Variability analysis for traffic distribution of typical network services

Network traffic characteristics analysis is an important prerequisite for network traffic engineering, traffic monitoring and controlling. Our analysis draws upon massive data obtained from a large Internet service provider (ISP) network. By analyzing traffic change rate features and the maximum rate of change, we extract and summarize the variation distribution characteristics of main services. Then we used Euclidean distance algorithm to research the similarity between the bi-direction change rates. Finally we used the change rate data for integral summation analysis. By these work, we present some interesting conclusions that characteristics of inbound traffic are more changeable than outbound, and the maximum value time intervals of typical services are fixed and similar, in addition the similarity between outbound and inbound change rate is not large, but the span among different services is large. These conclusions are useful information for providing insights of planning and managing networks.