Wavelength Requirements for a Scalable Single-hop WDM Optical Network
Document Sample


(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No. 5, August 2010
Wavelength Requirements for a Scalable Single-hop
WDM Optical Network
Rabi W Yousif and Borhanuddin Mohd Ali Mohd Khazani Abdullah
Faculty of Engineering, Universiti Putra Malaysia Significant Technologies Sdn. Bhd.
Selangor, Malaysia Serdang, Malaysia
rabi.habash@gmail.com
Kamaruzzaman Bin Seman Mohd Dani Baba
Faculty of Science and Technology Faculty of Electrical Engineering
Universiti Sains Islam Malaysia, Nilai, Malaysia University Teknologi Mara, Shah Alam, Malaysia
Abstract—In this paper, we present a method for designing a to 100G. WDM optical networks can efficiently support
passive optical based single-hop wavelength division multiplexing multicasting since splitting light is inherently easier than
multicast architecture that can achieve a scalable structure and copying data into an electronic buffer. Applications of
form the basis of a wavelength efficient single-hop WDM multicasting include multimedia conferencing, distance
network. The proposed architecture minimizes the number of education, video distribution, distributed games and many
wavelengths required for efficient multicast service and also others [3, 4]. For cost reasons each node in single–hop WDM
minimizes tunability requirement of the transceivers. The networks deploys a rather small number of transceivers which
network size scalability is achieved by adding transmitters and is typically smaller than the number of wavelengths available
receivers to the designated groups. We show that the proposed for data transmission/reception. To increase the network
system can accommodate large tuning delays and keeps with
efficiency all wavelengths should be used at any given time.
suitable throughput when the number of wavelength is equal to
the number of nodes. We also show that the design can lead to a In single hop communication, the network must be able to
scalable structure while minimizing the number of wavelengths establish any possible connection in one hop, without
and tunability of the transceivers required for an efficient intermediate relaying or routing. This in effect implies that the
multicast service resulting in an improved system throughput network will have to change the connections it supports at
and delay performance. different times. Multi-hop networks have the ability to
circumvent the network capacity limitations. Each node is
Keywords-multicasting, wavelength-division multiplexing,
connected to only a few other nodes, as such only few
single-hop passive optical network
wavelengths are required per node. This greatly reduces the
wavelength bottleneck.
I. INTRODUCTION
In recent years, the Internet traffic has increased II. KEY DESIGN REQUIREMENTS
tremendously, because multimedia traffic such as video
streaming service, high resolution images, digital video and When designing a WDM network architecture and
audio conferencing, and business data distribution becomes protocol, the following key requirements and properties have
prevalent in the Internet. Some multimedia applications to be satisfied [5 - 8]:
require strict quality-of-service (QoS) or multicasting. Provide point–to–multipoint connections in order to
Current state-of-the-art dense WDM systems are using support multicast applications such as
narrow 50-GHz (0.4 nm) channel spacing. In such systems, videoconferences and distributed games in an
functions traditionally performed by electronics, such as economical and bandwidth–efficient manner.
switching, signal amplification, etc, are performed in the Add or remove network nodes in an easy and
optical domain, therefore achieve signal transparency. Thus, nondisruptive way without significantly degrading the
the capability for multicast transmission has become a very network performance.
important requirement for access networks [1, 2].
Traffic should not have to traverse a large number of
WDM technology has the potential to satisfy the ever- intermediate nodes to ensure smaller resource
increasing bandwidth needs of network users on a sustained requirements and smaller propagation delays.
basis. Today, optical backbones with a transmission speed of
40 gigabits per second are deployed. This technology is Provide some level of assurance that the service
reliable and will meet bandwidth needs for the next few years. requirements for different types of traffic, e.g., for
However, considering that traffic is growing by 40 percent a delay–sensitive, real–time, and interactive applications,
year on average, even 40G networks will have to be expanded are satisfied.
20 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No. 5, August 2010
Allocate network resources to all nodes which need to All stations can communicate with one another. In
send data. In networks with fair channel access control addition, a pair of fixed transceivers and control receiver both
each node ready to send data should have an equal are tuned to the control channel is dedicated for pre-
opportunity to transmit. transmission co-ordination. However, communication between
two nodes is possible only when the transmitter of the source
To cope with the resulting increased local traffic, metro node and receiver of the destination node are tuned to the
networks have to be easily upgradeable. Advanced same channel during the period of information transfer.
technologies, e.g., tunable transceivers with a wider
tuning range and a smaller tuning time, have to be
incorporated without network service disruption and IV. SYSTEM ASSUMPTION
reconfiguration. The behavior of the system is characterized by the
following assumptions:
III. SYSTEM DESCRIPTION There are N network nodes and W wavelength
The system under study is based on a broadcast-and-select channels in the system.
WDM architecture consisting of N network nodes connected
via optical fibers to a passive star coupler (PSC) as shown in Each node has a single–packet buffer, i.e., each node
can store at most one data packet at any given time.
Figure 1. There are W wavelength channels, where W N. The
bandwidth of a fiber is divided into W +1 channels, where W ≤ After transmitting a data packet in a given frame the
N. One of the channels, 0, is used as a control channel which buffer becomes empty at the end of that frame.
is shared by all nodes. The rest of the channels, 1,…,W, are
used as data channels. Each message is multicast to a set of l receivers where
l W N.
Users
Whenever the receivers of a multicast group are ready
Networks Station
A station possible architecture
to receive a data packet the source node's transmitter is
0 1 -W ready to transmit.
FR TR
Users
Data A packet that arrives at the start of a slot can be
2
1
Protocol To/From
transmitted during that slot to any one of the other (N
Networks PSC
Station
1
Control Processing Users
1) nodes with equal probability.
Data
Users FT TT
A node sends out its control packet in a frame with
A pair of
Optical Fibers 0 (i)
probability p, not only for retransmissions but also for
Networks Station
first–time transmissions.
PSC: passive star coupler
Multicast FT/R: fixed transmitter/receiver
Unicast TT/R: tunable transmitter/receiver Random selection of a destination node among the (N
Figure 1. A broadcast-and-select star-based WDM optical system.
1) nodes is renewed for each attempt of transmitting
a control packet.
Each node in the network is connected to the PSC by a
transmitting and receiving fiber, and each message is V. SYSTEM MODEL
addressed (multicast) to a number of receivers (destination set
size), randomly chosen from the N network nodes and each A. Node Structure
receiver tunes to one of the wavelengths that has a message The proposed architecture aims to define a minimum group
addressed to it. Also, each node has one fixed transmitter and of network nodes for a local structure, assign a unique
one fixed receiver in order to access the control channel. wavelength to a transmitter, and identify, for each transmitter,
Moreover, in order to access data channels, each node has one the minimum set of additional wavelengths needed to achieve
tunable transmitter and one tunable receiver, so that full communication with every other node in the local cluster and
connectivity can be achieved by tuning transmitters to the hence all the nodes in the network.
different wavelengths.
Figure 2 shows the node structure of the system. Each
Tuning times are not negligible with respect to the slot receiver is able to tune to all the wavelengths assigned to the
time. A centralized network controller allocates slots in a transmitters having direct links to it. Each processor can
WDM frame according to (long-term) bandwidth requests transmit data on a fixed number of wavelengths, but can
issued by users. When W ≤ N, two or more nodes share one receive data on a range of wavelengths by dynamically tuning
data channel. Each node is equipped with a buffer in which to the wavelength of a transmitting station. All the processors
arriving data packets are stored. Deploying tunable are synchronized at the optical coupler. The use of the same
transmitters and receivers at each node allows for load structure for both the transmitter and receiver is strategic [6 -
balancing since traffic between a given pair of nodes can be 9]. This will greatly simplify the coupling of the local
sent on any wavelength. In particular for nonuniform traffic, structure. Each node can switch channels (wavelengths) during
load balancing increases the channel utilization and improves execution by dynamically changing the injection current to the
the throughput–delay performance of the network. laser.
21 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No. 5, August 2010
Partitioning is desirable to design a hierarchical structure
using a cluster-based approach. This becomes quite obvious
N000 N001
since it is intended that the network should be scalable, flexible
and implemented with wavelength division multiplexing
techniques. Each local unit should be autonomous, so that
wavelength reuse can be achieved.
Node Structure N111 N010
0 1-W
C. Scalable Approach
F-Rx T-Rx If there are more than m access nodes, where m is the
Data
N110 N011 desired number of access nodes representing the regular local
Control
Protocol
Processing structure, a partition can then be accomplished by defining a
Data
minimum set of access nodes as the local structure and
F-Tx T-Tx N101 N100 applying the partition mechanism that is explained above to
0 (i) achieve the partition set. The transmitter in a group needs to be
placed according to the partitioning mechanism and same also
Figure 2. A network node structure. for the receiver. Scalability here has two aspects. First, the
transmitter and receiver of the new access node must be
Additionally, transmission and reception can be performed physically connected to the optical medium and second, the
on different channels. The single star topology consists of n added access node must be incorporated into the MAC
inputs, to which one transmitter is connected, and n outputs, to mechanism that controls the single hop connections [8-10].
which one receiver is connected. To incorporate the added access nodes in the MAC
mechanism, it requires only modifications for the control
B. Connection Establishment and Partitioning channel. This means that the number of added nodes must
There are basically two ways to achieve connections increase the number of control slots. To correctly reach each
between nodes in an optical network, path multiplexing and added node, all transmitters must be informed about its
link multiplexing. In the first the same wavelength has to be receiver configuration and its address.
assigned all to the links between source and destination, while
in the second, different wavelengths can be assigned on D. Wavelength Allocation
different links. To achieve single-hop connectivity, a We assign wavelengths such that the tunability for the
wavelength allocation mechanism needed to determine a path transmitters is minimum and optimal while the receivers must
for a new request. Each transmitter group can have direct links be able to tune to the maximum number of wavelengths used
to exactly two receiver groups. For any transmitter group, the in the entire network. Higher number of tuning disrupts the
two receiver groups that do not have a direct link to it consist network due to the retuning time. Our goal is to minimize the
of those that contain receivers with the same index notation as number of tuning so that the reconfiguration does not suspend
one of the transmitters in the transmitter group. For any the operation of the network.
transmitter/receiver group, there are two receiver/transmitter
groups that can have direct links to it and two others that do Since computer networks traffic changes rapidly, there is a
not have direct links to it [9]. The transmitter/receiver groups need for a good mechanism to change the current situation of
not having direct links to the same receiver/transmitter groups the network in terms of wavelength allocation (i.e., the current
are mutually exclusive. Finally, half of the number of wavelength assignment into a new wavelength assignment).
transmitters/receivers can be directly connected to half of the However, the number of channels is a limiting factor in a
number of receivers/transmitters simultaneously. Figure 3 WDM network, and is typically less than the number of nodes
shows the connection establishment procedures among the in the network. Therefore, more than one receiver is assigned
network nodes in order to achieve single-hop communication. to one channel. This problem is called wavelength assignment
problem.
T-Tx001
T-Tx 111 Wavelength blocking is a major problem with path
T-Rx000 T-Tx 000 multiplexing. One obvious disadvantage with the link
T-Rx001 multiplexing is the use of wavelength converters at
T-Tx 001
T-Rx010
intermediate nodes to eliminate blocking. This however,
T-Tx 110 T-Rx 011
T-Rx011
increases the cost and complexity of the system.
T-Tx 101
T-Tx011 In single-hop communication, all the nodes can reach any
T-Rx100 T-Tx 010
other node directly. This means that the transmitted data are
T-Rx101
not passed through any intermediate routing stages and remain
T-Tx 100
in optical form all the way from the source node to the
T-Tx 011 T-Tx 111
T-Rx110
destination node. In such mode of communication, a lightpath
T-Rx111
is established before a communication starts and the data
transmission is carried out in a pure circuit-switched manner.
Figure 3. Connection establishment procedures among the nodes.
22 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No. 5, August 2010
With dynamic traffic demands, new lightpaths need to be Now assume that R and R' represent a current wavelength
added to the logical topology each time an arriving connection assignment and a new wavelength assignment respectively.
request cannot be accommodated. The WDM techniques Based on bj for the N receivers where j = 1,…, N, the received
enable extraction of a larger amount of usable bandwidth. traffic for each node can be calculated. The wavelength
Routing and wavelength assignment algorithms fine-tune the assignments R and R' refers to the association between the
overall process by achieving orders of magnitude of nodes and the wavelengths, then
performance improvement. The goal is to present an efficient
dynamic distributed routing and wavelength allocation method
that minimizes path latency, wavelength blocking and the R = {Rc , w = 1,…, W} (2)
number of wavelengths applied [12-13].
where Rc is the set of receivers that are assigned to wavelength
E. Multicast Scheduling w, w = 1,…, W in the current wavelength assignment, i.e.,
The multicast scheduling problem can be described as
follow: given N stations, W available wavelengths for data Rw = {j | (j) = w} w = 1,…, W (3)
transmission, L slots global cycle and a W L slots allocation
matrix D, each station is equipped with a pair of tunable where (j) is the channel to which the receiver j is assigned.
transceiver and each needs time slots for tuning from i to j,
i j. For example, if (3) = 2 , it means that node 3 receiver is
assigned to channel 2. And
For a setup request, find a new feasible slot allocation
matrix Dnew with a new global cycle length Lnew such that setup
request is arranged into Dnew and all the QoS requirements of R' = {R'w, w = 1,…, W} (4)
accepted multicast in D are not affected.
where R'w is the set of receivers that are assigned to w in the
The proposed system follows a broadcast-and-select new wavelength assignment, i.e.,
methodology. It is assumed that each processor has a fixed-
tuned transmitter and a fixed-tuned receiver for transmitting
and receiving on the control channel. In addition, each R'w = {j | (j) = ' w} (5)
processor has a tunable wavelength transmitter and a tunable
wavelength receiver for transmitting and receiving on the data Now, assume number of tunings is D(R, R'), i.e., the
channel. Each processor can transmit data on a fixed number number of receivers that need to be retune to take the network
of wavelengths, but can receive data on a range of from its current wavelength assignment R to the new
wavelengths by dynamically tuning to the wavelength of a wavelength assignment R' and can be given by
transmitting station. The broadcast function is achieved using
a free-space optical coupler, while the select function is
achieved by using the wavelength tunable photodetector with W
filters. All the processors are synchronized at the optical
coupler. The transmitted signals from each transmitter are
D(R, R') = N - R
w1
w R'w (6)
routed to a free-space optical star coupler via optical
waveguides. The signals are also routed back to the receivers Figure 4 shows the current wavelength assiginment for a
via optical waveguides. network with 6 network nodes and 3 wavelength channels,
then R = {R 1, R 2, R 3} where R1 = {4, 5}, R2 = {1} and R3 =
When the traffic demand changes in a network, the current
{2, 3, 6}.
wavelength assignment becomes unbalanced and requires to
be changed. A new wavelength assignment is calculated
depending on the new traffic demand and a different 1
wavelength assignment will be produced, which requires a
number of tunings. However, if the current wavelength 2
allocation is taken into account when calculating the new one, 6
2
the required number of tunings can be reduced with an 3
acceptable wavelength assignment [4, 11].
3
The traffic in the network is represented by a traffic matrix 1
T = {tij} where tij is the rate of traffic from node i to node j
where i, j = 1,…, N. From the traffic matrix the total
bandwidth requirement bj of receiver j can be computed, 5
3 3
which equals the sum of the elements of the jth column of T: 1
4
N
Figure 4. Current wavelength allocation for a network with 3 wavelengths
b j t ij (1) and 6 nodes.
i 1
23 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No. 5, August 2010
When the traffic demand of the network changes, the
wavelength assignment is updated as shown in Figure 5.
R' = { R'1, R'2, R'3} where R'1= {4, 5}, R'2= {3} and,
R'3 ={1, 2, 6}. Then R1 = R'1
R2 R'2 , R3 R'3
W
Since D(R, R') = N - R
w 1
w R ' w , therefore,
D(R, R') = 6 – 4 = 2.
Figure 6. Number of tunings vs. number of wavelengths for a network with
50, 80, and 130 nodes and without partitioning.
As a result, two nodes in the network need to tune their
receivers.
1
6 3 2
3
3
1
1
5 1 2 3
4
Figure 5. Updated wavelength allocation for a network with 3 wavelengths
and 6 nodes. Figure 7. Number of tunings vs. number of wavelengths for a network with
100, 150, and 200 nodes with 4 partitions.
Figure 8 and Figure 9 show the effect of partitioning on the
VI. RESULTS AND DISCUSSION network performance in terms of the required number of
Figures 6 and Figure 7 show a trade-off between the wavelengths and number of tunings. It can be observed that
required number of tunings and the number of wavelengths the number of tunings decreases when the number of partitions
with and without partitioning for different network population. increases. However, the performance is not affected by
The tuning here is carried out to equalize as much as possible partitioning the network where the network performs almost
the busy channel and the free channel. The difference in the same for all cases and for most of the time the performance
number of tunings comes from the difference in the improves when number of partitions increases.
probability of exchanging two network nodes between two
different loaded channels in which this probability decreases
when the number of channels increases.
This is because when the number of channels increases, the
number of receivers assigned to channels decreases. This in
turns means that the maximum number of tunings is inversely
proportional to the frequency of tuning, so that, the most
important factor that affects number of tuning is how
frequently it is applied.
Moreover, the number of channels affects the number of
tunings, but in small scale. When the number of channels
increases, the total number of tunings decreases because fewer
receivers are assigned for the busy channels. Hence, there is a
small chance that the nodes are exchanged between two
Figure 8. Number of tunings vs. number of partitions for a network with 100,
different loaded channels.
150, and 200 nodes and 50 wavelengths.
24 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No. 5, August 2010
number of wavelengths required for efficient multicast service
and also minimizes tunability requirement of the transceivers.
The network size scalability is achieved by adding transmitters
and receivers to the designated groups. Wavelength scalability
is achieved through wavelength spatial re-use.
The problem of updating the wavelength assignment in
single-hop WDM networks is considered where the traffic
demand changes frequently and changing the channel
assignment becomes necessary. Minimizing the number of
tunings required can be achieved by exchanging one of the
receivers, which is assigned to the channel with high load,
with the appropriate receiver in the channel with minimum
load. Tuning is carried out to equalize as much as possible the
most loaded channel and the least loaded channel. In this
environment, the problem is finding an allocation of
wavelengths to receivers such that the number of
Figure 9. Number of wavelengths vs. number of partitions for a network transmissions of a multicast message is minimized. Since the
with 200 nodes and 1, 2, 4, 6 partitions. number of wavelengths is limited by technology, the problem
then becomes in finding the best partitioning scheme for the
Figure 10 shows the average packet delay versus receivers in the network. The proposed system can
throughput characteristics of multicast, transmitter, and accommodate large tuning delays and keeps with suitable
receiver for a PSC based single-hop WDM network with 200 throughput when the number of wavelength is equal to the
network nodes and retransmission probability equals to 0.5. number of nodes. When the number of wavelengths is
The network receivers are divided into two groups allowing comparable to the number of users the tuning time influence
wavelength reuse, which is possible during the reservation on the packet delay increases.
phase, i.e., the first slots of every frame when the control
In the context of wavelength allocation, a study on
packets are transmitted.
increasing the number of exchanges by taking into account the
As can be seen the transmitter throughput is affected by the channel which comes right after the most loaded channel can
many destination conflicting multicast transmissions. be considered for future work. Typically, this will include the
However, with two partitions multicast copies destined to the optimization of the communication, analysis of the
coupler will likely experience receiver conflicts since on communication patterns, and connection scheduling and
average each copy is destined to more receivers for two communication phase analysis. Also a study the impact of
partitions than for more partitions. large tuning delays on reconfiguration process and on the
network performance is required.
REFERENCES
[1] Han, K.E. and Yang, W.H, “Design of AWG-based WDM-PON
architecture with multicast capability”, In the Proceedings of the
HSN’2008 Workshop, vol. 1, USA, 2007.
[2] Zhone Technologies. Zhone WDM Optical Transport Equipment Chosen
to Extend Fiber Capacity and Increase Reliability for Pure Packet
Network. Press Release, 2007. [online] :
http://www.zhone.com/about/news/2007/on-telecoms-greece.en-
us.n2n?print=1
[3] Huang, S., Dutta, R., and Rouskas, G.N, “Traffic grooming in path, star,
and tree networks: complexity, bounds, and algorithms”, IEEE journal
on Selected Areas in Communications, vol. 24, no. 4, 2006.
[4] Nina, S, “Multicast routing and wavelength assignment in WDM
networks: a bin packing approach”, Optical Society of America
Publications, 2006.
[5] Maier, M., Scheutzow, M., and Reisslein, M, “The arrayed-waveguide
Figure 10. Delay vs. throughput (multicast, transmitter, and receiver) with 200 grating based single-hop WDM network: an architecture for efficient
network nodes and, 2 partitions, and retransmission probability equals to 0.5. multicasting”, IEEE Journal on Selected Areas in Communications,
21(9), 2004.
[6] Okorafor, E., and Lu, M, “Efficient distributed control routing and
VII. CONCLUDING REMARKS AND FUTURE WORK wavelength assignment mechanism for a scalable hierarchical single-hop
WDM all-optical interconnection network”, In Proc. 6th Joint Conf. on
The proposed method of designing a PSC based single-hop Information Sciences (JSIC), pp. 276 – 282, 2004.
WDM multicast architecture can achieve a scalable structure [7] Hamad, A. Tao, W. Kamal, A., Somani, A.K., “Multicasting in
that can form the basis of a wavelength efficient single-hop wavelength-routing mesh networks”, Computer Networks 50 (16): 3105-
WDM network. The proposed architecture minimizes the 3164, 2006.
25 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No. 5, August 2010
[8] Rouskas, G.N. Hierarchical Traffic Grooming, Chapter Book, 2008. PON) technologies for broadband access: a review”, Journal of Optical
[9] Rouskas, G.N., “Optical layer multicast: rationale, building blocks, and Networking, vol. 4, no. 11, 2005.
challenges”, IEEE Network, vol. 17, no. 1, pp. 60-65, 2005. [13] Luekijna, K. and Saivichit, C., “Multicast traffic reconfiguration in
[10] Pointurier, Y. and Brandt-Pearce, M, “Routing and wavelength WDM network for single node failure design”, In the Proceedings of
assignment incorporating the effects of crosstalk enhancement by fiber ICACT’2007, pp. 1833-1838, 2007.
nonlinearity”, 2005 Conference on Information Sciences and Systems,
The Johns Hopkins University, 2006. AUTHORS PROFILE
[11] Alfouzan, I. Jayasumana, A., "Dynamic reconfiguration of wavelength- Rabi W. Yousif received the B.S. degree in Electronics and Communications
routed WDM networks", 26th Annual IEEE Conference on Local Engineering from Mosul University, Mosul, Iraq in 1996, and the M.S. degree
Computer Networks Proceedings, pp. 477 - 485, 2002. in Computer Systems Engineering from Universiti Putra Malaysia, Serdang,
[12] Banerjee, A., Park, Y., Clark, F., Song, H., Kim, K., and Mukherjee, B., Malaysia in 2000. His current research interests include WDM optical
“Wavelength-division-multiplexed passive optical network (WDM- networks architecture, protocols, and network security.
26 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
Related docs
Other docs by ijcsis
Comparative Analysis between Split and HierarchyMap Treemap Algorithms for Visualizing Hierarchical Data
Views: 39 | Downloads: 0
Non-Preemptive Multi-Constrain Scheduling for Multiprocessor with Hopfield Neural Network
Views: 5 | Downloads: 0
Reliable Multipath Routing Protocol (RMRP) For Mobile Ad Hoc Networks Using Adaptive Video Compression
Views: 22 | Downloads: 1
Single CCTA-Based Four Input Single Output Voltage-Mode Universal Biquad Filter
Views: 71 | Downloads: 0
Get documents about "