Dynamic Features Measurement and Analysis for Large-Scale Networks

Abstract

Detecting and measuring the changes of temporal traffic patterns in large scale networks are crucial for effective network management. This paper presents the concept of region flow to aggregate traffic packets. Regions are defined by the IP prefix, and a region flow is a group of packets with the same source and destination region during a time interval. In this way, the number of flows can be reduced significantly and a better extraction of pivotal traffic metrics is generated. Three traffic features: source connection degree, destination connection degree and packet distribution ratio are proposed to capture the dynamic change of the flow patterns between regions and the Renyi cross entropy are applied to measure and detect the changes. The experimental results show that the method proposed in this paper can capture the dynamic traffic features effectively for 10Gbps backbone networks, and can be used for detecting abnormal network behaviors.