Pathsift: a library for separating the effects of topology, policy, and protocols on IP routing

Routing in IP networks is a computation distributed across many routers and subnetworks with inputs specified as low-level instructions in a device-by-device manner. Predicting the impact of input changes is extremely difficult; thus, network designers must not only decide what policies to deploy, but also must encode those policies in a low-level configuration and then run the protocols to see the result. Although several tools exist for the last step, almost none exist to help with the first two, particularly when the network of interest spans multiple domains. In this paper we present Pathsift, a Python library for generating and comparing sets of paths, designed to permit the exploration of high-level routing policies. Our approach permits evaluation of path quality resulting from different routing policies without packet-level protocol simulation. Our library supports computation of interand intra-domain routes and generates visualizations of custom metrics on sets of paths. It thus brings together into one toolchain key components of graph libraries, network simulators, and visualization tools.