As application requirements continue to push the need for capacity higher and higher, optical networks and systems are playing an increasingly important role in the communications infrastructure worldwide. This special issue features eight papers that are extended versions of papers invited from those presented at the Optical Networks and Systems Symposium at GLOBECOM 2022 held in Rio de Janeiro, Brazil. The papers span a wide range of topics stretching from resource allocation to the use of novel machine learning techniques to solve a variety of problems.

The special issue starts off with a paper by George Rouskas et al. entitled “Symmetry-free algorithm for spectrum allocation: parallel implementations and evaluation.” This paper proposes a spectrum allocation algorithm for elastic optical networks that is inherently faster than conventional algorithms because of its ability to recognize spectrum combinations that are permutations of others. The paper further proposes a parallel implementation of the algorithm and conducts a thorough performance evaluation.

In the second contribution entitled, “Minimizing the cost of hierarchical optical transport network traffic grooming boards in metro networks,” Aryanaz Attarpour et al. address the problem of minimizing deployment cost for metro networks in which nodes utilize a hierarchical node architecture that consists of multiple stacked OTN traffic grooming boards supporting both coherent and non-coherent transmission technologies. They propose novel techniques based on genetic algorithms, simulated annealing, and knowledge-domain heuristics that result in a significant reduction in equipment cost.

In the third paper titled “Deep reinforcement learning for proactive spectrum defragmentation in elastic optical networks,” Ehsan Etezadi et al. study the problem of spectrum fragmentation in elastic optical networks that support dynamic traffic. This leads to the formation of several unoccupied spectral gaps that are not large enough to accommodate the spectrum requirements of service requests. In order to minimize spectrum fragmentation, a deep reinforcement learning (DRL) based framework is proposed. This framework enables the network operator to decide which existing connections can be reconfigured and when. A comparative study with other heuristic algorithms is presented; this shows that the DRL framework performs much better than the other existing schemes.

In the fourth contribution, entitled “Networking assessment of ROADM architecture based on a photonics integrated WSS for 800G multi-band optical transport,” Muhammad Umar Masood et al. propose a novel ROADM architecture that adopts integrated multi-band (S + C + L) WSSs. These WSSs consist of micro-ring resonator based filters implemented by photonic integrated circuit technology. Thus these WSSs are more compact than current WSSs based on MEMS and LCoS technologies and have modular design capability. A network performance analysis with a GSNR estimation, assuming 800/400 Gbps channels, is conducted for German and Italian topologies. Substantial network capacity enhancement is verified through the analysis.

In the fifth contribution entitled, “Optical identification using physical unclonable functions,” Pantea Nadimi Goki et al. introduce the concept of optical identification (OI), in which each network device or sub-system is assigned a unique optical fingerprint and digital signature through a physical unclonable function based on Rayleigh backscattering. They describe applications of OI, such as the identification of paths in an optical network and the authentication of users in a quantum key distribution system.

In the sixth contribution, entitled “Impact of the band upgrade sequence on the capacity and capital expenditure of multi-band optical networks,” Ningning Guo et al. provide an upgrade method of conventional networks to next-generation multi-band optical networks. The method includes the optimal selection of bands to be used for next upgrade timings, a routing and spectrum assignment algorithm, an OSNR-aware traffic grooming algorithm, and a power allocation algorithm which aims at network capacity maximization. The authors elucidate that the “C + L + E + O + S + U” scenario can achieve 3 times the capacity of the “C + L” even though the transmission characteristics on these bands are not uniform.

In the seventh paper of this issue, entitled “Machine learning framework for timely soft-failure detection and localization in elastic optical networks,” Sadananda Behera et al. leverage the capabilities of an encoder-decoder learning framework to predict the evolution of soft failures over a long time-horizon. They utilize the quality of transmission (QoT) as an indicator of failure and show that failure detection schemes using rule-based fixed QoT margins could lead to premature failure prediction and unnecessary early repair or failure to detect altogether. It is shown that their proposed approach performs better by predicting failures in a timely manner.

In the last paper in the feature issue, entitled “Hybrid amplifier placement in mesh DWDM networks using integer linear programming models,” the authors João Pedro and Nelson Costa tackle an optimization problem on the placement of hybrid erbium-doped-fiber/Raman amplifiers. Being more expensive than conventional EDFAs, the selection of locations of hybrid amplifiers is an important problem in mesh networks. The problem is made challenging by the diversity of optical channel conditions that arise in mesh networks. The paper presents integer linear programming methods to solve the problem.

All extended manuscripts underwent a separate rigorous peer review process for the special issue, and we thank all the reviewers whose meticulous reviews were very helpful in enhancing the articles. We hope you enjoy reading about these latest advances in optical networking and systems that were presented at GLOBECOM 2022.