A Quantum Formulation for Clustering Edge Topologies

Edge topologies usually comprise many different computational nodes, each with limited resources. Moreover, they manage varying workloads that can access the system from different entry points. These key characteristics make the problem of selecting the nodes dedicated to a specific workload quite heavy. In this context, clustering, that is, organizing nodes into smaller, more manageable groups, is a critical first step to support effective decision making. This paper introduces Min-Distance Based Clustering (MDBC), a novel quantum-based model for clustering edge topologies that utilizes quantum annealing. By recasting the clustering task as a quantum optimization problem, we can easily identify the appropriate clusters of edge nodes and nearby demand sources in a spatially aware manner. Our approach enables fast, periodic clustering executions, allowing the edge topology to adapt quickly to continuously changing conditions. We thoroughly evaluate the proposed method and demonstrate clear gains in scalability and execution times while maintaining almost optimal solution quality.