OPTIMIZATION OF ROUTING AND WAVELENGTH ASSIGNMENT IN PASSIVE OPTICAL NETWORKS ROSHNI.V.V. Register No:14MAE015 MASTER OF ENGINEERING

Transcription

1 OPTIMIZATION OF ROUTING AND WAVELENGTH ASSIGNMENT IN PASSIVE OPTICAL NETWORKS A PROJECT REPORT Submitted by ROSHNI.V.V Register No:14MAE015 in partial fulfillment for the award of the degree of MASTER OF ENGINEERING in APPLIED ELECTRONICS Department of Electronics and Communication Engineering KUMARAGURU COLLEGEOF TECHNOLOGY (An autonomous institution affiliated to Anna University, Chennai) COIMBATORE ANNA UNIVERSITY: CHENNAI APRIL 2016

2 BONAFIDE CERTIFICATE Certified that this project report titled OPTIMIZATION OF ROUTING AND WAVELENGTH ASSIGNMENT IN PASSIVE OPTICAL NETWORKS is the bonafide work of ROSHNI.V.V. [Reg. No. 14MAE015] who carried out the research under my supervision. Certified further, that to the best of my knowledge the work reported here in does not form part of any other project or dissertation on the basis of which a degree or award was conferred on an earlier occasion on this or any other candidate. SIGNATURE SIGNATURE Ms.R.HEMALATHA Dr. A.VASUKI PROJECT SUPERVISOR PROFESSOR AND HEAD Associate Professor Department of ECE Department of ECE Kumaraguru College of Technology Kumaraguru College of Technology Coimbatore Coimbatore The Candidate with Register No. 14MAE015 was examined by us in the project viva voice examination held on... INTERNAL EXAMINER EXTERNAL EXAMINER ii

3 ACKNOWLEDGEMENT First, I would like to express my praise and gratitude to the Lord, who has showered his grace and blessings enabling me to complete this project in an excellent manner. I express my sincere thanks to the management of Kumaraguru College of Technology and Joint Correspondent Mr. Shankar Vanavarayar for the kind support and for providing necessary facilities to carry out the work. I would like to express my sincere thanks to our beloved Principal Dr.R.S.Kumar Ph.D., Kumaraguru College of Technology, who encouraged me in each and every steps of the project. I would like to thank Dr. A. Vasuki Ph.D., Head of the Department, Electronics and Communication Engineering, for her kind support and for providing necessary facilities to carry out the project work. In particular, I wish to thank with everlasting gratitude to the Project Coordinator Ms.S.Umamaheswari M.E.(Ph.D), Associate Professor, Department of Electronics and Communication Engineering,for her expert counseling and guidance to make this project to a great deal of success. I am greatly privileged to express my heartfelt thanks to my Project Guide Ms.R.Hemalatha M.E.(Ph.D), Associate Professor, Department of Electronics and Communication Engineering, throughout the course of this project work and I wish to convey my deep sense of gratitude to all teaching and non-teaching staff of ECE Department for their help and cooperation. Finally, I thank my parents and my family members for giving me the moral support and abundant blessings in all of my activities and my dear friends who helped me to endure my difficult times with their unfailing support and warm wishes. iii

4 ABSTRACT Routing and Wavelength Assignment (RWA) problem is one of the important optimization problems in optical networks. RWA problem are of two types, static and dynamic. In static RWA the set of connections is known in advance where as in dynamic RWA connection request arrive sequentially. In the proposed work the dynamic routing and wavelength assignment problem is examined. The goal is to minimize the number of wavelengths and blocking probability. Evolutionary programming algorithms are used to optimize the routing and wavelength assignment. The RWA problem can be fixed by number of algorithms like PSO, ACO etc. Genetic Algorithm and Shuffled Frog Leaping Algorithm (SFLA) have been implemented in optical networks to fix the RWA problem. Cost, number of wavelengths, hop count and blocking probability are the optimization parameters. In WDM network, for the given set of connection requests, routing and wavelength assignment problem involves the task of establishing lightpaths (routing) and assigning a wavelength to each connection request. The problem is analyzed for different wavelength assignment methods such as first fit, random, round robin, wavelength ordering and FWM priority based assignment. Fitness function is calculated in terms of cost, number of wavelengths, hop count and setup time. The experimental result shows that the network has better blocking performance when using Shuffled Frog Leaping Algorithm with FWM aware priority based wavelength assignment. SFLA algorithm produces less blocking probability, less cost and less computational complexity than existing methods. iv

5 TABLE OF CONTENTS CHAPTER NO. TITLE PAGE NO. ABSTRACT LIST OF TABLES LIST OF FIGURES LIST OF ABBREVIATIONS iv vii viii ix 1. INTRODUCTION OPTICAL AMPLIFIERS WDM ROUTING AND WAVELENGTH ASSIGNMENT GENETIC ALGORITHM SHUFFLED FROG LEAPING ALGORITHM 4 2. LITERATURE SURVEY 5 3. EVOLUTIONARY PROGRAMMING METHOD BLOCK DIAGRAM NETWORK MODEL ROUTING MODEL Fixed Path Routing Fixed Alternate Routing Adaptive Routing Traditional Adaptive RWA Physically Aware Adaptive RWA WAVELENGTH ASSIGNMENT MODEL GENETIC ALGORITHM Flow Chart of Genetic Algorithm 20 v

6 3.5.2 Create a Random Initial Population Evaluate Fitness Produce Next Generation Next Generation or Termination Advantages of Genetic Algorithm SHUFFLED FROG LEAPING ALGORITHM RESULTS AND DISCUSSIONS FITNESS FUNCTION MEAN BLOCKING PROBABILITY(W.R.T. CHANNEL REJECTION RATIO) AVERAGE FITNESS SCORE MEAN BLOCKING PROBABILITY(W.R.T. GENERATIONS) MEAN EXECUTION TIME COMPARISON OF PERFORMANCE MEASURES IN GA AND SFLA CONCLUSION 39 REFERENCES 40 LIST OF PUBLICATIONS 43 vi

7 LIST OF TABLES TABLE NO. TITLE PAGE NO. 4.1 Fitness function of GA and SFLA with respect to time Mean blocking probability (wr.t. channel rejection ratio) of GA Mean blocking probability (wr.t. channel rejection ratio) of SFLA Average fitness score for GA and SFLA Mean blocking probability (wr.t. Generations) of GA Mean blocking probability (wr.t. Generations) of SFLA Mean execution time of GA Mean execution time of SFLA Comparison of different wavelength assignment techniques with respect to four performance measures in GA 4.10 Comparison of different wavelength assignment techniques with respect to four performance measures in SFLA vii

8 LIST OF FIGURES FIGURE NO. TITLE PAGE NO. 1.1 Principle of AON Principle of PON Principle of WDM Block diagram of evolutionary method Architecture of a wavelength routing node Model of a transmission lightpath Flow chart of GA Flow chart of SFLA Fitness function of GA and SFLA Mean blocking probability(w.r.t. channel 26 rejection ratio) of GA 4.3 Mean blocking probability(w.r.t. channel 28 rejection ratio) of SFLA 4.4 Average fitness for fixed network Mean blocking probability(w.r.t. generations) of 31 GA 4.6 Mean blocking probability(w.r.t. generations) of 32 SFLA 4.7 Mean execution time of GA Mean execution of SFLA 35 viii

9 LIST OF ABBREVIATIONS AON PON WDM OLT ONT RWA DRWA EDFA ABC GA GASP FA GOF SFLA WRN XCS ILP PABR FWM Active Optical Network Passive Optical Network Wavelength Division Multiplexing Optical Line Terminal Optical Network Terminal Routing and Wavelength Assignment Dynamic Routing and Wavelength Assignment Erbium Doped Fiber Amplifier Artificial Bee Colony Genetic Algorithm Grooming Adaptive Shortest Path Algorithm Firefly Algorithm Generic Objective Function Shuffled Frog Leaping Algorithm Wavelength Routing Node Cross Connect Switch Integer Linear Program Physically Aware Backward Reservation Four Wave Mixing ix

10 CHAPTER 1 INTRODUCTION 1.1 OPTICAL NETWORKS Optical networks are high capacity telecommunication networks based on optical technologies and components that provide routing, grooming and restoration at the wavelength level as well as wavelength based services. Fiber optics uses light signals to transmit data. As this data moves across a fiber, there needs to be a way to separate it so that it gets to the proper destination. There are two important types of systems that make fiber-to-the-home broadband connections possible. These are active optical networks and passive optical networks. Each offers ways to separate data and route it to the proper place, and each has advantages and disadvantages as compared to the other [12]. An active optical system uses electrically powered switching equipment, such as a router or a switch aggregator, to manage signal distribution and direct signals to specific customers. This switch opens and closes in various ways to direct the incoming and outgoing signals to the proper place. In such a system, a customer may have a dedicated fiber running to his or her house. Fig 1.1 shows the principal of Active Optical Network (AON). A passive optical network, on the other hand, does not include electrically powered switching equipment and instead uses optical splitters to separate and collect optical signals as they move through the network. A passive optical network shares fiber optic strands for portions of the network. Powered equipment is required only at the source and receiving ends of the signal. Fig 1.2 shows the principle of Passive Optical Network (PON). Fig 1.1 Principle of AON 1

11 Fig 1.2 Principle of PON 1.2 WDM In fiber-optic communications, wavelength division multiplexing (WDM) is a technology which multiplexes a number of optical carrier signals onto a single optical fiber by using different wavelengths (i.e., colors) of laser light. This technique enables bidirectional communications over one strand of fiber, as well as multiplication of capacity. The term wavelength-division multiplexing is commonly applied to an optical carrier (which is typically described by its wavelength). Fig 1.3 Principle of WDM A WDM system as shown in Fig 1.3 uses a multiplexer at the transmitter to combine the signals together from different sources operating at different wavelengths and a demultiplexer at the receiver to split them apart. With the right type of fiber it is possible to have a device that does both simultaneously and can function as an optical add-drop multiplexer. Wavelength-division multiplexing (WDM) have high band width demand. Traffic grooming, Optimal routing and wavelength assignment, 2

12 survivability, Quality of service (QoS) routing, physical layer impairment aware (PLI aware) routing and wavelength assignment are different problems that exist in optical wavelength division multiplexing (WDM) [1]. 1.3 ROUTING AND WAVELENGTH ASSIGNMENT (RWA) In the WDM networks, there is a tight coupling between routing and wavelength selection. A path of links between the source and destination nodes is selected and a particular wavelength on each of these links is reserved for the lightpath. Thus for establishing an optical connection select a suitable path and allocate an available wavelength for the connection. The resulting problem is called routing and wavelength allocation (RWA) problem. The routing and wavelength allocation problem is subject to the following two constraints: Wavelength continuity constraint and distinct wavelength constraint. There are two variations in the problem: Static RWA : The traffic requirements are known in advance Dynamic RWA: The sequence of lightpath requests arrive in some random fashion. The methods that have been employed to solve RWA problem include classical approaches and heuristics or metaheuristics-based approaches. Conventional techniques are able to give accurate results for simple problems. But to solve complex problems, these techniques have too much computational time [3]&[6]. Multiobjective evolutionary algorithms are used to solve the RWA problem which is based on swarm intelligence in real-world optical networks [4]&[5]. 1.4 GENETIC ALGORITHM Genetic algorithm (GA) is a search algorithm based on the mechanics of natural selection and natural genetics. GA works with individuals, each representing a solution to the problem being tackled. A fitness function is defined in order to estimate the goodness of a solution. An initial population of individuals is created and then evolved by means of genetic operators, such as cross over and mutation, to form a new population (the next generation) that is hoped to be fitter than the last one. The evolution process is repeated a predefined number of iterations or until another criterion is met [2]&[7]. 3

13 The crossover operator is applied to pairs of individuals in order to interchange their genetic material, imitating natural reproduction. By applying this operator to the fittest individuals, good properties should propagate down the generations. The mutation operator makes a random change in genetic material of a single individual, allowing the GA to explore new corners of the search space. Since individuals from the population become fitter throughout the generations, the final population will contain an optimal or near optimal solution. 1.5 SHUFFLED FROG LEAPING ALGORITHM Shuffled frog leaping algorithm (SFLA) is a meta-heuristic optimization method which is based on observing, imitating and modeling the behavior of a group of frogs when searching for the location that has the maximum amount of available food. Shuffled frog leaping algorithm is a population based random search algorithm inspired by nature memetics. Instead of using genes in GA, SFLA uses memes to improve spreading and convergence ratio. Meme is a contagious information pattern that alters human/animal behavior. The actual contents of a meme, called memotype, are analogous to the genes of a chromosome. The main difference between a gene and a meme is related to its transmission ability. Genes can only be transmitted from parents or a parent in the case of asexual reproduction to offspring. Memes can be transmitted between any two individuals. SFLA combines the benefit of the local search tool of Particle Swarm Optimization (PSO) and the idea of mixing information from parallel local searches, to move towards a global solution [8]. The whole population of frogs is distributed within a different subset called a memeplex. Each memeplex is considered a different culture of frogs, performing an independent local search. After a defined number of memetic evolutionary steps, frogs are shuffled among memeplexes, enabling frogs to interchange messages among different memeplexes and ensuring that they move to an optimal position, similar to particles in PSO. The local search and the shuffling processes continue until defined convergence criteria are satisfied [9]. 4

14 CHAPTER 2 LITERATURE SURVEY This chapter deals with review of literature about routing and wavelength assignment using several natural inspired algorithms and comparisons in terms of performance and computational complexity. [1] A Metaheuristic Approach for Optical Network Optimization Problems Urmila Bhanjaa et al proposed a metaheuristic approach for optical network optimization problems such as QoS routing and DRWA problems in their paper. Genetic algorithm is used to solve different optimization problems by designing problem specific fitness functions. The initial search space is very small since the initial population consists of only a single chromosome. The evolutionary algorithm depends on the mutation operator alone for creating and exploring the search space.encodings of the chromosomes are random and simple and these are of variable length. [2] Distributed Grooming, Routing, and Wavelength Assignment for Dynamic Optical Networks Using Ant Colony Optimization X. Wang et al made a comparison between ant colony optimization algorithm and a centralized heuristic algorithm, a grooming adaptive shortest path algorithm (GASP) for routing and wavelength assignment in optical networks. In their examination although GASP shows better efficiency in terms of blocking probability, ACO shows great robustness and adaptivity to varying network and traffic conditions. 5

15 [3] A Comparative Study on Multi objective Swarm Intelligence for the Routing and Wavelength Assignment Problem A lvaro Rubio-Largo presented a comparative study on swarm intelligence to solve the RWA problem. They have evaluated three multi objective metaheuristic based on the behavior of honey bees (MO-ABC), on the law of gravity and mass interactions (MO-GSA), as well as on the flash pattern of fireflies (MO-FA).They concluded that MO-FA is a very suitable approach to solve the RWA problem. [4] Routing and wavelength assignment in optical networks using Artificial Bee Colony algorithm Yousef S. Kavian et al introduced Artificial Bee Colony algorithm for routing and wavelength assignment. Every food source represented one of the K possible and feasible paths between each node pair in optical network. The positions of food sources were modified by artificial bees and evaluated by the fitness function. In their analysis ABC is faster than GA to solve RWA problem in real-world EUROPEAN and NSFNET optical networks. [5] Evolutionary Algorithms for Solving Routing and Wavelength Assignment Problem in Optical Networks: A Comparative Study Arash Rashedi et al described applications of some intelligent algorithms such as GA, PSO and ABC algorithms for solving routing and wavelength assignment problem in optical networks. The performances of proposed algorithms were compared for both convergence speed and accuracy using NSFNET test-bench network considering randomly generated connection requests. The convergence speed of ABC algorithm is much better than other two algorithms to reach near-optimum solution in acceptable processing time. 6

16 [6] A New Proposal of an Efficient Algorithm for Routing and Wavelength Assignment in Optical Networks Afonso Jorge F. Cardoso presented a RWA algorithm based on a Generic Objective Function (GOF) which aims to establish a base from which it is possible to develop a standard or multiple standards for optical networks. The GOF algorithm introduces the concept of implicit constraint, which guarantees a simple solution to a problem not as trivial as the RWA. In GOF no restriction point is considered, i.e. no explicit restrictions are considered. RWA is solved and creates the possibility that the algorithm GOF serve as a standard for optical WDM networks. [7] Shuffled Frog-Leaping Algorithm: A Memetic Meta-heuristic for Discrete Optimization Muzaffar et al proposed a new efficient natural inspired metaheuristic approach, called Shuffled Frog Leaping Algorithm (SFLA) for discrete optimization. SFLA is a population-based method and uses a population of solutions to proceed to the global solution. SFLA is a population-based cooperative search metaphor inspired by natural memetics. The algorithm contains elements of local search and global information exchange. The SFLA consists of a set of interacting virtual population of frogs partitioned into different memeplexes. The virtual frogs act as hosts or carriers of memes where a meme is a unit of cultural evolution. The algorithm performs simultaneously an independent local search in each memeplex. SFLA performance was compared with a GA for a series of test problems. The results for 11 theoretical test problems (functions) and two applications show that the SFLA performed better than or at least comparable with the GA for almost all problem domains and was more robust in determining the global solution. Four realistic engineering problems were also solved and results compared with literature results from several optimization algorithms. 7

17 [8] A Modified Shuffled Frog Leaping Algorithm for Long-Term Generation Maintenance Scheduling G. Giftson Samuel et al discussed a modified Shuffled frog leaping algorithm to Long term Generation Maintenance Scheduling to Enhance the Reliability of the units. The algorithm has been tested on thirty two generating unit system. The proposed method has been compared with other methods. The result obtained is compared with the results of other method such as DP, LR and PSO. From the result it is shown that the proposed algorithm provides true optimal solution for minimum fuel cost and computation timing in all cases. [9] An Ant-Based Algorithm for Distributed Routing and Wavelength Assignment in Dynamic Optical Networks Joan Triay et al proposed the use of an ant colony optimization (ACO) algorithm to solve the intrinsic problem of the routing and wavelength assignment (RWA) on wavelength continuity constraint optical networks. In optical burst switching the forward ants are implemented as burst control packets, whereas feedback ants, which gather information about the positive or negative delivery of the bursts, are a special type of acknowledgment control packet. The algorithm takes into account both the path length and the congestion in the network to update the values of the pheromone trails. It has been evaluated through extensive simulations with very promising results, particularly on highly congested scenarios where the load balancing capabilities of the protocol become especially efficient. 8

18 [10] A Genetic Algorithm for Shortest Path Routing Problem and the Sizing of Populations C.W. Ahn et al presented a genetic algorithmic approach to the shortest path (SP) routing problem. Crossover and mutation together provide a search capability that results in improved quality of solution and enhanced rate of convergence. The crossover is simple and independent of the location of crossing site. Consequently, the algorithm can search the solution space in a very effective manner. The mutation introduces, in part, a new alternative route. The population-sizing equation appears to be a conservative tool to determine a population size in the routing problem. [11] A Novel Solution to the Dynamic Routing and Wavelength Assignment Problem in Transparent Optical Networks Urmila Bhanjaa et al discussed an evolutionary programming algorithm for solving the dynamic routing and wavelength assignment (DRWA) problem in optical wavelength-division-multiplexing (WDM) networks under wavelength continuity constraint. They assume an ideal physical channel and therefore neglect the blocking of connection requests due to the physical impairments. They implemented three types of wavelength assignment techniques, such as First fit, Random, and Round Robin wavelength assignment techniques. 9

19 CHAPTER 3 EVOLUTIONARY PROGRAMMING METHOD 3.1 BLOCK DIAGRAM Fig 3.1 Block Diagram of Evolutionary Programming Method The evolutionary programming method is organized in four models namely Network model, Routing model, Wavelength Assignment model and Optimization Algorithm [10] as shown in Fig NETWORK MODEL The N node network can be modeled as a graph G (V, E), in which V is the set of nodes representing routers or switches and E is the set of edges representing connectivity between the nodes. The link existing between a pair of nodes is assumed to be bidirectional in nature, that is, the existence of a link e = (i,j) from node i to j implies the existence of another link e = (j,i) for any pair of nodes (i,j) E. For the DRWA problem with ideal as well as non-ideal physical layers, V is the set of nodes representing routers or WRNs, and E is the set of fiber links representing physical connectivity between the nodes. Each link is assumed to be bidirectional with fixed number of wavelengths per fiber. For the physical impairment aware DRWA problem, each wavelength routing node (WRN) consists of a cross connect switch (XCS), 10

20 transmitter and receiver arrays, optical taps and erbium doped fiber amplifiers (EDFA) as in Fig.3.2. Fig 3.2 Architecture of a wavelength routing node The wavelength routing switches (WRSs) in the XCS are assumed to employ nonblocking active splitter/combiner architecture. The XCSs transfer each wavelength in an input fiber into the same wavelength in one of the output fibers. A tap is present at the input and output of each XCS to monitor the signal condition [16]. The EDFA at the input side compensates for the fiber loss and the tap loss and the EDFA at the output side compensates for the switch loss. Fig 3.3 Model of a transmission lightpath 11

21 In Fig. 3.3, WRN (1) represents the source node, WRN (m) represents the destination node, and WRN (k) represents the kth intermediate node. Array of transmitters and receivers are present in each of the nodes for locally adding or dropping the traffic. In the adopted network model, each XCS consists of an array of demultiplexers followed by a set of WRSs and a set of multiplexers. All the signals that are demultiplexed and have identical wavelengths are directed to the corresponding WRS tuned to the same wavelength and the switch redirects the signal to the desired output port; the multiplexers then combine signals with different wavelengths and redirect them to the output fibers. The number of WRSs in an XCS depends on the number of input wavelengths, and the number of input and output ports of a WRS depends on the number of input and output fibers. 3.3 ROUTING MODEL The routing and wavelength assignment (RWA) problem is an optical networking problem with the goal of maximizing the number of optical connection. Each connection request must be given a route and wavelength. The wavelength must be consistent for the entire path, unless the usage of wavelength converters is assumed. Two connection requests can share the same optical link, provided a different wavelength is used [11]&[17]. The routing models used for all the four problems are nearly identical. For the QoS constrained routing problem F is assumed to be the set of flows existing at any time and f is any unicast flow or request belonging to F. The variable I f i,j is set to 1 if link (i,j) is used by flow f; otherwise, it is set to 0. A Path from the source S to destination D for a flow f is represented as Path (f) and is the collection of all the links belonging to the flow from S to D. Any link e E has a bandwidth, bandwidth (e): E R+, associated with it, R+ being the set of positive real numbers. The bandwidth from any source to any destination, for a flow f, is denoted by bandwidth(f), and is defined as: 12

22 bandwidth (f) = min {bandwidth (e) e Path(f )}, f F Fitness function is to maximize f W W W x x x x kx 1 x Hi, j Tx C gx( j), gx( j 1) ( i, j) E j 1 (3.1) In the fitness function, W x is the free wavelength factor. If the same wavelength is available in all links of the path x then it is one and zero otherwise. The other term in the fitness function defines the sum of the link costs in the path. The denominator of the second term represents the total number of hops the path passes through. The variable H x i,j equals one if link (i, j) is a part of path x; otherwise, it is equal to zero. The variable T x represents the set up time of path x. The variable k x represents the length of the x-th chromosome or number of memeplexes. A route is considered to be optimal when it maximizes this objective function while satisfying the following constraints: lp I Iij 1, if i=s, lp LP (3.2) lp ij ( i, j) E ( j, i) E lp I Iij 1, if i=d, lp LP (3.3) lp ij ( i, j) E ( j, i) E lp I Iij 0, if i S, i D, lp LP (3.4) lp ij ( i, j) E ( j, i) E lp Iij 1, if i D, lp LP (3.5) i j ( i, j) E lp Iij 0, if i=d, lp LP (3.6) i j ( i, j) E ( i, j) E I lp ij h, for t T 0 13 (3.7)

23 lp h0 Iij ( N 1), for t > T (3.8) ( i, j) E The flow conservation constraint, given from equation (3.2) to (3.4) and the loop constraint from equation (3.5) to (3.6) guarantee that the solutions obtained represent valid paths from S to D and that the lightpath has no loops. Equations (3.7) and (3.8) represent the hop count constraint, which, however, is a soft constraint. For a threshold time T, the number of hops traversed by the lightpath is initially limited to prevent excessively long paths causing delay and is required to be less than or equal to an upper bound h0. However after this time, the bound is relaxed up to a maximum value of (N - 1). In the initialization phase of the proposed algorithm, the threshold time is set to T1 and during the mutation phase, the threshold time is set to T2 [18]. The variable, t represents the algorithm execution time. Since dynamic RWA is more complex than static RWA, it must be the case that dynamic RWA is also NP-complete. The RWA problem is further complicated by the need to consider signal quality. Many of the optical impairments are nonlinear, so a standard shortest path algorithm can't be used to solve them optimally even if we know the exact state of the network. This is usually not a safe assumption, so solutions need to be efficient using only limited network information. Given the complexity of RWA, there are two general methodologies for solving the problem: The first method is solving the routing portion first, and then assigning a wavelength. Three types of route selection are Fixed Path Routing, Fixed Alternate Routing and Adaptive Routing. The second approach is to consider both route selection and wavelength assignment jointly. 14

24 3.3.1 Fixed path routing Fixed path routing is the simplest approach to find a lightpath. The same fixed route for a given source and destination pair is always used. Typically this path is computed ahead of time using a shortest path algorithm, such as Dijkstra's Algorithm. While this approach is very simple, the performance is usually not sufficient. If resources along the fixed path are in use, future connection requests will be blocked even though other paths may exist. The SP-1 (Shortest Path, 1 Probe) algorithm is an example of a Fixed Path Routing solution. This algorithm calculates the shortest path using the number of optical routers as the cost function. A single probe is used to establish the connection using the shortest path. The running time is the cost of Dijkstra's algorithm: O(m + nlog n), where is the number of edges and is the number of routers. The running time is just a constant if a predetermined path is used. This definition of SP-1 uses the hop count as the cost function. The SP-1 algorithm could be extended to use different cost functions, such as the number of EDFAs Fixed alternate routing Fixed alternate routing is an extension of fixed path routing. Instead of having just one fixed route for a given source and destination pair, several routes are stored. The probes can be sent in a serial or parallel fashion. For each connection request, the source node attempts to find a connection on each of the paths. If all of the paths fail, then the connection is blocked. If multiple paths are available, only one of them would be utilized. The SP-p (Shortest Path, p Probes, p>1) algorithm is an example of Fixed Alternate Routing. It calculates the p shortest paths using the number of optical routers as the cost function. The running time using is O(pn m + nlog n ) where m is number of edges, n is the number of routers, and p is the number of paths. The running time is a constant factor if the paths are precomputed. 15

25 3.3.3 Adaptive routing The major issue with both fixed path routing and fixed alternate routing is that neither algorithm takes into account the current state of the network. If the predetermined paths are not available, the connection request will become blocked even though other paths may exist. Fixed Path Routing and Fixed Alternate Routing are both not quality aware. For these reasons, most of the research in RWA is currently taking place in Adaptive algorithms. Adaptive algorithms fall into two categories: traditional and physically aware. Traditional adaptive algorithms do not consider signal quality, however, physically aware adaptive algorithms do Traditional adaptive RWA The routing algorithm is to route connection requests away from congested areas of the network, increasing the probability that connection requests will be accepted. This is accomplished by setting the cost of each link to be cost l = β usage (l) where is parameter that can be dynamically adjusted according to the traffic load and usage(l) is the number of wavelengths in use on link. A standard shortest path algorithm can then be used to find the path. This requires each optical switch to broadcast recent usage information periodically. Note that LORA does not consider any physical impairment. When is equal to one, the LORA algorithm is identical to the SP algorithm. Increasing the value of will increase the bias towards less used routes. The optimal value can be calculated using the well-known hill climbing algorithm. The optimal values of were between 1.1 and 1.2 in the proposal Physically aware adaptive RWA The physically aware backward reservation algorithm (PABR) is an extension of LORA. PABR is able to improve performance in two ways: considering physical impairments and improved wavelength selection. As PABR is searching for an optical path, paths with an unacceptable signal quality due to linear impairments are pruned. In 16

26 other words, PABR is LORA with an additional quality constraint. PABR can only consider linear impairments. The nonlinear impairments, on the other hand, would not be possible to estimate in a distributed environment due to their requirement of global traffic knowledge. PABR also considers signal quality when making the wavelength selection. It accomplishes this by removing from consideration all wavelengths with an unacceptable signal quality level. The approach is called Quality First Fit. 3.4 WAVELENGTH ASSIGNMENT MODEL Two of the most common methods for wavelength assignment are First Fit and Random Fit. First Fit chooses the available wavelength with the lowest index. Random Fit determines which wavelengths are available and then chooses randomly amongst them. The complexity of both algorithms is Ow, ( ) where w is the number of wavelengths. First Fit outperforms Random Fit. An extension to First Fit and Random Fit was proposed in to consider signal quality. Quality First Fit and Quality Random Fit eliminate from consideration wavelengths which have an unacceptable signal quality. The complexity of these algorithms is higher though, as up to calls to estimate the Q-factor are required. There are several other wavelength assignment algorithms: Least Used, Most Used, Min Product, Least Loaded, Max Sum, and Relative Capacity Loss. Most Used outperforms Least Used significantly, and slightly outperforms First Fit. Min Product, Least Loaded, Max Sum, and Relative Capacity Loss all try to choose a wavelength that minimizes the probability that future requests will be blocked. A significant disadvantage of these algorithms is that they require a significant communication overhead, making them impractical to implement unless you have a centralized network structure. In the proposed fitness function, a free wavelength factor, W X, is updated after the wavelength assignment phase. In the wavelength assignment model, the variable I ij lp is equal to one when the link (i, j) is used by the lightpath lp, and zero otherwise. The 17

27 additional variables used are, I lp ijw, the lightpath wavelength indicator that shows whether the lightpath lp uses wavelength W on link (i, j) and I lp(x,y) ijw, the lightpath wavelength link indicator that is one when the lightpath uses wavelength W on link (i, j) between the nodes x and y, and, l (x,y) which equals one if a physical link exists between the nodes x and y. The wavelength continuity constraints are I lp ij W 1 I, (i,j) (3.9) w 0 lp ijw I lp( x, y) ijw I, (i,j), (x,y), w (3.10) lp ijw lp( x, y) Iijw 1, (x,y), w (3.11) i, j W 1 W 1 lp( x, y) ( x, y) lp( y, x) ( y, x) lp Iijw l Iijw l Iij w 0 x w 0 x, y=j (3.12) W 1 W 1 lp( x, y) ( x, y) lp( y, x) ( y, x) lp Iijw l Iijw l Iij w 0 x w 0 x, y=i (3.13) W 1 W 1 lp( x, y) ( x, y) lp( y, x) ( y, x) Iijw l Iijw l w 0 x w 0 x 0, y i, y j (3.14) The binary variable lp I ijw is the lightpath wavelength indicator, which is one lp( x, y) whenever the lightpath lp uses wavelength w on link (i, j); I is another binary variable called the lightpath wavelength link indicator, which is one when the lightpath lp between the nodes x and y uses wavelength w on link (i, j); and the variable l ( xy, ) is one if a physical link exists between nodes x and y; otherwise, it is zero. Equations (3.9) and (3.10) together imply that the wavelength used by a lightpath is unique. Equation (3.9) dictates that the same wavelength is used in all the links traversed by a lightpath. On the 18 ijw

28 other hand, equation (3.10) implies that only a single lightpath using the link (i, j) can use the wavelength w. Equation (3.11) guarantees that two lightpaths using the same link are not assigned the same wavelength, and equations from (3.12) to (3.14) ensure the conservation of wavelengths at the end nodes of the physical links traversed by a lightpath. 3.5 GENETIC ALGORITHM In the field of artificial intelligence, genetic algorithm (GA) is a search heuristic that mimics the process of natural selection. This heuristic (also sometimes called a metaheuristic) is routinely used to generate useful solutions to optimization and search problems. Genetic algorithms belong to the larger class of evolutionary algorithms (EA), which generate solutions to optimization problems using techniques inspired by natural evolution, such as inheritance, mutation, selection, and crossover [13]. Genetic algorithms are not too hard to program or understand, since they are biological based. Thinking in terms of real-life evolution helps. The general algorithm for a GA is: Generate a large set of possible solutions to a given problem (initial population) Evaluate each of those solutions, and decide on a "fitness level" ("survival of the fittest") From these solutions breed new solutions (the next generation) o The parent solutions that were more "fit" are more likely to reproduce o While those that were less "fit" are more unlikely to do so Solutions are evolved over time, by repeating the process each generation. Terminate when a solution has been found or other termination criteria has been met 19

29 3.5.1 Flow Chart of Genetic Algorithm Creation of Initial Population (single chromosome generation) Mutation (Offspring generation from single chromosome) Selection of one best chromosome depending on the fitness function Is termination criteria reached? Stop Fig 3.4 Flow chart of Genetic Algorithm Fig.3.4 shows the flow of the genetic algorithm to solve the routing and wavelength assignment problem Create a Random Initial Population An initial population is created from a random selection of solutions. These solutions have been seen as represented by chromosomes as in living organisms.the genetic information defines the behaviour of the individual. A chromosome is a packet of 20

30 genetic information organized in a standard way that defines completely an individual (solution). The genetic principles (way in which that information encodes the individual) enable the individuals to evolve in a given environment. The genetic structure (way in which that information is packed and defined) enables the solutions to be manipulated. The genetic operands (way in which that information can be manipulated) enable the solutions to reproduce and evolve [14]&[15] Evaluate Fitness A value for fitness is assigned to each solution (chromosome) depending on how close it actually is to solve the problem. Therefore, define the problem, model it and simulate it or have a data set as sample answers. Each possible solution has to be tested in the problem and the answer is evaluated (or marked) on how good it is. The overall mark of each solution relative to all the marks of all solutions produces a fitness ranking Produce Next Generation Those chromosomes with a higher fitness value are more likely to reproduce offspring. The population for the next Generation will be produced using the genetic operators. Reproduction is by Copying or Crossing Over and Mutation is applied to the chromosomes according to the selection rule. This rule states that the fitter and individual is, the higher the probability it has to reproduce Next Generation or Termination If the population in the last generation contains a solution that produces an output that is close enough or equal to the desired answer then the problem has been solved. This is the ideal termination criterion of the evolution. If this is not the case, then the new generation will go through the same process as their parents did and the evolution will continue. This will iterate until a solution is reached or another of the termination criteria is satisfied. A termination criterion that always must be included is Time-Out (either as computing time or as number of generations evaluated).since one drawback of 21

31 Evolutionary Programming is that, it is very difficult (impossible most of the time) to know if the ideal termination criterion is going to be satisfied and/or when Advantages of Genetic Algorithm It can solve every optimisation problem which can be described with the chromosome encoding. It solves problems with multiple solutions. Since the genetic algorithm execution technique is not dependent on the error surface, it can solve multi-dimensional, non-differential, non-continuous, and even non-parametrical problems. Structural genetic algorithm gives the possibility to solve the solution structure and solution parameter problems at the same time by means of genetic algorithm. Genetic algorithm is a method which is very easy to understand and it practically does not demand the knowledge of mathematics. Genetic algorithms are easily transferred to existing simulations and models. 3.6 SHUFFLED FROG LEAPING ALGORITHM Shuffled Frog Leaping Algorithm (SFLA) is a natural inspired metaheuristic algorithm. The most distinguished benefit of SFLA is its fast convergence speed.the Shuffled frog leaping algorithm combines the advantages of the both the genetic-based memetic algorithm and the behavior-based Particle Swarm Optimization(PSO) algorithm. In the Shuffled frog leaping algorithm, possible solutions are defined by a group of frogs which is referred to as population. The group of frogs is partitioned into several communities referred to as memeplexes. Local search is performed by each frog in the memeplexes. The individual frog s behavior can be influenced by behaviors of other frogs within each memeplex and it will develop through a process of memetic evolution. The memeplexes are forced to mix together after a certain number of memetics evolution 22

32 steps and new memeplexes are formed through a shuffling process. The local search and the shuffling processes continue until convergence criteria are satisfied. The flowchart of Shuffled frog leaping algorithm is illustrated in Fig.3.5. The various steps are as follows: (1) The Shuffled frog leaping algorithm involves a population P of possible solution, defined by a group of virtual frogs(n). (2) Frogs are sorted in descending order according to their fitness and then partitioned into subsets called as memeplexes (m). (3) Frogs i is expressed as X i = (X i1, X i2,..x i3 ) where S represents number of variables. (4) Within each memeplex, the frog with worst and best fitness is identified as Xw and Xb. (5) Frog with global best fitness is identified as Xg. (6) The frog with worst fitness is improved according to the following equation. D i =rand ( ) (X b -X w ) (3.16) X neww =X oldw + D i (3.17) where rand is a random number in the range of [0,1]. D i is the frog leaping step size of the i-th frog and D max is the maximum step allowed for change in a frog s position. If the fitness value of new X w is better than the current one, X w will be accepted. If it isn t improved, then the calculated frog leaping step size D i and new fitness X neww are repeated with X b replaced by X g. If no improvement becomes possible in the case, a new X w will be generated randomly. Repeat the update operation for a specific number of iterations. After a predefined number of memetic evolutionary steps within each memeplex, the solutions of evolved memeplexes are replaced into new population. This is called the shuffling process. The shuffling process promotes a global information exchange among the frogs. Then, the population is sorted in order of decreasing performance value and updates the population best frog s position, repartition the frog 23

33 group into memeplexes and progress the evolution within each memeplex until the conversion criteria are satisfied. Start Initialize parameters: Population size (P) Number of memeplexes (m) Number of iterations within each memeplex Generate random population of P solutions (frogs) Calculate fitness of each individual frog Sorting population in descending order of their fitness Divide P solutions into m memeplexes Local Search Shuffle evolved memeplexes No Termination = true Yes Determine the best solution End Fig 3.5 Flow chart of Shuffled Frog Leaping Algorithm 24

34 CHAPTER 4 RESULTS AND DISCUSSIONS The optimization algorithms have been carried out in MATLAB R2012b. In the simulation work, Fig.4.1 depicts the fitness of the genetic algorithm and shuffled frog leaping algorithm with the execution time. The fitness function involves cost, number of hop counts and holding time. Better fitness is achieved for a smaller execution time. 4.1 FITNESS FUNCTION Fig 4.1 Fitness function of GA and SFLA 25

35 Table 4.1. Fitness function of GA and SFLA with respect to time Parameter Genetic Algorithm Shuffled Frog Leaping Algorithm Fitness Function The average fitness value in GA and SFLA for the execution time 0.4 seconds are and as given in Table MEAN BLOCKING PROBABILITY (w.r.t. CHANNEL REJECTION RATIO) Fig.4.2 and 4.3 shows the variation in the blocking probability assuming different values of adjacent wavelength rejection ratios for GA and SFLA respectively. In each case by executing the program several times and then by computing the average, mean blocking probability is estimated. In FWM aware priority based wavelength assignment, the mean blocking probability decreases for a reduction in each of the adjacent wavelength rejection ratio. Fig 4.2 Mean blocking probability for a fixed network load using GA 26

36 Table 4.2. Mean blocking probability of GA using different wavelength assignment techniques Wavelength Assignment Techniques Mean Blocking Probability (w.r.to Channel Rejection Ratio(dB) First Fit Random Round Robin Wavelength Ordering Table 4.2 shows the mean blocking probability of Genetic Algorithm with four wavelength assignment techniques first fit, random, round robin and wavelength ordering. Wavelength ordering gives less blocking probability for GA compared to other wavelength assignment techniques. 27

37 Fig 4.3 Mean blocking probability for a fixed load using SFLA Table 4.3. Mean blocking probability (w.r.to Channel Rejection Ratio) of SFLA using different wavelength assignment techniques Wavelength Assignment Techniques Mean Blocking Probability (w.r.to Channel Rejection Ratio(dB)) First Fit Random Round Robin Wavelength Ordering

38 Table 4.3 shows the mean blocking probability of Shuffled Frog Leaping Algorithm with four wavelength assignment techniques first fit, random, round robin and wavelength ordering. For SFLA wavelength ordering gives less blocking probability with respect to channel rejection ratio. 4.3 AVERAGE FITNESS SCORE Fig.4.4. Average fitness score for GA and SFLA Fig.4.4 depicts the rate of convergence of genetic algorithm and shuffled frog leaping algorithm for first fit, random, round robin, wavelength ordering and FWM aware priority based wavelength assignment techniques. By randomly selecting an individual and fixing the best fitness value, the curves can be plotted. The average fitness score decreases with increase in generations. Table 4.4 shows the average fitness score for GA and SFLA using different wavelength assignment techniques. Average fitness score for GA and SFLA are approximately same. FWM priority based assignment has higher average fitness score. 29

39 Table 4.4.Average fitness score for GA and SFLA using different wavelength assignment techniques Wavelength Assignment Techniques Average Fitness Score (w.r.to Generations) First Fit Random Round Robin Wavelength Ordering FWM priority based 1.179e e 1.18e e e Assignment +07 e+07 e MEAN BLOCKING PROBABILITY (w.r.to GENERATIONS) Fig.4.5 and 4.6 show the mean blocking probability exhibited by the Genetic Algorithm and Shuffled Frog Leaping Algorithm which is a performance metrics of dynamic routing and wavelength assignment. The mean blocking probabilities obtained by GA and SFLA for the three wavelength assignment techniques are plotted assuming exponential holding times distribution. 30

40 Fig.4.5. Mean blocking probability of GA Table 4.5. Mean blocking probability (w.r.to Generations) of GA using different wavelength assignment techniques Wavelength Assignment Mean Blocking Probability(w.r.to Generations) Technique First Fit Random Round Robin

41 Fig.4.6. Mean blocking probability of SFLA Table 4.5 and 4.6 show the mean blocking probability of Genetic Algorithm and Shuffled Frog Leaping Algorithm for three wavelength assignment techniques first fit, random and round robin. For both GA and SFLA round robin gives less blocking probability with respect to generations. Table 4.6 Mean blocking probability (w.r.to Generations) of SFLA using different wavelength assignment techniques Wavelength Assignment Mean Blocking Probability(w.r.to Generations) Technique First Fit Random Round Robin

42 4.5 MEAN EXECUTION TIME Fig.4.7.Mean execution time of GA For all wavelength assignment techniques mean execution time exhibited by Genetic Algorithm is depicted in Fig.4.7 and the values are shown in Table 4.7. FWM aware priority based wavelength assignment technique provides the least mean execution time for different network loads. 33

43 Table 4.7 Mean Execution Time of GA using different wavelength assignment techniques Wavelength Assignment Mean Execution Time (w.r.to network load(erlang)) Techniques First Fit Random Round Robin Wavelength Ordering FWM priority e e e e e- 2.94e- based Assignment 34

44 Fig.4.8. Mean execution time of SFLA Mean execution time exhibited by Shuffled Frog Leaping Algorithm is plotted in Fig.4.8 and the values are shown in Table 4.8 for five types of wavelength assignment techniques. For different network loads, FWM aware priority based wavelength assignment technique provides the least mean execution time. 35