Deliverable 2.3: Final DAEMON Network Intelligence framework and toolsets

This is the third and last public deliverable of WP2 of the DAEMON project. It builds upon the material of
the previous deliverable of WP2, i.e., D2.2 [1], and on activities and results achieved during the second
iteration of the project in WP3 D3.2 [2], WP4 D4.2 [3], and WP5 D5.2 [4]. As a result, the document
describes the following content.
First, it provides the final update on the functional and non-functional requirements of the eight NIassisted
functionalities (Reconfigurable Intelligent surfaces control - RISC, Multi-timescale Edge resource
management – MTERM, In-backhaul support for service management – IBSSI, Compute-aware radio
scheduling – CAWRS, Energy-aware VNF control and orchestration – EAWVNF, Self-learning MANO –
SLMANO, Capacity forecasting – CFORE, and Automated anomaly response – AARES) tackled by
DAEMON at the end of the WP2. Although no new updates were added to the functionalities, we assess
the risks to achieve the requirements and its current completion status. For the requirements that were
not finalized at the time of this deliverable, we also specify what is required to successfully finalize it and
in which deliverable (e.g., WP3 D3.3, WP4 D4.3, or WP5 D5.3) the results will be provided.
Second, it presents the final updates of the Network Intelligence Plane (NIP), a collection of modules and
interfaces responsible for managing NI within the network. In this deliverable, the NIP has evolved, and it
is presented as a unified framework that brings together (i) the operational hierarchy of NI components
and their orchestration and (ii) the N-MAPE-K representation of the NI components. By doing so, we make
another step forward toward the vision of a complete NIP initially presented in D2.2 [1].
Third, in addition to the unified DAEMON framework, we also identify and present in detail the specific
needs that NI algorithms pose on the NIP. Moreover, we analyze their specificity in terms of challenges
towards the procedures for NI management at the Network Intelligence Orchestrator (NIO) level. We
also devise and describe the functionalities that the NIO shall provide to support such requirements and
how they fit the whole architecture together. The architectural design is complemented by presenting
and discussing the interfaces required to allow communication between NIP components and with
external entities such as the RAN controller and the 5G Core systems. These interfaces are also enablers
for designing the set of procedures that address the needs and challenges introduced in this document.
Fourth, this document provides the final, comprehensive overview of the literature review carried out by
the project, focused on the integration of machine learning and NI in mobile network management. The
survey highlights key trends in current research and showcases the distinctive contributions made by the
DAEMON project. The insights that originated from this analysis also support our final updates to the
project guidelines, including new ones, for practical NI design. As in D2.2 [1], these guidelines focus on
two main directions: i) NI design tailored to the needs of B5G network management, orchestration, and
control, and ii) NI design that considers the use of more traditional, more straightforward, or interpretable
models to avoid overburdening the system with data-heavy models and promotes the utilization of
models that are easier to understand and interpret.
We closed this document with additional closing remarks and two appendices containing
complementary information related to the functional requirements and the literature review.