Estudo de algoritmos RWA em redes GMPLS sobre DWDM

Abstract

The increase in the number of users and the emergence of sophisticated network applications motivate the research to improve the data-communication technologies, and thus to meet the requirement for bandwidth. In optical networks, one of the main advances is wavelength division multiplexing (WDM). An essential issue in designing of optical networks is how the network will be controlled, that is, which type of signaling will be used for resources reservation, routing and wavelength assigning, and how physical impairments will be dealfish. This implies in the necessity of several devices to guarantee the good functioning of the network. The diversity and complexity of administration of these devices stimulate the creation of GMPLS (Generalized Multiprotocol Label Switching), in order to provide control, for domains not only based on packages (as it makes the MPLS - Multiprotocol Label Switching), but also on time slots, wavelength and fiber. However, the labels switching on the wavelength multiplexing can generate a NP-complete problem involving the routing and wavelength assignment: if both do no succeed, the connection blocking will occur with and loss of network performance. Another relevant question is that, in all-optical networks, there is some physical impairment that influence the blocking probability e must be taken in consideration in algorithms RWA. This work presents the study of the control plan GMPLS and RWA algorithms for the setting up of connections in alloptical WDM networks. Moreover, the study of two physical impairments (polarization mode dispersion and optical signal noise rate) was done, inserting them in RWA algorithm to reduce the blocking probability due them. The results had shown that the MostUsed algorithm has better performance than the BestFit and the Random one when considered the connection blocking due to resources (lambdas). Moreover, the use of impairments-aware algorithms influence significantly in the reduction of the blocking probability. Therefore, the more advantageous performance was observed in the use of impairment-aware algorithms simultaneously with the MostUsed algorithm.