NET1 – Network Softwarisation
Wednesday, 3 June 2026, 8:30-10:00, room Dirección de Certámenes (1st floor)
Session Chair: Anna Brunstrom (Karlstad Univ., SE & Univ. Malaga, ES)
Edge Computing Coordination for Latency-Aware and Fair Coalition Formation
Polyzois Soumplis (National Technical University of Athens, Greece); Aristotelis Kretsis (NTUA, Greece); Panagiotis Kokkinos (National Technical University of Athens & University of Peloponnese, Greece); Emmanouel Varvarigos (National Technical University of Athens, Greece)
Edge computing resources are often fragmented and underutilized, as individual operators struggle with bursty demand spikes that cause service level objective (SLO) violations, while nearby providers have idle capacity. Forming coalitions to share resources can yield significant multiplexing gains, but stable cooperation among self-interested operators remains a major challenge. We address this by modeling inter-operator resource sharing as a bilevel cooperative game. At the inner level, a convex program defines the coalition's economic value by optimally allocating heterogeneous workloads across shared infrastructure, accounting for non-linear congestion costs and end-to-end latency SLOs. The outer level employs a welfare-improving best-response mechanism that terminates at a partition with no unilateral deviation that is both individually profitable and strictly welfare-improving, where each operator's move is executed only if it is both individually profitable and strictly increases total social welfare. A nominal entry fee controls coalition churn, and profits are divided using a capacity-proportional surplus rule that is budget-balanced, individually rational, and closely approximates exact Shapley values (mean deviation <3.1%). We validate the mechanism on a synthetic metro topology with nine operators under diurnal and flash-load traffic scenarios. The proposed mechanism discovers geographically-aware coalitions that improve social welfare by 5-29% over non-cooperative operation and consistently outperform centralized planning, which suffers from cross-zone latency penalties. Sensitivity analysis over SLO penalty severity, entry fees, and load intensity confirms the robustness of these gains across a wide parameter range.
Performance Analysis of Approximation Strategies for Cell Switch-off Problem in Ultra-Dense Network Scenarios
Francisco Luna (Universidad de Málaga, Spain); Diego Rossit (Universidad Nacional del Sur, Argentina); Jesús Galeano-Brajones (Universidad de Extremadura, Spain)
Network sustainability is a major issue in ultra-dense deployments required for next generation wireless networks. This paper deals with the formulation of the selective deactivation of cells as an optimization problem to reduce power consumption while maintaining acceptable levels of quality of service. The results show that simple decision strategies suffer severe performance degradation as density increases, with average distances to the best-known solution exceeding 80% in low-density scenarios and frequently surpassing 90% in high-density regimes. Approximation-based InvL models achieve significantly better solution quality with minimal computational overhead, while non-linear formulations exhibit limited scalability in dense deployments. These findings highlight the structural impact of densification on cell switch-off decision-making and motivate the use of scalable, density-aware optimization frameworks.
AI-Paging: Lease-Based Execution Anchoring for Network-Exposed AI-as-a-Service
Merve Saimler (Ericsson Research, Turkey); Mohaned Chraiti (Sabanci University, Turkey)
With AI-as-a-Service (AIaaS) now deployed across multiple providers and model tiers, selecting the appropriate model instance at run time is increasingly outside the end user's knowledge and operational control. Accordingly, the6G service providers are envisioned to play a crucial role in exposing AIaaS in a setting where users submit only an intent while the network helps in the intent-to-model matching (resolution) and execution placement under policy, trust, and Quality of Service (QoS) constraints. The network role becomes to discover candidate execution endpoints and selects a suitable model/anchor under policy and QoS constraints in a process referred here to as \emph (by analogy to cellular call paging). In the proposed architecture, AI-paging is a control-plane transaction that resolves an intent into an AI service identity (AISI), a scoped session token (AIST), and an expiring admission lease (COMMIT) that authorizes user-plane steering to a selected AI execution anchor (AEXF) under a QoS binding.AI-Paging enforces two invariants: \textbf (without \COMMIT, no steering state is installed) and \textbf to support continuity and reliability of AIaaS services under dynamic network conditions. We prototype AI-Paging using existing control- and user-plane mechanisms (service-based control, \QoS\ flows, and policy-based steering) with no new packet headers, ensuring compatibility with existing 3GPP-based exposure and management architectures, and evaluate transaction latency, relocation interruption, enforcement correctness under lease expiry, and audit-evidence overhead under mobility and failures.
Towards Agentic Test-Driven Quality Assurance for 6G Networks
Christos Tranoris and Besiana Ioanna Agko (University of Patras, Greece); Kostis Trantzas (Netonomiq PC, GREECE); Irene Denazi (NetonomIQ PC, Greece)
This work proposes an agentic, intent-driven end-to-end (E2E) orchestration framework that combines intent co-creation with a test-driven quality assurance paradigm: agents iteratively refine a user's initial intent into a confirmed, auditable specification and automatically derive validation tests from the confirmed intent before provisioning, so that deployment SLA compliance moves progressively analogous to Test-Driven Development in software engineering. The approach is grounded in a standards-aligned knowledge representation based on TM Forum information models and catalogs, enabling deterministic graph traversal from Product Offerings to Service/Resource specifications and associated test specifications. We prototype the architecture by extending OpenSlice with a message-driven, multi-agent pattern, integrating MCP-enabled tool access for Knowledge retrieval. We evaluate a first set of agents using multiple open-source LLM backends connected to the TMF-based knowledge base, showing substantial variance in tool-use reliability and hallucination behavior.
uDCN: A High-Performance, ML-Orchestrated Data-Centric Networking Architecture for Edge Environments
Abdulrahman M Ben Mansor (University of Genoa, Italy); Alessandro Carrega (UNIGE, Italy & CNIT, Italy); Ramin Rabbani (CNIT, Italy); Franco R. Davoli (University of Genoa & National Inter-University Consortium for Telecommunications (CNIT), Italy)
Modern edge computing environments demand networking architectures that are capable of handling diverse application requirements with efficiency and reliability. This paper introduces uDCN (micro Data-Centric Networking), a high-performance networking architecture designed specifically for edge environments. uDCN combines the content-addressing paradigm of Named Data Networking (NDN) with the performance benefits of eBPF/XDP for kernel-level packet handling, Rust with QUIC for efficient and safe data transport, and Python with TensorFlow Lite for ML-based network orchestration. We present the design, implementation, and evaluation of uDCN, demonstrating its ability to dynamically adapt to changing network conditions through ML-driven MTU optimization. Our experiments show that uDCN achieves up to 30% higher throughput and 25% lower latency compared to traditional NDN implementations. This performance improvement, combined with uDCN's modular architecture and security features, makes it a promising solution for data-centric networking at the edge.
