E4.6.2 final design of zero-touch and proactive network management
Year: 2024
Type: article
Abstract: Deliverable E4.6.2, titled Final Design of Zero-Touch and Proactive Network Management, marks the completion of task A4.2 in the 6GENABLERS project (IA for 6G: Ultra Automation and Optimization, reference number TSI-063000-2021-10). Building on E4.3.2, this deliverable presents an enhanced approach for proactive, zero-touch management in future 6G networks. It leverages Neuro-Symbolic (NeSy) AI tools, particularly Logic Tensor Networks (LTN), to develop an efficient CPU resource prediction service that balances the trade-off between over-provisioning and under-provisioning through the use of smooth logical predicates in the model’s loss function. These predictions will inform the intelligent orchestrator for anticipatory scaling actions during the integration phase. The designed algorithm is highly adaptable to both online and offline telemetry data metrics, aligning with ETSI ZSM’s closed control loop framework.The report begins with an exploration of the rationale behind CPU prediction and its critical connection to network performance, followed by an introduction to NeSy’s concept and its transparency. It then outlines the model architecture and describes the O-RAN experimental dataset used. The deliverable delves into the NeSy formulation, presenting key results that show NeSy’s superior performance over the state-of-the-art DeepCog model by minimizing both under- and over-provisioning. Additionally, a complexity analysis is conducted, comparing NeSy to baseline models. The report concludes with a discussion on adherence to the ETSI ZSM closed-loop framework, as well as the integration of OPTARE’s telemetry system and OpenNebula orchestrator.
Author: i2CAT
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Award: G2828604718
OpenAlex ID: W6893010302
FWCI:
Cites:
Cited by:
Related to: