Multicast Topology Inference from End-to-end Measurements

Abstract

This paper describes a class of statistical estimators of multicast tree topology based on end-to-end multicast traffic measurements. This approach allows the determination of the logical multicast topology without assistance from the underlying network nodes. We provide six instances of the class, variously using measurements of loss or delay, or marks set by network elements. We compare their accuracy and computational cost, and recommend the best choice in each of the light and heavy traffic load regimes.