Large-Scale Emulation Network Topology Partition Based on Community Detection With the Weight of Vertex Similarity.

The article focuses on LENTP, a large-scale emulation network topology partition method designed to efficiently partition emulation network topologies composed of millions of nodes and edges using virtualization technology. LENTP leverages community detection based on vertex similarity weights, specifically employing an improved Louvain algorithm combined with a tree-structured area compression technique to reduce topology scale and enhance partitioning efficiency. The method aims to minimize the number of subnets and remote links while respecting virtual resource constraints, and it includes a repartitioning and merging stage to optimize resource utilization. Experimental results on network topologies ranging from 10,000 to over 1 million nodes demonstrate that LENTP improves running-time efficiency and partition quality compared to traditional methods like METIS, making it suitable for building large-scale network emulation scenarios.