Experimental Study on Diversity and Novelty of Interdomain Paths

Exploiting Internet path diversity to enhance communication reliability and performance is an important research field. Although the Internet has enormous physical diversity in the underlaying infrastructure, inter-domain routing protocol and routing policies highly limit this path diversity in the Internet. In order to investigate the effects of routing policies on inter-domain paths, we develop a measurement study framework to characterize and classify inter-domain paths. In this framework, we define the concepts of the transit strategy and the valley and valley-free paths, and give a formal model of strategylabeled inter-domain paths. Then we develop three associated algorithms: Topology extraction (TE), Multi-path computation algorithm (MCA) and Classification of path pattern (CPP), which are used to compute and classify inter-domain paths. Experimental results show two useful and important observations that valley paths have averagely more than 73% in all inter-domain paths, and Class 1 paths account for 64.6% of all valley paths. The observations may guide the design of inter-domain routing protocol, especially the design of multi-path routing, to achieve higher performance and reliability.