Research Article Performance Analysis of AODV, OLSR and GPSR MANET Routing Protocols with Respect to Network Size and Density

Transcription

1 Research Joural of Applied Scieces, Egieerig ad Techology (4): , 205 DOI: /rjaset..794 ISSN: ; e-issn: Maxwell Scietific Publicatio Corp. Submitted: April 9, 205 Accepted: May 2, 205 Published: October 05, 205 Research Article Performace Aalysis of AODV, OLSR ad GPSR MANET Routig Protocols with Respect to Network Size ad Desity Muthaa Najim Abdulleh ad Salma Yussof College of Iformatio Techology, Uiversiti Teaga Nasioal, Malaysia Abstract: The aim of this study is to compare the performace of Optimized Lik State Routig (OLSR), Ad hoc O-demad Distace Vector (AODV) ad Greedy Perimeter Stateless Routig (GPSR) routig protocols with respect to etwork size ad desity. Each of these protocols represets the three categories of MANET routig protocols which are proactive, reactive ad geographical routig protocol, respectively. The evaluatio was doe through simulatio ad the performace was measured i terms of throughput, average Ed-to-Ed (E2E) delay, Packet Delivery Fractio (PDF) ad Normalized Routig Load (NRL). The results of the simulatios show that the GPSR protocol is superior to OLSR ad AODV i most cases. The results also show that throughput, ed-to-ed delay ad packet delivery fractio are largely affected by the etwork size, while ormalized routig load is largely affected by the umber of odes i the etwork. Keywords: AODV, GPSR, MANET, OLSR, routig protocols, simulatio INTRODUCTION Mobile Ad hoc Network (MANET) is a type of etwork that does ot require a fixed ifrastructure (Coti ad Giordao, 204). Istead, it cosists of a group of odes that ca trasmit ad receive data amogst themselves. A importat characteristic of MANET odes is that they ca move radomly but ca still commuicate with oe aother at ay time. Aother importat characteristic of MANETs is that routes betwee two hosts might compose of hops through aother host i the etwork. Give a coditio where a seder ode is beyod the trasmissio rage whe data trasmissio is iitiated, commuicatio ca still occur if there are hosts betwee the sedig ad receivig odes that are willig to forward data packets to the receiver ode (Elgohary et al., 204). This is kow as multi-hoppig, a distictive characteristic of MANET. Routig protocols for MANETs are desiged to provide a route betwee odes. MANET routig protocols is categorized ito 3 mai groups: proactive, reactive ad geographical routig. Proactive routig protocols or table-drive routig protocols have a up-to-date topological map that eables latest routes to be maitaied. Therefore, whe a ode eeds to trasfer data to other ode, destiatio of the path is readily kow ad ca be immediately used. I reactive routig protocols, the topological map is ot updated ad thus routes through the etwork are ot maitaied. A additioal procedure, called a route discovery procedure, eeds to be carried out before data packets ca be trasmitted. This ca be accomplished by sedig a query to the etwork (Komai et al., 204). Protocol for reactive route is also kow as o-demad routig protocol. The geographical routig protocol is suited to sesor etworks that use locatio iformatio to search for a efficiet directio from the ode s source to the destiatio. Protocol geographical route scales much better i ad hoc etwork maily for two reasos. There is o eed of havig the latest routig table ad global etwork topologies view ad its chages. Protocol geographical route scales is useful for large multi-hop wireless etwork topologies because the physical iformatio of the odes is tracked via usig GPS or other types of positioig services. Sice the odes are movig radomly throughout the etwork ad the positio of the odes alters cotiuously (Shi et al., 20). Accordig to So et al. (2004), this protocol works by havig each ode forward a packet to the eighborig ode earest to the destiatio (So et al., 2004). This is kow as the greedy mechaism. The objective of this study is to ivestigate the effect of etwork size ad desity o the performace of MANET routig protocols usig simulatio. LITERATURE REVIEW Ad hoc O-demad Distace Vector (AODV) routig protocol: The AODV routig protocol is a reactive routig protocol. A reactive routig huts for routes whe data eeds to be set by a ode. Hece, routes are formed whe eeded. The AODV routig protocol cosists of four cotrol packets: hello messages, Route Replies (RREPs), Route Error Correspodig Author: Muthaa Najim Abdulleh, College of Iformatio Techology, Uiversiti Teaga Nasioal, Malaysia This work is licesed uder a Creative Commos Attributio 4.0 Iteratioal Licese (URL: 400

2 messages (RERRs) ad Route Requests (RREQs). These cotrol packets are used i two protocol mechaisms, route maiteace ad route discovery. All odes i the AODV protocol maitai a routig table to store iformatio regardig active routes from source to destiatio. The iformatio stored cosists of umber of hops, ext hop, destiatio sequece umber, active eighbours for a route ad the destiatio of a route table etry ad its time of expiry. Route etry timeouts are updated whe used. To prevet loopig i distace vector routig, a sequece umber is set with RREQs ad RREPs, both of which are kept i the routig table. Whe a ode receives multiple replies, the reply with the higher sequece umber is used. The AODV mechaism specifies that whe two routes possess similar sequece umber, the shorter route is used (Fehker et al., 202). Optimized Lik State Routig (OLSR): OLSR is a proactive routig protocol ad therefore keeps latest routes to other odes i the etwork. It is importat to sustai curret routig iformatio; proactive routig protocols eed to sed cotrol messages periodically which will geerate a large amout of routig overhead. However, OLSR is desiged to miimize this overhead (Sigla ad Paag, 203). Therefore, whe eeded, data ca be set without delay. OLSR routig protocol cosists of three geeral elemets: a mechaism for the effective floodig of cotrol traffic, oe for eighbor sesig ad a mechaism to determie how to choose ad publish adequate topological iformatio i the etwork i order to provide the best routes (Ahlgre et al., 202). I this protocol, the iformatio regardig etwork topology chages periodically through lik state messages. A hop-by-hop mechaism is utilized to forward packets (Saputro et al., 202). A Multipoit Relay (MPR) strategy is used to miimize the quatity of rebroadcastig odes ad the cotrol message size durig every route update (Ah ad Lee, 204). Nodes ca periodically exchage topological iformatio. MPR creates a uique route from the give source to the destiatio. These odes sese each other ad, i situatio ivolvig symmetrical liks, will cosider every ode a eighbor. Furthermore, lik sesig ad MPR selectio ca be carried out through hello messages. All iformatio related to the ode that seds the hello message ad its eighbourig odes is i the message. Each ode has the ability to obtai routig iformatio to reach two hops from a hello message. It ca also determie a subset of oe hop symmetric eighbour odes as its MPR set. This MPR set is ackowledged i its ext broadcastig of a hello message. I the first phase, eighbour odes are detected by use the hello messages. The exchage of the hello messages i OLSR permits the selectio of the MPR odes. The routig path to the kow destiatio of each ode is updated ad recalculated whe the Res. J. App. Sci. Eg. Techol., (4): , updated iformatio is received (Guo ad Wag, 204). The TC message broadcasts topological iformatio throughout the etwork, but these messages is oly forwarded through MPR odes. With MPR, odes are able to exchage topological iformatio i a periodical maer without havig to geerate a large amout of traffic. Greedy Perimeter Stateless Routig (GPSR): GPSR is a geographical routig protocol. Such protocols utilize positio-based routig, where a ode must kow where its immediate eighbour is located (Seok ad Saxea, 203). GPSR routig protocols use periodic beacoig to maitai updated geographical locatio iformatio of eighbourig odes withi their trasmissio rage (Jaiswal ad Khilar, 20). Greedy forwardig decisios are made by GPSR with the iformatio of the router s istat eighbors i the etwork topology. Whe a packet reaches where greedy forwardig is ot possible, the packet is forwarded aroud the perimeter of the regio, keepig status iformatio of local topology. GPSR scales best tha the shortest path ad ad hoc routig protocols as the umber of etwork destiatios odes grows (Alsaqour et al., 202). Related works: A large umber of research papers published i recet years have simulated ad equated to the MANET routig protocols performace. I this sectio, we discuss a series of past studies that compared various MANET routig protocols performaces. The authors i Issariyakul ad Hossai (20) have compared the performace of three routig protocols: Destiatio Sequeced Distaced Vector (DSDV) routig, AODV ad Dyamic Source Routig (DSR) (Vathaa ad Prakash, 204). The simulator tool NS-2 was used to lik the three routig protocols performaces. The parameters used by them were pause time, umber of coectios ad packet size, whereas the performace metrics used were packet loss, average ed-to-ed delay ad throughput. The simulatio results showed that AODV is the most suitable protocol for Trasmissio Cotrol Protocol (TCP) ad real-time traffic, sice AODV outperformed both the DSR ad DSDV protocols for all simulatio parameters. Aother study performed by Niraj ad Arora (202) compared AODV, DSR ad DSDV performaces usig NS-2. The performace parameters used for evaluatig the protocols were pause time, quatity of odes ad packet size ad the performace metrics used were throughput, ormalized routig load, packet delivery fractio ad average ed-to-ed delay. The results of the simulatio showed that DSR is superior to DSDV ad AODV based o the throughput ad packet delivery fractio. The regular ed-to-ed delay ad the ormalized routig load for AODV were foud to be better tha those for DSR ad DSDV for varyig umbers of odes. Furthermore, Niraj ad Arora (202)

3 reported that the routig overhead i AODV was better tha that of DSDV ad DSR with varyig values of pause time. Authors i Sigla ad Paag (203) have compared the performace of AODV, OLSR, DSR ad the Zoe Routig Protocol (ZRP) used the simulator tool OPNET. The parameter used i the simulatio was pause time ad the metrics used were throughput, retrasmissio attempts, etwork load ad media access delay. The results of the simulatio showed that ZRP outperformed the other routig protocols i term of throughput, retrasmissio attempts ad etwork load, whereas OLSR was superior to all other protocols i term of media access delay. METHODOLOGY Performace metrics: I MANET simulatios, the umber of performace metrics are commoly use to evaluate how the routig protocols perform (Beigh ad Peer, 202). I this study, we used the followig four performace metrics. Packet Delivery Fractio (PDF) is data packets fractio that effectively arrives at its destiatios odes. The packet delivery ratio idicates the efficiecy of a protocol i trasferrig packets from source to destiatio. A higher value meas that packet delivery is more successful (Aastasi et al., 2003): PDF = recvs 00 seds Throughput is the amout of data effectively set to the destiatio withi a specified time. It is ormally measured i bytes per secod. Throughput ca be affected by several factors, icludig badwidth, power, etwork topology ad reliability of commuicatio (Bai ad Helmy, 2004): Throughput recv = PkTduratio Average Ed-to-Ed delay (E2E) idicates packet trasmissio s iterruptio i from the mai ode to the destiatio. The total iterruptio is a accumulatio of several small delays i the etwork. This comprises possible delays due to a buffer i route discovery latecy, delays i liig up at the iterface, Media Access Cotrol (MAC) retrasmissio delays ad trasfer time ad propagatio delays. The average E2E delay of a received packet ca be calculated Res. J. App. Sci. Eg. Techol., (4): , by obtaiig the time variace betwee the trasmissio ad respose of the packet at a Costat Bit Rate (CBR) ad dividig the time differece by the total umber of CBR trasmissios. Lower ed-ed delays idicate better performace (Siha ad Se, 202): 2 = Normalized Routig Load (NRL) is the total of routig packets passedper data packet delivered to the destiatio ode. This metric is used to measure the overhead geerated by the routig protocol durig its routig operatio A low-value overhead meas a lower umber of cotrol packets is created by the protocol, which leaves additioal etwork resources available to trasmit real data packets (Yussof et al., 2009): NRL RPge = recvs SIMULATION RESULTS We simulated the AODV, GPSR ad OLSR routig protocols usig Network Simulator NS2 (versio 2.33). The simulatio was executed for two scearios ad the differece betwee them was i terms of the simulatio parameter evaluated. I the first sceario, the simulatio parameter that is varied was the umber of odes while the other parameters were kept costat. Icreasig the quatity of odes escalates the desity of the etwork. The details of the simulatio parameters for sceario are listed i Table. The simulatio parameter varied i the secod sceario was etwork size. By icreasig etwork size, the area i which odes ca travel becomes larger. The details of the simulatio parameters for sceario 2 are listed i Table 2. For each sceario, we measured the performace of AODV, GPSR ad OLSR usig four performace metrics: average ed-toed delay, packet delivery fractio, throughput ad ormalized routig load. Sceario : The impact of umber of odes: This sceario was simulated multiple times, where a differet umber of odes were used i each simulatio. The umber of odes used for each simulatio is listed i Table. The simulatio was ru 0 times, each with a differet seed, for every variatio i the umber of odes. The results preseted here were obtaied by calculatig the average of the simulatio results.

4 Res. J. App. Sci. Eg. Techol., (4): , 205 Table : Simulatio parameters for sceario Number of odes 30, 50, 70, 90, 0, 30, 50 odes Simulatio time 900 sec Map size m Max speed 20 m/sec Mobility model Radom waypoit Traffic type Costat Bit Rate (CBR) Packet size 52 bytes Coectio rate (omial radio rage) 4 pkts/sec Pause time 20 sec Number of coectios 5 Badwidth of liks 2 Mbit MAC layer type IEEE 805 Seed 5, 20, 44, 50, 64, 7, 80, 89, 9, 0 Table 2: Simulatio parameters for sceario 2 Number of odes 50 odes Simulatio time 900 sec Map size , , , , , Max speed 20 m/sec Mobility model Radom waypoit Traffic type Costat Bit Rate (CBR) Packet size 52 bytes Coectio rate (omial radio rage) 4 pkts/sec Pause time 20 sec Number of coectios 5 Badwidth of liks 2 Mbit MAC layer type IEEE 802. Seed 5, 20, 44, 50, 64, 7, 80, 89, 9, 0 Fig. 2: Packet delivery fractios for varyig umbers of odes for the three protocols Fig. 3: The average ed-to-ed delay for varyig umbers of odes for the three protocols Fig. : Throughput for varyig umbers of odes for the three protocols We simulated this etwork for each routig protocol ad the results are show i Fig. to 4. Figure shows the impact of the umber of odes o the throughput for each routig protocol. I geeral, the thought for all three protocols remai relatively similar regardless of the umber of odes i the etwork. However, GPSR has much higher throughput tha that of AODV ad OLSR. This is maily due to the behavior of the GPSR protocol where packets are simply set to the eighbor that is the earest to the fial ode. No routig packets to search for a path eed to be geerated. This low overhead causes more badwidth to 403 Fig. 4: Normalized routig load for varyig umbers of odes for the three protocols be available for data trasfer ad this cotributes to the higher throughput. The throughput for the AODV routig protocol was higher tha that for OLSR because AODV has a lower routig overhead tha OLSR sice it searches for paths o-demad ad does ot eed to sustai the latest routig table. The lower overhead allows more badwidth to be used for the data packets. OLSR recorded the worst throughput because it cosumes a sigificat amout of etwork badwidth because of the frequet eed to sed update messages. Figure 2 shows the effect of the umber of odes o the packet delivery fractio for each protocol route. As the quatity of odes grows, the packet delivery fractio

5 Res. J. App. Sci. Eg. Techol., (4): , 205 Fig. 5: Throughput results for etworks of differet sizes for the three protocols iformatio to forward the packet to aother ode that is closer to the destiatio. Doig this requires very short amout of time. Figure 4 shows the effect of the quatity of odes o the ormalized load route. As the quatity of odes grows, we see a slight growth i AODV ormalized routig load. This is due to the low demad for badwidth eeded to maitai the route betwee the source ode ad the destiatio ode. The ormalized routig load icreases for GPSR with icreasig umber of odes because there are more beacos that eed to be processed i order to update the iformatio regardig the geographic locatios of the eighbourig odes. With larger umbers of odes, the ormalized routig load for OLSR escalates tremedously because the odes eed to process more update messages. Fig. 6: Packet delivery fractio results for differet etwork sizes for the three protocols teds to slightly icrease. The packet delivery fractio for the GPSR routig protocol was higher tha that of the AODV ad OLSR protocols. As the etwork becomes deser (i.e., the etwork cotais more odes), GPSR attais a packet delivery fractio higher tha that of traditioal protocols, such as AODV ad OLSR. The performaces of AODV are just a little but lower tha that of GPSR. The OLSR routig protocol has the lowest value for packet delivery fractio. This is because i OLSR, odes eed sed frequet updates, which ca reduce the amout of etwork resources available to sed data. This may cause some data packets to be dropped, thus lowerig the packet delivery fractio. Figure 3 shows how the umber of odes o average affects ed-to-ed delay as the umber of odes grows. The average E2E delay for the OLSR routig protocol grows with the escalatig umber of odes. Icreasig the quatity of odes causes a chage i the etwork topology, which i tur causes more update messages to be set. These update messages ca cogest the etwork, causig a high delay for data packets. AODV performs relatively well with respect to E2E delay, where the performace is oly slightly less tha that of GPSR. The GPSR routig protocol delivered the best performace i terms of average E2E delay because the odes oly eed to use locatio 404 Sceario 2: Impact of etwork size: This sceario is simulated multiple times ad a differet etwork size is used each time. The etwork size used for each simulatio is listed i Table 2. For each etwork size, the simulatio was ru 0 times, each with a differet seed. The results preseted here were obtaied by calculatig the average of the simulatio results. The results of the simulatio i terms of throughput for AODV, GPSR ad OLSR i sceario 2 are show i Fig. 5. It shows that throughput results for all three routig protocols decrease as etwork size icreases. This is because whe etwork size icreases, odes have greater freedom to move, which leads to chages i etwork topology. This makes it more difficult to fid a routig track to the edpoit, regardless of the protocol. Some destiatio odes may ot eve be reachable. The reaso why GPSR performs better tha AODV ad OLSR is the same as the oe described i sceario above. Figure 6 shows GPSR, AODV ad OLSR protocol route packet delivery fractio decrease as the etwork topology icreases. This is because as the etwork gets larger, the odes are capable of movig further from each other. As a result, liks betwee odes may break more easily as the odes are mobile. Some odes could become iaccessible, which reduces packet delivery fractio. Of the three routig protocols, GPSR provided the highest packet delivery fractio ad OLSR provided the lowest. Figure 7 shows the result of the average E2E delay for AODV, GPSR ad OLSR. Obviously, the E2E delay gets higher as the etwork gets larger, especially startig from etwork size m. However, the GPSR protocol is the least affected by the etwork size, providig a much lower E2E delay compared to AODV ad GPSR, especially at the largest etwork size of m. This could be attributed to the low overhead of GPSR, which causes the etwork to be less cogested as compared to AODV ad OLSR.

6 Res. J. App. Sci. Eg. Techol., (4): , 205 Fig. 7: Results for average ed-to-ed delay for etworks of differet sizes for the three protocols route protocols perform with regards to the etwork size ad desity. The etwork performace was measured based o the throughput, average Ed-to-Ed (E2E) delay, Packet Delivery Fractio (PDF) ad Normalized Routig Load (NRL). The results of the simulatios show that GPSR is superior to OLSR ad AODV i most cases. This is maily attributed to GPSR s routig mechaism where iformatio used by greedy decisios are forwarded usig the router s earest eighbors i the etwork topology. This mechaism has low overhead ad this cotributes to its good performace. The simulatio results also shows that the rise i the odes umber affects the ormalized routig load, while the icrease i etwork size has a large effect o throughput, ed-to-ed delay ad packet delivery fractio. Fig. 8: Results for ormalized routig load for varyig etwork sizes for the three protocols Figure 8 shows the results for the ormalized routig load for the GPSR, AODV ad OLSR protocol routes are based o the size of the etwork. The ormalized routig load for OLSR icreases as the etwork topology becomes bigger. This is as the etwork topology gets larger, the odes ted to move more ad therefore more update messages eed to be geerated by OLSR to preserve up-to-date routig iformatio. For AODV, the ormalized load route slightly grows as the etwork size itesifies because more RREP ad RREQ messages eed to be geerated to search for odes that have moved further away from the source. GPSR o the other had is ot very much affected by the etwork size. I fact, the ormalized routig load ted to slightly decrease as the etwork size gets larger. This is because i GPSR o routig messages eed to be set to far away odes. The routig mechaism relies o the locatio of eighborig odes ad this mechaism works pretty much the same way regardless of the etwork size or the odes locatio. CONCLUSION I this study, performaces of OLSR, AODV routig ad GPSR protocolsis each compared to proactive, reactive ad geographical routig protocol, respectively. We used simulatios to assess the how the 405 REFERENCES Ahlgre, B., C. Daewitz, C. Imbreda, D. Kutscher ad B. Ohlma, 202. A survey of iformatiocetric etworkig. IEEE Commu. Mag., 50(7): Ah, J.H. ad T.J. Lee, 204. Multipoit relay selectio for robust broadcast i ad hoc etworks. Ad Hoc Netw., 7: Alsaqour, R.A., M.S. Abdelhaq ad O.A. Alsukour, 202. Effect of etwork parameters o eighbor wireless lik breaks i GPSR protocol ad ehacemet usig mobility predictio model. EURASIP J. Wirel. Comm., 7(): -5. Aastasi, G., E. Borgia, M. Coti ad E. Gregori, IEEE 802. ad hoc etworks: Performace measuremets. Proceedig of the 23rd Iteratioal Coferece o Distributed Computig Systems Workshops, pp: Bai, F. ad A. Helmy, A Survey of Mobility Models. Wireless Adhoc Networks. Uiversity of Souther Califoria, USA, pp: 206. Beigh, B.M. ad M. Peer, 202. Performace evaluatio of geographical routig protocols: A empirical study. Proceedig of the Iteratioal Coferece o Computer Commuicatio ad Iformatics (ICCCI, 202). Coimbatore, Idia, pp: -6. Coti, M. ad S. Giordao, 204. Mobile ad hoc etworkig: milestoes, challeges ad ew research directios. IEEE Commu. Mag., 52(): Elgohary, A., T.S. Sobh, S.A. Nouh ad M. Zaki, 204. A efficiet ad depedable protocol for critical MANETs. J. High Speed Netw., 20(3): Fehker, A., R. Va Glabbeek, P. Höfer, A. McIver, M. Portma ad W.L. Ta, 202. Automated aalysis of AODV usig UPPAAL. I: Flaaga, C. ad B. Köig (Eds.), TACAS, 202. LNCS 724, Spriger-Verlag, Berli, Heidelberg, pp:

7 Res. J. App. Sci. Eg. Techol., (4): , 205 Guo, J. ad A. Wag, 204. Study o itegratio OLSR protocol i mobile ad hoc etwork. Proceedig of the 9th Iteratioal Symposium o Liear Drives for Idustry Applicatios. Spriger, Berli, Heidelberg, 4, Issariyakul, T. ad E. Hossai, 20. Itroductio to Network Simulator NS2. Spriger Sciece+Busiess Media, New York. Jaiswal, J. ad P.M. Khilar, 20. Fault tolerat greedy perimeter stateless routig i wireless etwork. Proceedig of the Iteratioal Coferece o Commuicatio, Computig ad Security, pp: Komai, Y., Y. Sasaki, T. Hara ad S. Nishio, 204. K NN query processig methods i mobile ad hoc etworks. IEEE T. Mobile Comput., 3(5): Niraj, M. ad M. Arora, 202. Performace evaluatio of routig protocols for qos measures i MANETS. It. J. Maag. IT Eg., 2(4): Saputro, N., K. Akkaya ad S. Uludag, 202. A survey of routig protocols for smart grid commuicatios. Comput. Netw., 56(): Seok, K.K. ad N. Saxea, 203. Aalysis of a ovel advaced greedy perimeter stateless routig algorithm. Proceedig of the Iteratioal Coferece o ICT Covergece (ICTC, 203), pp: Shi, Z., C. Beard ad K. Mitchell, 20. Competitio, cooperatio, ad optimizatio i multi-hop csma etworks. Proceedig of the 8th ACM Symposium o Performace Evaluatio of Wireless Ad Hoc, Sesor, ad Ubiquitous Networks. ACM, pp: Sigla, S. ad T.S. Paag, 203. Evaluatig the performace of maet routig protocols. It. J. Electro. Commu. Eg. Techol., 4(): Siha, S. ad B. Se, 202. Effect of varyig ode desity ad routig zoe radius i ZRP: A simulatio based approach. It. J. Comput. Sci. Eg., 4: So, D., A. Helmy ad B. Krishamachari, The effect of mobility-iduced locatio errors o geographic routig i mobile ad hoc sesor etworks: Aalysis ad improvemet usig mobility predictio. IEEE T. Mobile Comput., 3(3): Vathaa, S. ad V. Prakash, 204. Comparative study of proactive ad reactive adhoc routig protocols usig Ns2. Proceedig of the World Cogress o Computig ad Commuicatio Techologies (WCCCT, 204), pp: Yussof, S., H.S. Jassim, T.S. Kiog, S.P. Koh ad R. Ismail, A routig protocol based o trusted ad shortest path selectio for mobile ad hoc etwork. Proceedig of the IEEE 9th Malaysia Iteratioal Coferece o Commuicatios (MICC, 2009), pp: