Sie sind auf Seite 1von 7

Design of Optimized WDM Networks using Heuristic Algorithms

G. Ramesh1, S. Sundara Vadivelu2 and Jose Anand3


1. Research Scholar, Anna University, Chennai
2. Professor/ECE, SSN College of Engineering, Chennai
3. Assistant Professor/ECE, JAYA Engineering College, Chennai
Email: - ramesh6_2000@yahoo.com, sundaravadivelus@hotmail.com, joseanandme@yahoo.co.in

Abstract – Advances in optical networks, together with layer than in the electrical domain. To build an
the rapid rising demand for network bandwidth, are economical network, the following elements are
fueling an increasing amount of research in the field of required:
optical networks. Optical communication has become a
promising networking choice to meet increasing • Flexibility — involves significant interchanges to
demand on bandwidth. Some of the emerging offload traffic in various ways.
bandwidth-intensive computing and communication • Optics adapted to the distances — systems must
applications such as data browsing on the web, video allow the optical signal to go several thousand
conferencing, e-commerce, high-definition video/audio kilometers without requiring regeneration.
on-demand processing, data mining, database and However, cheap optics for the shorter distances are
decision-support transactions etc. One of the essential preferred. Also, to save cost, each road where the
components for the dynamic provisioning of optical traffic is heavy must be able to carry maximum
networks across multiple domains is the routing traffic, which leads to the final requirement.
strategy adopted. Future optical networks will require • Option for a number of signals to share the
new protocols in order to route and support the on- same optical fiber — enable the best use of the
demand provisioning of optical networks between fiber, if this is important. Installing cable is
different domains. This paper presents a heuristic expensive, especially on long distances. Getting
algorithm for the issues of minimizing the cost of the right of way and digging takes time and
optical cables and to achieve the optimal route in the money, and this is not an attractive proposition for
design of Wavelength Division Multiplexing (WDM) a telecom carrier that hopes to get its network
networks. running and producing revenues.

Keywords:- optical networks, WDM, heuristic The remaining paper is organized as follows.
algorithms, routing, topology, NSFNET. Section 2 describes the review of literature on WDM
networks. Section 3 gives the method to find the cost of
I. Introduction the optical cables in the network and section 4
describes the way to optimize the route in WDM
To cope up with the recent rapid growth of data optical networks. Section 5 illustrates with analysis on
traffic volume, it is required to establish large capacity experimental results on three different networks.
optical networks. In network links, the transmission Finally conclusion is added in section 6.
capacity can be increased by using various optical
modulation formats with the wavelength, phase, II. Related Work
amplitude and polarization parameter [8]. An approach
of selecting destinations without regard for resource Ramamurthy et al [20] presented Integer Linear
availability or location may potentially limit Program (ILP) formulations for the routing and
performance in distributed computing applications, wavelength assignment problem are developed for a
since alternative destinations may result in lower cost. static traffic demand for both path and link protections
schemes. Ho et al [11] proposed that the primary path
The future data and telecommunication networks is divided into several overlapped segments. The
are likely to consist of elements such as routers, calculation of the backup path for each sub-domain is
switches, DWDM systems, add/drop multiplexers done individually. Redundant trees are used to provide
(ADMs), optical cross-connects, that will use rapid recovery is presented by Medard et al [16]. Their
Generalized Multi-protocol Label Switching (GMPLS) algorithm constructs two trees in such a fashion that
to dynamically provision resources and to provide each destination vertex is connected to the source by at
network survivability using protection and restoration least one of the directed trees when any vertex in the
techniques. Switching and routing high capacity graph is eliminated.
connections are much more economical at the optical
Hu et al [12] provided a decomposition method that
Anand et al [1] presented the performance of sub- divides the traffic grooming problem into two smaller
path protection scheme in terms of capacity utilization problems and then solved each problem independently.
and recovery time, compared with path and link Antonakopoulos et al [2] proposed the approximation
protection schemes. Zang et al [24] developed an on- algorithms for minimizing the total equipment cost and
line network control mechanism to manage the for minimizing the lightpath count. Chen et al [6]
connections in WDM mesh networks using path provided a hierarchical framework for traffic grooming
protection schemes. They use the two-step approach to in a WDM mesh network.
route the connections. A new multiplexing technique
called primary-backup multiplexing is proposed by In the study of survivable routing Modiano et al
Mohan et al [18] to improve resource utilization. This [17] deal with the problem of routing lightpaths, so that
technique allows a primary lightpath to share the same the virtual topology remains connected when there is a
wavelength with some backup lightpath. single fiber cut. The problem is formulated as an
Integer Linear Program (ILP) and solved for regular
Sen et al [21] proposed to use the link-disjoint path topologies and ring topologies. Madhyastha et al [15]
pair, whose longer path is shortest among all such pairs addressed the multicast traffic grooming problem in
of paths, for path protection so that the delay on the metropolitan WDM ring networks with the objective of
backup path is minimized. They prove that the problem minimizing electronic copying. The authors presented
of finding such a pair of paths is NP-complete, and an ILP formulation and then developed a heuristic
they use the one-step approach as the approximation approach that consists of three phases: routing, circle
solution. Xin et al [22] attempted to optimize the construction and grouping of circles.
network resource utilization of each call by minimizing
the overall cost of the primary and backup path. The Billah et al [5] considered the multicast traffic
paths are selected from K precomputed candidate route grooming problem in WDM mesh networks with
pairs. sparse nodal light splitting capability. Chowdhary et al
[7] proposed a non-linear programming formulation
Bernstein et al [4] proposed the Domain-to- was introduced as an analytical model for the multicast
Domain Routing Protocol (DDRP), which is basically a traffic grooming problem in splitting networks
hierarchical extension of OSPF-TE that is supported by followed by a number of heuristic solutions. The
a modified version of Dijkstra’s algorithm. For authors addressed the problem of many-to-one traffic
instance, it returns a single path at a time, so grooming in WDM mesh networks with the objective
complementary algorithms need to be adopted for path of minimizing the number of wavelengths and SONET
protection and diverse routing purposes. Yannuzzi et al ADMs. A survey of advances in multicast traffic
[23] proposed a very pragmatic approach. The proven grooming is presented by Kamal [14].
scalability of path-vector protocols in the current
Internet suggests that, if OBGP were deployed in the III. Cost of Optical Cables
future, it would probably operate and scale similarly as
BGP does today. They introduced a set of minor In this section, we discuss the strategies to
modifications to a path-vector protocol that can be minimize the cost of optical cables. Normally by
feasibly implemented in large scale multi-domain properly organizing the optical network units, the cost
optical settings, as in the case of OBGP, and show that of the cables can be reduced. The objective is to
these modifications are sufficient to drastically minimize the sum of total optical cable cost. Optical
improve the performance of OBGP. cables are used to connect between the optical network
units and the distributed arrayed gratings in deploying
Din [10] considered the routing and wavelength optical distribution networks using WDM. This cabling
assignment problem for anycast (ARWA) in will form a tree structure.
wavelength division multiplexed networks. A hybrid
method, which combines simulated annealing and As deployed in a tree structure the vertex set of the
genetic algorithm techniques, was proposed to solve optical cable tree is a subset of the vertex set of the
the ARWA problem. Anycast in OBS networks has construction tree [13]. The edge connecting (u, v) in
been studied by Leenheer et al [9] a proposed an OBS- the optical cable tree is actually the route connecting
based architecture for supporting Grid computing. vertex u and vertex v in the construction tree. The total
optical cable cost is the sum of the optical cable cost in
Zhu et al [25] considered the traffic grooming all sub-trees. The length of the edge (u, v) will be
problem in a WDM mesh network. They introduced an represented by ǀ(u, v)ǀ and the weight of the edge (u, v)
ILP formulation and then developed heuristic solutions. will be represented by w(u, v).
Generally u will be the parent node of v. Number Min-Cost Resource First algorithm
of child nodes of node v is given by w(u, v). The fibers
in the WDM optical networks will transmit the In the MCF algorithm, all the computing resources
upstream and downstream signals in bidirectional are given a rank based on the link cost from the source
format. The cost of optical cables in the sub-tree is node. The min-cost destination among the destination
given by, node is first selected by the source node. Now a
shortest path is established between the source and
destination nodes. Then the next min-cost destination
q(w ) ǀ(u, v)ǀ (1) which is not on the established path will be selected
(u, v) and the tree grows. This procedure will continue until
all the available resources are found. Figure 1
where, E(t) is the set of edges of tree t and q(x) is the illustrates the flow chart of the MCF algorithm.
unit price of optical cables containing x optical fibers.

IV. Route Optimization

Routing is the process of selecting paths in a


network along which to send network traffic. Routing
is performed for many kinds of networks, including the
telephone network, electronic data networks (such as
the Internet), and transportation networks. Routing are
or two types namely source routing and hop-by-hop
routing. In source routing, the optical network route is
calculated at the edge node.

Source routing has the advantage of knowing the


path length in advance. Due to this the optimal value of
the offset time can be evaluated correctly. Hop-to-hop
routing is used as the output link at every intermediate
node. This has the disadvantage of not knowing the
length of the path in a prior way and the offset must be
estimated [3].

For the route optimization, a problem in multi-


resource manycast with mesh network is considered
with 6 node, 10 node and 14 node as shown in figure 3,
4 and 5 respectively. In the network, G = (V, E)
represent a graph, in which V is the set of vertices and
E is the set of edges. The number of computing
resources at the each vertex i be ri. In this the available Fig. 1Flowchart of MCF
resources can be reserved by other vertex too. The
objective is to minimize the cost of the tree connecting Most Available First algorithm
with at least k computing resources, which is defined
as, In the MAF algorithm, the destination is first
selected based on the vertex which is having the most
Ʃ cij (2) available resources. Then the shortest path is
established between the source node and the
destination. In this algorithm all the computing
where, cij is the cost of the edge eij between the vertex i resources on the path will be included. Now the next
and vertex j on the tree. Now \for the multi-resoure destination with most available computing resources
manycast problem has two heuristic algorithms based which is not on the selected path is selected. This
on the shortest part tree algorithm are used namely, procedure is stopped until all the required computing
Min-Cost Resource First (MCF) and Most Available resources are included. Figure 2 illustrates the flow
First (MAF) [19]. Both MCF and MAF algorithms are chart of the MAF algorithm.
based on the available computing resources at the
vertex.
Fig. 3 A 6-node NSFNET

Fig. 4 A 10-node NSFNET

Fig. 2 Flowchart of MAF

V. Experimental Results

In this section, the performance of the proposed


algorithms is evaluated in a small, medium and large
sized network with 6-node, 10-node and 14-node
NSFNET, which is shown in figure 3, 4 and 5
respectively. Five set of experiments are conducted on Fig. 5 A 14-node NSFNET
small network with number of sessions in each
experiment, number of members in a session and no. of Figure 6, 7 and 8 illustrates the performance
members are randomly selected between [2, 4], [2, 5] analysis of average tree cost for the heuristic
and [0, 5] respectively. Another five set of experiments algorithms in the small, medium and large sized
are conducted on small network with number of network with 6-node, 10-node and 14-node NSFNET.
sessions in each experiment, number of members in a The average cost for the heuristic algorithms for all the
session and no. of members are randomly selected network is increased with the increase in required
between [4, 6], [2, 5] and [0, 9] respectively. Another resources. From the analysis it is shown that for all the
five set of experiments are conducted with large network size MAF achieves the highest tree cost. As
network. For large network the number of sessions in the MAF selects the vertex with the most available
each experiment is randomly selected between [6, 8]. resource each time, the cost of the path may be too
The number of members in a session is randomly high. The performance of MCF is close to the
selected between [2, 5], while members are randomly performance of MAF, since both the two approaches
selected between [0, 13]. Finally, traffic demand of consider the vertex with the minimum cost.
members in a session, in all the fifteen experiments is
randomly selected between [1, 16].
networks. In future characteristic such as reliability,
survivability and coverage can be improved by a
network topology that is scalable and enables traffic
routing. It is believed that the optical Internet provides
new and strong incentives to build a flexible, efficient
and bandwidth-abundant fiber-optic network
infrastructure capable of providing ubiquitous IP
services.

References

[1]. Anand V., Chauhan S. and Qiao C., “Sub-path


Fig. 6 Analysis of Average Tree Cost with 6 node protection: Anew framework for optical layer
survivability and its quantitative evaluation”,
Department of Computer Science and
Engineering, State University of New York at
Buffalo, Technical Report2002-01, 2002.
[2]. Antonakopoulos S. and Zhang L.,
“Approximation algorithms for grooming in
optical network design”, in Proceedings of IEEE
INFOCOM’09, pp. 1548-1556, 2009.
[3]. Aracil J., Akar N., Bjornstad S., Casoni M.,
Christodoulopoulos K., Careglio D., Fdez-
Palacios J., Gauger C., Gonzalez de Dios O., Hu
G., Karasan E., Klinkowski M., Morato D.,
Nejabati R., Overby H., Raffaelli C.,
Simeonidou D., Stol N., Tosi-Beleffi G. and
Fig. 7 Analysis of Average Tree Cost with 10 node Vlachos K., “Research in optical burst switching
within the e-Photon/One network of
excellence”, Journal of Optical Switching and
Networking, Elsevier, 4, pp. 1-19, 2007.
[4]. Bernstein G., et al., “Domain to domain routing
using GMPLS, OSPF extension V1.1 (Internet
Draft), OIF2002.23.06, July 2002.
[5]. Billah A., Wang B. and Awwal A., “Multicast
traffic grooming in WDM optical mesh
networks”, in proceedings of GLOBECOM’03,
Vol. 5, pp. 2755-2760, Dec. 2003.
[6]. Chen B., Rouskas G. and Dutta R., “On
hierarchical traffic grooming in WDM
networks”, IEEE/ACM Transaction on
Networking, Vol. 16, No. 5, pp. 1226-1238,
Fig. 8 Analysis of Average Tree Cost with 14 node
October 2008.
[7]. Chowdhary G. and Murthy C. S. R., “Grooming
VI. Conclusion
of multicast sessions in WDM mesh networks”,
Modern high-performance computing systems in Workshop on Traffic Grooming, 2004.
require optical networks with high capacity, extremely [8]. Cisco Systems Inc., “Cisco visual networking
high throughput and low latency in order to pass index – forecast and methodology”, 2007-2012,
messages between thousands of processors and l Cisco 2008 white paper, updated June 2008,
network elements. This paper presents a heuristic http: //www.cisco.com/.
algorithm for the issues of minimizing the cost of [9]. De Leenheer M., et al., “An OBS-based grid
optical cables and to achieve the optimal route in the architecture”, in Proceeding of the Globecom
design of WDM networks. Among the two heuristic Workshop on High-Performance Global Grid
algorithms MAF and MCF, MCF is the optimized one Networks, November 2004.
with minimum cost in all the three experimented
[10]. Din D. R., “A hybrid method for solving ARWA [22]. Xin C., Ye Y., Dixit S. and Qiao C., “A joint
problem on WDM network”, Computer lightpath routing approach in survivable optical
Communications, 30 (2), pp. 385-395, 2007. networks”, Proceedings of the SPIE Asia-Pacific
[11]. Ho P.H. and Mouftah H. T., “SLSP: a new path Optical and Wireless Communications, Vol.
protection scheme for the optical internet”, 4585, pp. 139-146, Nov. 2001.
Proceedings of OFC’01, Anaheim, CA, Vol. 2, [23]. Yannuzzi M., Masip-Bruin X., Fabrego G.,
March 2001. Sanchez-Lopez S., Sprintson A. and Orda A.,
[12]. Hu J. Q. and Leida B., “Traffic grooming, “Toward a new route control model for multi-
routing and wavelength assignment in optical domain optical networks, IEEE
WDM mesh networks”, in Proceedings of IEEE Communications Magazine 46 (6), pp. 104-111,
INFOCOM’04, pp. 495-501, 2004. 2008.
[13]. Jingjing Zhang and Nirwan Ansari, “Minizing [24]. Zang H and Mukherjee B., “Connection
the Arrayed Waveguide Grating Cost and the management for survivable wavelength-routed
Optical Cable Cost in Deploying WDM Passive WDM mesh networks”, SPIE Optical Networks
Optical Networks”, Journal of Optical Magazine, Vol. 2, No. 4, pp. 17-28, July 2001.
Communication Networking, Vol. 1, No. 5, pp. [25]. Zhu K. and Mukherjee B., “Traffic grooming in
352-365, October 2009. an optical WDM mesh network”, IEEE Journal
[14]. Kamal A. E., “Algorithms for multicast traffic on Selected Areas Communication, Vol. 20, No.
grooming in WDM mesh networks”, IEEE 1, pp. 122-133, January 2002.
Communication Magazine, Vol. 44, No. 11, pp.
96-105, Nov. 2006.
[15]. Madhyastha H., Chowdhary G., Srinivas N. and G. Ramesh is working as Assistant
Murthy C. S. T., “Grooming of multicast Professor in the Department of
sessions in metropolitan WDM ring networks”, Information Technology at KLN College
Journal of Computer Networks, Vol. 49, pp. of Engineering, Madurai, Tamil Nadu,
561-579, 2005. India. He has completed graduation in
[16]. Medard M., Finn S. G., Barry R. A. and Electrical and Electronics Engineering and Post
Gallager R. G., “Redundant trees for preplanned Graduation in Computer Science and Engineering. He
recovery in arbitrary vertex-redundant or edge- is doing research work in optical networking.
redundant graphs”, IEEE/ACM Transactions on
Networking, Vol. 7, No. 5, pp. 641-652, Oct. S. Sundara Vadivelu is working as
1999. Professor in Department of Electronics
[17]. Modiano E. and Narula-Tam A., “Survivable and Communication Engineering at SSN
lightpath routing: A new approach to the design College of Engineering, Kalavakkam,
of WDM-based networks”, IEEE Journal on Chennai, Tamil Nadu India. His research
Selected Areas in Communications, 20 (4), pp. interest is in Optical Communication and
800-809, 2002. Networks. He has published many international and
[18]. Mohan G., Siva Ram Murthy C. and Somani A. national journals in optical domain.
K., “Efficient algorithms for routing dependable
connections in WDM optical networks”, Jose Anand received his Diploma from
IEEE/ACM Transactions on Networking, Vol. the state board of Technical Education,
9, No. 5, pp. 553-566, Oct. 2001. Tamil Nadu, Bachelor of Engineering
[19]. Qingya She, Nipatjakorn Kannasoot, Jason P. Degree from Institution of Engineers
Jue and Young-Chon Kim, “On finding (INDIA), Calcutta, Master of
minimum cost tree for multi-resource manycast Engineering in Embedded System Technologies from
in mesh networks”, Journal of Optical Switching Anna University, Chennai, Master of Arts in Public
and Networking, Elsevier, 6, pp. 29-36, 2009. Administration from Annamalai University, and
[20]. Ramamurthy S. and Mukherjee B., “Survivable Master of Business Administration from Alagappa
WDM mesh networks, part I-protection”, University. He is a member of CSI, ISTE, IEI, IETE
Proceedings of IEEE INFORCOM’99, Vol. 2, and INS. He received State 3rd Rank in Bachelor of
(New York), pp. 744-751, March 1999. Engineering. He published/presented several technical
[21]. Sen A., Shen B. H. and Bandyopadhyay S., papers in various Conferences and Journals, and also
“Survivability of lightwave networks-path published several books for various polytechnic
lengths in WDM protection scheme”, Journal of subjects in Electrical, Electronics and Computer
High Speed Networks, Vol. 10, pp. 303-315, Science Disciplines.
2001.

Das könnte Ihnen auch gefallen