Converter placement in all-optical networks using genetic algorithms
Introduction
All-optical networks consist of access stations interconnected by a network of optical fibers and routing nodes/switches. In the access stations (or hosts), the data in electronic form is converted to optical form and transmitted out in optical fibers, while the incoming data in optical form is converted from optical to electronic form. In order to transfer data from one access station to another a connection is established and data is transmitted over it. The access stations consisting of electronic computers can switch only at a few Gbps at the most, whereas the bandwidth offered by the optical fiber is about 40 Tbps. Hence, to make better use of this enormous bandwidth and to cater to the exponential growth of bandwidth demands on wide area networks, several connections/channels are set up simultaneously on the fibers. Here, each channel is on a different wavelength to avoid mixing/interference of signals and operates at peak bandwidth of (1–10 Gbps), compatible with electronic switching speeds. In such networks, known as Wavelength Division Multiplexed (WDM) networks, routing can be done based on the wavelengths at the intermediate nodes. However this places a constraint, that throughout the connection from source to destination, in all fibers a single wavelength must be used. This constraint known as wavelength continuity constraint, imposes a limitation on the wavelength reusability and hence on the overall performance of the network. This is because the particular wavelength cannot be used by any other connection, along the entire route. This can be overcome by converting the wavelengths at the intermediate nodes and maintaining different wavelengths for different connections/channels. This technique known as wavelength conversion is done by wavelength converters placed at the routing nodes/switches. The wavelength converters apart from being costly, introduce distortion which increases with the extent of conversion. Hence they have to be used in an optimal fashion by placing them at a few strategic nodes to improve performance, keeping the cost and distortion at the minimal level.
This paper explores the impact of placement of converters on the blocking probability performance of the network for different number of converters whose locations are obtained by genetic algorithm and for different conversion degrees. Heuristics have been used to obtain starting solutions for the genetic algorithm. The routing technique reduces the number of cascaded wavelength conversions and limits the degree of the wavelength conversion enhancing the optical signal quality.
The paper is organized into the following sections. Section 2 describes the wavelength converter technology and its related components. Section 3 presents a brief review of Genetic Algorithms (GAs) and the parameters of GA that have been used in our work. The network model, mathematical formulations and related work are described in Section 4. Section 5 gives the simulation data and results for the 12-node ring network and 14-node NSFNET. Conclusions are presented in Section 6.
Section snippets
Wavelength converters
A wavelength converter converts an input wavelength w1 to an output wavelength w2. The converters are placed in the switching nodes of the WDM wavelength routed optical network. Converters can be provided for every channel, a bank of converters can be shared by all channels on a link or a bank of converters can be shared by each node. The last method known as share-per-node is used in our simulation studies. Converters can also be classified as opto-electronic optical converters and all-optical
Genetic algorithms
A genetic algorithm (GA) is a parallel, stochastic search algorithm which can be applied to a variety of problems. Its speed and simplicity of implementation have immensely contributed to its success.
GA is an iterative optimization procedure. Instead of working with a single solution in each iteration, GA works with a number of solutions in each iteration. These are collectively known as population. In the absence of any knowledge of the problem domain, a GA begins its search from a random
Formulations for the routing and assignment of wavelength problem and converter placement problem
In this section mathematical formulations for the static and dynamic Routing and Assignment of Wavelength (RAW) problem are presented for lowering the number of conversions.
Simulation results
Simulation was performed on NSFNET shown in Fig. 4 and the 12-node ring network shown in Fig. 3 using GA over the network simulation algorithm for finding optimal/near optimal placements of converters. For each load the GA generates different converter placements which are used by the network simulator to compute the blocking probability. The fitness of each placement string is inversely proportional to the blocking probability for that placement. Based on these fitness values the GA does
Conclusions
In this work new ILP formulations were proposed for the static and dynamic RAW problem with a view to reducing the number of conversions. The performance of the WDM network was studied with the layered-graph-based solution to the dynamic RAW problem with near optimal placement of full and partial converters obtained using genetic algorithm for 12-node ring network and 14-node NSFNET.
It was found from these simulations that
- •
Placement of converters at a few nodes as suggested by the GA gives
References (15)
- S. Ramamurthy, B. Mukherjee, Wavelength conversion in WDM networking. IEEE Select Areas in Communication, Department of...
- C. Chen, S. Banerjee, A new model for optimal routing and wavelength assignment in wavelength division multiplexed...
- et al.
Routing and wavelength assignment in all-optical networks
IEEE/ACM Transactions in Networking
(1995) - et al.
Models of blocking probability in all-optical networks with and without wavelength changers
IEEE Journal on Selected Areas in Communication
(1996) - et al.
All optical networks with sparse, wavelength conversion
IEEE/ACM Transactions in Networking
(1996) - K. Bala, E. Bouillet, G. Ellinas, Benefits of minimal wavelength interchange in WDM rings, Proceedings of the OFC ’97,...
- K.R. Venugopal, M. Shiva Kumar, P. Sreenivasa Kumar, Placement of limited range wavelength converters in wavelength...
Cited by (26)
Multi-objective routing and wavelength converter allocation under uncertain traffic
2015, Optical Switching and NetworkingCitation Excerpt :The approaches based on simulations can be classified in turn as pos-simulated and pre-simulated. In pos-simulated approaches, the blocking probability of each candidate solution is obtained by simulation; so, there are as many simulations as solutions to be evaluated [61–64]. On the other hand, in the pre-simulated approach, statistical information on the used wavelength converters is calculated considering full-wavelength conversion networks where each node has a complete conversion capacity.
Differential evolution optimization applied to the wavelength converters placement problem in all optical networks
2012, Computer NetworksCitation Excerpt :Their study includes both, static and dynamic light path establishments. Vijayanand et al. [9] proposed new integer linear program (ILP) formulations for the static and dynamic RWA problems to reduce the number of conversions. In their study, a genetic algorithm (GA) was used for placing limited-range wavelength converters in an arbitrary mesh WRON.
Strictly nonblocking grooming of dynamic traffic in unidirectional SONET/WDM rings using genetic algorithms
2003, Computer NetworksCitation Excerpt :In [14], Ali et al. employed it to optimize the power assignment in optical networks. In [15], Vijayanand et al. attacked the converter placement problem in all-optical networks with GAs. In [8] Xu et al. made use of this approach to solve the static traffic grooming problems etc.
Traffic grooming in unidirectional WDM ring networks using genetic algorithms
2002, Computer CommunicationsCitation Excerpt :They are also applicable to the optimal design of all-optical networks. In Ref. [10], Saha et al. used GAs to optimize the virtual topology design problem; in Ref. [11], Gazen et al. used this approach to optimally design the logical multihop lightwave networks; in Ref. [12], Ali et al. used it to optimize the power assignment in optical networks; in Ref. [13], Vijayanand et al. attacked the converter placement problem in all-optical networks using GAs etc. Applications of GAs to many other optimal design problems in optical networks can also be found in the references cited in the earlier papers.
On the behavior of parallel genetic algorithms for optimal placement of antennae in telecommunications
2005, International Journal of Foundations of Computer Science