Quantifying Resource Savings from Low-Margin Design in Optical Networks with Probabilistic Constellation Shaping

Karandin, Oleg; Musumeci, Francesco; Ayoub, Omran; Ferrari, Alessio; Pointurier, Yvan; Tornatore, Massimo

doi:10.1109/ecoc52684.2021.9605845

Cited by 17 publications

(19 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A flow diagram describing the behaviour of the Genetic Algorithm (GA-FF) is illustrated in Fig. 3 [40], [41]. The parameters of GA-FF have been finetuned in order to get the best trade-off between computing times and solution quality.…”

Section: Illustrative Numerical Resultsmentioning

confidence: 99%

On Deep Reinforcement Learning for Static Routing and Wavelength Assignment

Cicco

Mercan²,

Karandin³

et al. 2022

IEEE J. Select. Topics Quantum Electron.

Self Cite

View full text Add to dashboard Cite

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.

show abstract

Section: Illustrative Numerical Resultsmentioning

confidence: 99%

On Deep Reinforcement Learning for Static Routing and Wavelength Assignment

Cicco

Mercan²,

Karandin³

et al. 2022

IEEE J. Select. Topics Quantum Electron.

Self Cite

View full text Add to dashboard Cite

show abstract

“…The answer is challenging as savings depend on the algorithm employed, on the network characteristics and on the considered metric. Some quantitative estimations of the resource savings obtained using different metaheuristics are provided in Table II, showing saving ranging from 5 [20] Total used spectrum 15-20 Genetic Algorithm [21] Maximum number of occupied slots in fibre 5-15 Genetic Algorithm [18] 1. Total used spectrum 2.…”

Section: A Current Algorithmic Solutions For Flexible Resource Alloca...mentioning

confidence: 99%

“…In the previous examples, resource allocation is optimized for a static traffic scenario, neglecting traffic growth and/or traffic fluctuations. Regarding traffic growth, it has been shown that highly optimized resource allocation at deployment ("at year 0") leads to lower excess capacity, and hence reduced savings in the following years, calling for new optimization techniques considering traffic anticipation during resource allocation [18]. Regarding traffic fluctuations, new flexible optical-network technologies offer several new opportunities to dynamically adjust and reconfigure capacity of existing lightpaths, which, in turn, can increase network capacity (readers are referred to [19] for a comprehensive description of all reconfiguration actions that can be take to address traffic changes with flexible technologies).…”

Section: A Current Algorithmic Solutions For Flexible Resource Alloca...mentioning

confidence: 99%

Flexible Technologies to Increase Optical Network Capacity

et al. 2022

Self Cite

View full text Add to dashboard Cite

Increased global traffic puts tough requirements not just on fibre communications links, but on the entire network. This manifests itself in multiple ways including how to optimise wavelength routing around the network; how to maximise the benefits arising from fine-control DSP with increasingly accurate real-time monitoring; and how to best deploy Multiband or multiple fibre connectivity. This paper will summarise research into all these areas to present a full picture of how future optical networks will play their role in supporting the continuing traffic demands of broadband, 5G and associated applications.

show abstract

“…To perform RSA, we employ the multi-objective Genetic Algorithm, based on the concept of Pareto dominance, as in [10], that selects lightpath routing among 7 shortest paths, with the objective of minimizing the number of occupied frequency slots and the number of transponders in the network. We assume that an aggregate request between two optical nodes can be split into separate lightpaths, possibly going across different routes.…”

Section: Routing and Spectrum Assignment (Rsa) Algorithm In C And C+l...mentioning

confidence: 99%