Discovery of hotspots evolution based on community detection method

How the hotspot or congestion area evolves in a large scale complex networks is still not clear. The prediction of such behavior is more difficult. In this paper, the classical Fast-Newman algorithms for community detection is improved by considering node weight and edge weight in the network model. The evolution of the communities are reconstructed from the network trace. The relation between the hotspot evolution trends and the network characters including node count, dynamic path selection and load factor are investigated. The results show the hotspot area have rich dynamic behavior patterns such as merging, splitting, enlarging and shrinking behaviors, and multiple hotspot distributions is affected by the node count, dynamic path selecting policy and load factors. This work can help improving the congestion control policies widely existed in computing networks and transportation networks.