Network tomography based on flow level measurements

Internet traffic primarily consists of packets from elastic flows, i.e., Web transfers, file transfers (FTP), and e-mail, whose transfers are mediated via the transmission control protocol. We develop a conditional sampling technique to analyze throughput correlations among elastic flow classes based on flow level measurements from current network traffic monitoring tools. The primary contributions of the paper are: (1) a demonstration of throughput correlation among temporally overlapping flows on congested resources by using analytical/simulation models; (2) application of a multivariate statistical method (principal components) to infer network properties, such as the number of resources shared by flows in the network, from non-intrusive, flow level measurements collected at a single site. Our proposal for using flow level measurements to infer network properties differs significantly from previous network tomography research that has employed end-to-end packet level measurements for making inferences.