Experimental Translucent-Oriented Routing for Dynamic Lightpath Provisioning in GMPLS-Enabled Wavelength Switched Optical Networks

摘要

In the evolution from opaque networks, using 3R regenerators and OEO conversions for every wavelength channel, towards transparent networks, relying on end-to-end all-optical connections, translucent networks is considered as an intermediate step. A translucent network is a cost-efficient infrastructure between opaque and transparent networks, that aims at attaining an adequate trade-off between network construction cost (i.e., due to the high cost of 3R regenerators) and the service provisioning performance (i.e., end-to-end optical signal quality). This article addresses the dynamic provisioning of connections within a GMPLS-enabled translucent wavelength switched optical network with sparse 3R regenerators. To this end, translucent-oriented GMPLS protocol extensions are proposed to, on the one hand, disseminate required per node regenerator availability and OSNR-related link and node attributes and, on the other hand, allocate regenerator resources when setting up connections. An on-line OSNR-based impairment-aware RWA algorithm is devised to compute routes aiming at satisfying two constraints: the wavelength continuity constraint and the end-to-end optical signal quality. If either constraint cannot be satisfied, available regenerators along the route are used. To the best of our knowledge, the novelty and progress of this work is that, for the first time, both the proposed translucent GMPLS protocol enhancements and the performance of the RWA algorithm are combined within an experimental control plane network scenario (CTTC's ADRENALINE testbed). This allows validating and evaluating the feasibility of the described translucent GMPLS routing strategy when being potentially deployed into real WSON networks. Several configurations with numbers of 3R regenerators and wavelength channels per link, are experimentally evaluated and compared mainly in terms of the connection blocking probability. Although concrete details about the measured average setup delay are al-nso discussed.