You are on page 1of 8

2010 5th International Symposium on Wireless Pervasive Computing (ISWPC)

WiMAX spectrum efciency: considerations and simulation results


Aymen Belghith
Universit de Rennes I - IRISA, Campus de Beaulieu, e 35042 Rennes Cedex, France Email: aymen.belghith@univ-rennes1.fr

Oyumaa Norovjav, and Li Wang


TELECOM Bretagne, 2 rue de la ch taigneraie, a CS 17607, 35576 Cesson-S vign Cedex, FRANCE e e Email: {rst.last}@telecom-bretagne.eu

AbstractScheduling algorithms are very important in WiMAX allowing an efcient radio resources distribution to the uplink and downlink connections. However, the IEEE 802.16 standard does not specify which scheduling algorithm(s) should be used. In this paper, we propose a review of scheduling algorithms proposed for WiMAX / IEEE 802.16 system. We also propose some considerations for the spectrum efciency computations of scheduling algorithms in WiMAX. Then, we calculate the satisfaction percentage and spectrum efciency for three scheduling approaches. Figures are given for three scenarios (pessimistic, optimistic and realistic) allowing interesting observations about WiMAX radio efciency. These results can be used as an order of magnitude for more sophisticated studies on this important topic. The paper ends with some interesting comments about WiMAX scheduling.

I. I NTRODUCTION WiMAX is a Broadband Wireless Access (BWA) technology. It is based on the IEEE 802.16 standard [1], [2]. This technology promises a large coverage, high throughput and high spectrum efciency. The main objective of WiMAX is to have a highly-efcient use of the radio resources. The radio resources are distributed to subscriber stations (SS) having different types of services. Each service has its Quality of Service (QoS) parameters such as the trafc priority, maximum sustained trafc rate, minimum reserved trafc rate and maximum latency. Some capacity estimations of the IEEE 802.16 technology were studied in [3] and [4]. The IEEE 802.16 physical (PHY) species ve layers: WirelessMAN-SC, WirelessMAN-SCa, WirelessMANOFDM, WirelessMAN-OFDMA, and WirelessHUMAN PHYs. In this paper, we focus on the WirelessMAN-OFDM PHY which is based on the Orthogonal Frequency Division Multiplexing (OFDM) technology. The IEEE 802.16 Medium Access Control (MAC) species ve types of scheduling service classes or QoS classes: Unsolicited Grant Service (UGS), real-time Polling Service (rtPS), extended real-time Polling Service (ertPS), non realtime Polling Service (nrtPS), and Best Effort (BE) QoS classes. Each QoS class has its mandatory QoS parameters and requirements. The Base Station (BS) must distribute the radio resources to the uplink and downlink connections taking into account the corresponding QoS classes.

The IEEE 802.16 standard also denes the QoS parameters and management messages. However, it lets the scheduling algorithm as an open issue. It is up to vendors and operators to choose or the scheduling algorithm(s) to be used. A number of studies have been published about the scheduling algorithms for WiMAX. Some of these schedulers are previously well-known. There are also schedulers specically proposed for WiMAX tacking into account some characteristics of the different QoS classes. These schedulers consider more or less complex details of the topic and we present a state of the art in Section III. In this paper we also try to highlight considerations for WiMAX scheduling algorithm. The gures obtained give way to some interesting comments. The system model is described in the following Section. Section III summarizes some WiMAX scheduling algorithms. Section IV presents the computations made for WiMAX scheduling algorithms. Section V presents the numerical results. Finally, concluding remarks are made in Section VI. II. S YSTEM M ODEL A. System Model Details The IEEE 802.16 standard denes the physical (PHY) and Mac layers. The PHY layer denes ve physical interfaces. In our system model, only the Orthogonal Frequency Division Multiplexing (OFDM) PHY is considered. The IEEE 802.16 MAC layer denes two modes of transmission; the point-tomultipoint (PMP) and Mesh modes. In the PMP mode, the trafc occurs only between BS and SSs. While in the Mesh mode, the trafc can take place between different SSs. In our case, we consider only the PMP mode. The MAC layer also denes ve QoS classes briey described in the following. UGS supports real-time service ows that have xed-size data packets on a periodic basis. rtPS supports real-time service ows that generate variable data packets size on a periodic basis. ertPS is built on the efciency of both UGS and rtPS. The BS provides unicast grants in an unsolicited manner like UGS. Whereas the UGS allocations are xed in size, the ertPS allocations are dynamic. nrtPS is designed to support non real-time service ows that require variable size bursts on a regular basis. BE is used for best effort trafc where no throughput or delay guarantees are provided. The different characteristics of the ve QoS classes

978-1-4244-6857-7/10/$26.00 2010 IEEE

503

Authorized licensed use limited to: UR Rennes. Downloaded on June 18,2010 at 09:14:27 UTC from IEEE Xplore. Restrictions apply.

TABLE I E SSENTIAL C HARACTERISTICS OF THE Q O S C LASSES QoS Classes (Scheduling Services) UGS rtPS ertPS nrtPS BE Unicast Polling PM, Poll-Me can be used Allowed Allowed Allowed Allowed Contentionbased Polling Not allowed Not allowed Allowed Allowed Allowed

QoS Parameters Periodic, xed size data grants Minimum data rate Periodic data grants Slightly better than BE Best Effort. No unicast request polling obligation for the BS

PiggyBack Grant Request Allowed Not allowed Extended piggyBack Not allowed Not allowed

Example T1/E1 circuit voice MPEG VoIP FTP, HTTP Email, Fax

TABLE II N UMBER OF U SEFUL B ITS PER OFDM S YMBOL Modulation BPSK QPSK QPSK 16-QAM 16-QAM 64-QAM 64-QAM Fig. 1. OFDM downlink frame structure

Coding rate 1/2 1/2 3/4 1/2 3/4 2/3 3/4

Received SIR (dB) 3.0 6.0 8.5 11.5 15.0 19.0 21.0

Number of useful bits per OFDM symbol 192 1 1/2 = 96 192 2 1/2 = 192 192 2 3/4 = 288 192 4 1/2 = 384 192 4 3/4 = 576 192 6 2/3 = 768 192 6 3/4 = 864

are summarized in Table I. Only the UGS and BE QoS classes are considered in the system model of this paper. The IEEE 802.16 standard supports both the frequencydivision duplex (FDD) and time division duplex (TDD) modes. In FDD mode, the downlink and uplink data are transmitted in two different frequencies. Whereas in the TDD mode, the downlink and uplink data are transmitted using the same frequency and differentiated in a duplexing frame. In our system model, only the FDD mode is considered, where we study the scheduling of the downlink. The downlink FDD frame structure contains a preamble, a Frame Control Header (FCH), broadcasted MAC management messages, and downlink bursts (see Fig. 1). B. Considerations for Spectrum Efciency Computations The spectrum efciency indicates the amount of information that can be transmitted over the channel. It is equal to the throughput divided by the bandwidth and expressed in bit/s/Hz. The throughput is computed as follows: Throughput = fs * Nused / NF T T * 1 / (1+G) * d * c where: fs : represents the sampling factory. It is equal to the nominal channel bandwidth BW multiplied by the sampling factor n. Nused : represents the number of useful data subcarriers. For the OFDM PHY, only 192 subcarriers are used for useful data transmission (Nused = 192). NF T T : represents the total number of subcarriers. For the OFDM PHY, the total number of subcarriers is equal to

256 (NF T T = 256). G: represents the ratio of the guard time to the useful symbol time. d: represents the number of bits per symbol of modulation. c: represents the code rate of the Forward Error Correction (FEC).

The throughput directly depends on the Modulation and Coding Scheme (MCS) used. In WiMAX, the following modulations can be used: binary phase shift keying (BPSK), quaternary PSK (QPSK), 16-quadrature amplitude modulation (16QAM), and 64-quadrature amplitude modulation (64-QAM). Each subcarrier of an OFDM symbol is coded with different number of bits according to the MCS used. In Table II, we remind the number of useful bits per OFDM symbol depending on the MCS used. We assume that the SIR is equal to the Signal-to-Noise Ratio (SNR). As the OFDM symbol consists of a useful symbol time, called Tb , and a guard time, called Tg , the throughput also depends on the ratio of Tg to Tb . This ratio is denoted G. The smaller the ratio G is, the lower the data rate. The guard time is used to prevent multipath. Then, a small value of Tg is not recommended. The time duration of an OFDM symbol (Ts ) can be computed as follows: Ts = Tb + Tg Ts = Tb + G * Tb Ts = 1 / (one subcarrier spacing) * (1 + G) Ts = (1 / f ) * (1 + G) Ts = [1 / (fs / NF T T )] * (1 + G) Ts = [1 / (n * BW / NF T T )] * (1 + G)

504
Authorized licensed use limited to: UR Rennes. Downloaded on June 18,2010 at 09:14:27 UTC from IEEE Xplore. Restrictions apply.

Given the values of BW, n, and G, the OFDM symbol duration can be computed. In the proposed system model, the main parameters are: the frame duration is equal to 10 ms, the nominal channel bandwidth is equal to 7 GHz, the sampling factor is equal to 8/7, and the ration of Cyclic Prex (CP), noted G, is equal to 1/8. Therefore, the OFDM symbol duration is equal to 36 s. Since the frame duration is equal to 10 ms, the total number of symbols per frame is equal to 277 symbols. Given the number of the OFDM symbol and the MCS used, we can compute the theoretic throughput. As the OFDM symbols are not used only for data transmission, the effective throughput is less than the theoretic throughput. For example, OFDM symbols are taken for the preamble, Frame Control Header (FCH), and broadcasted MAC management messages such as the downlink map (DL-MAP) and uplink map (ULMAP). One byte is also reserved for the channel coding. Indeed, a 0x00 tail byte is introduced by the FEC (see section 8.3.3.1 of [1]). III. W I MAX SCHEDULING ALGORITHMS As already mentioned, the IEEE 802.16 standard does not specify the scheduling algorithm to be used. The vendors and operators have the choice among many existing scheduling techniques; they can also propose their own scheduling algorithms. In the WiMAX PMP mode, the MAC architecture is centralized at the BS. The BS scheduler is responsible for the whole access control of the different wireless subscribers in the downlink and uplink directions. Therefore, scheduling has to be applied to the downlink and uplink directions in the Base Station (BS). In order to indicate the assignment of the downlink and uplink transmission intervals (or bursts) in each frame, the BS transmits the DL-MAP and UL-MAP MAC management messages, respectively. These messages are transmitted at the beginning of the downlink subframe. When it receives an UL-MAP management message, the SS determines if it can access to the uplink channel during the current frame. Since the SS may have several simultaneous connections, an uplink scheduler is required in each SS. In the WiMAX mesh mode, either a distributed (like in [5] and [6]) or centralized (like in [7] and [8]) scheduler can be used: When using a distributed scheduler, the mesh BS and SSs shall coordinate their transmissions in their two-hop neighborhood. The two-hop neighborhood represents the set of all the extended neighbors of the neighborhood. The neighborhood contains all stations with which a node has direct links. The BS and SSs shall also broadcast their requests, available resources, and grants to all their neighbors. The allocation of resources is performed using a three-way handshake: A station broadcasts its bandwidth request using a Mesh distributed scheduling (MSH-DSCH):Request message.

When receiving an MSH-DSCH:Request message, a station having available resources that t with the request sends an MSH-DSCH:Grant message including all the suggested availabilities. The original requester sends an MSH-DSCH:Grant message in order to conform the schedule. When using a centralized scheduler,the mesh BS is responsible for the bandwidth allocation of a certain number of hop (h) neighborhood. The BS gathers the bandwidth requests from its h-hop neighborhood. Then, it determines the granted resources for each downlink an uplink links. Finally, the BS informs all SSs belonging to its h-hop neighborhood about its decision. In this paper, we consider only the PMP mode. We can distinguish between two types of scheduling algorithms. The rst type is made of the known scheduling methods. It contains the most widely used. The scheduling algorithm can distribute the radio resource between the different connections without taking into account their radio channel conditions or their QoS requirements as in the Round Robin (RR), Weighted Round Robin (WRR) [9], Decit Round Robin (DRR) [10], and prorate schedulers. The scheduling algorithm can take into account the radio channel condition as in the maximum Signalto-Noise Ratio (mSIR) scheduler and Temporary Removal Scheduler (TRS) [11]. The scheduler can take into account some QoS parameters like the latency in the Earliest Deadline First (EDF) [12] scheduler. Some of well-known schedulers can be applied in WiMAX system like in [13] and [14]. The second type represents the schedulers specically proposed for WiMAX. The scheduler can be specically proposed for a single WiMAX QoS class like the Adaptive Bandwidth Allocation Scheme (ABAS) [15] which is proposed for the BE class or the adaptive Polling Service (aPS) [16] which is proposed for the rtPS class. We interested in the rtPS QoS class and our proposition can be applied to all the rtPS schedulers [17]. The scheduler can also take into account the characteristics of many WiMAX QoS classes such as the uplink packet scheduler with Call Admission Control (CAC) mechanism [18], the cross-layer scheduling with QoS support [19], the hybrid scheduling algorithm [20], the Frame Registry Tree Scheduler (FRTS) [21], and the scheduler ensuring QoS requirements. Many optimization criterions can be considered for the scheduling algorithms such as the total maximum data rate, fairness, and operator revenue optimization. The description of some scheduling techniques is presented in the following. A. Known Scheduling Algorithms 1) Systematic schedulers: In this Section, we present some schedulers that systematically serve all subscribers without taking into account their QoS characteristics neither their radio channel conditions. Those systematic schedulers allocate quantum of resources for each subscriber. The quantum can be the same for each SS like for the Round Robin (RR) scheduler. It can depend on the number of symbols requested by all SSs like for the prorate scheduler. The quantum can also be

505
Authorized licensed use limited to: UR Rennes. Downloaded on June 18,2010 at 09:14:27 UTC from IEEE Xplore. Restrictions apply.

xed like for the Decit Round Robin (DRR) scheduler. The throughput of each SS can be different and it depends on its corresponding quantum of resources and its MCS used. a) Round Robin Scheduler: The RR scheduler, also called cyclic scheduler, equitably distributes the channel resources to the multiplexed packet data calls. This technique is adapted if the subscribers have the same trafc and radio characteristics. b) Prorate Scheduler: The prorate scheduler allocates a quantum of resources Qi for each SS i. This quantum depends on the resource demands of all SSs. Let NSS be the total number of SSs, Nsymbols be the total number of symbols, and Si be the number of symbols requested by SS i. Qi is equal Si Nsymbols . Then, the higher the demand of a to NSS subscriber is, the more symbols are allocated to this subscriber. c) Decit Round Robin Scheduler: The DRR scheduler [10] is packet by packet basis. At each round, the DRR scheduler virtually grants a quantum of resource for each connection. The quantum of SS i is never used by other users. Therefore, each connection can use the non-used part of its quantum in the next rounds. For that, the DRR scheduler associates to each queue i: Q(i): represents the xed quantum granted to queue i at each round. DC(i): represents a decit counter. This counter is used for fair purposes. At the start of each round, DC(i) is incremented by Q(i) for each queue i. The head of queue i is eligible to be dequeued if DC(i) is greater than the length of the packet waiting to be sent, noted L(k). In this case, DC(i) is decremented by L(k). At each round, one packet at most can be dequeued for each queue. 2) Radio channel conditions-aware schedulers: In this Section, we present some schedulers that take into account the radio channel conditions. a) Maximum Signal-to-Interference Scheduler: The mSIR scheduler allocates the radio resources to subscriber stations (SS) having the highest Signal-to-Interference Rate (SIR). Then, this scheduler offers high spectrum efciency. Nevertheless, subscribers having a SIR that is always small may never be served. b) Temporary Removal Scheduler: The main objective of the temporary removal scheduler (TRS), dened in [11], is the determination of the set of subscribers that can be scheduled. This set is called scheduling list. The SSs having bad radio conditions are temporarily blocked. The TRS performs as follows. It identies the packets under a denite power radio threshold. These packet calls are temporarily removed from the scheduling list for a certain adjustable time period TR . If TR expires, the temporarily removed packet is checked again. If the radio conditions are still poor, this packet is temporarily removed for another time period TR . The whole process is repeated up to L times. When the packet is removed for a period of L TR , it is added to the scheduling list independently of the current radio conditions. Then, a penalty
k=1

Fig. 2.

Allocation of symbols for rtPS connection.

Sk

time Tp prevents the packet call from being immediately selected once more. If an improvement is observed in the radio channel, the packet could be topped up in the scheduling list again. The TRS can be useful especially when the channel radio conditions are uctuant. Since this scheduler only determines the SSs that can be scheduled, it shall be combined with another scheduler in order to determine the number of symbols that will be allocated to each SS belonging to the scheduling list. B. Scheduling Algorithms Specically Proposed for WiMAX In this Section, we present some Full and Partial WiMAX QoS-Aware schedulers. The schedulers are specically proposed for WiMAX systems and take into account the characteristics of the QoS classes. 1) Schedulers proposed for a specic WiMAX QoS class: a) Adaptive Polling Service: A novel adaptive Polling Service (aPS) mechanism for the rtPS QoS class is dened in [16]. The main idea of the proposed mechanism is to adjust the polling period. The polling period represents the period of the sending of two successive unicast request opportunities. The BS initializes the polling period with Tmin . Parameter Tmin is determined using the average packet arrival rate. If the BS does not receive bandwidth request after N polls, it exponentially increases the polling period until reaching Tmax . Parameter Tmax is determined using the tolerable delay of the connection since the aPS mechanism is dened for real-time applications. When the BS receives a bandwidth request, it resets the polling period to Tmin . b) Proposals for the rtPS QoS class: We highlight a problem that may exist with the rtPS QoS class. If the BS allocates unicast request opportunities and resource grants for rtPS connections in the same frame, the BS cannot immediately take into account the new length of the uplink data connection of the subscriber. The reason is that the BS allocates symbols for rtPS connections before receiving the latest unicast bandwidth request (see Fig. 2). In [17], we propose that the BS only serves the subscribers who do not have unicast request opportunities in the same frame. This proposition can be applied to all rtPS schedulers. In order to assess our proposition, we modify, at a rst stage, a scheduler that provides higher throughput. We choose

506
Authorized licensed use limited to: UR Rennes. Downloaded on June 18,2010 at 09:14:27 UTC from IEEE Xplore. Restrictions apply.

the mSIR scheduler and our modied scheduler is called maximum Signal-to-Interference Ratio (mmSIR). The proposed scheduler modies the classical mSIR scheduler in order to decrease the mean sojourn time while not degrading the throughput. The main steps of the proposed scheduler are described as follows. The BS sorts the SSs with decreasing SNR values. If the next SS to serve has a unicast opportunity in the next frame, the BS does not allocate an uplink burst to this subscriber. Otherwise, the BS serves the SSs having the highest SNR in order to increase the capacity of the WiMAX system. The performance of the proposed scheduler is evaluated using our QoS-including WiMAX module [22]. We add all WiMAX QoS classes as well as the WiMAX mechanisms used for the addition of connections as specied by the IEEE 802.16 standard. We also implement the mechanisms required to send bandwidth requests in order to specify the amount of radio resource to reserve. As the IEEE 802.16 standard supports many MCSs and the MCS to use depends on the channel quality of the subscriber, we add a link adaptation mechanism. We then implement some scheduling algorithms in order to compare between them and estimate the WiMAX capacity that depends on the scheduler used and the radio channel characteristics. 2) Schedulers proposed for WiMAX QoS classes: a) Uplink packet scheduler with Call Admission Control (CAC) Mechanism: The Call Admission Control (CAC) mechanism of [18] is based on token bucket principle. The token bucket is a mechanism used to control network trafc rate. The uplink packet scheduler algorithm proposed in [18] works as follows. First, All the UGS connections are granted. Then, the CAC is applied to the rtPS packets. The deadline of these packets is calculated. An Earliest Deadline First (EDF) scheduler is applied for the attribution of the rtPS connections. The EDF scheduler attributes priorities to different packets according to their deadlines. The closer is the deadline of a packet; the higher is its priority. Then, the nrtPS connections are allocated if there is remaining bandwidth and the bandwidth requirements are below a threshold, called TnrtP S . Finally, the BE connections are allocated if there is remaining bandwidth and the bandwidth requirements are below a threshold, called TBE . If there is remaining bandwidth, the nrtPS connections and then the BE connections are granted until use of all the available bandwidth. b) Cross-layer Scheduling Algorithm with QoS Support: In [19], a novel scheduling algorithm is proposed for WiMAX networks. This scheduler is based on affecting a priority to each connection. For the UGS connections, the scheduler must guarantee a xed quantum of radio resources. The UGS connections are characterized by a constant number of time slots allocated. Therefore, the transmission mode is selected and remains the same during the whole service time. For the rtPS connections, the scheduler must guarantee the

latency. The scheduler allocates all the residual time slots to the connection that has the maximum value of a dened priority function (PRF). For the nrtPS connections, the scheduler must guarantee the minimum reserved rate i . The scheduler allocates all the residual time slots to the connection that has the maximum value of a dened PRF. For the BE connections, the scheduler has no QoS parameters to guarantee. The scheduler allocates all the residual time slots to the connection that has the maximum value of a dened PFR. The PRFs, for rtPS, nrtPS, and BE connections, are dened in [19]. c) Hybrid Scheduling Algorithm: The hybrid scheduling algorithm for QoS in WiMAX in [20] works as follows. The Earliest Due Date (EDD) scheduler is used for the real time services while the Weighted Fair Queue (WFQ) scheduler is used for the non-real time services. This is then a hybrid scheduling algorithm. The EDD scheduler is based on dynamic priority. In an EDD queue, the packets are classied in order of their deadline values. The expected deadline time of a packet is calculated by adding the packet arrival time and maximum service time of this packet. The WFQ scheduler provides a required throughput rate for each service. The delay in WFQ for a service is computed as follows: DW F Q =
n i=1

wi / (R * wi )

where: wi : represents the weight given to the queue i. n: represents the number of services. R: represents the link transmission rate. 3) Synthesis: The RR, mSIR, and Prorate scheduler are not suitable to use in a WiMAX context for most of the cases. Indeed, the RR scheduler is only suitable for subscribers who have the same characteristics such as the SNR, QoS requirements, and the trafc load. The mSIR and Prorate schedulers may block subscribers having a poor SIR value or less data quantity, respectively. Some schedulers can be used only in one direction. For example, the DRR scheduler can be used only in the downlink trafc because the BS does not know the queue length of the subscribers. The uplink scheduler with CAC mechanism is only proposed for the uplink trafc since the CAC is performed to control the connections that are transmitted by the subscribers. The TRS, cross-layer, and hybrid schedulers can be used in the downlink and uplink trafc. A use of the hybrid (EDD+WFQ) schedulers requires the determination of the suitable weights. The removal time, penalty time, and maximum number of repetitions are the main parameters of the TRS scheduler. Different values of these parameters completely change the behavior and performance of these schedulers. We also note that some schedulers are proposed specically for a single WiMAX QoS class. For example, the aPS and our

507
Authorized licensed use limited to: UR Rennes. Downloaded on June 18,2010 at 09:14:27 UTC from IEEE Xplore. Restrictions apply.

mmSIR schedulers are proposed for the rtPS QoS class in the uplink direction. The parameters of the stochastic algorithm of the adaptive rtPS scheduler have to be dened. Our proposed scheduler requires only to specify the period of the unicast request opportunities for each SS. There are schedulers proposed for many WiMAX QoS classes. Some of them take into account only the deadline of the real-time applications and do not provide their minimum reserved rates like the cross-layer and Hybrid (EDD+WFQ) schedulers. The minimum reserved rates of nrtPS connections are not also taken into account by these schedulers. Some of them may block the non-real-time applications like the uplink scheduler with CAC mechanism and scheduler ensuring QoS requirements. IV. C OMPUTATIONS FOR W I MAX S CHEDULER The choice of schedulers is challenging because of per connection translation of the QoS requirements to the scheduler conguration. We propose in a rst stage to study three scheduling techniques at the base station in the downlink direction and for UGS and BE QoS classes. Those scheduling techniques are the RR, mSIR, and Prorate schedulers. They provide interesting reference gures. A program is written to make the computations for the WiMAX scheduling. The language used is Java. The frames are generated as dened in the IEEE 802.16 standard in the FDD mode. The downlink subframe contains preamble, FCH, broadcast messages and downlink bursts. The DL-MAP and UL-MAP broadcast management messages use the BPSK 1/2 MCS. We focus on the downlink subframe because we only consider the downlink scheduler in this paper. The main program parameters are shown in Table III. The program produces the distribution of symbols per user and calculates the relevant parameters needed for analyzing the performance of the scheduling techniques. The network is made of one BS, three SSs that use UGS connection and nine SSs that use BE connections. In our scheduling algorithms, the UGS connections are served rst as this QoS class has the highest priority. The, the BE connections are scheduled in three different scheduling algorithms: the RR, mSIR, and Prorate schedulers. The admission control is such that all the UGS connections can be satised. In addition, we choose the network demand such that at least one BE connection is not completely satised in order to estimate our scheduling algorithms behavior. The main procedure of scheduling is described in the following. First, the SIR is randomly generated for all the SSs. The uniform distribution is used and the SIR values are between 3 dB (that represents the minimum threshold to use the worst efcient MCS, BPSK 1/2) and 24 dB (that is upper than the minimum threshold to use the most efcient MCS, 64-QAM). Then the modulation and coding rate is chosen for every SS according to Table II. With the given system parameters (modulation type, user number, required trafc rate, etc), we determine the total symbols number that are available for data trafc and the required symbols number for

TABLE III M AIN PARAMETERS OF THE S IMULATIONS Type PHY layer Parameters Bandwidth Modulation type Values 7 MHz BPSK 1/2, QPSK 1/2, QPSK 3/4, 16-QAM 1/2, 16-QAM 3/4, 64-QAM 2/3, and 64-QAM 3/4. 1/8 8/7 maximum SIR, Prorate, and Round Robin schedulers. 10 ms 3 9 UGS1 = 50 kbit/s, UGS2 = 1 Mbit/s, and UGS3 = 90 kbit/s. BE1 = 2 Mbit/s, BE2 = 4.5 Mbit/s, BE3 = 1.9 Mbit/s, BE4 = 3 Mbit/s, BE5 = 5 Mbit/s, BE6 = 4 Mbit/s, BE7 = 2 Mbit/s, BE8 = 2 Mbit/s, and BE9 = 2 Mbit/s

MAC layer

G (ratio of CP time to useful time) Sampling factor Type of scheduling Frame duration Number of UGS connections Number of BE connections Required trafc rates per UGS connection Required trafc rates per BE connection

General

each SS. After the allocation of the necessary symbols to the UGS connections, the BE connections have to be scheduled. This is done in three different scheduling techniques. With the RR technique, it should be noted that the trafc which requires less symbol is the rst be completely served. Then, the symbols are uniformly distributed for the remaining BE connections until all remaining symbols have been used. With the mSIR technique, we sort the BE connections in the decreasing order of the SIR. The connection which has the higher SIR is served at rst, and so on, until all the available symbols are allocated. With the Prorate technique, the number of symbols that are allocated for each BE connection is a function of the percentage of necessary symbols to the total number of available symbols. The different steps of the three scheduling techniques are summarized in Fig. 3. Recall that all the implemented schedulers consider the management MAC messages such as DL-MAP and UL-MAP. These massages affect the throughput since their lengths depend on the number of served SSs. V. N UMERICAL R ESULTS We consider three simulation scenarios in order to compare the scheduling techniques. We study the behavior of the RR, mSIR, and Prorate schedulers in three scenarios: bad radio conditions (pessimistic), ideal radio conditions (optimistic), and random radio conditions (realistic). In the pessimistic scenario, bad radio conditions are considered. All the SSs use the most robust MCS (BPSK 1/2). This MCS is applied when the SIR is between 3 dB and 6 dB (see Table II).

508
Authorized licensed use limited to: UR Rennes. Downloaded on June 18,2010 at 09:14:27 UTC from IEEE Xplore. Restrictions apply.

TABLE V S ATISFACTION P ERCENTAGE AND T HROUGHPUT Scenario Scheduler Prorate Pessimistic mSIR RR Prorate Optimistic Fig. 3. Different steps of scheduling mSIR RR Prorate Realistic mSIR RR 3.169 1.214 3.194 2.380 3.169 1.828 Satisfaction percentage Values between 8.14%, and 8.62%. BE1 = 100%, BE2 = 6.19%, and BE3-9 = 0%. Values between 5%, and 12.45%. Values between 80%, and 83.34%. BE1-6 = 100%, BE7 = 70.84%, and BE8-9 = 0%. Values between 58.63%, and 100%. Values between 30.19%, and 31.43%. BE1-3 = BE6-7 = 100%, 91.43%, and BE1 = BE3 = BE6-8 = 0%. Values between 12.92%, and 100%. Throughput 2.22 Mbit/s 2.28 Mbit/s 2.22 Mbit/s 21.94 Mbit/s 22.11 Mbit/s 21.94 Mbit/s 8.26 Mbit/s 16.42 Mbit/s 12.56 Mbit/s

TABLE IV S PECTRUM E FFICIENCY VALUES Scenario/Scheduler Pessimistic scenario (SIR = 4dB) Optimistic scenario (SIR = 22dB) Realistic scenario (SIR = random) Prorate 0.352 Maximum SIR 0.360 Round Robin 0.352

In the optimistic scenario, ideal radio conditions are considered. All the SSs use the most efcient MCS (64-QAM 3/4). This MCS is used when the SIR is greater than 21 dB (see Table II). Therefore, it offers the highest spectral efciency. In the realistic scenario, random radio conditions are considered. Hence, the SSs may have different MCSs. The purpose of this scenario is to test the provisioning of the QoS guarantees when the subscribers use different MCSs. Since the allocation of the necessary number of symbols to the UGS QoS class is mandatory, all the UGS connections must be satised rst in all the scenarios. The throughput of the BE connections is then related to the amount of free resources left when the QoS requirements of all the UGS class SSs are satised. The numerical results are summarized in Table IV and Table V. The spectral efciency values obtained are shown in Table IV. For more details, Table V presents the satisfaction percentage and throughput for BE connections. For the pessimistic and optimistic scenarios, all the SSs have the same SIR. Table IV shows that the RR and Prorate schedulers provides the same spectrum efciency. This is expected as the MCS used in these cases is the same for all the SSs. Moreover, these schedulers serve the same number of SSs per frame. Then, the number of symbols used for the preamble, FCH, and broadcast management messages is the same. For the pessimistic and optimistic scenarios, although the RR and Prorate schedulers provide the same throughput, they have different satisfaction percentages (see Table V). This is due to the fact that the RR scheduler fairly distributes the symbols independently of the required trafc while the distribution

of the symbols, using the Prorate scheduler, depends on the radio resources requirements. In spite of the same SIR used in the pessimistic and optimistic scenarios, Table IV shows that the mSIR scheduler slightly outperforms the other schedulers. This is because the mSIR scheduler serves only the SSs that have higher SIR while the RR and Prorate schedulers serves all the SSs in every frame. Moreover, the higher the number of served SSs per frame, the smaller is the number of useful data symbols. Table IV also shows that the spectrum efciency is less than 0.36 bit/s/Hz for the pessimistic scenario. As it is expected, this scenario has the lowest throughput value due to the use of the BPSK 1/2 MCS. Using the most efcient MCS, the optimistic scenario provides the highest spectrum efciency with values higher than 3.169 bit/s/Hz. We also note from Table V that the total throughput of the optimistic scenario is nearly ten times better than that of the pessimistic scenario. Indeed, the number of useful bits per OFDM symbol when the MCS used is 64-QAM 3/4 is nearly ten times better that when the MCS used is BPSK 1/2 (see Table II, section II). In the realistic scenario, we recall that we study resources allocation when different MCSs are used in SSs. All the UGS connections are always provided with the number of symbols they need. Table V shows that, using the Prorate scheduler, the BE connections are served at a percentage between 30.19 % and 31.43 %. The satisfaction percentage values are very close because the required trafc is taken into account. Table Valso shows that, using the mSIR scheduler, ve BE connections are satised entirely, one connection is satised at 91.43 %, and the others connections are not served. This is because the mSIR scheduler may never serve SSs having small SIR. Using the RR scheduler, the BE connections have great differences

509
Authorized licensed use limited to: UR Rennes. Downloaded on June 18,2010 at 09:14:27 UTC from IEEE Xplore. Restrictions apply.

between the satisfaction percentage values. These values are between 12.92 % and 100 %. We notice that the RR scheduler is not a proper choice especially with the fact that the data rate for each of the BE connection is not the same. This for sure will not have a good fairness. It be better to use the WRR scheduler which distributes the available bandwidth among BE connections according to their trafc weight. In the realistic scenario, Table IV shows that the mSIR scheduler provides the highest spectrum efciency with a value of 2.38 bit/s/Hz. This is due to the taking into account the SIR of the different subscribers. From Table IV, we conclude that the Prorate scheduler is the least efcient with a spectrum efciency value equal to 1.214 bit/s/Hz. Indeed, a subscriber can have high trafc and low SIR. On the other hand, the BS takes into account only the trafc requirements when it schedules the radio resources. VI. C ONCLUSION In this paper, we have studied the scheduling techniques in the WiMAX technology. The choice of the scheduling algorithm is highly dependent on the transmission service type and the trafc shape in addition to other QoS requirements. We have proposed a synthesis for some of the proposed scheduling algorithms. The efciency of a scheduling algorithm can be estimated through simulations. We have compared between three scheduling algorithms: the RR, mSIR and Prorate schedulers. Our goal was to highlight some order of magnitudes useful for more sophisticated WiMAX scheduling algorithms studies. We have estimated the difference between the realistic spectrum efciency (between 1.214 and 2.38 bit/s/Hz) and the optimistic spectrum efciency (3.194 bit/s/Hz). In our results, we have compared the minimum and maximum gures of the spectrum efciency. We have veried that the mSIR scheduler has the highest spectrum efciency. The more sophisticated scheduling algorithms, some of them being cited in this paper, should optimize one or another of the optimization criterions or a combination of many criterions with respect to the multi-QoS level environment of WiMAX. In this work, we do not consider Adaptive Antenna Systems (AAS). AAS use leads to average SIR increase. MultipleInput Multiple-Output (MIMO) systems use allows the use of a better MCS for the same SIR values. Consequently, the use of intelligent antennas techniques would give a higher spectrum efciency gure for the realistic scenario. The study of scheduling algorithms in a WiMAX environment with intelligent antennas can be a future topic of research. R EFERENCES
[1] IEEE 802.16-2004, IEEE Standard for local and metropolitan area networks, Air Interface for Fixed Broadband Wireless Access Systems, Oct 2004. [2] IEEE 802.16e, IEEE Standard for local and metropolitan area networks, Air Interface for Fixed Broadband Wireless Access Systems, Amendment 2: Physical and Medium Access Control Layers for Combined Fixed and Mobile Operation in Licensed Bands and Corrigendum 1, Feb 2006 (Approved: 7 Dec 2005).

[3] L. Nuaymi and Z. Noun, Simple Capacity Estimations in WiMAX/802.16 System, the 17th Annual IEEE International Symposium on Personal Indoor and Mobile Radio Communications, PIMRC2006, Helsenki, 11 - 14 September 2006. [4] A. Belghith and L. Nuaymi, WiMAX capacity estimations and simulation results, IEEE 67th Vehicular Technology Conference, VTC2008Spring, Marina Bay, Singapore, 11 - 14 May 2008. [5] H. Zhu, Y. Tang and I. Chlamtac, Unied collision-free coordinated distributed scheduling (CF-CDS) in IEEE 802.16 mesh networks, IEEE Transactions on Wireless Communications, Vol. 7, October 2008. [6] D. Teng, S. Yang, W. He, and Y. Hu, TEOS: A Throughput-Efciency Optimal Distributed Data Subframe Scheduling Scheme in WiMAX Mesh Networks Wireless Communications, the 4th International Conference on Wireless Communications, Networking and Mobile Computing, WiCOM 08, Dalian, China, 12 - 14 October 2008. [7] J. El-Najjar, B. Jaumard, and C. Assi, Maximizing Network Stability in a Mobile WiMax/802.16 Mesh Centralized Scheduling, the 4th IEEE International Conference on Wireless and Mobile Computing, Networking and Communications, WIMOB 08, Avignon, France, 12 - 14 October 2008. [8] A. Al-Hemyari, N. K. Noordin, A. Ismail, S. Khatun, Y. H. Tahir, Y. A. Qassem, Centralized scheduling, routing tree in WiMAX mesh networks, Innovations in Information Technology, IIT 2008, Madras, India, 16 - 18 December 2008. [9] J. Liebeherr, Packet scheduling, Technical report, Lab 3, February 2007. [10] T. P. Lee and G. Mercankosk, Decit round robin favors longer documents, TENCON 2006, IEEE Region 10 Conference, Hong Kong, China, 14 - 17 November 2006. [11] C.F. Ball, F. Treml, X. Gaube, and A. Klein, Performance Analysis of Temporary Removal Scheduling applied to mobile WiMAX Scenarios in Tight Frequency Reuse, the 16th Annual IEEE International Symposium on Personal Indoor and Mobile Radio Communications, PIMRC2005, Berlin, 11 - 14 September 2005. [12] K. M. Alsayed, Enhancing the End-to-End Schedulability Condition of EDF Scheduling for Real-Time Applications, ATM Workshop Proceedings, 1998 IEEE, May 1998. [13] C.F. Ball, E. Humburg, K. Ivanov, and F. Treml, Comparison of IEEE802.16 WiMAX Scenarios with Fixed and Mobile Subscribers in Tight Reuse, the 14th IST Mobile & Wireless Communication Summit, Dresden, 19 - 23 June 2005. [14] J. Chen, W. Jiao, and Q. Guo, An Integrated QoS Control Architecture for IEEE 802.16 Broadband Wireless Access Systems, Global Telecommunications Conference, 2005, GLOBECOM05, 28 November - 2 December 2005. [15] C. Chiang, W. Liao, and T. Liu, Adaptive Downlink/Uplink Bandwidth Allocation in IEEE 802.16 (WiMAX) Wireless Networks: A Cross-Layer Approach, Global Telecommunications Conference, 2007, GLOBECOM 07, Washington, DC, USA, 26 - 30 November 2007. [16] C. Nie, M. Venkatachalam, and X. Yang, Adaptive Polling Service for Next-Generation IEEE 802.16 WiMAX Networks, Global Telecommunications Conference, 2007, GLOBECOM 07, Washington, DC, USA, 26 - 30 November 2007. [17] A. Belghith and L. Nuaymi, Comparison of WiMAX scheduling algorithms and proposals for the rtPS QoS class, 14th European Wireless 2008, EW2008, Prague, Czech Republic, 22 - 25 June 2008. [18] T. Tsai, C. Jiang, and C. Wang, CAC and Packet scheduling Using Token Bucket for IEEE 802.16 Networks, journal of communications, Vol. 1, No. 2, May 2006. [19] Q. Liu, X. Wang, G. B. Giannakis, and A. Ramamoorthly, A CrossLayer Scheduling Algorithm With QoS Support in Wireless Networks, IEEE Transactions on Vehicular Technology, 3 May 2006. [20] K. Vinay, N. Sreenivasulu, D. Jayaram, and D. Das, Performance Evaluation of End-to-end Delay by Hybrid Scheduling Algorithm for QoS in IEEE 802.16 Network, Wireless and Optical Communications Networks, 2006 IFIP International Conference on, 11 - 13 April 2006. [21] S. A. Xergias, N. Passas, and L. Marekos, Flexible Resource Allocation in IEEE 802.16 Wireless Metropolitan Area Networks, the 14th IEEE Workshop on Local and Metropolitan Area Networks, LANMAN 2005, Chania, Greece, 18 - 21 September 2005. [22] A. Belghith and L. Nuaymi, Design and implementation of a QoS included WiMAX module for NS-2 simulator, First International Conference on Simulation Tools and Techniques for Communications, Networks and Systems, SIMUTools 2008, Marseille, France, 3 - 7 March 2008.

510
Authorized licensed use limited to: UR Rennes. Downloaded on June 18,2010 at 09:14:27 UTC from IEEE Xplore. Restrictions apply.

You might also like