Interference-Aware IEEE WiMax Mesh Networks

Transcription

1 in Proceeing of 61 t IEEE Vehicular Technology Conference (VTC 2005 pring), tockholm, ween, May 29-June 1, 2005 Interference-Aware IEEE WiMax Meh Network Hung-Yu Wei, amrat Ganguly, Rauf Izmailov Broaban & Mobile Networking Department NEC Laboratorie America Princeton, New Jerey, UA {hungyu, amrat, rauf}@nec-lab.com Zygmunt J. Haa chool of Electrical an Computer Engineering Cornell Univerity Ithaca, New York, UA haa@ece.cornell.eu Abtract The IEEE WiMax tanar provie a mechanim for creating multi-hop meh, which can be eploye a a high pee wie-area wirele network. To realize the full potential of uch high-pee IEEE meh network, two efficient wirele raio reource allocation extenion were evelope. The objective of thi paper i to propoe an efficient approach for increaing the utilization of WiMax meh through appropriate eign of multi-hop routing an cheuling. A multiple-acce interference i a major limiting factor for wirele communication ytem, we aopt here an interference-aware cro-layer eign to increae the throughput of the wirele meh network. In particular, our cheme create a tree-bae routing framework, which along with cheuling i interference aware an reult in a much higher pectral efficiency. Performance evaluation reult how that the propoe interference-aware cheme achieve ignificant throughput enhancement over the baic IEEE meh network. Keywor-IEEE , Meh Network, WiMax, Cro-layer Deign an Optimization, Multi-hop Routing, Meh Moe I. INTRODUCTION The IEEE Working Group create a new tanar, commonly known a WiMax, for broaban wirele acce at high pee, at low cot, which i eay to eploy, an which provie a calable olution for extenion of a fiber-optic backbone. WiMax bae tation can offer greater wirele coverage of about 5 mile, with LO (line of ight) tranmiion within banwith of up to 70 Mbp. Beyon jut proviing a ingle lat hop acce to a broaban IP, WiMax technology can be ue for creating wie-area wirele backhaul network. When a backhaul-bae WiMax i eploye in Meh moe, it oe not only increae the wirele coverage, but it alo provie feature uch a lower backhaul eployment cot, rapi eployment, an reconfigurability. Variou eployment cenario inclue citywie wirele coverage, backhaul for connecting 3G RNC with bae tation, an other. In aition to the ingle-hop IEEE PMP (point-tomultipoint) operation, IEEE a tanar [1] efine the baic ignaling flow an meage format to etablih a meh network connection. ubequently, the Meh moe pecification were integrate into the IEEE reviion [2]. Although ingle hop WiMax provie high flexibility to attain Quality of ervice in term of ata throughput, achieving the ame in multi-hop WiMax meh i challenging. One of the major problem i ealing with the interference from tranmiion of the neighboring WiMax noe. Cro-layer eign an optimization i known to improve the performance of wirele communication an mobile network [3,4]. In orer to eign a pectrally efficient IEEE meh network, we opt for joint eign an optimization that relie on Application Layer loa eman information, Phyical Layer interference information, a well a cheuling an route election mechanim in Data Link Layer. We believe that interference in wirele ytem i one of the mot ignificant factor that limit the network capacity an calability of wirele meh network. Conieration of interference conition uring raio reource allocation an route formation procee impact the eign of concurrent tranmiion cheme with better pectral utilization while limiting the mutual interference. We evelope an interference-aware IEEE framework with a eign goal of achieving overall high utilization of the WiMax Meh network. Our propoe cheme inclue a novel interference-aware route contruction algorithm an an enhance centralize meh cheuling cheme, which conier both traffic loa eman an interference conition. Thi interference-aware eign approach will lea to better patial reue an thu higher pectral efficiency. The ret of the paper i organize a follow: in ection II, we briefly overview the IEEE Meh moe operation. In ection III, we preent the interference-aware framework that inclue route contruction an cheuling algorithm. In ection IV, the IEEE meh network performance i evaluate via imulation an compare to theoretical upper boun, which we foun uing linear programming. Finally, we ummarize our reult an provie ome concluing remark in ection V.

2 II. OVERVIEW OF WIMAX MEH A. Motivation an Problem Overview In comparion to IEEE a/b/g bae meh network, the bae WiMax meh provie variou avantage apart from increae range an higher banwith. The TDMA bae cheuling of channel acce in WiMax-bae multi-hop relay ytem provie fine granularity raio reource control, a compare to RT/CT-bae a/b/g ytem. Thi TDMA bae cheuling mechanim allow centralize lot allocation, which provie overall efficient reource utilization uitable for fixe wirele backhaul network. (The WiMaxbae meh backhaul application iffer from the a/b/gbae meh, which target mobile a hoc network.) However, the interference remain a major iue in multi-hop WiMax meh network. To provie high pectral uage, an efficient algorithm for lot allocation i neee, o a to maximize the concurrent tranmiion of ata in the meh. The level of interference epen upon how the ata i route in the WiMax network. In thi paper, we conier the following cenario of WiMax-bae meh eployment. A meh network i manage by a ingle noe, which we refer to a Meh B. Meh B erve a the interface for WiMax-bae meh to the external network. We provie an algorithm for interference-aware multi-hop route election for a given capacity-requet matrix, which lea to efficient cheuling. B MH N CN MH-NCFG MH-NENT MH-CCH MH-CCF TABLE I. IEEE MEH MODE ACRONYM Bae tation ubcriber tation Meh ponoring Noe Caniate Noe Meh network Configuration Meage Meh Network Entry Meage Meh Centralize cheuling Meage Meh Centralize cheuling Configuration Meage B. IEEE Meh Moe Operation In IEEE Meh moe, a Meh bae tation (B) provie backhaul connectivity of the meh network an control one or more ubcriber tation (). When centralize cheuling cheme i ue, the Meh B i reponible for collecting banwith requet from ubcriber tation an for managing reource allocation. We will firt introuce the Meh network entry proce (i.e., a proce by which a new noe join the meh), an then we ecribe the network reource allocation requet/granting proceure. In IEEE Meh moe, Meh Network Configuration (MH-NCFG) an Meh Network Entry (MH-NENT) meage are ue for avertiement of the meh network an for helping new noe to ynchronize an to joining the meh network. Active noe within the meh perioically avertie MH-NCFG meage with Network Decriptor, which outline the baic network configuration information uch a B ID number an the bae channel currently ue. A new noe that plan to join an active meh network can for active network an liten to MH-NCFG meage. The new noe etablihe coare ynchronization an tart the network entry proce bae on the information given by MH-NCFG. Among all poible neighbor that avertie MH-NCFG, the joining noe (which i calle Caniate Noe in the Meh moe terminology) elect a potential ponoring Noe to connect to. A Meh Network Entry meage (MH-NENT) with NetEntryRequet information i then ent by the Caniate Noe to join the meh. The IEEE Meh moe MAC upport both centralize cheuling an itribute cheuling. Here, we focu on the centralize meh cheme to etablih high-pee broaban meh connection, where the Meh B coorinate the raio reource allocation within the meh network. In the centralize cheme, every Meh etimate an en it reource requet to the Meh B, an the Meh B etermine the amount of grante reource for each link an communicate. The requet an grant proce ue the Meh Centralize cheuling (MH-CCH) meage type. A ubcriber tation capacity requet are ent uing the MH- CCH:Requet meage to the ubcriber tation parent noe. After the Meh B etermine the reource allocation reult, the MH-CCH:Grant i propagate along the route from Meh B. To ieminate the link, noe, an cheuling tree configuration information to all participant within the meh network, the Meh Centralize cheuling Configuration (MH-CCF) meage i broacate by the Meh B an then re-broacate by intermeiate noe. III. INTERFERNCE-AWARE DEIGN WITH MEH A. Interference-Aware Route Contruction To achieve efficient pectral utilization an high throughput in meh network, the route contruction within the meh network i crucial. To thi en, we propoe an interference-aware route contruction algorithm that conier interference conition in the meh network. The concept of blocking metric B(k) of a given route from the Meh B towar an noe k i introuce to moel the interference level of route in the meh. The blocking metric B(k) of a multihop route inicate the number of blocke (interfere with) noe by all the intermeiate noe along the route from the root noe towar the etination noe k. We alo efine the blocking value b(η) of a noe η, a the number of blocke (interfere with) noe when noe η i tranmitting. Therefore, the blocking metric of a route will be the ummation of the blocking value of noe that tranmit or forwar packet along the route. A imple example of blocking metric computation i hown in Figure 1. In thi example, a noe i blocke when it i within the tranmiion range of the tranmitting noe. Thu, noe η i

3 blocking value b(η i ) equal the number of neighboring noe of η i. The blocking metric along a route towar k, B(k) i equal to the ummation of all the blocking value b(η i ) for every noe η i (incluing the ource noe an all the forwaring noe) on the route k. In the example in Figure 1, B(k) i compute by aing b(), b(n1), b(n2), an b(n3). A hown in Figure 1, noe are blocke if their tranmiion woul interfere with the currently receiving noe. imilarly, Figure 2 how an example of blocking metric computation of a route with larger interference than the example in Figure 1. Our eign approach in the propoe interference-aware cheme i to elect the route with le interference. For the clarity of illutration in Figure 1 an 2, the blocking metric computation preente here how a implifie cae, where only noe within the tranmiion range of a tranmitting noe are blocke. In variou cenario, a tranmitting noe coul interfere with noe that are in a larger itance away. Other type of blocking metric (uch a etaile propagation moel or meaurement with receiver enitivity) coul be efine bae on the information availability an ytem eign traeoff. b(1)=2 n1 b(2)=4 n2 b(3)=3 Figure 1. An exmple of blocking metric B(k)= =13 b (1)=2 b (2)=4 b (3)=5 Figure 2. Blocking metric of an alternative route B(k)= =15 n3 b(4)=4 b (4)=4 We how the interference-aware route contruction algorithm in Figure 3. Beginning with a ingle Meh B noe, we a one into the meh at a time. The time equence of noe η joining the meh i repreente by σ(η). When the noe η join the meh, it will elect the ponoring Noe with the minimum blocking. In the interference-aware route contruction cheme, the blocking metric information i incorporate into the Network Decriptor of a MH-NCFG meage. When a new noe i canning for active network uring the network entry proce, the new noe chooe the potential ponoring Noe bae on the blocking metric information to reuce the interference of the multihop route an hence to improve the throughput. {0} //noe 0 i the root noe; Initialize the et of electe noe N {1,2,..., n} //Initialize the et of unelecte noe p( i), i {1,2,..., n} //Initialize parent noe for noe 1,2,,n Do if END N η argmax σ ( i) //Noe η with high σ(η) value join firt i N Neighbor ( ) C( η) Neighbor ( η) //All noe within tranmiion itance of η W( η) C( η) //Caniate parent noe of noe η p( η) argmin B( i) //elect the noe with minimum blocking A η to i W ( η) Remove η from N Figure 3. Interference-Aware Route Contruction Algorithm B. Interference-Aware cheuling The eign goal of the propoe interference-aware cheuling i to exploit concurrent tranmiion opportunity to achieve high pectral utilization an hence high ytem throughput. The interference-aware cheuling eek to maximize the number of concurrent tranmiion, without creating exceeing interference for other imultaneou tranmiion. Thi i achieve by taking into the conieration the traffic capacity requet of each. We enote the capacity requet of an noe k from the Meh B a D(k). The Meh B grant raio reource accoring to the Application Layer capacity requet, D(k)-, of all noe an the route information of the meh network. We how the interference-aware cheuling algorithm in Figure 4. With the obtaine route information from network entry an the initialization proce, the noe capacity requet D(k) can alo be equivalently repreente in term of link eman Y(j) for every link j. The cheuling algorithm iteratively etermine ActiveLink(t), which i the et of active link at the time t. In each allocation iteration t, a link with the highet unallocate traffic eman i electe for next allocation of a unit traffic. The cheuling algorithm i eigne to fin the maximum number of concurrent tranmiion. To atify the INR contraint of concurrent tranmiion, the Blocke_Neighbor(k) function i ue to exclue interfering link that are locate in the neighborhoo of k. The iterative allocation continue until there i no unallocate capacity requet.

4 t 1 While exit anyy(j)>0 for any link j k arg max Y(j) //elect link k B // et of blocke link in thi iteration A // et of electe active link in thi iteration While k A k to A A Blocke_Neighbor(k) to B k arg max Y(j) En while j A B; Y ( j) > 0 En while ActiveLink(t) A t ( t + 1) Y(j) Y(j)-1 A Figure 4. Interference-Aware cheuling Algorithm IV. PERFRORMANCE EVALUATION We have evaluate the ytem throughput of IEEE meh in two cenario: a linear chain topology an a ranom meh topology. The propoe cheme are compare to the no patial reue MH-CCH cheuling example given in IEEE a tanar [1], which we refer to in thi paper a the baic cheme. To ee how the meh cheme perform, we alo invetigate a theoretical upper-boun bae on linear programming. The network throughput of the baic cheme an the propoe interference-aware cheme are imulate on the Matlab platform. A. Optimal olution From Linear Program A linear programming algorithm i formulate to moel the network throughput upper boun of IEEE meh network. imilar to the moeling technique emontrate in [5], the network activity i moele in term of the normalize time fraction a a real number between 0 an 1. The et repreent all the poible tranmiion cheme in term of link banwith R (bit/econ). In a given tranmiion cheme j with active tranmitting noe t an receiving noe r, (t,j) i et to -R(t,r), the link banwith between the noe pair (t,r), if noe t i tranmitting to noe r, an (r,j) i et to +R(t,r). All other (x, j) are et to 0 if noe x i neither tranmitting nor receiving in the j-th tranmiion cheme. In a tranmiion cheme, multiple concurrent tranmiion are allowe, if all receiving conition are atifie. The variable x(j) repreent the normalize time fraction that the j-th tranmiion cheme i active in unit time. At all receiver, it i require that INR(j)>γ, which i the minimum reception INR threhol. The objective function of thi linear programming algorithm i to maximize the overall network throughput. A reource allocation contraint i impoe to enure fair throughput allocation among uer. En-to-en throughput i proportionally allocate bae on the parameter L; i.e., uer α will be allocate the en-to-en throughput proportional to L(α). In aition, the ummation of all tranmiion time fraction x(j) houl be le than unit time. The optimal olution of the linear programming algorithm, i ue a a performance comparion benchmark for the propoe interference-aware meh network cheme. ince the optimal linear programming algorithm i compute bae on arbitrarily licing of time fraction of all feaible tranmiion combination, the performance of the icretetime event riven eign of propoe cheme i cappe by the performance of the linear programming algorithm. Thu the performance of the linear programming algorithm erve a an upper boun. Maximize uch that i> 1 x( j) ( i, j) L( α ) / x( j) ( α, j) = L( β ) / x( j) ( β, j), α, β {1, 2,..., N} x( j) 1, j {1, 2,..., M} 0 x( j) 1 Figure 5. The Linear Programming Algorithm B. Throughput Performance in Chain Topology The firt cenario conier only the cheuling problem in a chain topology IEEE multihop network. In thi chain topology, route contruction i traightforwar; i.e., ata packet are alway forware along the chain. Here, we invetigate the effectivene of interference-aware cheuling. The throughput of the baic meh network without interference-aware cheuling i compare to that of the propoe cheme an to the upper boun obtaine from the linear programming algorithm. A hown in Figure 6, the propoe interference-aware cheuling cheme performance approache the upper boun, while outperform the baic meh moe ignificantly. Alo a hown in the figure, the number of noe in thi multihop network affect the poibility of concurrent tranmiion an hence averely affect the network throughput. In the baic meh cheme, the network throughput rop ignificantly a the number of noe increae, becaue of the limite patial reue. On the other han, with the interference-aware concurrent tranmiion, the normalize overall throughput egrae ignificantly le a the number of noe increae. The propoe interference-aware cheuling cheme i more calable than the baic cheme. A the length of relay route increae with the number of noe in the network, the overall network throughput ecreae ue to the fact that a packet nee to be forware everal time. If the increae in the egree of patial reue i le than the increae of number of hop, the network throughput will ecreae. Depening on the number of relay hop an the network topology, there are limitation on the egree of patial reue that coul be achieve. By comparing the reult of the optimal linear programming algorithm, we conclue that our propoe

5 cheme coul achieve a near-optimal network throughput an patial utilization. Figure 6. Overall throuhgput of a chain topology network ranomly elect a ponoring Noe in the ranom routing cheme. Thi type of route contruction i enote a ranom routing in Figure 7. Both, the propoe interference-aware cheuling an the baic meh cheuling were imulate. imilar to the chain topology cenario, we icu the overall network throughput ecreae a the number of noe increae. A hown in Figure 7, the cheme with both, the interference-aware routing an the cheuling achieve the highet network throughput. The interference-aware routing only cheme ha the econ bet performance. The interference-aware cheuling only cheme alo outperform the baic meh cheme. A a reult, we conclue that the propoe interference-aware framework coul effectively enhance the baic IEEE Meh moe operation. In the eign proce of cheuling an route contruction in meh network, one houl aopt the interference-aware eign concept an hence exploit the benefit of concurrent tranmiion with le interference. Conequently, the pectral utilization in meh network i enhance with le interference an more patial reue. C. Throughput Performance in Ranom Topology In the econ cenario, we coniere both cheuling an routing in a ranom-topology meh network. Location of a et of meh noe are ranomly generate. The orer of noe joining the meh i not correlate with the noe location, but i ranomly etermine. The meh formation begin with the Meh B noe. Then the noe join one-byone. Any noe that ha alreay joine the meh network coul become a ponoring Noe. When a new noe join the exiting meh network, epening on the number of caniate N within the new noe tranmiion range, it may hear multiple MH-NCFG avertiement meage. V. CONCLUION Allowing concurrent tranmiion to achieve high patial reue i eential for calable wirele meh network eign. We propoe an interference-aware reearch framework for the emerging IEEE Meh moe to improve pectral utilization. Uing thi framework, we introuce an interference-aware route contruction algorithm for meh network initialization proce to improve the network throughput by electing route with minimal interference to exiting noe. In aition, a loa-aware an interferenceaware cheuling algorithm for centralize cheuling in IEEE Meh moe i alo icue. imulation reult how that the propoe cheme effectively improve the network throughput performance in IEEE meh network an achieve high pectral utilization. REFERENCE Figure 7. Throuhgput performance of a ranom topology meh In the interference-aware routing cheme, the new noe elect the ponoring Noe a the caniate N with the minimum blocking metric. For comparion, the new noe [1] IEEE t a-2003, "IEEE tanar for Local an metropolitan area network--part 16: Air Interface for Fixe Broaban Wirele Acce ytem--amenment 2: Meium Acce Control Moification an Aitional Phyical Layer pecification for 2-11 GHz," 2003 [2] IEEE t (Reviion of IEEE t ), "IEEE tanar for Local an Metropolitan Area Network Part 16: Air Interface for Fixe Broaban Wirele Acce ytem," 2004 [3] M. Conti, G. Maelli, G. Turi, an. Giorano, "Cro-layering in mobile a hoc network eign," IEEE Computer, vol. 37, pp , [4]. hakkottai, T.. Rappaport, an P. C. Karlon, "Cro-layer eign for wirele network," IEEE Communication Magazine, vol. 41, pp , 2003 [5]. Toumpi an A.J. Golmith, Capacity Region For Wirele A hoc Network, IEEE International Conference on Communication (ICC), 2002