In search for an appropriate granularity to model routing policies

Routing policies are typically partitioned into a few classes that capture the most common practices in use today[1]. Unfortunately, it is known that the reality of routing policies[2] and peering relationships is far more complex than those few classes[1,3]. We take the next step of searching for the appropriate granularity at which policies should be modeled. For this purpose, we study how and where to configure per-prefix policies in an AS-level model of the Internet, such that the selected paths in the model are consistent with those observed in BGP data from multiple vantage points. By comparing business relationships with per-prefix filters, we investigate the role and limitations of business relationships as a model for policies. We observe that popular locations for filtering correspond to valleys where no path should be propagated according to inferred business relationships. This result reinforces the validity of the valley-free property used for business relationships inference. However, given the sometimes large path diversity ASs have, business relationships do not contain enough information to decide which path will be chosen as the best. To model how individual ASs choose their best paths, we introduce a new abstraction: next-hop atoms. Next-hop atoms capture the different sets of neighboring ASs an AS uses for its best routes. We show that a large fraction of next-hop atoms correspond to per-neighbor path choices. A non-negligible fraction of path choices, however, correspond to hot-potato routing and tie-breaking within the BGP decision process, very detailed aspects of Internet routing.