Deep Reinforcement Learning (DRL) is rising as a promising tool for solving optimization problems in optical networks. Though studies employing DRL for solving static optimization problems in optical networks are appearing, assessing strengths and weaknesses of DRL with respect to state-of-theart solution methods is still an open research question. In this work, we focus on Routing and Wavelength Assignment (RWA), a well-studied problem for which fast and scalable algorithms leading to better optimality gaps are always sought for. We develop two different DRL-based methods to assess the impact of different design choices on DRL performance. In addition, we propose a Multi-Start approach that can improve the average DRL performance, and we engineer a shaped reward that allows efficient learning in networks with high link capacities. With Multi-Start, DRL gets competitive results with respect to a state-of-the-art Genetic Algorithm with significant savings in computational times. Moreover, we assess the generalization capabilities of DRL to traffic matrices unseen during training, in terms of total connection requests and traffic distribution, showing that DRL can generalize on small to moderate deviations with respect to the training traffic matrices. Finally, we assess DRL scalability with respect to topology size and link capacity.

The success of novel multimedia services such as Video-on-Demand (VoD) is leading to a tremendous growth of Internet traffic. Content caching can help mitigate such uncontrolled growth by storing video content closer to the users in core, metro and access network nodes. So far, fixed, and, especially, mobile access networks have evolved independently, leveraging logically (and often also physically) separate infrastructures. This means that mobile users cannot access caches placed in the fixed access network (and vice-versa) even if they are geographically close to them, and energy consumption implications of such undesired effect must be investigated. In this paper we perform an evaluation of energy-efficient VoD content caching and distribution under static and dynamic traffic in converged networks as well as in non-converged networks. We define an Integer Linear Programming optimization problem modeling an energy-efficient placement of caches in core, metro and fixed/mobile access nodes where energy is minimized by powering-on and-off caches located in different segments of the network and by performing an energy-efficient VoD-request routing. To deal with problem complexity, we propose an energy-efficient content caching and VoD-request routing heuristic algorithm, which is also adopted under dynamic traffic scenarios. Our results show how deploying caches in the access and metro network segments can reduce the overall energy consumption of the network. Moreover, results show how the evolution towards a Fixed-Mobile Converged metro/access network, where fixed and mobile users can share caches, can reduce the energy consumed for VoD content delivery.

No abstract

Filterless Optical Networks (FONs) represent a novel cost-effective solution for metro optical networks, that allows to achieve equipment-cost savings by removing expensive optical-switching components from network nodes. In this study, we investigate how to further reduce equipment cost in FONs by minimizing amplifiers' cost. We propose a Genetic Algorithm (GA) for placing boosters, inline amplifiers and pre-amplifiers in FONs with the objective of minimizing amplifiers cost. We provide two versions of the GA and compare their performance against a baseline amplifier placement in terms of amplifiers cost and quality-of-transmission (QoT), i.e., lightpaths OSNR and received power. Moreover, we provide a comparison between filterless and wavelength-switched architectures. Simulative results achieved over realistic network topologies show significant amplifier cost savings, up to 60% compared to baseline approaches.

Filterless optical networks based on broadcast-and-select nodes have been proven to be a cost-effective alternative to active photonic network solutions in core networks. However, due to the emergence of novel metro-based highbandwidth cloud-based services (e.g., Virtual Reality, 4K Video-on-Demand, etc.), filterless solutions have started to attract research attention also in the metro area. In this paper, we evaluate the performance of fully-filterless and semi-filterless (i.e., hybrid solutions between fully-filterless and active photonic architectures) optical-network architectures in terms of cost of network elements and spectrum utilization, in a metro-network scenario. Our evaluations show that, due to the ring-based hierarchical nature of metro networks, fully-filterless architectures tend to require excessive spectrum utilization as the broadcast effect spreads among all hierarchical rings. On the contrary, semi-filterless network architectures seem more promising due to the presence of filters that fend off the propagation of unfiltered channels. The results also show that it is more advantageous to deploy filters at nodes of the lower network levels than at nodes of higher network levels.