A Topology Partition Algorithm Based on Link Coarsening in Parallel Network Simulation

Today´s networks are facing more and more serious large-scale network attack such as worm and botnet. While such attacks can´t and shouldn´t be reproduced in the real network, network simulation is become more and more popular during the research. As the prerequisite of network simulation, topology partition influences the performance of simulation significantly. We suggest a new topology partition algorithm in parallel network simulation. This algorithm is a kind of agglomerative hierarchical clustering methods based on link coarsening. The experiment shows that our algorithm can finish partition on current computers in acceptable time for even around millions of vertices. Furthermore, the links between different sub-domains are fewer and the connectivity in each sub-domain is also guaranteed. Compared with the results of KMETIS, the new algorithm can reduce the edgecut more than 90% in general.