Optimal Design for Shared Backup Path Protected Elastic Optical Networks Under Single-Link Failure

Transcription

1 Shen et al. VOL. 6, NO. 7/JULY 2014/J. OPT. COMMUN. NETW. 649 Optimal Design for Shared Backup Path Protected Elastic Optical Networks Under Single-Link Failure Gangxiang Shen, Yue Wei, and Sanjay K. Bose Abstract This paper considers the network protection technique of shared backup path protection (SBPP) in comparison with path protection for elastic optical networks. We develop integer linear programming (ILP) models to minimize both the required spare capacity and the maximum number of link frequency slots (FSs) used. We consider transponder tunability that corresponds to the condition of whether or not the same set of FSs is required to be used for both the working and protection lightpaths. We also apply the bandwidth squeezed restoration technique to obtain the maximum restoration levels for the affected service flows, subject to a limited FS capacity on each fiber link. Our studies show that the proposed SBPP technique requires much lower spare capacity compared to the traditional path protection approach. The flexibility of allowing the working and protection lightpaths to use different sets of FSs (i.e., with full transponder tunability) has the advantage of reducing both the number of FSs needed and the spare capacity redundancy required. Index Terms protection; BSR technique; Elastic optical network; ILP model; SBPP. I. INTRODUCTION F lexi-grid elastic optical networks are receiving wide attention due to their flexibility in bandwidth allocation and high efficiency of fiber spectrum utilization [1,2]. Most of the studies performed for the design and performance evaluation of this type of network [3 14] have focused on unprotected lightpath services. Only some [15 25] have dealt with the design of survivable elastic optical networks even though survivability would be of paramount importance in an optical transport network that carries a large amount of traffic. Of the many network protection techniques available, shared backup path protection (SBPP) is considered as one of the most promising ones due to its combined advantages of operational simplicity, speed, and efficiency [26]. SBPP is a failure-independent path-oriented technique in which the protection route is identified in advance and Manuscript received November 22, 2013; revised April 28, 2014; accepted May 23, 2014; published June 30, 2014 (Doc. ID ). G. Shen ( shengx@suda.edu.cn) and Y. Wei are with the School of Electronic and Information Engineering, Soochow University, Suzhou, Jiangsu Province , China. S. K. Bose is with the Indian Institute of Technology, Guwahati, India. the protection capacity is cross-connected on the protection route in real time. Its protection capacity efficiency is realized by protection capacity sharing on the common links of the protection lightpaths whose corresponding working lightpaths do not share any common link. Thus, compared to other network protection techniques, SBPP has two key advantages: (1) more efficient spare capacity sharing and (2) failure-independent network protection that does not need to identify the failure location and can therefore use much simpler network control and operation. Most prior works on SBPP have focused on SDH/SONET networks and, more recently, on dense wavelength division multiplexing (DWDM) optical networks [26]. This paper applies the SBPP technique to elastic optical networks [2]. A unique and challenging new feature of our study (in comparison with the existing SBPP studies) is that it also includes special constraints relevant for elastic optical networks, i.e., (a) frequency slot neighboring (or spectrum contiguity) in the spectrum domain for each lightpath connection and (b) spectrum continuity along a lightpath route. Specifically, the constraint of spectrum contiguity requires all the frequency slots (FSs) that make up an optical channel to be contiguous, and the constraint of spectrum continuity requires the assigned contiguous spectra on all the fiber links traversed by a lightpath to be the same when none of nodes in the network is capable of spectrum conversion. This study focuses on the full recovery of any single-link failure because the probability of simultaneously occurring multi-link failures is much lower compared to a single-link failure. We develop an integer linear programming (ILP) model for the SBPP technique. Moreover, in order to evaluate the capacity efficiency of the SBPP technique, we compare its performance with that of the 1 1 path protection technique in terms of both protection capacity and link spectrum usage in the network. We also apply the bandwidth squeezed restoration (BSR) [22,23] capability to the design of SBPP and 1 1 path protection techniques for the elastic optical network and evaluate how the bandwidth squeezing ratio can affect the network spectrum capacity required. To the best of our knowledge, this is the first study that is dedicated to a comprehensive design of the SBPP and 1 1 path protection techniques for an elastic optical network. The rest of this paper is organized as follows. In Section II, we introduce the SBPP and 1 1 path /14/ $15.00/ Optical Society of America

2 650 J. OPT. COMMUN. NETW./VOL. 6, NO. 7/JULY 2014 Shen et al. protection techniques based on elastic optical networks. In Section III, we present the ILP models for optimal designs of SBPP and 1 1 techniques based on elastic optical networks. In Section IV, we show the results obtained and compare the spare capacity efficiency for the SBPP and 1 1 path protection techniques. Section V concludes the paper. II. SHARED BACKUP PATH PROTECTION AND 1+1PATH PROTECTION This section introduces the concept of SBPP in contrast to the 1 1 path protection technique in the context of elastic optical networks. Considering the tunability of optical transponders at the source and destination nodes of each lightpath [27], we introduce the concept of the SBPP technique based on two cases, i.e., (a) working and backup paths must use the same set of contiguous FSs, and (b) they can use different sets of contiguous FSs. The current design is general to consider only the assigned number of FSs, each of which can be used to carry different modulation formats and baud rates. As shown in Figs. 1(a) and 1(b), when a link fails, both cases find a replacement path directly between the end nodes of the path. For example, if link (6-8) fails, working path ( ) is affected and a switch-over is performed to direct the affected working traffic flow onto a predefined protection route ( ) for failure recovery. Similarly, if link (0-1) fails, working path (0-1-4) is affected and a switch-over is performed onto a predefined route (0-2-4) for failure recovery. We can see that the two cases are similar in terms of finding backup paths and the switchover between the working and backup paths. However, they differ in the way that the spectrum is assigned on the backup path. Specifically, if the transponder is fully tunable as shown in Fig. 1(a) (i.e., it can be tuned from any one spectrum segment to another spectrum segment), then there is no constraint on the selected starting index of FSs on the backup path. Thus, this case can allocate any spectrum segment on the fiber link subject only to the conditions that the number of assigned FSs is sufficient to recover the affected working flows and that all the FSs are contiguous. In particular, on protection route ( ), FSs whose indices range from 3 to 6 are assigned, and on protection route (0-2-4), FSs whose indices range from 2 to 4 are assigned, which are both different from the FSs of their corresponding working lightpaths. In contrast, for the case in which the transponder is not tunable as shown in Fig. 1(b), (i.e., the working and backup lightpaths must use the same set of contiguous FSs), not only should the assigned FSs on the backup path be contiguous and sufficient to recover the affected working flows, but we must also guarantee that the starting index of FSs of the backup path is the same as that of the working path. These two cases will also arise under the 1 1 protection technique (i.e., working and backup paths use the same or different sets of contiguous FSs) to enable 100% failure recovery. An important feature of SBPP is its ability to share protection capacity on the common links traversed by multiple protection paths, if their corresponding working paths do not share any common link. For example, in Figs. 1(a) and 1(b), links (0-2) and (2-4) are both traversed by the two backup paths ( ) and (0-2-4) whose corresponding working paths ( ) and (0-1-4) do not share any common link. Therefore, these two backup paths can share protection capacity on their links to reserve only five FSs, and that will be sufficient to guarantee full recovery as long as the links (0-1) and (6-8) do not fail simultaneously. In contrast, the 1 1 protection technique cannot share spare capacity on backup paths. If we consider the same examples in Figs. 1(a) and 1(b), seven FSs should be reserved by the 1 1 protection technique on the links (0-2) and (2-4) commonly traversed by the two protection paths. III. ILP DESIGN MODELS FOR SBPP AND PATH PROTECTION Fig. 1. SBPP examples with/without tunable transponders in the elastic optical network. (a) SBPP with fully tunable transponder. (b) SBPP with nontunable transponder. In this section, we present the ILP design models for both the SBPP and the 1 1 path protection techniques. We first introduce the models for SBPP, which is followed by the models for 1 1 protection. For each of the protection techniques, we will consider both of the cases with or without tunable transponders (i.e., whether or not the working and protection paths must use the same set of contiguous FSs). Given a set of lightpath traffic demands with each requiring a predefined number of FSs, we minimize both the required protection capacity and link spectrum usage in the

3 Shen et al. VOL. 6, NO. 7/JULY 2014/J. OPT. COMMUN. NETW. 651 network. We assume that the elastic optical networks are operated under the flexi-grid mode [2]; this assumes that each fiber spectrum is divided into many FSs with a constant small granularity. We also assume that between each pair of nodes there is only one single shortest route employed to establish the working lightpath, but there may be several routes (link-disjoint from the working route) that can be selected to establish the protection lightpath. A. Model for SBPP In this part, we first introduce the model for SBPP with tunable transponders, where we allow the backup path to use a different set of FSs from the working path. The sets and parameters of the model are as follows. 1) Sets: S: Set of network links. R: Set of node pairs in the network. B r : Set of backup routes between node pair r. We assume that there is only a single fixed working path and that all the backup paths in the set are link-disjoint from the working path. 2) Parameters: η r i : A binary parameter that equals 1 if the working path of node pair r is affected when link i fails; 0, otherwise. ζ r;b j : A binary parameter that equals 1 when the bth eligible backup route for node pair r crosses link j; 0, otherwise. ε t r: A binary parameter that equals 1 when the working path of node pair r and the working path of node pair t share a common link; 0, otherwise. γ b;t r : A binary parameter that equals 1 when the working path of node pair r and backup path b of δ b;t a;r: node pair t share a common link; 0, otherwise. A binary parameter that equals 1 when backup path a of node pair r and backup path b of node pair t share a common link and their corresponding working paths also share a common link; 0, otherwise. d r : Working demand units (in FS) between node pair r. : A large value. α: A weight factor. 3) Variables: x t r: A binary variable that equals 1 when the starting FS index of the working lightpath between node pair r is larger than that of the working lightpath between node pair t, i.e., f r >f t ; 0, otherwise. r : A binary variable that equals 1 when the starting FS index of the working lightpath between node pair r is larger than that of protection lightpath b y b; t z b;t a;r : between node pair t, i.e., f r >e t b ; 0, otherwise. A binary variable that equals 1 when the starting FS index of protection lightpath a between node pair r is larger than that of protection lightpath b between node pair t, i.e., e r a >e t b ; 0, otherwise. s j : The total number of spare capacity units (in FS) that should be reserved on link j. c: The maximal index of used FSs. 4) Objective: 5) Constraints: Minimize X j S s j α c: (1) c f r d r r R; (2) c e r b dr r R; b B r ; (3) X b B r ρ b;r 1 r R; (4) e r b ρ b;r r R; b B r ; (5) X η r i ρ b;r ζ r;b j d r s j i; j S; i j; (6) r R;b B r f t f r 1 x t r 1 ε t r 1 r; t R; r t; (7) f r d r f t x t r 1 ε t r r; t R; r t; (8) e t b f r 1 y b;t r 2 ρ b;t γ b;t r 1 b B t ; r; t R; r t; (9) f r d r e t b yb;t r 2 ρ b;t γ b;t r b B t ; r; t R; r t; (10) e t b dt f r 1 y b;t r 2 ρ b;t γ b;t r b B t ; r; t R; r t; (11) ρ b;r : f r : e r b : A binary variable that equals 1 if the bth eligible backup route of node pair r is chosen; 0, otherwise. An integer variable denoting the assigned starting FS index of the working lightpath between node pair r. An integer variable denoting the assigned starting FS index of backup path b of node pair r. e t b er a 1 z b;t a;r 3 ρ b;t ρ a;r δ b;t a;r 1 a B r ; b B t ; r; t R; r t; (12) e r a d r e t b zb;t a;r 3 ρ b;t ρ a;r δ b;t a;r a B r ; b B t ; r; t R; r t: (13)

4 652 J. OPT. COMMUN. NETW./VOL. 6, NO. 7/JULY 2014 Shen et al. Objective (1) is to minimize the total required spare capacity in units of FS and the maximal index of FSs used in the network. We use both of the criteria to measure the spectrum efficiency of the network, and it is generally more efficient if fewer FSs are required to accommodate all the lightpath traffic demands. We set α to be a small value (i.e., 0.01) to ensure that minimizing the total spare capacity is the first priority. Constraints (2) and (3) ensure that the maximal FS index in the whole network should always be greater than the ending FS index of the lightpath between any node pair. Constraint (4) ensures that there is only one backup route selected for any node pair. Note that this constraint is important as all the restored FSs must travel together along the same route and thus only one restoration route can be employed to recover the affected working flow. This is also true in the traditional model of SBPP. Constraint (5) says that a backup path between a node pair can be assigned with a starting FS index only if the backup path is selected in constraint (4). Constraint (6) counts the total number of spare capacity units in FS that should be reserved on link j so as to ensure successful establishment of all restoration paths. Constraints (7) and (8) ensure that the allocated spectra for the working lightpaths between different node pairs do not overlap on any common link. Specifically, if the starting FS index of working lightpath A is larger than that of working lightpath B, the starting FS index of lightpath A should also be larger than the ending FS index of lightpath B. Figure 2 shows the situation in which the working lightpaths between node pairs r and t overlap on a common link. Constraints (7) and (8) ensure that the allocated spectra on the two working lightpaths do not overlap on the common link. While constraints (7) and (8) ensure that the allocated spectra for the working paths to different node pairs should not overlap on a common link, constraints (9) (11) ensure that the working path of a node pair does not overlap the backup paths of other node pairs in the spectra. Similarly, constraints (12) and (13) ensure that the allocated spectra for the backup paths to different node pairs should not overlap on a common link. Constraints (9) (11) are a version of constraints (7) and (8) for a pair of working and backup lightpaths that share common link(s). For any single-link failure, if the selected backup path for recovery of a working lightpath shares common link(s) with another working lightpath that is not affected by the link failure, then the backup path Fig. 3. Node pairs whose working path and backup path share common span(s). and the second working lightpath should not overlap in their assigned spectra. Figure 3 shows the situation in which when a link fails, the selected backup path b for recovery of the working lightpath between node pair t shares a common span with another working lightpath between node pair r that, however, does not cross the failed link. Constraints (9) (11) ensure that the backup path b and the second working lightpath between node pair r do not overlap in their assigned spectra. Constraints (12) and (13) are modified versions of constraints (9) (11) for any pair of restoration paths that share common link(s) and whose corresponding working lightpaths also share common span(s) as in Fig. 4. These constraints ensure that the pair of restoration paths does not overlap in their spectra. Here, we do not consider the situation in which a pair of backup paths shares common link(s) and whose corresponding working lightpaths are mutually disjoint. This is because the SBPP strategy can share spare capacity on the common link(s) crossed by the corresponding backup paths in this situation. Note that this is different from the 1 1 path protection technique, which does not have this feature of sharing spare capacity. The ILP model for SBPP without tunable transponders (which would require the working and backup paths to use the same set of FSs) can be extended from the earlier scheme without incorporating spectral limitation. For this model, we have the same model sets as those for the one allowing the two paths to use different sets of FSs. We need to replace constraint (5) with a new constraint (i.e., e r b f r, r R, b B r ). This constraint ensures that the working and protection paths have the same starting FS index. In addition, parameter δ b;t a;r and variable z b;t a;r are not required, and constraints (12) and (13) become redundant. The reason for this is that when the situation in which a pair of backup paths shares common link(s) and meanwhile their Node pair r Node pair t Backup path a Common link Backup path b Fig. 2. Node pairs whose working paths share common link(s). Fig. 4. Two backup paths of the two working lightpaths that are affected by a common span failure.

5 Shen et al. VOL. 6, NO. 7/JULY 2014/J. OPT. COMMUN. NETW. 653 corresponding working paths share common spans occurs, constraints (7) and (8) make constraints (12) and (13) redundant under the requirement that the working and backup paths must use the same set of contiguous FSs. B. Model for Path Protection The ILP models for 1 1 path protection can be obtained by extending the models of SBPP. For the model of 1 1 path protection that allows the working and backup paths to use different sets of contiguous FSs, we have the same model sets as those of the SBPP case. However, the parameter η r i is not required for the model of 1 1 path protection. In addition, since spare capacity sharing is not allowed under the 1 1 path protection, constraint (6) should be replaced with the new constraint (6 ) given below: X ρ b;r ζ r;b j d r s j j S: 6 0 r R;b B r In addition, we need to redefine the parameter δ b;t a;r as follows. δ b;t a;r: A binary parameter that equals 1 when backup path a of node pair r and backup path b of node pair t share a common link; 0, otherwise. Because of the dedicated protection in the 1 1 path protection scheme, if any pair of backup paths shares common link(s), δ b;t a;r equals 1. Constraints (12) and (13) ensure that the pair of backup paths does not overlap in their spectra at any time and would also include the case in which the corresponding working lightpaths are mutually disjoint, as shown in Fig. 5, under which the spectra of the backup paths can be shared under SBPP. This illustrates the most important difference between SBPP and 1 1 path protection; i.e., spare capacity sharing is not allowed under the 1 1 path protection technique. The ILP model for 1 1 protection that requires the working and backup paths to use the same set of contiguous FSs may be obtained by further extending the 1 1 path protection model that allows the two paths to use different sets of contiguous FSs. The same sets and parameters are used as before. The differences between the constraints are as follows. The constraint dimension of constraints (12) and (13) is changed. Due to the dedicated protection of the backup path in the 1 1 path protection technique, constraints (12) and (13) had ensured that the pair of backup paths does not overlap in their spectra regardless of whether the corresponding working lightpaths are mutually disjoint or if they intersect. In addition, for the case of 1 1 path protection using a backup path without tunable transponders that requires the working and protection lightpaths to use the same set of FSs, we also need to replace constraint (5) with the new constraint (i.e., e r b f r, r R, b B r ). C. SBPP and Path Protection With Bandwidth Squeezed Restoration We have introduced the ILP models for 100% failure recovery for different protection cases. In an elastic optical network, when sufficient bandwidth is not available for complete restoration, we may partially restore the affected service flows. This technique is referred to as BSR [22,23]. The BSR technique would be especially useful in a network that does not have sufficient protection capacity for full restoration. Figure 6 shows an example of BSR. When span (3-6) fails, working path ( ) is affected. Without BSR, eight FSs are required on the protection path for failure recovery. However, by applying BSR, we can partially recover from the failure, e.g., by providing only four FSs along the backup path shown in Fig. 6 even if full restoration bandwidth was not available on that path. To plan for the SBPP and 1 1 path protection techniques with BSR, we extend the previous optimization models for the two protection techniques. To implement BSR, we define β 0 β 1 as the ratio of bandwidth squeezing. This would imply that for all the service flows we need to ensure the recovery of at least a fraction β of the original bandwidth. Specifically, β 0 corresponds to the case of no restoration, while β 1 corresponds to the case of full restoration. For the example in Fig. 6, β is 0.5 as four FSs can only recover 50% of the original eight-fs working capacity. The model objectives under BSR is the same as objective (1), i.e., to minimize both the total spare capacity required in units of FS and the maximal FS index used in the whole network. Some modifications to the constraints for each of the cases are required as follows. Node pair r Backup path a Common link Backup path b Node pair t Fig. 5. Two backup paths whose corresponding working lightpaths are mutually disjoint share a common link. Fig. 6. Path protection with bandwidth squeezing.

6 654 J. OPT. COMMUN. NETW./VOL. 6, NO. 7/JULY 2014 Shen et al. For the case of SBPP with tunable transponders that allows the backup path to use a different set of FSs from the working path, we replace constraints (2), (3), (6), (8), (10), (11), and (13) with new constraints (2 ), (3 ), (6 ), (8 ), (10 ), (11 ), and (13 ) as follows. These constraints ensure that only d r β units of traffic demand are restored: X c f r d r β r R; 2 0 c e r b dr β r R; b B r ; 3 0 r R;b B r η r i ρ b;r ζ r;b j d r β s j i; j S; i j; 6 00 f r d r β f t x t r 1 ε t r r; t R; r t; 8 0 f r d r β e t b yb;t r 2 ρ b;t γ b;t r b B t ; r; t R; r t; (10 ) D. Computational Complexities of the ILP Models The ILP models for the two techniques have different computational complexities. We count the dominant number of variables and constraints to evaluate their complexities. For the model of the SBPP technique with tunable transponders, the dominant number of variables is, and the dominant number of constraints is, where j Bj is the average number of link-disjoint shortest routes between each pair of nodes, and jrj is the number of node pairs in the whole network. The model without tunable transponders has smaller dominant numbers of variables and constraints than those of the model with tunable transponders because the former lacks constraints (12) and (13). Specifically, the dominant numbers of variables and constraints in the model without tunable transponders are both O j Bj jrj 2. For both models of the 1 1 path protection technique, they have the same dominant numbers of variables and constraints, i.e., both. The computational complexities in terms of dominant numbers of variables and constraints for all four ILP models are summarized in Table I. e t b dt β f r 1 y b;t r 2 ρ b;t γ b;t r b B t ; r; t R; r t; (11 ) e r a d r β e t b zb;t a;r 3 ρ b;t ρ a;r δ b;t a;r a B r ; b B t ; r; t R; r t: (13 ) For the case of SBPP without tunable transponders, because constraints (12) and (13) are redundant, we just need to replace constraints (2), (3), (6), (8), (10), and (11) with the new constraints (2 ), (3 ), (6 ), (8 ), (10 ), and (11 ) in the original model. For the 1 1 protection case, in addition to (2 ), (3 ), (8 ), (10 ), and (11 ), we need to replace constraint (6 ) with the new constraint (6 ) given below: X ρ b;r ζ r;b j d r β s j j S: r R;b B r In contrast to minimizing the total spare capacity required in units of FS subject to the condition that all the lightpath demands are recovered, under BSR we may also plan to maximize service restorability (corresponding to bandwidth squeezing ratio β) given a limited spectrum capacity on each fiber link (i.e., a limited number of FSs, c). In this design scenario, β becomes a variable and the maximal index of used link FSs c changes to a given parameter. The optimization models have the same sets of constraints as those for the BSR capacity minimizing designs, while maximizing the bandwidth squeezing ratio (or restoration level) β becomes the objective to replace objective (1) for all four protection cases. IV. RESULTS AND PERFORMANCE ANALYSES To evaluate the performance of the SBPP and 1 1 path protection schemes in elastic optical networks, we consider three test networks: (a) a 6-node, 8-link network (n6s8, average nodal degree 2.7), (b) the 11-node, 26-link COST239 network (average nodal degree 4.7), and (c) the 14-node, 21-link NSFNET network (average nodal degree 3.0). 1 The three test networks are shown in Fig. 7. For all the networks, the traffic demand on each node pair is random with a uniform distribution within a certain range. We set a maximum number of FSs, X, and each node pair can choose any number (between 1 and X) of FSs. In this study, we set X to be 3, 4, and 5, respectively. In addition, we employed the K-disjoint shortest path algorithm to find all eligible protection routes for each node pair. All these routes are link-disjoint from the shortest working route. We used the commercial software AMPL/Gurobi [28] to solve all the ILP models; the version of Gurobi is 5.5. The MIPGAPs of all the optimization models were set to be smaller than 1%. For the different cases, the computation times for the optimization models were different. The running time of the case without tunable transponders is always longer than that with tunable transponders for each of the protection techniques under the three test networks. In addition, the running time of the 1 1 path protection technique is always shorter than that of the 1 For all the test networks, we can obtain the optimal results. However, if the network size becomes larger, then the optimization models become intractable since the planning problem of SBPP itself is NP-complete. For a larger network, we need to refer to an efficient heuristic algorithm.

7 Shen et al. VOL. 6, NO. 7/JULY 2014/J. OPT. COMMUN. NETW. 655 TABLE I COMPUTATIONAL COMPLEXITIES OF THE FOUR ILP MODELS Computational Complexities Models SBPP with tunable transponder SBPP without tunable transponder with tunable transponder without tunable transponder Dominant Number of Variables O j Bj jrj 2 SBPP technique. Specifically, the running time for both cases of the 1 1 path protection technique is no more than 4000 s, and that of the SBPP technique is no more than 6500 s. A. Spare Capacity Efficiency Dominant Number of Constraints O j Bj jrj 2 We evaluated the performance of the SBPP and 1 1 path protection techniques based on elastic optical networks in terms of their spare capacity redundancy and maximal FS index used. The two cases, with and without Fig. 8. Spare capacity redundancies of different test networks. tunable transponders, were separately considered for each of the protection techniques. Figure 8 show the results of spare capacity redundancy (which is defined as the ratio of the total protection capacity to the total working capacity in the whole network) for SBPP and 1 1 path protection. Note that we did not set a limit on the maximal number of FSs on each fiber link, and minimizing the total spare capacity is considered as the first optimization objective. Therefore, the two cases (i.e., with/without tunable transponders) of each protection technique show the same spare capacity redundancy, and each result point on the curves represents the spare capacity redundancies for both cases. For all the test networks, we find that SBPP shows lower spare capacity redundancies than the 1 1 protection technique. This is reasonable since SBPP allows protection capacity sharing among multiple protection lightpaths whose corresponding Spare capacity redundancy Spare capacity redundancy n6s8 NSFNET COST Maximal number of demand units per node pair, X (a) NSFNET n6s8 COST Maximal number of demand units per node pair, X (b) Fig. 7. Three test networks: (a) 6-node, 8-link (n6s8) network. (b) 11-node, 26-link COST239 network. (c) 14-node, 21-link NSFNET network. Fig. 9. Spare capacity redundancies of the SBPP and 1 1 path protection techniques with different numbers of demand units per node pair. (a) SBPP. (b) 1 1 path protection.

8 656 J. OPT. COMMUN. NETW./VOL. 6, NO. 7/JULY 2014 Shen et al. working lightpaths do not share any common link, while 1 1 does not. In addition, comparing the spare capacity redundancies of the three test networks under the SBPP protection technique as shown in Fig. 9(a), we find that the redundancy of the COST239 network is the lowest among the three networks. This is because the COST239 network has the highest average nodal degree, which increases spare capacity sharing opportunities and therefore helps reduce spare capacity redundancy. For the 1 1 technique, comparing the spare capacity redundancies of the three test networks as shown in Fig. 9(b), we find that the redundancy of the COST239 network is the lowest and that of the NSFNET is the highest, while the n6s8 network lies in between. This is attributed to the fact that a denser network provides Fig. 10. Maximal numbers of FSs required by the SBPP and 1 1 path protection techniques with different maximal numbers of demand units per node pair. (a) n6s8. (b) COST239. (c) NSFNET. shorter protection routes, which therefore helps reduce the sum of link protection capacity and consequently spare capacity redundancy. Because the average hop-length of the lightpaths in the COST239 network is the shortest and in the NSFNET network is the longest, we see that the COST239 and NSFNET networks show the lowest and highest spare capacity redundancies, respectively. B. Maximal Number of Frequency Slots Used Figures 10(a) 10(c) show the maximal number of FSs required for accommodating all the traffic demands under the SBPP and 1 1 path protection techniques. We can see that for all the test networks and for both of the cases (with and without tunable transponders) SBPP always requires a smaller number of FSs than the 1 1 path protection technique. This is attributed to the fact that SBPP allows Maximum number of used FSs Maximum number of used FSs Maximum number of used FSs _Non-tunable SBPP_Non-tunable 1+1_Fully-tunable SBPP_Fully-tunable 30% 50% 70% 100% Bandwidth squeezing ratio (a) 1+1_Non-tunable SBPP_Non-tunable 1+1_Fully-tunable SBPP_Fully-tunable 30% 50% 70% 100% Bandwidth squeezing ratio (b) 1+1_Non-tunable SBPP_Non-tunable 1+1_Fully-tunable SBPP_Fully-tunable 30% 50% 70% 100% Bandwidth squeezing ratio (c) Fig. 11. Maximal numbers of FSs required under different bandwidth squeezing ratios. (a) n6s8. (b) COST239. (c) NSFNET.

9 Shen et al. VOL. 6, NO. 7/JULY 2014/J. OPT. COMMUN. NETW. 657 protection capacity sharing among multiple protection lightpaths, while the 1 1 path protection technique cannot do that. We also observe that for each protection technique, the case with tunable transponders requires a smaller number of FSs than that without tunable transponders. This is because the tunable transponders allow the working and protection lightpaths to use different sets of FSs, which provides more flexibility in spectrum allocation and therefore improves the efficiency of spectrum usage. C. Impact of BSR In this section, we evaluate how the BSR technique can impact the design of SBPP and 1 1 path protection techniques based on elastic optical networks. We also consider the two cases, with and without tunable transponders, for each of the protection techniques. All the results were obtained under the assumption that the traffic demand intensive X is 5. Given different levels of bandwidth squeezing ratio β,we minimize the total number c of the spectrum slots required. The corresponding results for each case of the two protection techniques are shown in Figs. 11(a) 11(c) for the three test networks, respectively. For all the protection cases, we find that a larger bandwidth squeezing ratio requires a larger number of FSs to accommodate all the lightpath traffic demands. In addition, comparing the results of the four cases, we see that under BSR, the 1 1 protection technique always requires more FSs than the SBPP technique. This is again because SBPP allows protection capacity sharing among multiple protection lightpaths. In addition, comparing the cases with and without tunable transponders, we see that, because with tunable transponders a pair of working and protection lightpaths does not need to use the same set of FSs, the required number of FSs under tunable transponders is smaller than that under nontunable transponders for both of the SBPP and 1 1 path protection cases. Assuming a limited number of FSs on each fiber link, we can also maximize the restorability of network services, which is measured as a maximum achievable bandwidth squeezing ratio. The corresponding results are shown in Figs. 12(a) 12(c) for the three test networks, respectively. We can see that with increasing numbers of FSs on each fiber link, the maximal restoration level or bandwidth squeezing ratio grows accordingly. Such an observation is in line with our expectation that more network services would be restored with increasing network bandwidth. In addition, we notice that for the 1 1 path protection without tunable transponders, 100% restorability is not always achievable. This is because as shown in Figs. 11(a) 11(c), for the three test networks, the maximal number of FSs required for realizing 100% restorability is always larger than the values we set. V. CONCLUSION Fig. 12. Restoration percentages (or bandwidth squeezing ratios) under different maximal numbers of FSs on each fiber link. (a) n6s8. (b) COST239. (c) NSFNET. We considered the SBPP and 1 1 path protection techniques for elastic optical networks. For each of the techniques, we further considered two cases, i.e., with or without tunable transponders, which requires whether a pair of working and backup lightpaths should use the same sets of contiguous FSs. For all the cases, we developed the corresponding ILP optimization models to minimize both the required spare capacity and the maximum number of FSs used in the network. We also extended the four ILP models to design elastic optical networks with the BSR technique. Due to the advantage of spare capacity sharing, our results indicate that the SBPP technique always requires much lower spare capacity compared with the traditional 1 1 protection technique. In addition, it was observed that a higher nodal degree can help improve spare capacity redundancy for SBPP, which is attributed to

10 658 J. OPT. COMMUN. NETW./VOL. 6, NO. 7/JULY 2014 Shen et al. the fact that a denser network provides more opportunities for spare capacity sharing. Our results also show that, as expected, systems with tunable transponders require fewer FSs than the case without tunable transponders. This indicates that the tunablity of optical transponders can provide good flexibility in spectrum allocation and therefore improve the spectrum efficiency of the elastic optical network. Under the BSR technique, we made similar observations on how the feature of spare capacity sharing and transponder tunability can improve spectrum efficiency. For a larger number of given FSs on each fiber link, a higher restoration level or bandwidth squeezing ratio can be expected for all the protection cases. ACKNOWLEDGMENTS Part of the paper was presented in ACP 2012 [29]. This work was jointly supported by the National 863 Plans Project of China (2012AA011302), the National Natural Science Foundation of China (NSFC) ( , ), and the Natural Science Foundation of Jiangsu Province (BK , BK ). REFERENCES [1] M. Jinno, H. Takara, B. Kozicki, Y. Tsukishima, Y. Sone, and S. Matsuoka, Spectrum-efficient and scalable elastic optical path network: Architecture, benefits, and enabling technologies, IEEE Commun. Mag., vol. 47, no. 11, pp , Nov [2] G. Shen and M. Zukerman, Spectrum-efficient and agile CO-OFDM optical transport network: Architecture, design, and operation, IEEE Commun. Mag., vol. 50, no. 5, pp , May [3] Y. Wang, X. Cao, and Q. Hu, Routing and spectrum allocation in spectrum-sliced elastic optical path networks, in Proc. IEEE ICC, [4] Y. Wang, X. Cao, Q. Hu, and Y. Pan, Towards elastic and fine-granular bandwidth allocation in spectrum-sliced optical networks, J. Opt. Commun. Netw., vol. 4, no. 11, pp , Nov [5] K. Christodoulopoulos, I. Tomkos, and E. Varvarigos, Elastic bandwidth allocation in flexible OFDM-based optical networks, J. Lightwave Technol., vol. 29, no. 9, pp , May [6] X. Wan, L. Wang, N. Hua, H. Zhang, and X. Zheng, Dynamic routing and spectrum assignment in flexible optical path networks, in Proc. OFC/NFOEC, [7] M. Klinkowski and K. Walkowiak, Routing and spectrum assignment in spectrum sliced elastic optical path network, IEEE Commun. Lett., vol. 15, no. 8, pp , Aug [8] S. Shakya and X. Cao, Spectral defragmentation in elastic optical path networks using independent sets, in Proc. OFC/NFOEC, [9] L. Velasco, M. Klinkowski, M. Ruiz, and J. Comellas, Modeling the routing and spectrum allocation problem for flexgrid optical networks, Photon. Netw. Commun., vol. 24, no. 3, pp , Dec [10] X. Wan, N. Hua, and X. Zheng, Dynamic routing and spectrum assignment in spectrum-flexible transparent optical networks, J. Opt. Commun. Netw., vol. 4, no. 8, pp , Aug [11] Y. Liu, N. Hua, X. Wan, X. Zheng, and Z. Liu, A spectrum-scan routingschemeinflexibleopticalnetworks, inproc. ACP, [12] Y. Liu, N. Hua, X. Zheng, H. Zhang, and B. Zhou, Discrete spectrum-scan routing based on spectrum discretization in flexible optical networks, in Proc. OFC/NFOEC, [13] B. Chen, J. Zhang, Y. Zhao, H. Chen, S. Huang, W. Gu, and J. P. Jue, Minimized spectral resource consumption with rescaled failure probability constraint in flexible bandwidth optical networks, in Proc. OFC/NFOEC, [14] R. C. Almeida, Jr., A. F. Santos, K. D. R. Assis, H. Waldman, and J. F. Martins-Filho, Slot assignment strategy to reduce loss of capacity of contiguous-slot path requests in flexible grid optical networks, Electron. Lett., vol. 49, no. 5, pp , Feb [15] A. N. Patel, P. N. Ji, J. P. Jue, and T. Wang, Survivable transparent flexible optical WDM (FWDM) networks, in Proc. OFC/NFOEC, [16] M. Liu, M. Tornatore, and B. Mukherjee, Survivable traffic grooming in elastic optical network-shared path protection, in Proc. ICC, [17] X. Shao, Y. K. Yeo, Z. Xu, X. Cheng, and L. Zhou, Shared-path protection in OFDM-based optical networks with elastic bandwidth allocation, in Proc. OFC/NFOEC, [18] B. Chen, J. Zhang, Y. Zhao, C. Lv, W. Zhang, S. Huang, X. Zhang, and W. Gu, Multi-link failure restoration with dynamic load balancing in spectrum-elastic optical path networks, Opt. Fiber Technol., vol. 18, no. 1, pp , Jan [19] K. D. R. Assis, R. C. Almeida, Jr., and H. Waldman, MILP formulation for squeezed protection in spectrum-sliced elastic optical path networks, in Proc. SPECTS, [20] S. Kosaka, H. Hasewaga, K. Sato, T. Tanaka, A. Hirano, and M. Jinno, Shared protected elastic optical path network design that applies iterative re-optimization based on resource utilization efficiency measures, in Proc. ECEOC, [21] Y. Wei, G. Shen, and S. You, Span restoration for CO-OFDMbased elastic optical networks under spectrum conversion, in Proc. ACP, [22] Y. Sone, A. Watanabe, W. Imajuku, Y. Tsukishima, B. Kozicki, H. Takara, and M. Jinno, Bandwidth squeezed restoration in spectrum-sliced elastic optical path networks (SLICE), J. Opt. Commun. Netw., vol. 3, no. 3, pp , Mar [23] L. Ruan and N. Xiao, Survivable multipath routing and spectrum allocation in OFDM-based flexible optical networks, J. Opt. Commun. Netw., vol. 5, no. 3, pp , Mar [24] Y. Wei and G. Shen, Span restoration for flexi-grid optical networks under different spectrum conversion capabilities, in Proc. DRCN, [25] Y. Wei, G. Shen, and S. K. Bose, Applying ring cover technique to elastic optical network, in Proc. ACP, [26] W. D. Grover, Mesh-Based Survivable Networks. Upper Saddle River, NJ: Prentice Hall PTR, 2003, ch. 5. [27] G. Shen and W. D. Grover, Dynamic path-protected service provisioning in optical transport networks with a limited number of add/drop ports and transmitter tunability, J. Sel. Areas Commun., vol. 25, no. 6, pp , Aug [28] Gurobi [Online]. Available: [29] G. Shen, Y. Wei, and Q. Yang, Shared backup path protection (SBPP) in elastic optical transport networks, in Proc. ACP, Gangxiang Shen [S 98 M 99 SM 12] received his B.Eng. degree from Zhejiang University, China; his M.Sc. degree from Nanyang Technological University, Singapore; and his Ph.D. degree

11 Shen et al. VOL. 6, NO. 7/JULY 2014/J. OPT. COMMUN. NETW. 659 from the University of Alberta, Canada, in January He is a Distinguished Professor with the School of Electronic and Information Engineering of Soochow University in China. Before he joined Soochow University, he was a Lead Engineer with Ciena, Linthicum, Maryland. He was also an Australian ARC Postdoctoral Fellow with University of Melbourne. His research interests include integrated optical and wireless networks, spectrum efficient optical networks, and green optical networks. He has authored and coauthored more than 80 peer-reviewed technical papers. He is a Lead Guest Editor of the IEEE Journal on Selected Areas in Communications (J-SAC) Special Issue on Next-Generation Spectrum-Efficient and Elastic Optical Transport Networks and a Guest Editor of the IEEE JSAC Special Issue on Energy-Efficiency in Optical Networks. He is an Associate Editor of the Journal of Optical Communications and Networking, and an editorial board member of Optical Switching and Networking. He is a Secretary for the IEEE Fiber-Wireless (FiWi) Integration Sub-Technical Committee. He received the Young Researcher New Star Scientist Award in the 2010 Scopus Young Researcher Award Scheme in China. He was a recipient of the Izaak Walton Killam Memorial Award from the University of Alberta and the Canadian NSERC Industrial R&D Fellowship. He is a Senior Member of the IEEE. Yue Wei is currently a graduate student in the School of Electronic and Information Engineering, Soochow University. His research interest focuses on optical network survivability. Sanjay Kumar Bose got his B.Tech. degree from IIT Kanpur in 1976 and his Master s and Ph.D. from S.U.N.Y. Stony Brook, USA, in 1977 and 1980, respectively. After working with the Corporate R&D Centre of the General Electric Co. in Schenctady, NY, till 1982, he joined IIT Kanpur as an Assistant Professor and became a Professor there in He left IIT Kanpur in 2003 to join the faculty of the School of EEE, NTU, Singapore. In December 2008, he left NTU to join IIT Guwahati, where he is currently a Professor in the Department of EEE and the Dean, Alumni Affairs and External Relations. Prof. Bose has been working in various areas in the field of computer networks and queueing systems and has published extensively in the area of optical networks and network routing. Prof. Bose is a Senior Member of the IEEE, a Fellow of IETE (India), and a member of Sigma Xi and Eta Kappa Nu. More details on Prof. Bose can be found on his Web page at