Int. J. Communications, Network and System Sciences, 2009, 2, 888-894
doi:10.4236/ijcns.2009.29103 Published Online December 2009 (http://www.SciRP.org/journal/ijcns/).
Optimizing WiMAX: A Dynamic Strategy for Reallocation
of Underutilized Downlink Sub-Frame to
Uplink in TDD Mode
Abdul Qadir ANSARI1, Abdul Qadeer K. RAJPUT2, Adnan Ashraf ARAIN2, Manzoor HASHMANI2
1Wireless Core Network, Pakistan Telecommunication Company Limited, Pakistan
2CREST-Research Group, IICT, Mehran University of Engineering and Technology, Jamshoro, Pakistan
Received August 20, 2009; revised September 22, 2009; accepted November 1, 2009
Abstract
WiMAX networks experience sporadic congestion on uplink when applications running at subscriber sta-
tions need more bandwidth to transmit than allocated. With the fast proliferation of mobile Internet, the
wireless community has been looking for a framework that can address the issue of impediment on uplink.
Due to asymmetric behavior of Internet applications downlink sub-frame is expected to have longer duration
as compared to uplink. According to IEEE 806.16 standard for WiMAX the segmentation of TDD frame
between uplink and downlink can be dynamically redefined even at runtime. Research contributions so far
lack in addressing an optimal strategy for readjustment of uplink and downlink sub-frame boundaries; based
on traffic statistics. In this paper, we introduce a mechanism that allows uplink sub-frame to grow, borrowing
resources from the downlink sub-frame, if the uplink utilization is high and the downlink is being underuti-
lized. We present here, a framework to dynamically demarcate the TDD frame-duration between uplink and
downlink. Proposed algorithm takes into account the present utilization of downlink and reallocates a certain
quantum of free resources to uplink. This occurs when uplink observes shortage of bandwidth to transmit.
We simulate some test scenarios using OPNET Modeler with and without dynamic reallocation capability.
The results of our simulation confirm the effectiveness of proposed algorithm which observes a remarkable
decrease in end-to-end packet delay. Also, we observe an improvement in throughput at uplink such that, the
performance of downlink remains unaffected.
Keywords: WiMAX, Time Division Duplex (TDD), Quality of Service (QoS), End-to-End Packet Delay,
Network Throughput
1. Introduction
The IEEE 802.16 family of standards specifies the air
interface of fixed and mobile broadband wireless access
(BWA) systems that support multimedia services. The
IEEE 802.16-2004 standard, which was previously called
802.16d or 802.16-REVd, was published for fixed access
in October 2004. Good reviews of the standard can be
found in [1–3]. The standard has been updated and ex-
tended to the 802.16e standard for mobile access, Mobile
WiMAX, as of October 2005 [4]. Mobile WiMAX is a
broadband wireless solution that enables convergence of
mobile and fixed broadband networks. Mobile WiMAX
technology is designed to be able to scale to work in dif-
ferent channelizations from 1.25 to 20 MHz to comply
with varied requirements.
The fundamental premise of the IEEE 802.16e MAC
architecture is QoS on the move. With fast air interface,
asymmetric downlink/uplink configuration capability,
fine resource granularity and a flexible resource alloca-
tion mechanism, Mobile WiMAX can meet QoS require-
ments for a wide range of data services and applications
[5].
IEEE 802.16e standard includes QoS support frame-
work; however, it left undefined the details to ensure
QoS guarantees, scheduling algorithms, uplink (UL) and
downlink (DL) sub-frame allocation; for vendors as a
motivation to device effective scheduling and resource
allocation mechanisms to deliver QoS guarantees, espe-
cially for the real-time traffic.

A. Q. ANSARI ET AL. 889
WiMAX networks support two types of duplexing
modes to separate UL and DL communication; i.e. Time
Division Duplex (TDD) and Frequency Division Duplex
(FDD). In this paper we have focused on TDD mode
where both UL and DL share same frequency and to
separate downlink and uplink, time division multiple
access (TDMA) is used.
The duration of DL and UL sub-frames may be de-
cided once based on average traffic statistic expectations.
However, it is also possible to tune the network configu-
ration through real-time monitoring, and may readjust
the uplink and downlink boundaries.
According to the standard, this segmentation can be
dynamically adjusted even at runtime. Unfortunately, re-
search contributions so far lacks in addressing an optimal
strategy towards readjustment of UL and DL boundaries
dynamically; while keeping the current traffic statistics
in account. It is important to remember that asymmetric
behavior of Internet applications intuitively ask for more
duration of DL sub-frame as compared to UL. Also DL
traffic behaviors could be controlled at serving Base Sta-
tion (BS); but it is not true for UL.
We have introduced a mechanism in order to allow the
UL sub-frame to “grow”, borrowing resources from the
DL sub-frame, if the UL utilization is high and the DL
utilization is low. Our strategy is to keep the perform-
ance graph of downlink traffic unaffected by monitoring
the downlink utilization and requirement. Moreover re-
sources borrowed from DL will be relinquished as and
when required at DL. The mechanism is tested in a con-
trolled environment for it effectiveness. Simulation re-
sults confirmed the positive impact of this new capability
on throughput and packet end-to-end delay on UL.
2. WiMAX-Time Division Duplex
The IEEE 802.16e-2005 supports both time division du-
plexing (TDD) and frequency division duplexing (FDD)
modes. In TDD mode, the uplink and downlink trans-
mission share the same frequency but do not transmit
simultaneously. The frame, in Figure 1, is flexibly di-
vided into a downlink sub-frame and an uplink sub-
frame. The downlink sub-frame used to transmit data
from a BS to SS. The uplink sub-frame carries SS traffic
to the BS. The sub-frame is divided into mini slots,
which is the minimum unit of data transfer in this level.
Frames are broadcasted and during the downlink sub-
frame, the SS picks up the data addressed to it. Media
Access Protocol (MAP) messages are broadcasted at the
beginning of each downlink sub-frame. There are two
types of MAP messages, DL-MAP and UL-MAP. The
DL-MAP described the usage of the downlink sub-frame
whereas the UP-MAP tells which mini slot and how many
mini slots are allocated to the specified SS for its trans-
mission during the uplink sub-frame.
Mobile WiMAX profiles only consider the TDD mode
of operation for the following reasons:
1) It allows dynamic reallocation of DL and UL radio
resources to effectively support asymmetric traffic pat-
tern that is common in Internet applications.
2) The allocation of radio resources in DL and UL is
determined by the DL/UL switching point(s).
3) Both DL and UL are in the same frequency channel
to yield better channel reciprocity and to better support
link adaptation.
4) A single frequency channel in DL and UL can pro-
vide more flexibility for spectrum allocation.
As shown in Figure 2, the connection between a SS
and BS is identified by a unique connection identifier
(CID). One CID can correspond to an individual applica-
tion or a number of applicants bundled together such as a
group of users in the same building. CID also specifies
polling schemes provided to the connection by the BS,
which will result in QoS for the connection who owns
this CID.
Figure 1. WIMAX TDD frame.
Figure 2. Connection (CID)-based WiMAX MAC layer [6].

A. Q. ANSARI ET AL. 890
3. Quality of Service Support in WiMAX
WiMAX with provisioned quality of service (QoS) for
digital multimedia applications to mobile end users over
wide area networks is the new frontier of telecommuni-
cations industry.
Before providing a certain type of a service, the base
station and user-terminal first establish a unidirectional
logical link between the peer MACs called a connection.
The outbound MAC then associates packets traversing
the MAC interface into a service flow to be delivered
over the connection. The QoS parameters associated with
the service flow define the transmission ordering and
scheduling on the air interface. The connection-oriented
QoS therefore, can provide accurate control over the air
interface. Since the air interface is usually the bottleneck,
the connection-oriented QoS can effectively enable the
end-to-end QoS control. The service flow based QoS
mechanism applies to both DL and UL to provide im-
proved QoS in both directions.
The service flow parameters can be dynamically
managed through MAC messages to accommodate the
dynamic service demand. IEEE 802.16 MAC is connec-
tion-oriented. BS controls the access to the medium,
bandwidth is granted to SS on demand. At the beginning
of each frame, the BS schedules the uplink and downlink
grants to meet the negotiated QoS requirements. Each SS
learns the boundaries of its allocation under current up-
link sub-frame via the UL-MAP message. The DL-MAP
delivers the timetable of downlink grants in the downlink
sub-frame [7].
4. Related Research Review
During last couple of years, many proposals for QoS
service support in WiMAX networks were published
[8–14]. Most of them are the solutions on the bandwidth
allocation (with preset allocation of UL and DL sub-
frame size [13]), flow scheduling and Adaptive Modula-
tion and Coding Schemas [14] to optimize the perform-
ance of WiMAX network. In [15] a three-tier QoS
framework is introduced where a Pre-scale Dynamic
Resource Reservation (PDRR) is proposed to allocate
frame bandwidth to UL sub-frame and DL sub-frame
with pre-scaled bounds. In [16] the presented baseline
network model is examined for fixed and dynamic real-
location, but under dynamic reallocation the resources
are relinquished completely from UL and allocated back
to DL when a certain preset threshold for DL occupancy
is met. We have modified the algorithm and proposed a
step-by-step allocation and similarly a step-by-step de-
allocation of UL resources back when certain threshold
of DL occupancy is met.
5. Proposed Strategy to Enable Reallocation
Capability
We provide here a basic mechanism that allows uplink
sub-frame to grow, borrowing resources from the DL
sub-frame, if the uplink utilization is high and the down-
link is underutilized. Our strategy ensures that resources
borrowed should be relinquished as and when required
by DL in order to ensure that with this introduced capa-
bility the DL performance should not be affected. Our
strategy is to observe the utilization of DL resources and
if DL is underutilized and UL is starving for bandwidth;
DL sub-frame duration may be reduced and UL sub-
frame duration may be increased by a certain quantum of
time.
5.1. The Proposed Algorithm
1) Baseline network scenario is simulated using OPNET
Modeler 14.5 (Wireless Suite) to test the impact of new
capability to redefine the boundaries of UL and DL
sub-frames.
2) Observe the frame allocation information and the
traffic behavior prior to the addition of the new capabil-
ity. For example; a) Observe sub-frame utilization, and b)
Application Performance in terms of throughput and
end-to-end packet delay on UL
3) Implement a mechanism to change the uplink and
downlink partition dynamically.
4) Implement an algorithm that performs re-allocation
of the sub-frames.
5) Ensure the performance of DL traffic should remain
unaffected with introduction of new capability.
6) Observe and analyze the results with the new capa-
bility.
7) Compare, analyze and summarize the simulation
results with the baseline network results.
5.2. Flowchart of our Proposed Algorithm
Here, we set certain thresholds for UL and DL sub-frame
utilization to decide whether or not the sub-frame dura-
tion need to be dynamically adjusted. The same is shown
here, in the flowchart of our proposed algorithm in Fig-
ure 3.
The above flow chart shows how the evaluation of the
sub-frames utilization is used to determine whether in-
crease the UL sub-frame size or prohibit any readjust-
ment of sub-frame duration.
5.3. Pseudo Code of Proposed Algorithm
1) Check DL Sub-Frame Utilization.
a) Is DL already reduced

A. Q. ANSARI ET AL. 891
Figure 3. Proposed scheme- flow chart.
b) Is DL utilization is below threshold
c) If DL is already reduced but utilization is still
under threshold go to Step 2.
d) Else revert back last reduction in DL
e) Reset the UL and DL sub-fame boundaries to the
previous values for each respectively and go to Step 4.
2) Check UL Sub-Frame Utilization.
a) Is UL sub-frame utilization above threshold?
b) Is DL under utilized?
c) If both conditions a and b are TRUE then check
condition in d.
d) Check DL sub-frame length < minimum allow-
able length
e) If condition fails in d then go to Step 4.
f) Else go to Step 3.
3) Increase UL Frame Size by One Step.
a. Update UL and DL sub-fame boundaries
4) EXIT.
6. Our Proposed Setup for Simulation
The baseline network is composed of one WiMAX cell
with four SS nodes. All SS nodes have an uplink appli-
cation load of 250 Kbps for a total of 1 Mbps. At specific
times, the Server generates 600 Kbps of application traf-
fic directed to SS-0 and SS-1; this creates a total
downlink application load of 1.2 Mbps. The cell uses
Scalable OFDMA frame with 512 sub-carriers and of 5
milliseconds duration. Uplink sub-frame is set to 12
symbol times (i.e. frame columns). Downlink sub-frame
is assigned 34 symbol times. For QPSK ½, the capacity
expected is Uplink: ~ 0.6 Mbps. Downlink: ~2.5 Mbps.
6.1. Preset Sub-frame Allocation between UL and DL
6.1.1. DL Traffic Behavior
Following graph (Figure 4) shows MAC load and
throughput for DL. It can be easily observed that DL
MAC load and throughput is same i.e. 1.2 Mbps. Total
DL capacity is 2.5 Mbps; thus there is enough capacity at
DL to successfully transport the DL load.
Moreover the ETE packet delay remains between
6~10 milliseconds which is fairly under acceptable range,
as shown in Figure 5.
Figure 4. DL traffic load and throughput.
Figure 5. End-to-end packet delay (DL).

A. Q. ANSARI ET AL. 892
6.1.2. UL Traffic Behavior
Following graph (in Figure 6) shows MAC load and
throughput for UL. We observe that UL MAC load is ~ 1
Mbps and throughput is 0.56 Mbps. UL capacity is 0.6
Mbps; thus the offered load is exceeding the capacity;
which results in high application delays ranging from 3.5
to 3.75 seconds (Figure 7). Thus UL is running out of
resources and do not have enough capacity to accommo-
date any further load.
6.1.3. Statistics of Usage and Usable Sizes (UL and DL)
Statistics results are also collected, in Figure 8, for data
burst percentage utilization and sub-frame usable size for
UL and DL.
From above diagram it is evident that DL usage is
about 60%; however UL usage is 100%. Also usable
Figure 6. UL traffic load and throughput.
Figure 7. End-to-end packet delay (UL).
Figure 8. Data burst usage and sub-frame usable size.
sub-frame size for UL is merely ~3.3 K symbols and
same for DL is nearly 11.3 K Symbols.
6.2. Dynamic Sub-Frame Allocation between UL
and DL
6.2.1. DL Traffic Behavior
The performance of DL traffic remained unaffected as
desired; i.e. 1.2 Mbps MAC load and throughput (Figure
9). Comparison could also be found in second frame.
Application end-to-end packet delay is observed to re-
main under ~8 milliseconds (Figure 10).
6.2.2. UL Traffic Behavior
Here we can confirm the major improvement in through-
put at UL. Comparison is presented between constant
and adaptive schemas. It is evident that load and
throughput are almost aligned on UL as well (Figure 11).
Now UL transports almost all the traffic originated form
MS and directed to BS.
This improvement has also resulted in remarkable de-
crease in packet end-to-end delay on UL; which has now
reduced to ~7 milliseconds from 3~4 seconds (Figure 12).
6.2.3. Statistics of Usage and Usable Sizes (UL and
DL)
From UL and DL data burst usage and Usable size (Fig-

A. Q. ANSARI ET AL. 893
Figure 9. DL traffic load and throughput. Figure 10. End -to -end packet delay (adaptive).
Figure 11. UL traffic load and throughput. Figure 12. End-to-end packet delay (UL).
7. Conclusions
ure 13) statistics it is evident that our scheme worked
well and effectively utilized the free DL bandwidth at
UL. With our strategy UL usage has improved because
the UL usable capacity increases to maximum when DL
utilization is lowest.
This paper presents a dynamic strategy that takes advan-
tage of underutilized DL sub-frame, allowing the UL
sub-frame to acquire temporarily free resources of DL

A. Q. ANSARI ET AL. 894
Figure 13. Data burst usage and sub-frame usable size.
when UL observes shortage of resources. Our proposed
mechanism ensures the performance of DL to remain
unaffected by continuous monitoring of the DL utiliza-
tion. In case of requirements, the sub-frame boundaries
could be readjusted and DL will be prioritized. The pro-
posed mechanism is tested under controlled environment
using OPNET Modeler for correctness and effectiveness.
In our proposed work, an observable improvement is
seen in both throughput and end-to-end packet delay at
UL without affecting the performance of DL.
8. References
[1] C. Eklund, et al., “IEEE standard 802.16: A technical
overview of the wireless MAN air interface for broad-
band wireless access,” IEEE Communication Magazines,
pp. 98–107, June 2002.
[2] A. Ghosh, et al., “Broadband wireless access with Wi-
MAX/802.16: Current performance benchmarks and fu-
ture potential,” IEEE Communication Magazine, pp.
129–36, February 2005.
[3] 802.16d Task Group, “IEEE standard for local and met-
ropolitan area networks part 16: Air interface for fixed
broadband wireless access systems,” IEEE, IEEE802-
16-2004 Version, 802.16d or Fixed WMAN, June 2004.
[4] 802.16e Task Group, “IEEE standard for local and metro-
politan area networks part 16: Air interface for mobile
broadband wireless access systems,” IEEE, Active Inter-
national Standard, 802.11e or Mobile WMAX, 2005.
[5] WiMAX Forum: Mobile WiMAX – Part I: A Technical
Overview and Performance Evaluation, June 2006.
[6] X. Yang, M. Venkatachalam, and S. Mohanty, “Exploiting
the MAC layer flexibility of WiMAX to systematically
enhance TCP performance,” IEEE Mobile WiMAX
Symposium, 2007.
[7] K. C. Chen and J. Roberto B. de Marca, “Mobile Wi-
MAX,” IEEE PRESS, IEEE Communications Society,
Sponsor, John Wiley & Sons, Ltd, April 2008.
[8] J. Sun, Y. Yao, and H. Zhu, “Quality of service scheduling
for 802.16d broadband wireless access systems,” Pro-
ceedings of IEEE 63rd International Conference of Ve-
hicular Technology 2006, VTC 2006-Spring, Vol. 3, pp.
1221–1225, 2006.
[9] G. Chu, D. Wang, and S. Mei, “A QoS architecture for the
MAC protocol of IEEE 802.16d BWA system,” Pro-
ceedings of IEEE International Conference on Commu-
nications, Circuits and Systems, Vol. 1, pp. 435–439,
June 2002.
[10] K. Wongthavarawatn and A. Ganz, “Packet scheduling
for QoS support in IEEE 802.16d broadband wireless ac-
cess systems,” International Journal of Communication
Systems, Vol. 16, pp. 81–96, 2003.
[11] J. Chen, W. Jiao, and H. Wang, “A service flow man-
agement strategy for IEEE 802.16d broadband wireless
access systems in TDD mode,” Proceedings of IEEE In-
ternational Conference on Communications 2005, Vol. 5,
pp. 3422–3426, May 2005.
[12] R. Iyengar, P. Iyer, and B. Sikdar. “Delay analysis of
802.16 based last mile wireless networks,” IEEE Globe-
com, USA, 2005.
[13] R. Pries, D. Staehle, and D. Marsico, “IEEE 802.16 capac-
ity enhancement using an adaptive TDD split,” Vehicular
Technology Conference, pp. 1539–1543, May 11–14,
2008.
[14] A. Q. Ansari, A. Q. K. Rajput, and M. Hashmani; “Wi-
MAX network optimization-analyzing effects of adaptive
modulation and coding schemes used in conjunction with
ARQ and HARQ,” Proceedings of 7th International Con-
ference on Communication Networks and Services Re-
search, pp. 6–13, 2009.
[15] M. Ma, J. Lu, S. Kumar, and B. Chong, “A three-tier
framework and scheduling to support QoS service in
WiMAX,” Proceedings of ICICS, 2007.
[16] “Understanding WiMAX model internals and interfaces,”
Opnetwork, 2008.