The dawn of 5G is pushing operators to deploy high-capacity, agile networks capable of adapting to time-varying traffic patterns, especially into metro sections. ROADMs are key enablers for agility in the optical layer, however the benefits of this agility do not always compensate for increased costs. As such, filterless optical networks are emerging as a cost-effective and reliable solution compared to active photonics, thanks to a winning combination of coherent transponders and passive splitters/couplers. However, spectrum allocation management policies are of paramount importance to maximize the overall network throughput. In this paper, we focus on a filterless metro network where the hourly variation of the demands traffic is known, coming from historic data estimations. Then, we observe how the knowledge of the traffic profiles can be exploited. To assess this, we evaluate the performance, in terms of throughput, of three different spectrum management approaches: (i) fixed, where lightpaths remain static along time once allocated; (ii) semi-elastic, where lightpath-bandwidth vary according to current traffic requirements, but central frequency remains fixed; and (iii) hitless full-elastic, where any lightpath parameter may be reconfigured without disrupting the traffic. Besides, we consider two transponder types equipped with (i) shared or (ii) independent tunable lasers for transmission and reception, which affects to spectrum allocation of bidirectional connections. According to our results, the semi-elastic approach clearly outperforms the fixed approach (23-33% more throughput) with a reduced gap to the hitless full-elastic case (10-24% less throughput), especially considering that the latter is not commercially available yet. Interestingly, using dual-laser transponders only yields a 10% gain with respect to single-laser transponders for the semi-elastic scenario, and thus may not justify the extra hardware.

Peer Reviewed