Efficient Inference of AS-Level Paths in the Internet

Abstract

Routing protocols maintain connectivity of Internet routers and hosts and determine the path that a packet traverses. Inferring Internet paths is critical for evaluating the performance of Internet ap plications and services, such as cloud services and content delivery. Unlike intra-domain routing protocols, which typically use the shortest paths , inte-domain routing protocol apply local policies for selecting routes/paths and propagating routing information. These routing policies are typically constrained by contractual commercial agreements between ASes. It is well-known that routing policies can impact the AS path that an AS may select for delivering a packet. Thus, the ability to infer the AS-level paths is critical to evaluate the imp act of routing policies on the performance of Internet applications and services. In addition, inferring AS-level paths is also imp ortant for content providers, such as Google and Amazon, to determine routing policies to ensure small latency in delivering content. However, inferring AS-level paths is challenging. Internet path selection largely depends on routing policies, which in turn are determin ed independently by network administrators and are considered as confidential information. In this paper, we present thre e common routing policies in the Internet and formulate the problem of inferring routing policy conforming AS-level paths. We present efficient algorithms for inferring the Internet AS-level paths. The algorithms are proved to be optimal in terms of the ability of derive the policy-conforming AS-level paths. We further quantify the efficiency of these algorithms.