Internet-wide multipath: a scalability analysis of path identification schemes

In the present Internet, inter-domain routing is based on BGP-4 which prevents the use of multiple paths, but rather selects a single path per destination prefix. However, as the global Internet connectivity graph displays vast inherent path diversity, multipath routing has been identified as an important feature for the Future Internet, for both robustness and traffic engineering purposes. Some works propose to enable the propagation of multiple inter-domain routes per prefix [1], [2], [3]. While these works propose the use of a new packet field (which we call Path-ID) to specify the path which is to be followed by packets, they do not discuss about the scalability of such an approach. Indeed, as such a field is globally significant, it can lead to some scalability issues which could make such proposals impossible to implement. We propose the study of the scalability of such a field in order to answer the recurring but un-answered question: "Will Internet-wide multipath make the FIB explode?". To this end, we propose some Path-ID global identification schemes, identify their weaknesses and propose simple and realistic assumptions which resolve their scalability issues. At the end, we show that these simple assumptions are sufficient to enable vast multipath in the Internet while remaining manageable.