DeepDefrag: A deep reinforcement learning framework for spectrum defragmentation

Abstract

Exponential growth of bandwidth demand, spurred by emerging network services with diverse characteristics and stringent performance requirements, drives the need for dynamic operation of optical networks, efficient use of spectral resources, and automation. One of the main challenges of dynamic, resource-efficient Elastic Optical Networks (EONs) is spectrum fragmentation. Fragmented, stranded spectrum slots lead to poor resource utilization and increase the blocking probability of incoming service requests. Conventional approaches for Spectrum Defragmentation (SD) apply various criteria to decide when, and which portion of the spectrum to defragment. However, these polices often address only a subset of tasks related to defragmentation, are not adaptable, and have limited automation potential. To address these issues, we propose DeepDefrag, a novel framework based on reinforcement learning that addresses the main aspects of the SD process: determining when to perform defragmentation, which connections to reconfigure, and which part of the spectrum to reallocate them to. DeepDefrag outperforms the well-known Older-First First-Fit (OF-FF) defragmentation heuristic, achieving lower blocking probability under smaller defragmentation overhead.

Demonstration of DRL-based intelligent spectrum management over a T-API-enabled optical network digital twin

“…The aim is to obtain a representation of the network state resulting from long-term operation (i.e., steady state representation). The DeepDefrag [4] DRL agent making the defragmentation decisions during the demonstration is trained separately beforehand for practical purposes.…”

“…Intelligent and adaptable techniques, such as those based on machine learning (ML), are needed to meet the network operators' quest for efficient and automated network management. DeepDefrag [4] , a recently proposed SD framework based on deep reinforcement learning (DRL), has been shown to outperform existing deterministic algorithms in SBR minimisation. DeepDefrag performs proactive SD by deciding on the reconfiguration timing, the concerned subset of connections, and their target spectrum.…”

DRL-based RMSCA for SDM Networks with Core Switching in Multi-Core Fibres

“…However, heuristic algorithms may lead to far-fromoptimal solutions due to their inability to assess the impact of current RMSCA decisions on the provisioning results of future connection requests. Recently, deep reinforcement learning (DRL) showed promising results in solving the dynamic connection provisioning/reconfiguration problem in single-core elastic optical networks (EONs) [4]- [7]. In these works, DRL employs deep neural networks (DNNs) to extract network state information and optimise a long-term cumulative reward.…”