Characteristic measurement of the connection degree for network monitoring

Monitoring and measuring the dynamic changes of the traffic patterns are important for network management. In this paper, we propose the bidirectional flow model to aggregate the traffic packets, which is more efficacious to reflect the alternation character of the users. Then we select the one order and second order connection degree to capture the characteristic of the traffic patterns. The K-L divergence is proposed to measure the changes between the distributions of the degrees at the adjacent time point and decide whether there are abnormal changes. Finally, the China Reminder Theory is employed to design the sketch method for connection degree calculation and the abnormal connection holders query for real time monitoring. The experimental results in an actual networks show that the methods proposed in this paper are efficacious for dynamic change measurement and monitoring in the nowadays network. The remainder sketch methods can calculate the degree quickly and the abnormal connection degree holders can be located exactly, which are important for network management and active defense.