Towards Easy-to-Monitor Networks: Network Design and Measurement Path Construction

In the era of Industry 4.0, networking has undergone a significant transformation driven by technologies like artificial intelligence. This shift necessitates the design of new network typologies to address the challenges posed by the exponential growth in network scale and complexity. Alongside this, network monitoring has gained increased importance due to the need for real-time operations. Consequently, there is a requirement for new network topologies that offer enhanced monitorability. To meet these evolving demands, this paper leverages a quantitative measure of network monitorability called $k$-identifiability, which forms the basis for two new network design problems: topology design and end-to-end measurement design. Both problems, however, are intractable. To tackle these problems, we conduct a systematic analysis of topological features that can simplify the design process. Based on this analysis, we propose an integrated algorithm that utilizes a novel dual-heuristic approach to generate well-formed topologies. Additionally, we develop an algorithm for creating measurement paths that accurately pinpoint failures. Through evaluation, we compare our dual-heuristic algorithm to theoretical optimal solutions in small-scale networks where brute-force search is feasible, demonstrating its near-optimality. We also showcase the effectiveness of our method in both topology and measurement design for large-scale networks.