Reproducible Datasets and Simulation Results for QoS-Aware Internet Traffic Engineering: Exact Optimization versus Adaptive Multi-Agent Routing

This Zenodo record contains the datasets, complete experimental output files, master seeds, and consolidated results associated with the manuscript: “Exact Optimization versus Adaptive Multi-Agent Routing for QoS-Aware Internet Traffic Engineering”. The experimental data were generated from synthetic network topologies under controlled connectivity, capacity, and traffic-demand constraints, together with benchmark topologies from the SNDlib library. The repository includes: • Input datasets used for both the exact CPLEX-based optimization approach and the MAS/NetLogo approach• Generated synthetic network topologies and traffic-demand instances• SNDlib benchmark instances used in the experimental evaluation• Complete output files for the reported CPLEX experiments• Complete output files for the MAS/NetLogo experiments• Five independent MAS runs per dataset when applicable• Master seeds and consolidated baseline experimental results• Parameter-sensitivity experiments for the threshold granularity δ and RetryLimit• Ablation experiments evaluating the individual contributions of the three-mode threshold controller and the multi-group routing retry strategy• Deterministic heuristic baseline experiments using Standard Widest-Path and Yen k-shortest paths (k = 2)• From-scratch CPLEX re-optimization experiments for representative dynamic-traffic perturbation and link-failure cases• Consolidated Excel workbooks reporting objective values, routing feasibility, CPU times, iteration counts, solver statuses, reference-value hit rates, heuristic routing rates, and other performance indicators used in the manuscript The same baseline input datasets are used for the exact CPLEX formulation and the MAS/NetLogo framework, enabling direct comparison under identical traffic-engineering objectives and network conditions. The Excel workbooks report the master seeds used for reproducibility. Topology-generation, traffic-generation, MAS-scheduling, exploration, perturbation, and link-failure sub-seeds are deterministically derived from the reported master seeds according to the procedure described in the manuscript appendix. The experimental materials cover: • Static routing experiments• Scalability analysis• Dynamic traffic perturbation scenarios• Link-failure scenarios• Threshold-granularity sensitivity analysis• RetryLimit sensitivity analysis• MAS component ablation study• Deterministic routing heuristic baselines• From-scratch CPLEX re-optimization after traffic perturbations and link failures Benchmark instances from the SNDlib library were also used in the study and are publicly available at:https://sndlib.put.poznan.pl/home.action The source code is not included in this Zenodo record and is available from the author upon reasonable request, as stated in the associated manuscript. The provided materials support reproduction, verification, and independent inspection of both the main experimental campaign and the supplementary analyses reported in the revised manuscript.