Szczegóły publikacji
Opis bibliograficzny
MARINA: multi-agent reinforcement learning-based routing with intelligent network adaptation in SDN / Hamed Nazari, Nashid Shahriar, Piotr BORYŁO // W: 2025 Canadian Conference on Electrical and Computer Engineering (CCECE) [Dokument elektroniczny] : 26–29 May 2025, Vancouver, Canada. — Wersja do Windows. — Dane tekstowe. — Piscataway : IEEE, cop. 2025. — ( Conference proceedings (Canadian Conference on Electrical and Computer Engineering) ; ISSN 2576-7046 ). — Dod. ISBN: 979-8-3315-2258-2 (print on demand). — e-ISBN: 979-8-3315-2257-5. — S. 631–637. — Wymagania systemowe: Adobe Reader. — Bibliogr. s. 637, Abstr.
Autorzy (3)
- Nazari Hamed
- Shahriar Nashid
- AGHBoryło Piotr
Słowa kluczowe
Dane bibliometryczne
| ID BaDAP | 166181 |
|---|---|
| Data dodania do BaDAP | 2026-03-11 |
| Tekst źródłowy | URL |
| DOI | 10.1109/CCECE64018.2025.11364348 |
| Rok publikacji | 2025 |
| Typ publikacji | materiały konferencyjne (aut.) |
| Otwarty dostęp |  |
| Wydawca | Institute of Electrical and Electronics Engineers (IEEE) |
| Czasopismo/seria | Conference proceedings (Canadian Conference on Electrical and Computer Engineering) |
Abstract
Optimizing routing in Software-Defined Networks (SDNs) to meet Quality of Service (QoS) demands presents notable challenges, particularly with traditional methods that struggle with the dynamic nature of SDNs. Existing techniques often fail to adapt efficiently to varying traffic patterns, resulting in suboptimal network performance. Furthermore, the existing routing methods typically prioritize either the infrastructure provider or service provider benefits, neglecting a balanced approach that ensures efficient utilization of network resources while meeting QoS metrics for clients, such as guaranteed latency and throughput. This paper proposes a multi-agent Deep Reinforcement Learning (DRL)-based QoS-aware routing solution in SDN environments with intelligent network adaptation, called MARINA. MARINA dynamically adjusts routing paths for both existing and new incoming traffic flows in the network. Moreover, MARINA simultaneously optimizes network resource usage and meets QoS constraints such as latency and guaranteed throughput in an SDN. MARINA is extensively evaluated on real-world GEANT2 network topology. The results indicate that our algorithm outperforms Open Shortest Path First (OSPF), Equal Cost Multiple Path (ECMP), and single-agent DRL in key performance metrics. Specifically, MARINA meets 51.4%, 29.4%, and 11.1% more QoS requirements than OSPF, ECMP, and single-agent DRL-based approaches, respectively.
