EUVIMEDEUVIMED entdecken
Crossref

Machine-learning-based deferred protection for deadline-driven traffic in elastic optical networks

Nilvan Santana Souza, Augusto O. Perin, Suresh Subramaniam, Gustavo B. Figueiredo

Journal of Optical Communications and Networking · 01.07.2026 · 18 (7) · 663

DOI: 10.1364/jocn.588508 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Optical Communications and Networking, online erschienen am 10.06.2026. Verfasst von Nilvan Santana Souza und 3 weiteren Autor:innen. Die Arbeit nennt 22 Literaturangaben. Gefördert von: National Council for Scientific and Technological Development, Fundação de Amparo à Pesquisa do Estado da Bahia.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

In inter-data center elastic optical networks (IDC-EONs), even brief transmission interruptions can result in substantial data loss, which is particularly critical given the rapid growth of machine learning and data-intensive workloads in transport networks. These applications drive highly dynamic and deadline-sensitive traffic patterns, reinforcing the need for protection strategies that ensure both reliability and efficient spectrum use. Traditional protection assigns a working and a backup path simultaneously, guaranteeing recovery but consuming spectrum rigidly. However, most inter-data center requests are deadline-driven requests (DDRs), which only require transmission completion before a deadline. This flexibility enables deferred protection, where backup resources are scheduled in time slots subsequent to the start of the primary path, improving spectrum availability since they can be released if failures do not occur. The existing resource allocation strategy for Deferred Dedicated Path Protection (DDPP) is heuristic and does not adapt to the network state. To address this, we propose ANN-RAND, a machine-learning-based allocation policy for DDPP that leverages spectrum availability while applying randomized tie-breaking. Numerical results on NSFNET and EURO28 topologies demonstrate that ANN-RAND significantly reduces blocking probability compared to baseline methods in both the Dedicated Path Protection (DPP) and DDPP paradigms, highlighting its suitability for ML-driven traffic in modern transport networks.

Förderung

  • National Council for Scientific and Technological Development
  • Fundação de Amparo à Pesquisa do Estado da Bahia

Zitieren

APA

Souza, N. S., Perin, A. O., Subramaniam, S., & Figueiredo, G. B. (2026). Machine-learning-based deferred protection for deadline-driven traffic in elastic optical networks. Journal of Optical Communications and Networking, 18(7), 663. https://doi.org/10.1364/jocn.588508

Vancouver

Souza NS, Perin AO, Subramaniam S, Figueiredo GB. Machine-learning-based deferred protection for deadline-driven traffic in elastic optical networks. Journal of Optical Communications and Networking. 2026;18(7):663. doi:10.1364/jocn.588508

BibTeX
@article{Souza2026,
  author = {Souza, Nilvan Santana and Perin, Augusto O. and Subramaniam, Suresh and Figueiredo, Gustavo B.},
  title = {Machine-learning-based deferred protection for deadline-driven traffic in elastic optical networks},
  journal = {Journal of Optical Communications and Networking},
  year = {2026},
  volume = {18},
  number = {7},
  pages = {663},
  doi = {10.1364/jocn.588508}
}
Kostenlos starten →