JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 31, NO.

2, JANUARY 15, 2013

257

Priority Based Dispersion-Reduced Wavelength Assignment for Optical Networks

Bijoy Chand Chatterjee, Nityananda Sarma, and Partha Pratim Sahu
AbstractDispersion in optical ber degrades the quality of signal in optical networks. Although the use of dispersion compensating ber (DCF) reduces the effects of dispersion, it is expensive and has more propagation loss compared to step-index ber (SIF). In this paper, we propose a priority based dispersion-reduced wavelength assignment (PDRWA) scheme to reduce overall dispersion in optical network. In this scheme, the connection requests having a same source-destination (s-d) pair are groomed rst to avoid intermediate optical-electrical-optical (O/E/O) conversation and then these groomed connection requests with longer lightpath are assigned the wavelengths having lesser dispersion and the wavelengths having a higher dispersion are assigned to the lightpaths with shorter distance. If the connection requests are assigned to the wavelengths using such constraint on dispersion, the overall dispersion in the network can be reduced to a great extent, which will in turn lead to better performance of the network in terms of overall signal quality (Q-factor) without increasing network setup cost. The performance analysis of the proposed scheme using SIF is conducted with different channel speeds (10, 40 and 100 Gbps) in terms of total dispersion and compared the same with the use of DCF. Furthermore, we have studied the overall Q-factor in the network with different channel speeds and considering polarization mode dispersion (PMD) effect using SIF and DCF. Index TermsOptical network, dispersion, PMD, grooming, RWA, Q-factor.

I. INTRODUCTION N recent years, to fulll the ever increasing demand of bandwidth, research interests have grown towards the high speed optical networks. The optical network consists of nodes linked with optical ber where dispersion creates the signal distortions during transmission. In a wide area optical network, dispersion of a signal increases with increase of ber length which results in degradation of system performance. Although, dispersion compensating devices like Dispersion Compensating Fiber (DCF), Optical Phase Conjugation, Pulse Prechirping and Duobinary Transmission are usually used to reduce dispersion, they are expensive. As an alternative solution, routing and wavelength assignment (RWA) scheme can incorporate mechanisms to reduce the overall dispersion in the network. In RWA, many conventional wavelength assignment approaches such as First-Fit (FF), Random Wavelength Assignment, Least-Used (LU), Most-Used (MU), Min-Product (MP), Least-Loaded (LL),
Manuscript received March 09, 2012; revised October 08, 2012; accepted November 13, 2012. Date of publication November 26, 2012; date of current version December 28, 2012. This work was supported by IETE Research Fellowship Scheme (No. IETE/J-282-10/BOR/2011) of The Institution of Electronics and Telecommunication Engineers. B. C. Chatterjee and N. Sarma are with the Department of Computer Science and Engineering, Tezpur University, Assam 784028, India (e-mail: bijoycc@tezu.ernet.in, nitya@tezu.ernet.in). P. P. Sahu is with the Department of Electronics and Communication Engineering, Tezpur University, Assam 784028, India (e-mail: pps@tezu.ernet.in). Digital Object Identier 10.1109/JLT.2012.2228628

MAX-SUM, and Relative Capacity Loss (RCL) have been reported in [1][3], [29], but none of these approaches take care of the reduction of overall dispersion in the network. In this direction, Zulkii et al. [4] has reported a dispersion optimized impairment constraint (DOIC) based RWA where DCF is used to optimize the dispersion. Use of DCF is not very effective because DCF is very expensive and its propagation loss is very high compared to step-index ber. On the other hand majority of the connection requests are typically in the range of Mbps where a single wavelength channel can support in Tbps range and hence a new opportunity has opened up as a form of trafc grooming. It is possible to incorporate trafc grooming [5], [6], [12] mechanism with RWA approach in which a number of low speed connection requests are multiplexed onto a high capacity wavelength channel to enhance overall channel utilization. Various studies [5], [6] on trafc grooming have been focusing on employing optical-electrical-optical (O/E/O) conversation at every node of the optical path to regenerate transmitted data signals, which lead to increasing cost of the optical network. To the best of our knowledge, no priority based study has been conducted using RWA approach with trafc grooming, which does not use dispersion compensating devices to improve the overall signal quality in the network. In this paper, we have proposed a priority based dispersionreduced wavelength assignment (PDRWA) scheme with incorporation of trafc grooming mechanism to reduce the total dispersion in the network and compared its performance with similar non dispersion-reduced wavelength assignment (NDRWA) scheme based on FF method. The rest of the paper is organized as follows. Section II formally denes the problem and the dispersion-reduced node architecture is presented in Section III. The proposed priority based dispersion-reduced wavelength assignment scheme with incorporation of trafc grooming mechanism is presented in Section IV. Section V evaluates the performance of the proposed scheme and nally, Section VI concludes the paper. II. PROBLEM STATEMENT We model the physical topology of an optical network as a , where and are directed connected graph the set of nodes, set of bi-directional optical ber links and set of wavelengths in each ber link respectively in the network. has a nite number of wavelengths. In the Here each link network, a non-negative cost (distance between adjacent nodes) is assigned to every . The link cost between two nodes is considered to be , if they are not directly connected. The following assumptions are considered in the model. Each ber link can carry an equal number of wavelengths and the network is without wavelength conversion capabilities. All the lightpaths using the same ber link must be allocated distinct wavelengths.

0733-8724/$31.00 2012 IEEE

258

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 31, NO. 2, JANUARY 15, 2013

TABLE I USED NOTATIONS AND SYMBOLS

The holding time of the connection requests is exponentially distributed and we also assume the same holding time for all the connection requests having the same s-d pairs for the sake of simplicity. However, differences in holding times for trafc requests can be handled by taking holding time as the maximum (similar to [13]) of holding times for all the trafc with the same s-d pairs. For the remainder of this paper, the used symbols and notations are summarized in Table I. Objective Function As dispersion increases with channel bit rate [7], we have formulated the total dispersion in the network with consideration of channel bit rate. Our objective in this paper is to minimize , total dispersion in the network. (1) In the above equation (1), the total dispersion of the network is computed for all the connection requests such as . For each connection request , we use a lightpath-link-indicator denoted by . III. DISPERSION-REDUCED NODE ARCHITECTURE First, we dene some basic terms that are used throughout the paper. The total dispersion [8] of a wavelength denoted by , is the combination of material and waveguide dispersions. Material dispersion denoted by , occurs because the refractive index varies as a function of the optical wavelength, whereas waveguide dispersion denoted by , occurs due to the wavelength dependence of the group velocity on the mode. The total dispersion in optical ber can be computed as (2) where

For the following three equations, the value of

is 1 and 2.

Each node can work as both an access and a routing node. A xed number of tunable transceivers are equipped with each node. Each node is capable of multiplexing/de-multiplexing as many connection requests (having a same source-destination (s-d) pair) within the channel capacity. All the channels have the same bandwidth. The connection requests arrive in the system randomly based on a Poisson process.

In order to reduce the total dispersion in the optical network, a dispersion-reduced node is being designed to switch the signals according to the PDRWA scheme (which is discussed in Section IV). Fig. 1 shows the logical architecture of the dispersion-reduced node which uses a number of devices such as number of wavelength division multiplexers (WDMs)/wavelength division de-multiplexers (WDDMs) [9], number of

CHATTERJEE et al.: PRIORITY BASED DISPERSION-REDUCED WAVELENGTH ASSIGNMENT FOR OPTICAL NETWORKS

259

Algorithm 1: Priority based dispersion-reduced wavelength assignment with trafc grooming (PDRWATG) Input : Network conguration and set of connection requests. Output : Wavelengths assignment and total dispersion of the network. Step 1: The connection requests with same s-d pair are groomed within channel capacity.

is the set of connection requests and indicates the bandwidth of a connection request from source to destination . represents the groomed connection request from source to destination . Step 2: Enqueue all the groomed connection requests in the priority queue. Step 3: Compute numbers of the shortest paths (including primary path) for each of the groomed connection requests on the basis of link state information and sort them in descending order of their primary path lengths. such that where GR is the ordered set of groomed connection requests and the priority order of groomed connection requests is assigned according to their positions in GR. indicates the length of the shortest lightpath of groomed connection . request Step 4: Arrange the wavelengths of each ber link in the increasing order of their dispersion, estimated using (2). is the ordered set of wavelengths and indicates the dispersion of the wavelength . Step 5: For each of the groomed connection requests in GR, selected based on their priority order, perform the following in the given sequence: a) First, try to assign a wavelength with less dispersion to the primary path. b) If no wavelength assignment is possible in step 5(a), consider the alternate paths in the ascending order of their lightpath distance for assigning a wavelength (with similar constraint on dispersion like in step 5(a)) until one alternate path is assigned a wavelength. c) If no wavelength assignment is possible either in step 5(a) or step 5(b) within , the groomed connection request is treated as blocked one. Otherwise, compute the dispersion (with the assigned wavelength) for the groomed connection request and add this dispersion to the total dispersion of the network. d) Drop the groomed connection request from the network. where

where

Fig. 1. Dispersion-reduced node architecture.

thermo-optic switches (TOSWs) [10], number of transceivers, number of add-drop multiplexers (ADMs) [11], SONET STS-192 multiplexer/SONET STS-192 de-multiplexer and wavelength router based on PDRWATG algorithm. The functionality of a dispersion-reduced node is explained below. In the gure, initially, a number of connection requests arrive at the system randomly based on a Poisson process. The connection requests (as per bandwidth) having a same s-d pair are then groomed/grouped with hierarchical time division multiplexer SONET STS-192 (as for example, if the connection requests of bandwidth 622.08 Mbps are groomed then a maximum 16 number of connection requests are accommodated with SONET STS-192). The groomed connection requests are assigned the wavelengths using PDRWATG algorithm. The signals of the assigned wavelengths are sent by using transmitters and added to TOSWs through ADMs. Then, the wavelengths are switched by TOSWs and nally are multiplexed by WDMs to the output ber link specied by wavelength router based on the PDRWATG algorithm to deliver to the destination node. Further, the wavelengths from input ber links are de-multiplexed by WDDM which are then switched by TOSWs and then multiplexed to the corresponding output ber link to deliver the signals to the destination node. The wavelength carrying the signal for the node itself is dropped through ADM and de-multiplexed by SONET STS-192 de-multiplexer for getting signals to the users. IV. PROPOSED PRIORITY BASED DISPERSION-REDUCED WAVELENGTH ASSIGNMENT WITH TRAFFIC GROOMING The aim of connection requests grooming is to enhance the effective utilization of a given capacity optical network, whereas the motivation of priority order estimation is to serve higher priority groomed connection request rst. In this paper, the connection requests having a same s-d pair are groomed rst to avoid intermediate optical-electrical-optical (O/E/O) conversation and then these groomed connection requests are served for routing and wavelength assignment according

260

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 31, NO. 2, JANUARY 15, 2013

to their priority order. The priority order of all the groomed connection requests is estimated according to descending order of their primary lightpath lengths. Using this criterion, longer lightpath groomed connection requests are always given higher priority compared to groomed connection requests having shorter lightpath distance. Our goal in this paper is to reduce the overall dispersion in the network without using dispersion compensating devices and hence the overall quality (Q-factor) of the signal in the optical network is improved without increasing network setup cost. To achieve our goal, the groomed connection requests of higher priority (longer lightpath) are assigned the wavelengths having lesser dispersion and the wavelengths having a higher dispersion are assigned to the groomed connection requests with shorter lightpath distance. Use of a conventional RWA approach may lead to a situation where the groomed connection requests with longer lightpath are assigned the wavelengths having a higher dispersion and the wavelengths having lesser dispersion are assigned to the groomed connection requests with shorter lightpath distance. As a result the overall dispersion in the network may increase which will degrade the signal quality. Therefore, if the priority order of connection requests is estimated using the above criterion and assign the wavelengths with such constraint on dispersion, the overall dispersion in the network can be reduced to a great extent, which will in turn lead to better performance of the network in terms of overall signal quality without increasing network setup cost. The details of RWA approach are numbers of paths are given in the Algorithm 1, where the computed dynamically on the basis of link state information for the entire session of a connection request. The computational complexity of the algorithm obtained as , where the 1st, 2nd, 3rd, and 4th terms of the complexity are used to groom number of connection requests, compute number of the shortest paths for all the groomed connection requests and sort them in descending order of their primary path lengths, arrange all the wavelengths in the network according to increasing order of their dispersion and perform wavelength assignment for number of groomed connection requests using alternate path respectively. The functionality of the proposed scheme is explained with a sample example network shown in Fig. 2(a). It consists of 6 nodes, 10 directed optical links and each link has three wavelengths (bandwidth 9953.28 Mbps) and . We assume the dispersions of such as and as 20, 18 and 19 in [ps/(nm km)] respectively. We also assume few randomly generated connection requests (maximum bandwidth 622.08 Mbps), . Applying Step 1 of the algorithm, connection requests are groomed within the channel capacity and enqueued them in a priority queue to estimate their priority order. According to Step 3 of the algorithm, the priority order of groomed connection requests are estimated and the sequence of ordered set, is formed. Then, wavelengths are arranged in the increasing order of their dispersion such that the dispersions of and are 18, 19 and 20 in [ps/(nm km)] respectively. Finally, groomed connection requests are established on the basis of

Fig. 2. (a) Physical topology of sample network and virtual topology of sample network using (b) NDRWA and (c) PDRWA schemes.

Fig. 3. Dependence of and on , obtained by using PDRWA for (a) Indian network and (b) NSFNET (10 Gbps channel speed).

priority order as per the constraint on dispersion given in Step 5(a) of the algorithm. For comparison purpose, the virtual topology of the sample network using NDRWA and PDRWA scheme are shown in Fig. 2(b) and (c) respectively. Therefore, the total dispersion in the sample network using PDRWA scheme is estimated as 91000 in [ps/nm] which is 3600 [ps/nm] less than that of using NDRWA scheme. V. PERFORMANCE ANALYSIS In this section, we have evaluated the performance of the proposed scheme using simulation. We set up the experiment for 14 nodes with 24 bi-directional physical links of Indian Network [12] and 14 nodes with 21 bi-directional physical links of NSFNET [1]. The following assumptions are made for simulation purpose. Measurements can be performed on materials contemplated for core 13.5 GeO :86.5 SiO compositions and for claddings Quenched SiO respectively [14]. Wavelength range is considered from 1.520 m to 1.590 m because of lower propagation loss in the optical ber link. The spacing between two wavelengths is 0.8 nm for 100 GHz frequency spacing (ITU G.694.1) [15]. The connection requests are generated randomly based on a Poisson process and the arrival time between the two successive requests follows an exponential distribution. We chose the Poisson model because the burstiness of trafc on the backbone is usually suppressed by the huge amount of aggregation of services and the actual trafc distribution remains unknown. The maximum bandwidth of a connection request is 622.08 Mbps (as per SONET STS-12 [16], [17]). As the quality of the signal (Q-factor) varies with channel speed, we consider wavelength channels having three different capacities [16] such as 10 Gbps (STS-192), 40 Gbps (STS-768) and 100 Gbps respectively. For these capacities, each wavelength channel can accommodate a maximum 16, 64 and 160 number of connection requests respectively belonging to the same s-d pair. The used system parameters are summarized in Table II. We have simulated the proposed scheme with connection requests varying from 1000 to 5000, distributed randomly

CHATTERJEE et al.: PRIORITY BASED DISPERSION-REDUCED WAVELENGTH ASSIGNMENT FOR OPTICAL NETWORKS

261

Fig. 4. versus speeds (inner graph,

, obtained by using PDRWA and NDRWA for (a) Indian network and (b) NSFNET under non-blocking condition with different channel versus , obtained by using PDRWA and NDRWA for (a) Indian network and (b) NSFNET with 10 Gbps channel speed).

TABLE II SYSTEM PARAMETERS AND THEIR VALUES IN THE MODELS

among all the possible s-d pairs and compared its performance with similar non dispersion-reduced wavelength assignment (NDRWA) scheme. In NDRWA, the connection requests having a same s-d pair are groomed rst and then groomed connection requests are assigned to wavelengths based on the FF method [3] and alternate path routing. In this scheme, we use the FF method because among other wavelength assignment approaches FF is better in terms of blocking probability and its computational complexity is also low. For our simulation study, we have estimated the dispersion of SIF using (2) and also incorporated the experimental results of dispersion of DCF developed by PureGuide [26]. It is observed that the dispersion using SIF increases with increase of wavelength but it is almost constant for DCF. Although, it is found that the dispersion of DCF is lower than SIF, the propagation loss of DCF is much higher than that of SIF. Therefore, we use SIF with our proposed PDRWA scheme and compare the same with DCF. Fig. 3(a) and (b) show the dependence of dispersion and the number of wavelengths on the number of alternate paths, obtained by using PDRWA for Indian network and NSFNET respectively (with 3000 numbers of connection requests and 10 Gbps channel speed). It is seen that in both networks, the number of wavelengths decreases with increase of the number of alternate paths. This is because more paths are used to establish connection requests successfully. It is also evident from both the gures that as the number of paths increases, the total dispersion also increases due to increase of path length. The number of wavelengths and total dispersion of the network and for NSFNET and Indian cross each other at network respectively. The value of Indian network is greater than NSFNET because bi-directional optical links of Indian network are more than that of NSFNET. Further, we also observe that in both networks the average setup time [6] increases

Fig. 5. versus for SIF and DCF with different channel speeds (10, 40 and 100 Gbps) of Indian network under non-blocking condition.

Fig. 6. BP versus , obtained by using PDRWA and NDRWA for (a) Indian network and (b) NSFNET with 10 Gbps channel speed.

with the number of alternate paths. Moreover, it is also seen (not shown in the gure) that for 40 and 100 Gbps wavelength channel speeds, values remain almost same. Therefore, for further study, we consider and 3 alternate path routing in NSFNET and INDIAN network respectively. A. Non-Blocking Case Fig. 4(a) and (b) show total dispersion versus number of connection requests, obtained by using PDRWA and NDRWA for Indian network and NSFNET respectively with different channel data speeds (10, 40 and 100 Gbps) under non-blocking condition where all the connection requests in the network are established successfully. It is seen that in both networks, total dispersion increases with channel bit rates but the rate of increase of total dispersion for PDRWA is less than that of NDRWA. This is because in PDRWA, the connection requests with longer lightpath are assigned to the wavelengths having lesser dispersion. We also observe that the total dispersion in NSFNET is more than that of Indian network. This is due

262

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 31, NO. 2, JANUARY 15, 2013

Fig. 7. Q-factor versus

, obtained by using PDRWA and NDRWA for (a) Indian network and (b) NSFNET (without PMD effect and 10 Gbps speed).

to the total path length during RWA for NSFNET is more than that of Indian network. Total ber loss versus number of connection requests (with 10 Gbps channel speed) are also shown for Indian network and NSFNET in Fig. 4(a) and (b) respectively. It is seen that in both networks, the variation of total ber loss with number of connection requests for PDRWA is almost close to that for NDRWA. So, in both networks for all channel speeds, the total dispersion signicantly reduces by using PDRWA without increasing total ber loss. Furthermore, it is reported [27] that for higher bit rate channel ( 10 Gbps) the pulse-broadening effect caused by PMD, degrades the signal quality. Therefore, we have studied PMD effect in our scheme with use of SIF and DCF. The PMD in a path with hops is expressed as [7], [19], [27] (3) Fig. 5 shows the total polarization mode dispersion (PMD) , for SIF and versus the number of connection requests DCF with different channel speeds (10, 40 and 100 Gbps) of Indian network. It is evident from the gure that as the number of connection requests increases, the PMD of the network increases for all channel capacities. The PMD using DCF is more than that of SIF for the channels with data speed greater than 10 Gbps, as the PMD coefcient of DCF is double than that of SIF [25]. It is also seen in the gure that the PMD in the network increases with increase of channel speed. It is also observed (not shown in the gure) that the PMD of NSFNET is greater than that of Indian network due to more path length in NSFNET and the same nature of curves of PMD is obtained in case of NSFNET. B. Blocking Case We have also studied the performance of PDRWA under blocking condition when some connection requests are blocked or rejected due to unavailability of wavelength. Fig. 6(a) and (b) show blocking probability versus total number of wavelengths, obtained by using PDRWA and NDRWA for Indian network and NSFNET respectively (with 3000 numbers of connection requests and 10 Gbps channel speed). It is seen that the blocking probability decreases with number of wavelengths in both networks but the decrease of blocking probability in PDRWA is almost close to that for NDRWA. We have also observed that the blocking probability of Indian network is less as compared to NSFNET. This is because alternate path routing is used to establish groomed connections in the Indian alternate path routing is network; whereas in NSFNET, used. It is also observed (not shown in the gure) that in both

Fig. 8. Q-factor (with PMD effect) versus , obtained by using PDRWA and NDRWA for Indian network with different channel capacities.

the networks, the total dispersion of PDRWA using SIF with 15% blocked connection is almost close to that of NDRWA in which DCF is used without blocking. C. Q-Factor Analysis As dispersion in optical ber degrades the quality of signal, it is also required to study the overall Q-factor of signals in the network which is dened as [18], [21], [28] (4) (5) where

Fig. 7(a) shows the overall Q-factor (without PMD effect and 10 Gbps channel speed) versus number of connection requests, estimated by using PDRWA and NDRWA for Indian network with use of step-index ber and dispersion compensating ber whereas Fig. 7(b) indicates the same for NSFNET. It is seen from the gures that in all cases of Indian networks and NSFNET, the Q-factor of the signal decreases with increase of number of connection requests due to increase of the total path length to accommodate these connection requests. In the gures, the Q-factor for NDRWA scheme with using SIF is lower than other schemes and using our proposed PDRWA scheme with SIF the Q-factor increases due to assignment of less dispersion wavelength to the connection requests having longer paths and the wavelengths having a higher dispersion are assigned

CHATTERJEE et al.: PRIORITY BASED DISPERSION-REDUCED WAVELENGTH ASSIGNMENT FOR OPTICAL NETWORKS

263

to the lightpaths with shorter distance. Further, it is observed that Q-factor in our scheme with DCF is almost close to that of our scheme with SIF. This is because more Amplied Spontaneous Emission (ASE) noise is introduced due to increase of the number of EDFA to compensate enhanced propagation losses with the use of DCF in the network. We have also compared our scheme with an existing NDRWA scheme using DCF. It is seen from the gures that Q-factor of our scheme with SIF is also more than that for NDRWA scheme using DCF. We have also found from Fig. 7(a) and (b) that in NSFNET the Q-factor is less than that of Indian network because the total path length in NSFNET is more than that of Indian network. Considering channel data speed greater than or equal 10 Gbps, we have studied the overall Q-factor (with PMD effect) versus number of connection requests estimated by using our PDRWA and the existing NDRWA schemes for Indian network with use of step-index ber and dispersion compensating ber, as shown in Fig. 8. It is seen in the gure that the overall Q-factor decreases with increase of channel speed because mainly, dual imaging of the bits at the receiver occurs due to pulse broadening of shorter bit signal. Although, the Q-factor without PMD effect (shown in Fig. 7(a)) of signal using PDRWA scheme with DCF is more than others but the overall Q-factor (with PMD effect) of signal using PDRWA scheme with SIF is more than that using DCF because of higher PMD coefcient of DCF. It is also evident from the gure that in PDRWA scheme with SIF, the rate of decrease of Q-factor with number of connection request is lower than that of higher channel speed, mainly due to mainly less power penalty in SIF. VI. CONCLUSION In this paper, we have proposed a priority based dispersionreduced wavelength assignment (PDRWA) scheme to reduce the overall dispersion in optical networks. The effectiveness of the proposed scheme is examined through its performance evaluation in two optical networks namely-Indian network and NSFNET. It is seen that although the dispersion with the use of DCF is lower than that of the SIF, the ber loss in DCF is 3 times higher than that of SIF. The results of our simulation studies show that using PDRWA scheme with SIF signicantly reduces the total dispersion in the network without increasing blocking probability and total propagation loss. It is observed that the total dispersion increases with channel speed but the rate of increase of total dispersion using PDRWA is less than that of NDRWA. It is also seen that Q-factor (with PMD effect) while using our scheme with SIF is more than that of using our scheme with DCF due to the higher PMD coefcient. Therefore, using our scheme with SIF will lead to a cost-effective solution with better quality of signal for high speed optical network design. REFERENCES
[1] B. Mukherjee, Optical WDM Networks. New York: Springer, 2006. [2] R. M. C. Siva and G. Mohan, WDM Optical Networks: Concepts, Design and Algorithms. New York: PHI, 2003. [3] H. Zang, J. P. Jue, and B. Mukherjee, A review of routing and wavelength assignment approaches for wavelength-routed optical WDM networks, Optical Networks Mag., vol. 1, pp. 4760, 2000.

[4] N. Zulkii, C. Okonkwo, and K. Guild, Dispersion optimised impairment constraint based routing and wavelength assignment algorithms for all-optical networks, in Proc. 2006 IEEE Int. Conf. Transparent Optical Networks, 2006, pp. 177180. [5] K. Zhu and B. Mukherjee, Trafc grooming in an optical WDM mesh network, IEEE J. Sel. Areas Commun., vol. 20, no. 1, pp. 122133, 2002. [6] P. P. Sahu, New trafc grooming approaches in optical networks under restricted shared protection, Photon. Netw. Commun., vol. 16, no. 3, pp. 233238, 2008. [7] A. Rahbar, Dynamic impairment-aware RWA in multiber wavelength-routed all-optical networks supporting class-based trafc, IEEE/OSA J. Opt. Commun. Netw., vol. 2, no. 11, pp. 915927, 2010. [8] G. Keiser, Optical Fiber Communications. New York: McGraw-Hill, 1991. [9] A. K. Das and P. P. Sahu, Compact integrated optical devices using high index contrast waveguides, in Proc. 2006 IEEE IFIP Int. Conf. Wireless and Optical Communications Networks, 2006. [10] P. P. Sahu and A. K. Das, Polarization-insensitive thermo-optic mach Zehnder device based on silicon oxinitride waveguide with fast response time, Fiber Integr. Opt., vol. 29, no. 1, pp. 1020, 2010. [11] P. P. Sahu, Polarization insensitive thermally tunable add/drop multiplexer using cascaded Mach-Zehnder coupler, Appl. Phys. B: Lasers Opt., vol. 92, no. 2, pp. 247252, 2008. [12] B. C. Chatterjee, N. Sarma, and P. P. Sahu, Priority based routing and wavelength assignment with trafc grooming for optical networks, IEEE/OSA J. Opt. Commun. Netw., vol. 4, no. 6, pp. 480489, 2012. [13] S. Du, S. Zhang, Y. Peng, and K. Long, Power-efcient rwa in dynamic WDM optical networks considering different connection holding times, Science China Information Sciences, pp. 19, 2012. [14] J. Fleming, Material dispersion in lightguide glasses, Electron. Lett., vol. 14, no. 11, pp. 326328, 1978. [15] G.694.1: Spectral Grids for WDM Applications: DWDM Frequency Grid. [Online]. Available: http://www.itu.int/rec/T-REC-G.694.1/ [16] SONET/SDH Technical Summary. [Online]. Available: http://www. techfest.com/networking/wan/sonet.htm [17] B. C. Chatterjee, N. Sarma, and P. P. Sahu, A heuristic priority based wavelength assignment scheme for optical networks, Optik-Int. J. Light Electron Opt., vol. 123, no. 17, pp. 15051510, 2012. [18] G. Agrawal, Nonlinear Fiber Optics. New York: Academic Press, 2001. [19] H. Pereira, D. Chaves, C. Bastos-Filho, and J. Martins-Filho, OSNR model to consider physical layer impairments in transparent optical networks, Photon. Netw. Commun., vol. 18, no. 2, pp. 137149, 2009. [20] D. Mahgerefteh and C. Menyuk, Effect of rst-order pmd compensation on the statistics of pulse broadening in a ber with randomly varying birefringence, IEEE Photon. Technol. Lett., vol. 11, no. 3, pp. 340342, 1999. [21] V. Alwayn, Optical network design and implementation, Cisco Systems, 2004. [22] L. Spiekman, J. Wiesenfeld, A. Gnauck, L. Garrett, G. Van Den Hoven, T. Van Dongen, M. Sander-Jochem, and J. Binsma, 8x10 Gb/s DWDM transmission over 240 km of standard ber using a cascade of semiconductor optical ampliers, IEEE Photon. Technol. Lett., vol. 12, no. 8, pp. 10821084, 2000. [23] S. Liaw, K. Huang, W. Chen, Y. Hsiao, and G. Lai, Investigate c+ l band EDFA/Raman ampliers by using the same pump lasers, in Proc. 9th Joint Conf. Information Sciences (JCIS2006), 2006, Paper PNC-11. [24] E. Ip and J. Kahn, Compensation of dispersion and nonlinear impairments using digital backpropagation, J. Lightw. Technol., vol. 26, no. 20, pp. 34163425, 2008. [25] Dispersion-Compensating Fiber: Precision and Repetition. [Online]. Available: http://documents.exfo.com/appnotes/anote122-ang.pdf [26] M. Nishimura, Optical bers and ber dispersion compensators for high-speed optical communication, J. Opt. Fiber Commun. Reports, pp. 115139, 2005. [27] J. Strand, A. Chiu, and R. Tkach, Issues for routing in the optical layer, IEEE Commun. Mag., vol. 39, no. 2, pp. 8187, 2001. [28] S. Pachnicke, T. Gravemann, M. Windmann, and E. Voges, Physically constrained routing in 10-Gb/s DWDM networks including ber nonlinearities and polarization effects, J. Lightw. Technol., vol. 24, no. 9, p. 3418, 2006. [29] B. C. Chatterjee, N. Sarma, and P. P. Sahu, Review and performance analysis on routing and wavelength assignment approaches for optical networks, IETE Tech. Rev., vol. 30, no. 01, 2013, accepted. Author biographies not included by author request due to space constraints.