This deliverable, the first outome of WP2, includes the initial review of standard control, management and orchestration specifications and systems, and the initial review of the spectrum of learning and optimization tools, and of the emerging techniques in fundamental research on AI models that are relevant to concrete networking problems. This D2.1 deliverable lays the technical foundation of other work packages, in particular WP3 and WP4, as they are in charge of developing the NI-assisted functionalities for B5G problems, in the area of real-time control and network function intelligence (WP3), and in the area of service and resource management and orchestration (WP4). In addition, D2.1 presents DAEMON’s vision for a network intelligence framework with a more systematic and deeper integration of NI, thereby providing means to overcome the limits of current control planecentric closed-loop approaches. This deliverable provides insights into the eight NI use cases/functionalities and a thorough analysis of their functional and non-functional requirements. Furthermore, it discusses the NI workflows, which are presented as a closed-loop that consists of data management, control, and training, thus, providing traditional network management and orchestration frameworks with automation capabilities enforced by optimized decision-making processes. D2.1 also presents DAEMON’s initial guidelines on the i) selection, ii) configuration, and iii) orchestration of all NI instances deployed in the network. All three aspects are driven by the time constraints in NI functions execution, considering time as one of the most influential factors when making decisions on selection, configuration, and orchestration, of NI functions and algorithms. The specifications of AI for NI that are provided in Section 6 are studied from the perspectives of the network, and the service, thus the specifications are separately presented as network-driven, and service-driven ones. For the network-drive specficitations there are trade-offs that an NI native framework shall be able to tackle are discussed in terms of the infrastructure monitoring data, the timely decision and the enforcement of such a decision in the infrastructure. For the service-driven specifications, this document identified three three use cases where the specifications of AI for NI depend on the application-level target. The future work in WP2 is based on an iterative process that will incorporate the feedback provided by WP3/WP4 (algorithms) and WP5 (performance evaluation), so the DAEMON’s NI framework and tools keep evolving and are aligned with the NI-assisted functionality designed and developed in WP3 and WP4. In addition, this deliverable D2.1 contains three appendixes, Appendix A presents the functional and non-functional requirements of the NI use cases and verification scenarios. Appendix B presents an analysis on existing frameworks and architectures, and finally Appendix C presents MLOps methodology that combines Machine Learning (ML) with software development operations (DevOps).
This project has received funding from the European Union's Horizon 2020 Research and Innovation Programme under grant agreement No. 101017109.