 0.73 99.1 2.68 2.91 3.38 0.214 PDR: packet delivery ratio; Thrput: throughput; EC: eergy cost 2.1 Simulatio results ad aalysis As show i Table 2, outperforms i terms of packet delivery ratio ad eergy cost. Three factors cotribute to the success of. To begi with, iitiates the route discovery mechaism oly whe ecessary, avoidig the use of stale routes as well as periodic routig iformatio broadcast. Secodly, if a lik breaks dow i the data trasmissio process, the upstream ode may buffer the lost packets ad activate the local lik repair mechaism which icreases the umber of data packets that are able to be delivered. Last, but ot the least, broadcasts route iformatio packets periodically ad those packets may collide with data packets. However, outperforms i other three metrics, delay, jitter ad throughput. The key reaso for this is the proactive ature of. is able to establish route much more quickly by searchig routig tables which is updated periodically. Istead, iitiates a route discovery process o demad which takes more time. 2.2 Problem statemet For a etwork operator who strives to offer reliable packet delivery service, is better a solutio compared to. O the cotrary, for a time sesitive applicatio, is preferred. However, whe the umber of QoS-metrics icreases, just as i may real applicatios, the problem becomes much more complicated [6]. 3 Problem Solutio The proposed SAW-AHP model ivolves maily two steps as show i Fig.2. Hierarchical structure compositio Weights computatio Performace evaluatio Fig.2 Dyamic switch Adaptive process SAW-AHP diagram 3.1 Performace evaluatio Optimal selectio The performace evaluatio process ca be further divided ito three steps ad the first step is to decompose the decisio problem ito a hierarchy structure, composed of a objective layer, a criteria layer ad a alterative layer so that a hard problem ca be more easily uderstadable. 3.1.1 Hierarchy structure The objective i this paper is to balace five QoS metrics ad thus fid the optimal routig give the preferece of a umber of QoS metrics which are treated as criteria. For simplicity but without loss of geerality, two alteratives, which is a proactive routig, ad, which is a reactive are selected. Fig.3 shows the hierarchy structure with three layers, the objective layer, criteria layer ad alterative layer. Objective layer Criteria layer Alterative layer Rak alterative s PDR Delay Jitter Thruput EC Fig.3 Hierarchy structure 3.1.2 Weights computatio Oce the hierarchy is built, the decisio makers compare elemets i a pair-wise fashio with predefied rules based o which the compariso matrices are obtaied. 3.1.2.1 Weight for QoS metric A decisio maker is assumed to be able to compare ay two elemets, say E i ad E j, at the same level of the hierarchy structure ad provide a umerical value e ij accordig to his/her preferece, e ij > 0 for ay i=1,2,, ad j=1,2,,. The reciprocal E-ISSN: 2224-2864 373 Volume 13, 2014
property e ji =1/ e ij holds. The fudametal scales for pair-wise compariso could serve as a good basis ad they are itemized i Table 3. Table 3. Importace ad defiitio Importace Defiitio Explaatio 1 Equal importace Two elemets cotribute equally to the objective 3 Experiece ad Moderate judgmet slightly favour importace oe elemet over aother 5 Strog importace Experiece ad judgmet strogly favour 7 9 Very strog importace Extreme importace oe elemet over aother Oe elemet is favoured very strogly over aother; The evidece favourig oe elemet over aother is of the highest possible order of affirmatio Itesities of 2,4,6 ad 8 ca be used to express itermediate values. Prior to obtaiig the pair-wise compariso matrix for criteria, several assumptios are made for the relative importace of criteria i this paper. They are as follows: (I)Packet delivery ratio is moderately more importat tha delay; (II)Packet delivery ratio is moderately more importat tha jitter; (III)Packet delivery ratio ad throughput are equally importat; (IV)Packet delivery ratio is moderately more importat tha eergy cost; (V)Delay ad jitter are equally importat; (VI)Delay ad eergy cost are equally importat; (VII)Jitter ad eergy cost are equally importat; (VIII)Throughput is moderately more importat tha delay; (IX)Throughput is moderately more importat tha jitter; (X)Throughput is moderately more importat tha eergy cost. Oe thig to ote is that these parameters are applicatio depedet ad the choices here are for a specific applicatio sceario. Accordig to Table 3, the above 10 assumptios lead to the compariso matrix for criteria as follows PDR Delay C= Jitter Thrput EC PDR Delay Jitter Thrput EC 1 3 3 1 3 1/3 1 1 1/3 1 1/3 1 1 1/3 1 1 3 3 1 3 1/3 1 1 1/3 1 where PDR, Thrput ad EC deote packet delivery ratio, throughput ad eergy cost respectively. There are several methods to derive weights from a compariso matrix of which Geometric Mea Method (GMM) is a straight forward ad reliable (1) alterative [7]. I GMM, the ormalized weight is computed firstly via 1 1 ω = ( ) / ( ) i a ij a ij j= 1 i =1 j= 1 (2) where a ij (i,j=1,2,,) deotes the value of ij th elemets i compariso Matrix (1) ad is umber of elemets i the row. Combiig Eq. (2) with Matrix (1), the ormalized weights for criteria are obtaied i Table 4. Table 4. Normalized weights for criteria Criteria PDR Delay Jitter Thruput EC Weight 0.333 0.111 0.111 0.333 0.111 As observed, the weights for packet delivery ratio ad throughput are equal, idicatig the same importace of those two metrics. Delay, jitter ad eergy cost have the same weight which accouts for oe third of that for packet delivery ratio, revealig that they are less importat compared to packet delivery ratio. Qualitatively, a that has a better performace i terms of packet delivery ratio ad throughput is more likely to be selected. A decisio maker may give icosistet judgmets for the compariso matrix ad therefore SAW-AHP is desiged with capability of measurig the cosistecy based o the idea of cardial trasitivity. A matrix M is cosistet if ad oly if a ik a kj = a ij, where a ij is the ij th elemet i Matrix (1). However, this coditio ca rarely be satisfied i practice, especially i scearios with a large umber of criteria or alteratives. The violatio level of cosistecy chages with perso or cotext. I SAW-AHP, a metric Cosistecy Ratio (C.R.), developed by Satty [5], is employed to idicate the extet to which the cosistecy is violated as follows 1 ( Cω) i ( )/[( 1) ( RI..)] > 2 CR.. = i = 1 ωi 0 = 1, 2 where C ad ω i deote the pair-wise compariso matrix ad weight for the i th elemet respectively, represets the umber of elemets ad R.I. is the radom idex of a pair-wise compariso matrix that depeds o the umber of elemets i the matrix as itemized i Table 5. Table 5. Radom icosistecy idex (R.I.) Number of elemets 3 4 5 6 7 Radom Idex (R.I.) 0.58 0.90 1.12 1.24 1.32 (3) E-ISSN: 2224-2864 374 Volume 13, 2014
The C.R. of Matrix (1) equals 0, idicatig that Matrix (1) is cosistet. 3.1.2.2 Weight for alteratives Istead of usig scales i [5], simulatio results obtaied i Table 2 are employed to costruct the pair-wise compariso matrices for alteratives for the sake of accuracy. However, the attributes ad uits of metrics are differet. Table 6 summarizes the attributes of metrics i this paper. As see, two metrics, packet delivery ratio ad throughput, are grouped ito the the larger the better category while the other three metrics, delay, jitter ad eergy cost, are allocated to the the smaller the better category. Table 6. Metric ad attributes Metric PDR Thruput Delay Jitter EC Attribute the larger the better the smaller the better I SAW-AHP, The value of the correspodig elemet i the pair-wise compariso matrix for alteratives equals (4) orm a = d / d ij i orm j orm where di = di/max{ di} for metrics that are the orm larger the better ad d i =mi{ di}/ di for the parameters that are the smaller the better. Criterio Table 7. Weights for alteratives Weights PDR Delay Jitter Thruput EC 0.489 0.575 0.547 0.521 0.227 0.511 0.425 0.453 0.479 0.773 Table 7 itemizes the weights for alteratives uder differet metrics. As see, has larger weights i terms of packet delivery ratio ad eergy cost, idicatig its better performace over i those two metrics. O the cotrary, the weights for exceed those for i three other metrics, revealig s better performace i delay, jitter ad throughput. Sice there are oly two elemets i the compariso matrices for alteratives, those matrices are cosistet. 3. Sythetic weights The fial step is to sythesize the weights for criteria via (5) sω = cω ( i, j = 1,..., ) j i ij i= 1 where sω j deotes the sythetic weights for the j th alterative, c i symbolize weights for the i th metric ad ω ij represets the weight for the j th alterative uder the i th metric. The alterative with the largest sythetic weight is cosidered to the optimal oe. Table 8. Sythetic Weights Protocol Sythetic weight Rakig order 0.49 0.51 > As show i Table 8, the weight of is larger tha. Therefore, is preferred i this case. 3.2 Adaptive process Based o the rakig order, is selected i this case. Four sets of simulatios are carried out as show i Fig.4 to validate the reliability as well as efficiecy of the adaptive process. Referece sim#1 sim#2 Fig.4 Referece sim#3 sim#4 Simulatios for validatio As show, both sim#1 ad sim#3 cotiue to employ the same whereas the other two switch to a differet. Sim#1 ad sim#2 are combied to determie the effect of switch from to whereas sim#3 ad sim#4 are combied to reveal the effectiveess of the switch to. The results are itemized i Table 9. Table 9. Simulatio results metric sim#1 sim#2 sim#3 sim#4 PDR (%) 94.7 99.1 99.1 94.8 Delay (ms) 1.98 2.68 2.68 1.99 Jitter (ms) 2.41 2.91 2.91 2.41 Thruput (Mb/s) 3.68 3.38 3.38 3.68 EC (J/pkt) 0.73 0.214 0.214 0.72 3.3 Performace Improvemet Ratio A metric, the performace improvemet ratio deoted by PIR, is developed to specify the level of differece betwee two alteratives uder certai metrics. PIR is defied as the quotiet of the differece betwee the referece ad target s for a value of the referece. For E-ISSN: 2224-2864 375 Volume 13, 2014
metrics that are the larger the better, PIR ref-tar is computed via PIR P P P target referece target ref tar = = 1 (6) P P referece referece where P target ad P referece deote the performace of the target ad referece s respectively. For the smaller the better metrics, PIR ref-tar is 1/ Ptarget -1/ Preferece PIRref tar = = P / P 1 (7) referece target 1/ P referece A positive PIR suggests the performace improvemet while a egative oe reveals the deterioratio. PIRs may be aggregated by cosiderig the weights for metrics i a applicatio via (8) AIRi = ci PIRi where AIR i deotes the aggregated improvemet ratio for the i th metric ad c i deotes the weight for i th metric. AIR reflects the impact of performace improvemet/deterioratio of a metric o the overall QoS satisfactio. AIRs are sythesized to obtai the sythetic improvemet ratio idex (SIRI) SIRI = AIR (9) i=1 A positive SIRI is desired sice it idicates system improvemet whe a target is selected. O the cotrary, a egative SIRI reveals performace deterioratio if the target is selected. Referece sim#1 sim#2 SIRI=0% SIRI=20.8% Fig.5 SIRI results i Referece sim#3 sim#4 SIRI=0% SIRI=-0.14% As see i Fig.5, a positive SIRI is achieved which demostrates the effectiveess of the switch from to. O the cotrary, whe replaces the origial, the overall performace deteriorates. Therefore, it is cocluded that is more suitable for the case of 2 traffic streams, which is idetical with results i Table 8. 4 Coclusio I spite of various attributes ad uits for differet QoS metrics, the proposed SAW-AHP is able to balace competig QoS metrics ad thus rak alterative s ad efficietly ad reliably. Based o the performace evaluatio results, the system is able to switch to the optimal adaptively. Extesive simulatios show that the performace of the whole etwork may improve as much as 20.8% by adoptig the SAW- AHP model. Despite oly oe case beig studied i this paper usig the SAW-AHP method, it is geeric to other cases with differet QoS requiremets. 5. Future work The SAW-AHP model is appropriate for scearios where the decisio maker is certai about his/her preferece o the performace metrics ad oly the average value is cosidered. I the future, the SAW- AHP model will be fuzzified to icorporate the stadard deviatio of simulatio results as well as the ucertaity of the decisio maker. Ackowledgemet The project was fuded by the State Key Laboratory of Advaced Optical Commuicatio Systems Networks ad Suzhou Sciece ad Techology Fud (SZS201304). Referece [1] Cisco visual etworkig idex: global mobile data traffic forecast update, 2012-2017, http://www.cisco.com/e/us/ solutios /collateral/s341/ s525/s537/s705/s827/ white_paper_c11-520862.pdf, 2013, pp.1-30. [2] Bakht, H., Survey of Routig Protocols for Mobile Ad-hoc Network, Iteratioal Joural of Iformatio ad Commuicatio Techology Research, vol.1 No. 6, 2011, pp.258-270. [3] Che, W. et al., A survey ad challeges i routig ad data dissemiatio i vehicular ad hoc etworks, Wireless Commuicatios ad Mobile Computig, vol.11 No.7, 2011, pp.787-795. [4] Wag Y., A simple additive weightig method for timeseries multi-idices decisio makig ad its applicatios, Joural of systems egieerig ad electroics, vol.10, 1999, pp.21-26. [5] Saaty, T. L. How to make decisio: The Aalytic Hierarchy Process, Europea Joural of Operatioal Research, vol.48, 1990, pp.9-26. [6] Papadimitriou C. H., Computatioal Complexity, 1 st ed., Addiso-Wesley, USA, 1993. [7] Xu Z., O cosistecy of the weighted geometric mea complex judgemet matrix i AHP, Europea Joural of E-ISSN: 2224-2864 376 Volume 13, 2014
Operatioal Research, vol.126 No.3, 2000, pp.683-687. E-ISSN: 2224-2864 377 Volume 13, 2014