A Genetic Algorithm Based on Cell Loss for Dynamic Routing in ATM Networks

Abstract

Abstract. New B-ISDN will have to ensure a high standard of quality of service measured as delay in the communications, guarantee of safety communications and of course no loss of information. Although the routing optimization problem has been previously dealt with in the bibliography, not many works have been presented according to the routing in ATM networks using cell loss as the overall criterion. Here a new model based on cell loss is presented for an ATM network with matrix switches and queues at the end. A quickly evaluated genetic algorithm is proposed to solve the model.

A Tabu Search algorithm for dynamic routing in ATM cell-switching networks

“…The Tabu Search algorithm was compared to the Genetic Algorithm first described in Cortés et al (2002), whose pseudocode is shown in the appendix. The solutions provided by the MSwP algorithm are also presented, showing the improvement reached by the TS and GA heuristics respectively.…”

On the enrouting protocol problem under uncertainty

“…A flow formulation is considered in the paper, instead of a required path formulation. The approach presented in the paper only considers the call scale, instead of considering also the cell scale and the burst scale as it is mandatory for real time routing algorithms and, then, the approach is less ambitious than the approach presented in Medova (1998), even Cortés et al (2002). We refer to these two papers for a deepen more about ATM network performance and also Tsybakow's papers, see e.g.…”

“…1 shows a representation of the information flow at a terminal node and its relationship with the variables of the model. Cortés et al (2002) and Medova (1998) path oriented formulations. A potential drawback for this type of formulations can occur if the number of paths is big enough.…”

“…These routes force only one outgoing arc from each (terminal or switch) node for the information flow directed to each destination terminal node; in another context, see Paramés et al (2000). See also Medova (1998) and Cortés et al (2002). So, an enrouting protocol can be viewed as a set of coordinated reverse arborescences, one for each destination terminal node, probably with common nodes and arcs, where each of the destination terminal nodes is an inverse root node and the related origin terminal nodes are the inverse leaves of the tree, see Gamvros et al (2001) and Gouveia (1999).…”

“…The problem belongs to the category of path and capacity allocation problems for a network presented in Medova (1998), see also Cortés et al (2002). The goal is to determine the enrouting protocol (i.e., the information path to be followed by the traffic flow among each origin and destination nodes) in all time instants of the time horizon.…”