Towards a Statistical Characterization of the Interdomain Traffic Matrix
Identifying the statistical properties of the Interdomain Traffic Matrix (ITM) is fundamental for Internet techno-economic studies but challenging due to the lack of adequate traffic data. In this work, we utilize a Europe-wide measurement infrastructure deployed at the GÉANT backbone network to examine some important spatial properties of the ITM. In particular, we analyze its sparsity and characterize the distribution of traffic generated by different ASes. Our study reveals that the ITM is sparse and that the traffic sent by an AS can be modeled as the LogNormal or Pareto distribution, depending on whether the corresponding traffic experiences congestion or not. Finally, we show that there exist significant correlations between different ASes mostly due to relatively few highly popular prefixes.