You are on page 1of 10

ISSN:2277-9043 International Journal of Advanced Research in Computer Science and Electronics Engineering, Volume 1, Issue 1 March 2012

A Novel Approach for Light Trail Minimization Using Genetic Algorithm in Optical Networks
Soumya Paul, Inadyuti Dutt

Abstract: Recently a new architecture called light trail has been proposed for carrying data traffic in optical networks and has emerged as a promising candidate for enabling IP over WDM networks. Optimizing the number of light trails that must be setup to service connection requests as they arrive, is an important light trail design problem. Most of the existing light-trail assignment algorithm,s adopt ILP (Integer Linear Programming) approach. Due to high time complexity such algorithms are not scalable. In this paper, a heuristic approach with Connection Matrix (CM) and Genetic Algorithm (GA) is proposed for reducing the number of light trails for given a graph (network) and a number of connection requests, The proposed heuristic gives an optimal solution with reasonably lower complexity and reduces blocking probability. Keywords/Index Terms: Heuristic; Internet Protocol (IP); Wavelength Division Multiplexing (WDM); Lightpath; Light-trail; Integer Linear Programming (ILP); Genetic Algorithm (GA).
I. INTRODUCTION

optically throughout the course of the lightpath. This requires that an appropriate path (route) and a free wavelength on this path be determined such that this lightpath connects the sending and receiving clients optically. The process of lightpath determination may be repeated till either a connection is established or all the options have been exhausted. Normally it is required that the same wavelength be allocated on all the fiber links in the lightpath. This is known as the wavelength continuity constraint. This lightpath approach does not yield good bandwidth usage due to the burst nature of data traffic and the fact that most users generate relatively lower bit rate traffic compared to the bandwidth of the lightpath. A light trail is a unidirectional optical bus between a convener node (i.e. light-trail head) and an end node. It is implemented using a wavelength, and allows the intermediate nodes to share the bandwidth by adding and dropping traffic, provided that the total traffic load carried is not more than the bandwidth of the wavelength. At each intermediate node, this is achieved by splitting a percentage of transponder. Meanwhile, the rest of the optical signals continue to propagate along the light-trail. Thus a light trail is similar to a lightpath, yet it is different in the sense that intermediate nodes can also access the unidirectional connection for data transfer, resulting in better source utilization. One of the light trail design problem is to minimize the number of light trails needed to serve a given number of connection requests. This is an integer linear programming (ILP) problem; due to high complexity such algorithms are not scalable. The proposed heuristic finds a set of light-trails to cover the given requests in survivable optical networks such that the total number of light-trails required is minimized.
II. SURVEY STUDY OF EXISTING ALGORITHMS

Optical networks are high-capacity telecommunications networks based on optical technologies and components that provide routing, grooming, and restoration at the wavelength level as well as wavelength-based services. In an all-optical network (AON), all network-to-network interfaces are based on optical transmission, all user-tonetwork interfaces use optical transmission on the network side of the interface and all switching and routing within AON network nodes is performed optically. In a wavelength routed optical network, data transfer between clients requires setting up a unidirectional lightpath (channel) in the optical layer. Such a channel allows source-to-destination transfer

All Rights Reserved 2012 IJARCSEE

ISSN:2277-9043 International Journal of Advanced Research in Computer Science and Electronics Engineering, Volume 1, Issue 1 March 2012

Ashwin Gumaste and Imrich Chlamtac [1] propose the concept of light-trails to alleviate the problems in sustaining IP centric communication at the optical layer. With the principle of access to the all-optical path at any node, a light-trail offers to provide full unidirectional optical connectivity, while avoiding the need for dynamic, burst type, optical paths establishment. For setting up light-trails as well as signaling purposes within light-trails an out of band communication channel called optical service channel (OSC) exists which is dropped and processed at each node. Michael T. Frederick, Nathan A. VanderHom and Arun K. Somani [2] discuss the advantage of light-trail over MPLS (Multiprotocol Lebel Switching), OBS (Optical Brust Switching) and OPS (Optical Packet Switching) as solutions to realizing an all-optical network. Proposed solutions for all-optical networks such as MPLS and OBS monopolize a wavelength throughout connection duration and do not take advantage of switches that are already configured. OPS is a technology that is not mature and most likely will not be in the near future. Light trail technology attempts to address the shortcomings of theses various technologies by allowing intermediate connection nodes to also use the unidirectional bus, as well as maximize the reuse of already configured optical switches. Ashwin Gumaste, Gabriel Kuper and Imrich Chlamtac [3] used the concept of clustered light-trail (CLT). While ring networks, due to the fact that their nodal degree is 2, can support only linear light-trails, mesh networks, on the other hand can support tree shaped light-trails which are called clustered light-trails (CLT). A Clustered Light-trail is a tree rooted at a node called the convener. The end nodes of the branches may be distinct, and so the CLT can have multiple end nodes. Wensheng He, Jing Fang, Arun K. Somani [4] focus on the optimal design of survivable light trail optical network. Two protection schemes, namely connection based protection and link based protection, that can achieve 100% protection against single link failure are proposed and compared. Connection based protection scheme is more practical for light trail architecture where the hop-length is limited due to power loss. Jing Fang, Wensheng He, Arun K. Somani [5] present light-trail architecture and its optimal design. A two-step approach for solving the light trail design

problem -the first step is called traffic matrix preprocessing; it divides single longhop paths into several shorter paths that satisfy the hop-length constraint. In the second step, the light trail design problem is formulated as an integer linear programming (ILP) optimization problem. Bin Wu and Kwan L. Yeung [6] introduce an efficient heuristic algorithm to solve static light-trail assignment problem. The problem is solved based on three key factors. These factors include the request discreteness, the shortest path length of each request, and the flow volume. The importances of the three factors are ranked in the same order as they are mentioned. This algorithm also adopts a reference node mechanism to address the request discreteness issues. Weiyi Zhan, Guoliang Xue, Jian Tang, Krishanaiyan Thulasiraman [7] consider the dynamic light-trail routing problem where the connection request come and leave dynamically. For each dynamically arriving connection request a light-trail is found to carry it with the objective of consuming a minimum number of free wavelengths. Protection issues in light-trail routing in WDM networks with dynamic traffic also has been considered. Akhil Lodha, Ashwin Gumaste, Paresh Bafna and Nasir Ghani [8] consider an optimization technique, the uncertain nature of traffic is called stochastic optimization. Stochastic optimization is based on a multistage model reducible to a two-stage model. The stages involve computation of an optimal solution followed by estimation and recourse taking future uncertainties into consideration. Arun K. Somani [9] discusses an important problem in optical networks with wavelength-division multiplexing is that of traffic grooming. Traffic grooming is a technique for multiplexing different sub wavelength capacity traffic requirements onto a single wavelength so that the wavelength and hence the capacity requirements of the whole network are minimized. WDM grooming networks can be classified into two categories: dedicatedwavelength grooming (DWG) networks and sharedwavelength grooming (SWG) networks. Dzmitry Kliazovich, Fabrizio Granelli, Hagen Woesner and Imrich Chlamtac [10] present a novel solution of bidirectional high speed communications for IP traffic transport over WDM networks. Bidirectional

All Rights Reserved 2012 IJARCSEE

ISSN:2277-9043 International Journal of Advanced Research in Computer Science and Electronics Engineering, Volume 1, Issue 1 March 2012

light-trail (BDLT) is an organization of two separate light-trails connecting a set of nodes in two directions (downlink and uplink) allowing bidirectional communication. The nodes of the BDLT are always aware of the bandwidth requirements of other nodes. Broadcasting and multicasting in BDLT architecture is performed by simultaneous transmission in both directions for any node of the light-trail.

2. Assignment constraint: Each request is assigned to one and only one light-trail. 3. Light Trail Capacity Constraint: The aggregated request capacity on a light trail should not exceed the full wavelength capacity. 4. Link-disjoint Light trail Constraint: In order to provide connection based protection, primary and backup connection request should be accepted and maintained as link-disjoined light trails. (iii) ALGORITHM A. LIGHT-TRAIL IMPLEMENTATION WITH CONNECTION MATRIX(CM) INPUT: Graph G(V,E); the set of existing light-trails LT from request matrix rqst_mtrx; connectivity matrix cntv_mtrx; the maximum hop-length of a light-trail Lmax. OUTPUT: The set of light-trails (final_lt) that cover the given requests having following properties: 1. The length of each light-trail is no longer than Lmax. 2. Total number of light-trails required is minimized. And the reduction of blocking is calculated. Step1: Checking of blocking 1. if ,any light-trail of rqst_mtrx is blocked, keep it in blocked_lt . else, keep the light-trail in nonblocked_lt. 2. Split each link of each light-trail of blocked_lt. 3. Check, whether these links are covered by the light-trails of nonblocked_lt either directly or via nodes. if not, the links are kept in blocked_links. Step2: Minimization of light-trails of nonblocked_lt with CM. 1. Determine the hop-length of each light-trail of nonblocked_lt. 2. Compare the hop-length of each light-trail with Lmax. (i) if hop-length < Lmax , select the corresponding light-trail. (ii) else if hop-length ==Lmax , add the corresponding light-trail to temp1. 3. Concatenate the light-trails selected from step2.2(i) with each other and store the result in temp2. 4. Each node of the light-trails of temp2 is checked. If, two consecutive nodes are same (same node number), consider them as one node (uniqueness). 5. Again count the hop-length of each light-trail of temp2. 6. Compare the hop-length of each light-trail of temp2 with Lmax. (i) if hop-length > Lmax, cancel out the corresponding light-trail.

III. PROPOSED HEURISTIC IN SURVIVAL OPTICAL NETWORK

(i) PROBLEM DEFINITION Given a graph (network) and a number of connection requests, the problem is to find a set of lighttrails to cover the given requests, such that the total number of light-trails required is minimized. Additionally the blocking problem of the network has been considered. Number of blocked links or connections has been reduced with minimized light-trails and the percentage of reduction of blocking is calculated here. (ii) PROBLEM DESCRIPTION Consider G(V,E) is a graph (unidirectional or bidirectional) which represents a topology, where V represents the set of n nodes and E represents the set of m links. The maximum length (in hop) of a light-trail is denoted by Lmax. The set of existing request for the network is taken in a request matrix which is denoted by rqst_mtrx and the set of connections among the nodes of the network is represented by another matrix denoted by connectivity matrix or shortly cntv_mtrx. Each request in rqst_mtrx is served by only one light-trail and one lighttrail is used to serve only one request. So, here the number of light-trail is equal to the number of existing request in rqst_mtrx. So, the set of light-trails denoted by LT is actually the set of existing requests in rqst_mtrx. ASSUPMTIONS: 1. Traffic-matrix Preprocessing: Normally due to power losses, there is a limit on the maximum hop length of a connection. The hop length is generally restricted to a maximum of 5. Thus if the number of hops in the shortest path between a pair of nodes exceeds this limit, then a connection request between them must be broken down into two or more connection requests. The algorithm that preprocesses the connection request is discussed in [5]. It was not implemented as the part of the work.

All Rights Reserved 2012 IJARCSEE

ISSN:2277-9043 International Journal of Advanced Research in Computer Science and Electronics Engineering, Volume 1, Issue 1 March 2012

(ii) else, the remaining light-trails are kept in temp3. 7. Check the connection of each light-trail of temp3 by cntv_mtrx. (i) if the connection of light-trail is not valid according to cntv_mtrx, cancel out the light-trails. (ii) else, keep it in temp4. 8. Check uniqueness of the light-trails of temp4. If, two or more light-trails are same consider them as one unique light-trail. 9. Selection of proper light-trails from temp4 and kept in temp5. (i) Check for the requests which are served by only one light-trail, select the light-trails first. (ii) Find out other requests which are also served by these selected light-trails and cancel out the other light-trails which serve only these requests. (iii) Now if any request is left that is served by two or more light-trails select any one from them. 10. Match each lighttrail of temp5 with each lighttrail of temp1. (i) if a match is found the corresponding lighttrail is cancel out from temp5. (ii) else keep the light-trail of temp5 in new_lt. 11. Add temp1 and new_lt and also blocked light-trails to final_lt. 12. Check whether all the requests in the rqst_mtrx are served by the light-trails of new_lt. If not, then the request is added to the final_lt. 13. Return final_lt. Step3: Reduction of blocking with minimized light-trail. 1. Check whether the links of blocked_links are covered by the light-trails of new_lt either directly or via nodes. if not, the links are kept in finally_blocked. 2. Calculate the percentage of reduction of blocking using the following equation. Consider, total number of links of blocked_links = bl total number of links of finally_blocked =fb So, reduction of blocking(%)=

Here, the size of the rqst _mtrx is 8x4. So, at the beginning the total number of light-trails needed to serve all the requests is 8. Now consider the first two light-trail of rqst_mtrx are blocked. So each link of these two light-trails is blocked.

The rest of the light-trails of rqst_mtrx are kept in nonblocked_lt. The links of blocked light-trails which are not covered by the light-trails of nonblocked_lt (either directly or via nodes) are considered as blocked_links. Here, [2 3] is covered by 2nd light-trail of nonblocked_lt.

(bl - fb) bl

X 100

Example Illustrating Proposed Heuristic with CM The unidirectional graph G, representing mesh topology is depicted in figure1. The request matrix denoted by rqst_mtrx containing all the requests and the connectivity matrix denoted by cntv_mtrx representing all the connections between the nodes of network G are given as: All Rights Reserved 2012 IJARCSEE

ISSN:2277-9043 International Journal of Advanced Research in Computer Science and Electronics Engineering, Volume 1, Issue 1 March 2012

In this example, the value of Lmax is considered 3. The first two light-trails (that served first four requests) having Lmax=3, will be a part of final result. But the rest four having hop-length <3, will be concatenated with each other. The steps are shown below:

Now [0 1] link of the blocked_links is covered by 1 light-trail of new_lt through node 3 and [1 2] link of the blocked_links is covered by 2nd light-trail of new_lt through node 5. finally_blocked will contain the links of blocked_links which is not covered by new_lt.
st

So, finally_blocked = 4 1 Now request 3-1-5 is only served by the light-trail 0-3-1-5. So this light-trail is considered first. 0-3 and 1-5 are also served by this light-trail. So, we dont need to consider the light-trail 1-5-0-3 any more. Now 2-4 can be served by both 1-5-2-4 and 2-4-1-5. So, any one of them can be selected. The minimized light-trails are kept in new_lt. Total number of links of blocked_links (bl)=3; Total number of links of finally_blocked (fb)=1. So, reduction of blocking(%)=((3 -1)/3)*100=66.67% Similarly another example is given for NSF network shown in figure2. Hence also the value of Lmax is 3.

new_lt =

0 3 1 5 1 5 2 4

So, in this example finally the total number lighttrails required to serve all the requests is reduced from 8 to 6. Fig 2.NSF network

Total node number: 14 Total bi-directional links: 21 Average nodal degree: 3 Average hop distance: 2.14286

The request matrix rqst_mtrx and the connectivity matrix cntv_mtrx of this network are given below:

All Rights Reserved 2012 IJARCSEE

ISSN:2277-9043 International Journal of Advanced Research in Computer Science and Electronics Engineering, Volume 1, Issue 1 March 2012

Now [5 7] link of the blocked_links is covered by 2nd and [8 9] link of the blocked_links is covered by 3rd light-trail of new_lt. 1 2 So, finally_blocked =

Total number of links of blocked_links (bl)=4; Total number of links of finally_blocked (fb)=2 So, reduction of blocking(%)=((4 -2)/4)*100=50%

B. LIGHT-TRAIL MINIMIZATION USING GENETIC ALGORITHM Step1 and step 3 are same as previous algorithm. So, they are not mentioned here. Only step 2 is implemented using GA. Step2: Minimization of light-trails of nonblocked_lt with Genetic Algorithm. (i) Set generation number t 0; Set maximum generation max_gen 10; Set string length xlen (Lmax +1); Set crossover probability pcross 0.98; Set mutation probability pmut 0.001; Set crossover site xsite 2; Now consider 2nd and 4th light-trail of rqst_mtrx is blocked. So each link of these two light-trails is blocked. Using the proposed heuristic the minimized light-trails are given below (the steps are not shown here): new_lt = (ii) Initial population: (a) Calculate the value of WL for each lighttrail in rqst_mtrx, where WL=

3-10-13-12 4-5-7-9 0-8-9-13

and

1, if there exists a unidirectional link from node i 0, otherwise

The total number of light-trails required to serve all the requests in the rqst_mtrx is reduced from 14 to 11.

(b) The light-trail which satisfy WL<Lmax, is considered as a string Si. (c) ini_pop {S1,S2,.,Spop_size} as initial population, where pop_size denotes total number of strings. (d) Keep the rest of light-trail of nonblocked_lt in temp1. (iii) Compute fit(Si) for each string Si (1 ipop_size) of initial population, where fitness function fit(Si) = | Lmax + WL| (iv) Selection or Reproduction: The selection operation described below on the string of ini_pop, generate a mating pool mate_pool of size pop_size. (a) Calculate the probability Pi of selection of Si (1ipop_size) Pi =

All Rights Reserved 2012 IJARCSEE

ISSN:2277-9043 International Journal of Advanced Research in Computer Science and Electronics Engineering, Volume 1, Issue 1 March 2012

(b) Calculate the expected number of copy E.C. (expected count) of Si (1ipop_size).

E.C=

where,

pop_size

(viii) Check, whether the requests, served by the light-trails of ini_pop are now all served by the light-trails of new_lt. if, yes, then (a) Light-trails of temp1 and new_lt are kept in final_lt. The blocked lighttrails of blocked_lt are also added to the final_lt. (b) Return final_lt. else if t< max_gen, (a) Set t t+1; (b) Rename new_pop as ini_pop; (c) Go to step(iii).

(c) Generate a random number rj from [0,1] for j=[1,2,,pop_size] and compute actual number of copies (actual count) of Si (1ipp_size). if rj > (1- fractional remainder of E.C.), then the additional copy is awarded to the ith string. (v) Crossover: The crossover operation is described below on the strings of mate_pool and obtains a population of temp_pop of size pop_size. (a) Randomly select two strings as pair from mate_pool such that the total number of each string is equal to the corresponding actual count and form (pop_size/2) number of paires. (b) Generate a random number rk from [0,1] for k=[1,2,,(pop_size/2)] for each pair such that, if rk pcross, the crossover will undergoes at the cut point xsite.

EXAMPLE ILLUSTRATING PROPOSED HEURISTIC WITH GA WL values of each light-trail of nonblocked_lt of 1st example (fig.1.) are shown in table1. The last four lighttrails having WL<Lmax (i.e WL<3 as Lmax=3) forms the initial population. Table2. computes fitness value of each string.

rq no. 1. 2. 3. 4. 5. 6.

rqst_mtrx 4 5 3 0 1 2 1 2 1 3 5 4 2 3 3 1 5 - - - -

W(Li) 3 3 2 1 1 1

Initial population

Table 1: Evaluating WL value of each light-trail r q n o. 1. 2. 3. 4. rqst_mtrx fit(Si)

(vi) Mutation: Mutation operation is described below on the string of temp_pop and obtains a population new_pop of size pop_size. Generate a random number rm from [0,1] for m=[1,2,..,pop_size X xlen] for each node nj (1jxlen) of each string Si (1ipop_size) of temp_pop such that , if rm pmut, exchange node nj of Si with any other randomly selected node nk of Si where 1kxlen and k j. (vii) Compute fit(Si) for each string Si (1ipop_size) of new_pop. if, fit(Si)=2Lmax, then Si is kept in new_lt.

3 0 1 2

1 3 5 4

5 - - - -

5 4 4 4

Table 2: Evaluating fitness value of each string


The considerations are: (i) Population size (pop_size) = 4; (ii) Maximum number of generation (max_gen) = 10; (iii) String length = 4; (iv) Crossover probability (pcross) = 0.98; (v) Mutation probability (pmut) =0.001; (vi) Crossover site (xsite) = 2.

All Rights Reserved 2012 IJARCSEE

ISSN:2277-9043 International Journal of Advanced Research in Computer Science and Electronics Engineering, Volume 1, Issue 1 March 2012

Selection procedure of individual string is referred in table3. Crossover and offspring generation is shown in table4. Here, fitness value of each offspring is evaluated. S tr in g n o. 1. 2. 3. 4. Initial popula tion fit( Si) Selectio n probabil ity Expec ted count Ran dom Ac tu al co un t 1 1 1 1

31503 -15 -24 --

5 4 4 4

0.295 0.235 0.235 0.235

1.18 0.94 0.94 0.94

0.67 0.28 0.55 0.98

Similarly this concept is also applied on NSF network (fig2) with same rqst_mtrx.

Sum

17

1.000 4.25 0.25

4.00 1.00

4 1
rq no. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. rqst_mtrx 0 3 5 13 9 1 4 12 9 8 3 6 11 11 6 1 13 4 8 9 12 4 0 W(Li) 3 3 3 3 3 3 2 1 1 1 1 1

Average

Table 3: Selection (or Reproduction) Parents (with xsite=2) - 3 1 5 0 3 - 1 5 - - - 2 4 New population (Offspring) 0 5 3 3 1 6 3 1 5 2 4 fit(Si) New_lt

6 3

0 3 1 5

4 3 3 10 4 5 7 9 13 12 0 8 9 13

Initial population

1 5 2 4 - - - -

Table 5: Evaluating WL value of each light-trail

Table 4: Generating new population Strings of new population having fitness value equal to 2Lmax (here Lmax=3, so 2Lmax=6) are considered as light-trails of new_lt. Now the light-trails of new_lt have served all the requests which were served by the light-trails of ini_pop. So, we dont need to go to the next generation. The total number of light-trails required to serve all the requests in the ini-pop is reduced from 4 to 2. So, total number of light-trail in rqst_mtrx is minimized from 8 to 6.

rq no. 1. 2. 3. 4. 5. 6.

rqst_mtrx 3 10 4 5 7 9 13 12 0 8 9 13 13 -

fit(Si) 5 4 4 4 4 4

Table 6: Evaluating fitness value of each string

All Rights Reserved 2012 IJARCSEE

ISSN:2277-9043 International Journal of Advanced Research in Computer Science and Electronics Engineering, Volume 1, Issue 1 March 2012

Str in g no . 1. 2. 3. 4. 5. 6.

Initial population

fi t( S i)

Sele ction prob abilit y

Expe cted coun t

Ran dom

3 10 13 4 5 7 9 13 12 0 8 9 13 -

5 4 4 4 4 4 0.2 0.16 0.16 0.16 0.16 0.16 1.2 0.96 0.96 0.96 0.96 0.96

0.67 0.28 0.55 0.98 0.47 0.71

Ac tu al co un t 1 1 1 1 1 1

IV.SIMULATION AND RESULT Table.9 summarizes the result. The 2nd and 3rd rows give the number of light-trails required by the Proposed Heuristic with CM and GA respectively. Though both generate same result, but there is no certainty that the later will always produce an optimal solution. Here reduction of blocking is also similar as both yield same minimized light-trails for a given set of requests for same blocked light-trails. (Fig.1.) N=6 (Fig.2.) N=14 (Unidirectional ) (Bidirectional)

Sum Average

25

1.0

6.00

6 1
CM GA

4.17 0.167 1.00

6 6

11 11

Table 7: Selection (or Reproduction) Parents (with xsite=2) 3 10 13 - - 13 12 4 5 - - - 7 9 0 8 - - - 9 13 New population (Offspring) 3 10 13 12 - 13 4 5 7 9 - - - fit(Si) New_lt Table.9 Comparison of CM with GA Complexity of the proposed heuristic is O(n.k), where n is the number of light-trails assigned to the network to serve the requests and k is the number of nodes between source and destination of each light-trail or request, whereas the complexity of second approach is slightly increases due to formulation of initial population. V. CONCLUSION A heuristic approach with connection matrix and Genetic Algorithm is proposed for minimizing the number of light trails for given a graph (network) and a number of connection requests. Here, mesh topology assuming it as unidirectional and NSF network which is bidirectional are considered to illustrate the proposed heuristic. Both the approaches can solve it in polynomial time. But the former is few steps ahead than the later in terms of complexity and also assures to always return an optimal solution. Above all, we have tried to develop an algorithm of light-trail optimization problem which is very efficient and user friendly such that it will always return an optimal or a sub-optimal solution with having reasonable lower complexity than other existing algorithms.

6 3 6 3 6 3

3 10 13 12

4 5 7 9

0 8 9 0 8 9 13 13 - - - Table 8: Generating new population Now the light-trails of new_lt have served all the requests which was served by the light-trails of initial population. The total number of light-trails required to serve all the requests in the ini-pop is reduced from 6 to 3. So, total number of light-trail in rqst_mtrx is minimized from 14 to 11.

All Rights Reserved 2012 IJARCSEE

ISSN:2277-9043 International Journal of Advanced Research in Computer Science and Electronics Engineering, Volume 1, Issue 1 March 2012

REFERENCES
[1] Ashwin Gumaste and Imrich Chlamtac, Light-trails: A NovelConceptual Framework for Conducting Optical Communications,Proceedings of HPSR 2003 Torino Italy, June 2003. Michael T. Frederick, Nathan A. VanderHom and Arun K. Somani, Light Trails: A Sub-Wavelength Solution for Optical Networking, HPSR2004,pp. 175-179. Ashwin Gumaste, Gabriel Kuper and Imrich Chlamtac, Optimizing Light-trail Assignment to WDM Networks for Dynamic IP Centric Traffic, IEEE LANMAN 2004, Apr. 2004,pp 113-118. Wensheng He, Jing Fang, Arun K. Somani, On Survivable Design in Light Trail Optical Networks, proceeding of 8th IFIP Working Conference on Optical Network Design and Modeling, Feb,2004. Jing Fang, Wensheng He, Arun K. Somani, Optimal Light Trail Design in WDM Optical Networks, IEEE Intl Conf on Commun (ICC) 2004 Paris.

[6]

Bin Wu nad Kwan L. Yeung, Light-trail assignment in WDM optical networks, IEEE Proc of Globecom 2006 San Francisco CA. Weiyi Zhan, Guoliang Xue, Jian Tang, Krishanaiyan Thulasiraman,Dynamic light-trail routing and protection issue in WDM optical network, IEEE Globecom 2005, vol. 4, pp 1963-1967. Akhil Lodha, Ashwin Gumaste, Paresh Bafna and Nasir Ghani,Stochastic Optimization of Light-trail WDM Ring Networks using Benders Decomposition, Indian Institute of Technology, Bombay, India,Tennessee Technological University Cookeville USA. Arun K. Somani, Survivability and Traffic Grooming in WDM Optical Networks, Cambridge University press. Dzmitry Kliazovich and Fabrizio Granelli, Bidirectional LightTrails for Synchronous Communications in WDM Networks, IEEE Globecom 2005.

[7]

[2]

[8]

[3]

[4]

[9]

[10]

[5]

Soumya Paul, Assoc. Professor and Head, Department of Computer Application in B. P. Poddar Institute of Management & Technology, Kolkata, has been in teaching and research for over 12 years. He holds a Masters Degree in Technology, Computer Application as well as in Mathematics and has gathered vast experiences in the same. He received his M.Sc. (Mathematics) from Visva Bharati University and stood 1st class 1st. He received MCA from National Institute of Technology, Rourkella and M. Tech (CSE) from AAI-Deemed University and pursuing Ph. D in Computer Science and Engineering. He served as a faculty member and visiting faculty member in various Institutes and Universities like RCCIIT, Visva Bharati University, University of Calcutta, Bardhaman University, West Bengal University of Technology etc. He has delivered numerous lectures across India in the field of his research interest, Optical Networks and Genetic Algorithms. He is an author/co-author of several published articles in International Journals and International Conferences. He has chaired an International Conference technically supported by IEEE communication. He has more than 17 research publications and currently Reviewer and Member, Editorial Board in many conferences and journals like International Journal of Data Modeling and Knowledge Management.

Department of Computer Application of B. P. Poddar Institute of Management & Technology, Kolkata, West Bengal, India. .Earlier, she held several technical positions in National Informatics Centre, Kolkata and Semaphore Computing Networks Pvt. Ltd. respectively. She has earned Masters Degree in Computer Application and currently pursuing her research in Computer Science and Engineering. She has more than 17 publications to her laurels and her research interest is specifically in the field of Optical Networking, Security and Genetic Algorithms. She has also been Member, Editorial Bord in several journal publications like International Journal of Software

Inadyuti Dutt, has been in the field of academics and research for more than ten years and is currently the Assistant Professor in the

All Rights Reserved 2012 IJARCSEE

10

You might also like