Next-generation wireless systems are increasingly heterogeneous, comprising multiple Radio Access Technologies (multi-RATs) that coexist within a small geographic area. In dense, traffic-congested environments, a single macro base station cannot sustain adequate throughput or reliability. Network densification through small cell deployment, combined with Software-Defined Networking (SDN), offers a flexible control plane but leaves vertical handover and traffic offloading as critical open problems. Traditional single-criterion handover policies fail to capture the diversity of service requirements and network conditions, often degrading performance. To address these limitations, we propose TACM, a Traffic-Aware, Configurable Multiple-Criteria Decision-Making (MCDM) framework to jointly manage vertical handover and offloading. TACM integrates the Analytic Hierarchy Process (AHP) for adaptive weighting of key metrics (network load, SNR, user velocity) and Technique for Order of Preference by Similarity to the Ideal Solution (TOPSIS) for ranking candidate cells. This combination yields unified, policy-driven decisions that can be tuned for different traffic classes without redesigning the algorithm. Simulation in a dense SDN-based multi-RAT testbed demonstrates that TACM attains 90% offloading efficiency while maintaining a handover success rate of 99.89% (only 0.11% below the SNR-optimal scenario). Throughput lies within 1% of the load-optimal baseline, and packet loss is reduced by up to 93%. These results confirm TACM's effectiveness in balancing divergent performance objectives. The performance of all methods also improves as the number of WiFi access points increases due to network densification and increased offloading opportunities.
Managing security policies across hybrid multi-cloud environments requires coordinating vendor-specific rules across heterogeneous platforms, a process that is manual, errorprone, and inconsistent. We present a vendor-agnostic policy fusion framework that translates a single high-level intent specification into five heterogeneous enforcement formats spanning container orchestration, service mesh, cloud security groups, policy engines, and firewall access control lists. The framework employs Answer Set Programming (ASP) for declarative policy reasoning with a plan compiler that computes model diffs to produce minimal update plans. From a single 5-policy intent, the system generates semantically consistent policies and deploys them to a hybrid testbed comprising Kubernetes with Istio, AWS emulation via LocalStack, and OPA. Scale benchmarks across 25–200 services demonstrate that incremental plan compilation reduces plan size by 98.4–99.0% compared to batch recomputation (Wilcoxon, $p<0.001)$, with sub-10ms median solve latency at 200 services. Drift detection identifies 100 % of injected configuration violations (10 of 10) with a mean detection time of 1.25ms and unit-sized repair plans.
The near-real-time RAN Intelligent Controller (near-RT RIC) is responsible for real-time monitoring and control functions in O-RAN systems through the processing of E2 indications and xApp-based control logic under sub-second operational constraints. As the number of connected E2 nodes increases, periodic indication traffic may introduce significant queueing delay and system-level runtime overhead, leading to stale control decisions and degraded responsiveness on resource-constrained platforms. However, the practical processing limits of near-RT RIC systems under sustained indication-driven workloads remain insufficiently understood. This work establishes a baseline upper-bound characterization for a single-core near-RT RIC using a simplified queueing model and measurement-driven analysis. We model the indication-handling path as a single-server queueing system and validate the analysis using an open-source near-RT RIC implementation under controlled periodic workloads with varying numbers of connected E2 nodes. Experimental results show that practical saturation emerges substantially earlier than the analytical stability bound due to runtime overhead, reduced CPU headroom, and increasing system-level contention. Once overload conditions appear, queueing delay and overload probability grow rapidly, resulting in early performance collapse under sustained workload. The results provide a measurement-grounded reference for interpreting the gap between analytical stability bounds and practical operating limits in near-RT RIC systems.
Reactive SDN installs flow rules on demand, coupling data- and control-plane delays. We derive a closed-form expression for the steady-state mean packet delay in reactive SDN with idle timeout-based rule eviction. Using M/M/1 queuing theory, Poisson thinning, and renewal-theoretic residual-life analysis, we decompose delay into three components: baseline switch sojourn, miss-triggered installation penalty, and residual waiting for packets arriving during active installations. We identify three stability conditions, including a novel installation-load constraint. Extensive simulation and emulation validates the model, showing close agreement with analytical predictions. The closed-form expression enables direct optimization of idle timeout to minimize delay, supporting practical configuration decisions in latency-sensitive SDN deployments.
The growing demand for deterministic communication in Industry 5.0 has reinforced the role of IEEE Time-Sensitive Networking (TSN) as the backbone of next-generation Ethernet. Traditional synchronous mechanisms face scalability and clock-drift issues, which IEEE 802.1Qcr mitigates through the Asynchronous Traffic Shaper (ATS). ATS regulates frame timing by computing eligibility times without clock alignment, ensuring low delay and deterministic performance across multiple traffic streams. This paper focuses on the metering phase of ATS, analyzing both per-stream eligibility computation and the resource utilization of metering modules as key factors to guarantee fairness and efficiency. We design a Field Programmable Gate Array (FPGA)-based implementation combining token bucket logic with ATS scheduling, enabling precise evaluation of delay, and hardware cost under realistic industrial workloads. Results show that per-stream ATS metering provides deterministic behavior while maintaining minimal hardware utilization, making the design well-suited for large-scale industrial environments.
