Content uploaded by Abdou Ahmed
Author content
All content in this area was uploaded by Abdou Ahmed on Jan 02, 2016
Content may be subject to copyright.
A New Bandwidth Allocation Algorithm for
EPON-WiMAX Hybrid Access Networks
Abdou Ahmed
The University of Western Ontario
London, ON N6A 5B9, Canada
E-mail: aahmed84@ uwo.ca
Abdallah Shami
Alfaisal University
Riyadh, Saudi Arabia
Email: abchami@ieee.org
Abstract—Integration between Ethernet Passive Optical Net-
work (EPON) and Worldwide Interoperability for Microwave
Access (WiMAX) is a promising solution for next generation
access networks. In this paper, we devise a new architecture
framework for EPON-WiMAX hybrid networks that is more
reliable and extend the service coverage range. In addition, we
propose a new bandwidth allocation algorithm for the proposed
architecture that provides per-stream QoS protection, bandwidth
guarantee for real-time flows and improves the overall system
performance. Through intensive simulations, we show the effec-
tiveness of the proposed architecture and bandwidth allocation
algorithm.
I. INTRODUCTION
The complementary features of Ethernet Passive Optical
Network (EPON) and Worldwide Interoperability for Mi-
crowave Access (WiMAX) make the integration of these tech-
nologies a superior solution for access networks [1], [2], [3].
There are several important factors that motivate such integra-
tion. This integration takes advantages of the both technolo-
gies; such as bandwidth and reliability of optical networks,
and the mobility and flexibility of wireless networks.
In [1], [4] authors proposed four architectures for in-
tegration of EPON and WiMAX. These architectures in-
clude independent architectures, hybrid architectures, unified
connection-oriented architectures, and microwave-over-fiber
(MoF) architectures. Also they discussed how most of design
and operational issues such as: bandwidth allocation and QoS
support, network survivability, packet forwarding, handover
operation, and network design and planning; can be accessed
in these architectures. Integration of EPON and WiMAX were
described in [5] and [6] as well. The hybrid networks in [6], [5]
consist of a large WiMAX network that transmit its data
through a passive optical network. Other types of optical
wireless access networks were also proposed in [7], [8]. In
theses architectures, a wireless Base Station (BS), can be
attached directly to gateways/ONUs and sends data over an
Optical Network Unit (ONU) as in the architectures proposed
in [1], [6], or can be connected to gateways over through
intermediate wireless BSs by taking advantage of wireless
mesh networking. For this optical wireless networks, authors
mainly discussed the issues of routing, load balancing, packet
forwarding, and the placement of wireless BSs in wireless
mesh networks.
To date, some scheduling and Bandwidth Allocation mech-
anisms have been proposed to support QoS and improve
performance for delay sensitive traffic in EPON-WiMAX
networks [2], [5], [9], [10], [11], [12]. QoS-based Dynamic
Bandwidth Allocation (QDBA) [11] is incorporated with the
prediction-based fair excessive bandwidth allocation (PFEBA)
scheme in EPON to enhance the system performance. In
addition to QDBA, the authors in [11] propose a queue-based
scheduling scheme that efficiently satisfy the demand for band-
width request and enhance the efficiency of the system. The
DBA scheme in [2] enables a smooth data transmission across
optical and wireless networks, and an end-to-end differentiated
service to user traffics of diverse QoS requirements. This
QoS-aware DBA scheme supports bandwidth fairness at the
ONU-BS level and class-of-service fairness at the WiMAX
subscriber station (SS) level [13]. Bandwidth allocation and
the support of different service flow in [12] changed the EPON
MAC layer mechanism to adopt connection-oriented MAC
layer structure familiar with WiMAX.
In this work, we propose a new architecture for the
integrated WiMAX-EPON networks that overcomes the draw-
backs of earlier architectures and extend the converge range
of the network. A new bandwidth allocation algorithm that
supports QoS for different services types is proposed for the
proposed architecture.
The remainder of this paper is organized as follows.
The new hybrid network architecture is presented in Section
II. In Section III, a new bandwidth allocation for the new
architecture is described. The performance evaluation of the
proposed scheme is discussed in Section IV. Finally, Section
V concludes this work and outlines the future work.
II. PROPOSED HYBRID NETWORK
ARCHITECTURE
Access network architecture should be scalable, resilient,
support packet routing and forwarding, enable smooth protocol
adaption, and allow QoS support with efficient bandwidth
sharing.
The wireless-optical broadband-access network (WOBAN)
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.
architecture [7] consists of a wireless network at the front
end, supported by an optical network at the back end. This
architecture has many PON segments supported by a telecom
Central Office (CO). Each PON segment connects an Optical
Line Terminal (OLT), located at the CO, with a number
of ONUs, which are connected to wireless BSs from the
other side. These wireless BSs form the wireless portion of
WOBAN.
In the network architecture proposed in [8], the optical
backhaul consists of a ring and multiple tree networks. Each
tree network connects between one OLT and multiple ONUs.
OLTs of all tree networks are connected to the ring network.
Each ONU of any tree network is connected to a gateway
router of a Wireless Mesh Network (WMN). Both architectures
have many advantages, but the common drawback of both
architectures is that the entire network is under the same
management system; thus, the Central hup/central Office may
become the system bottleneck. Moreover, in the event of link
or node failures in architecture [7], the optical-wireless loses
full or a large portion of network connectivity, where users’
traffic will need to be re-routed. Architecture in [8] offers fault
tolerance against a single link or node failure but not for two
or more failures.
These drawbacks can be addressed by using a distributed
management system and by making the architecture more
immune against failures. In order to make optical part of
the network immune against failures we can use idea of the
reliable optical network solution was proposed in [14]. This
architecture is known as the Dual-Router Architecture. In
this architecture, the main access point (MAP) has dual large
backbone routers (BRs). BRs collect traffic from access routers
(ARs) and send it to other MAPs. Remote access routers are
often dual-homed over diversely routed unprotected/protected
fiber facilities. As AR-BR links are doubled and there are
two routers in each MAP; IP routers can restore traffic under
a variety of failures including fiber cuts, and router failures.
Instead, if each AR is connected to only a single BR, in the
case of link failure the AR will be isolated from the network.
In addition to overcoming the mentioned drawbacks of
architectures in [7], [8], there is a need for a hybrid network
architecture to extend the network converge distance of optical
network beyond the typical 25 km, allowing more end-users to
share an OLT link. This can be done by inserting an interme-
diate network between the backhaul and front end networks.
In urban areas where fiber is deeply deployed, an Intermediate
Network can be an additional optical network using the two
stages optical network design proposed in [15]. This will form
the Optical-Optical-Wireless (OOW) architecture, where OLT
is connected to a group of subOLTs instead of ONUs, and each
subOLT performs the functions of OLT in its own network
segment. Namely, each subOLT is connected to a group of
ONUs and each ONU is connected to a wireless BS (or more).
The Optical-Optical-Wireless architecture is shown in Fig. 1.
The splitter in the OOW architecture can be a WDM
Fig. 1. Optical-Optical-Wireless architecture
splitter, TDM splitter, or hybrid WDM/TDM implementation.
Although this work is considering TDM splitter in order to
simulate the OOW architecture and measure its performance,
we recommend WDM/TDM implementation to give the net-
work the capability to be upgraded to WDM extension.
As the subOLT is connected to the two routers at the OLT,
one of these routers is the primary gateway of the subOLT
and the second router works as the secondary gateway. The
subOLT stores information about its connection status with
the primary and secondary gateways, and each gateway keeps
track of the connections status with all subOLTs for whom it
works as primary gateway. When a subOLT has a packet in
its queue, it sends the packet to its primary gateway unless
the connection with this gateway is down or the Gateway is
highly congested, in this case the subOLT sends the packet to
the secondary gateway. When an OLT receives a packet for
a subOLT, packet received by the subOLT’s primary gateway,
to forward the packet, primary gateway checks the connection
with the subOLT, if it is up and has a reasonable transmission
time it sends the packet; otherwise the packet is sent to the
subOLT through the secondary gateway. The same operation
is done for the router that is connected to the splitter, as it
is connected to both the primary and secondary subOLTs.
Also the splitter should have the ability to de-multiplex in
downstream direction and multiplex in upstream direction to
select between primary and secondary subOLT connections.
Only the first stage (CO to Splitters) of the architecture
is more reliable in case of failures, as the second stage
(Splitters to ONU/BS) compensate for failure by user mobil-
ity in the wireless part. In the case of failure in ONU/BS
or its fiber connection, users served by this ONU/BS can
move to another working ONU/BS. The connection between
optical and wireless networks can be according to one of
the integration architectures in [1]; independent architecture,
hybrid architecture, unified connection-oriented architecture,
or microwave-over-fiber (MoF) architecture.
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.
III. BANDWIDTH ALLOCATION
In this section we present a new bandwidth allocation
algorithm for the Optical-Optical-Wireless hybrid network
architecture in Fig. 1; here the ONU and the WiMAX BS
are integrated in a single system box (ONU-BS), according to
hybrid architecture in [1].
As users are mostly serviced through WiMAX part of the
network, the bandwidth allocation algorithm should support
all service types defined in WiMax standard, namely Unso-
licited Grant Service (UGS), real-time Polling Service (rtPS),
extended real-time Polling Service (ertPS, defined in 802.16e),
non-realtime Polling Service (nrtPS) and best-effort (BE).
The WiMax part of the hybrid network works as the
traditional WiMax network. The BS manages the bandwidth
allocation and scheduling. However, the wireless part in the
proposed protocol differs from the stand-alone WiMAX net-
work in three aspects: First, a stand-alone WiMAX serves
connection admission requests in first-come-first-serve man-
ner, whereas the proposed protocol for the hybrid architecture,
serves these requests on priority based. Second, in the stand-
alone WiMAX, the scheduler manages packets, by scheduling
all the packets of the first SS in the time slot assigned to
this SS; then it schedules all the subsequent SSs packets in
sequence until it consumes all available bandwidth. Whereas
in the proposed protocol, the scheduler allocate the highest
priority type’s packets from all SSs first, followed by next
level priorities until it schedules all packets or reaches the end
of available bandwidth. Third, the stand-alone WiMAX does
not change its frame duration to meet the delay requirement
of the connection whereas in the proposed protocol frame size
is not fixed; the schedule changes the frame size if this does
not affect the current active connections.
Bandwidth allocation is performed according to the total
capacity (in bps) allocated for BS (CBS ). In hybrid network,
CBS is different than that in the traditional WiMAX network.
The traditional WiMAX network assumed that the backhaul
capacity that connects BS to the rest of the network (optical
line), CBS bh, is greater than traditional wireless channel
capacity. Hence CBS depends only on wireless capacity of the
BS, CBS wl . In hybrid network CBS bh varies from cycle to
cycle dependent on the bandwidth allocated to the ONU to
which the BS is attached to, therefore
CBS =min(CBS wl,C
BS bh).(1)
Bandwidth Allocation is a three level algorithm, the first
level of Bandwidth allocating runs at the BS, the second level
runs at the ONU, and the third level runs at the subOLT of
the OOW architecture. As the subOLTs connect to OLT in
a point-to-multipoint manner, OLT does not need to run a
bandwidth allocation algorithm. Only a switching mechanism,
like Ethernet switching, is needed to multiplex subOLTs’
traffics in upstream direction and de-multiplex traffic to one
of subOLTs in downstream direction.
A. BS Bandwidth Allocation (BSBA)
First step in BSBA is to set the frame size (Fl)oftheBS:
Fl=min(2Cd
i)∀i∈N. (2)
Where Nis the group of connections currently running at
the BS; Cd
iis the delay limit of the connection i.Flmust be
related to the min delay requirement of all connections.
After setting the frame size, different types of connections
are scheduled as follows:
•BSBA assigns bandwidth according to the strict priority
principle, where service types priorities from highest
to lowest are UGS, ertPS, rtPS, nrtPS, then BE. To
stop higher priority connections from monopolizing the
network, traffic policing is included in each SS. This
policing forces the connection’s bandwidth demand to
stay within its traffic contract.
•To provides per-stream QoS guarantee, BSBA allocates
each stream a bandwidth that meets its QoS requirements.
•UGS traffic: Each UGS connection is assigned a constant
bandwidth (fixed time duration) based on its fixed band-
width requirement. This policy is determined by the IEEE
802.16 standard.
•ertPS traffic: BSBA allocates requested bandwidth (fixed
time duration) based on their fixed period requirement.
•rtPS traffic: We apply the earliest deadline first (EDF)
service discipline to this service flow. Bandwidth needed
to transmit packets with earliest deadline is assigned to
each SS first. Then the bandwidth for other packets is di-
vided among SSs. The packets’ deadlines are determined
by the packet’s arrival time and delay requirement of the
connection.
•nrtPS traffic: BSBA applies weight fair queue (WFQ)
service discipline to this service type. Each nrtPS gets
bandwidth based on the weight of the connection (ratio
between the connection’s nrtPS average data rate and total
nrtPS average data rates).
•BE traffic: The remaining bandwidth is equally allocated
to each BE connection.
B. Bandwidth allocation at the ONU (ONUBA)
Bandwidth allocation for the Optical part of the hybrid
network is initiated at the ONU. The ONU receives data from
the BS(s) and the users connected directly to the ONU. It
classifies data based on their QoS requirements to suitable
queues and then sends a bandwidth request to the subOLT.
Each ONU has 8 priorities Queues (PQ). These PQs have
different priority levels and are described as follows:
a) UGS queue: holds data of UGS connections and holds
the highest priority.
b) ertPS queue: holds ertPS connections data and has
second level of priority. The size of this queue is the
actual bandwidth request of ertPS connections and it
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.
differs from the minimum amount reserved for ertPS as
we will discuss later.
c) rtPS-s-dead queue: this queue has the third level of
priority and holds data packets of rtPS connections with
deadline time in the next cycle.
d) rtPS-l-dead queue: holds data packets of rtPS connec-
tions with deadline time later than next cycle and comes
in forth level of priority. This queue will be scanned
periodically to move packets with deadline in next cycle
to the rtPS-s-dead queue.
e) nrtPS queue: for data of nrtPS connections and comes
in fifth level of priority.
f) Under-test queue: holds data of the connections that are
accepted by the BS and its performance is monitored to
check if it is eligible for final admission or not. This
queue holds data for all types of connections and sorts
them according to their priority levels, e.g. UGS, ertPS,
rtPS, then nrtPS. It is in sixth level of priority.
g) New-connections queue: holds bandwidth requirements
on new connections that can not be admitted by the BS.
It contains 2 elements for each connection; one element
for bandwidth requirement and second for frame size
required to satisfy delay requirement of connection. Data
of connections in queue are sorted in ascending order
according to required frame size. This queue has seventh
level of priority.
h) BE queue: holds data of BE connections and comes in
last level of priority.
In addition to these queues, ONU stores information about
BS; total data rates of all UGS connections, total of minimum
data rates of all ertPS connections, and total of mean data rates
of all rtPS connections. This information is updated when a
new connection is finally admitted by BS and when one of the
running connections completes service.
When an ONU requests bandwidth from the subOLT, it
sends a report message with 10 fields; please note that one
report can carry up to 13 fields. The essential bandwidth
information (UGS queue size + minimum ertPS rates + mean
rtPS rates) is available in the first field, and the difference
between the size of the second queue and min ertPS band-
width is available in second field. The difference between size
of the third queue and mean rtPS bandwidth is filled in the
third field. The size of fourth queue is stored in the fourth
field. The fifth and sixth fields of the report message carry the
expected rates for the ertPS and rtPS queues; here the ONU
does not only request bandwidth for existing data of ertPS
and rtPS, but also requests additional bandwidth for predicted
upcoming data between sending report message and arriving of
grant message. The remaining four fields of the report message
include the sizes of the queues from five to eight, respectively.
The subOLT grants bandwidth to the ONU; the ONU divides
this bandwidth among PQs.
C. Bandwidth Allocation at subOLT (subOLTBA)
First the subOLT sets its cycle time ( Tscycle )tothe
minimum frame size of all BSs.
The subOLT allocates bandwidth to the ONUs based on
total capacity that it is assigned by the OLT. The subOLT
allocates Bandwidth to the ONUs as follows:
a) First the subOLT assigns basic bandwidth part Bmin for
each ONU, where
Bmin =BUGS +Bmin
ertP S +Bsdl
rtPS .(3)
where BUGS ,Bmin
ertP S , and Bsdl
rtPS are the bandwidth
requested for UGS, ertPS-min, and rtPS-c-dead queues,
respectively.
b) Then, the subOLT tries to satisfy the bandwidth requests
for the rest of the queues as follows:
i) First calculate the remainder capacity Crem ; then
tries to satisfy the requirements of ertPS; every
ONU iwill be granted Brem
i,ertP S . This allocation
is based on the unsatisfied portion of the ertPS
requests Crem
i,ertP S and Crem hence,
Brem
i,ertP S =min(Crem
i,ertP S ,Crem
i,ertP S ×Crem
kCrem
k,ertPS
).
(4)
ii) Step (i) will be repeated, in sequence, for the rtPS,
predicted ertPS, predicted rtPS, nrtPS, under-test,
new-connections, and BE requests.
c) After assigning all requests to all queues, if Crem >0,it
is divided among ONUs according to their total requests
weight.
d) ONU is granted Btotal
ONU that is the sum of all component
grants in previous steps.
e) If all new connections requests can be satisfied by
the subOLT available bandwidth, CsOLT total , these
connections requests will be allocated to the requested
ONU and the connection will be placed under-test. Any
connection that can be accepted will be removed from
the queue. If it requires a frame size that cannot be satis-
fied by current cycle time, the cycle time will be changed
according to new frame size. Conversely, any connection
that cannot be admitted based on CsOLT total,is
rejected.
IV. PERFORMANCE EVALUATION
To simulate the proposed OOW architecture and the pro-
posed Bandwidth Allocation algorithm for this architecture,
we use ns-2 simulation software [16] and WiMAX module for
ns-2 [17] developed by The National Institute of Standards
and Technology. We simulate a segment of the network that
consists of 6 ONU/BS connected to a subOLT through 10 Gb/s
fiber optic. In this network, each BS serves 4 SSs and each SS
has 7 UGS, 8 ertPS, 7 rtPS, 9 nrtPS, and 5 BE connections.
Although we do not simulate the OLT - subOLT segment of
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.
12345678910
0
10
20
30
40
50
60
70
80
System Throughput Average Throughput of rtPS Service Type
Throughput (kb/s)
Number of rtPS Connections per SS
OOW System/200
EPMAX System/200
OOW rtPS Avrg.
EPMAX rtPS Avrg.
Fig. 2. System and Average of erPS Throughput
the OOW architecture in this work, but presented results give
a reasonable measure of the architecture’s performance. This
is due to the fact that the OLT - subOLT segment provides
almost fixed capacity.
Our objective is to evaluate the performance of the pro-
posed OOW setup and compare it with non-integrated EPON-
WiMAX network. We call this EPON-WiMAX network as
(EPMAX).
The average throughput of rtPS traffic and the system
throughput are shown in Fig. 2. Fig. 2 shows that the rtPS’
throughput in OOW outperforms that in EPMAX (30% to
44% improvement). The average throughput of rtPS in OOW
is constant while average throughput of rtPS in EPMAX de-
creases slightly as the number of connections per SS increases.
Moreover, Fig. 2 shows that the total system throughput in
OOW is generally greater than the total system throughput
in EPMAX (47% to 69% higher). Furthermore, the system
throughput in EPMAX decreases slightly or at least remains
constant as the number of connections per SS increase. Mean-
while, the system throughput in OOW increases as the number
of connections per SS increases. This shows better OOW
bandwidth utilization.
Average and max delays of rtPS are shown in Fig. 3. From
Fig. 3 we see that the maximum delay of rtPS in OOW is less
than the maximum delay of rtPS in EPMAX. The average
delay in EPMAX is almost double of that in OOW. Average
delay in OOW decreases slightly as the number of connections
per SS increases. Whereas in EPMAX, average delay increases
as the number of connections increase. Delay decreases as
number of connections increases because more packets from
each service type’s queue are transmitted together in the same
frame and this minimizes the delay. The number of packets
that can be transmitted in the same frame is limited by both
frame size and traffic rate. So the average delay in OOW
firstly decreases as number of connections increases, when the
maximum number of packets that can be accommodated in one
12345678910
0
10
20
30
40
50
60
70
80
90
100
Averag and Max. Delay of rtPS Service Type
Delay (ms)
Number of rtPS Connections per SS
OOW Avrg.
EPMAX Avrg.
OOW Max. /10
EPMAX Max. /10
Fig. 3. Average and Max. Delay of rtPS type
2 4 6 8 10 12
0
50
100
150
200
250
300
350
400
Rejection in UGS and All Service Types
Number of Rejected Connections
Number of UGS Connections per SS
OOW All
EPMAX All
OOW UGS
EPMAX UGS
Fig. 4. Rejection in UGS and all service types
frame is reached; the delay remains constant. In EPMAX, the
average delay decreases for the same reasons as in OOW. But
also, delay increases due to the fact that when many packets
are transmitted from each SS and as EPMAX transmits SSs
packets in sequence. Therefore, higher priorities packets of
SSs not serviced early in the cycle have to wait longer. So
the average delay firstly increases as number of connections
increases, when the decreasing factor become more effect than
increasing factor; the delay starts in decreasing.
UGS connections in both EPMAX and OOW are granted
fixed bandwidth. However, as network congestion affects on
the connections admission in the network, so it is reasonable
to compare the number of rejected connections in OOW and
EPMAX. The number of rejected connections is shown in
Fig. 4. Fig. 4 shows that OOW admits more connections
than EPMAX. Jitter distribution [18] of connection 3 in SS1
(randomly selected) is measured and shown in Fig. 5. The
graph shows that in OOW the jitter is centered at 40 ms
whereas in EPMAX the jitter is distributed over the rang from
few milliseconds to 300 ms.
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.
−0.1 −0.05 00.05 0.1 0.15 0.2 0.25 0.3 0.35
0
0.5
1
1.5
2
2.5
3x 105Jitter Probability Density Function
Probability Density
Jitter (s)
OOW 3 UGS /SS
OOW 9 UGS /SS
EPMAXx10000 3 UGS /SS
EPMAXx10000 9 UGS /SS
Fig. 5. Jitter pdf of connection 3 SS1.
0.9 11.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8
0
2
4
6
8
10
12
14
16
18
20
Rejection DUe to Delay Limit
Number of Rejected Connections per BS
Low limit of Delay (ms)
OOW
EPMAX
Fig. 6. Rejection in UGS Type due to Delay Limit
Computing the number of rejected connections due to
delay requirement of the connection is a good measure of
performance. Here number of connections is kept fixed and
their data rates guaranteed to be satisfied by bandwidth of
the network. Minimum delay of a connection will vary and
the number of connection rejected is measured. Min delay re-
quirement of each connection is generated randomly between
Min Delay and 10 mS. in Fig. 6, the number of rejected
connections are plotted as Min Delay is changed from 0.9
to 1.8 ms. The Number of rejected connections in EPMAX is
higher than the number of rejected connection in OOW. This
because OOW can change the frame size to meet the delay
requirements of a connection request, but EPMAX does not.
V. C ONCLUSION
EPON-WiMAX integration is a promising solution for
next generation access network. In this paper, we have pro-
posed a new Optical-Optical-Wireless architecture for EPON-
WiMAX hybrid network. In addition, we have proposed a
new bandwidth allocation algorithm. Through intensive sim-
ulation experiments, we showed the effectiveness of the pro-
posed scheme. In future work, another architecture (Optical-
Wireless-Wireless) will be proposed for EPON-WiMAX hy-
brid network. Joint MAC for both Optical-Optical-Wireless
and Optical-Wireless-Wireless architectures and take care
about channel errors in WiMAX will be proposed.
REFERENCES
[1] G. Shen, R. Tucker, and C. Chae, “Fixed mobile convergence archi-
tectures for broadband access: Integration of epon and wimax,” IEEE
Communications Magazine, vol. 45, pp. 44–50, August 2007.
[2] K. Yang, S. Ou, G. Ken, and H.-H. Chen, “Convergence of EthernetPON
and IEEE 802.16 Broadband Access Networks and its QoS-Aware
Dynamic Bandwidth Allocation Scheme,” IEEE Journal of Selected
Areas in Communications (JSAC), vol. 27, pp. 101 – 116, Feb 2009.
[3] X. Bai, A. Shami, and Y. Ye, “Robust QoS Control for Single Car-
rier PMP Mode IEEE 802.16 System,” IEEE Transactions on Mobile
Computing, vol. 7, pp. 416–429, April 2008.
[4] G. Shen and R. Tucker, “Fixed Mobile Convergence (FMC) Archi-
tectures for Broadband Access: Integration of EPON and WiMAX
(invited),” in Network Architectures, Management, and Application V,
vol. 6784, (Wuhan, China), pp. 678403–1–678403–13, SPIE, Nov 2007.
[5] Y. Luo, S. Yin, T. Wang, Y. Suemura, S. Nakamura, N. Ansari, and
M. Cvijetic, “QoS-Aware Scheduling over Hybrid Optical Wireless
Networks,” in Optical Fiber Communication and the National Fiber
Optic Engineers, pp. 1–7, 25–29, March 2007.
[6] Y. Luo, T. Wang, S. Weinstein, M. Cvijetic, and S. Nakamura, “Inte-
grating Optical and Wireless Services in the Access Network,” in OFC,
(Anaheim, CA), p. NThG1, March 2006.
[7] S. Sarkar, S. Dixit, and B. Mukherjee, “Hybrid Wireless-Optical
Broadband-Access Network (WOBAN): A Review of Relevant Chal-
lenges,” LIGHTWAVE TECHNOLOGY, vol. 25, pp. 3329 – 3340, Nov
2007.
[8] W. Shaw, S. Wong, N. Cheng, K. Balasubramanian, X. Zhu, M. Maier,
and L. Kazovsky, “Hybrid Architecture and Integrated Routing in a Scal-
able Optical-Wireless Access Network,” LIGHTWAVE TECHNOLOGY,
vol. 25, pp. 3443 – 3451, Nov 2007.
[9] F. Alsolami and P. Ho, “A Novel Two-level Hierarchical Scheduling
Scheme for Integrated WiMAX and EPON,” in ICS 2008, DSpace FC
University, Jan 2009.
[10] Y. Yan, H. Yu, H. Wang, and L. Dittmann, “Integration of EPON and
WiMAX Networks: Uplink Scheduler Design,” in SPIE 2008, vol. 7137,
pp. 713704–1, International Society for Optical Engineering, 2008.
[11] I. Hwang, “Advanced Dynamic Bandwidth Allocation and Scheduling
Scheme for the Integrated Architecture of EPON and WiMAX,” in 10th
Mobile Data Management: Systems, Services and Middleware, pp. 655
– 660, MDM ’09, 2009.
[12] T. Tang, G. Shou, Y. Hu, , and Z. Guo, “Performance Analysis of
Bandwidth Allocation of Convergence of WiMAX and EPON,” in
Networks Security, Wireless Communications and Trusted Computing,
vol. 1, pp. 662 – 665, NSWCTC ’09, 2009.
[13] X. Bai, A. Shami, and C. Assi, “On the Fairness of Dynamic Bandwidth
Allocation Schemes in Ethernet Passive Optical Networks,” Journal of
Computer Communications, vol. 29, pp. 2123–2135, July 2006.
[14] G. Li, D. Wang, J. Yates, R. Doverspike, and C. Kalmanek, “Detailed
study of IP/reconfigurable optical networks,” in First International
Conference on Broadband Networks, BROADNETS’04, 2004.
[15] A. Shami, X. Bai, N. Ghani, and C. Assi, “QoS Control Schemes for
Two-Stage Ethernet Passive Optical Access Networks,” IEEE Journal
On Selected Areas In Communications, vol. 23, Aug 2005.
[16] “http://www.isi.edu/nsnam/ns/.” Web available.
[17] “http://www.antd.nist.gov/seamlessandsecure/toolsuite.html.” Web avail-
able.
[18] X. Bai, A. Shami, C. Assi, and N. Ghani, “Jitter Performance in Ethernet
Passive Optical Networks,” IEEE/OSA Journal of Lightwave Technology,
vol. 23, pp. 1745–1753, April 2005.
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.