A Compressive Sensing-Based Reconstruction Approach to End-to-End Network Traffic

Abstract

This paper studies the reconstruction method of end-to-end network traffic based on compressing sensing. The existing technologies about end-to-end traffic measurements need many more network resources, and thus they are prohibitive in practice. We take advantage of a subset of router interfaces to collect the volume of partial origin-destination flows. Different from previous approach, we exploit a novel method derives from compressive sensing to recover the rest of the origin-destination flows according to the partial origin-destination flows. The random walk method is used to build the measurement matrix for compressive sensing. And then, we exploit this measurement matrix to select a subset of origin-destination flows for directly measuring and calculate the observe results. Finally, we reconstruct all of the origin-destination flows from the observe results. Simulation results from the real backbone networks states that our method can reconstruct the end-to-end network traffic more accurately than previous methods.