Fair Bandwidth Assignment in Hierarchical Scheduling for Mobile WiMAX System

Ali Mohammed Alsahag*1, Borhanuddin Mohd Ali*1,2, Nor Kamariah Noordin*3,

and Hafizal Mohamad+4 *Dept. Computer & Comms. Systems Engineering, University Putra Malaysia, Selangor, Malaysia

2Institute of Advanced Technology, Universiti Putra Malaysia, Selangor, Malaysia +Wireless Communications, MIMOS Berhad, Technology Park Malaysia, 57000 Kuala Lumpur, Malaysia

1alkazmi_jo@yahoo.com, 2borhan@eng.upm.edu.my, 3nknordin@eng.upm.edu.my, 4hafizal.mohamad@mimos.my

Abstract— The increasing demands for multimedia applications with various QoS requirements arouse the interest of researchers in the Fourth-generation wireless networks such as WiMAX. In order to ensure that the QoS requirements of these applications are met, effective scheduling algorithms must be designed. Even though it may be trivial to ensure that the minimum QoS of all service classes is attained, this often results in a marked degradation of the overall system throughput. In this paper, we propose a fair bandwidth assignment algorithm that allocates the bandwidth among different services classes based on a hierarchical scheduler. By taking the overall system throughput and the QoS requirements into consideration, our proposed algorithm dynamically assigns the available bandwidth to the various service classes in such a way that the network resource utilization is optimized. Simulations result showed that the proposed algorithm optimize the overall system throughput and assigns bandwidth effectively to the different service classes while ensuring that the QoS requirements are satisfied.

Keywords: Mobile WiMAX, MDRR, QOS, Scheduler

I. INTRODUCTION With the increasing popularity of broadband wireless

access networks and multimedia applications the demand for Quality-of-Service (QoS) becomes a key objective in the design and deployment of wireless communication networks. IEEE 802.16 which has the ability to provide QoS requirements for a variety of applications; the QoS are fairness, throughput, packet loss, delay and delay jitter have been developed to fulfill these characteristics[1, 2].

Scheduling algorithms and resource management are the

main factors to provide the required QoS. Several bandwidth allocation schemes have been proposed to tackle the drawbacks of class service applications in IEEE 802.16[3-5]. The paper in[6] described a two-tier hierarchical scheduling; in the first tier a technique called Deficit Fair Priority Queuing (DFPQ) assigns the total available bandwidth for downlink(DL) and uplink(UL) services, Since DFPQ cannot ensure QoS requirements for real time services such as Voice Over IP(VoIP) streaming, a variant of this algorithm has been proposed in[7],[8] where DFPQ gives more bandwidth to Real-time Polling Services (rtPS) but This consequently starves Best Effort (BE) traffic. On the other hand , the authors in[9], defined three levels of priority. Prioritizing additional rtPS flow may solve the problem of interrupting rtPS packets but, this

may lead to a situation where rtPS flows gains arbitrarily large access at the expenses of BE and Non-real time Polling Service (nrtPS) flows. To overcome this shortcoming,[10] proposed a Customize Deficit Round Robin (CDRR) with the technique of additional queue, it considers only the real-time packets, which are just prior the deadline. However, using additional queue for the rtPS applications results in additional access latency and bandwidth allocation disorder where errors occur and the flow becomes backlogged, which is undesirable, especially if the flow is of type a high priority.

There are several scheduling algorithms with different design goals [11-13]. However, up-to-date trivial algorithms have been performed to create effective scheduler structure with significant performance in the two-tier hierarchical scheduler, which requires further optimization.

In this paper, we propose a fair bandwidth assignment algorithm, for the first-tier of the hierarchical scheduler, which is an Adaptive Deficit Round robin (ADRR). The idea is to assign required amount of bandwidths with respect to the weight demands and load traffic. Based on the QoS requirements, an amount of bandwidth is assigned to the real-time classes of service while ensuring QoS satisfaction for non real-time class of services.

The organization of the rest of this paper is as follows: Section II describes two-tier hierarchical scheduler and the proposed algorithm of the scheduler. The setting parameters for the simulation are presented in Section III performance evaluation presented in Section IV followed by results and discussion in section V. Section VI concludes the paper.

II. HIERARCHICAL SCHEDULER STRUCTURE Figure 1 shows the design of a two-tiered hierarchical

WiMAX systems scheduler in a base station (BS).

In the first-tier, a scheduling algorithm distinguishes and grants priority to downstream and upstream traffic.

In the second-tier, conventional or hybrid algorithms can be used to schedule the different QoS service classes. Incoming packets are categorised by the connection classifier against the connection identifications (CID), which are then sent to their corresponding service class queues. The scheduler algorithm provides adequate bandwidth for each service class based on its minimum requirements and priority. Since incoming traffic of the same service class share similar QoS requirements, each

2011 17th Asia-Pacific Conference on Communications (APCC) 2nd – 5th October 2011 | Sutera Harbour Resort, Kota Kinabalu, Sabah, Malaysia

978-1-4577-0390-4/11/$26.00 ©2011 IEEE 743

class scheduler selects a scheduling algorithm that best fulfills the QoS requirements. In this way, the two-tier scheduler exhibits a superior scheduling mechanism over the one-tier scheduler.

Figure 1: Uplink Hierarchical Scheduler

A. Inter-Class Schedulers: The bandwidth allocation algorithm plays a critical role on

the resource utilization of the class service schedulers. If the first-tier algorithm does assign adequate bandwidth to the second-tier scheduler, there is no assurance that the required QoS of the corresponding service class will be met. In contrast, if the scheduler assigns excessive bandwidth to the class scheduler, bandwidth will be wasted. So the design for a fair bandwidth scheduler algorithm must be effective. The proposed Adaptive Deficit Round Robin (ADRR) ensures that the BE and nrtPS are not starved of bandwidth and offers better performance to the overall system throughput.

• Adaptiv Dificit Round Robin:

ADRR is a variant of the Deficit Round Robin (DRR) algorithm[14]. The key feature of ADRR is the low-latency queue, which allows strict priority queue with delay-sensitive data such as voice to be dequeued and start first before packets in other queues are dequeued. The ADRR operates in two modes: Alternate Mode, when the queue is alternated between the no low-latency queues and strict priority mode, where no other queue can be served until the low-latency queue is totally served. The Maximum Transmission Unit (MTU) is the maximum packet size that could be served in one round while the quantum value (Q) is defined as the average number of bytes served in each round. This can be calculated using equation (3). Q MTU w 1 Δ 3

Where Δ is a constant equal to 512, w is the weight of the actual queue defined by equation (4). w R ∑ R N 4 Q should be greater than zero in order to give the opportunity to send at least one packet in the first round this is shown in equation (5) [15]. Finally, if the queue is empty, the deficit counter will be reset to zero avoiding a waste of symbols in the MAP creation [16]. MTU Q 51.200 MTU 5

In a normal Modified Round Robin (MDRR), fixed weight is assigned to every connection[17]. However this causes complication to the service class scheduling in IEEE 802.16e, which results in deteriorated QoS. In this paper, we propose a fair scheduling assignment scheme based on the ADRR algorithm; we adopt a dynamic weight assigning strategy which is a modification of Q and w to work with our algorithm in order to handle the overall system throughput while maintaining the QoS requirements. The queue is prioritized based on the weight assigned to it. The diverse service flows are assigned different weights using our scheme.

The main objective of this proposed bandwidth assignment scheme is to dynamically allocate the available bandwidth to the various service classes in such a way that the overall system throughput is optimized without sacrificing their QoS requirements. Specifically, the following objectives must be met: to maximize the number of subscribers, to fulfill their QoS requirements and the overall system throughput.

Since Unsolicited Grant Scheme (UGS) scheduling class has been defined by the standards we assign the required bandwidth based on their constant-bit-rate (CBR) requirements

NUGS i UGS RB 6

Where RB is the required bandwidth by UGS connection i. Let NT be the total bandwidth in each frame, then the remaining bandwidth after assigning UGS class will be distributed to the other service classes as equation (7). NR NT – NUGS 7

For real-time class services Extended real-time Polling Service (ErtPS) and rtPS, the amount of bandwidth assigned is based on the number of requests. The bandwidth formula assignment for class services is as follows:

• Step 1: for each round, count the number of requests and get their constraint information (minimum bandwidth requirements and delay deadline).

• Step 2: assign the minimum required bandwidth for ErtPS and rtPS class according to equation (8) R 8

• Step 3: calculate the number of requests for nrtPS and BE class, and then assign the minimum requirements

• Step 4: After the scheduler has finished assigning the minimum requirements for bandwidth for all the requested service classes, it looks up the residual bandwidth. If this bandwidth is larger than the amount of bandwidth required by the rtPS and ErtPS class services, the scheduler assigns excessive bandwidth to ErtPS and rtPS classes respectively based on their minimum required weight according to equation(9). R 2 R2 9

Where R is the minimum requirement for each connection.

744

• Step 5: the residual bandwidth is assigned to nrtPS and BE classes, respectively.

B. Intra-class algorithms In the following part we assign the bandwidth for each

service class.

1) Allocating bandwidth to UGS services: UGS is defined by the standard to serve the real-time application with fixed data size such as T1/E1 and VoIP. BS provides grants for this type in unsolicited behavior.

2) Allocating bandwidth to (ErtPS, rtPS) services: The applications for ErtPS and rtPS services are delay sensitive and has strict delay requirement such as Moving Pictures Expert Group (MPEG) video, We allocate bandwidth for ErtPS and rtPS using Earlier Deadline First (EDF) algorithm which is suitable for applications that are sensitive to deadline.

The weight required for all connections is calculated as follows: j is the number of subscribers, i the required weight for this subscriber, w is a summation of all weights w

10

3) Allocating bandwidth to nrtPS services: The nrtPS

services such as File Transfer Protocol (FTP) is achieved even with long delays, but it needs a satisfactory system throughput. The simplicity at the BS traffic is preferred to QoS; therefore, simple scheme with low complexity such as Weight Round Robin (WRR) algorithm is used to allocate nrtPS application.

4) Allocating bandwidth to BE services: The QoS are not necessarily satisfied for BE applications such as Hypertext Transfer Protocol (HTTP). We use the Round Robin (RR) algorithm which has simplest complexity O(1) to allocate bandwidth to BE applications.

III. SIMULATION MODEL In order to evaluate the proposed bandwidth assignment

algorithm against other classical algorithms, simulations were carried out using QualNET version 5.

A. System Model In the simulation, the uplink of a single-cell WiMAX

OFDM/TDD system of 1.5 km radius is considered. For each run, randomly placed subscriber stations (SSs) with uniform distribution are deployed in the network. The SSs move randomly with a speed of 100km/h as defined by the standard [2]. Table I summarizes the system parameters used in the simulation. We assume that the BS has perfect knowledge of channel state information (CSI).

TABLE 1: THE PARAMETE SETTINGS FOR SIMULATION

Parameters Value System OFDMA/TDD,TDM Frequency band (GHz) 2.4 Channel bandwidth (MHz) 20 Frame duration (ms) 20 FFT size 2048 Cell radius 1.5 km Number of MS 10-50 Number of BS 1 BS transmit power 15 w User distribution Mobility Model Propagation Model

Uniform Random Two Ray

Max user speed 100km/h Beam pattern Omni-directional

B. Traffic Model In order to evaluate the proposed algorithm we evaluate the

real-time traffic type such as VoIP for rtps, videoconference for ertps, File Transfer Protocol (FTP) served in nrtPS class and HTTP evaluated as BE class. Each subscriber is able to send one or more applications to the BS. VoIP traffic is modeled as a two-state Markov ON/OFF source[18]. A videoconference session consists of a VoIP and a video session [18].

IV. PERFORMANCE EVALUATION

• Performance Metrics Since, the performance evaluation parameter of UGS

applications are defined by the standard and BE application do not have any particular QoS requirements, we only focus on the performance evaluation of ertps, rtps and nrtps. For ertps and rtps services, the average packet delay and the average jitter are the main performance metrics, whereas for nrtps service, average throughput is the main consideration. In order to evaluate the bandwidth assignment efficiency for the service classes, we use the total system throughput to evaluate the proposed ADRR algorithm against the conventional algorithms.

V. RESULTS The efficiency of the proposed fair bandwidth assignment

algorithm is compared with well known conventional schemes, e.g., strict priority, WRR, Weighted Fair Queuing (WFQ), Self-Clocked Fair Queuing (SCFQ) and RR. Figure 2 shows the total system throughput of the proposed scheme and five conventional schemes identified above. From the graph we can observe that the throughput fluctuated with increased number of subscribers in the system, particularly when the number of connections in the system is less than thirty seven the overall system throughput in the conventional schemes are fair. After that point, all except one of the conventional schemes experience bandwidth degradation, whereas ADRR experiences increased throughput when the subscriber connections exceed.

745

Figure 3 shows the-end-to-end delay for the proposed ADRR against five conventional schemes. The graphs show sharp increase in end-to-end delay with respect to the subscriber connections when there are more than forty users in the system, while in the proposed algorithm (ADRR) the delay is constant. This is because ADRR algorithm adapts the bandwidth assignment among various service class algorithms in order to improve the overall system throughput while maintaining the QoS requirements.

From Figure 4, we can see that all the algorithms meet the delay jitter constraint when the subscriber smaller than thirty 30, after that point the jitter increases sharply. In contrast, the performance of ADRR can adjust to the increased load in the system and make jitter fairly constant.

Figure 2: Total throughput of the compared scheduling

algorithms

Figure 3: Total End-To-End Delay for the six scheduling algorithms

Figure 4: Average jitter for six scheduling algorithms

Figure 5: Total throughput for the various service classes

based on ADRR

Figure 5 shows the various service classes throughput of proposed ADRR where the overall system throughput in each service class improves with the number of subscribers. The service classes for all applications in the system achieved their required bandwidth by assigning all the available bandwidth dynamically. The ADRR assigns adequate amount of bandwidth for ertPS and rtPS applications to maintain its delay constraint in the second-tier scheduler to avoid packets reaching to the deadline, therefore the performance of the class service algorithm scheduler becomes more effective to serve the relevant applications without sacrificing the QoS requirement, thus extensively enhancing the utilization of bandwidth to improve the overall system capacity.

VI. V. CONCLUSIONS In this paper, a fair bandwidth assignment algorithm

scheduler in two-tier hierarchical QoS for WiMAX systems is proposed to improve the total system throughput while maintaining various QoS requirements for each application. The proposed algorithm called ADRR considers the overall system throughput and QoS requirements as the main performance metric when assigning the requested bandwidth. Simulations have been carried out to evaluate the performance of the proposed algorithm in terms of end-to-end delay, throughput, and delay jitter for UGS,ertPS, rtPS, nrtPS and BE applications. It shows that the overall system throughput of ADRR is optimized in comparison to the conventional algorithms. Our algorithm improves the performance of the various service class algorithms. This can maintain and dynamically assign bandwidth various service class applications.

REFERENCE [1] "Draft IEEE Standard for Local and metropolitan area networks

Corrigendum to IEEE Standard for Local and Metropolitan Area Networks - Part 16: Air Interface for Fixed Broadband Wireless Access Systems (Corrigendum to IEEE Std 802.16-2004)," IEEE Std P802.16/Cor1/D2, 2005.

[2] "IEEE Standard for Local and Metropolitan Area Networks Part 16: Air Interface for Fixed and Mobile Broadband Wireless Access Systems Amendment 2: Physical and Medium Access Control Layers for Combined Fixed and Mobile Operation in Licensed

746

Bands and Corrigendum 1," IEEE Std 802.16e-2005 and IEEE Std 802.16-2004/Cor 1-2005 (Amendment and Corrigendum to IEEE Std 802.16-2004), pp. 0_1-822, 2006.

[3] A. Sayenko, O. Alanen, J. Karhula, and T. Hamalainen, "Ensuring the QoS requirements in 802.16 scheduling," in Proceedings of the 9th ACM international symposium on Modeling analysis and simulation of wireless and mobile systems Terromolinos, Spain: ACM, 2006.

[4] H. Safa, H. Artail, M. Karam, R. Soudah, and S. Khayat, "New Scheduling Architecture for IEEE 802.16 Wireless Metropolitan Area Network," in Computer Systems and Applications, 2007. AICCSA '07. IEEE/ACS International Conference on, 2007, pp. 203-210.

[5] O. Iosif, E. R. Cirstea, I. Banica, and S. Ciochina, "Performance analysis of uplink resource allocation in WIMAX," in Communications (COMM), 2010 8th International Conference on, pp. 351-354.

[6] C. Jianfeng, J. Wenhua, and W. Hongxi, "A service flow management strategy for IEEE 802.16 broadband wireless access systems in TDD mode," in Communications, 2005. ICC 2005. 2005 IEEE International Conference on, 2005, pp. 3422-3426 Vol. 5.

[7] I. C. Msadaa, F. Kamoun, and F. Filali, "An Adaptive QoS Architecture for IEEE 802.16 Broadband Wireless Networks," in Mobile Adhoc and Sensor Systems, 2007. MASS 2007. IEEE Internatonal Conference on, 2007, pp. 1-3.

[8] J. Freitag, N. L. S. da Fonseca, and J. F. de Rezende, "Admission control in IEEE 802.11 networks," in Global Telecommunications Conference Workshops, 2004. GlobeCom Workshops 2004. IEEE, 2004, pp. 258-265.

[9] K. Wongthavarawat and A. Ganz, "IEEE 802.16 based last mile broadband wireless military networks with quality of service support," in Proceedings of the 2003 IEEE conference on Military communications - Volume II Boston, MA: IEEE Computer Society, 2003.

[10] E. Laias and I. Awan, "An interactive QoS framework for fixed WiMAX networks," Simulation Modelling Practice and Theory, vol. 18, pp. 291-303.

[11] I. P. Hsieh, F.-M. Chang, and S.-J. Kao, "A slot-based BS scheduling with maximum latency guarantee and capacity first in 802.16e networks," International Journal of Communication Systems, pp. n/a-n/a.

[12] D.-J. Deng, L.-W. Chang, C.-H. Ke, Y.-M. Huang, and J. M. Chang, "Delay constrained uplink scheduling policy for rtPS-ertPS service in IEEE 802.16e BWA systems," Int. J. Commun. Syst., vol. 22, pp. 119-133, 2009.

[13] L. Yi-Neng, W. Che-Wen, L. Ying-Dar, and L. Yuan-Cheng, "A Latency and Modulation Aware Bandwidth Allocation Algorithm for WiMAX Base Stations," in Wireless Communications and Networking Conference, 2008. WCNC 2008. IEEE, 2008, pp. 1408-1413.

[14] M. Shreedhar and G. Varghese, "Efficient fair queueing using deficit round robin," SIGCOMM Comput. Commun. Rev., vol. 25, pp. 231-242, 1995.

[15] G. D. Castellanos and J. Y. Khan, "Performance of WiMAX packet schedulers for multi-class traffic," in Communications (LATINCOM), 2010 IEEE Latin-American Conference on, pp. 1-6.

[16] C. Ravichandiran, P. Raj, and Vaithiyanathan, "Analysis and Modification of Scheduling Algorithm for IEEE 802.16e (Mobile WiMAX)," in Proceedings of the 2010 Second International Conference on Communication Software and Networks: IEEE Computer Society.

[17] L. Lenzini, E. Mingozzi, and G. Stea, "Bandwidth and latency analysis of modified deficit round robin scheduling algorithms," in Proceedings of the 1st international conference on Performance evaluation methodolgies and tools Pisa, Italy: ACM, 2006.

[18] C. Cicconetti, L. Lenzini, E. Mingozzi, and C. Eklund, "Quality of service support in IEEE 802.16 networks," Network, IEEE, vol. 20, pp. 50-55, 2006.

747