Sharing Spectrum and Straddling p-Cycle FIPP for

Protection Against Two Simultaneous Failures in

SDM Elastic Optical Networks

Helder M. N. S. Oliveira and Nelson L. S. da Fonseca

Institute of Computing - State University of Campinas, Brazil
Email: [email protected], [email protected]

Abstract—The introduction of Space division multiplexing impairments that reduces the spectrum usability.
(SDM) in optical networks brings new challenges for protection
of networks since a lightpath can span multiple cores. In this The Internet backbone is composed by optical fibers span-
paper, we investigate the problem of dynamic protection against ning long distances and high transmission rates. In optical
two simultaneous failures in spacial division multiplexing elastic transport networks that carry huge amounts of traffic, redun-
optical networks. For that, we propose a new path-protection dancy is the most adequate approach to augment the end-to-end
sharing spectrum and straddling p-cycle FIPP algorithm called (E2E) service availability. p-Cycle is an attractive protection
Sharing Slot and Straddling p-Cycle FIPP (SSSPF). In SSSPF,
each connection is assigned one primary path and one link-
schemes, and has been intensively investigated in the past
disjoint backup path. SSSPF is the first algorithm in literature, years. p-Cycles combine the properties of ring-like recovery
to provide protection against two simultaneous failure in SDM speed and efficiency of restorable mesh networks. p-Cycles
elastic optical network. protect the working capacity on the span they cover, as shared
protection rings, and, unlikely rings, they protect the working
Keywords—Protection, Multi-core Fiber, Elastic Optical Net- capacity of off-cycle spans which have their end-points on
work with Space Division Multiplexing, p-cycle FIPP.
the p-cycle (straddling spans). A type of p-cycle of special
interest is the failure-independent path protecting p-cycle (FIPP
I. I NTRODUCTION p-cycles) which provide fully pre-connected protection paths
in optical networks. FIPP p-cycles offer all the advantages of
Over the last few years, the capacity limitation of sin- SBPP and in addition the protection path is pre-configured.
gle core optical fibers has motivated the definition of new
techniques to increase the traffic capacity in optical fibers In elastic optical networks, traffic grooming is a technique
leading to the emergence of spacial division multiplexing that combines multiple connections in an optical path without
(SDM). Spacial division multiplexing employs multiple single needing guard bands between them [2]. Sharing spectrum is a
mode cores placed in a single fiber structure. Space division technique in which two backup lightpaths can use the same
multiplexing can be realized using multimode fiber (MMF), cores, links and spectrum, since the working paths of the
multicore Fiber (MCF) or few-mode multicore fiber. In MMF, two connections are physically disjoint [3]. The combination
the number of modes supported by a fiber depends on the core of traffic grooming and spectrum sharing allows significant
size and the refraction index of the fiber cladding. In MCF, gain in spectrum utilization, which decreases the blocking of
each core acts as a single mode fiber. connections.
The routing and spectrum assignment (RSA) problem is In this paper, we propose an algorithm called Sharing
a fundamental problem in elastic optical networks (EON). In Slot and Straddling p-Cycle FIPP (SSSPF) for providing
RSA, there are constraints assuring contiguous and continuous FIPP p-cycle protection in SDM-EONs. The algorithm creates
allocation of the spectrum on all links of the selected route [1]. protection paths against two simultaneous failures, using the
However, in SDM, it is possible to allocate one or more cores straddling FIPP p-cycle technique, sharing spectrum and traffic
for the establishment of a connection. The inclusion of the grooming. Results show that the proposed algorithm pro-
space degree of freedom adds another dimension to the RSA motes protection effectively without compromising networking
problem becoming the routing, spectrum and core allocation blocking. The key advantages of p-cycles are pre-configured
(RSCA) problem. Moreover, in RCSA additional issues such protection, switching speed and operational simplicity similar
as inter-core crosstalk should be taken into account. Inter-core to ring networks. Therefore, FIPP p-cycle protection has great
crosstalk happens when the same spectrum propagates through potentiality to play a key role in SDM- EON protection.
adjacent cores in MCF. Elastic optical networks with SDM
promises to provide much larger capacity when compared to The rest of the paper is organized as follow. Section II
conventional single mode fiber systems. If on one hand, Space- reviews related work. Section III introduces the proposed al-
Division Multiplexing (SDM) technology allows the increase gorithm. Section IV evaluates the performance of the proposed
of network capacity, on the other hand, MCF produces physical algorithm and Section V concludes the paper.
978-1-5386-2098-4/17/$31.00 c 2017 IEEE
 II. R ELATED W ORK solution. In these graphs, (Fig. 1d) an ∞ label value means
that at least one slots is already allocated whereas the value 1
The emergence of spacial division multiplexing elastic
means that all slots are available for allocation.
optical networks has motivated several investigations, mainly
on RCSA algorithms but only recently protection schemes have The following notation will be used to describe the algo-
been proposed and no other algorithm has been proposed for rithm:
protection against two failures.
s: source node;
Saridis et.al. [4] reviewed research progress on spacial
division multiplexing fibers and network components. They d: destination node;
introduced two figures of merit aiming for quantitative eval- b: bandwidth demand;
uation of technologies such as amplifiers, fan-in/fan-out mul-
tiplexers, transmitters, switches, and SDM nodes. In [6], it N : number of slots between two nodes;
is introduced a Routing, Core and Spectrum Assignment C: number of cores;
(RCSA) algorithm based on the Connected Component La-
belling (CCL) algorithm. Spectrum fitting policies are also V : set of nodes;
proposed to be jointly employed with the CCL algorithm. The
authors in [7] introduced an algorithm based on p-cycle to eu,v,n : the nth edges connecting u and v;
provide failure-independent path protection in elastic optical E = {eu,v,n }: set of edges;
networks with spacial division multiplexing. However shared
slot is not considered. In [8] [9], it is proposed an algorithm G = (V, E, W ): labeled multigraph composed by a set
to provide Failure-independent path protecting p-cycle with of nodes V , a set of edges E and a set of edge weight W ,
minimum interference for path protection in elastic optical |E| = C · N · |V |. The edges connecting two vertices of G
networks using space division multiplexing. Hirota et.al. [10] represent the N slots in the link connecting two network nodes;
divides the RSCA problem into the routing, and Core and r(s, d, b): request from the node s to the node d with
Spectrum Assignment (SCA) problems, and introduces a K- bandwidth demand b;
shortest path based pre-computation method as the routing
solution. They proposed SCA methods with crosstalk aware- δ(G, r(s, d, b)): shortest path between s and d in G that
ness. In [11], it is proposed an algorithm to provide protection satisfies the request of b slots ;
using p-cycle FIPP and modulation. The authors evaluated
w(eu,v,n ): weight of the edge eu,v,n ; w(eu,v,n ) = 1 if
the energy efficiency of the algorithm combining p-cycle and
the nth slot in the link connecting OXC u and v is free and
adaptive modulation. Sasaki et.al. [12] numerically analyzed
w(eu,v,n ) = ∞ if the slot is already allocated;
the crosstalk behaviors over various effective index differences
between non-identical cores. The authors in [13] evaluated W = {w(eu,v,n )}:set of edge weights
the advantages of using the extra dimension introduced by
space-division multiplexing (SDM) for dynamic bandwidth- Ve = V : set of nodes;
allocation purposes in a flexible optical network. In [14], a eeu,v ∈ E:
e edge connecting u
e and ve;
routing, spectrum and core allocation (RSCA) problem for
flexgrid optical networks is proposed for network planning eeue,ev = {eu,v,n } ∈ E is a chain such that eu,v,n is the
problem using integer linear programming (ILP) formulation least ordered edge, eu,v,n+b is the greatest ordered edge and
as well a heuristic. The spectrum overlap and p-cycle FIPP |e
eu,v | = b;
was studied in [15] for protection in elastic optical networks.
w
en (e
eue,ev ): weight of the edge eeue,ev ;
III. T HE SSSPF A LGORITHM W
f=w
en (e
eue,ev );
The algorithm introduced in this subsection, called Sharing G
e n,b = (Ve , E,
e Wf ): the nth labeled graph such that E
e is the
Slot and Straddling p-Cycle FIPP (SSSPF), decides on the set of edges connecting {e u, ve} ∈ V and W is the set of costs
e f
establishment of lightpaths in protected networks. A lightpath
associated to E.e The edges in E e correspond to the mapping
is established if and only if it can be protected by a shared
of b edges in G starting at the nth edge;
path against two failures.
The proposed algorithm models the spectrum availability in σ = |{G e n,b }| = C × (N − b + 1): number of graphs
the network as labeled multigraph (Fig. 1a). A label on an edge extracted from the multigraph;
represents the availability of a slot. In Fig. 1b, the multigraph τ (G, C, b) = {G
e n,b }: function which produces all σ graphs
is divided into C multigraphs, where C is the number of cores. from G;
Each of these multigraphs is transformed into multigraphs with
N − b + 1 edges, (Fig. 1c) where b is the bandwidth demand Pn : chain of Ge n,b such that the source node s is the least
in slot. Then, each of these multigraphs is transformed into ordered node and d is the greatest ordered node;
N − b + 1 graphs. In other words, the original multigraph P
W (Pn ): eeu,e eeue,ev : the weight of the path Pn (the
e v ∈{Pn }
(Figure 1c) is transformed into C × (N − b + 1) graphs (Fig. sum of the weights of all the edges in the chain);
1d). Each edges in these graphs represent a combination of
b slots. This representation assures spectrum contiguity to the WPs,d = weight of the shortest path between s and d;
(a) Network with 3 cores and 4 slots. (b) The Multigraph, separated by (c) The Multigraph in that set edges (d) Graphs generated.
cores, each one representing 4 are mapped in to one edge, contigu-
slots. ity constraint.

Fig. 1: Transforming multigraph in graphs

κ(G, C, b, Pn ) = {G e n,b }: function which produces all Algorithm 1 SSSPF

graphs from G, considering that slots of protection can be 1: τ (G, C, b)
shared, since the working paths (Pn ) of the connections are 2: (W (Pn ), Pn ) = δ(G e n,b , r(s, d, b)) ∀n ∈ σ
physically disjoint (sharing slot); 3: WPs,d = W (Pn )| ∀i W (Pn ) ≤ W (Pi )
4: if WPs,d = ∞ then
Tn : chain of G
e n,b such that the source node s is the least
5: block r(s, d, b)
ordered node and d is the greatest ordered node; 6: else
7: if ∃$(Pn , Ts,d , r(s, d, b)) then
Tu,v : set of all backup path between vertices u and v in 8: establish r(s, d, b) as Pn and Ts,d
G; 9: W (eeu,v,i ) = ∞ ∀{u, v} ∈ Pi
PTu,v : set of all paths protected by backup path Tu,v ; 10: else
11: κ(G, C, b, Pn )
T = {Tu,v }: set of all established backup paths; 12: (W (Tn ), Tn ) = ρ(G e n,b , Pn , r(s, d, b))
13: WTs,d = W (Tn )| ∀i W (Tn ) ≤ W (Ti )
$(Pn , Tu,v , r(s, d, b)): backup path in Tu,v which PTu,v 14: if WTs,d = ∞ then
are link disjoint to Pn and PTu,v are link disjoint to Tu,v and 15: block r(s, d, b)
satisfies the request of bandwith b; 16: else
17: establish r(s, d, b) as Pn and Tn
ρ(Ge n,b , Pn , Tu,v , r(s, d, b)): shortest straddling p-cycle in
18: W (eeu,v,i ) = ∞ ∀{u, v} ∈ Pi
Tu,v which PTu,v are link disjoint to Pn and satisfies the 19: W (eeu,v,i ) = ∞ ∀{u, v} ∈ Ti
request of bandwith b; 20: end if
P 21: end if
W (Tn ): e e v ∈{Tn }
eu,e eeue,ev : the weight of the p-cycles Tn 22: end if
(the sum of the weights of all the edges in the chain);
WTs,d = weight of the p-cycle which protects the path
between s and d; protect the lightpath to be established should be created (Line
12). In case no path can be created to protect the lightpath
The algorithm SSSPF is introduced in Algorithm 1. Line then the request is blocked (Line 15). In case a path can be
1 transforms the multigraph into C × (N − b + 1) graphs. created, the primary path as well as the backup path (Line 17)
Line 2 computes the shortest path for all graph G e n,b and are established to satisfy the request and the corresponding
chooses the least costs one. If the weight of the shortest path edges in the multigraph G have their weight changed to ∞
is ∞, it was not possible to find a path under the contiguity (Lines 18 and 19) meaning that the slots were allocated to the
constraint for the demand b. Line 3 selects the path among newly established lightpath.
all shortest paths that has the lowest weight value. In case The complexity of the SSSPF algorithm is analyzed as
the weight of all shortest path is ∞ (Line 4), there is no follows. The complexity of transforming the original multi-
path in the network that satisfies the request of b slots under graph in graphs is O(E + V ). In the worst case, a shortest
the contiguity constraint. If there is no path available then straddling p-cycle algorithm is executed in C × N − b graphs,
the request is blocked (Line 5). Otherwise, another path to O(E + V + (C × N × (kEk + kV k × logkV k))), since C and
protect the lightpath to be established is searched (Line 7). In N values can be expressed as constant, then the complexity
case there exists a path, the lightpath is established (Line 8) is: O(kEk + kV klogkV k).
and the corresponding edges in the multigraph G have their
weight changed to ∞ (Line 9) meaning that the slots were
IV. P ERFORMANCE E VALUATION
allocated to the newly established lightpath. Otherwise, Line
11 transforms the multigraph into C × (N − b + 1) graphs, To assess the performance of the SSSPF algorithm in multi-
considering the spectrum sharing for protecting slots. A path to core networks, simulation experiments were employed using
the FlexGridSim [16] simulator. In each simulation, 100,000
requests were generated as input and simulations for all the n{1 − exp(−(n + 1) · 2 · h · L}
XT = (2)
algorithms used the same set of seeds. Seven types of requests 1 + n{exp(−(n + 1) · 2 · h · L)}
were employed 25 Gbps, 50 Gbps, 125 Gbps, 200 Gbps, 500
Gbps, 750 Gbps and 1 Tbps. The links were composed by Eq. 2 uses the mean crosstalk increase per unit length (Eq.
MCFs with 7 core and each core was divided in 320 slots. 1), the length of the fiber (L) and n represents the number of
Confidence intervals were derived using the independent repli- neighboring cores.
cation method with 95% confidence level. Requests follows a
Poisson process and are uniformly-distributed among all node- We assumed a maximum n value of 6, k = 2 × 10−5 , R
pairs of network. The topology used in the simulations were = 50 mm, β = 4 × 106 e D = 45 µm [4]. In this study, the
the USA (Figure 2a) and the NSF (Figure 2b) topologies. The worst case of crosstalk is always be the reached at central core
NSF topology has 16 nodes and 25 links whereas the USA (or any other core that has the largest number of neighboring
topology has 24 nodes and 43 links (Fig. 2). The numbers on cores), since it receives undesired interference from all others
the links represent the length of the link in kilometers. adjacent cores. In our simulation, we assume that the spectrum
of each core is fully utilized.
In the figures, the curves labeled "SSCA" show the re-
sults for networks using the algorithm based on the methods
proposed in [10] which uses a K-shortest paths algorithm to
compute routes, we use K = 3. The curves labeled "FIPPMC"
show the results for networks using the algorithm FIPPMC [7].
The FIPPMC decides on the establishment of lightpaths in an
FIPP p-cycle protected network. The curves labeled "MIFMC"
show the results for networks using the algorithm MIFMC [8],
[9]. The MIFMC algorithm prioritizes the use of straddling
p-cycles in order to generate minimum interference to reduce
rejections of future requests. The curves labeled "SSSPF" show
the results for networks using the proposed algorithm.
(a) USA Topology
SSSPF SSCA
FIPPMC MIFMC
1
Bandwidth Blocking Ratio

0,1

0,01

0,001
(b) NSF Topology
0,0001
Fig. 2: Topologies
1e−05

1e−06
The inter-core crosstalk is a type of interference in which 0 50 100 150 200 250 300 350 400 450 500
one core causes in another core of the same link, i.e., the ratio Load (erlangs)
of the optical power inserted from adjacent cores to the one
divided by the power of the signal already in that core and Fig. 3: Bandwidth blocking ratio for the USA topology
measured in dB [4]. To calculate the crosstalk (XT) from one
core in relation to n neighboring cores, in a homogeneous MCF
fiber, we used Eq. 1. Considering the coupled-power theory [4] Fig. 3 shows the bandwidth blocking ratio (BBR) as a
[5], and using Eq. 1 leads to Eq. 2, which was used to ensure function of the load for the USA topology. While SSSPF
the quality of transmission of the connections. starts blocking requests under loads of 25 erlangs, FIPPMC
and MIFMC start blocking only under loads of 125 erlangs,
and SSCA starts blocking requests under loads of 50 erlangs.
2 · k2 · R The SSSPF algorithm produces the highest BBR, since it
h= (1) produces double protection. SSSPF takes advantage of the high
β·D
connectivities of nodes in the USA topology and the BBR
increases smoothly as a function of the load increase. Under
Eq. 1 expresses the mean crosstalk increase per unit length; loads of 125 erlangs, the BBR produced by the SSSPF is
h is the mean crosstalk increase per unit length, k, β, R, D similar to that produced by the SSCA algorithm, since it does
are coupling coefficient, propagation constant, bend radius and not use spectrum sharing and traffic grooming. Under high
core-pitch, respectively. loads of 275 erlangs, the BBR values produced by the SSSPF
is lower than those produced by the FIPPMC, and similar to by dividing the total traffic demand successfully served in the
those produced by MIFMC. These results show that the SSSPF network by the total power network consumption. The energy
algorithm produces acceptable blocking for SDM with multi efficiency produced by SSSPF is higher than that produced by
core fibers in despite of protecting two simultaneous failure. the FIPPMC and MIFMC algorithms, since it uses spectrum
sharing and traffic grooming. There is not much difference
SSSPF SSCA between the energy efficiency produced by FIPPMC and that
FIPPMC MIFMC
produced by MIFMC. Until 325 erlangs, the energy efficiency
0,6
produced by the SSSPF algorithm is higher than that produced
0,5 by the SSCA algorithm.
Crosstalk per Slot Ratio

0,4 SSSPF SSCA

FIPPMC MIFMC
0,3 1

Bandwidth Blocking Ratio

0,2 0,1

0,1
0,01
0
0 50 100 150 200 250 300 350 400 450 500 0,001
Load (erlangs)

Fig. 4: Crosstalk per slot ratio for the USA topology 0,0001

1e−05
The use of seven cores generates intercore crosstalk. Fig. 4 0 50 100 150 200 250 300 350 400 450 500
shows the “Crosstalk per Slot” (CpS) as a function of the load Load (erlangs)
for the USA topology. The crosstalk value for each spectrum
slot is defined as the ratio of actual crosstalk index to the max- Fig. 6: Bandwidth blocking ratio for the NSF topology
imum value of crosstalk index. The crosstalk ratio is defined
as the average value considering all spectrum slots [1]. The
CpS is not considered when the slot is reserved but not used. Fig. 6 shows the bandwidth blocking ratio (BBR) as
The generated CpS by the SSSPF algorithm starts at a 0.04 a function of the load for the NSF topology. The SSSPF
and increases until 0.51 while that generated by the FIPPMC algorithm starts blocking requests under low load. The highest
algorithm starts at 0.01 and increases until 0.37. The generated BBR produced by the SSSPF algorithm is due to the cost of
CpS for the MIFMC algorithm starts at 0.01 and increases until protecting against two simultaneous failures. While FIPPMC
0.43. The SSSPF algorithm produces the highest CpS values, and MIFMC start blocking requests under load of 125 erlangs,
as a consequence of the high utilization produced. The SSCA SSSPF starts blocking only under loads of 25 erlangs. Such
algorithm produces the lowest CpS values, as a consequence trend is a consequence of the FIPPMC and MIFMC algorithms
of the high blocking and low utilization generated. protecting against a single failure only. Under loads of 125
erlangs, the difference between the BBR values produced by
SSSPF SSCA the SSSPF algorithm and those given by the FIPPMC and
FIPPMC MIFMC MIFMC algorithms is almost two order and three order of
100 magnitude, respectively. Under high loads of 275 erlangs, the
Energy Efficiency (Mbits/Joule)

95 BBR values produced by the SSSPF is similar to the values

produced by the other algorithms. The low node degree in this
90 topology leads to the creation of bottlenecks as well as a rapid
85
increase in blocking when compared to the blocking for the
USA topology.
80
Fig. 7 shows the “Crosstalk per Slot” (CpS) as a function
75 of the load for the NSF topology. The CpS generated when
70 employing the NSF topology is higher than that produced
when employing the USA topology. The CpS generated by
65 the SSSPF algorithm starts at a 0.06 value and increases until
0 50 100 150 200 250 300 350 400 450 500
0.62. The CpS generated by the FIPPMC algorithm starts at a
Load (erlangs) 0.02 value and increases until 0.55. The CpS generated by the
Fig. 5: Energy Efficiency for the USA topology MIFMC algorithm starts at a 0.02 value and increases until
0.56. The SSCA algorithm produces the lowest CpS values
when compared to the CpS values generated by the other three
Fig. 5 shows the energy efficiency as a function of the algorithms, as consequence of producing high blocking and
load for the USA topology. The energy efficiency is obtained low utilization.
 SSSPF SSCA R EFERENCES
FIPPMC MIFMC
0,7
[1] S. Fujii, Y. Hirota, H. Tode, and K. Murakami, “On-demand spectrum
0,6
and core allocation for reducing crosstalk in multicore fibers in elastic
Crosstalk per Slot Ratio

0,5 optical networks,” Optical Communications and Networking, IEEE/OSA

Journal of, vol. 6, no. 12, pp. 1059–1071, Dec 2014.
0,4 [2] G. Zhang, M. De Leenheer, and B. Mukherjee, “Optical traffic grooming
in ofdm-based elastic optical networks [invited],” Optical Communica-
0,3 tions and Networking, IEEE/OSA Journal of, vol. 4, no. 11, pp. B17–
B25, Nov 2012.
0,2
[3] M. Liu, M. Tornatore, and B. Mukherjee, “Survivable traffic grooming
0,1 in elastic optical networks -shared protection,” Lightwave Technology,
Journal of, vol. 31, no. 6, pp. 6230–6234, March 2013.
0
0 50 100 150 200 250 300 350 400 450 500 [4] G. M. Saridis, D. Alexandropoulos, G. Zervas, and D. Simeonidou,
Load (erlangs) “Survey and evaluation of space division multiplexing: From technolo-
gies to optical networks,” IEEE Communications Surveys Tutorials,
Fig. 7: Crosstalk per slot ratio for the NSF topology vol. 17, no. 4, pp. 2136–2156, Fourthquarter 2015.
[5] T. Hayashi, T. Taru, O. Shimakawa, T. Sasaki, and E. Sasaoka,
“Uncoupled multi-core fiber enhancing signal-to-noise ratio,” Opt.
SSSPF SSCA Express, vol. 20, no. 26, pp. B94–B103, Dec 2012. [Online]. Available:
FIPPMC MIFMC http://www.opticsexpress.org/abstract.cfm?URI=oe-20-26-B94
130
[6] P. M. Moura and N. L. S. da Fonseca, “Routing, core and spectrum
Energy Efficiency (Mbits/Joule)

125 assignment based on connected component labelling for sdm optical

networks,” in 2016 IEEE International Conference on Communications
120 (ICC), May 2016, pp. 1–6.
115 [7] H. M. N. S. Oliveira and N. L. S. da Fonseca, “Algorithm for protection
110 of space division multiplexing elastic optical networks,” in 2016 IEEE
Global Communications Conference (GLOBECOM), Dec 2016, pp. 1–
105 6.
100 [8] ——, “The minimum interference p-cycle algorithm for protection of
space division multiplexing elastic optical networks,” in 2016 8th IEEE
95 Latin-American Conference on Communications (LATINCOM), Nov
90 2016, pp. 1–6.
0 50 100 150 200 250 300 350 400 450 500 [9] ——, “The minimum interference p-cycle algorithm for protection
Load (erlangs) of space division multiplexing elastic optical networks,” IEEE Latin
America Transactions, vol. 15, no. 7, pp. 1342–1348, 2017.
Fig. 8: Energy Efficiency for the NSF topology
[10] H. Tode and Y. Hirota, “Routing, spectrum and core assignment for
space division multiplexing elastic optical networks,” in Telecommuni-
cations Network Strategy and Planning Symposium (Networks), 2014
16th International, Sept 2014, pp. 1–7.
Fig. 8 shows the energy efficiency as a function of the
load for the NSF topology. The energy efficiency produced [11] H. M. N. S. Oliveira and N. L. S. d. Fonseca, “Proteção de redes
Ópticas elásticas com multiplexação espacial baseada em modulação,
by the SSSPF algorithm is the highest one as a consequence p-cycle FIPP e interferência mínima,” in 2017 Brazilian Symposium on
of using spectrum sharing and traffic grooming. As for the Computer Networks and Distributed Systems, May 2017, pp. 831–844.
USA topology, there is not much difference between the energy
[12] Y. Sasaki, Y. Amma, K. Takenaga, S. Matsuo, K. Saitoh, and
efficiency produced by FIPPMC and MIFMC algorithms. Until M. Koshiba, “Investigation of crosstalk dependencies on bending radius
325 erlangs, the SSSPF algorithm produces energy efficiency of heterogeneous multicore fiber,” in 2013 Optical Fiber Communica-
higher than does that produced by SSCA algorithm. tion Conference and Exposition and the National Fiber Optic Engineers
Conference (OFC/NFOEC), March 2013, pp. 1–3.
V. C ONCLUSION [13] P. S. Khodashenas, J. M. Rivas-Moscoso, D. Siracusa, F. Pederzolli,
B. Shariati, D. Klonidis, E. Salvadori, and I. Tomkos, “Comparison
Protection is a fundamental problem in optical networks, of spectral and spatial super-channel allocation schemes for sdm net-
especially in SDM elastic optical networks. This paper focused works,” Journal of Lightwave Technology, vol. 34, no. 11, pp. 2710–
on the problem of dynamic protection in spacial division multi- 2716, June 2016.
plexing elastic optical networks mesh networks and proposed a [14] A. Muhammad, G. Zervas, D. Simeonidou, and R. Forchheimer,
new algorithm called SSSPF to recover from two simultaneous “Routing, spectrum and core allocation in flexgrid sdm networks with
link failures. The SSSPF algorithm provides 100% protection multi-core fibers,” in Optical Network Design and Modeling, 2014
for two failures. Results indicate that the overhead demanded International Conference on, May 2014, pp. 192–197.
by the SSSPF algorithm is quite acceptable. [15] H. M. N. S. Oliveira and N. L. S. da Fonseca, “Traffic grooming
and spectrum overlap in fipp p-cycle for protection of elastic optical
networks,” in 2016 8th IEEE Latin-American Conference on Commu-
ACKNOWLEDGMENTS nications (LATINCOM), Nov 2016, pp. 1–6.
The authors would like to thank the grant 165446/2015-3, [16] P. M. Moura and A. C. Drummond, “FlexGridSim: Flexible Grid Optical
CNPq for the financial support. Network Simulator,” http://www.lrc.ic.unicamp.br/FlexGridSim/.