Zoom: 1x 1.5x 2x

ABSTRACT: The traffic volume that originates at one node and is destined for another is described as a traffic demand, capturing how much traffic "wants" to go from one place to another. Traffic matrices describe the traffic demands for all pairs of nodes in a network, and thus represent a network-wide view of the traffic carried on the whole network. Direct measurement of these network-wide traffic descriptions is considered too costly, and thus a number of interesting inference techniques have been proposed to tackle the problem of estimating traffic matrices from a limited set of measurements. In this overivew talk, we summarize the progress in this field in the last 5 years. We begin by presenting inference techniques from each of the three generations of methods. This body of research has thus resulted in the development of a number of interesting traffic models for origin-destination flows. Modeling such flows presents additional challenges beyond the traditional flow modeling, as it now becomes important to incorporate spatial and temporal correlations, as well as non-stationarity. Because traffic matrices are the inputs to many network design problems, this reseach has lead to an increase of applications using traffic matrices. We illustrate examples such as routing analysis, failure planning and network security. To further enhance benchmarking and performance evaluation of such network problems, we discuss the problems of synthetic traffic matrix generation used to create testing scenarios.

AUTHOR BIO: Nina Taft is a senior researcher at Intel Research Berkeley. After completing her PhD at UC Berkeley, Nina did a postdoc at the University of Paris VI. She then spent 4 years at SRI working on routing and congestion control. Following that she spent 5 years in Sprint's IP Research group, where her worked focused on traffic engineering problems, including traffic characterization, prediction, and estimation as well as capacity planning and routing. At Intel, her work is currently focused on enterprise traffic modeling, distributed monitoring, and network security problems. She is an editor for the IEEE Transactions on Networking, serves on the steering committee of the Internet Measurement Conference, and is PC co-chair for ACM SIGCOMM 2007.