Internatonal Journal of Computer Networks & Communcatons (IJCNC), Vol., No.4, July 1 COMARISON OF DIFFERENT BROADCAST SCHEMES FOR MULTI-HO WIRELESS SENSOR NETWORKS 1 S. Mehta and K.S. Kwak UWB Wreless Communcatons Research Center, Inha Unversty Incheon, 4-751, Korea suryanand.m@gmal.com ABSTRACT In ths paper, we present the performance of dfferent broadcast schemes for multhop sensor networks based on mathematcal modelng. In near future many applcatons wll demand multcast (Broadcast) communcaton feature from the sensor networks. Ths broadcast feature does not use vrtual carrer sensng but reles on yscal carrer sensng to reduce collson. For ths paper, we analyze the dfferent broadcast schemes for multhop wreless sensor networks and also calculated the achevable throughput. KEYWORDS Wreless Sensor networks, Broadcast, MAC protocol, Mult-hop communcaton. 1. INTRODUCTION The IEEE 8.11 standard [1] s wdely used and deployed n wreless systems. IEEE 8.11 MAC protocol allows multple nodes to share the wreless medum wthout any central coordnator. If two nodes that are near by each other transmt frames at the same tme, the frames collde and the channel bandwdth wll not utlzed. The MAC protocol tres to avod ths stuaton usng a mechansm called Multple Sensng Access wth Collson Avodance (CSMA/CA). CSMA/CA mechansm frst lsten the channel for a partcular duraton (a slot tme), whenever a node wants to transmt frame. If the channel s deal for a partcular duraton, the node transmts the frame. The node dffer ts transmsson and wats for a random delay tme (back-off nterval) before retryng f the channel s busy. Ths channel sensng mechansm s well-known as yscal carrer sensng. hyscal carrer sensng does not avod the collson from the hdden termnal problem f we assume the carrer sensng range s equal to the recevng range. To avod the collson from hdden termnal, a hdden termnal should not transmt a frame for partcular tme perod, ths tme perod s known as a vulnerable perod. So for broadcastng, only yscal carrer sensng s used. Vrtual sensng s not drectly applcable for broadcast transmsson because CTS messages sent by multple recevers wll result n a collson. All prevous studes are for uncast communcaton; they do not consder broadcast communcaton. In [], authors presented mathematcal model and a defnton for broadcastng scheme, and also the numercal results for IEEE 8.11 DCF. In ths paper, we use the same mathematcal model and defnton of broadcastng as n [] [5], and also extend the numercal results for slotted aloha and threshold condtons based MAC protocols [ex. IS-MAC] for sensor networks [3][4]. In [4], authors used threshold condtons for transmttng, and these threshold 1 A part of ths paper was publshed n NEXT 7 Conference, Seoul,Korea [7]. DOI : 1.511/jcnc.1.41 1
Internatonal Journal of Computer Networks & Communcatons (IJCNC), Vol., No.4, July 1 condtons are also useful for broadcast transmsson. Here, broadcast transmsson refers successful only when all of the sender s neghbors receve the broadcast message correctly. Relable broadcast can be used for number of applcatons, such as base applcaton, nformaton dstrbuton, and a bass for supportng dstrbuted protocols. The man contrbutons of ths paper are as follows We present the performance of dfferent broadcastng schemes based on mathematcal models as n [][3]. Our comparson of dfferent broadcastng schemes s very useful for a sensor network desgner to set tradeoff between spatal reuse of channel and hdden node area. The rest of the paper s organzed as follows: In secton, we present the numercal analyss of broadcastng scheme. In secton 3, we present numercal results from our analyss. Fnally, we conclude n secton 4.. NUMERICAL ANALYSIS Frst of all we analyze the performance of IEEE 8.11 broadcast scheme, the hdden node problem n a broadcast scenaro, and then we extend our analyss for slotted aloha and threshold condtons based MAC protocols. As shown n fgure 1(a), node A s n recevng range of node B but not n the recevng regon of node C, may cause hdden termnal problem. Ths area defned as potental hdden node area. For uncast communcatons, the sze of the potental hdden node area calculated usng the dstance between sender and recever. However, same calculaton s not appled for broadcast communcaton. The potental hdden node area for broadcast communcaton depends on recevng range of all the neghbourng nodes as shown n fgure 1(b). So t s dffcult to exactly compute the sze of ths area. Moreover, as explaned earler, varyng the carrer sensng area also change the form of ths area. When there are nfnte numbers of node at the edge of the sender s transmsson range, the potental hdden node area s maxmzed for the worst case. Let R denotes the transmsson range of a node. As shown n fgure 1(b) maxmum sze of potental hdden node area can be ( ) π R π R = 3π R. Thus, n case of broadcast, the potental hdden termnal area can be dramatcally larger than that of uncast. (a) Uncast (b) Broadcast Fgure 1 otental hdden nodes area We use the same mathematcal models as derved n [][5] to acheve the average throughput for multhop sensor networks. To make mathematcal model tractable, we assume followngs for the mult-hop wreless network model.
Internatonal Journal of Computer Networks & Communcatons (IJCNC), Vol., No.4, July 1 1. All nodes n the networks are two-dmensonally osson dstrbuted wth densty λ,.e., the probablty p(, A) of fndng nodes n an area of sze A s gven by ( λ ) λ A p(, A) = A e!. All nodes have the same unform transmsson and recevng range of radus R. N s the average number of neghbor nodes wthn a crcular regon ofπ R. Therefore, we have N = λπ R. 3. A node transmts a frame only at the begnnng of each slot tme; however, IS-MAC protocol (Threshold condtons based MAC) based node transmts only f mnmum threshold condton gets satsfed [4]. The sze of a slot tme,τ, s the duraton ncludng transmt-to-receve turn-around tme, carrer sensng delay and processng tme. 4. The transmsson tme or the frame length s the same for all nodes,.e., the same packet length. 5. When a node s transmttng, t cannot receve at the same tme. 6. A node s ready to transmt wth probablty p ; however, for IS-MAC protocol transmttng probablty p also depends on the node s buffered sze [4]. Let p denote probablty that a node transmts n a tme slot. If p s ndependent at any tme slot, t can be obtaned by p = p.rob{channel s sensed dle n a slot} p. I. Where I s the lmtng probablty that the channel s sensed to be dle. 7. The carrer sensng range s assumed to vary between the range R~ R. Fgure Markov chan model for the channel From the above mentoned assumptons, the channel process modeled can be represented as a two-state Markov chan shown n fgure. As shown n the fgure ths model has states and descrbed as follows Idle: Ths s the state when the channel around node x s sensed dle, and ts duraton T dle sτ. Busy: Ths s the state when a successful DATA transfer s done. The channel s n busy state for the duraton of DATA transfer, thus the busy tme, T busy, s equal to the transmsson δ. ( T busy = δ ). In MAC scheme, all nodes should stay at least deal for one slot tme, tme after the channel becomes dle. Thus, the transton probablty probablty of the neghbour nodes transmsson, and s gven by b s 1. s the transton 3
Internatonal Journal of Computer Networks & Communcatons (IJCNC), Vol., No.4, July 1 λπ R R p N ( ) ( ) λπ 1 = p e = e! = Let, Φ and Φ b denote the steady-state probabltes of dle and busy states respectvely. From fgure, we have Snce Φ = 1 Φ, we have b Φ = Φ + Φ bb = Φ + Φ b 1 1 Φ = = e Now the lmtng probablty I can be obtaned by p N. I τ = p N ( δ )(1 e ) + τ Accordng to the relatonshp between and, s gven by τ p = p N ( δ )(1 e ) + τ Fgure 3 Markov chan model for the node For the throughput calculaton we need to calculate the probablty of a successful transmsson. As shown n fgure 3 the transmsson state of node x can be modelled by three states Markov chan model. Wat, succeed, and collson states represents the node s transmsson dffer, successful DATA transmsson, and collson state condtons, respectvely. At the begnnng of each tme slot, node x leaves the wat sate wth probablty p. Thus, the translaton probablty s gven by ww = 1 p. ww And, the duraton of a node n wat sate T wat s τ (Ths watng tme s only after node satsfy the mnmum threshold condton [4]). The duraton of success and collson sates are equal to the frame transmsson duraton tme, hence, T succ and T coll are δ + τ. After executng the desred acton n succeed and collson state, node x always returns to the wat sate. Therefore, sw and cw equals to 1. 4
Internatonal Journal of Computer Networks & Communcatons (IJCNC), Vol., No.4, July 1 Let Φw, Φs, and Φ c represents the steady-state probabltes of wat, success, and collson states respectvely. From the fgure 4 we have Hence, we have Φ = Φ + Φ + Φ = Φ + 1 Φ. (1) w w ww s sw c cw w ww w 1 Φ w = ww Based on the above condton, transton probablty ws cab be ws = () 1 3 Where, 1 = rob{node x transmts n a slot} = rob{all of node x's neghbor nodes do not transmt n thesameslot} = rob{nodes n potental hdden area do not transmt for δ +τ} 3 Last term represents the vulnerable perod that s equal to δ aloha based MAC protocols [6][3].Obvously, 1 = p and + τ and δ s be gven by + τ n case of slotted λπ R R p R p N ( ) ( ) λπ 1 λπ = p e = e = e! = Fgure 4 Transmsson area, addtonal carrer sensng area, and potental hdden nodes area To calculate, we frst approxmate the number of node n the potental hdden node area. Let A tx, A cs, and A represent the transmsson area, addtonal carrer sensng area, and potental hdden node area, respectvely. As shown n fgure 4 addtonal carrers sensng area s the yscal carrer sensng area s larger than transmsson range and smaller than potental hdden node area. Hence, we have π Acs 3 R. And, the potental hdden node area s gven by 5
Internatonal Journal of Computer Networks & Communcatons (IJCNC), Vol., No.4, July 1 Hence, ( ) A = π R A A = π ( R) π R A = 3 π R A. tx cs cs cs π A 3 R. Let N represent the number of node n potental node area. As we assumed, nodes are unformly dstrbuted, hence, N s gven by N = λ A N λ3π R N 3N (3) Wth eq.3 3 s gven by ( N ) 3 = ( 1 p) e 1 = e =! N ( δ ) p N ( δ τ ) + τ + Therefore, eq. s gven by = p e e ws p N p 3 N ( ) δ + τ. From the fgure 4, we have ws = 1 ww wc and cw = sw = 1. Hence, the steady state probablty of succeed state, Φ, s gven by s ws Φ s = Φ wws = 1+ p Accordng to the defnton of throughput from [6], the throughput equals the fracton of tme n whch the channel s engaged n successful transmsson of user. Therefore, the throughput Th s equal to the lmtng probablty that the channel s n success state. Th Φ δ s = (4) Φ + Φ + Φ T T T s succ c coll w wat p ( N + 3 N ( )) ( δ + τ wsδ p e ) δ Th = = p T + T τ + p ( δ + τ ) coll wat 3. NUMERICAL RESULTS In ths secton, we present numercal results based on the model presented n the prevous secton. We frst study the performance of IEEE 8.11 broadcast scheme by varyng the average number of neghbourng node (N) and transmsson attempt probablty (p ). In ths analyss, we fx the frame as 1τ. Fgure 5 shows the throughput results of the IEEE 8.11 broadcast scheme wth dfferent potental hdden node area ( R ). From the fgure 6 t s 6
Internatonal Journal of Computer Networks & Communcatons (IJCNC), Vol., No.4, July 1 cleared that, as the percentage of A reduces to, throughput performance ncreases to maxmum value. Ths means that, by achevng maxmum value of A cs, the IEEE 8.11 broadcast scheme mnmzes the probablty of hdden node problem. Fgure 6 shows the throughput results of the threshold condtons based broadcast scheme wth dfferent potental hdden node area. These threshold condtons help to mprove the throughput by not lettng all the nodes to transmt n the same slot at the same tme. Wth these threshold condtons, we acheve nearly twce of the throughput compared to IEEE 8.11 broadcast scheme. Fgure 7 shows the throughput results of the slotted aloha based broadcast scheme. Slotted aloha based broadcast scheme wth threshold condtons gve qute good throughput. Fgure 8 shows the combned results of all the schemes wth all the varatons n hdden node area. From fgure 9 t s clear that, the threshold condtons based broadcast scheme wth % hdden area gves the hghest throughput. So t s benefcal to set the large carrer sensng range for broadcast communcaton. However, for uncast communcaton, a large carrer sensng range leads to reduce spatal reuse, so mnmzng hdden node effect and ncreasng spatal reuse becomes a tradeoff whch must be studed further. Slotted aloha and threshold condtons based broadcast schemes help us to acheve a good tradeoff between spatal reuse and hdden node effect. Our results reveal the performance of broadcastng communcatons under the mpact of hdden termnals and open some new drectons for further research...18.16.14.1.1.8.6.4..4 1 3 4 5 6 p': robablty of transmt n a tme slot x 1-3 (a) 1% Hdden node area.35.3.5..15.1.5 1 3 4 5 6 7 8 9 p'=robablty of transmt n a tme slot x 1-3 (b) 5% Hdden node area 7
Internatonal Journal of Computer Networks & Communcatons (IJCNC), Vol., No.4, July 1.1.9.8.7.6.5.4.3..1.5.1.15..5.3 p':robabltyof transmt n a slot (c) % Hdden node area Fgure 5 IEEE 8.11 based broadcast scheme.4.35.3.5..15.1.5.7.1..3.4.5.6.7.8.9.1 p : robablty of transmt n a tme slot (a) 1% Hdden node area.6.5.4.3..1..4.6.8.1.1.14.16.18. p : robablty of transmt n a tme slot (b) 5% Hdden node area 8
Internatonal Journal of Computer Networks & Communcatons (IJCNC), Vol., No.4, July 1.16.14.1.1.8.6.4..5.1.15..5.3.35.4.45.5 p : robablty of transmt n a tme slot (c) % Hdden node area Fgure 6 Threshold condtons based broadcast scheme.4.35.3.5..15.1.5.1..3.4.5.6.7.8.9.1 p : robablty of transmt n a tme slot (a) Slotted aloha type scheme wthout threshold condtons.1.1.8.6.4..5.1.15..5.3 p : robablty of transmt n a tme slot (b) Slotted aloha type scheme wth threshold condtons Fgure 7 Slotted aloha type broadcastng scheme 9
Internatonal Journal of Computer Networks & Communcatons (IJCNC), Vol., No.4, July 1.16 1% hdden area wthout prorty condtons 5% hdden arae wthout prorty condtons.14.1 % hdden area wthout prortycondtons 1% hdden area wth prortycondtons 5% hdden area wth prorty condtons % hdden area wth prorty condtons Slotted aloha type wthout prorty condtons Slotted aloha type wth prorty condtons.1.8.6.4. 3 4 5 6 7 8 9 Average number of neghbor nodes (N) Fgure 8 Combned results 4. CONCLUSIONS In ths paper, we present the performance of dfferent broadcastng schemes based on a smple markov chan model. The results show that overall performance of dfferent broadcastng schemes degrades rapdly when the number of competng nodes allowed wthn a regon ncreases. Our comparson of dfferent broadcastng schemes s very useful for sensor network desgner, and also helps us to set tradeoff between spatal reuse of channel and hdden node area. In future, we want to extend our study for mult-channel hdden termnals and non-unform recevng/transmttng range problems of a broadcastng node. ACKNOWLEDGEMENT Ths research was supported by the MKE(The Mnstry of Knowledge Economy), Korea, under the ITRC(Informaton Technology Research Center) support program supervsed by the NIA(Natonal IT Industry romoton Agency) (NIA-1-C19-111-7). REFERENCES [1] IEEE 8.11 Workng Group, (1999),Wreless LAN Medum Access Control (MAC) and hyscal Layer (HY) specfcatons. [] J.M. Cho, J. So, and Y. B. Ko, (5) Numercal Analyss of IEEE 8.11 Broadcast Scheme n Multhop Wreless Ad-hoc Networks, ICOIN, Januaray, pp:1-1. [3] L. G. Roberts, (197) Aloha acket Systems wth and wthout Slots and capture, ASS Note 8, Stanford Research Insttute, Advance Research rojects Agency, Network Informaton Center. [4] S.Mehta and J. Km, (6) Improved Sensor MAC rotocol for Sensor Networks, Internatonal Journal of lateral Computng (IJLC), publshed by MacMllan Inda ltd. ISBN-1:143-931-1. [5] L. Wu and. K. Varshney, (1999) erformance analyss of CSMA and BTMA protocols n,multhop networks(1). Sngle channel case, Elsever Informaton Scence, Vol.1,pp-159-177. [6] R. Rom and M. Sd, (1989) Multple Access rotocols: erformance and Analyss, Sprongerverlag, 1989. [7] S.Mehta and K.S. Kwak, (7) Numercal Analyss of Dfferent Broadcast Schemes for Mult-hop Sensor Networks, n proceedng of NEXT 7 Conference, September, pp.34-34. 1
Internatonal Journal of Computer Networks & Communcatons (IJCNC), Vol., No.4, July 1 Authors S.Mehta receved the B.E. and M.S degrees both n Electroncs Engneerng from Mumba Unversty, Mumba, Inda, and Ajou Unversty, Korea n and 5, respectvely. He s currently pursung the h.d. degree n Telecommuncaton engneerng from the Inha Unversty, Korea. Hs research nterests are n the area of performance analyss of wreless networks and RFID systems. K. S. Kwak receved the B.S. degree from Inha Unversty, Korea n 1977, and the M.S. degree from the Unversty of Southern Calforna n 1981 and the h.d. degree from the Unversty of Calforna at San Dego n 1988, respectvely. From 1988 to 1989 he was a Member of Techncal Staff at Hughes Network Systems, San Dego, Calforna. From 1989 to 199 he was wth the IBM Network Analyss Center at Research Trangle ark, North Carolna. Snce then he has been wth the School of Informaton and Communcaton, Inha Unversty, Korea as a professor. He had been the charman of the School of Electrcal and Computer Engneerng from 1999 to and the dean of the Graduate School of Informaton Technology and Telecommuncatons from 1 to at Inha Unversty, Inchon, Korea. He s the current drector of Advanced IT Research Center of Inha Unversty, and UWB Wreless Communcatons Research Center, a key government IT research center, Korea. He has been the Korean Insttute of Communcaton Scences (KICS) s presdent of 6 year term. In 1993, he receved Engneerng College Young Investgator Achevement Award from Inha Unversty, and a dstngushed servce medal from the Insttute of Electroncs Engneers of Korea (IEEK). In 1996 and 1999, he receved dstngushed servce medals from the KICS. He receved the Inha Unversty Engneerng aper Award and the LG aper Award n 1998, and Motorola aper Award n. Hs research nterests nclude multple access communcaton systems, moble communcaton systems, UWB rado systems and ad-hoc networks, hgh-performance wreless Internet. 11