Conference PaperPDF Available

Simulated Annealing algorithm for optimization of elastic optical networks with unicast and anycast traffic



The growing number of services like Content Delivery Networks (CDNs) and cloud computing triggers a sustainable growth of data transfer and consequently leads to an increasing interest in anycasting that is an efficient way to provision network requests related to Data Centers. On the other hand, Elastic Optical Network (EON) is an effective and cost-efficient solution for optical networks, which can support bandwidth-demanding needs ranging beyond 100 Gb/s. This paper focuses on joint optimization of unicast and anycast traffic in EONs including both Routing and Spectrum Allocation (RSA). We propose a novel heuristic algorithm to solve the RSA problem. The algorithm - named VSA - is a hybrid method that combines a Simulated Annealing (SA) approach and a simple greedy algorithm. To evaluate our approach, we run a wide range of experiments on real network topologies. First, we tune VSA to find the best values of the algorithm's parameters. Next, we compare the performance of VSA against CPLEX (optimal results) and other heuristics. According to our experiments, VSA can improve the solution provided by a greedy algorithm by more than 10% and on average is about 4% worse in comparison to the optimal results.
ICTON 2014 Th.B3.6
Simulated Annealing Algorithm for Optimization of Elastic
Optical Networks with Unicast and Anycast Traffic
0LFKDá$LELQ and Krzysztof Walkowiak*, Member, IEEE
The growing number of services like Content Delivery Networks (CDNs) and cloud computing triggers
a sustainable growth of data transfer and consequently leads to an increasing interest in anycasting
that is an efficient way to provision network requests related to Data Centers. On the other hand, Elastic Optical
Network (EON) is an effective and cost-efficient solution for optical networks, which can support bandwidth-
demanding needs ranging beyond 100 Gb/s. This paper focuses on joint optimization of unicast and anycast
traffic in EONs including both Routing and Spectrum Allocation (RSA). We propose a novel heuristic algorithm
to solve the RSA problem. The algorithm – named VSA – is a hybrid method that combines a Simulated
Annealing (SA) approach and a simple greedy algorithm. To evaluate our approach, we run a wide range of
experiments on real network topologies. First, we tune VSA to find the best values of the algorithm’s
parameters. Next, we compare the performance of VSA against CPLEX (optimal results) and other heuristics.
According to our experiments, VSA can improve the solution provided by a greedy algorithm by more than 10%
and on average is about 4% worse in comparison to the optimal results.
Keywords: elastic optical networks, anycast traffic, simulated annealing.
In recent years, we can observe a continued growth of data transfer, which triggers the demand to develop an
efficient and scalable optical transport platform for a capacity beyond 100 Gb/s [1]. Moreover, a substantial
increase can be seen in anycast traffic defined as one-to-one-of-many transmission techniques. This increase is
caused by the introduction of new services, such as CDNs, cloud computing, IPTV and Video On Demand
(VOD) [2], [4]. One of the technologies, which enable improved use of flexible optical network, is a scalable and
efficient architecture called SLICE. This approach permits sharing links with a capacity beyond 100 Gb/s into
smaller slices and combines them into larger groups. As a result we can achieve a greater variety of spectrum
selection scenarios [3]. The SLICE approach, also known as Elastic Optical Network (EON), is an evolution of
current optical networks. In contrast to current solutions, EON allows the resources to be assigned accordingly to
the size of the required bandwidth [2].
A new optimization problem, known as Routing and Spectrum Allocation (RSA), occurs with the development
of the EON technology. In this paper, we propose a novel metaheuristic algorithm for an RSA problem with joint
unicast and anycast traffic. The proposed VSA algorithm is a hybrid method that improves a solution yielded by
a simple greedy algorithm by using Simulated Annealing approach. An important part of metaheuristics is tuning
of the algorithm. Therefore, we pay special attention to evaluate selection of algorithm’s parameters. According
to the best of our knowledge, this is the first paper that proposes a SA algorithm in the optimization of EONs
with joint unicast and anycast flows.
The remainder of the paper is organized as follows. In Section 2, we describe a heuristic algorithm – VSA.
In Section 3, we present numerical results for algorithm tuning, followed by our results and lastly we conclude
our findings.
In this Section, we present a novel VSA algorithm proposed to solve RSA problems with joint unicast and
anycast traffic. In particular, we solve two ILP models with two different objective functions. The first one –
called MaxSpec – minimizes the maximum spectrum usage and was formulated in [4]. The second model called
AvgSpec focuses on minimization of average spectrum usage and was presented in [7]. The VSA method is
based on the SA approach, which is a generic probabilistic heuristic for the global optimization of a given
function in a large search space.
In VSA, a solution of the optimization problem is represented as a sequence (ordering) of demands. To
calculate the objective function of a particular solution, we allocate the demands in the network one by one
according to the particular sequence using a following procedure. For each demand, we analyze all possible
candidate paths provided in advance (HJ., using N-shortest path algorithm). For each examined path, using the
First Fit approach we calculate the value of the considered objective function, assuming that the demand is
assigned to a particular path. Finally, we select a candidate path that guarantees the best value of the objective
function. In the case of anycast demands, the candidate paths leading to all possible Data Center (DC) nodes are
analyzed. Below, we present a pseudo-code of the algorithm.
978-1-4799-5601-2/14/$31.00 ©2014 IEEE 1
ICTON 2014 Th.B3.6
Algorithm 1 VSA
2: while (LLPD[ and 7!0.01):
3: G1
rand(D), G2
4. Swap(G1G2); calculate FXUUHQW5
5. WHPS$
6. if (WHPS$ < EHVW):
7: EHVW
9: if (GHOWD < 0):
11: else:
12: [
13: if ([ < exp-GHOWD7)):
15: else:
16: Swap(G1G2);
17: 7
18: return EHVW;
In line 1, we calculate an initial solution (sequence of demands and value of the objective function) denoted as
FXUUHQW0 using the MSF algorithm introduced in [5]. This solution is selected as the best solution – EHVW and
current solution – FXUUHQW5. Also, in line 1, we calculate initial temperature 7. We present an innovative
approach for the calculation of this parameter that is fully automated. We take the result (number of slices)
generated by MSF in line 1 and multiply it by Pparameter, which is an input tuning parameter of VSA. The
simulation is being processed until we fulfill conditions from line 2 (number of iterations Lis smaller than the
maximum number LPD[and the temperature 7 has not reached absolute temperature, which is equal to 0.01). In
line 3, we randomly select two demands, swap them in the sequence of demands and calculate a new result. We
save the number of the current solution (number of slices) (countAllTakenSlices()) into WHPS$ variable and then
process it by using the SA algorithm. If condition in line 6 is fulfilled, we assign a temporary result to best result
and calculate GHOWD – difference between FXUUHQW5DQGWHPS$. Next, in line 9 we review if the current solution is
better than the best by checking the value of GHOWD and accordingly we assign FXUUHQW as WHPS$. If it is not
fulfilled, we randomly pick a value of [ (line 12) and check if it is smaller than Boltzmann function of
probability distribution (line 13). If it is fulfilled, we repeat the same operation as in line 10. Alternatively, (LH., [
is larger or equal then Boltzmann function), we swap the demands to restore the previous state without saving
the result. Finally, we reduce the temperature 7 with the M parameter, named as a cooling rate parameter. When
we fulfill conditions from the loop in line 2, we return best result as a result of an algorithm. For better
understanding of the algorithm refer to [6].
In this Section, we discuss the results of computational experiments. The goal of the experiments is threefold.
First, we evaluate the performance of the VSA by tuning the three input parameters – jump parameter, number of
iterations and the initial temperature. Second, we focus on the trade-off between the execution time and the
effectiveness of the algorithm. Finally, we compare two objective functions considered in optimization.
3.1 Experiments design
The experiments were performed using four network topologies - Euro28, Euro16, UBN24 and NSF15 (Fig. 1).
In the experiments, we apply an anycast ratio ($5) parameter that is related to a ratio between anycast and
unicast traffic in the network. In detail, we assume that K$Q\ and K8QL denote the overall volume of all anycast and
unicast demands, respectively. Next, let K$OO = K$Q\ + K8QL denote the overall demand in the network. The $5
parameter is defined as the volume (capacity) of all anycast demands divided by the volume of all demands in
the network, LH., $5 = K$Q\ / K$OO.
We run various simulation scenarios. Parameters that we changed in each scenario are as follows: ratio of
anycast to unicast traffic – 0%, 20%, 40%, 60%, 80%, 100% ($5); number of DC (replica) nodes – 1, 2, 3, (4 –
for UBN24 and Euro28) and candidate paths – 2, 3, 5, 10, 30. For each value of $5we generate 5 demand sets,
which were tested for 12 (16 for bigger topologies) different scenarios of location and number of DCs. This gave
us the overall number of 360 (480 for bigger topologies) separate experiments. Since CPLEX can find optimal
results of considered ILP models only for relatively small instances, to obtain optimal results we use only
smaller topologies (NSF15 and Euro16), with overall demand K$OO equal to 2.5 Tbps, and the number of candidate
paths is N = 2. Afterwards we compare the optimal results given by CPLEX 11.0 solver [8] to the following
reference algorithms: FF [3], MSF [5] and LSF [5]. However, for experiments presenting the relationship
between both maximum spectrum and average spectrum, we use all larger topologies with overall demand K$OO
ranging from 40 Tbps to 50 Tbps. Concerning EONs assumptions, we use the half distance law, as in [4], [5] and
[9], for selecting modulation levels for lightpath connections.
3.2 Tuning of the VSA Algorithm
To tune VSA, we examine Euro16 topology for 15 separate cases different in terms of DC nodes numbers,
amount of anycast and unicast traffic and number of candidate paths. We conduct simulations to tune the
following three parameters of VSA algorithm: number of iterations, initial temperature 7calculated according to
ICTON 2014 Th.B3.6
the m parameter and the cooling rate parameter – M Firstly, we tune cooling rate (M) and define initial temperature
using m parameter. We consider all combinations of three cases of M and three cases of m resulting in 9 scenarios
summarized in Table 1.
Each individual result is averaged over 15 cases and 10 repetitions of the algorithm for each case. The number
of iterations is limited to 15,000. As we can see in Table 1, the best combination of tuning parameters is obtained
in scenario S5 – the average optimality gap of VSA for this scenario is 2.12%. According to additional
simulations carried out to determine the number of iterations, we set this parameter to 10,000, since this value
provided the best trade-off between quality of results and execution time. As a result of the tuning process, we
select the following values for further experiments: number of iterations – 10,000,m = 5%,M=0.99. The results
of the tuning process presented above refer to the MaxSpec function, however in the case of the second AvgSpec
function, the results of tuning process were comparable.
3.3 Optimality Gap of VSA in MaxSpec and AvgSpec ILP models
In this Section, we present optimality gaps of VSA and other heuristic algorithms. As we can see in Table 2,
VSA is much better than other algorithms, with stable results in terms of confidence intervals. We can easily
notice that the optimality gap of VSA for the AvgSpec model presents similar trends as for the MaxSpec model.
Regarding the execution time, VSA outperforms CPLEX providing optimal results, LH., for Euro16 network
VSA needs about 60 seconds, while CPLEX requires on average 400 seconds. It should be underlined that VSA
can significantly improve the initial solution provided by MSF (about 10% for MaxSpec and about 20% for
Finally, we report additional results to show the difference between average and maximum spectrum usage of
unicast and anycast traffic in EON, evaluated by VSA for bigger networks UBN24 and Euro28. Note that in
Figs. 2 and 3, the number next to the name of function (MaxSpec/AvgSpec) refers to a number of Data Center
nodes in the network. As we can see in Fig. 2, the difference between two examined objective functions grows
with the increase of anycast traffic ratio, however different performance is observed for various number of DC
nodes, In general, with the increase of anycast traffic ratio, the average spectrum usage decreases.
A corresponding trend for maximum spectrum is different, especially for 2 DCs. This follows mostly from the
fact that maximum spectrum denotes the spectrum usage in the most congested link – in the case of 100% of
anycast traffic and with only 2 nodes, the links adjacent to DC nodes are strongly congested and causes high
values of this performance metric. Nevertheless, in the same case, the average utilization of the spectrum is
much smaller, which is the main advantage of anycast traffic. In Fig. 3, we compare two cases in terms of the
number of DC nodes. We can observe that with the increase of anycast traffic ratio, the difference between 1 and
4 DC nodes significantly grows up to factor of five. This is due to the fact that using more DC nodes decrease
Scenario No. M mOptimality
S1 0.9 10% 2.76%
S2 0.99 10% 2.92%
S3 0.999 10% 3.14%
S4 0.9 5% 2.72%
S5 0.99 5% 2.12%
S6 0.999 5% 2.89%
S7 0.9 2.5% 2.18%
S8 0.99 2.5% 2.21%
S9 0.999 2.5% 2.84%
NSF15 45.1% 13.1% 18.1% 3.8%
Euro16 48.6% 11.5% 14.3% 4.3%
NSF15 2.09% 1.51% 1.78% 0.69%
Euro16 2.15% 1.43% 1.56% 0,92%
NSF15 51.4% 26.7% 22.5% 3.7%
Euro16 50.5% 24.4% 19.0% 4.0%
NSF15 2.42% 1.59% 1.66% 0.91%
Euro16 2.34% 1.02% 1.26% 0,74%
ICTON 2014 Th.B3.6
the lightpaths’ lengths, what is especially visible when there is more anycast traffic, Moreover, this effect is
amplified but possibility to use higher modulation formats which are available to implement as a result of
smaller distance.
In this paper, we have focused on anycast-oriented EONs. In particular, we have proposed a novel metaheuristic
algorithm named VSA. We have presented detailed results related to tuning for the VSA algorithm. To assess the
algorithm performance, we have compared VSA with other reference algorithms and with optimal results
generated by CPLEX for smaller networks. The numerical experiments have shown that VSA outperforms the
reference heuristics as well as demonstrates significantly lower execution times and brings radically better
scalability than CPLEX. Moreover, we have performed additional experiments to show the difference between
average and maximum spectrum usage in EON. Furthermore, we have shown that increasing the number of DCs
can bring savings in the spectrum usage, especially for large values of the anycast ratio. In future work, we plan
to include in our research on EONs other objective functions such as network CAPEX/OPEX cost and power
consumption, as well as develop a SA algorithm for these new optimization problems. Moreover, we would like
to formulate a SA method for optimization of multicast flows in EONs.
This work was supported by The Polish National Science Centre (NCN) under Grant DEC-
2012/07/B/ST7/01215 and statutory funds of the Department of Systems and Computer Networks,
Wroclaw University of Technology.
[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”, ,(((
&RPPXQLFDWLRQV0DJD]LQH vol. 47, no. 11, pp. 66-73, 2009.
[2] B. Kozicki, H. Takara, Y. Tsukishima, T. Yoshimatsu, K. Yonenaga, and M. Jinno, “Experimental
demonstration of spectrum-sliced elastic optical path network (SLICE)”, 2SWLFV([SUHVV, vol. 18, no. 21,
pp. 22105-22118, Oct. 2010.
[3] M. Jinno, H. Takara, B. Kozicki, A. Hirano, Y. Tanaka, Y. Sone, and A. Watanabe, “Distance-adaptive
spectrum resource allocation in spectrum-sliced elastic optical path network”, ,(((&RPPXQLFDWLRQV
0DJD]LQH, vol. 48, no. 8, Aug. 2010.
[4] K. Walkowiak and M. Klinkowski, “Joint anycast and unicast routing for elastic optical networks:
Modeling and Optimization”, in 3URF,(((,QWHUQDWLRQDO&RQIHUHQFHRQ&RPPXQLFDWLRQV,&&,
Budapest, Hungary, Jun. 9-13, 2013.
[5] K. Christodoulopoulos HWDO., “Elastic bandwidth allocation in flexible OFDM based optical networks”,
,(((-/LJKWZ7HFKQRO, vol. 29, no. 9, pp. 1354-1366, May 2011.
[6] S. Kirkpatrick, C.D. Gelatt, and M.P. Vecchi, “Optimization by simulated annealing”, 6FLHQFH vol. 220,
pp. 671-680, May 1983.
[7] K. Walkowiak, R. *RĞFLHĔand M. Klinkowski, “On minimization of the spectrum usage in elastic optical
networks with joint unicast and anycast traffic”, in 3URF$VLD&RPPXQLFDWLRQVDQG3KRWRQLFV&RQIHUHQFH
$&3 Sep. 2013.
[8] ILOG AMPL/CPLEX software:
[9] L. Velasco HWDO., “Modeling the routing and spectrum allocation problem for flexgrid optical networks”,
3KRWRQ1HWZ&RPPXQ, vol. 24, pp. 177-186, Dec. 2012.
... Thus, slot 3 is assigned to it in order to avoid overlapping with d 1 . Finally, the shortest path to satisfy the demand d 3 is the arc v 1 v 2 , but since it has only 2 free slots, then in the proposed solution d 3 is satisfied using the rank [1,3] of an alternative path. ...
... However, there are a few works [163] that assume multiple destinations per demand, i.e., the scenario known as manycast, and also works where the messages are delivered to any one out of a group of nodes, typically the one nearest to the source using a one-to-one-of-many association where datagrams are routed to any single member of a group of potential receivers that are all identified by the same destination address. This last case is known as anycast [3,149,150]. ...
... A few works can be found where the authors apply Lagrangian decompositions and column generation techniques [21,42,43,51,58,75,121,122,147]. The most common approach is to decompose the RSA problem into a routing problem and the SA problem [1,3,7,9,10,12,33,34]. The latter approach corresponds to the interpretation of the RSA problem as an interval coloring problem in an edge-intersection graph, resembling a bandwidth allocation problem already studied in [86,87,88]. ...
Full-text available
One of the most promising solutions to deal with huge data traffic demands in large communication networks is given by flexible optical networking, in particular the flexible grid (flexgrid) technology specified in the ITU-T standard G.694.1. In this specification, the frequency spectrum of an optical fiber link is divided into narrow frequency slots. Any sequence of consecutive slots can be used as a simple channel, and such a channel can be switched in the network nodes to create a lightpath. In this kind of networks, the problem of establishing lightpaths for a set of end-to- end demands that compete for spectrum resources is called the Routing and Spectrum Allocation problem (RSA). Due to its relevance, this problem has been intensively studied in the last decade. It has been shown to be N P-hard, and nontrivial computational techniques are to be applied in order to tackle the solution of practical instances. Since integer programming techniques are known to provide successful practical ap- proaches for several combinatorial optimization problems, the aim of this work is to further the exploration of such an issue, by applying integer linear programming techniques to RSA. The first objective is to explore several integer programming models for RSA, ana- lyzing their effectiveness over known instances. We resort to several modeling techniques in order to find natural formulations of this problem. Having compared these models, then we make a detailed analysis of one of the most promising of them in order to understand its properties, strengths, and weaknesses, taking combinatorial properties into account. Based on these explorations, we develop integer linear programming based computational procedures for RSA including the use of cutting planes and primal heuristics, with the final objective of solving to optimality or near-optimality real-world instances of RSA. This work also contains the development of an instance generator based on real topolo- gies, and the most extensive survey to date listing and classifying the works that treat the RSA and its closest variations from the point of view of combinatorial optimization. Spanish resource:
... It is shown that RSA and RMSA are NP-complete [11,12]. Several Integer Linear Programming (ILP) RSA formulations, heuristics, and metaheuristics are proposed for their solving [4,8,11,[13][14][15][16][17][18][19]. In general, the majority of the proposed ILP models aim at minimizing the total or the maximum number of occupied spectrum slices in a network. ...
... Constraints (17) indicate that loops are forbidden in the network, i.e., links i-j and j-i cannot be the part of the established lightpath concurrently. Constraints (18) enforce that there can be up to one input link a-j and up to one output link j-b for the intermediate node j of the lightpath k-s. These constraints indicate also that there can be up to two adjacent links of the intermediate node j that belongs to the lightpath k-s. ...
Full-text available
In this paper, the problem of content provisioning over elastic optical network (EON) with possible cloud migration is addressed. We perform the analysis that encompasses the problem of content placement into cloud data centers together with lightpath routing and spectrum assignment in EON. We propose a novel mixed integer linear programming model to solve the considered optimization problem. Multiple attribute decision-making methods are used to determine the relevance of each criterion in the objective function. The developed model achieves appropriate placement of content groups into cloud data centers and determines the lightpaths depending on a given bandwidth demand for content provisioning, thus optimizing cloud migration, spectrum utilization and lightpath length, simultaneously. The model is evaluated in two realistic optical network scenarios: the medium-sized INT9 network and the Euro28 network topology. Extensive optimization experiments are performed and the analysis of the network resources utilization is provided. © 2018 Springer Science+Business Media, LLC, part of Springer Nature
... We also tuned the out-of-box MFO, including different spiral curves, considering the unrestricted use of surgical beds between different procedures and the optimization of algorithm speed using different sorting algorithms to improve its performance with success. For future work, we plan to focus on resource-constrained emergency surgery scheduling problems and further optimization of the algorithm itself by introducing a hybrid MFO model connected to Simulating Annealing, similar to [17]. ...
Conference Paper
Full-text available
In this paper, we propose an optimization approach based on an improved Moth Flame Optimization (MFO) algorithm for solving emergency operating room scheduling problems. The purpose of the MFO is to minimize the maximum span of operations, ensuring patients receive their surgeries in a timely manner. This nature-inspired algorithm stimulates the moth's special navigation method at night called transverse orientation. The moth uses the moonlight to sustain a fixed angle to the moon, therefore, guaranteeing a straight line. However, a light source can cause a useless or deadly spiral fly path for moths. The results show that MFO has advantages over Grey Wolf Optimization (GWO) and Genetic Algorithm (GA), particularly when comparing the performance of the algorithms under different spiral curves when considering the unrestricted use of surgical beds between different procedures and the optimization of algorithm speed. Index Terms-cloud computing, moth-flame algorithm, scheduling I. BACKGROUND With increasing congestion in urban areas, the number of seriously injured people is high and healthcare facilities are often overwhelmed. Hence, an efficient surgical scheduling system is needed to maximize operating room efficiency, save the lives of the wounded, and implement appropriate resource allocation. As a cost and revenue center for a hospital, the operating room has a significant impact on the profitability of the entire hospital [1]-[3]. A review of the existing literature confirms the current hospital operating room optimization scheme mainly focuses on optimizing surgical scheduling [4]-[11]. Their contributions are summarized below. First, they try to reschedule the remaining surgical time by taking into account the new information available in each scheduling decision. Next, they built a mathematical model of the rescheduling process at any decision point. The models are proposed for joint optimization of surgical ordering and allocation. Finally, they try to quantify the benefits and analyze the performance of their algorithms. These studies mainly consider reducing operating time through maximal use of operating room resources under normal circumstances (no emergencies). However, there is no report on minimizing the operating time and maximizing resource utilization faced by the numerous injured people that require emergent surgeries. The main contributions of this paper are as follows. We first identify the features of emergent surgeries. We consider the preoperative, intraoperative and postoperative stages. We examine multiple resource constraints and the integration of surgical upstream and downstream resources. Surgical scheduling is usually described as an extended multi-resource, flexible flow-shop scheduling problem [12]. We modelled our batch surgical scheduling using a similar approach. Also, we propose an optimization approach based on an improved Moth Flame Optimization (MFO) algorithm [13]. Finally we check the efficiency of the algorithm through empirical analysis. The entire solution is implemented as a Cloud Computing service, and its benefits are discussed at the end of the paper. To the best of our knowledge, this is the first paper that proposes an MFO algorithm for emergency operating room scheduling purposes [14]. The overall goal is to minimize the maximum span of operations, ensuring patients receive their surgeries in a timely manner. We also improve various aspects of the algorithm, as discussed in the following sections.
... To overcome this difficulty, we solve the problems using algorithms based on simulated annealing. This technique is a randomized, search-based, global optimization algorithm frequently employed to solve network problems [50], [51], [52], [53]. Simulated annealing algorithms do not ensure optimality, but they generally deliver good solutions in a suitable computation time. ...
Full-text available
Network reliability has become an important concern to network administrators and service providers, and is prominently considered in network design. Particularly, 0-day vulnerabilities are an increasing threat to software-based networking systems. When shared between node appliances, they can be exploited simultaneously and compromise large portions of the network. Moreover, it has been observed that the number of 0-day vulnerabilities discovered yearly in node appliances tends to increase over time. Thus, we can expect that the reliability to 0-day exploits of a network implemented with these appliances will also worsen over time. In this work, we treat network reliability to 0-day exploits as a service, where the network provider agrees to deliver a reliability-based level of service over time. We propose a network reliability metric based on network connectivity and discovered appliance vulnerabilities. We formulate a strategy to guarantee a reliability value over time, based on heterogeneous networking and periodically running cost-effective partial node migrations. We use numerical evaluations to test our methodology on two software-defined wide-area networks based on known backbone IP topologies. Our significant findings are the following: First, when the network reliability becomes worse than the service guarantee, it can be restored in most cases by combining appliance reallocation and node migration. Second, our evaluations show a direct relationship between a network reliability value and the cost incurred to guarantee it. Third, we noted that, when using our appliance-to-node allocation strategy to guarantee the same reliability on different networks, their post-failure connectivity depends on the underlying network topology.
... Unlike other algorithms where the routing is addressed first and then followed by spectrum and core assignment [22]- [24], the algorithm we will introduce would first map the network slots to verify the spectrum availability. Then it would find the most efficient way to allocate the resources for each request. ...
Conference Paper
Full-text available
Due to the increase in data transmission worldwide, the capacity of a single-core network will soon not meet the transmission demand of higher traffic. New solutions to optical networks, such as Spatially-Spectrally Flexible Optical Networks, are proposed and investigated to solve this problem. This paper introduces a new algorithm for the Routing, Modulation, Core and Spectrum Assignment (RMCSA) problem based on the adaptive slot mapping procedure. The main goal is to reduce the Bit Rate Blocking (BRB). We compare our approach with three others. The result shows that our algorithm reduces BRB by an average of 20-25%.
... Nos últimos anos, tem-se observado um contínuo crescimento na transferência de dados. Esta alta demanda é ocasionada pelo aumento de assinantes e pelo aparecimento de novos serviços que exigem maiores taxas de transmissão, tais como vídeos sob demanda (VOD, do inglês Video On Demand ), IPTV, jogos 3D on-line, cursos on-line (e-learning) e a computação na nuvem (cloud computing) [1,[7][8][9]. ...
Full-text available
The growth in data traffic is mainly caused due to both the increasing number of subs- criber and new applications that require high transmission rates. Optical networks have been the most suitable way to support this demand. One of the biggest challenges faced by these networks is optimizing the use of the available spectrum. The technologies that may lead to better use of the available optical spectrum is the one that allows to divide the devide the spectrum narrow slices. This approach is known as Elastic Optical Networks (EON), EONS are an evolution of WDM (Wavelength Division Multiplexing) networks. Moreover translucent elastic optical network may be designed from a transparent network by the additon of certain amount of regenerators 3R distributed in a strategically way in the network. In a translucent network, some nodes are transparent and other ones are opaque or translucent. Therefore, in such networks the question of deciding which nodes are translucent or transparent arises. This problem is known as Regenerator Placement (RP). The RP consists of to decide at which of EONs will be deployed 3R regenerators and as well as the number of 3R should be placed at each node. After the RP problem is solved, the regenerator allocation algorithm (RA) is responsible to determine wheter to use, or not, the 3R regenerators deployed in each intermediate node of a given route. In EONs, the use of regenerators can be done by lack of quality in the transmission, by spectrum conversion or to improve modulation format. In this thesis, three algorithms to perform regenerator assignment for EONs are proposed. EONs with sparse regeneration and dynamic traffic are considered. The proposed algorithms are: The FLR-RA (First Longest Reach Regenerator Assignment), which tries to use as less as possible regene- rators along a route, the FNS-RA (First Narrowest Spectrum Regenerator Assignment), which tries to use as less as possible spectrum along a route, and the exhaustive, that returns either the solution that uses the fewest possible number of regenerators or occupy fewest possible number of slots along a given route. The algorithms a simulated in two network topologies. The results show, in the simulated scenarios, that the FLR-RA rea- ches smaller blocking propabilities when the number of regenerators deployed in network is small, while the FNS-RA reaches smaller blocking propabilities in the cases in which many regenerators are placed in network. The FLR-RA and FNS-RA heuristics show similar results to the exhaustive, meaning that the heuristics present a good performance.
... In contrast, using random values took 3666 runs to find an equally optimal solution. SA is an excellent candidate for decreasing BP in regards to both spectrum and regenerators in a dynamic environment as it can quickly create new neighbor values to optimize routing [18]. The temperature can be reset in order to find new optimum values once network traffic patterns change significantly. ...
Conference Paper
Full-text available
The goal of this work is to reduce bandwidth blocking probability in cloud-ready optical networks. We employ data analytics to optimize parameters of Adaptive Modulation and Regenerator-Aware (AMRA) algorithm used for solving the Routing, Modulation, and Spectrum Assignment problem in Elastic Optical Networks. Once data analytics part is finished, we used the acquired knowledge to update the core parameters of AMRA. Our approach allows to optimize the AMRA algorithm execution, thus, enables more traffic provision in the network, using the same resources.
Elastic optical networks (EONs) using optical orthogonal frequency division multiplexing (O-OFDM) technique are considered to be promising optical networks due to their ability to achieve flexible spectrum management. This paper studies dynamic anycast in EONs. The rapid growth of service demand has brought enormous challenges to the network environment, resulting in the increase of blocking rate. In order to reduce the traffic blocking rate, we propose a spectrum-aware ant colony optimization (SA−ACO) algorithm to optimize the anycast transmission problem in EONs and improve the spectrum utilization in the network. Different from the traditional anycast allocation algorithm, the proposed SA−ACO algorithm can adapt to the real-time changes of the system. We add bandwidth factor into the node pseudo-random selection rule, and add a weight factor to weigh the impact of bandwidth on the system. In the found candidate path set, the link distance and link spectrum availability are processed to find the optimal path suitable for the current service transmission. In terms of pheromone update, we delay the secretion of ant pheromone, and secrete pheromone on the path where the relatively optimal solution is located in the solution space found by the ant colony. Through extensive simulation experiments, we evaluate the performance of the algorithm. By comparing with the reference algorithms in previous literature, our proposed algorithm significantly outperforms other methods in reducing bandwidth blocking probability performance.
Conference Paper
Full-text available
The rapid increase in bandwidth demand has driven the development of flexible, efficient, and scalable optical networks. One of the technologies that allows for much more flexible resource utilization is Elastic Optical Network. However, there is a need to solve the Routing, Modulation and Spectrum Assignment (RMSA) problem. In this paper, we use reinforcement learning to improve the efficiency of the routing algorithm. More specifically, we implement an off-policy Q-learning and compare it with the state-of-the-art algorithms. The results confirm that Q-learning is highly effective when optimal results need to be found in a large search space.
Increased global traffic puts tough requirements not just on fiber communications links but on the entire network. This manifests itself in multiple ways, including how to optimize wavelength routing around the network, how to maximize the benefits arising from fine-control DSP with increasingly accurate real-time monitoring, and how to best deploy multiband or multiple fiber connectivity. This article will summarize research into all these areas to present a full picture of how future optical networks will play their role in supporting the continuing traffic demands of broadband, 5G, and associated applications.
Conference Paper
Full-text available
The paper proposes a Tabu Search algorithm for Routing and Spectrum Allocation with joint unicast and anycast traffic in Elastic Optical Networks (EONs). Numerical experiments are presented to verify the algorithm and show main features of EONs with anycast traffic.
Conference Paper
Full-text available
Anycast communication is gaining much interest due to the deployment and the growing importance of new network services, including Content Delivery Networks (CDNs) and grid/cloud computing, among others. Concurrently, elastic optical networks (EONs), being a very promising solution for future optical transport networks, are considered as effective and costefficient approach for supporting such bandwidth-demanding applications. In this article, we focus on anycast communication in EON. Specifically, we formulate a joint anycast and unicast Routing and Spectrum Allocation (RSA/JAU) off-line optimization problem and solve it by means of both Integer Linear Programming (ILP) and dedicated heuristic algorithms. We evaluate the performance of algorithms and assess the effectiveness of anycast communication under various network scenarios. The main conclusion is that anycast routing brings significant spectrum savings in EON.
Full-text available
Flexgrid optical networks are attracting huge interest due to their higher spectrum efficiency and flexibility in comparison with traditional wavelength switched optical networks based on the wavelength division multiplexing technology. To properly analyze, design, plan, and operate flexible and elastic networks, efficient methods are required for the routing and spectrum allocation (RSA) problem. Specifically, the allocated spectral resources must be, in absence of spectrum converters, the same along the links in the route (the continuity constraint) and contiguous in the spectrum (the contiguity constraint). In light of the fact that the contiguity constraint adds huge complexity to the RSA problem, we introduce the concept of channels for the representation of contiguous spectral resources. In this paper, we show that the use of a pre-computed set of channels allows considerably reducing the problem complexity. In our study, we address an off-line RSA problem in which enough spectrum needs to be allocated for each demand of a given traffic matrix. To this end, we present novel integer lineal programming (ILP) formulations of RSA that are based on the assignment of channels. The evaluation results reveal that the proposed approach allows solving the RSA problem much more efficiently than previously proposed ILP-based methods and it can be applied even for realistic problem instances, contrary to previous ILP formulations.
Full-text available
Orthogonal Frequency Division Multiplexing (OFDM) has recently been proposed as a modulation technique for optical networks, because of its good spectral efficiency, flexibility, and tolerance to impairments. We consider the planning problem of an OFDM optical network, where we are given a traffic matrix that includes the requested transmission rates of the connections to be served. Connections are provisioned for their requested rate by elastically allocating spectrum using a variable number of OFDM subcarriers and choosing an appropriate modulation level, taking into account the transmission distance. We introduce the Routing, Modulation Level and Spectrum Allocation (RMLSA) problem, as opposed to the typical Routing and Wavelength Assignment (RWA) problem of traditional WDM networks, prove that is also NP-complete and present various algorithms to solve it. We start by presenting an optimal ILP RMLSA algorithm that minimizes the spectrum used to serve the traffic matrix, and also present a decomposition method that breaks RMLSA into its two substituent subproblems, namely 1) routing and modulation level and 2) spectrum allocation (RML + SA), and solves them sequentially. We also propose a heuristic algorithm that serves connections one-by-one and use it to solve the planning problem by sequentially serving all the connections in the traffic matrix. In the sequential algorithm, we investigate two policies for defining the order in which connections are considered. We also use a simulated annealing meta-heuristic to obtain even better orderings. We examine the performance of the proposed algorithms through simulation experiments and evaluate the spectrum utilization benefits that can be obtained by utilizing OFDM elastic bandwidth allocation, when compared to a traditional WDM network.
Full-text available
We describe experimental demonstration of spectrum-sliced elastic optical path network (SLICE) architecture. We employ optical orthogonal frequency-division multiplexing (OFDM) modulation format and bandwidth-variable optical cross-connects (OXC) to generate, transmit and receive optical paths with bandwidths of up to 1 Tb/s. We experimentally demonstrate elastic optical path setup and spectrally-efficient transmission of multiple channels with bit rates ranging from 40 to 140 Gb/s between six nodes of a mesh network. We show dynamic bandwidth scalability for optical paths with bit rates of 40 to 440 Gb/s. Moreover, we demonstrate multihop transmission of a 1 Tb/s optical path over 400 km of standard single-mode fiber (SMF). Finally, we investigate the filtering properties and the required guard band width for spectrally-efficient allocation of optical paths in SLICE.
Full-text available
There is a deep and useful connection between statistical mechanics (the behavior of systems with many degrees of freedom in thermal equilibrium at a finite temperature) and multivariate or combinatorial optimization (finding the minimum of a given function depending on many parameters). A detailed analogy with annealing in solids provides a framework for optimization of the properties of very large and complex systems. This connection to statistical mechanics exposes new information and provides an unfamiliar perspective on traditional optimization problems and methods.
The sustained growth of data traffic volume calls for an introduction of an efficient and scalable transport platform for links of 100 Gb/s and beyond in the future optical network. In this article, after briefly reviewing the existing major technology options, we propose a novel, spectrum- efficient, and scalable optical transport network architecture called SLICE. The SLICE architecture enables sub-wavelength, superwavelength, and multiple-rate data traffic accommodation in a highly spectrum-efficient manner, thereby providing a fractional bandwidth service. Dynamic bandwidth variation of elastic optical paths provides network operators with new business opportunities offering cost-effective and highly available connectivity services through time-dependent bandwidth sharing, energy-efficient network operation, and highly survivable restoration with bandwidth squeezing. We also discuss an optical orthogonal frequency-division multiplexing-based flexible-rate transponder and a bandwidth-variable wavelength cross-connect as the enabling technologies of SLICE concept. Finally, we present the performance evaluation and technical challenges that arise in this new network architecture.
The rigid nature of current wavelength-routed optical networks brings limitations on network utilization efficiency. One limitation originates from mismatch of granularities between the client layer and the wavelength layer. The recently proposed spectrum-sliced elastic optical path network (SLICE) is expected to mitigate this problem by adaptively allocating spectral resources according to client traffic demands. This article discusses another limitation of the current optical networks associated with worst case design in terms of transmission performance. In order to address this problem, we present a concept of a novel adaptation scheme in SLICE called distance-adaptive spectrum resource allocation. In the presented scheme the minimum necessary spectral resource is adaptively allocated according to the end-to-end physical condition of an optical path. Modulation format and optical filter width are used as parameters to determine the necessary spectral resources to be allocated for an optical path. Evaluation of network utilization efficiency shows that distance-adaptive SLICE can save more than 45 percent of required spectrum resources for a 12-node ring network. Finally, we introduce the concept of a frequency slot to extend the current frequency grid standard, and discuss possible spectral resource designation schemes.