Enhanced Crankback Signaling for Multi-Domain Traffic Engineering

Abstract

Multi-domain traffic engineering is a major focus area for carriers today and crankback signaling offers a very promising and viable alternative here. Although some initial crankback studies have been done, there is still significant latitude for improving multi-domain crankback performance. To address these concerns, this paper studies realistic IP/MPLS multi-domain networks and proposes a novel solution for joint intra/inter-domain signaling crankback. Namely, dynamic intra-domain link-state routing information is coupled with inter-domain path/distance-vector routing state to improve the overall search process. Mechanisms are also introduced to limit setup signaling overheads/delays. The performance of the proposed solution is then analyzed using simulation and compared against other crankback techniques as well as hierarchical inter-domain routing strategies.

Enhanced crankback signaling for multi-domain IP/MPLS networks

“…Finally, the results in Figs. 7 and 8 show that the proposed crankback solution (with moderate counter values, i.e., H 1 = 2/H 2 = 2) can even outperform the other, more complex hierarchical routing strategies, extending upon the findings in [18]. In particular, resultant BBR values are always lower than those yielded by simple node abstraction, and for the case of NSFNET, even lower than those yielded by more advanced full-mesh abstraction.…”

“…Along these lines, this paper proposes a novel enhanced crankback solution for multi-domain networks based upon the standard resource reservation (RSVP-TE) protocol. Note that this solution extends upon our recently published work in [18] by presenting more detailed discussions of the proposed scheme as well as wider range of simulation analysis results. Specifically, two-levels of crankback are defined-at the intra and inter-domain levelsand active crankback history (failure state) is also leveraged.…”

“…5 and 6. This extends upon the work in [18] which only considers one topology. Here the 10-domain topology has an average of 2.5 links/domain whereas the NSFNET topology has an average of 1.56 links/domain, i.e., slightly lower inter-domain connectivity.…”

“…The solution was recently tabled in [18] and this effort details a more expanded discussion thereof. The framework assumes realistic settings with full link-state routing at the intra-domain level (e.g., OSPF-TE) and more scalable path/distance vector routing at the inter-domain level.…”