A New Queue Length based Scheduling Strategy for nrtPS Service

Class in IEEE 802.16 Networks

Pallavi Grover

and Marcio Andrey Teixeira

M.E Research Scholar, Comp Sc. & Engg., UIET Panjab University, Chandigarh, India

Department of Computing, Federal Institute of Education, Science and Technology of São Paulo, Catanduva, Brazil

Keywords: IEEE 802.16 Networks, nrtPS, QoS, Scheduling Algorithms.

Abstract: IEEE 802.16 standard is designed to provide services to various types of multimedia applications. It supports

real-time and non-real-time service classes. With such a large volume of traffic, a new strategy needed to be

developed for the non-real-time service class, since there have been limited studies in this area. In this paper,

a new queue length based scheduling strategy for the non-real-time service class is proposed. The proposed

algorithm is developed on the basis of virtual queue’s and counter scheme’s, aiming at ensuring minimum

bandwidth for non-real-time applications.

1 INTRODUCTION

In recent years, the development of wireless

technology has increased rapidly, particularly in the

field of broadband wireless networks. WiMAX

(IEEE, 2013) is intended to offer low-cost, high-

speed internet access to a wide variety of devices. The

advantages of WiMAX networks include high

transmission speed, scalable bandwidth, link layer

retransmission, robust security, high peak data rates,

and an efficient QoS mechanism for data, voice, and

video. It is very challenging task to assure QoS

requirements of every type of traffic in a wireless

network. The challenges associated with WiMAX

(IEEE, 2013) are due to limitations of wireless

networks such as strong attenuation with increasing

distance, Rayleigh fading, limited scalability etc. The

protocol stack of IEEE 802.16 standard which defines

characteristics of physical layer’s as well as media

access control (MAC) layer’s.

The working architecture of the IEEE 802.16

network is composed of 2 components: Base Station

(BS) and Subscriber Station (SS). The BS is a core

element that acts as an interface between the

infrastructure network such as the Internet Service

Provider (ISP) and the SS. The SS further extends

internet services to its users. The SSs connects with

the BS to provide a variety of services to its users.

WiMAX has defined five service classes: Unsolicited

Grant Service (UGS), real-time Polling Service

(rtPS), non-real-time Polling Service (nrtPS), best-

effort service (BE) and an extended real-time Polling

Service (ertPS). Each of them is associated with a

certain set of QoS parameters such as delay,

throughput, and jitter.

Real-time services such as UGS, ertPS, and rtPS

are always given higher priority by the scheduler,

whereas non-real-time traffic, which amounts to the

majority of internet traffic, is always neglected. With

such large amounts of traffic volume, there is a need

to develop a dedicated strategy for the non-real-time

service class. There have been very limited studies in

this regard.

The IEEE 802.16 standard does not specify any

scheduling algorithm, it is left to the vendors as to

whether or not they will employ their own scheduling

algorithm. Scheduler defines the distribution of

allocated resources in the form of slots, which are

further mapped into sub channels. Logically, it

calculates the number of slots for each class and

physically, it selects the sub channels and time

intervals. There are 3 distinct scheduling processes:

two are for the BS (uplink and downlink), and one is

for the SS (uplink). The proposed algorithm is

developed to act in the uplink scheduling at BS,

because the uplink scheduling is more complex than

the downlink scheduling. In the downlink scheduler,

the BS has complete knowledge of the queue status

whereas in the uplink scheduling, the input queues are

located in the SSs and hence, are separated from the

BS. The BS does not have any information about the

Grover, P. and Teixeira, M.

A New Queue Length based Scheduling Strategy for nrtPS Service Class in IEEE 802.16 Networks.

DOI: 10.5220/0005989601650171

In Proceedings of the 13th International Joint Conference on e-Business and Telecommunications (ICETE 2016) - Volume 6: WINSYS, pages 165-171

ISBN: 978-989-758-196-0

165

arrival time of packets in the SS queues, however, the

BS makes the bandwidth allocation based on the

bandwidth requests (BW-REQ), messages sent by the

SSs. The BW-REQ message indicates the queue

status in the SS.

This paper presents a survey of scheduling

algorithm to the nrtPS connections and proposes a

new scheduling algorithm for nrtPS service class

based on virtual queue and counter scheme. The

proposed algorithm is applied directly to the uplink

virtual queue in the BS aiming at ensuring minimum

bandwidth for nrtPS service class. The scheduling is

done by assigning priority to the connection having

larger queue size. Counter is attached to each virtual

queue to prevent starvation of connections having

lower queue size. The proposed algorithm has been

evaluated by means of modeling and simulation. The

simulations experiments have shown satisfactory

results.

The remainder of this paper is organized as

follows: Section II resumes the related work. Section

III describes the proposed work. Section IV defines

the network scenario and the main parameters used in

the simulation. Section V shows the numerical

results. Finally, Section VI concludes the paper.

2 RELATED WORK

The literature survey of non-real-time services can be

traced back to conventional algorithms such as

proportional fairness and modified largest weighted

delay first algorithms. Proportional fairness was

originally designed for downlink traffic to increase

the throughput of the system, as well as to provide

fairness among multiple queues. It calculates a

priority function which is the ratio of the current rate,

to the average rate, and then schedules different

queues accordingly. It is very simple and efficient,

but it does not take into consideration saturated

queues, when dealing with non-real-time traffic.

Qingwen Liu et al. (Liu, 2006) proposed a cross

layer scheduling algorithm providing QoS support in

IEEE 802.16 networks. A priority is assigned to each

connection which is updated dynamically on basis of

wireless channel quality and QoS satisfaction. For

each nrtPS connection, a minimum reserved rate (ŋ)

is defined. The proposed algorithm ensures average

transmission rate should be greater than ŋ. Scheduling

of nrtPS connections is done with the help of priority

function, which is dependent on nrtPS-class

coefficient and rate satisfaction indicators. The rate

satisfaction indicator is the ratio of the average

transmission rate over the minimum reserved rate. If

the value of the indicator is greater than 1, then the

requirement is satisfied; otherwise, packets should be

sent as soon as possible. This scheduler offers

flexibility, scalability, and low implementation

complexity. The major drawbacks include fairness

issues among same service classes and imperfect

channel conditions arising, due to errors in estimation

and feedback latency.

Authors Fen Hou et al. (Hou, 2009) presented a

simple scheduling structure for non-real-time

services in the IEEE 802.16 networks. It is a cross

layer algorithm, which considers selective automatic

repeat request mechanisms at MAC which layers and

uses adaptive modulation and coding schemes at the

physical layer. It tries to ensure minimum throughput

requirements of the nrtPS class, and at the same time,

maintains flexibility between resource allocation and

packet scheduling. To achieve this flexibility, two

parameters, m and n are defined, where m represents

the number of SSs selected in each MAC frame, and

n represents bandwidth, which is granted to SS when

it is being serviced. In the beginning of each frame m,

SSs, which have superior channel conditions, are

selected and n, amounts to the number of resources

that are given to them. The value of m=1 assures

opportunistic scheduling, leading to maximum

resource utilization. When m = the total number of

SSs, then it leads to minimum resource utilization

with lesser delivery delays. This paper focuses on

scheduling unicast nrtPS applications, but there is a

dire need to concentrate on multicast multimedia

applications.

Ali Mohammed Alsahag et al. (Alsahag, 2014)

proposed a fuzzy based adaptive deficit round robin

uplink scheduler that adjusts the weights of the

service queues for real-time as well as non-real-time

applications. The allocation of bandwidth is done on

the basis of deadline based schemes. To compute the

deadline authors have used maximum latency for

real-time and throughput for non-real-time as input

variables. The overall mechanism can be divided in

three fundamental phases: fuzzification, fuzzy

inference, and defuzzification. In the fuzzification

process, we use two input variables as real-time

maximum latency (RTML) and non-real-time

throughput (NRTTHR). These input variables are

processed with the help of a rule base in the fuzzy

inference phase, and then finally in the

defuzzification phase, where crisp numerical values

are obtained, which determines the weights that need

to be used, as an indication for priority. In the

bandwidth assignment process, several queues are

maintained which are associated with a DC value. In

each round DC is incremented by a value, which is

WINSYS 2016 - International Conference on Wireless Networks and Mobile Systems

166

determined by the fuzzy system, keeping in

consideration overall capacity of the system.

Transmission of queue takes place when the DC value

is equal to the amount of requested bandwidth. The

proposed algorithm optimizes the overall systems

utilization, but gives more consideration to satisfying

maximum latency requirements of real-time traffic.

D. David Neels Pon Kumar et al. (Kumar, 2011)

proposed a neural network based on fuzzy priority

scheduling algorithms. Fuzzy is used to calculate

priority, which itself is composed of a primary fuzzy

scheduler and a dynamic fuzzy scheduler. The

primary fuzzy scheduler takes an input Expiry time

(E), Waiting time (W), Queue length (Q), Packet size

(P) and gives priority index as output. After this

priority index is feed as an input to the dynamic fuzzy

scheduler, which calculates the final priority, by

taking certain types of services into consideration.

Scheduling is done using an artificial neural network,

based on prioritized input received from DPFS

(dynamic priority fuzzy scheduler).Artificial neural

network (ANN) consists of three layers: input layer,

modified Kohonen layer and the Grossberg layer. The

input layer processes prioritized output received from

DPFS and arranges them in order of priority. After,

the output is fed to modify the kohonen layer, which

checks whether the value is in the range of the

threshold. If it is given as an input to the grossberg

layer it has less chance to be rejected. This algorithm

improves fairness, and prevents starvation of the low

priority traffic, but it has not considered bursty traffic

conditions. The major drawback is that processing

time has increased.

Dusit Niyato et al. (Niyato, 2007) presented a

survey of the game theory techniques for

management of radio resources in different wireless

networks and also proposed bandwidth allocation and

connection admission control schemes for IEEE

802.16 networks. The bandwidth allocation helps the

non-cooperative game in which three parameters are

considered: players, strategies, and pay off’s. Players

refer to rtPS and nrtPS connection, strategies which

imply the amount of bandwidth that is to be allocated

with a new connection and payoff’s that refer to the

total utility of currently running rtPS and nrtPS

connections, plus the utility of new connections. The

solution of this game is provided with the help of nash

equilibrium which is calculated through best response

function. It maximizes the payoff of BS which defines

QoS requirements of connections. The simulation

results of adaptive scheme shows that it is not able to

satisfy delay and throughput necessities when load is

high.

3 PROPOSED SCHEDULING

ALGORITHM

The proposed scheduling algorithm aims at

guaranteeing the efficient utilization of the bandwidth

resources, and thus, promotes the effective use of the

wireless link. Figure 1 shows the architecture used to

develop the proposed solution.

Figure 1: Architecture used in development of the proposed

solution.

As shown in Figure 1, the BS receives from the

SS bandwidth request (BW-REQ) message, which

reports the current queue size of each connection. The

algorithm is applied directly to the nrtPS bandwidth

request which queues in the BS (Teixeira, 2012).

Figure 2 shows the pseud-code of the proposed

scheduling algorithm.

The scheduling process of the nrtPS is explained

below, in the following steps:

1) Initially, the algorithm checks the amount of

bandwidth requested by the nrtPS connections, and

stores them in a virtual queue at the BS (lines 1 - 7).

2) It then verifies the actual amount of bandwidth

requested by the SSs and sorts the bandwidth requests

by the largest queue size (lines 8 - 15).

3) After sorting the bandwidth requests, a

counter is assigned to each virtual queue, which

prevents the starvation of connections, which

ultimately prevents having lower bandwidth requests

(line 16).

4) On completion of the first round of allocation,

the proposed algorithm verifies if the bandwidth

requests are satisfied or not. If it is not satisfied, the

algorithm checks if there are more symbols to allocate

(lines 18 - 25). However, in this case, the scheduling

sorts the connections by the highest counter number.

5) If more symbols are available, then the

proposed algorithm allocates these available symbols

A New Queue Length based Scheduling Strategy for nrtPS Service Class in IEEE 802.16 Networks

167

to nrtPS connections, and its counter value will be

decreased (line 29). When the counter decreases to

zero, the count is initiated (lines 26-27).

__________________________________________

1: Verifies the bandwidth messages request at the BS queue;

2: begin

3: for BWrequest i at the BS queue do

4: begin

5: if (BWrequest[i] = nrtPS) then

6: begin

7: BW[i] += BWrequest[i].length;

8: if (connections_numbers > 1) then

9: begin

10: if ( BWrequest[i].length < BWrequest[i+1].length)

then

11: begin

12: Temp = BWrequest[i].length;

13: BWrequest[i].length = BWrequest[i+1].length;

14: BWrequest[i+1].length = Temp;

15: end;

16: BWrequest[i].counter = 3;

17: UL-MAP = request order by the length;

18: if (BwToAllocate > 0) then

19: begin

20: if ( BWrequest[i].counter <

BWrequest[i+1].counter) then

21: begin

22: Temp = BWrequest[i].counter;

23: BWrequest[i].counter = BWrequest[i+1].counter;

24: BWrequest[i+1].counter = Temp;

25: end;

26: if (BWrequest[i

].counter == 0) then

27: BWrequest[i].counter =3

28: else

29: BWrequest[i].counter --;

30: UL-MAP = request order by the counter;

31: end;

32: end;

33: end;

34: end;

35:end;

_____________________________________________________

Figure 2: Proposed Scheduling Algorithm.

Figure 3 shows the Flowchart of the proposed

scheduling algorithm.

4 MODELING AND SIMULATION

The simulation studies and evaluates the properties of

the proposed scheduling algorithm for non-real-time

traffic. It has been implemented in the Network

Simulator 2 (NS-2) version 2.34 (Network, 2016)

along with the WiMAX module. The simulation

scenario consists of a BS, and also of several SSs

distributed around the BS in a random mode. The

Table 1 shows the main parameters used in the

simulation.

Determine all nrtPS connections

The SS has a

bandwidth

request ?

YES

Store the BW-REQ

length

Arrange the connections

by the queue length

Are there

more symbols

to allocate?

Is there another

nrtPS connection?

YES

Arrange the connections

by the lowest

counter(ini tially counter=3)

YES

UL-MAP =Process

requests by lowest counter

For all other

SSs

counter == 0 ?

YES

Counter --

Counter = 3

For all SSs in BS range

Set Counter =3 for those

which are scheduled

Figure 3: Flowchart of the proposed scheduling algorithm.

Table 1: Main parameters used in the simulation.

Parameters Values

Frequency band (MHz) 5

Transmit antenna gain 1

Received antenna gain 1

Frame duration (ms) 20

Cyclic prefix 0.25

Simulation time (s) 100

Downlink bandwidth (Mbps) 5.4

Uplink bandwidth (Mbps) 10.5

WINSYS 2016 - International Conference on Wireless Networks and Mobile Systems

168

The sources of traffic used in the simulation were,

voice, video, web and a file transfer, which

corresponds to the UGS, rtPS, BE, and nrtPS class.

The voice traffic is modeled as an on/off scheme with

a mean of 1.2 s and 1.8 s for “on” and “off” periods

respectively. During “on” period packets, 66 bytes are

generated every 20 ms, which follows an exponential

distribution. The voice traffic was modeled on a basis

of an exponential traffic model in NS-2.

The video was modeled using a traffic source that

generates packets periodically with variable sizes,

simulating the MPEG traffic. The packets size vary

between 450 to 1500 bytes. The web traffic was

modeled using hybrid Lognormal/Pareto distribution.

FTP traffic was generated from a source which follows

exponential distribution within a mean of 512 Kbytes.

The performance of the proposed algorithm was

evaluated to take into account the follow performance

metrics:

Average end-to-end-delay: This is the time taken

by packets to transmit from source to destination.

There can be various reasons for the delay such as

propagation and network delay, source and

destination processing delay etc. It can be calculated

as followed:

()

∑

−

sConnectionofNumber

timestartPackettimearrivalPacket

DelayAverage

(1)

Throughput: This measures the data rate which

is generated by the application in terms of bits per

second. It can be calculated as followed:

()

∑

−

timestartPackettimearrivalPacket

SizePacket

Throughput

(2)

The nrtPS scheduler should guarantee minimal

bandwidth to the users. It is used with the Throughput

metric. However, the average delay performance

metric is used to analyze whether the proposed

scheduler algorithm is influencing the other service

classes (UGS, rtPS BE).

5 NUMERICAL RESULTS

5.1 Experiment 1

The first experiment analyzes the performance of the

nrtPS connections and verifies how the BS allocates

bandwidth to the nrtPS service class. The simulated

network consists of one BS and a number of nrtPS

SSs, varying between 2 and 12. The transmission rate

of each nrtPS connection is 512 kbps. There are 2

UGS, 4 rtPS, and 8 BE connections as background

traffic, along with nrtPS connections. Each UGS, rtPS

and BE connection generates 40, 320 and 510 kbps

respectively. The Figure 4 shows the behavior of

average delays of nrtPS SSs.

Figure 4: Average delay of the UGS, rtPS, nrtPS and BE

connections.

It is possible to see in Figure 4 that the average

delay rises with an increase in the number of nrtPS

connections. The increase in the average delay has

occurred due to link saturation and the total traffic of

the nrtPS connections has increased. The proposed

algorithm prioritizes transmission of nrtPS

connections on basis of queue length, which results in

lesser average delay. The average delay of the UGS

and rtPS classes remains constant, as defined by the

standard. The increased of the nrtPS load traffic does

not interfere in the UGS and nrtPS classes. This is

expected because the UGS and rtPS Traffic have high

priority over nrtPS and BE service classes. The

average delay of the BE is high, however, the BE

service class does not have QoS parameters.

Figure 5 shows the throughput of the UGS, rtPS,

nrtPS and BE connections.

Figure 5: Throughput of the UGS, rtPS, nrtPS and BE

connections.

A New Queue Length based Scheduling Strategy for nrtPS Service Class in IEEE 802.16 Networks

169

Figure 5 shows the increase of the nrtPS traffic

load also does not interfere with the UGS and rtPS

traffic. The BS distributed the resources to all service

classes. However, with the increase of the nrtPS

traffic load, the throughput of the BE connections

decrease. The BE connections have lower priority

among the other service classes, which causes them

to receive less resource (slots).

5.2 Experiment 2

The second experiment compares the performance of

the proposed algorithm with the traditional Round

Robin (RR) scheduling algorithm. This simulated

network consists of one BS a number of nrtPS

connections varying from 2 to 12. There are 2 UGS,

2 rtPS and 4 BE connections as background traffic

along with nrtPS connections. Figure 6 shows the

throughput of the RR and the proposed algorithm.

As we can see in Figure 6, the proposed algorithm

has better performance than the RR scheduling

algorithm. This happens because the proposed

algorithm organizes the uplink frame in accordance

with the queue length. Moreover, the algorithm

verifies the counter of the nrtPS connections to avoid

starvation. The RR scheduler makes the scheduling

and does not consider the queue length.

Figure 6: Throughput of the RR and nrtPS connections.

5.3 Experiment 3

The third experiment verifies the performance of the

nrtPS connections in the presence of high and low

priority traffic. The input traffic has increased by a

ratio of 2:2:2:2 i.e. 2 UGS, 2 rtPS, 2 nrtPS and with 2

BE connections. In such a scenario, performance

metrics are analyzed by increasing traffic inputs up to

10 connections of each service class. This experiment

helps in analyzing the performance of the scheduler

in presence of diverse traffic classes. Figure 7 shows

the throughput of the connections.

Figure 7: Throughput of the UGS, rtPS, nrtPS and BE

connections.

It is possible to see in Figure 7 that the throughput

of the nrtPS connections decreases with the increase

of the traffic load in the system. However, the

scheduler distributes the bandwidth to guarantee

minimum bandwidth to the nrtPS connections. This

decrease in throughput can be justified due to the

presence of high priority traffic, along with nrtPS

connections. Increasing demand for these classes

forces the system to allocate more amounts of

bandwidth to them instead of the nrtPS connections.

However, the proposed algorithm allocates

bandwidth efficiently to the nrtPS connections in

presence of UGS and rtPS connections, which are

highest in priority. When the number of connections

increases, the scheduler distributes the resources to

connections with high priority and the BE

connections receives fewer resources once the BE

connections do not have any QoS requirements.

The same experiment has been completed, but this

time, with the RR algorithm, in order to compare the

performance between RR and the proposed

algorithm. The Figure 8 depicts the throughput of the

proposed algorithm and the RR.

Figure 8: Throughput of the RR and nrtPS connections.

WINSYS 2016 - International Conference on Wireless Networks and Mobile Systems

170

As is shown in Figure 8, the proposed algorithm

allocates the bandwidth more efficiently than the RR

algorithm. The resources were distributed more

efficiently when the scheduler was made using the

information about the queue length. Furthermore, the

counter scheme helps the scheduler to avoid the

starvation to the nrtPS connections that have lower

queue length.

6 CONCLUSIONS

In this paper, a queue length based scheduling

strategy for an nrtPS service was proposed. The

proposed algorithm was verified by performing

different experiments with diverse traffic scenarios.

The performance metrics used to evaluate this

proposed scheduling algorithm were average delay

and throughput. Scheduler is expected to allocate

bandwidth effectively to the nrtPS class. It should be

able to satisfy the constraint of minimum throughput

for the nrtPS class, even in the presence of high

priority traffic classes such as UGS and rtPS. It has

been observed that bandwidth allocation to nrtPS

class is done efficiently because it leads to increased

quality and more user satisfaction. The proposed

algorithm also shows better performance than the RR

scheduling algorithm.

REFERENCES

IEEE Std. 802.16n 2013. IEEE standard for local and

metropolitan area networks Part 16: Air interface for

fixed and mobile broadband wireless access systems––

amendment 2: Higher Reliability Networks. IEEE

Computer Society and IEEE Microwave Theory and

Techniques Society. p. 1–132. ISBN 978-0-7381-8425-

8, 2013.

Liu Q, Wang X, Giannakis G.B., 2006. A Cross-Layer

Scheduling Algorithm With QoS Support in Wireless

Networks, IEEE Transactions on vehicular technology;

55(3), 839 – 847.

Hou F, Ho P.H.,2009. A Flexible Resource Allocation and

Scheduling Framework for Non-real-time Polling

Service in IEEE 802.16 Networks, IEEE transactions

on Wireless Communications; 8(2): 766 – 775.

Alsahag A.M, Ali B.M, Noordin N.K, Mohamad H., 2014.

Fair uplink bandwidth allocation and latency guarantee

for mobile WiMAX using fuzzy adaptive deficit round

robin, Journal of network and computer

applications;39:17-25.

Kumar D.D.N.P, Murugesan K, Raghavan S, Suganthi M.,

2011. Neural Network based Scheduling Algorithm for

WiMAX with improved QoS Constraints In:

Proceedings of international conference on Emerging

trends in electrical and computer technology, 23-24

March, Tamil Nadu. p. 1076 – 1081.

Niyato D, Hossain E., 2007. Radio resource management

games in wireless networks: an approach to bandwidth

allocation and admission control for polling service in

IEEE 802.16, IEEE Journal of Wireless

Communications;14(1):27–35, Feb.

Teixeira, M. A.; Guardieiro, P. R., 2012. Adaptive packet

scheduling for the uplink traffic in IEEE 802.16e

networks. International Journal of Communication

Systems (Print).

The Network Simulator-ns2.

A New Queue Length based Scheduling Strategy for nrtPS Service Class in IEEE 802.16 Networks

171