Revealing MPLS tunnels obscured from traceroute

Benoit Donnet, Matthew Luckie, Pascal Mérindol, Jean-Jacques Pansiot
Appears in: 
CCR April 2012

Operators have deployed Multiprotocol Label Switching (MPLS) in the Internet for over a decade. However, its impact on Internet topology measurements is not well known, and it is possible for some MPLS configurations to lead to false router-level links in maps derived from traceroute data. In this paper, we introduce a measurement-based classification of MPLS tunnels, identifying tunnels where IP hops are revealed but not explicitly tagged as label switching routers, as well as tunnels that obscure the underlying path. Using a large-scale dataset we collected, we show that paths frequently cross MPLS tunnels in today's Internet: in our data, at least 30% of the paths we tested traverse an MPLS tunnel. We also propose and evaluate several methods to reveal MPLS tunnels that are not explicitly flagged as such: we discover that their fraction is significant (up to half the explicit tunnel quantity) but most of them do not obscure IP-level topology discovery.

Public Review By: 
Yin Zhang

Multiprotocol Label Switching (MPLS) has been widely deployed in the Internet. It is well known that MPLS clouds may potentially lead to inaccurate and incomplete Internet maps when performing active measurements using traceroute. In particular, MPLS tunnels may either obscure the underlying path or be incorrectly classified as direct IP links (between IP routers that are not physically adjacent). However, the quantitative impact of MPLS tunnels on Internet topology measurements is not well understood. The paper uses two features to characterize MPLS tunnels: (i) the ttl-propagate option on ingress label edge routers (LERs) and (ii) RFC4950. The ttl-propagate option allows traceroute to reveal IP hops within an MPLS tunnel. RFC4950 implementation provides all information related to a label switched path (LSP). The paper provides a taxonomy of MPLS tunnels: (i) explicit tunnels, which use both ttl-propagate and RFC4950, (ii) implicit tunnels, which use only ttl-propagate, (iii) opaque tunnels, which use only RFC4950, and (iv) invisible tunnels, which use neither option. The paper then develops new inference mechanisms to identify different types of MPLS tunnels and thus reduce the impact of MPLS on IP topology discovery. The focus is on explicit, implicit and opaque tunnels. Based on the new inference techniques, the authors collected a largescale measurement dataset. According to this dataset, a significant fraction (at least 30%) of the paths traverse an MPLS tunnel. Moreover, a significant fraction of MPLS tunnels are not explicitly flagged, but fortunately most of these MPLS tunnels do not obscure IP-level topology discovery. Overall, a nice measurement paper. The topic is timely. The proposed measurement techniques are sound. The findings are interesting and shed light on the impact of MPLS tunnels on IPbased topology discovery. I also expect the techniques developed in this paper to be applied by others in future research on IP-based topology discovery.