Multiple Downlink Fair Packet Scheduling Scheme in Wi-Max

International Journal of Computer Applications Technology and Research
Volume 2– Issue 4, 389 - 395, 2013
www.ijcat.com 389
Multiple Downlink Fair Packet Scheduling Scheme in
Wi-Max
Pooja Gupta
Lingaya’s University
Faridabad,India
Brijesh Kumar
Faridabad,India
B.L. Raina
Faridabad,India
Abstract: IEEE 802.16 is standardization for a broadband wireless access in network metropolitan area network (MAN). IEEE 802.16
standard (Wi-Max) defines the concrete quality of service (QoS) requirement, a scheduling scheme and efficient packet scheduling
scheme which is necessary to achieve the QoS requirement. In this paper, a novel waiting queue based on downlink bandwidth
allocation architecture from a number of rtps schedule has been proposed to improve the performance of nrtPS services without any
impaction to other services. This paper proposes an efficient QoS scheduling scheme that satisfies both throughput and delay guarantee
to various real and non-real applications corresponding to different scheduling schemes for k=1,2,3,4. Simulation results show that
proposed scheduling scheme can provide a tight QoS guarantee in terms of delay for all types of traffic as defined in WiMax standards.
This process results in maintaining the fairness of allocation and helps to eliminate starvation of lower priority class services. The
authors propose a new efficient and generalized scheduling schemes for IEEE 802.16 broadband wireless access system reflecting the
delay requirements.
Keywords: QoS,IEEE 802.16,WiMax,rtps,nrtps,Fair Scheduling.
1. INTRODUCTION
IEEE 802.16 based wireless access scheme (commonly
known as WiMAX) is considered as one of the most
promising wireless broadband access for communication
networks in metropolitan areas today. Since IEEE802.16
standard defines the concrete quality of service(QoS)
requirement, a scheduling scheme is necessary to achieve the
QoS requisite level. Many Scheduling schemes have earlier
been proposed with the purpose of throughput optimization
and fairness enhancement. However, few scheduling
algorithm support the delay requirement. Here delay refers to
unpredictably longer time for packets to reach the destination
due to unavailability of network resources. IEEE802.16 has
strict delay requirement, the lower the delay the better the
QoS. To achieve the objective of providing a fair and efficient
allocation of bandwidth to all the services without much of the
delay different algorithms have been proposed by taking care
of different parameters. Moreover, we have four service
classes of IEEE802.16 which have different delay
requirements for example for UGS service class cannot
tolerate delay, while for BE service class is delay tolerable.
Sometime the priority of service classes leads to the starvation
problem for lower priority service class. The priority goes
from higher to lower like UGS>rtps>nrtps>BE respectively.
In [1], the authors propose a hybrid of Earliest Due Date
(EDD) and Weighted Fair Queue (WFQ). In EDD, all the
arriving packets get a deadline stamp and are scheduled
according to the increasing order of deadlines. The algorithm
intends to serve the real time traffic first and only if real time
buffer is empty will they consider BE traffic. This will
certainly lead to starvation of lower priority requests. In [3],
the authors consider two types of queues. The first type is
used to schedule data grants for UGS and allocate request
opportunities for rtPS and nrtPS. These grants are scheduled
in a first in first out (FIFO) manner. Once the first queue type
has been served, the scheduler will consider the second type
leading to scarcity. The authors in [4] propose an architecture
consisting of three schedulers. The first scheduler is
concerned with UGS, rtPS and ertPS flows. The second
scheduler is concerned with flows requiring a minimum
bandwidth mainly nrtPS. The third scheduler is used for BE
traffic that comes into picture only when the first two
schedulers have become free. In [5], the authors suggest
downlink bandwidth allocation algorithms based on flow type
and strict priority from highest to lowest - UGS, rtPS, nrtPS
and BE. Here an Opportunistic fair scheduling was used. Here
BE traffic is served whenever an opportunity is available, but
for most of the time BE starves for bandwidth. For
guaranteeing the QoS for different service classes, priority
based schemes could be used in WiMAX scheduler [6]. The
priority order used in real time schedulers is: UGS, ertPS,
rtPS, nrtPS and BE. Here the priority is maintained but the
least priority ones starve for bandwidth. Delay based
algorithm is specifically designed for both real time traffic
and non real time traffic, where the delay tolerance is the
primary QoS parameter. Earliest Due Date (EDD) is the basic
algorithm for scheduler to serve the connection based on the
deadline [7, 12]. However, this algorithm does not guarantee
the throughput for higher priority service viz.UGS. So [13]
proposed a scheduler that combines both EDD and priority for
assuring nrtPS and BE better performance while trying to
satisfy the other classes. Even though there are lots of

Volume 2– Issue 4, 389 - 395, 2013
www.ijcat.com 390
conventional scheduling algorithms they are not meeting all
the required QoS parameters. The performance affecting
parameters like fairness, bandwidth allocation, throughput,
average delay and delay jitter are studied and found out that
none of the algorithms perform effectively for delay
requirement of nrtps service class connections with respect to
other service classes.
2. SYSTEM MODEL
Point to Multi Point(PMP) mode and mesh mode are the
two types of operating modes define for IEEE802.16. In the
PMP mode SSs are geographically scattered around the BS.
The performance of IEEE 802.16 in the PMP mode is verified
in[8][9]. Our system model is based on a time-division-duplex
(TDD) mode. The IEEE 802.16 frame structure is illustrated
in Fig.1[2]. The downlink subframe starts with preamble
followed by frame control header (FCH), downlink map (DL-
MAP), uplink map (UL-MAP) messages and downlink burst
data. The DLMAP message defines the start time, location,
size and encoding type of the downlink burst data which will
be transmitted to the SSs. Since the BS broadcasts the
DLMAP message, every SS located within the service area
decodes the DL-MAP message and searches the DL-MAP
information elements (IEs) indicating the data bursts directed
to that SS in the downlink subframe. After the
transmit/receive transition gap (TTG), the uplink subframe
follows the downlink subframe. IEEE 802.16 provides many
advanced features like adaptive modulation coding (AMC),
frame fragmentation and frame packing. In the current work,
the focus is on the downlink scheduling scheme.
3. MULTI USER SCHEDULER OF THE
MAC LAYER
In this section, a multiuser scheduler is designed at the
medium access control (MAC) layer. Delay requirement is
taken into account in the scheduler design. The AMC, packet
fragmentation and packet packing have not been considered.
In case of the UGS traffic, the required bandwidth is reserved
in advance. Hence, only rtPS, nrtPS and BE connections are
focused in the design.
3.1 Proportional Fair Scheduling
The proportional fair scheduling [10] has shown an
impressive guideline in scheduler design because it
maximizes the total sum of each SS’s utility. In the
proportional fair scheduling, the metric for each connection is
defined as follows
ϕi(t)=DRCi(t)/Ri(t). (1)
where DRCi [12] is the rate requested by the SSi and Ri is
the average rate received by the SSi over a window of the
appropriate size Tc [2][12]. The average rate Ri is updated as
Ri(t+1)=(1-1/Tc)Ri(t)+1/Tc*current transmission rate.
(2)
where Tc is the window size to be used in the moving
average. The proportional fair scheduler selects the
connection that has the highest metric value.
Fig. 1. IEEE 802.16[2] frame structure
3.2 Proposed Fair Scheduling(FS)
In the proportional fair scheduling, the strict fairness is
guaranteed, however the QoS requirement is not reflected. To
the knowledge of authors rtps connections for QoS have been
discussed in the literature with regard to one specified nrtps
connection ,Kim et. al.[2].Pooja gupta et al. [14]have
generalized this concept by associating various parameters
such as scheduling ratio xi of rtps class parameter
associated to k number of nrtps class.Thus, the general
scheduling scheme is being introduced that satisfies the delay
requirement. In this paper we have generated a number of fair
scheduling schemes corresponding to the parameter k so that
the delay requirements are minimized with regard to
corresponding nrtps schemes as mentioned below.
The metric value of the rtPS connections with the delay
requirement should be increased as the queuing delay
increases because the scheduler selects the connection with
the highest metric value with nrtps connections, because nrtps
connections are in the lowest priority. For the above
mentioned conditions the equations for rtps and ,nrtps are
proposed by the authors in paper [2].Here we are generalizing
the above equation by proposing a new scheduling scheme
based on the following metrics for rtPS, nrtPS and BE
connections are given as :
Φrt,i(t)= 1/Rrt,i(t)+C(1+2/π*arctan(|d|)). if qi >0 and d ≥ dmin.
(3)

Volume 2– Issue 4, 389 - 395, 2013
www.ijcat.com 391
=1/Rrt,i(t)+ C. if qi >0 and d< dmin.
=0 if qi =0
Φnrt,i(t)= 1/Rnrt,i(t)+ C. if qi >0 (4)
=0 if qi =0
ΦBE,i(t)= 1/RBE,i(t). if qi >0 (5)
=0 if qi =0
The parameter d is the queuing delay and C means the
intensity of the delay requirement in the rtPS connection. The
parameter dmin is the minimum delay that triggers the service
differentiation between the rtPS connection and nrtPS
connection, and qi means the queue length of the connection i.
Note that Rrt, Rnrt and RBE are updated in the same manner as
in the proportional fair scheduling, that is
Rrt,i(t+1)= (1-1/Tc)R rt,i(t) + r/Tc if connection i is scheduled.
(6)
=(1-1/Tc) R rt,i(t) otherwise
where Tc is the window size to be used in the moving average
and r is the current transmission rate requested by the SS. The
long-term rate is the average sum of the previously scheduled
transmission rates during the time window Tc, where the high
Tc value means that the long-term rate changes slowly
because the average is taken over many previous transmission
rates. The long-term rate of a connection decreases
exponentially before the connection is scheduled, and it
increases when the connection is scheduled. We do not
consider the AMC, so r is a constant. On every frame, the
scheduler selects the connection that has the highest metric
value. Owing to the delay requirement term in the rtPS metric,
rtPS connections are served more frequently than other
connections when the queuing delay increases.
3.3 Determination of Novel Parameters with Analysis
The scheduling ratio x as the average number of scheduling
times for each of rtPS to k nrtPS connections where k<=x has
been defined as given by the following two cases:
If rtPS and nrtPS connections are scheduled equally, the
scheduling ratio x equals k, and if rtPS connection is
scheduled more frequently than nrtPS connection, the
scheduling ratio x becomes greater than k. Now the average
scheduling interval in the rtPS connection is ((x+k)/x) frames
because, on the average, k nrtPS schedule corresponding to x
rtPS connections. As a result of this, the average scheduling
interval in nrtPS connection is (k+x) frames. At the steady
state, the average long-term rates of rtPS and nrtPS
connections at the scheduling instance are as follows:
̅̅̅̅ = ̅̅̅̅(1-(1/Tc))(k+x)/x
+ (r/Tc), at the steady state, we
obtain
̅̅̅̅=(r/Tc)/ (1-(1-(1/Tc))(k+x)/x
(7)
Analogously, Since ̅̅̅̅̅̅ = ̅̅̅̅̅̅(1-(1/Tc))(k+x)
+ (r/Tc) at
the steady state, we obtain
̅̅̅̅̅̅̅̅=(r/Tc)/ (1-(1-(1/Tc))(k+x)
(8)
We consider the same assumption as in [11] that the average
metric value for each connection at the scheduling instance
becomes similar to each other. Hence,
1/̅̅̅̅(1-(1/Tc))(k+x)/x
+
C(1+(2/π)arctan(d)).
≈1/̅̅̅̅̅̅(1-(1/Tc))(k+x)
+ C. (9)
From (7) and (8) , (9) can be written as
(1-(1-(1/Tc))(k+x)/x
)*Tc/ (r*(1-(1/Tc))(k+x)/x
) +
C(1+(2/π)arctan(d)).
≈ (1-(1-(1/Tc))(k+x)
)*Tc/( r*(1-(1/Tc))(k+x)
) + C. (10)
Put (1-1/Tc)=X, L=1+(2/π)arctan(d), therefore from above
equation we have
(1-(X)(k+x)/x
)*Tc/(r*(X(k+x)/x
) + C*L
(1-(X)(k+x)
)*Tc /( r*(X(k+x)
) + C
i.e. C*(L-1)=(Tc/r)*((1-(X)(k+x)
/X(k+x)
– (1-(X)(k+x)/x
)/X(k+x)/x
))
C*(2/π)* =(Tc/r)*((X(k+x)/x
– X(k+x)
)/X((x*x+k*x+k+x)/x)
)
(11)
Now with the help of L and X as defined above and with little
algebra, the set of values of delay represented by d=di
correspond to different sets of values of x,k and C, from
equation (11) we have for d≥0,
d= tan(((π*Tc)/(2*r*C)*[ ((1-1/Tc)(k+x)/x
– (1-1/Tc)(k+x)
)]/(1-
1/Tc)((x*x+k*x+k+x)/x)
] (12)
Now generalizing the above equation if di represents the
various delays for i iterations corresponding to the above
parameters associated to number of rtps,nrtps and intensity
such that d≥0.Thus we have the main result as :
di=tan(((π*Tc)/(2*r*C)*[ ((1-1/Tc)(k+x)/x
– (1-1/Tc)(k+x)
)]/( 1-
1/Tc)((x*x+k*x+k+x)/x)
], However di≥0 (13)
here xi=i,0≤i≤10.However,di, Cj, kt all will take real values
under the investigation as given below:
Now we determined the solution set (di) corresponding to the
various parameters Cj, xi and kt. As the parameter Ci

Volume 2– Issue 4, 389 - 395, 2013
www.ijcat.com 392
increases, the delay di decreases because di and Ci are
inversely proportional to each other. For, k=1, we derive the
delays di corresponding to different values of C=Ci and plot
the graph as given below.When we compare these with the
ones derived by Kim et. al.[2], we notice here that our results
turn out to be almost closer with their results.
4. SIMULATION RESULTS OF FOUR
FAIR SCHEDULING SCHEMES
We now compare the delays of rtPS and nrtPS corresponding
to a number of scheduling schemes corresponding to
k=1,2,3,4 and as such analysis has been done using Matlab for
values of di (delays) corresponding to different prescribed
values of xi, kt, and Cj. ,for 1≤ i ≤10,j=1,2,3,t=1,2,3,4.
CASE I a) x:k=1:1 that is number of rtps connections is equal
to number of nrtps connections.In this case x=5,k=5.
Table 1. Delay requirement of rtps connections for k=1
corresponding to various values of C=Cj , 1≤ j ≤3
We now draw the graph for delay di , 1≤ i ≤10, for various
values of x corresponding to one nrtps scheduling i.e. k=1
and compare these delays with the once obtained by Kim et.
al.[2].The analysis follows as under:
Case I.1
Fig 2. Delay against Scheduling ratio x when k=1 at C=0.1
Case I.2
Fig 3. Delay against Scheduling ratio x when k=1 at
C=0.05
Case I.3
C=0.01
From Fig. 2., Fig. 3. and Fig. 4,it has been observed that our
Eq.(13) coincides with Kim et. al. equation[2]’s equation.
Hence our equations proves to be correct for k=1 at
C=0.1,0.05 and 0.01.However, it has been analysed that as x
increases, the delays corresponding to each of five rtPS, five
nrtPS increase. However, we notice that for first five values
of x, respective delays of nrtPS.>rtPS and then there follows
transition and for the next five values the delays reverse such
that nrtPS<rtPS justifying doing away with the starvation of
nrtps.
Case II:
Delay requirement of rtps connection for k=2,such that k ≤ x.
corresponding to various values of C=Cj ,1≤ j ≤3

Volume 2– Issue 4, 389 - 395, 2013
www.ijcat.com 393
C=0.05
C=0.01
Case III:
C=0.05
C=0.01
Case IV:
C=0.1
C=0.05

Volume 2– Issue 4, 389 - 395, 2013
www.ijcat.com 394
C=0.01
Table 5. Simulation Information
Parameter Value
Packet Size 1500 bytes
Number of nodes 10
Delay requirement 30ms
Table 6. represent the comparison of all the analysis that has
been observed for Kim et. al.[2] value’s at k=1.Our’s at
k=1,2,3,4 at different values of C corresponding to different
values of x where 1≤x≤10.
Table 6. Anaylsis of result of Kim et. al. at k=1,Our’s at
k=1 ,k=2, k=3 and k=4 at different values of C
corresponding to different values of x where 1≤x≤10 .
5. ANALYSIS OF DELAYS OF rtps and
nrtps SERVICE CLASSES
Now in particular, we give the comparison of delay of
different classes with regard to the 10 nodes for the cases C=
0.1,C=.05 and C=0.01 Kim et. al.[2] at k=1 with our’s
Equation(13) and analyze the comparison of the downlink
services within rtPS, nrtPS, as given below:
Fig. 14. Difference of delay d of Kim et. al.[2] with our’s
Equation(13) when k=1 and C=0.1
From Fig. 14, it has been observed that the difference of delay
when k=1 and C=.1 first decreases till sixth value of x and
then become steady. The value of difference of delay is
approx. 0.001 which is quite less. Similarly, from Fig. 15 it
has been observed that the difference of delay when k=1 and
C=0.05 first decreases till sixth value of x then increases and
decreases at 10 but again the value of difference of delay is
quite less. From Fig. 16 it has been observed that the
difference of delay when k=1 and C=.01 is showing the
variation of increase and decrease but that will not not effect
much because of the lesser difference.
6. CONCLUSION
In this paper, an effort has been done to remove the
starvation problem and improvise the proportional fair
scheduling scheme by managing the value of k. The suggested
model has been simulated using Matlab and the results have

Volume 2– Issue 4, 389 - 395, 2013
www.ijcat.com 395
been discussed. To support the QoS requirement the delay
requirement term in the proportional fair scheduling scheme
has been added. The main contribution of this paper is that a
method has been proposed which will generalize the delay
requirement by associating various parameters of xi defined as
(various) rtps connections to the parameter ki associated to the
(various) nrtps connections. The suggested general scheduling
scheme satisfies the delay requirement. One can find the
appropriate parameter C according to the traffic condition of
the networks. After fine tuning of the operating parameter, the
delay requirement can be satisfied without excessive sacrifice
in the nrtps connection performance.
7. REFERENCES
[1]. Guojun D., Dai J., (March 2007), “An Improved Handover
Algorithm for Scheduling Services in IEEE802.16e”,
Published in Mobile WiMAX Symposium,Orlando,FL,
IEEE,pp 38-42.
[2]. Kim T., Lim J.T.,”Quality of service supporting downlink
scheduling scheme in worldwide interoperability for
microwave access wireless access systems”, Published in
IET.Communications, Vol. 4, Iss. 1, pp. 32–38,2010.
[3]. Howon L. , Taesoo K. and Dong-Ho Cho, (Sept. 2004),
“An efficient uplink scheduling algorithm for VoIP
services in IEEE 802.16 BWA systems”, 60th IEEE
Vehicular Technology Conference, vol.5, pp. 3070 –
3074.
[4]. Kun y., Jie z. and Hsiao-hwa c., (March-April2007), “A
Flexible QoS aware service Gateway for heterogeneous
wireless networks”,Published in IEEE NETWORK, vol.
21,issue 2,pp 6-12 .
[5]. Mehrjoo M.,Dianati M., Shen X. and Naik K.,(2006),
“Opportunistic Fair Scheduling for the Downlink of IEEE
802.16Wireless Metropolitan Area Networks”, Published
in Proceeding Qshine’06 proceedings of the 3rd
International Conference on QoS in heterogenous Wired
/Wireless Networks, vol.191.
[6]. Wang Y., et al., (May 2008), “Priority-Based Fair
scheduling for Multimedia WiMAX Uplink Traffic”, Proc.
of IEEE International Conference on Communication,
China.
[7]. Perumalraja R., et al., (2006), “Multimedia Supported
Uplink Scheduling for IEEE 8021.6d OFDMA Networks”,
Proceedings of Annual India Conference.
[8]. Kumar S.,”Hybrid scheduling For QoS in
WiMAX”,Master Thesis , Thapar University,Patiala,June
2009.
[9]. “IEEE Standard for Local and metropolitan area
networks”, Part 16: Air Interface for Broadband Wireless
Access Systems.
[10]. CDMA-HDR a high efficiency-high data rate
personal communication wireless system’. Proc. IEEE
Vehicle Technology Conf., Tokyo, May 2000, vol. 3,pp.
1854–1858.
[11]. HOU F., HO P., SHEN X., CHEN A. “A novel QoS
scheduling scheme in IEEE 802.16 networks”. Proc. IEEE
WCNC, Hong Kong, March 2007, pp. 2457–2462.
[12]. Wongthavarawat K., and Ganz A., (2003), “Packet
Scheduling for QoS Support in IEEE 802.16 Broadband
Wireless Access Systems”, International Conference. on
Journal of Commn. Systems., pp.81-86.
[13]. Perumalraja R., et al., (Sept. 2010), “QoS Assured
Uplink Scheduler for WiMAX Networks”, Published in
Vehicular Technology Conference Fall,IEEE 72nd
,pp-1-5.
[14]. Gupta P., Kumar B., Raina B.L.,”QoS Support
downlink for WiMAX Network”, Published in QSHINE
2013,LNICST(Lecture Notes of the Institute for Computer
Sciences, Social Informatics and Telecommunications
Engineering) Springer 115 proceeding ,pp- 989-997.

Multiple Downlink Fair Packet Scheduling Scheme in Wi-Max

More Related Content

What's hot (15)

Viewers also liked (15)

Similar to Multiple Downlink Fair Packet Scheduling Scheme in Wi-Max (20)

More from Editor IJCATR (20)

Recently uploaded (20)

Multiple Downlink Fair Packet Scheduling Scheme in Wi-Max